Ilfll
Jl '' >' Vt ' ^! ill
f }<!
.Is, ,
fyxull ^nivrnxi^ pihMg
THE GIFT OF
..^....f/-
f>%yi^rj/rr<^P!hi^^^..
..d..:...!.J?7.'+:/»r..r?r ^J^J.Al.l.iAS,..
I Cornell University Library
j P 221. S43
' Elements of experimental phonetics
3 1924 026 451 967
Cornell University
Library
The original of tiiis book is in
tine Cornell University Library.
There are no known copyright restrictions in
the United States on the use of the text.
http://www.archive.org/details/cu31924026451967
gale 05icentenntal f^utilication0
THE ELEMENTS OF
EXPERIMENTAL PHONETICS
gale 'Bicentennial faublicationjs
JVith the approval of the President and Fellows
of Tale University, a series of volumes has been
prepared by a number of the Professors and In-
structors, to be issued in connection with the
Bicentennial Anniversary, as a partial indica-
tion of the character of the studies in which the
University teachers are engaged.
This series of volumes is respectfully dedicated to
THE ELEMENTS OF
EXPERIMENTAL PHONETICS
-/;
BY
EDWARD WHEELER SCRIPTURE
W/T/f THREE HUNDRED AND FORTY-EIGHT ILLUSTRATIONS
AND TWENTY-SIX PLATES
NEW YORK: CHARLES SCRIBNER'S SONS
LONDON : EDWARD ARNOLD
1902
Copyright, WOS,
By Yale University
Published July, igo2
W}
UNIVERSITY PRESS • JOHN WILSON
AND SON • CAMBRIDGE, U.S.A.
PREFACE
This book is an attempt to collect the most valuable
experimental data concerning the voice in song and speech.
I believe that the science of phonetics cannot be confined
to a study of the physics and physiology of speech sounds,
and that the problems of speech perception, of the psychology
of language, of rhythm and verse, etc., can all be treated by
experimental methods and must be included.
The book owes much to many friends. Prof. Hanns
Oertel (philology, Yale Univ.), Prof. J. Gbddbs (modern
languages, Boston Univ.), Prof. E. C. Sanfoed (psychology,
Clark Univ.), Prof. W. H. Burnham (pedagogy, Clark Univ.)
and Dr. F. M. Josselyn (experimental phonetics, Boston
Univ.) have made many suggestions that were incorporated
into Part II. Prof. H. L. Swain (laryngology, Yale Univ.),
Prof. H. B. Ferris (anatomy, Yale Univ.) and Mr. C. C.
Clarke (French, Yale Univ.) have aided in Part III. Dr.
Jonathan Weight (laryngology, Brooklyn) has furnished
many valuable suggestions for the chapters on the larynx.
Prof. C. H. Graisidgbnt (Romance languages. Harvard
Univ.), Prof. A. M. Elliott (editor of Mod. Lang. Notes,
Johns Hopkins Univ.), Prof. E. S. Dana (editor of Amer.
Journ. Sci., Yale Univ.) and Dr. T. R. French (laryn-
gology, Brooklyn) have permitted me to use blocks from
their publications. Prof. E. B. Barker (translator of
Spalteholtz's anatomy, Chicago Univ.) consented to the
viii PREFACE
reproduction of some plates. W. B. Saunders (publisher
of the Amer. Textbook of Physiology) furnished electro-
types for Figs. 51 to 56, and Ginn & Co. (publishers,
Boston) those for Plates XVII to XXVI. Many of the
illustrations in this book are to be credited to the Studies from
the Yale Psychological Laboratory, Vols. VII and X.
My deepest obligation is to Mr. E. H. Ttjttlb (Yale
Univ.). His wide acquaintance with the phonetics of vari-
ous languages has enriched the book with many examples.
During the preparation of the manuscript and the correction
of the proof I have relied constantly on his technical knowl-
edge and rare critical skill.
E. W. SCEIPTURE.
Yale TJniversitt,
New Haven, Conn., April, 1902.
CONTENTS
Part I. CURVES OF SPP^ECH
Chapter Page
I. Vibratory Movement . . . 1
II. Phonautograph Curves and Manometric Flames . 17
III. Phonograph Records .... 32
IV. Gramophone Records . .....' 52
V. Immediate Analysis op Speech Curves . . 62
VI. Harmonic Analysis . . . . 72
Part II. PERCEPTION OF SPEECH
VII. The Organ of Hearing .... 76
VIII. Perception of Sounds ... 89
IX. Perception of Speech Elements . . . 113
X. Speech Ideas . . .... . 126
XI. Association of Ideas 135
XII. Habits of Association ... . .... 152
XIII. Special Associations in Speech . . . 163
XIV. Formation of Speech Associations 175
Part III. PRODUCTION OF SPEECH
97
XV. Voluntary Action and the Graphic Method . . 18
XVI. Breathing . . . . . . 21
XVII. Vocal Organs 229
XVIII. Structure and Observation of the Larynx . 239
XIX. Action of the Larynx . . . . 251
XX. Tones of the Vocal Cavities . . . . 281
XXI. Tongue Contacts : Methods of Palatograph y ; /
American, Irish and Hungarian Records . . . 296 V
1
X CONTENTS
Chapter Pagb
XXII. Tongue Contacts : German Records . 308
\ XXIII. Tongue Contacts: French and Italian Records 312
^ XXIV. Tongue Positions and Movements .... 325
XXV. Pharynx, Nose, Velum, Lips and Jaw . 338
XXVI. Simultaneous and Successive Speech Movemknts 357
^- XXVII. Vocal Control . . .379
Part IV. FACTORS OF SPEECH
399
432
446
462
... 472
488
503
506
517
XXXVII. Speech Rhythm . 537
APPENDICES
Appendix I. Fourier Analysis ... . . 561
" II. Studies of Speech Curves 575
" III. Free Rhythmic Action ... . 602
Additions and Corrections . ... 607
List op Phonetic Symbols . . . ... 608
Musical Notation . 2 o /ace 610
Index . . 611
Plates
XXVIII.
Vowels . ...
XXIX.
Liquids and Consonants
, XXX.
Sound Fusion . .
^ XXXI.
Progressive Change .
XXXII.
Melody . .
XXXIII.
Duration . . .
XXXIV.
Loudness . . .
XXXV.
Accent ....
XXXVI.
Auditory and Motor Rhythm
LIST OF ILLUSTEATIOI^S
Figure Page
1. Vibrating particle 1
2. Sinusoid vibration .... 3
3. Propagation of vibration 5
4. Frictional sinusoid 6
5. Magnetic vibrator 7
6. Apparatus for recording vibrations, drum, motor, resistance . . 8
7. Continuous-paper drum . . 8
8. Lantern recorder . . 9
9. Countershaft for reducing speed , 10
10. Countershaft for reducing speed . . . 10
11. Countershaft for reducing speed .10
12. Countershaft for reducing speed .11
13. Diagram of connections for experiments with magnetic vibrator . 11
14. Record from magnetic vibrator 13
15. Spherical resonator (Kcenig) ... 14
16. Adjustable resonator (Kcenig) 14
17. Electric fork (after IZimmermann) 15
18. Manometric-flame apparatus (Kcenig) ... . ... 25
19. Manometric flames, vowels (Kcenig) .... .... 26
20. Manometric flames, m and n (Kcenig) ... . . 27
21. Manometric flame, r (Kcenig) . . 28
22. Manometric capsule with oxygen cylinder (Nichols and
Merritt) ■ 28
23. Apparatus for photographing manometric flames (Nichols and
Merritt) 29
24. Manometric flame, o . . . . 29
25. Phonograph (Edison) . . . . ^^. . . . . . 32
26. Phonograph (Edison) 33
27. Phonograph recorder (Edison) .... 33
28. Phonograph knife cutting cylinder .... ... 34
29. Phonograph record (Boeke) . 34
30. Phonograph reproducer (Edison) . .... ... 34
31. Hermann's curves of lotg German vowels . 40
xii LIST OF ILLUSTRATIONS
Figure Page
32. Hermann's curves of short German vowels 41
33. Hkrmann's curves of consonants 42
34. Phonograph tracer (Beviek) 49
35. Bevier's curve of a 50
36. Gramophone recorder 52
37. Gramophone disc . . 54
38. Gramophone reproducer 54
39. Gramophone . . 55
40. Gramophone tracer (first model) . . 56
41. Gramophone tracer (tracing lever) ... 57
42. Gramophone tracer (second model) 60
43. Gramophone curves tested . . . 61
44. Curve of pitch for o in ' saw him ' . . . 66
45. Compounding of sinusoids 2 : 3 ... 67
46. Compounding of sinusoids 1 : 2 67
47. Compounding of fork vibrations (Kcenig) 68
48. Synthetic vibrator 70
49. Curve with 12 ordinates drawn .... 74
50. Harmonic plot . . . 75
51. General structure of ear (Quajn) 76
52. Middle ear (Testut) ... . . . 77
53. Scheme of labyrinth (Testut) . .... 79
54. Bonework of cochlea (Testut) . , .... 80
55. Section through canal of cochlea (Testut) ... .... 80
56. Terminal organs in cochlea (Testut) 81
57. Speech centers in cortex of cerebrum 83
58. Scheme of functional connections of speech centers 87
59. Siren with electrical connections .... 90
60. Pfeil marker (Langendorfp) . . 91
61. Deprbz marker (Verdin) 92
62. Contact wheel 93
63. Variator (Stern) 102
64. Audiometer ....'. 110
65. Pendulum chronoscope 153
66. Voice key 155
67. Diagram to illustrate lapses (Meringer and Mayer) . . . 165
68. Stimulation and record of muscle . . 189
69. Curve of muscular contraction . . . 189
70. Scheme of brain structure (after Auzoux) . . ... . 193
71. Tambours and clockwork recording drum 195
72. Recording tambour (Marey, Verdin) 196
73. Recording tambour (Carpentier, Verdin) . .... 196
74. Rectification of tambour records (Langendorfk) 197
LIST OF ILLUSTRATIONS xiii
FiGUEB Page
75. Adjustment of recording point to drum (Langendorff) . . 198
76. Records of pulls 201
77. Record of constant effort ... 202
78. Record of pulls with increasing haste 204
79. Apparatus for measuring reaction time . . 206
80. Scheme of lamp battery . . 209
81. Four-socket lamp battery ... 210
82. Caliper spirometer (Beut) 214
83. Belt tambour spirometer (Verdin) 215
84. Records of abdominal breathings . . 216
85. Records of mouth breaths 217
86. Records of mouth breaths in recitation . 218
87. Records of mouth breaths in different expressions of du
(Vietok) . . 218
88. Nasal olives (Rousselot) . . 219
89. Vocal tambour (Rousselot) . 219
90. Breath recorder (Gad) ... 220
91. Nose and mouth breaths (Goldschbider) 224
92. Manometer (Roy) ... 225
93. Vocal organs (Testut) . . . . .... 230
94. Muscles of the face, superficial layer (Testut) 231
95. Muscles of the face, deeper layer (Testut) . ... 231
96. Muscles of the lips (Spalteholtz) . . ... 232
97. Muscles around the pharynx (Testut) . . . . 233
98. Muscles of the tongue, superficial layer (Testut) .... 234
99. Muscles of the tongue, deeper layer (Testut) . . 235
100. Section across the tongue (Spalteholtz) . . . 236
101. Muscles of pharynx (Testut) .... . . . 237
102. Thyroid and circoid cartilages, front view (Spalteholtz) . 239
103. Thyroid and circoid cartilages, side view (Spalteholtz) . . 239
104. Laryngeal cartilages with vocal bands relaxed .... . 240
105. Laryngeal cartilages with vocal bands stretched .... 240
106. Horizontal section across larynx (after Spalteholtz) . . . 240
107. Side view of larynx (Testut) ... 241
108. Front view of larynx (Testut) , 241
109. Rear view of larynx (Testut) .... 242
110. Side view of larynx with portion of thyroid cartilage displaced
(Testut) .... 242
111. Vertical transverse section of larynx (Spalteholtz) . . . 243
112. Vertical transverse section of one side of larynx (Spalteholtz) 244
113. Diagrams showing various adjustments of the glottis .... 245
114. Laryngoscope equipment . 247
115. LTse of laryngoscope (Stokk) 248
xiv LIST OF ILLUSTRATIONS
Figure Page
116. Glottis during respiration (French) 251
117. A glottis in singing/"! (French) . . 252
118. A glottis in singing gi (French) 252
119. A glottis in singing /'( (French) 253
120. A glottis in singing cP (French) 253
121. A glottis in singing e^ (French) 254
122. Membrane pipe . 258
123. Action of vocal bands (Mtjsehold) 259
124. Laryngeal recorder (Rousselot) 267
125. Flame figure, unison (Hensen) 269
126. Flame figure, octave (Hensen) . 269
127. Flame figure, duodecime (Hensen) 270
128. Flame figure, fifth (Hensen) 270
129. Flame figure, fourth (Hensen) 270
130. Flame figure, third (Hensen) 270
131. View of larynx, breathy tone (Curtis) ........ 274
132. View of larynx, breathy tone (Curtis) 274
133. Larynx in whispering (Czermak) 274
134. Map of roof of mouth (Lenz) 296
135. Sagittal map of mouth cavity (Lenz) 297
136. Artificial palate (Kingsley) 298
137. Palatogram (Kingsley) ... 299
138-150. Palatograms of American sounds (Kingsley) 303
151-167. Palatograms of Hungarian sounds (Balassa) .... 306
168-172. Palatograms of German sounds (GRiJTZNER) 308
173-183. Palatograms of German sounds (Vietor) 310
184-205. Palatograms of Parisian sounds (Rousselot) . . . 313
206-215. Palatograms of Rousselot's sounds (Rousselot) . . . 319
216-240. Palatograms of Italian sounds (Josselyn) . . . 323, 324
241. Atkinson's tongue curve (Laclotte) .... ... 330
242. Tongue explorer (Atkinson) . . .331
243. Tongue explorer, details (Atkinson) 331
244. Tongue positions (Atkinson) . . 332
245. Exploratory bulbs (Rousselot) . .... 333
246. Tongue movement in dido, tito (Josselyn) . ... 333
247. Tongue movement in a (Josselyn) 333
248. Tongue movement in e (Josselyn) 334
249. Tongue movement in o and u (Josselyn) 334
250. Tongue movement in i (Josselyn) 334
251. Tongue movement in ini (Josselyn) 334
252. Tongue movement in popolo (Josselyn) . . .... 335
253. Geniohyoid tambour (Rousselot) 335
254-259. Tongue movements (Rousselot) 336
LIST OF ILLUSTRATIONS xv
Figure Page
260-266. Velum movements (Weeks) ... 345
267-273. Nose and mouth records (Josselyn) 347-351
274-277. Pharyngeal arches in four French vowels (Thudichum) . 352
278. Lip and breath recorder (Rocsselot) 354
279. Records from breath and cords in French and German pa, ba
(Rotjsselot) .... 357
280. Records from breath and cords in American paet, baed (Rous-
selot) 358
281-283. Records from nose, cords, lips in apa, aba, ama (RosA-
pelly) 358
284. Record of mparltelbje" (Rousselot) 359
285. Record of tuna^pjur (Rousselot) 361
286. Record of mapovfoem (Rousselot) 361
287. Record of Voivylnalelnalopnm (Rousselot) 362
288. Record of totale (Josselyn) 365
289. Record of slita (Josselyn) ... 366
290. Record of riordinare (Josselyn) . 366
291. Record of atjene (Josselyn) 367
292. Record of mowoder (Zwaardemaker) . 371
293. Record of mudsr (Zwaardemaker) . . 371
294. Tongue positions (Laclotte) 372
296. Record of ^ouKoXos (Laclotte) 373
296. Record of */3ou7ro'Xos (Laclotte) 374
297. Record of aiTToXoj (Laclotte) .... 374
298. Record of apa (Rousselot) . 375
299. Record of apa (Rousselot) 375
800. Sagittal diagram for s (Zund-Burguet) 394
301. Sagittal diagram for s (Zijnd-Burguet) 394
302. Tongue director (Ztjnd-Burguet) 395
303. Sagittal diagram for s (ZiJND-BuRGNET) 395
304. Double tongue tambour (Meunier) 397
305. Tambour indicators (Meunier) 397
306. Artificial larynx and resonator 418
307-323. Lenz's diagrams 435
324. Curve of speech energy in a phrase 448
325-333. Curves of melody in German vowels .... . 475, 477
334. Melody of a French phrase 477
335. Melody in French phrases 479
336. Voice records 509
337. Curves of breath pressure in Lithuanian and Lettic .... 514
338-343. Forms of rhythm 519,520
344. Record of arhythmic action . 524
345. Results of arhythmic action .... . . . . 524
xvi LIST OF ILLUSTRATIONS
PlGUKE PAGB
346. Noiseless key 629
347. Apparatus for experiments on rliythm 530
348. Harmonic plot 572
349. Rondet's abacus . . . . 572
350. Curve of ai 676
351. Vibrations in ai .... 577
352. Vibrations in ai 578
363. Curve of dai 582
354. Vibrations in dai ... . 683
355. Curve of dai .... . . ... 584
356. Curve of lai ... 685
357. C\xv^i%oi Who HI he tie parson 1 . . . .590
358. Modulation of parson . 595
359. Modulation of parson . ...... 595
360. Curve of difficulty in free rhythmic action . .... 605
Plates I-II. Curves from Cock Robin.
Plates III-XI. Curves from Rip Van Winkle's Toast.
Plates XII-XIII. Melody of Rip Van Winkle's Toast.
Plate XIV. Curves of frequency in rhythmic sounds (Miyake).
Plate XV. Curves of frequency for some sounds in Cock Robin.
Plate XVI. Curves of intensity for some sounds in Cock Robin.
Plates XVII-XXVI. Tongue and lip positions in German and English
sounds (Geandgent).
THE ELEMENTS OF EXPERIMENTAL
PHONETICS
THE
ELEMENTS OF EXPERIMENTAL PHONETICS
PART I
CUEVES OF SPEECH
CHAPTER I
VIBEATOEY MOVEMEKT
ViBEATOEY movement can be most conveniently studied
by considering the motion of a single particle under certain
hypotlietical conditions. Suppose a particle to be drawn to its
position of equilibrium 0 (Fig. 1) by some force that
increases as the particle is moved from 0. When dis-
placed from this position ' in the direction + I^ by a
momentary blow, it will move a certain distance with
steadily decreasing speed until it comes to rest, and
then on account of the force acting toward 0 will
return to its starting-point. In doing this it will
have acquired a velocity that will drive it bej'-ond the
original position toward—!^ This movement will be -^^^ j
opposed by the tendency toward the position of equi-
librium, and the particle will be again stopped and drawn
back, whereby it will gain new momentum and will pass the
point 0 again, but in the opposite direction. At 0 it will
again have its maximum velocity. The momentum with
which it passes 0 will carry it toward + Y and the move-
ment will begin over again. It will thus continue to oscil-
late across its center of equilibrium.
Since the distance y of the particle above or below the
center of equilibrium varies with the time t, it is said to be
a function of t; this is indicated by the expression y =f(ty
1
-Y
2 CURVES OF SPEECH
If the center of equilibrium in Fig. 1 is supposed to travel
from left to right at a definite rate, it will describe a straight
line representing the time that has elapsed. The oscillating
particle will describe a wave-line showing its position at each
moment. The shape of the wave and the form of the function
will depend on the nature of the force acting on the particle.
The attraction of a particle to its position of equilibrium
may vary directly as its distance from that position. If we
make the additional suppositions that the actual movement
is small and that there are no forces of viscosity or friction
tending to dissipate the energy, it is not difficult to get an
expression for the movement.
Let t/ be the position of the particle at any moment t ; then,
on the suppositions made, the equation of movement can be
shown to be
• o t
y=a.sm Stt-,
where a is the amplitude, or the maximum value assumed by
y, and T is the periodic time, or the time required for y to
pass through one complete set of its values. A vibration of
this kind is said to be ' harmonic ' or ' sinusoidal ; ' a curve
expressing it is called a 'sinusoid.' When the vibrating
particle moves without being affected by external forces, its
period T is said to be its 'period of free vibration,' or its
' natural period.'
t
lo
li
a.
2 a.
11 X
sin q
20 sin 5
t
lo
II
a.
00 SI,
II X
siny
20 sin g
0.00
0.0
0°
0.00
0.0
0.55
1.1
198°
— 0.31
-6.2
0.05
0.1
18
+ 0.31
+ 6.2
0.60
1.2
216
-0.59
-11.8
0.10
0.2
36
+ 0.59
+ 11.8
0.65
1.3
234
-0.81
-16.2
0.1.5
0.3
54
+ 0.81
+ 16.2
0.70
1.4
252
- 0.95
-19.0
0.20
0.4
72
+ 0.95
+ 19.0
0.75
1.5
270
-1.00
-20.0
0.2.5
0.5
90
+ 1.00
+ 20.0
0.80
1.6
288
-0.95
-19.0
0.30
0.6
108
+ 0.95
+ 19.0
0.85
1.7
306
-0.81
-16.2
0.35
0.7
126
+ 0.81
+ 16.2
0.90
1.8
324
-0.59
-11.8
0.40
0.8
144
+ 0.59
+ 11.8
0.95
1.9
342
-0.31
-6.2
0.45
0.9
162
+ 0.31
+ 6.2
1.00
2.0
360
0.00
0.0
0.50
1.0
180
0.00
0.0
VIBRATORY MOVEMENT 6
t
As an example we may take the curve y —'2>(i sin 27rQ-F ;
here the amplitude is 20"" and the period 0.5^ To express
this as a curve vs^e calculate the foregoing table.
The curve may be conveniently di-awn by plotting on
millimeter paper. On a heavy horizontal line a convenient
distance is chosen to indicate 0.10^. For example, let 60""
= 1.00s whence 3"" = 0.05% 6"" = 0.10% etc. For y
the most convenient arrangement is to let 1"" on the paper
represent 1"" of amplitude. At a; = 0 a point is placed on
the line for y —0. At a; = 3"" (representing t — 0.05^) a
point is placed just above y = 6"" (representing y = 6.2"").
In a similar manner the points of the curve indicated in the
table are marked on the paper and they are then joined by a
smooth line. The curve evidently continues to repeat the
same form, and further waves can be plotted directly from
the same table ; the result will be a curve like that shown
half-size in Fig. 2.
Fig. 2.
The curve crosses the horizontal axis whenever y = 0, that
is, wherever sin 27ry = 0; this occurs for sin 0°, sin 180°,
sin 360°, etc., that is (since it = 180°), for sin 0°, sin tt, sin 2m;
etc., or at the moments t = 0, t = \T,t ^ T, etc. The curve
will be at its maximum or minimum for sin 90°, sin 270°,
etc., that is, for sin^, sin -^ , etc., or at the moments
t = ^T, t = ^T, etc.
4 CURVES .OF SPEECH
The sine vibration is also expressed by «/ = « ■ sin 27rnt
where w = — - is the frequency, or the number of vibrations in
the unit time. The relation between T and n is readily un-
derstood by considering that when the particle makes a
complete vibration in ^ = 0.01", it must make n = 100
vibrations in one second.
It is also useful to know that the vibration may be ex-
pressed by
■ ^A
y^=a. sm y —t ,
^ 771
where s is the amount of the central force and m the mass of
the particle. The period of the vibration is thus
s
the period varies directly as the square root of the mass of
the particle and inversely as the square root of the strength
of the central force.
The condition of movement of the particle at any moment
is called its phase. Two vibrations are in the same phase
when the particles are at the same stage of the movement at
the same time. In the .case of the sinusoid curve the phases
are repeated at intervals of 27r (Fig. 2).
It is evident that the particle is moving with its least
velocity at the moment it changes its direction ; that
, . . TT Stt
IS, its velocity is zero at „, -^, etc. It moves with
its greatest velocity as it passes its center of equilibrium,
that is, at multiples of tt. The moment of greatest elonga-
tion is thus the moment of least velocity, and the moment
of no elongation is that of greatest velocity.
The vibration of a particle of air is communicated to the
neighboring particles and the disturbance travels in dry air
at the rate of 330.7™ per second at the sea-level pressure
of 760"™ and a temperature of 0° C. The successive
particles in a line of propagation will be in different phases
VIBRATORY MOVEMENT 5
of movement at any given moment (Fig. 3) ; the distance
between two particles in the same phase is known as the
Fig. 3.
wave-length. The wave-length I can be found from the
period Thy the known relation Z = 330.7 T, and from the
frequency n by
330.7
n
The wave-lengths are directly proportional to the periods
and inversely proportional to the frequencies. Thus,
the wave-length of a vibration with the period 0.01', or the
frequency 100, is I = 330.7°' x 0.01 = 3.307% and that
of a vibration with the period 0.02^, or the frequency 50,
is I = 330.7'" X 0.02 = 6.614°'.
A material point set in motion under the circumstances
described above would Continue to vibrate indefinitely about
its position of equilibrium if there were no forces of a dissi-
pative character to be overcome. There are, however, always
forces of this character which modify the movement ; such
forces can be grouped under the term friction.
The effect of friction is to reduce the motion of the particle
so that instead of vibrating indefinitely it gradually comes
to rest. The simplest supposition in regard to friction is
that its force is proportional to the velocity of the moving
particle. This supposition is in general well adapted to the
cases of actual experience. On this supposition it can be
shown that the amplitude of the vibration will steadily de-
crease at a rate expressed by dividing the initial amplitude
by the factor e*' where e is the number 2.71828, k the factor
of friction, and t the elapsed time. The amplitude at any
moment t will thus be -- or a . e~*', and the curve of vibra-
tion will be
y — a . e *'. sin 27r-— - ,
O CURVES OF SPEECH
where y is the displacement of the point at the moment
t, a the amplitude, e the constant 2.71828, k a factor
depending on the relation between the mass of the point and
the amount of the friction, and T^ the period under the given
circumstances. The period T^. depends only in a very slight
degree on the friction, and for nearly all purposes it can
be regarded as the same as the period without friction ; that
is, in general we may use I' (p. 2 ) for T^ , and we thus have
y = a. e~''* . sin 27r—
as the expression for the movement when friction is present.
The amplitude a is subjected to a steady decrease by the divisor
e*' , for in the expression a . e~^*- = ^ the amplitude will have
its greatest value only when A; = 0 or when there is no fric-
tion. Any friction will give a positive value to Ic and this
will reduce the value of a. When there is friction the
value of e*' will increase proportionately as time elapses;
thus a will be steadily reduced. A vibration with decreasing
Fjg. 4.
amplitude is shown in Fig. 4 ; the line along the tops of the
waves indicates the decrease according to the divisor e*'.
A model comprising a light flat wooden bob suspended
between two springs is of value in illustrating the laws of
vibration. The bob is moved to one side and then released ;
it executes vibrations of a definite period with slowly decreas-
ing amplitude. The effect of the mass of the particle can be
shown by adding lead weights to the bob and the effect of the
amount of the central force by increasing the tension of the
springs. A pair of light mica or cardboard vanes, adjustable
VIBRATORY MOVEMENT
7
at various angles to the direction of movement, render the
model available for illustrating the effects of friction.
Many of the phenomena of vibrating bodies can be accu-
rately studied in the action of vibrating springs or reeds. A
steel spring B (Fig. 5) clamped tightly in a small vise D on
the frame IJ bears at its end a recording point N of thin
steel ribbon. The frame also carries an adjustable electro-
magnet M clamped in place by L and a felt damper G ad-
justed as desired by the clamps F and H with their rod E.
Fig. 5.
The rod A is placed in a supporting standard (Fig. 6) which
is so adjusted that the recording point rests against the sur-
face of a smoked drum.
One of the most useful forms of a smoked drum is that
shown in Fig. 6. It is a light cylinder of metal with an axle
made so as to receive any of a set of pulleys or gears. A
sheet of glazed paper of the correct size to fit the drum re-
ceives a little paste along one edge ; it is stretched smoothly
around the drum and the paste-edge is lapped over. A gas
flame is then held close under it while the drum, with its
axle horizontal, is rotated against the direction of the flame ;
8
CURVES OF SPEECH
in this way an even coating of soot is deposited all over the
surface (light brown for ordinary work, black for reproducing
curves by photography).
Pig. 6.
Several of these drums may be smoked and kept ready for
use. Long strips of paper may be used over two drums. The
continuous-paper drum shown in Fig. 7 is suitable for very
long records. Two plates DE are held together by cross-
rods. At any points on the edges of these plates metal shafts
may be clamped, and two drums A,B with hollow axles
placed on them. A band of paper CC is fastened evenly
around the drums and tightened after the paste is dry by ad-
FiG. 7.
justing one of the shafts ; it is then smoked as usual. To
rotate the drums a loose pulley may be placed on one of the
shafts before or after the drum is on the support ; this pulley
VIBRATORY MOVEMENT 9
has a spring that catches one of the spokes of the drum.
The drums may be used horizontally as in Fig. 7 ; or vertically
by tipping the pair of plates on their edges.
Graphic records may also be made on a rotating glass wheel,
smoked over a candle flame. One form of such, a wheel,
shown in Fig. 8, is adapted for use with a projection lantern.
Fig. 8.
The shaft of the glass wheel W is connected to the pulley
shaft >S' by the bevel gear Q; the belt from the pulley P
passes to the motor. The legs AA are adjusted to fit over
the projection lantern in a manner to bring the lower part of
the glass wheel just in front of the condenser. The tracings
appear on the screen as they are made while the apparatus
itself can be seen through the smoked surface.
10
CURVES OF SPEECH
i
Tig. 9.
The drum is rotated by a small electric motor whose speed
is regulated by an appropriate resistance ; Fig. 6 shows both
a lamp resistance for large
I gJA changes in speed and an ad-
justable wire resistance for
smaller changes. A clock-
work motor may also be
used.
An adjustable counter-
shaft fastened to the base
of the motor allows the
speed to be reduced in
transmission. For very high
speeds the belt from a pul-
ley on the drum runs directly to a small pulley on the motor
axle. For more moderate speeds the countershaft is used
with a spur gear A on the motor axle and another B on the
countershaft (Fig. 9); the pulley C on
the countershaft runs at a lower speed
on account of the reduction AB. For
very low speeds a worm W is placed on
the motor axle and a worm gear F or a
spur gear on the countershaft (Fig. 10).
When the drum is used with its axis'
horizontal, a spur gear S (Fig. 11) may, if preferred, be placed
on its axle and made to connect with a spur gear T of any
desired size on the countershaft, which is run by a worm W on
the motor axle. For very low speeds
the spur gear S (Fig. 12) is run by a
worm X on the countershaft, which is
turned by the worm gear V in con-
nection with the worm W on the motor
axle. A collection of various sizes of
pulleys and gears makes it possible to
get almost any speed desired ; the finer
gradations are accomplished by the resistances and by slightly
pressing or loosening the motor brushes against the commutator.
Fig. 10.
Fig. 11.
VIBRATORY MOVEMENT
11
These types of recording apparatus are of constant use in
the most varied forms of phonetic work — the drums for
investigation and laboratory practice, the glass recorder for
class-demonstrations.
A blow on the spring B (Fig. 5) will cause it to draw a
sinusoidal line on the drum; the waves, however, slowly
decrease in amplitude owing
to loss of energy by friction
(p. 5). A quicker decrease
due to additional damping
can be obtained by placing
the surface of the felt dam-
per ((r, Pig. 5) more or less
tightly against its edge. A
curve of vibrations dying
away by friction due to damping was shown in Fig. 4 ; it was
made by the damped spring struck by a blow.
A vibratory body may receive a series of impulses. The
results of different natural periods of the vibratory point,
Fig. 12.
Fig. 13.
of frictional factors, of various strengths of impulse and of
different intervals of repetition, can be studied by means
of the vibrating spring. A series of impulses may be imparted
12 CURVES OF SPEECH
to the spring B (Fig. 5) by brief electric currents sent
through the magnet M. In a study of the action of such
impulses on a spring these impulses were obtained and re-
corded in the following way. A hard rubber contact wheel
C (Fig. 13) carried on its rim two pieces of metal £B'. A
pair of copper brushes H bearing against the rim were the
poles of a circuit through the magnet M (Fig. 5), indicated
by / in Fig. 13. As £ or £' passed across H, it closed the
circuit and sent a magnetic impulse to the spring. This
had the effect of a sharp blow. The strength of the blow
could be readily adjusted by varying the current or dis-
placing the magnet M. As it was desirable to have an
indication of the exact moment at which the impulse was sent
to the spring, a spark coil was made to register directly on
the line drawn by the vibrating point. A pair of copper
brushes G formed the poles of a circuit through the primary
coil i'' of a spark coil, whose secondary coil £ was connected
by the wires G to the metallic spring and the base of the re-
cording drum. A condenser D was connected around the
break at O. Whenever a metal piece £ or £' passed under
the brushes C, the circuit was closed. With an appropriate
adjustment of the current, a spark passed from the record-
ing point through the paper to the drum, removing the
smoke and making a white dot when the circuit was closed
and also when it was broken. The two pairs of brushes were
so adjusted that the sparks registered exactly the moments
at which the impulses were sent through the magnet and
those at which they ceased.
A record of an experiment in which the contact wheel was
revolved with steadily increasing rapidity is reproduced in
Fig. 14. The waves were drawn by the point iV (Fig. 5) ;
the pairs of dots marked the beginning and end of each
impulse. The figure shows that each impulse started a
vibration which died away by friction. If one impulse fol-
lowed the preceding one before the vibration was entirely
gone, its effect was increased or diminished according as the
phase of movement in which it occurred was the same as or
VIBRATORY MOVEMENT 13
opposed to the movement started by the impulse. When the
impulses occurred quite close, together and at exactly the right
phases, the summation of effects made the vibrations very
strong. In all such cases an increase occurred in amplitude
whenever the period t of the impulses became a multiple of
the natural period T of the spring. In all cases the
spring vibrated with the period T; only the amplitude was
affected by the variations of t.
Fig. u.
The condition of equal lengths of impulse could not be
illustrated with the arrangement just described, as the con-
tacts through B and B' (Fig. 13) lasted a constant fraction of
a revolution and the length of the impulse decreased propor-
tionately as the speed of revolution increased. The impulses
_ were weaker as they came faster. Nevertheless the increase
in amplitude whenever t was a multiple of I' appears strikingly
in Fig. 14. This increase in amplitude for harmonic relations
(that is, according to the simple ratios, 1:2:3: etc.) between
a natural period and impressed force is known as ' resonance.'
Resonance decreases with increase in damping, with devia-
tion from a harmonic relation and with distance apart in the
harmonic series.
The principles of resonance are of considerable importance.
Some of them can be readily illustrated by the use of acous-
14
CURVES OF SPEECH
Fig. 15'.
tical resonators. The spherical resonator (Fig. 15) is a hollow-
globe of glass or brass with a small opening arranged to fit
the ear and a larger one to receive the
vibrations of the air. When the reso-
nator is placed in the ear, it will be heard
to respond loudly whenever a certain
tone occurs in its neighborhood. This
tone is approximately the same as that
found, by tapping or blowing, to be its
natural tone. The adjustable resonator
(Fig. 16) can be made to answer to any tone within its'
limits. Resonators respond not only to vibrations of their
own period but in less degrees also to those of longer periods
in the harmonic series. They respond slightly or not at all
to vibrations not in this series. The natural period of a reso-
nator depends on its volume. The shape of the cavity is of
little influence. The size and shape of the opening are of great
effect. The resonance period of a bottle is altered by filling
it with water ; it does not change when the bottle is tipped,
but does change with any change of the size of the opening.
The resonance period of the mouth in speech probably de-
pends mainly on the size of the cavity and on the size and
shape of the labial, lingual and nasal apertures.
Owing to the unavoidable presence of friction all vibratory
bodies execute movements of decreasing amplitude unless the
Fig. 16.
loss of energy is replaced. For vibrating springs this may be
conveniently accomplished by having the spring regulate a
series of magnetic impulses sent to it ; this is the principle of
VIBRATORY MOVEMENT 15
the constantly used self-interrupting electric fork. One of
the many forms is shown in Fig. 17. The fork AA is held
in the block C by a nut R on its stem. The block 6< is on a
rod attached to the clamp D by which it can be adjusted on any
rod F. An electric current is brought to the binding post B
connected with the magnet coil M; it passes from the magnet
to the platinum disc P, from which it goes by a small piece of
platinum wire to the prong A and from another binding post
at Gr or elsewhere on the base of the fork back to the battery.
The current passing through the coil M makes it magnetic,
and the prongs AA are drawn inward. This movement
Fig. 17.
breaks the circuit at F, the magnetism ceases, the prongs fly
back, the circuit is again made, etc. Owing to self-induction
in the circuit the passing of the current through the magnet
is slightly retarded when the contact is made at Pas the prong
flies outward, and is somewhat prolonged when it is broken
at F as the prong flies inward. The pull of the magnet in-
ward is therefore somewhat longer when the prong is moving
in the direction of the pull than when it is moving against
it; this furnishes the extra energy required to compensate
the loss by friction and keep the fork vibrating indefinitely.^
When the point F is placed on the recording drum, it draws
a sinusoidal curve (Fig. 2) whose wave-length corresponds to
1 Rayleigh, Theory of Sound, § 64, 2d ed., London, 1894 ; Dvorak, Ueber
rerschiedene Arten selbstthdtiger Stromunterbrecher und deren Verwendung, Zt. f. In-
strumentenk., 1891 XI 423 ; Zusatz zu der Mittheilung, ' Ueber versch. Arten selbstth.
Stromunterbrecher,' Zt. Instrumentenk., 1892 XII 197.
16 CURVES OF SPEECH
the period of the fork. The period is lengthened by a rise in
temperature, but for the usual room-temperatures the differ-
ences are small. Increased pressure of N against P very
slightly shortens the period. The friction of the point B on
the record surface reduces the amplitude but does not appre-
ciably affect the period (p. 6).
References
For an elementary summary of the phenomena of vibration : Tyndall,
Sound, 5th ed., Loudon, 1893; Muller-Pouillet-Pfaundler, Lehrbuch
d. Physik, I, Braunschweig, 1886. For the mathematical treatment of
vibrations : Kayleigh, Theory of Sound, London, 1894 ; Helmholtz,
Math. Prinoipien d. Akustik, Leipzig, 1898. For the deduction of the
sinusoid: Tait and Steele, Dynamics of a Particle, § 88, London,
1878; Winkelmann, Handbuch d. Physik, I 692, Breslau, 1891 ; and the
usual works on dynamics. For apparatus to illustrate vibrations and
wave-movement: Frick, Physikalische Technik, I 566, Braunschweig,
1890; Wbinhold, Physikalische Demonstrationen, 3. Aufl., 219, Leipzig,
1899 ; MiJLLER-PouiLLET-PFAUNDLER, Lehrbuch d. Physik, 1 625, Braun-
schweig, 1886. For the technique of smoke records : Langendorff,
Physiologische Graphik, Leipzig-Wien, 1891; Scripture, New Psycho-
logy, Ch. V, London, 1897; New apparatus and methods, Stud. Yale Psych.
Lab., 1896 IV 76 ; Elementary course in psychological measurements, same,
108, 113.
For millimeter paper : Kayer, Paris ; Keuffel & Esser, New
York. For vibration model and recording drums : Chicago Labora-
tory Supply Co., Chicago. For recording drums and electric forks :
Zimmermann, Leipzig; Petzold, Leipzig; Heele, Berlin; Verdin,
Paris; SocriT:^ genevoise, Geneve. For forks with certificates of ac-
curacy from the Physikalisch-Teehnische Reichsanstalt in Berlin : Edel-
MANN, Miinchen; and the other German makers. For resonators and
electric forks : Kcenig, Paris. For resistances, switches, and other elec-
trical appliances for lamp circuits : Siemens & Halske, Berlin ; All-
GEMEiNE Elektricitats-Gesellschaft, Berlin ; Voigt & Hafner,
Bockenheim-Frankf urt a/M. For small motors and Lalande batteries :
Edison Manufacturing Co., Orange, N. J.
CHAPTER II
PHONATJTOGKAPH CTJEVBS AND MANOMBTEIC FLAMES
The first attempt at recording speech was made by Scott
in 1856. Scott was a proof-reader; noticing a picture of
the ear in the proof-sheets of a text-book of physics, he be-
lieved that he could get a record of speech by imitating the
structure of the ear.^ In Scott's phonautograph a large par-
abolic receiving trumpet carried at its end a thin membrane
whose movement caused a small recording lever to write upon
the smoked surface of a cylindrical drum. The sounds of
the voice passing down the receiver agitated the membrane and
caused the lever to draw the speech curve on the drum. The
instrument as improved by Kcenig was used by Dondeks
and others.^ It is the prototype of the later machines that
make speech records by registering the vibrations of a dia-
phragm on a moving surface by means of a lever.
The logograph of Baelow consisted of a trumpet or mouth-
piece ending in a thin membrane of rubber. A thin lever of
aluminum carrying a point dipped in color wrote the speech
curves on a band of paper.^
1 Scott, Inscription automatique des sons de I'air au moyen d'une oreille
artificielle, 1861 ; Phonautographe, Annales du Conservatoire des Arts et Metiers,
Oct. 1864; Phonautographe et fixation graphique de la voix. Cosmos, 1839 XIV
314 ; LiPPiCH, Studien iiber d. Phonautographen von Scott, Sitzb. d. Wien. Akad.,
Math.-naturw. Kl., 1864 L (II. Abth.) 397.
2 Bonders, Ueber d. Natur der Vokale, Arch. f. d. hoUand. Beitrage z. Natur-
u. Heilk., 1858 I 157 ; Zar Klangfarbe der Vokale, Arch. f. d. holland. Beitrage z.
Natur- u. Heilk., 1861 III 446 ; Zur Klangfarbe der Vokale, Ann. d. Phys. u. Chem.,
1864 CXXIII 527; De physiologie der spraakklanken, Utrecht, 1870; Schwan
tindPringsheim, Derfranzbsische Accent, Arch, f . d. Stadium d. neueren Sprachen,
1890 LXXXV 203.
2 Barlow, On the pneumatic action which accompanies the articulation of sounds
2
18 CURVES OF SPEECH
An improved phonautograph was used by Schneebeli ; ^ it
carried two points, one fixed to aid in comparison and the
other moving with tlie membrane. The inscription was made
on a strip of glass covered with a light coating of smoke and
drawn on a carriage rapidly in front of the recording points.
The tracings were measured with the aid of micrometric
screws. Schneebeli gave a number of the characteristic
curves of the vowels. Various similar methods have been
employed with constantly better results. The ear drum has
been used for the membrane by C. Blake.^ Preece and
Steoh used a thin membrane of rubber stretched by a cone
of paper. The cone was made to move a fine glass tube sup-
plied with an aniline ink, the record being taken on a band
of paper.^
Hensen's * phonautograph consisted of a membrane of gold-
beater's skin in a conical form produced by molding it over
a shape while moist and allowing it to dry before removal.
A single light lever attached to the center of the membrane
carried a fine glass thread as a recording point. It wrote the
curve on a thinly smoked strip of glass. The curves were
studied with a microscope.
Wendelee's curves,^ obtained with Hensen's phonauto-
graph, were observed through a microscope and copied by hand.
bi/ the human voice, as exhibited by a recording instrument, Proc. Roy. Soc. Lond.
1874 XXII 277 ; On the articulation of the human voice as illustrated by the logo-
graph, Proc. Roy. Soc. Dublin, 1880 N. S. II 153.
1 Schneebeli, Experiences avec le phonautographe. Arch, des Sciences phys. et
nat. de Geneve, 1878 (Nouvelle pe'riode) LXIV 79 ; Sur la the'orie du timbre et
parliculierement des voyelles. Arch, des Sciences phys. et uat. de Geneve, 1879 (III.
periode) I 149.
2 Blake, The use of the membrana tympani as a phonautograph and logograph
Archives of Ophthal. and Otol., 1876 V No. 1.
3 Pkeece and Stroh, Studies in acoustics, Proc. Roy. Soc. Lond., 1879
XXVIII 358.
* Hensen, Ueber die Schrift von Schallbewegungen, Zt. Biol., 1887 XXIII 29] ■
first described by Griitzner, Physiologic d. Stimme u. Sprache, 187, Hermann's
Handb. d. Physiol., I. Bd., II Theil, Leipzig, 1879.
^ Wendeler, Ein Versuch d. Schallbewegung einiger Consonanten u. anderer
Gerdusche mit d. Hensen'schen Sprachzeichner graphisch darzustellen, Diss. Kiel
1886 ; also in Zt. f. Biol., 1887 XXIII 303.
PHONAUTOGRAPH CURVES 19
' The curve of r was found to consist of small vibrations
with rather regular fluctuations of amplitude having long
periods ; the resemblance to the familiar curves of two tones
forming beats suggested the term ' pseudobeats ' for the
fluctuations of intensity observed in the r-curves. The curves
of r were similar to those afterwards obtained by Hermann
(see Chap. Ill below). The terminal portions of r were
thought to resemble the curves of the adjacent vowels ; the
r was defined as a rhythmically repeated weakening of the
vowel sound of the syllable to which the r belonged.
Wendeler also noted the vowel-like character of the
curves of 1, m and n. The vibrations in a vowel curve were
observed not to remain constant in form but to undergo
gradual changes ; these he attributed to changes in the pitch
of the resonance tone, assuming that the cord tone remained
constant. This view is inadequate. These changes should
be attributed rather to the changes in the relations among the
various resonance tones and the cord tone ; they appear in
all vowel curves and indicate that a vowel sound never re-
mains constant. I believe it is safe to say that a vowel
cannot be treated as a sound whose character is the same
throughout its duration. The various component tones are
continually changing both in pitch and intensity, and it is
highly probable that every typical vowel has typical forms
of change, and that these forms of change are as important
characteristics as the pitches and intensities of the component
tones.
Martens,-' with curves obtained by Hensen's phonauto-
graph, observed and charted the changes in the cord tone in
speech. The cord tone is continually changing, and it may
be said that a vowel is not spoken on a tone of a cer-
tain pitch but on a slide that may be quite extensive and
complicated.
Martens, like Wendeler, observed gradual changes in
the form of the vowel curve ; but he attributed them to the
1 Maktens, Ueber das Verhalten von Vohalen und Diphthongen in gesprochenen
Worten, Diss., Kiel, 1888 ; also in Zt. f. Biol., 1889 XXV 289.
20 CURVES OF SPEECH
changes in the cord tone. As stated above, the changes are
due to the changes in the relations between the various tones
of the vowel.
Curves of the German diphthongs au and ai showed grad-
ual changes extending throughout the sound. Such curves
show that the diphthong recorded is not to be considered as
the sum of two sounds united by a glide but as one sound of
a changing character, y It is undoubtedly true that in ai, for
example, various stages in the development of the first por-
tion might be picked out that resemble various forms of the
a-sound, and that similarly various forms of the i-sound
might be picked out at various moments in the later por-
tion; possibly resemblances to even other sounds could be
found at various moments. The attempt to lump the whole
effect into two portions leads to conflicting opinions, of which
all are partly, and none completely, correct. Diphthongs of
this character differ from long vowels only in undergoing
somewhat more change. The long vowels themselves are
not constant ; a sharp division between diphthongs of this
kind and long vowels cannot be made. The tendency of
long vowels to diphthongize is a familiar phenomenon ^ that
has made some of them, like o and e, as strongly dipthongal
as au and ai in Southern English.^ The unification of the
elements of a diphthong may go so far that it becomes, in the
history of the sound, a long vowel; it depends on circum-
stances ^ just what portion of the diphthong will remain
as the fundamental portion of the vowel.
An improvement was made in Hensen's recorder by Pip-
ping, 'who replaced the glass thread by a small diamond
which scratched the curve directly on the glass strip. With
this instrument Pipping has made a series of investigations,
chiefly on the vowels.* The curves on the glass strip were
measured with a microscope and analyzed.
1 Sweet, History of English Sounds, § 63, Oxford, 1888.
2 SOAMES, Introduction to Phonetics, § 87, London, 1899.
^ Sweet, as before, §§ 70, 71.
* Pipping, Oni Uangfdrgen hos sjunga vokaler, Diss., Helsingfors, 1890; also
PHONAUTOGRAPH CURVES
21
The curves of the Swedish vowels sung and spoken by
Pipping were found ^ to contain resonance tones that lay
for a at about g^ and c^; for e^ (Swedish close e) about/^
/3« and e*« ; for i about d?-, c** and /*» ; for o about / ; for
u about cP- to /I and about <F ; for y about cl^ and e* ; for o
(a) about b^ ; for . e^ (a) about g'^ and f^ ; for ccj (close 6)
about /I and g^ ; for oBj (open 6) about e^ and d^* . In musi-
cal notation these results are :
-OJf_
^:
62 OBj CE2
The Finnish vowels ^ found in the words 'Aamu,' '^erik,'
'vMsas,' 'taloon,' 'kitMsi,' 'p«/y,' 'saa,' and 'Toolo' were all
sung on g'^ by two tenor-barytones, one bass-barytone and two
bass voices, also all on c" and on 5'~-'*, and e again on g'^ by
one of the bass voices. A registration of words spoken by one
of the tenor-barytone voices included ' satama,' ' saadaan,'
'kuopio,' 'houreet,' etc.
The eight vowels sung on g~^^ showed a maximum reso-
nance near ^"S or c^- Since the Swedish vowels i, y and a
tested by resonators (p. 14 ) also showed a maximum reso-
nance neare^ and the eight German vowels of Hermann
as Zur Klangfarbe d. gesungenen Volcale ; Untersuchung mit Hensen's Sprachzeich-
ner, Zt. f. Biol., 1890 XXVII 1 ; Nachtrap zur Klangfarbe der gesungenen Volcale
Zt. f. Biol., 1890 XXVII 433 ; Om Hensen's fonautograf som ett hjalpmedel fOr
sprakvetenskapen, Helsingfors, 1890 ; Fonautografiska studier, Knlandska Bidrag,
till Svensk Sprak- och Folklifsforskning, 99, (Helsingfors, 1894) ; Zur Lehre
V. d. Vokalklangen, Zt. f. Biol., 1895 XXXI 524; Veher d. Theorie d. Vokale,
Acta Societatis Scientiarum Fennicae (Helsingfors), 1894 XX No. 11 ; Zur
Phonetik d. Jinn. Sprache, Unters. mit Hensen's Sprachzeichner, Ij/lem. de la Soc.
finno-ougrienne, XIV, Helsingfors, 1899.
1 Pipping, as before, Zt. f. Biol, XXXI.
2 Pipping, as before, Me'm. Soc. flnno-ougrienne.
22
CURVES OF SPEECH
(Chap. Ill) gave like results, Pipping concluded that this reso-
nance tone must arise from some portion of the acoustical
apparatus not influenced by the movements for the different
vowels. This constant resonance was attributed by Pipping to
the chest cavity. Since the tones to which the chest resonates
best in a male adult are nearly an octave below c^ and since
the capacity of the chest cavity is continually changing during
respiration, this supposition can hardly be accepted ; the res-
onance may preferably be attributed to the trachea which is of
a fairly constant capacity with an aperture adapted to such
a tone.
Pipping's computations and interpretations of the Fin-
nish vowels gave three resonances, which he attributed to the
chest, throat and mouth in the order of rising pitch as
shown in the accompanying table (a [y] indicates a when
followed by y~).
Chest
Throat
Mouth
a (a) sung by bass voice
gn
61 - c2
c3
62 (a) '■
'*
6=
O (0) "
ji»
a2_c3
oe (0) ■'
"
e3
ei (e) "
."
p~rt
n (m) "
about q'''%
p~an
y (y) "
P -/"«
i (i) "
^3_^3J
Chest
Throat
Mouth
a (a) spokeu
by tenor-barytone
rfi
p-r-i
rf3 _ s%^
62 («) ^ "
" "
dn - e2
Pt-g"
62 (aW) "
pt
9^
O (0) "
6lJt - c2
a2-c3
oe (6-) "
U li.
61 -c2
9^-gH
ei (e) "
a>%
a3-63
U («) , "
et if
about rfi
g2 - 62
y to) "
tt (t
"
^ ao
i (0 "
(t
a
an - c4
PHONA UTOGRAPII CUR VES
23
In musical notation the results are :
t -P- E t
i
i
E
ta
m
^ ^— "JL T X- 1: i:: ^
I I' I I L J I L L I L I I fc- I IL. I II. } II. I
Ba9S
If:
i^r
-:t-
t t
f: £
^
^^
=«hi
e2[y] o oe
Tenok-Bakytone
The vowels sung by one of the bass voices N and the
vowels spoken by one of the tenor-barytones E — as given in
the list above — ' form in a way the extremes between which
the other investigated sounds lie. Trustworthy exceptions
are almost exclusively the e2-sounds of W and L. The e^ of
W has a throat resonance /^* which otherwise occurs only
in 62 [y]. The e^ of L has a still lower mouth resonance, d^*,
than that of N, e^. The mouth resonance in u of J. appears
to lie below the lowest step otherwise found, f^ \ perhaps this
is also the case in u of W.
' As long as we confine ourselves to one individual and do
not compare sung vowels with spoken ones, the vowels may
be easily distinguished by their resonance tones. But when
we consider sung and spoken vowels of different persons, the
variations become considerable, so that the ranges of fluctua-
tion of different vowel resonances not seldom overlap. The
i of iV has a mouth resonance that is not only deeper than
the i resonance of U but is in fact deeper than his y reso-
nances. Since the vowels i and y belong to the same group
in respect to throat resonance, we may ask in what they are
to be distinguished from each other. The vowels oe and e
cannot be confused with y and i on account of the higher
24 CURVES OF SPEECH
throat resonance, but how does it happen that the mouth reso-
nance of e with iVlies lower than the resonance of oe with
U ? Can the resonances of a sink to c^ and e^ (iV) without
the vowel changing into an o ? ' Pipping's answer is that
the vowels that are designated by the same name are really a
group differing considerably from each other. Thus, the a of
a bass voice at a low pitch resembles an o. Again, according
to Pipping, the pitch of the maximum resonance tone is not
the only essential to the character of a vowel ; the range
through which the mouth can resonate to the impulses from
the cords is a special characteristic of each vowel.
Pipping found that in the word ' houreet ' the number of
flaps of the r was always 3, and in ' kiura ' 2 ; in ' houreet '
the resonance tone e^ during the r differed from the preceding
ff^ for the u and the following g^ for the i ; in ' kiuru ' the
resonance tone during the r was lower than that for the pre-
ceding and the following u. The r in these cases seems to
have been different from that of Wendeleb (p. 19). In the
word ' keihaita ' there was no interruption of the cord vibra-
tions at the time of the h; it was thus a sonant h, corre-
sponding probably to the soiiant h of the Sanskrit grammarians
and of later observers ^ and seemingly related to the usual whis-
per sound which has been shown to possess a slight degree
of sonancy.^'
The hindrances due to the inertia of material levers and to
the friction of a recording point were avoided by E. W. Blake,
who attached a mirror to a telephone plate in such a way that
a beam of light was deflected by each movement. A ray of
light from a heliostat was reflected through lenses upon a
photographic plate moving with a constant velocity. The
sound wave recorded a line on the plate.^ Rigollot and
Chavanon covered the wider end of a paraboloid with a
very thin membrane of collodion, to the center of which was
1 Meyer, Beitrage zur deutschen Metril; numbers 13 and 15 of Tofel, Neuere
Spraclien, 1899 VI 1 ; Metek, Stimmhaftes H, Neuere Spracheu, 1900 VIII 261.
2 Olivier, De la voix chuchote'e, La Parole, 1899 I 20.
^ Blake, A method of recording articulate vibrations ly means of photographu
Amer. Jour. Sci., 1878 XVI 55 ; also in Nature, 1878 XVIII 338.
PHONAUTOGRAPH CURVES
25
fixed a small mirror working on an axis of fine thread.^ The
ray of light was thrown on a rotating mirror and observed
on a screen. Lebedeff substituted a membrane of cork in
a similar apparatus.^
Samojloff^ used a phonautograph with a 1"™ thick cork
membrane to which a lever with a mirror was attached. The
deflections of the mirror were recorded by a ray of light fall-
FlG. 18.
ing on a photographic plate. The characteristic tones of the
vowels (Russian) were found to vary from a^ to g^ for a, h^ to d^
for o, c^ to g^ and c^ to e^ for u, h^ (?) to (P (?) and V to d^ for
e, ci (?) to g^ (?), and finally e^ (?) to e^ (?) and cZ* to e* for i.
A modification of the phonautograph idea is found in the
magnetic and carbon transmitters of the telephone and in the
various voice keys. These are used mainly for determining
1 RiGOLtOT ET Chavanon, Projection des ph^nomenes acoustiques, Journal de
physique, 1883 (2) II 553.
2 Lebedeff, Journal d. russ. physik.-chem. Qes., 1 894 XXVI 290, mentioned by
Samojlopf, Zur Vokal/rage, Arch. f. d. ges. Physiol. (Pfiiiger), 1899 LXXVIII 4.
" Samojlofp, Zur Vokalfrage, Arch. f. d. ges. Physiol. (Pfiiiger), 1899
LXXVIII 1, 27 ; Graphische JDarstellung d. Vokale, Physiologiste Russe, 1900 II 62.
26
CURVES OF SPEECH
the tone from the vocal cords; accounts will be found in
future chapters.
The manometric flame method was devised by Kceitig.^
'The vowel is sung or spoken into a trumpet leading to a
small box known as the ' manometric capsule.' This box is
divided in two parts by a thin rubber membrane. The details
of its construction are shown in Fig. 18 A. One part is a
!^^M '(MIMMi MMk MiM ^M
MMA '^Mk MMk 'USM ^ttl
'MMiiM ^/iA/yi 4iSi4,iiii MMid ^^^fefej
i$i$,^ dddLMm i$Si0d %MMM ^MM^:
jkMM. ^M^ 'iMdsdi (jiMkiL §MMi
§MMMi^ d^kA SSM 'udlMdhM MMiii
Fig. 19.
tight chamber through which illuminating gas is flowing ; the
gas is lighted at the end of the small jet. As the sound waves
descend they strike the rubber membrane, set it in vibration
and thus produce movements of the gas analogous to those of
the air in the sound waves. The vibrations of the flame can be
seen when the eye is suddenly moved sidewise ; owing to the
lag of sensation the image remains in the eye and the succes-
sive vibrations appear simultaneous. They are conveniently
1 KoENiG, Die manometrischen Flammen, Ann. d. Phys. u. Chem., 1872 CXLVI
161; Quelques experiences d'acoustique, 56, Paris, 1882.
MANOMETRIC FLAMES 27
studied, like other vibrating bodies, by means of revolving
mirrors ; the best form is a cube with mirrors on four sides set
in rotation by a handle (Fig. 18 M). The curves of the five
French vowels obtained with this apparatus and carefully
drawn by Kcenig are shown in Fig. 19. The pictures of
m and n (no difference detected) are given in Fig. 20.
The picture of r is given in Fig. 21.
Tig. 20.
The manometric flames can be photographed ^ by selecting
the right composition of the illuminating gas. Cyanogen gas
has been used. A mixture of hydrogen (or ordinary illumi-
nating gas) and acetylene burning in a chamber of oxygen
gives brilliant flames.^ The first two gases are mixed in a
1 Stein, in Maret, La methode graphlque, 647 ; Doumer, Mesure de la
hauteur des sons par lesjlammes manometriques, C. r. Acad. Sci. Paris, 1886 CIV
340 ; Hitudes du timbre des sons, par la methode des flammes manometriques, C. r.
Acad. Sci. Paris, 1887 CV 222 ; Des voyelles dont le caractere est tres aigu, C. r.
Acad. Sci. Paris, 1887 CV 1247 ; Hallock, Photography of manometric flames.
Physical Eeview, 1895 II 305 ; Marage, Studes des voyelles par la photograpkie ■
■des flammes manometriques, Bull, de I'Acad. de Med., 1897 XXXVIII 476.
^ Mekritt, On a method of photographing the manometric flame, with applications
to the study of the vowel A, Physical Review, 1893 I 166^; Nichols and Merritt,
Photography of manometric flames, Physical Eeview, 1898 VII 93.
28
CURVES OF SPEECH
Fig. 21.
i^
tank and supplied to the capsule M by the tube (?, Fig. 22.
The jet AA, with a platinum tip B, is fixed in an outer
tube T which re-
ceives oxygen at 0.
The flame F thus
burns in an atmos-
phere of oxygen with
a strongly actinic
light. The image
of the flame is fo-
cused on a photo-
graphic plate in a
camera. When this plate is moved rapidly sidewise, the
flame traces a curve showing its vibrations. Mbeeitt's re-
sults with a number of records of the vowel a
sung by different voices on different notes,
varying from a frequency of 102 with a bass
voice to 667 with a soprano, showed a reso-
nance tone averaging 736, practically inde-
pendent of the pitch of the voice.
In the experiments by NiOHOLS and Merkitt the image
of the flame was focused upon a sensitive celluloid film
mounted on a wheel D with
a circumference of nearly
1™ inside a light-tight box
(Fig. 23). The wheel was
rotated at a surface speed
of 1" in a second. The vi-
brations of the gas flame F
were thus photographed on
the film. The curves may
be reproduced by a photo-
gravure process ; as the
reproductions sometimes
shrink, the measurements
should be made in the original films. I have found it con-
venient to print the curve on blue prussiate paper, trace the
Fig. 22
MANOMETRW FLAMES
29
outline in Chinese ink and clear off the blue color by washing-
soda. Fig. 24 shows o of bo reproduced in this way.
Nagel and SamojlofpI used the ear in the head of a
freshly killed animal as a manometric capsule. ' The gas was
\
\
/
/
1
I
\
\
\
\
/
KiG. 23.
passed through the Eustachian tube into the middle ear and
out through a hole in the bone ; it burned in a small flame at
the end of a platinum point. The tympanic membrane formed
the membrane between the two compartments. On speak-
ing into the outer ear the sound waves agitated the tympanic
membrane, and this caused the flame to vibrate exactly as in
the usual manometric capsule. The flames have been photo-
graphed.2
Fig. 24.
In concluding this chapter it is necessary to consider a
few facts that concern all speech-recording machines.
Records on such machines have certain advantages over
those obtained by instruments attached to the speaker, as the
vocal organs act with less interference. When speaking
into a machine, a person may depart to a certain extent from
1 Nagel tjnd Samojloff, Einige Versuche ii. d. Uebertrayung v. Schallschwing-
ungen aufd. Mittelohr, Arch. f..Anat. u. Physiol. (Physiol. Abth.), 1898, 50.5.
2 Samojloff, Zur Vokalfrage, Arch. f. d. ges. Physiol. (Pfliiger), 1899
LXXVIII, Tafel I.
30 CURVES OF SPEECH
his conversational voice unless the records are made with the
proper precautions. It is highly desirable to have no one in
the room except the experimenter and the subject, as a
person, even in ordinary conversation, has a tendency to
speak somewhat differently when he feels himself observed.
With some informal conversation beforehand, almost any
person can be so put at his ease that when he turns to
speak into the phonautograph or phonograph he feels quite
at home and does not change his voice in any way. Much
experience with the phonograph has shown that it requires
only a little knack to put people at ease with the machine
and lead them not to think of it. The result may be com-
pared to that of the views of a kineto-camera (cinematograph)
taken when the person is not thinking about it. An un-
skilful or nervous experimenter may get results to be com-
pared rather with the usual photograph; still, even such
results give the fundamental facts in good approximation to
the truth. With the latest form of gramophone apparatus,
records can be made while the subject is unconscious of the
fact.
It is sometimes said that the speech machines do not
faithfully record the vibrations of the air; this is true to
a certain extent. Tlie speaking-tube and the diaphragm
reinforce or weaken some of the tones, but the influence is
chiefly on the very high ones. The modification of the
finer details has been studied in a series of curves of the same
sound made with different diaphragms. ^ The friction of the
recording point greatly reduces the size of the vibrations
and modifies them chiefly by rounding off the corners. The
inertia of the recording levers also has some influence.
Just how much detail has been lost and just how much dis-
tortion has occurred cannot be known with machines of the
phonautograph type because the records cannot be turned
back into speech. It is in any case a question of the degree
of approximation, which is to be discussed and specified as
1 Hermann, Phonophotographische. Untersuchungen, I., Arch. f. d. ges. Physiol.
(Pfluger), 1889 XLV 582.
MANOMETRIC FLAMES 31
in all scientific work. No measurements can ever be exact;
the progress in accuracy consists in increasing the degree of
approximation. The degree of approximation in the Hensbn
phonautograph' is far greater than in that of Scott ; that of
the manometric flame is still unknown. In any case it is a
matter of scientific detail, and the remark sometimes made
to the effect ' that most of the characteristics of a speech curve
may be due to the apparatus ' shows a lack of comprehension
of instrumental methods equalled only by that of a critic
who supposes the phonograph to be able mysteriously to add
the very strongest tones to a record and — practically — ^to
be able to change a vocal solo into an orchestral performance.
The student of experimental phonetics should endeavor to
learn the degree of accuracy in the action of each part and
of the whole apparatus and should ever bear in mind
that it is a machine whose every action is a matter of
mechanics.
Repeeences
For the Scott phonautograph and for mauometrio flame apparatus :
KcENiG, Paris. For the Hensen phonautograph : Zwickert (care of
Prof. Dr. V. Hensen), Kiel.
CHAPTER III
PHONOGRAPH EECOEDS
The original machine for reproducing speech seems to have
been the phonograph of Edison.^ The tin-foil phonograph was
afterwards superseded by the wax-cylinder form (Figs. 25, 26).
When a person talks into the speaking-tube of the phono-
graph, the air vibrations from the mouth are hindered from
spreading and are conducted along the
tube to the recorder at the further
end. In the recorder (Fig. 27) they
pass down the channel E to the thin
diaphragm of glass A held in the
frame F between thin rings of rubber
J. This diaphragm follows the pres-
sure by bending and possibly in some
cases by moving as a whole also. A
metal head D cemented to the center
of the diaphragm transfers the vibra-
tory movement by means of the link
G to the lever B which ends in the
sapphire recording knife K. The recording knife cuts a
groove in a surface of a special wax composition. The ac-
tion of the sapphire knife in cutting the surface of the cylin-
der is shown in Fig. 28. The depth of this groove depends
on the vibrations of the diaphragm. A portion of the groove
of the vowel a in the Dutch word 'daar' by Boeke^ is
shown in Fig. 29. The vibrations should never be so o-reat
as to remove the knife entirely from the wax. The work of
1 Patent of November, 1877.
2 BoEKE, Mikroskopische Phonogrammstudien, Arch. f.d. ges. Physiol. (Pflugerl
1891 L297. '
PHONOGRAPH RECORDS
33
cutting the wax naturally modifies the movement of the
knife. With the best wax composition this modification con-
sists almost entirely in a reduction of amplitude.
Fig. i:ti.
To reproduce the sound the recorder is replaced by a repro-
ducer (Fig. 30) which has a round sapphire point instead of the
sapphire knife, but which is otherwise closely like the
recorder in structure.
The rotation of the cylinder causes the round point to
follow the rise and fall of the groove, whereby it repeats the
movement of the recording knife. The glass diaphragm
connected to the reproducing point
is set in vibration and arouses air vi-
brations similar to the original ones in
the spoken sounds.
The changes of the cylinder under
the fluctuations of temperature are so
great that the reproducer must have
considerable sideplay in order that the
stylus may keep in the groove. The usual relation between
the size of the cutting edge of the recording sapphire, 1™™,
and the diameter of the spherical reproducing sapphire, 0.9™™,
has been shown to be the most favorable one.^
The accuracy with which the phonograph reproduces the
^ Hermann, Forlgesetzte Untersuchmgen liber die Konsonanten, Arch. f. d. ges.
Physiol. (Pfluger), 1900 LXXXIII 6.
Fig. 27.
34
CURVES OF SPEECH
original sound depends on a series of factors tha;t need care-
ful consideration.
In the first place the sound may be modified by the speak-
ing-tube. A narrow cylindrical tube with perfectly hard
Fig. 29.
Fig. 28.
walls will conduct the vibrations practically unchanged in
character, except when harmonic — or resonance — relations
(p. 14) occur between the natural period (p. 2) of the tube
and the periods of the voice vibrations ; a very short tube has
such relations only to the high
partials of the voice. Most
voice records are made with a
flexible mohair tube. A soft-
walled tube is unfavorable to the reflection of very high
tones ; this is possibly the reason why some voice records
seem ' muffled ; ' in such a case a metal horn, conical or
flaring, may be tried, the best one being a perfectly conical
horn 20 to 26 inches long and not more than 6 inches across
the end. The speaking-tube is a necessity as the vibrations
of the diaphragm would otherwise be too weak.
The glass diaphragm of the recorder is
from 0.003 in. to 0.009 in. in thickness
(0.08°™ to 0.23"'"). The thinner the dia-
phragm, the more sensitive it is. A loud
voice makes the diaphragm execute such
great excursions that the recording knife
leaves the wax at times ; this results in a rattling sound,
or 'blast,' when the record is reproduced. Some voices
produce rattling records with a certain diaphragm, but
not with a thicker one or a thinner one. If the rubber
Fig. 30.
PHONOGRAPH RECORDS 35
washers become hardened, the vibration of the diaphragm
will be hindered. The washers should not be pressed too
tightly or too loosely -by the screw ring / (Fig. 27); in
the former case the diaphragm will lose sensitiveness, in
the latter it will rattle. If the metal head D is not firmly
fixed, the record will give a dull, raspy sound. To re-
move this head from a diaphragm, allow one drop of
water to remain upon it for ten or fifteen minutes, remove
the screw on the edge and lift up the weight G. If the
glass diaphragm has been broken, the metal head should
be scraped clean. Both head and glass should be thoroughly
cleaned with benzine. To fasten the head apply a very small
drop of stratena to it and lower it upon the diaphragm ex-
actly in the center ; if it is found to twist the link C when
dry, loosen the ring I and turn the diaphragm.
The best glass diaphragms for reproducing are 0.004J,
0.005, 0.0051 and 0.006 inch in thickness. They should be
of uniform thickness. It is well to try several sizes when
first adjusting a phonograph.
The sapphire recording knife should be handled with great
care ; in order to preserve its sharp even cutting edge it
should never come into contact with any hard surface. It
should be examined with a magnifying glass, if the records
seem to be in any way poor.
A diaphragm favors certain tones by resonance (p. 30).
A diaphragm whose lowest period of free vibration (p. 2) is
very high is able to favor only the very high partials of a
note. For this reason a stiff or tightly stretched diaphragm
records with greater accuracy. The strong damping caused
by the knife cutting in the wax eliminates most of the influ-
ence of the free period of the diaphragm in the phonograph.
The wax cylinder is composed mainly of stearate of soda.
To make the surface of the cylinder as smooth as possible, it
is shaved by runniiig it at a very high rate of speed and turn-
ing it off by a sapphire shaving knife. Chips and dust are
cleared off by a camel's hair brush. The surface should
never be touched by the finger or blown upon. Any rough-
36 CURVES OF SPEECH
ness of the surface will appear as a strong noise in the record.
Large cylinders take better records than small ones ; owing
to the greater speed at which the surface travels under
the knife, the waves in the groove are more extended and
the sapphire ball of the reproducer can more readily follow
the finer fluctuations.
The machinery of the phonograph should be well oiled,
free from dust and in perfect order, as any noise from friction
transfers itself to the record. Records taken in a perfectly
bare room have a loud resonant quality; those taken in a
room with a few hangings are mellower but fainter. In
making a record the speaker should stand or sit immediately
in front of the instrument and should speak directly into it.
The articulation should be distinct but natural. In the case
of singing, the head should be drawn back when very high or
loud notes are sung. The speaker should be put at ease as
much as possible (p. 30).
The final test of the truthfulness of a record is made by
hearing it. The words spoken by the machine represent
what is on the record with close approximation. When a
record is found that speaks clearly in a natural voice, it can
be trusted for what it says, since it cannot say anything more
than is on it and cannot improve its own tracing. The
speech represented by a tracing from a record is the speech
of the record itself. How nearly this reproduces the original
speech can only be determined by comparing it by the ear
with the words of the original speaker. By skilful manip-
ulation records can be made whose speech cannot be distin-
guished from that of a living person except by their weakness
and by the scratching noise due to the friction of the tracing
point in the groove. Curves correctly traced from such a
record give exactly the curve of the speech spoken into it.
From the known velocity of rotation of the cylinder, lengths
on the surface can be translated into time. The rotations
may be counted for a number of seconds. For m rotations
in n seconds the time of one rotation is T= n/m seconds.
For a diameter of d millimeters the circumference is ird
PHONOGRAPH RECORDS 37
(tt = 3.1416). The time represented by l""-" is Tjird. Thus, if
the cylinder rotates 122 times in 60 seconds, T= 0.49', a figure
that is accurate owing to the long time during which the rota-
tion was counted. With an outside diameter — measured by
calipers — of 53°™ the circumference (use table for ttc?) is
53 X 3.1416 = 229.3363°'"'- The time represented by 1°™ cir-
0 49
cumference is then '^ = 0.0021', a figure that for a good
phonograph is generally reliable to half a thousandth of a sec-
ond when the spring is kept at about the same tension or the
motor is run by a tested storage battery. The constancy of the
speed can be roughly tested by sounding into the recorder from
time to time the note of some musical instrument of known
pitch, for example, a telephone connected with an electric fork.
The tracings on the phonograph have been observed through
^ microscope and sketched.^
The groove of a phonogram may be conveniently examined ^
by a corneal microscope, having a side illuminating tube that
concentrates the light of a small incandescent lamp on the rec-
ord. Measurements may be made by an ocular micrometer.
Measurements of phonograph tracings have been made
by BoEKE.^ The widths of the grooves were measured by a
microscope with an ocular micrometer; the shape of the
cutting surface of the recording sapphire being known, the
depths could be at once calculated.
The records on the phonograph cylinder may be enlarged
by amplifying levers recording on a smoked drum.*
1 Marichelle, La parole d'aprfes le trace du phonographe, Paris, 1897;
Mariohelle et Hemardingee ; Etudes des sons de la parole par le phono-
graphe, C. r. Acad. Sci, Paris, 1897 CXXXV 884; Gelle, L'audition, Paris, 1897.
2 Meyer, Zur Tonhewegung d. Volcals, Nenere Sprachen, 1897 IV, Beiblatt.
^ Boeke, Mededeeling omtrent onderzoehingen van klinkerindrushels op de was-
rollen van Edison's verheterden fonograaf, De natuur, 1890, July; Mikroslcopische
Phonogrammstudien, Arch. f. d. gas. Physiol. (Pfliiger), 1891 L 297 ; Mikroskopische
Phonogrammstudien, Arch. f. d. ges. Physiol. (Pfliiger), 1899 LXXVI 497;
M'Kendrick and Gray, On vocdl sounds, Schaefer's Text Book of Physiology,
II 1227, 1229, London, 1900.
* Mayer, Edison's talking machine. Nature, 1878 XVII 469; FiCK, Zur
Phonographik, Beitrage zur Physiologie Ludwig gewidmet, 23, Leipzig, 1887;
38 CURVES OF SPEECH
Highly accurate tracings of phonograph records have been
made by Heemann.i The desired record is made on the
phonograph and tested by reproducing. Owing to the
changes of the wax with the temperature, the tracing must
begin immediately; to keep the centering the wax must
not be removed from or displaced on the metal barrel. In
place of the reproducer a system of levers is so brought
to bear that a fine glass knob travels in the record and de-
flects a mirror with every movement. For consonants three
successive levers are used, for vowels only two levers. It
is well to have the bearings jeweled.
Above the system of levers a weak convex lens is fastened
with its center exactly over the mirror. The phonograph is
Jenkin and Ewing, The phonograph, and vowel theories. Nature, 1878 XVIII 167,
340, 394 ; On the harmonic analysis of certain vowel sounds, Trans. Roy. §oc. Edinb.,
1878 XXVIII 745 ; Lahk, Die Grassmann'sche Vokaltheorie im Lichte des Experi-
ments, Diss., Jena, 1885 ; also in Ann. d. Phys. u. Chem., 1886 XXVII 94 ;
Wagner, Ueber d. Verwendung d. Griitzner-Marey' schen Apparats u. d. Phono-
graphen zur phonetisclien Untersuchungen, Thonet. Studien, 1 890 IV 68 ; M'Ken-
DKiCK, On the tone and curves of the phonograph, Jour. Anat. and Physiol., 1896
XXIX 583 ; M'Kendrick, Murkay and Winoate, Committee report on the
physiol, application of the phonograph and on the form of the voice curves hy the in-
strument, Kept. Brit. Ass. Adv. Sci., 1896 669 ; M'Kendrick, Observations on the
phonograph. Trans. Roy. Soc. Ediu., 1897 XXXVIII 765; Demonstration of an
improved phonograph recorder, Fioc. TS,oy. Soc. Edin., 1896-97 XXI 194; Sound
and Speech Waves as revealed by the Phonograph, London, 1897 ; M'Kendrick
AND Gray, On Vocal Sounds, Schaefer's Text Book of Physiology, II 1229,
London, 1900.
1 Hermann, Phonophotographische Untersuchungen, I., Arch. f. d. ges. Physiol.
(Pfliiger), 1889 XLV 582; Ueber d. Verhalten d. Vokale am neuen Edison'schen
Phonographen, Arch. f. d. ges. Physiol. (Pfliiger), 1890 XL VII 42 ; Phonophoto-
graphische Untersuchungen, II., Arch. f. d. ges. Physiol. (Pfliiger), 1890 XL VII
44; Phonophotographische Untersuchungen, III., Arch, f . d. ges. Physiol. (Pfliiger),
1890 XL VII 347 ; Eemerkungen zur VoJcalfrage, Arch. f. d. ges. Physiol. (Pfliiger),
1890 XLVIII 181, 543 ; Phonophotographische Untersuchungen, IV., Untersuchungen
mittels des neuen Edison'schen Phonographen, Arch. f. d. ges. Physiol. (Pfliiger) 1893
LIII 1 ; Hermann und Matthias, Phonophotographische Mittheilungen, V., Die
Kurven d. Konsonanten, Arch. f. d. ges. Physiol. (Pfliiger), 1894 LVIII 255;
Hermann, Phonophotographische Untersuchungen, VI., Nachtrag zur Untersuchung
der Vohalkurven, Arch. f. d. ges. Physiol. (Pfliiger), 1894 LVIII 264- Weitere
Untersuchungen ii. d. Wesen d. Vokale, Arch. f. d. ges. Physiol. (Pfliiger), 1895
LXI 169 ; Fortgesetzte Untersuchungen iiber die Konsonanten, Arch. f. d. gei.
Physiol. (Pfliiger), 1900 LXXXIII 1 ; Ueber d. Zerlegung von Kurven in harmon-
ische Partialschwingungen, Arch. f. d. ges. Physiol. (Pfliiger), 1900 LXXXIII 33.
PHONOGRAPH RECORDS 39
tipped till the mirror is vertical. A vertical slit in front of
an arc lamp permits a ray to strike the mirror and be re-
flected so that its image, obtained by means of the lens, falls
across a horizontal slit in the recording box. As the mirror
is deflected back and forth sidewise, the vertical beam of
light swings along the horizontal slit, so that the eye on the
opposite side sees a bright point vibrating along a horizontal
line. Inside the light-tight recording box a cylindrical
drum is arranged to rotate on a horizontal axis so that
sensitive paper on its surface receives the point of light.
As the phonograph and the drum are set in motion, the point
of light traces the speech curve upon the sensitive paper.
The result is developed in the usual photographic manner.
Since the curve on the phonograph cylinder and the set of
levers travel axially, the point of light would soon fall
one side of the horizontal slit ; to avoid this the phonograph is
placed on rails and slowly moved sidewise by the hand.
Curves of the German vowels sung by Hermann (born in
Berlin, 1838) are given in Figs. 31 and 32. . The vowel is
indicated by the phonetic letter and the note (physical scale)
by the letter in parentheses.
Analyses of the curves show that they are composed of
several tones ; that in most vowels (a, o, o, . . . ) there
is no sinusoid (p. 2) corresponding to the tone on which the
vowel is sung ; that in all there are one or more frictional
sinusoids (p. 6) which remain constant for each vowel regardless
of the tone on which it is sung ; that in some cases (i, y, . . .)
the vowel shows a sinusoid for the tone on which it is sung,
the other vibrations having no relation to its period.
The tone of constant pitch for each vowel — independent
of the tone on which it is sung — is termed by Hermann its
' formant.'
The cord tone itself is seen in the Aibrations for i and y, but
otherwise seldom appears in the vowels. It seems evident
that in most cases the cords act by emitting a series of more
or less sudden explosions that set the air in the resonance
cavities in free oscillation. The periodical changes from
40 CURVES OF SPEECH
strong to weak in these oscillations produce the cord tone
as heard, just as a series of sharp noises from a card held
Fig. 31.
against a toothed wheel or puffs from a siren will produce a
note. The groups of similar vibrations indicate separate
PHONOGRAPH RECORDS 41
puffs from the cords. Thus the curve a (c") in Fig. 31 shows
three groups of vibrations that represent the free vibra-
tions of the resonance cavity, each aroused by a puff from
the cords. That there is no sinusoid for the cord tone in
most of the vowels does not mean that the phonograph is
' deaf to the cord tone,' as has been absurdly stated, for the
phonograph will speak the vowel on that tone ; but it does
mean that the cord tone from the human mouth is not pro-
duced by a sinusoid vibratory movement but by a series of puffs.
The series of puffs act on the air in the resonance cavity just as
the series of magnetic impulses act on a spring (p. 11). The
curves of some vowels (o, e, oe . . .) seem to indicate that
4:2
CURVES OF SPEECH
the puffs occur with an explosive abruptness quite like
that of the magnetic impulses (Fig. 14) ; those of other vowels
(a) indicate puffs with more gradual beginnings, while still
others (i, y) indicate puffs of so gradual rise and fall of in-
tensity that the curve of explosion is like a distorted sinusoid.
Fig. 33.
The short vowels in an, en, etc., alter their tones as they
are about to change into the following consonants. The ex-
planation is, I suggest, that the mouth starts on its movements
from the vowel position to the consonant position before the
vowel ends.
The so-called ' short' vowels are in ordinary German gen-
erally not briefer ' long ' vowels, but are really other sounds,
as appears clearly in Hermann's curves (Fig. 32). The short
e in ' Helm ' is much more like e^ (a) than like Cj. In
fact no difference could be heard in pronouncing ' Halm * and
'Helm.' Short o in 'Wort' resembles long o much more
PHONOGRAPH RECORDS 43
than long o. Short i as in ' Bild ' appears quite different
from long i. Essential differences are also found between long
and short u, ce (6'), y (ii). Short ce (6') can be heard to
resemble a form of long ce not usual in Germany but very-
common in France. The short a differs least from the long
a, but with some persons it can be heard to somewhat re-
semble 0 or ae.
Hermann's curves show that the cord tone is not constant
but continually fluctuating to a slight degree ; similar minute
fluctuation will also probably be found in the resonance
tones if they are studied.
With a triple recording lever working in jeweled bearings
Hermann has obtained highly magnified curves of several
consonants (Fig. 33).
The 1 curve resembles that for short i with no intervals of
weakening or cessation of the vibrations (pseudobeats, p. 19).
At the point where a vowel borders on 1 there is regularly a
weakening or a pause ; in very rapid speech this does not
occur. The shortest 1, occurring in 'AUallal,' lasted 0.075^
to 0.100°. The main vibration of the 1 has sharp points at
the extremes of oscillation with one extra oscillation on the
ascending portion. Aside from this extra oscillation the
curve bears some resemblance to the curve given by Kcenig ^
for the series of partial tones 1, 2, 3, 4, 5, 6, 7, 8 with the
respective amplitudes 1, J, I, }, ^^, ^, ^-g, y|^ but without
difference of phase. An excessive amplitude given to the
second partial would perhaps produce the form seen in
the 1. If this should happen, the curve of the 1 might
be interpreted to indicate a cord action producing such
a set of partials. The first partial, or fundamental (that
is, the cord tone) appears strongly in 1. On this main vi-
bration there appear regular small sharply pointed oscil-
lations with a constant period independent of the note on
which the 1 is sung; these represent the resonance tone
of the mouth cavity. It shows no influence of a preceding
or following vowel. The characteristic resonance tone of
1 Kcenig, Quelques experiences d'acoustique, 228, Paris, 1882.
44 CURVES OF SPEECH
the 1 lies near f^. The cord tone with its octave is promi-
nent in 1.
The main fluctuations in the curve of m resemble some-
what those of 1 but do not have the extra oscillation from the
second partial and do have a slightly different incline. The
resemblance to Kcenig's curve of the series of partials of
decreasing amplitude and of identical phase as given above
is a fairly close one ; it would be rather close to that of the
same series with slight differences of phase. The small vibra-
tions arise from the resonance of the mouth cavity. Har-
monic analysis of the curves of m sung on the note 6° indicates
a verj^ strong first partial, or fundamental (cord tone), a
strong fourth or fifth partial at e^ to ^^* and a strong eighth
or ninth partial at e^ — /^* and not a geometric decrease in
amplitude. The constant resonance tone has a pitch of
h^ to c** . The cord tone (without its octave) is prominent.
The curve of n appears to closely resemble that of m.
Harmonic analysis shows that the first partial, or fundamen-
tal, is relatively not so strong and that the fourth or fifth
partial is much stronger than for m, while the eighth or ninth
partial and the resonance tone are the same as in m. The
cord tone (without its octave) is prominent.
The depressions in amplitude (pseudobeats) need not be
numerous in the middle of a word in order to produce the r-
effect. The depression often becomes complete, showing
complete interruption of the sound. The characteristic form
of the vibrations during the r remains the same and is in-
dependent of the degree of depression. The form of the r-
vibration may be one peculiar to the r and not derived from
the preceding vowel. A vowel tinge can, however, be pur-
posely given to the r. The vibrations between a vowel and an
r show intermediate forms that indicate the change. In rapid
speech the r may perhaps partake largely of the vowel-sounds
bounding it. The characteristic tone for the lingual r of two
German voices lies at 6^, that for the uvular r of Hermann
at about /^. All Hermann's rs were continuous rolled so-
nants, no records of unrolled or of surd rs having been mada
PHONOGRAPH RECORDS 45
With p, t and k the durations of the silence and the follow-
ing explosion vary. In the cases where the explosive noise is
lacking (p followed by m and sometimes by n, or t followed
by m or n) the phonograph curves show no record during
the speaking of these consonants, and Hermann's conclu-
sions concern only the full consonants with explosions. Their
chief characteristic is the long period of silence ; this was
never less than O.lS very seldom less than 0.2' in ordinary
speecli, and generally about 0.3= to 0.4= for p and a trifle
longer for t and k. In groups like pipip, papap and
tetet without special stress on either syllable, the pause
after the second vowel is much longer than after the first
one. The time is lengthened unconsciously when stress is
given to the consonant. The explosion of these consonants
is recorded as a "noise curve, irregular in its vibrations, and
also often as a strong excursion due to the increased air
pressure.
The record of increased pressure is more marked for p than
for t and k. The pressure curve is very steep, the explosion
being very sharp for p and less sharp for t and k ; it is also
longest for p (0.03' to 0.1'). The explosion is not necessary to
the hearing of these consonants ; in several cases there was no
record of an explosion in the phonograph and yet the sounds
of p, t and k were heard from it as well as usual. Just what
makes the acoustic distinction between these sounds in such a
case Hermann does not suggest ; it is presumably the noise
made by the breath as the closure is formed, since this will
differ for each kind of closure. ' At the end of a syllable
before a pause the noise curve is more prominent than before
a vowel. The small vibrations during some of the occlusions
suggest that sometimes the closure is not perfect and that
they have a slight breathiness.
In the case of p the following vowel usually occurs imme-
diately after the explosive noise ; the noise curve and the
vowel curve often begin on the line of recovery from the
explosion ; this may, I suggest, be due to the fact that the ex-
plosion of the p has pressed the glass diaphragm so far inward
46 CURVES OF SPEECH
that its inertia prevents its recovery as a whole before it
vibrates for the vowel. Regularly in t and k and occasion-
ally in p there is a very short silent period between the noise
and the vowel. This second silent period varies according
to the vowel, being shorter in ka than in kl although in ki it
is less than 0.02^ Apparently, I may suggest, the mouth organs
require more time to change from the t and k positions to a
vowel position than from the p position, and the time varies
with the resemblance between the consonant position and the
vowel position.
The noise curves of the explosions are very irregular, yet
they show certain periodic prominences that presumably cause
tones to be heard. For Hermann's voice the tones were
near «•' or oP- for p, near /^ for k and near /^* (in one case
63) for t.
In the case of two successive stops or occlusives as in atka
a considerable silent interval occurs between them, — in one
case about 0.15^
The curve for Hermann's x (^^^ ^s in acK) shows vibra-
tions often as strong as those of vowels. In many cases the
entire long x, curve consists of almost regular equidistant
points whose amplitude, however, rises and falls. The fre-
quency of these vibrations lies near 1000, or 5^ — (i^*. At
many places more frequent vibrations (near 1300, or e^—f^')
are mixed with the others. The fluctuations of amplitude
occur often at fairly regular periods of about 200 a second ;
these oscillations sometimes appear to have a periodic altera-
tion in strength occurring about 30 to 40 times a second. In
the latter case the curve bears a resemblance to the r-curve,
as might be expected from a certain resemblance between
X and uvular r.
The curve for s shows very fine sharp oscillations, the
frequency varying between d^ and /* but chiefly around b^.
Sometimes these fine oscillations are superimposed on coarser
ones with a frequency of about 600.
The fainter s curve resembles most that of s, with a fre-
quency mainly between g^ and 6^, but also sometimes be-
PHONOGRAPH RECORDS 47
tween V^ und dH. The finest oscillations (BOOO frequency, or
(/■') are also here often superimposed on coarser ones (600
frequency).
The curves for c (ch as in icK) show intermitting groups of
rather large pointed oscillations (about 750-800 frequency,
in one case 456) with still finer oscillations upon them (various
frequencies: 1100, #«; 2280, # ; 2736, J^.
The curve of f, when spoken very energetically, usually
shows a very definite periodicity (from 150 to 250 frequency),
arising from the tone produced by the vibration of the
lips or of the lower lip against the upper teeth. In the
ordinary f this periodicity is lacking ; the curve is composed
of coarse and fine oscillations in which frequencies of 1300 to
1500 iP-g^) and of 1700 to 2000 (a^-c^') can be occasion-
ally picked out. Many details have probably been lost in
these curves ; the very high tones have probably been entirely
lost.
In the curves of p, v, z, z (as in ' Lo^is ') and j (as in
' Jahr ') the cord tone appears distinctly. The curves for p
and V vary greatly. Being spoken with the two lips only, p
produces a curve closely resembling that of u with fine zig-
zags superimposed. The ordinary v with the upper teeth on
the loAver lip produces a curve that varies essentially from
the sinusoid, rising more rapidly and falling more slowly,
with superimposed zigzags of rather irregular period but of
an average frequency generally that of c* — c?* but not seldom
that of/3* — a^*. These two tones agree closely with those
of the similar sound f. The curve of z shows the cord tone
with superimposed zigzags whose mean frequency is about
a^* and c* — e*. A still higher tone is probably present also.
The curve for z is similar to that for z with tones of a^# — h^
and c** —f*. The curve for j likewise shows the cord tone ;
the zigzags represent tones between c* and e*, the same as
those of the vowel i. When followed by a vowel, the j
always becomes i before the vowel begins.
The curves for b, d and g show waves from the cord vibra-
tions ; they are weaker than those for m and n. The fine zig-
48
CURVES OF SPEECH
zags on these waves show frequencies between # and a^ for all
three sounds. When passing into a vowel, the b curve often
shows a small pressure explosion ; d and g never do this.
When used as final consonants, as in bib, did, gig (spoken
distinctly, and not like p, t, k as often in German), the
curves show at the end of the consonant regular vowel vibra-
tions that indicate lower cord tones for g than for b and d and a
resonance tone about g^ — a^ for b and g and about b^ — c^
for d.
The following table gives the resonance- tones (formants)
found by Hermann :
Sneech sounds ^^S'°° ""^ ^^^^'""^ °^
opeecn sounas rfisonance tones
J^CCl^li
owuiiu.
resonance tones
"f
resonance
tone
s
U
long
ci-/i
rf2-e2
Tl.
lingual
62
O
c2 - dn
'"2.
uvular
P
0
"
f.2 -P
P
a" - «l
a
"
e^-gn
k
rfl
62 (a)
"
c2-c2
m - an
t
/3 or 63
ei (f)
«
d^ - c2
a^ - 68
X
62 - (/sj
e3-
-p
oe
"
.p-9'
s
(rf2)
rf3J
-f
y
"
a^-b^
s
62 - rf3J
93*
-63
1
"
e*-/*
9
61 - 52J
^3)t
-/*
u
short
c2 (1)
f
/3-c«
o
"
61 - c2)t
V
/3S-</*
0
"
61 - cH
z
a3J
C*-
-e*
a
"
c2 - c2J
z
a3f — 63
<^'«-
-r
62 (S)
"
about c2
about e3
i
c4-e«
ei(e)
"
ti
"
h
d3_a3
oe
Ci
e3 „ ysj
d
rf3_a3
y
"
fH-gn,
S
d3-a3
i
«
aH
b
(explosion)
o3 — a3
1
P
d
(explosion)
63 -c*
m
ti^-c*
g
(explosion)
j3_aS
n
63 -C4
With the understanding that there may be considerable
range around a note, these results may be indicated approxi-
mately as follows :
r t= f r
eE
t
^E:
^ESEt
--t=t
u o 3 a
Long
Short
PHONOGRAPH RECORDS
49
i
f, t t
t t t
It
=t
^
-t^
44
II:
-r-
k
ra
f:
t t
=E=t
^ :b * := t
-ti-
-^=^=^
I
ffi:
s cfvzzjbd g b-ex. d-ex. g-ex.
A careful inspection of Hermann's curves shows that
gradual changes of form occur even in sung vowels. The
interpretation is that which I have already given (p. 19),
namely, continual changes
in the relations among the
cord and resonance tones ;
it is undoubtedly ti'ue that
even in the best singing the
voice does not remain at
Fig. 34.
exactly the same pitch but fluctuates ; it is probably also true
that the resonance tones fluctuate likewise.
In some experiments by Bevier,i the diaphragm of a
1 Bevier, The aroiistic analysis of the vowels from a phonographic record, Phygi-
cal Review, 1900 X 193; Acoustic analysis of the vowel A, Neuere Sprachen,
1900 VIII 2, 65.
50
CURVES OF SPEECH
phonograph reproducer was removed ; a rigid arm U (Fig.
34) with an adjustable plane mirror F was fastened to the
tracing lever D ; a spring on this arm held the knob of the
reproducer against the furrow in the wax. A narrow beam
of light 5" reflected from the mirror and focused by the lens
J on bromide paper around the drum IC, registered the speech
curve in great magnification. An improved form of tracing
mirror has just been devised by Bbviee. The sound was
Fig. 35.
recorded on the phonograph ; it was tested by reproducing it,
and then the above tracing apparatus was immediately put in
action. The curves of the vowel a sung on different notes
by three adult barytone voices and one child's soprano voice
showed after analysis the presence of the fundamental in
varying degrees of strength, the first and second overtones
very weak, a lower resonance tone averaging about 675, or/^
(from 575 to 800, d"^ to g^^^, and a higher resonance tone
^ averaging about 1150, or d^ (from 1000 to 1300,
c^ to e^). In musical notation these two tones
are as indicated. The curves in Fig. 35 are
from three voices, A, B, D; the upper curve
is from A on a, tone of 181 frequency, the second
from B on 202, the third from B on 226, the fourth from A
on 226, the fifth from B on 226 and the sixth from D on
240.
4=
PHONOGRAPH RECORDS 51
Maeichblle 1 has given a series of drawings of the phono-
graph grooves for the French vowels.
A new form of speech machine, Poulsen's telegraphone,^
may perhaps be available for phonetic purposes. An electro-
lytic phonograph by Nernst and v. Lieben has recently
appeared.
Refekences
For phonographs : Edison Manuf. Co., Orange, N. J. ; Lioret,
Paris. For Hermann's lever systems : Valbntinowycks, Konigsberg.
1 Marichelle, La parole d'aprfea le tracd du phonographe, Paris, 1897.
2 The tekgraphone, Nature, 1901 LXIV 183.
CHAPTER IV
GRAMOPHONE EECOEDS
The gramophone is a development by Beelinek of the
idea contained in Scott's phonautograph in combination with
the idea of reproducing the sound in a special manner.^
Fig. 36.
One form of the recorder with which the air vibrations are
received is shown in Fig. 36. It comprises a thin mica
diaphragm held in a frame. The sound waves coming
down the speaking tube set the diaphragm in motion; this
diaphragm moves one arm of the stylus and the point at the
' The following account is condensed from Scriptuke, Researches in experi-
mental phonetics (first series). Stud. Yale Psych. Lab., 1899 VII 7.
GRAMOPHONE RECORDS 53
end of the other arm repeats this moTement. Various modi-
fications of this recorder are used for various kinds of records.
The impression disc is prepared by two methods. In one
method (Berllnee) a highly burnished zinc disc 1 8™ in di-
ameter is flowed with a saturated solution of wax in benzine ;
the film of wax thus deposited is so thin that the touch of a
camel's hair brush marks it perceptibly. The prepared disc
is placed on a revolving plate so that its surface is touched
by the point of the recording stylus. As the plate revolves,
the recorder is made to travel toward the center; thus its
point cuts a spiral groove through the wax. The vibrations of
the point make deflections in this groove. These deflections are
in the plane of the surface of the plate and not dug into it as
in the case of the phonograph. The record disc is then placed
in an etching bath similar to that used by photo-engravers.
The part of the zinc from which the wax has been removed by
the stylus is attacked by the acid and a permanent groove is
made. A copper matrix is then made from this by electrolysis.
In still another method (Berliner) a glass plate is
clamped on an axis by which it can be rotated. The under-
surface of the disc is carefully polished and dried and is then
covered with a thin film of linseed oil by means of a camel's
hair brush. A smoky flame held under the plate deposits
a fine layer of lamp-black, thus forming an amorphous ink
which covers the glass in an even, exceedingly thin layer.
The coating of ink does not flow spontaneously ; it requires
only a minute force to trace a line in it. The sound line is
drawn by the point of the recording stylus in a manner
similar to that just described. Copies of the disc are made
by placing it over a sensitized photographic plate and
proceeding by photo-engraving.
In a later method (Cheney) the point of the recorder
draws a groove in the surface of a viscous substance, from
which an electrotype is made as a matrix. The resistance
to the vibratory movement is very small and the sound
is recorded with increased truthfulness. Records by this
process have been called zonophone records.
54 CURVES OF SPEECH
The impression disc made in any of these ways is used to
form a copper matrix by electrolysis. This matrix contains
the sound line in relief. After
the matrix thus secured has been
backed in order to give it strength
and stability, its face is protected
by a layer of nickel, which answers
the double purpose of protection
and also of giving a polish to the
final record. A composition made
of a combination of shellac and
filler (the same material from which
doorknobs, billiard balls, etc., are
now being made in large quantities), in the form of a stiff
board, f of an inch thick, is heated to the required consist-
ency. The mold, which is the exact size of the matrix, is also
heated, the matrix itself is heated, and then, by a quick pro-
cess, the matrix is thrown into the mold, the material on top,
the face of the die — also heated — is thrown over the whole,
and all is subjected to a pressure averaging about 80,000
pounds. Water is then driven into the
press for the purpose of cooling the mass,
the pressure is removed, the die opened,
f^^ra^^B and the completed record plate taken
m^mM^M' from the matrix. It is a true copy of the
Ja^^S^ original disc. This is the record known
to commerce ; it appears as in Fig. 37.
„ The process is repeated, a single matrix
sometimes producing 2000 or 2500 hard
records before the wear is sufficient to interfere with its
efficiency.
To reproduce the sound, the permanent disc-record is
placed on a plate which can be rotated by some motor power.
The reproducing sound-box (Fig. 38) is so arranged that the
point of its stylus travels in the sound groove. The devia-
tions in the sound groove move the point of the stylus,
whereby a mica diaphragm is made to reproduce the sound
GRAMOPHONE RECORDS
55
waves. The reproducing sound-box differs only in detail
from the recording sound-box. The operation is simply in
reverse order from that of the recording box — the sound,
when the recorder was used, being conducted first through
the tube to the diaphragm, and thence communicated to the
needle point, whereas, with the reproducing box, the waves
are communicated first to the needle point, thence to the
diaphragm, and thence outwardly, through the tube, to the
amplifying horn. The complete re-
producing apparatus is shown in
Fig. 39.
The speed at which the plate trav-
els in the record-making machine is
about 70 revolutions a minute. This
stretches out the curves for the speech
sounds so that the variations in am-
plitude are visible through the mi-
croscope only in the case of musical
sounds and vowels. The method of
direct reading by the microscope is
therefore not available. The record
must be transcribed in such a way
that the relation between length and height, that is, between
time and amplitude, shall be changed. In the method about
to be described the amplitude was enlarged.
In my first transcribing apparatus (Fig. 40) the gramo-
phone plate was put on a metal disc E similar to that of the
original record-making machine. This disc was rotated at
a speed of 0.1 of a revolution a minute by a system of spur
and bevel clears. A miter gear a on the axle of the electric
motor fitted into another miter gear on the first axle of the
speed-reducing machine B. The first axle of the reducing
machine thus revolved at the speed of the motor, which was
800 revolutions per minute. The second axle carried a large
spur gear with 160 teeth which fitted into a small spur gear
with 16 teeth on the first axle ; thus the second axle made
80 revolutions per minute. In a similar way gear-transmis-
Fio. 39.
56
CURVES OF SPEECH
sion to a third axle reduced the speed to 8 revolutions, and
transmission to a fourth axle reduced it to 0.8 of a revolu-
tion. This fourth axle carried a spur gear of 20 teeth which
fitted into the 160 teeth of the final driving machine of the
disc, whose axle thus made 0.1 of a revolution a minute.
Fig. 40.
The axle of the final driving mechanism carried on one
end a tube C (Figs. 40, 42) with a slit in it. Within this
tube was a rod l""" in diameter with a thread of 96 turns
to the inch on its surface ; it was held by a nut correspond-
ingly threaded. A projection from the rod fitted into the
.slit in the tube ; thus the rod was forced to turn with the
tube. At the same time the thread on its surface forced it
to move lengthwise -^ of an inch for each revolution. The
rod bore on its end a carefully centered point and just back
of this point a miter gear. The point pressed against the
GRAMOPHONE RECORDS
57
disc-carriage. This carriage consisted of a bar of brass
running on a pair of rails and carrying the metal wheel E.
The metal wheel rested on- the carriage and its axle projected
through it. As the rod travelled forward it pushed the
carriage ahead of it. At the bottom of the axle there was
a second miter gear D bearing against the first one on the
rod ; this turned the metal- wheel in unison with the rod.
When a gramophone plate was clamped on the wheel with
proper centering, it turned once in 10 minutes and was
driven forward radially -^-^ of an inch per revolution. Thus
the speech curve on a plate would travel steadily under a
fixed point from beginning to end.
Fig. 41.
Just above the disc the amplifying lever F was adjusted
so that the soft steel point rested in the sound groove. The
arrangement is shown in Fig. 41. The distance from the
fulcrum to the point was 22°™. The lever possessed side
movement in order to transcribe the curve, and vertical move-
ment in order to follow the changes in the thickness of the
plate. The long arm of the lever reached 595°"° beyond the
fulcrum. The extreme part of it consisted of a recording
58 CURVES OF SPEECH
point of pendulum ribbon 3£ (Fig. 40) 152-°°' long. This
point traced the side movement on the smoked paper and
also yielded to the up and down fluctuations without any
noticeable effect on the records. The amplification was ap-
proximately 27 times.
The centering of the gramophone plate was not an easy
matter. The speech curve was made in the form of a spiral
around the center of rotation in the original macliine ; neither
the edge of the rubber disc with the record nor the hole in
its center coincided with this center. To center the spiral
accurately on the metal plate two methods could be used.
The microscope method proved somewhat the more con-
venient. The metal disc was moved away from the point of
the rod. A microscope or a large magnifying glass was fixed
so that it was focused on the spiral groove. As the disc
was turned the groove passed through the field of vision.
If the plate was not centered, it would move to one side or
the other during one half a revolution; it was adjusted by
the fingers until the groove did not appear to move back and
forth with every turn, but to maintain a steady side move-
ment amounting to once the width between lines for one
revolution. The other method consisted in turning the disc
with the recording point adjusted and noting the deviation
to one side for one half a revolution. The disc was then
moved radially until the point marked one half the deviation.
The steel point was pressed into the groove of the plate
by means of the rubber baud on the thread h ; the verticality
of the pressure was assured by the plumb line C.
The record was made on smoked paper moved by the
Baltzae kymograph K with side movement of the drum by
the driving mechanism G. To avoid jarring through the
floor the table was at a later date suspended from the ceiling
by springs. The jarring of the motor was avoided by placing
it on sand.
Of the speech curves that were made with this apparatus
only three sets have been used in this book : Cock BoMn,
Series I ; Self Help, Series I ; Lord's Prayer, Series I.
GRAMOPHONE RECORDS 59
Greater amplification, accuracy and convenience have been
attained by modifications of the preceding arrangement. A
vsrorm placed on the motor axle turns a vi^orm gear on the first
shaft of the speed-reducing mechanism. A set of lamps of
various resistances modifies the speed of the motor. The
gramophone disc is rotated about once in 5 hours when a
curve appears in the tracing, and very much faster when
there is no curve. As the speech curve on a gramophone
disc runs around from 100 to 200 times, requiring 500 to
1000 hours of tracing at the low speed, it is desirable to
save time by running the plate faster during pauses.
The latest tracing apparatus (Fig. 42) comprises a primary
lever FJ with a steel point kept in the speech groove by
a small weight. This lever is held in a gimbal joint on the
block S on the support / over the plate U just as in the
original machine (Fig. 40). It is connected with a second
lever § by a link and the gimbal joints L and JV. The second
lever, with its fulcrum at 0 fixed to the support P, carries a
fine recording point on an axle. The magnification by suc-
cessive levers can be raised to almost any amount.
The records are made by a light aluminum point on an
axle at the end of the rod H; the point writes on a band
of smoked paper S about 6 meters long stretched over two
drums (p. 8). As the drums are run by a belt from the
same shaft that turns the gramophone disc, any changes in
the speed of the motor affect the disc and the record alike.
The gear wheel Y by which the plate is driven and the
pulley X for the belt to the drum are shown in Fig. 42 on
the axle within the barrel C.
The accuracy with Which the machine reproduces the
vibrations in the groove on the gramophone plate may be
shown by a comparison of repeated tracings of the same
curve ; the pieces in Fig. 43 were cut from different tracings
and were reproduced directly by photography. The tracing
is thus done with an accuracy indicated by the likeness of the
two records. These differences are so small as to escape
anything but microscopic measurement. The fine vibrations
60
CURVES OF SPEECH
Fig. 42.
GRAMOPHONE RECORDS 61
in some of the consonants, which are lost in the tracing, are
smaller than these differences.
As this machine can be run continuously day and night
with no supervision except for changing the paper, great
Fig. 43.
quantities of tracings can be accumulated in spite of the low
speed. The entire tracing — with long silences omitted —
of Rip Van Winkle's Toast, spoken by Joseph Jeffersok, is
reproduced in plates at the end of this volume.
The statements (p. 29) concerning the accuracy of talking-
machine records apply to the gramophone also.
Refeeences
For gramophones: National Gramophone Corporation, New
York. For gramophone tracing machines : Psychological Labor-
atory OF Yale University, New Haven, Conn. (The machine be-
' longing to Yale University will as far as practicable be freely placed
at the disposal of investigators who wish to have plates traced off.)
CHAPTER V
IMMEDIATE ANALYSIS OF SPEECH CUE.VES
A CUEVE of speech is at first sight no more intelligible
than a line of Chinese ideographs. The knowledge of the
speech sounds to which a certain portion of a curve belongs
gives the general meaning of the curve but affords little
information concerning its details. A careful study of the
sound by the ear reveals some of the grosser characters of
the sound, but cannot indicate any of the finer details that
lie before the eye in the complexities of the curve. The
meaning of these details — the very essentials of the speech
sounds — is not apparent at first observation ; only by patient
and persistent unraveling of the tangled curve is an inkling
of it obtained.
A set of speech curves (Plate I) from the Oock Bohin
record (p. 58) will be used to illustrate the first steps taken
in analysis. The curve reads from left to right ; the itali-
cized letters indicate the sounds recorded.^ The speech
curves in the figure would naturally run along horizontal
lines. The slow fluctuations seen in the records are due
to irregularities in feeding the gramophone plate sidewise.
They in no way affect the accuracy of the records. In mak-
ing measurements of duration, however, the ruler should
always be horizontal.
To interpret the details of a sound the grouping of the
vibrations is first noticed. In a series of groups of the same
general form each group may usually be considered as arising
from one puff of the vocal cords. The minor vibrations
arise from the vibrations of the resonating cavities and from
the overtones of the cords.
1 This account is from Scripture, Speech curoes, I., Mod. Lang. Notes, 1901
XVI 71.
IMMEDIATE ANALYSIS 63
Many of the main features of the speech curves can be
obtained by inspection without measurement; very much
more can be obtained by simple measurements. Long dis-
tances may be measured by millimeter scales ; the tenths of
a millimeter may be estimated by the eye. Finer measure-
ments may be made with a scale graduated in tenths of a
millimeter ; the work is done with a watchmaker's eyeglass, .
or under a magnifying glass. When the curves are very
small, the measuring may be done by a microscope with a
micrometer object table or a micrometer eye-piece.
The calculations are all done by books of tables or with a
slide rule. The investigator should become familiar with
various books containing extensive multiplication tables,
tables of reciprocals, etc. A Chinese abacus is also very
convenient in adding.
The speech curves are frequently of such a nature that the
period of the cord tone may be found by measuring the distance
between two like points in two successive groups of vibrations.
The distance in millimeters is translated into time accord-
ing to the equation valid for the tracing. For all the
curves in Plate I except that of ' draw your ' the relation is
l-""" = 0.0016= ; for this curve it is 1°™ = 0.0007^ Thus, the
distance between the two high points in the last vibration in
the fourth line is 3.2""° ; at 1""° for 0.0016» (use Zimmee-
mann's table for 16) ; this gives a period of 0.01536= for the
cord vibrations at that instant. A period of 0.01536' is
the same as a frequency of 1 ^ 0.01536 (use Baklow for
reciprocals) or 65.1.
To illustrate the details a complete analysis will be given of
the words ' saw him ' which occur in the phrase ' Who saw him
die ? ' The words are run together in speech on the gramo-
phone so that the result is soim rather than sohim, the h
not being heard, and the two vowels being fused like a
diphthong. The record shows no trace of the s. The first
vibrations of the curve differ from the rest, and show chang-
ing relations between the resonance (or mouth) tone and the
cord tone ; they indicate that the cords have begun to vibrate
64
CURVES OF SPEECH
while the mouth is still changing from the s position to the
o position. After this the grouping of the vibrations in
threes indicates a cord tone with a resonance tone a duo-
decime higher ; this general relation is maintained throughout
the diphthong. That still other resonance tones are present
is indicated by the subordinate modifications of the small
vibrations. The sound o increases slowly in intensity, but di-
minishes again as it changes into i (middle of first line). The
i is quite strong but falls quickly as the sound changes to m.
The m vibrations slowly fade away. The relations between
o and i in this diphthong somewhat resemble those between a
and i in ai discussed in a later chapter ; they differ in the fall of
amplitude at the end of o before the i begins, whereby the sepa-
ration of the elements of the double sound is slightly marked.
The accompanying table shows the way in which the
course of the cord tone in reference to pitch is calculated. It
illustrates several important principles used in computing and
interpreting results.
A.
Period in
milli-
meters.
B.
Period in
seconds.
C.
Frequency.
A.
Period in
milli-
meters.
B.
Period in
seconds.
C.
Frequency.
3.8
0.0061
167
4.8
0.0077
130
3.8
0.0061
167
5.0
0.0080
125
3.9
0.0062
161
5.1
0.0082
122
4,0
0.0064
156
5.0
0.0080
125
4.0
0.0064
156
5.1
0.0082
122
2.6
0.0042
238
5.2
0.0083
120
4.2
0.0067
149
5.1
0.0082
122
4.2
0.0067
149
4.7
0.0075
133
4.1
0.0066
152
4.6
0.0074
135
4.0
0.0064
156
4.7
0.0075
133
4.2
0.0067
149
4.8
0.0077
130
4.3
0.0069
145
4.7
0.0075
133
4,3
0.0069
145
4.4
0.0070
143
4.2
0.0067
149
4.5
0.0072
139
4.3
0.0069
145
4.5
0.0072
139
4.3
0.0069
145
4.5
0.0072
139
4.3
0.0069
145
4.7
0.0075
133
4.1
0.0066
152
4.5
0.0072
139
4.2
0.0067
149
4.7
0.0075
133
4.3
0.0069
145
4.5
0.0072
139
4.5
0.0072
139
4.6
0.0074
135
4.5
0.0072
139
4.4
0.0070,
143
4.5
0.0072
139
4.6
0.0074
135
IMMEDIATE ANALYSIS 65
The figures in column A give the distances in millimeters
from apex to apex of the strongest vibrations in the successive
groups. The measurements were made by an assistant who
did not know the nature of the problem investigated. It is
very important to note the following :
1. The determination of the exact point to be called the
apex may be indefinite to the extent of one or two tenths of
a millimeter, owing (a) to the roundness of the apex, (J) to
the fact that the apex is sometimes slightly displaced by inter-
fering resonance tones.
2. The general character of muscular action indicates that
the changes in the voice proceed with some regularity; this
would indicate that the unusual figure 2.6 for the sixth period
does not give the proper period at that point but shows some-
thing else.
Using Zimmeemann's table for 16, the figures in column A
are turned into time as shown in column B. These are the
lengths of successive periods in the cord tone. Using a table
of reciprocals (Barlow or Zimmermann) these are turned
into the frequencies as in column C.
The curve of frequency is now to be plotted. This is best
done by supposing the speech curve to be laid off along the
horizontal or JT axis, so that the first vibration is at zero.
From zero the proper number of millimeters is counted up-
ward to indicate the frequency of the cord tone at the start.
Thus, if the duration of the first group is 0.12^, the frequency
will be 83 ; if lOO"™ have been assigned to each 100 of fre-
quency, the dot will be placed at 83°"° above the X axis.
Above the point on the X axis at which the second group
of vibrations would begin if the curve were laid upon it, the
frequency of the cord tone at this moment is indicated by a
dot at the proper height. In this manner a series of dots is
obtained, indicating the frequency of the cord tone at a suc-
cession of moments.
In the diagram of frequency the successive dots might be
connected by straight lines. We probably come nearer to the
true curve of frequency (see 1. and 2. above) by drawing a
66
CURVES OF SPEECH
smooth curve that evenly distributes the dots on either side.
This may be done with the free hand, by means of draughts-
man's curves or by a flexible rubber rule; the more general
reasons for this procedure may be found in works on the
methods of science .^ The curve of frequency of oi, plotted
from the table on p. 64, is shown in Fig. 44.
m
ISO
100
so
50
IDli
Fig. 44.
ISO
200
The curious interruption of the regular course of figures in
the table by 2.6 arises from the fact that the series of the
strongest vibrations used to mark off the groups is replaced
at this point by a series arising from one of the weaker vibra-
tions. In the first part of the curve there is some vibration ,
of a changing character that causes a change in the moment
of strongest vibration. The unusual figure indicates this latter
fact and not any sudden break in the cord tone. A similar
occurrence may be seen in the o of- ' bow ' at the middle of
line 2 (Plate I ) and in o of ' draw ' as indicated below.
The periods of the smaller, or resonance vibrations can fre-
quently be obtained by direct measurement. This occurs most
readily when these vibrations are of a simple form or of a
pitch much higher than the cord tone. The result becomes
more accurate when several successive resonance vibrations
1 Jevons, Principles of Science, Chap. XXII.
IMMEDIATE ANALYSIS
67
Fig. 45.
can be measured together. When the resonance vibrations
are simple in form and a place in the curve can be found
where a number of them exactly fill out a group period, the
length of the group period divided by the number of vibra-
tions will give the length of the resonance period.
At the beginning of the record of soim (the ' glide ' from
s to o) the smaller vibrations show a period of 0.0032' or a
frequency of 313. This resonance
tone quickly changes to one of
0.0024^ period, or 417 frequency. It
remains at this figure throughout
most of its course but becomes
0.0028' or 357, toward the end of o.
During the i it is 0.0032'.
The other curves in Plate I are
described in Appendix II.
The analysis of speech curves might be greatly facilitated
by an inspection of curves produced by compounding vibra-
tions of known characters.
Vibrations of any character may be compounded by tabu-
lating them or by plotting them separately, adding the results
and plotting the sums. The
synthesis of sinusoid curves of
the same amplitude and the
same phase at the start but of
the periods y and 1 7 is shown
in Fig. 45. The ordinates of
the constituent curves 1 and 2
are added at each moment to
give the ordinate of the result-
ant 3 / thus, m"n" = mn + m'n',
p"q" =pq + p'q' (^p'q' having a
negative value), etc. A syn-
thesis of two vibrations of the
periods Tand ^T with the same
amplitude a and with the phase differences at the start of 0,
IT, ^Tand fTis shown in Fig. 46.
The sum of two curves of any kind with any relations of
Fig. 46.
68
CURVES OF SPEECH
period, amplitude and phase can be drawn automatically by
Pearson's curve-adder.
The composition of two sinusoid vibrations can be per-
formed mechanically by moving the recording point of one
fork over a smoked plate attached to another fork. The
arrangement consists of
two large forks, one fixed
and the other movable
in the direction of its
axis by sliding its sup-
port along a guide. The
fixed fork carries a nar-
row strip of thin glass
coated with soot or cov-
ered with smoked paper.
Both forks are set in vi-
bration and the movable
fork is rapidly drawn
back. The periods of the
forks can be altered by
weights. The results of
the syntheses for several
relations of pitch ^ are
shown in Fig. 47. The
relations are indicated
by the figures at the left-
hand side. In the first
line the speed with which
the fork is drawn along
decreases from left to
right, in the other lines
it increases.
To avoid the labor of computing the synthesis of several
sinusoids in a harmonic series (p. 13), machines have been
devised to add such vibratory movements mechanically .^
1 KoLNiG, Quelques experiences d'acoustique, 13, Paris, 1882.
2 DoNKiN, On an instrument for the composition of two harmonic curves^ Proc.
Koy. Soc. Lond., 1874 XXII 196; Blake, A machine for drawing compound
Fig. 47.
IMMEDIATE ANALYSIS 69
Only the machine of Peeece and Stroh has been used to
imitate speech curves. The artificial curves thus produced
can hardly be said to bear any close resemblance to the actual
vowel curves.
Sinusoids not in a harmonic series may be added by plot-
ting, by the curve-adder, by adjusting one of two vibrating
bodies (second curve in Fig. 47) or by inserting inharmonic
discs into a curve-producing machine.
It is quite doubtful if either harmonic or inharmonic syn-
theses of simple sinusoids can give close approximations to
speech curves. It is quite certain that the component tones
in most speech sounds do not belong to a harmonic series.
Moreover, it is highly probable that each component represents
a vibratory movement of a more or less explosive character and
not a harmonic of constant amplitude ; its equation is (p. 2)
y ^ a . e"*'. sin Stt-^
rather than (p. 5)
y ^^^ a . sm iTT-pp-
These considerations have suggested the synthesis of free
frictional sinusoids. A free sinusoid is understood to ex-
press the movement of a body displaced by a sharp blow and
allowed to vibrate in its natural period ; its amplitude will
decrease according to the, amount of friction present. A
synthesis of two such frictional sinusoids may be accom-
plished by the arrangement shown in Fig. 48. The spring B is
the spring B of Fig. 5 (p. 7). Upon it there is placed the
slide V carrying the spring U and another shde R with
the electro-magnet S. The movement of B is recorded on a
smoked drum by the point N, that of Uhj the point T. The
magnet ilf of the spring B (Fig. 5) and *S' of the spring U
harmonic curve^^, Amer. .Tour, of Otology, 1879 I 81 ; abstract in Nature, 1879 XX
103 • Peeece and Stroh, Studies in acoustics, I. On the synthetic examination
of vowel sounds, Proc. Roy. Soc. Loud., 1879 XXVIII 358; Michei-son and
Stratton, a new harmonic analyzer, Amer. Jour. Sci., 1898 (4) V 1.
70 CURVES OF SPEECH
(Fig. 48) are connected with the contact wheel A (Fig. 13).
When the current passes through M alone, both points iVand
T draw the curve of vibration for B as in Fig. 14. When
sent through S alone, the point ? draws the curve of vibration
of U. In both cases the vibration is a free frictional sinusoid.
When the curve is sent through both M and S, the point T
draws the curve of the sum of the vibrations of B and U.
The relations of period may be altered by changing the lengths
of B and V, those of amplitude by shifting the magnets, those
of damping by adjusting the dampers. When the curve drawn
by T is like that found in a speech curve, it can be assumed
that the speech curve is the result of two vibratory move-
ments simultaneously aroused by a sudden blow, which have
relations of pitch, amplitude and damping like those in the
springs. The sudden blow is the puff from the cords heard in
the cord tone and the two free vibrations are those of the vocal
resonance cavities. Tables of typical combinations would be
useful. A third sinusoid might be added by placing another
spring and magnet on U in the same way as U and S on B.
Work on these problems is now in progress ; tables of curves
may be expected at some future date.
References
For mathematical tables ; Crelle, Rechentafeln, Berlin, 1857 ; First
English Edition, New York, 1888 ; Zimmeemann, Rechentafeln, Berlin,
1891 ; Barlow, Tables of Squares, Cubes, Square Roots, Cube Roots,
Reciprocals of all Integer Numbers up to lOOOD, Reprint Edition, London,
1897.
For measuring rules ; Societe genevoisb, Geneva (specially adapted
is a 'petite ^chelle en argentan divise d'un cotd en dixifemes de
IMMEDIATE ANALYSIS 71
millimetres' for 20 franca). For slide rules and similar calculating
instruments : Dennbrt & Pape, Altona ; W. F. Stanley, London ;
Beyerlen & Co., Stuttgart ; Tavernier-Gbavet, Paris ; Keuffel &
EssER, New York.
For microscopes with micrometer eye-pieces : Zeiss, Jena ; Bausch &
Lome, Rochester, N. Y. For micrometer object tables: Zimmermann,
Leipzig. For adding machines: Felt & Tarrant, New York City.
For calculating machines (most advantageous for multiplication and
division) : Burkhabdt, Glashiitte i/S ; Bruckner, Dresden ; Grimme,
Natalis & CiE., Braunschweig. For the curve-adder : Coeadi, Zurich.
CHAPTER VI
HARMONIC AJSIALYSIS
The tones represented in a period of a speech curve
may be determined to a certain extent by the harmonic, or
Fourier, analysis. ^
The hypotliesis on which this analysis rests can be readily
illustrated. A stretched string — that of a tonometer, a
violin, etc. — is made to sound. The edge of a piece of
blotting paper, the tip of the finger or any narrow object is
then applied exactly to its middle point. The main tone
of the string ceases at once, but the octave is heard to con-
tinue. Careful inspection shows that the string has ceased
to vibrate as a whole but continues to vibrate in halves.
After the experiment has been repeated a number of times,
the unaided ear can hear the octave in addition to the fun-
damental when the string vibrates freely. Similar results
occur when the string is touched at J, J, . . . of its length.
The note from the violin string can thus be analyzed into
a series of partial tones consisting of a fundamental tone
and its overtones. It is assumed that these partial tones
correspond to vibrations of the sinusoid form (p. 2) with
different periods and that the complex tone of the violin is
made up of a sum of these sinusoids. The series of tones
thus found in the complex tone from a violin have periods in
the relations of 1, J, ^, J, . . . and frequencies in relations of
1, 2, 3, 4, . . . They form a harmonic series (p. 13). No
tones outside of the harmonic series can be detected.
The analysis of a musical tone into such a series of har-
1 FouKiER, Th^orie analytique de la chaleur, Ch. Ill, Paris, 1822.
HARMONIC ANALYSIS 73
monies can be accomplished on the principle of resonance
(p. 13). A resonator will respond loudly to a tone of its own
pitch. Resonators tuned to different tones are held to the
ear in succession while the tone is sounding. The periods
of the resonators that respond are taken as giving with
fair approximation the periods of the partial tones present.
Spherical resonators (Fig. 15) answer very accurately to the
partials ; when they are used, the tone to be tested is adjusted
to the pitch of one of the set. Adjustable resonators (Fig.
16) can be accommodated to any tone ; they can also indicate
inharmonic partials, that is, partials whose frequencies do
not stand in simple relations to that of the fundamental.
The application of the resonators to the ear requires the tone
to be prolonged unchanged for a long time, or to be repeated
unchanged if many partials are to be determined. The
resonators may be connected to manometric flames and may
be mounted in sets.^
Although an analysis by resonators is useful for demon-
stration, it is practically valueless in the study of speech
because 1. speech sounds are not constant long enough for
the adjustment of the apparatus ; 2. a resonator responds in
some degree to other tones than its own (p. 14); 3. the
harmonic analysis of speech tones can at best be only an
approximation.
When a vibration is registered in the form of a curve that
can be accurately measured, it can be resolved into a series of
harmonics by means of the Foueiek analysis.
To apply this method the heights of a series of ordinates
are measured at equidistant points along a base line paral-
lel" to the axis of the curve. When the records are very large,
the base line may be drawn directly on the record-sheet and
all the measurements made with a ruler graduated in tenths
of a millimeter. Smaller records may be enlarged by the
precision-pantograph of Coeadi. A camera lengthened by a
wooden tube projecting in front and bearing a lens of short
focus — for example, a 20-diopter lens from an optician's test
1 KoENiG, Quelques experiences d'acoustique, 73, Fig. 31, Paris, 1882.
74
CURVES OF SPEECH
case — may also be used for enlarging. The lens is covered
with a card containing a circular hole of l""^ diameter to in-
crease the sharpness of definition. The record is placed in
front of the lens. If desired, a photograph may be ob-
tained in the usual way. To simply trace off the enlarged
curve, the ground glass of the camera is replaced by a sheet
of clear glass on which a piece of tracing paper is laid.
When the record may be cut out, a micrometer object table
under a microscope may be used. A piece of the tracing is
cut out, placed between glass plates and focused under the
microscope. The measuring is done by micrometer screws
that move the curve horizontally and vertically.
Fig. 49.
For the analysis of a curve so many ordinates must be
measured that the piece of curve cut off between any two
ordinates can be considered as a straight line, whereby it is
implied that no maxima or minima (no points or turns) of the
curve lie between two ordinates. Very smooth curves can be
handled with only 12 ordinates ; these will give the first few
partials with fair accuracy. A portion of a speech curve, en-
larged 20 times by a camera, is shown with 12 ordinates in
Fig. 49. More complicated curves require 20, 24, 36 or 40
ordinates. The utility of the method depends upon the
success with which the measuring and computing can be
kept within reasonable limits of time.
The measurements of the ordinates, when inserted into
certain formulas, give values that indicate the relative am-
HARMONIC ANALYSIS 75
plitudes of the sinusoids into which the given curve may be
analyzed. An analysis of the curve in Fig. 49 gave the rela-
tive amplitudes as indicated in Fig. 50.
On the supposition that the original curve (Fig. 49) rep-
resents a tone composed of harmonic partials the analysis
shows that the second partial (2,
Fig. 50) was the predominant tone,
that the third and first were much
weaker and the others very weak.
For speech" curves we cannot make
the preceding supposition, and the
results of the analysis do not indicate
the presence of component harmonic
partials, but do indicate the presence
of tones in certain relations to one
another. A tone that is not har- I—
monic to the fundamental appears by ' ^ * * *^ *
the FoTJKiBK analysis to reinforce ■^"'' ^°'
neighboring harmonics. The diagram in Fig. 50 thus seems
to indicate the presence of a tone slightly higher than the
second partial. The Foueier analysis is often the only, way
of locating the tone or tones in a complex, even though they
do not stand in harmonic relations.
The preceding account is probably sufficiently detailed for
an understanding of the objects of the Fourier analysis;
full instructions for performing the analysis will be found in
an Appendix.
The resolution of an empirically obtained speech curve
into a series of harmonics by the Fourier analysis seems
to have been first performed by Schneebeli (p. 18) ; it was
used by Pipping (p. 20), Boeke (p. 37), Hermann (p. 38)
and Bevier (p. 49) in obtaining their results.
References
For pantographs and harmonic analyzer : Cokadi, Zurich. For mi-
crometer object table : Zimmermann, Leipzig.
PART II
PERCEPTION OF SPEECH
CHAPTER VII
THE ORGAN OP HEAEING
The auricle (1 in Fig. 51) in man is of little aid in liear-
ing sounds. It strengthens them slightly by reflecting more
Fig. .'51.
of the wave into the ear canal (^, Fig. 51) ; it favors those
from the front ; by resonance it modifies slightly the partial
tones of a complex sound ; and it favors the hissing tones of
sounds like s.
THE ORGAN OF HEARING
77
The vibration of the air traveling down the external canal
(2, Fig. 51) reaches the ear-drnm, or tympanum, memhrana
tympani (3 in Fig. 51 ; 1 in Fig. 52). This membrane con-
sists mainly of radiating fibers in the central portions and of
circular fibers in the peripheral portions ; it thus has great
possibilities of adjustment and damping. It is of slightly
conical form. Its structure of radiating and circular fibers,
Fig. 52.
its conical shape and its damping by the ear-bones attached
to it (Fig. 52) permit it to repeat vibratory movements of
various pitches without reinforcing any of them greatly by
resonance (p. 13). At the tympanum the vibratory move-
ment of the air is transformed into vibratory displacements
of the tympanic membrane. Weak sounds are most favored
by the shape and position of the tympanum.
The cavity beyond the tympanum is known as the middle
■ear (3, Fig. 51). It communicates with the pharynx by the
78 PERCEPTION OF SPEECH
Eustachian tube (^, Fig. 51). The middle ear is shown
in Fig. 52 with the cavity b, the ear-bones 2, 5, 7, the
tympanum 1, and the oval window d to the inner ear c.
The inner side of the tympanum is attached to the end of
a small bone, the hammer, or malleus (2—3, Fig. 52), which
rotates on an axle and thereby repeats the movements of the
tympanum. The anvil, or incus (5-6, Fig. 52), is a small bone
fitting on the head (3, Fig. 62) of the hammer and pivoted in
such a way that its long arm repeats the movement of the
handle of the hammer with a somewhat lessened amplitude.
The stirrup, or stapes (7, Fig. 52), attached to the long arm
of the anvil, fits loosely into an oval opening, fenestra ovalis
(d. Fig. 52), between the middle ear and the inner ear (c, Fig.
52). It is held in this opening by a ligament running around
on all sides. In repeating the movements of the anvil it is
forced to twist because the ligament is more tense on the side
below than above. The movement executed by the tym-
panum and ear-bones is indicated by the white line in Fig.
52. The ear-bones together form a lever arrangement for
transforming the air vibrations into a movement of the liquid
behind the oval opening. As the arms of the lever are in
the relation of \^ to 1, and as the tympanum is nearly twenty
times the size of the oval opening, the movement is reduced
in amplitude but increased in energy to the extent of 1-^- x 20
= 30. The movements of the stirrup are communicated to the
membranous sac, or labyrinth, of the internal ear.
Two muscles act upon the chain of ossicles. The tensor
tympani from the handle of the malleus (at 4 iii Fig. 52)
serves to pull it inward (-^ in Fig. 52) thereby stretching the
tympanum and pushing the stirrup more strongly against the
internal ear. The former action reduces the amplitude of
the vibrations of the tympanum and, by shortening its period
of free vibration, makes it better fitted to transform those of
various periods with less resonance-effect. The latter action
stretches the membrane of the oval window and thus pro-
duces more opposition to the movement of the stirrup, the
other bones and the tympanum ; the effect is to reduce the
THE ORGAN OF HEARING
79
amplitude of the vibratory movement. The stapedius muscle
attached to the stirrup tv^ists it in a way to relieve the in-
ternal ear from pressure and to oppose the action of the tensor
tympani. It thus renders the tympanum, the ear-bones, and
the liquid of the labyrinth more sensitive to vibrations.
The internal ear consists of a complicated membranous
labyrinth ' (1, ^, J, 4-^ Fig. 53) inclosed in a bone labyrinth
Fig. 53.
(7, 5, 5, 10, 1%'). The portion belonging to the cochlea (Jf)
is specifically concerned in hearing ; the utricle (1) and saccule
(^) are of doubtful function ; the semicircular canals (one
shown at S") are not concerned in hearing.
The cochlea comprises a long canal (^, Fig. 53) wound
around a cone (Fig. 54) and divided longitudinally through
nearly its whole length by a partition partly of bone and partly
of membrane. One portion of the canal (A in Fig. 55) is
connected with the oval opening, the other (5) with the
80
PERCEPTION OF SPEECH
round one (^16, Fig. 53). The bone division between them
(i, Fig. 55) is continued across by a membrane, memhrana
Fig. 54.
basilaris (7), to the opposite wall. The membrane of Reiss-
NBE (5) is stretched across the canal B. The whole laby-
rinth is filled with a liquid. The pressure of the stirrup
on the liquid connected with one side (A) of the canal will
Fig. 5.5.
press the membrane to the other side (£') . The basilar mem-
brane is composed of transverse bands of different lengths
and presumably of different periods of free vibration (p. 2).
THE ORGAN OF HEARING 81
Being light in mass these bands are readily set in vibration ;
as they are well damped, the vibrations quickly die away
on account of friction (p. 5). A diagram of one of the bands
of the basilar membrane and its annexes is given in Fig. 56.
5
& «0 fl £ E
Fig. 56.
The band I carries on it sets of supporting cells F and E, the
rods of COETI A B, and the hair cells D D' ; these last are
held in place in the membrane G.
The vibratory movements of the liquid in the cochlea will
thus readily set in motion those bands whose periods of free
vibration are harmonic to its period or to the period of any of
its components. For musical notes the action of the mem-
brane is thus somewhat like that of a series of harmonic
resonators responding to a sound (p. 14). For vibrations of
the character of voice curves a similar principle of analysis
by resonance may be applied. The whole membrane is forced
to move very slightly in a manner corresponding to the curve
of vibration, but any fibers whose periods correspond to the
periods of impulses contained in the original vibration will
be made to vibrate strongly.^ Each fiber of the membrane
carries on it a set of cells (i) D', Fig. 56) with hairs that
rub ^ against a floating membrane (^IP) when the fibers vibrate.
This causes deformation of the cells and irritation of the
nerve endings (9, 9', 9"} around them. The irritation is
transmitted along the nerve fibers (3, Fig. 55) to the brain.
The number of fibers in the membrane, that of the hair cells,
1 Helmholtz, Lehre v. d, Tonempfindungen, 5. Aufl., 235, Leipzig, 1896.
2 TER KniLE, Die Uehertragung d. Energie von d. Grundmemhran auf d. Haar-
zellen, Arch. f. d. ges. Physiol. (Pfluger), 1900 LXXIX 146.
6
82 PERCEPTION OF SPEECH
and that of the nerve fibers is greater than that required
for the number of tones that can be distinguished by the
ear.i Other theories of the action of the basilar membrane
and the stimulation of the nerve endings have been proposed
but have failed of general acceptance.
A complicated vibratory movement arriving in the internal
ear is probably analyzed into a series of simultaneous nerve-
irritations that proceed to the brain. These irritations change
at every instant. On the supposition (which may have
to be modified) that the irritation passing along a nerve can
change only in its intensity, all variations in the vibratory
movement of the air result in variations in the intensity and
number of the nerve irritations aroused. Every change in
the air-wave produces changes in these irritations. Arriving
in the central nervous system these irritations are combined
with others from other parts of the body and from various
nerve cells. The manner of analysis by the nerve endings in
the internal ear is largely unknown; tones like those from
some musical instruments are supposed to be analyzed into a
harmonic series (p. 72) on the principle of resonance. The
coriiplicated vibrations in speech sounds, especially in the
consonants, present difficulties to such a harmonic analy-
sis ; these difficulties may not be fatal and the theory is
still a plausible one. Just what happens to the irritations
when they reach the central nervous system is still entirely
unknown.
From every sound the brain receives impulses of differ-
ent strengths that arrive along different nerve fibers. The
activities of the brain cells connected with sensations of
sound bear no resemblance whatever to the physical vibra-
tions that arouse them, although the two sets of phenom-
ena are related by definite laws. The sensations of sound
in consciousness, moreover, in no way resemble the activities
of the nerve cells, although most intimately connected with
them.
1 Snodgkass and M'Kendrick, in Schaefer's Textbook of Physiology, II
1184, Edinburgh and London, 1900.
THE ORGAN OF HEARING
83
The integrity of a relatively definite region of the outer
layer (cortex) of the left cerebral hemisphere is necessary for
the correct understanding and use of speech in its usual
forms.
The centers for controlling combinations of vocal move-
ments in speech are in the region (Broca's convolution)
indicated by 'Motor words' in Fig. 57. The centers for
memories of word-movements are mainly in the anterior por-
tion of the speech region. The region connected with the hear-
ing of sounds is in the first temporal convolution ('Hearing,'
Fig. 57.
Fig. 57), that for the perception of speech sounds in Wer-
nicke's convolution ('Auditory words '). The region for the
perception of visual objects lies in the occipital lobe (' Vision ').
When printed signs take on the character of intelligible words
through association with other speech elements, they probably
involve the activity of portions of the brain in the fore part of
the visual region ('Visual words,' also J.). The region for
arm movements lies in the middle portions of the two central
convolutions (' Arm '), that for the corresponding motor
ideas probably near the same place ('Written words').
Differing from those of other activities the centers of
84 ' PERCEPTION OF SPEECH
speech all lie on one side, on the left in right-handed persons,
although the muscles and sense organs lie on both sides ; it
has been suggested that such an arrangement arises from the
necessity of exact coordination of the movements and im-
pressions from the two sides.
Four fundamentally different types of disturbance occur in
the brain. The more elementary ones appear as disturbances
in movement and sensation on one side of the body, for
example, the inability to move the right arm as the result of
a lesion in the arm center on the left side of the brain.
Another group is characterized by the loss of definite,
complicated associations of movements and sensations; for
example, the inability to perform the movements involved
in sewing, drawing, etc. A third group involves the loss of
highly complicated associations like those involved in speech.
Finally, the distuibance may involve the most complicated of
all associations, such as those of logical thought, language,
etc.
The special characteristic of speech disturbances in the
surface of the brain lies in injury to or loss of the ability to use
the common language of expression, such as words, letters,
notes and other symbols, although the general ability to use
ideas is not notably gone and the action of the ear and
vocal organs is not seriously injured. Such disturbances are
generally grouped under the term 'aphasia.' The mildest
forms of aphasia appear in the difficulty of finding the correct
(though familiar) word to express an idea (this weakness
of word-memory occurs regularly in old age, in conditions of
fatigue, etc.) ; or in the difficulty of understanding spoken or
wiitten words.
Motor aphasia, or word-dumbness, is the term applied to
a condition in which the voluntary control of the speech
muscles is in general fairly complete and the production of
sounds and even of single monosyllabic words or syllables
is possible, but in which the speech names for the most com-
mon things and conditions cannot be found and used and the
words heard by the patient cannot be repeated. It results
THE ORGAN OF HEARING 85
from injury to the motor speech center (' Motor words,'
Fig. 57).
Auditory aphasia, or word-deafness, is characterized by re-
tention of hearing with loss of understanding of spoken
words. A word appears to the patient as a meaningless
noise or a word from an unintelligible language. The dis-
turbance is connected with lesions in Weknicke's convolu-
tion (' Auditory words,' Fig. 57) ; it may be due to resistances
in the conduction in the brain, whereby single portions of
words do not appear in consciousness with sufficient rapidity,
clearness or completeness to be grasped, or whereby they can-
not be held long enough in memory to be united with the
following ones. Word-deafness may be chiefly perceptive or
chiefly associative, according as there is difSculty in grasping
the elements, or in recognizing them by assimilation to past
experiences. The understanding of words is seldom com-
pletely lost ; specially familiar words are generallj^ still under-
stood. The difficulties of conduction occasion exchanges of
words (paraphasia). The word-deaf person gets the meaning
imperfectly or not at all of what he hears ; nevertheless he
answers every question with alacrity but quite inappropri-
ately. He has no full consciousness of his errors and gener-
ally shows indifference to them. His thoughts need not be in
any wise incorrect, but he misunderstands and misuses words.^
Agraphia ordinarily occurs whenever motor aphasia is
present ; it is not known to occur alone. It is characterized
by inability to write words, while the other arm movements
are still completely under control. It is connected with defi-
ciencies in the ideas of speech movements and speech sounds,
and not with special lesions of the arm center. Agraphia
occurs regularly with auditory aphasia. Generally it is a
word-agraphia, the power to write single letters being retained
though they do not correspond to the sounds which they are
supposed to indicate. In cases of complete auditory aphasia
the hand makes only irregular, meaningless strokes and signs,
a proof that the chief guidance of the hand in writing is not
1 V. MoNAKOw, Gehirnpathologie, 523, Wien, 1897.
86 PERCEPTION OF SPEECH
from the visual memories of the letters or from the motor
memories and sensations of the arm and hand, but from the
auditory ideas with which these movements are connected.^
Alexia is characterized by the inability to understand words.
It is of two forms. The one is usualty connected with audi-
tory aphasia; the other arises independently. The former
results from disturbances in the temporal convolutions and
the gyrus angularis (J. in Fig. 57), the other from subcorti-
cal lesions.
It can be considered as definitely established ^ that the con-
trol of the individual muscles occurs in the spinal cord and
the pons, whose centers can operate them for combined action ;
that the cortical centers are those of group action, the sepa-
rate muscles being represented only when they are often used
singly; that the arrangement of the group centers in the
cortex does not occur by chance or in reference to anatomical
relations but according to definite, varied, fundamental move-
ments ; that each form of use of a muscle-group in this or
that act is represented by a different center. This latter fact
may be illustrated by a case in which injury to a portion of
the cortex was followed by loss of the ability to extend the
right thumb alone, while the ability to do so in combination
with other movements remained.
This principle of group representation in the cortex pre-
sumably holds good for speech volitions and speech percep-
tions ; no other principle seems to agree with the phenomena
that appear in disturbances of language functions.
The diagram in Fig. 58, a development of one by Licht-
HEiM,^ indicates the supposed functional connections of the
speech centers. Eepetition of a word heard involves the
transmission to the hearing center and stimulation of the motor
speech center by the associating fibers. Reading aloud in-
volves the translation of the visual words into auditory words
and then into spoken words. It does not occur by direct
1 T. MotTAKOW, as before, 518.
2 V. MoNAKOW, as before, 382.
5 LiCHiHEiM, Ueber Aphasie, Deut. Arch. f. klin. Med., 1884-85 XXXVI 204.
THE ORGAN OF HEARING
87
connection between the visual and motor speech centers ex-
cept in the case of the deaf. In spontaneous writing the ideas
are first put into speech form and then translated into writing
movements, and are not expressed by the arm directly. In
writing from dictation the perceived words first arouse the
center for speech action and this arouses the writing center.
In copying by sight there is direct connection between the
reading center and the writing center. Although used to
indicate the accepted theory of the action of the cortical
'action """'
Mision of
\JiiiQn
£ye
Fig. 58.
centers, the diagram is essentially a scheme of the mental
processes involved; it cannot be fitted with any closeness
to the actual brain action.
Beyond this point our knowledge of what occurs in the
brain is limited. A study of the cases of disturbance of
the language functions of mind has been carefully made
with reference to brain action, and certain fundamental facts
have been established. The attempts of various psychologi-
cal writers to state the phenomena of consciousness — parti-
cularly ' the marginal or fringe processes that are seldom if
88 PERCEPTION OF SPEECH
ever within the scope of introspective observation' — 'in
neural terms ' are pure figments of the imagination whose
hypotheses are incompatible with the familiar facts of phj^sio-
logical action.
Systematic application of experimental methods to deter-
mine the laws of speech disturbance has not yet been exten-
sively undertaken.
The motor functions of the brain in controlling the vocal
organs will be considered in detail in Chap. XV.
Eepeeences
For the anatomy of the ear : Tbstut, Traits d'anatomie humaine,
III, Paris, 1899 ; Spalteholtz, Handatlas dei- Anatomie des Menschen,
Leipzig, 1901. For the physiology of the ear: Sewall, Hearing,
Howells's Amer. Textbook of Physiology, 2d ed., Philadelphia, 1900 ;
Hensen, Physiologie des Gehors, Hermann's Handbuch d. Physiol., Ill
(2), Leipzig, 1880 ; M'Kendrick and Gray, The ear, Schaefer's Text-
book of Physiology, II, Edinburgh and London, 1900. For the mono-
graph literature : v. Stein, Die Lehre von den Funktionen der einzelnen
Theile des Ohrlabyrinths, Jena, 1894 ; Catalog of the Surgeon-General's
Library in Washington. For a summary of the present knowledge of
the speech functions of the brain : v. Monakow, Gehirnpathologie, Wien,
1897. For disorders of speech : Kussmaul, Die Storungen der Sprache,
Leipzig, 1877 ; Gutzmann, Vorlesungen iiber d. Storungen der Sprache,
Berlin, 1893; Lif.bmann, Vorlesungen ii. Sprachstbrungen, Berlin, 1898-
1900. For yearly bibliograpliies of recent works on aphasia: Zt. f.
Psychol, u. Physiol, d. Sinnesorgane; Annee psychologique ; Psycholo-
gical Review ; Jahrebericht ii. d. Leistungen u. Fortschritte in d. ges.
Jled. For summary and literature concerning musical centers : Lario-
now, Ueber d. musikalischen Centren des Gehirns, Arch. f. d. ges. Physiol.
(Pfliiger), 1899 LXXVI 608.
For models of the ear : Montaudon, Paris. For mechanism to illus-
trate the action of the tympanum and ossicles : Kohl, Chemnitz ; Cam-
bridge Sci. Instr. Co., Cambridge, England.
CHAPTER VIII
PERCEPTION OF SOUNDS
Sounds are purely mental experiences, most of which are
the results of vibratory movements reaching the ear through
the air. ' Tone ' and ' noise ' are the two extremes of a men-
tal arrangement of sounds according to likeness. A pure tone
can be obtained from a well made tuning fork ; a vowel sung
by a good voice can be made nearly a pure tone without ad-
mixture of noise ; a whispered vowel combines tone and noise ;
f and s are mainly noises, but nevertheless resemble tones
to some extent and are heard to vary in pitch. Every ordi-
nary sound appears to have more or less of each element in it.
Tones have three necessary properties : pitch, duration, and
intensity. Other properties such as timbre, objectivity, emo-
tional tinge, etc. may be added.
Many of the fundamental facts concerning tones can be
illustrated by a siren.^ In its simplest form the siren consists
of a carefully balanced and trued disc with a circle of holes
pierced through it. When such a disc {A, Fig. 59) is rotated
through a jet of air from a tube U, the holes produce a series
of puffs at intervals depending on the speed of the disc.
This disc may be mounted on any rotating axle B; it is most
conveniently placed directly on the axle of a small electric
1 Cagniard-Latouk, Sur la sirene, nouvelle machine, d'acoustique destinee a
mesurer les vibrations de I'air qui constituent le son, Ann. de chim. et de phys.,
1819 XII 167; Seebbck, Beobachtungen iiber einige Bedingungen der Entstehung
von TSnen, Ann, d. Phys. u. Chem., 1841 LIU 417; Ueber die Sirene, Ann. d.
Pliys. u. Cliem., 1843 LX 449 ; Dove, Beschreibung einer Lochsirenef. gleichzeitige
Erregung mehrerer TBne, Ann. d. Phys. u. Chem., 1851 LXXII 596; Helmholtz,
Lehre v. d. Tonempfindungen, 5. Aufl., 269, Leipzig, 1896,
90
PERCEPTION OF SPEECH
motor Avhose speed is regulated by an appropriate resistance
(p. 10). As the speed of the siren disc is increased, the pufPs
come more rapidly and finally change into a tone that con-
tinues to alter in its property of pitch.
Since a series of puffs, as known mentally, gradually changes
to a ton6 of rising pitch as the frequency of the puffs increases,
we may say that a tone is composed of puffs and that the
property of pitch depends on the frequency of the puffs. We
may not be able to detect the puffs in a tone and the tone may
appear to us as a simple phenomenon, yet we are quite justi-
fied in considering it as a compound of the more elementary
sensations termed puffs. The numerical expression for the
pitch of a tone may be derived from the number of puffs that
compose it. With a siren it can be readily demonstrated that
at low frequencies one puff corresponds to the passage of one
hole before the blast tube ; this correspondence can be followed
as long as the puffs can be heard separately. When the puffs
fuse into a tone, we may assume that the correspondence
still remains ; the number of jets of air a second can thus
be taken as the figure for the pitch of a tone. The pitch of a
tone and the frequency of the puffs are thus correlated.
The property of pitch may be varied by changing the speed
of the disc, and making the holes pass the jet at different
PERCEPTION OF SOUNDS 91
rates. With a small number of holes the pitch is said to be
' low ' or ' grave ; ' with a large number it is said to be 'high '
or ' acute.' These terms are metaphors, having no physical,
physiological or psychological meaning except what has been
derived by association.
The property of duration in a tone can be illustrated by
changing the time during which the tone is produced ; the
property of intensity by making it louder or weaker.
The pitch of the tone from the siren at any moment can be
determined by placing on the axle of the motor a contact Q
(Fig. 59) consisting of a gear wheel with spaces filled by
vulcanized rubber, and adjusting a pair of copper brushes
D on its rim. A battery current Gr is sent through the
brushes, a make-key, IJ, and a
magnetic marker iV"; whenever the
knob H is pressed, the circuit is
closed at / and each closure at D
will register a check in the line of
the marker-point N on the drum.
To get a registration of the time the
marker 0 is connected to a fork M
(p. 1 5) in such a way that the break-
ing of the circuit at ^sets it vibrat-
ing. This can be conveniently done
by using the key as a shunt around „
the marker in a circuit coming from
the fork. A comparison of the checks in the line from the
marker N with the waves of known frequency from the
marker 0 will give the time between contacts at the brushes
I) ; from this the speed of the disc and the number of puffs
can be readily calculated.
The magnetic marker referred to appears in various forms.
The Pfbil marker is shown in Fig. 60. A current passing
through the coils m draws down the steel spring p and causes
the lever A (only partly shown in the figure) to record on
the drum. This marker in connection with an electric fork
can produce smooth waves of the sinusoid form, which is
92 PERCEPTION OF SPEECH
well adapted for time-comparisons ; by adjusting the weight
g, it may be tuned to the harmonic series of which the fork-
period is a member. The screw s moves the cores of the
magnets to and from the spring p and thus regulates the
amplitude of the action. The side screws 1 1 are for adjust-
ing the point of h accurately on the drum. The Depeez
marker (Fig. 61) is a form adapted to quick action on
account of the light mass of its armature V. The magnets
U U are connected to the binding posts S S. The armature
is held back against the cone C by the spring R, whose ten-
sion is regulated by a movable arm Y. The distance of the
armature from the magnet is regulated by moving the cone
C axially by means of the knob T. With appropriate adjust-
ments of the strength of the current, the tension of the
Fig. 61.
spring and the position of the armature, the rapidity of
action can be made so great that the marker loses only a
few ten-thousandths of a second in responding to a magnetic
impulse ; it may be used to record 500 impulses a second.
The marker is held on a rod by the screw P through the
barrel I) ; its stem is lengthened or shortened by the knob
Q moving the rack B by means of the pinion A.
The latent time of a marker should be measured frequently
when it is of importance. This is best done by placing on
the axle of the drum a rubber or fibroid wheel with a strip of
metal on its edge around half the circumference; the two
poles of the circuit are rested against the wheel as indicated
in Fig. 62 ; the marker is placed in the circuit, its point
being against the drum ; the drum is brought to rest just
as the contact is closed, whereby a check is made on
the smoked surface; the same is done for the point where
PERCEPTION OF SOUNDS 93
the circuit is broken. The drum is now turned rapidly; the
marker responds to each make and break by a check on the
line on the drum ; the distance between the check made by
bringing the drum to rest and that made while it is moving
represents the time lost by the marker; with a time hue
(p. 15) on the drum this latent time can be measured. It is
convenient to have such simple contact wheels with brushes
on all drums ; the determination of the latent time then takes
only a few moments before or after the experiments are
made. In the experiment with the siren this measurement is
not required.
The puff produced by the siren is accompanied by consider-
able noise. Most of this noise can be avoided by using a
vibrating steel reed. When such a
reed is clamped in a vise, it can be
set vibrating by the finger. When
its length is adjusted so that its fre-
quency is sufficiently high, a tone is
heard. As the length ia increased,
the tone begins to rumble and finally
breaks up into puffs of sound almost
entirely unaccompanied by noises.
Still clearer results may be obtained
by using large tuning forks with
heavy sliding weights.
The rise in pitch may be continuous between the lowest
and the highest limits. The changes by steps as used in
music are required neither by the voice nor by the ear.
The psychophysic law for tones in its simplest form as
given by Aristotle^ states that a tone corresponds to the
vibrations of a body (the statements of Pythagoras refer to
the relation between the length of a string and the pitch of
a tone and not to vibrations).^ The psychophysic law of
1 Aeistoteles, Problemata, XIX 27 ; De anima, II viii 420.
2 NiKOMACHOs, Harm, introd., 116 (direct source for the statements regard-
ing Pythagokas) ; Bookh, Philolaos des Pythagoreers Lehren nebst Bruch-
stucken seines Werkes, 65, Berlin, 1819.
9 J: PERCEPTION OF SPEECH
pitch was first formulated by Galilei ^ and Meesenne^:
the pitch of tones depends directly on the frequency of
the impulses.
Using siren discs with holes of different shapes Seebeck
demonstrated ^ a principle whose validity was long overlooked
in favor of an erroneous theory of vowel tones. Seebeck's
principle may be stated thus : a puff recurring with a con-
stant period produces a tone whose pitch is given by that
period independently of the character of the puff and of the
portion of the period occupied by it. The experiment has
been repeated in various developments. A strong tuning
fork may be held close behind a rotating disc with large holes
so that the tone is heard only when a hole passes the fork.*
The intermittent tone gives a series of puffs whose frequency
is that of the holes passing the fork. When the boles pass
rapidly enough, the puffs are heard as a tone of a pitch corre-
sponding to the number of holes passing in a second. The
tone of the fork and that of the series of holes are thus heard
simultaneously. When regular groups of holes are filled
(or omitted) in a siren disc, the changes between tone and
silence likewise produce two tones, one with a pitch corre-
sponding to the frequency of the holes within a group and
another to the frequency of the group.^ Similar experiments
may be readily made with a card or paper cone held against
the teeth of a gear wheel. The intermissions are produced
1 Galilei, Discorsi e dimostrazioni matematiche, Leida, 1638.
" Meksenne, Harmonie universelle (Harmonicorum libri XII), Paris, 1636;
Govi, Su un' antica dimostrazione del numero delle vihrazione che corrispondono ad
un suono data delta scala musicale, Rend. Ace. di Napoli, 1886 XXV 106 (refers to
Mersenne) ; Tatlok, De inventione centri osciUationis, Phil. Trans. Eoyal Soc.
Lond., 1713, XXVIII 11; Methodus incrementorum, London, 1715; Eulek,
Tentameri novae theoriae musicae, Petropoli, 1739.
' Sebbeck, Ueber die Erzeugung von Tonen durch getrennte Eindrilcke, mit
Beziehung auf d. Definition des Tones, Ann. d. Phys. u. Chem., 1844 LXIII
368.
^ KcENiG, Ueber den Zusammenhlang zweier Tone, Ann. d. Phys. u. Chem.,
1876 CLVII 231 ; Quelques experiences d'acoustique, 139, Paris, 1882.
* Des'sert, Akustisch-physiotogische Untersuchungen, Arch. f. Ohrenheilk., 1886-
87 XXIV 171.
PERCEPTION OF SOUNDS 95
by filling some of the teeth with wax.i The tone whose
pitch corresponds to the frequency of intermission is always
the loudest although the tone whose pitch is that of the
number of blows of the teeth on the card may be also
heard. When a tone through a telephone is regularly inter-
rupted by breaking the secondary circuit, the interruption
tone appears also.^ The fundamental law demonstrated by
the preceding experiments may be thus stated : every periodic
change within certain limits of frequency is perceived by the
sense of hearing as a tone.
Out of tones regarded as simple, other tones, or notes, can
be built up. "When two tones are sounded by blowing on
two series of holes of widely different frequencies in tbe
siren disc, the result appears different from either tone heard
separately. To the uneducated ear this tone appears just as
simple as the others though quite different in its character.
It may, however, be regarded as a complex or compound
tone composed of two simpler tones, just as each of these
simpler tones is to be regarded as composed of puffs.
The character of a compound tone varies with the relations
of pitch and intensity among its components. When the
components are of approximately equal intensity and when
they have certain simple relations of pitch, they are called
' chords ; ' thus, three tones of nearly equal intensity in the
relations of 2 : 3 : 5 form a major chord. The chords can be
demonstrated by a siren disc with several series of holes, in
the desired numerical relations, blown by several jets. A few
of the simple tones in compound ones can frequently be picked
out by listening for them. The notes of a chord may appear
simple to the unaccustomed ear, but the musician can by
attention separate them.
"When the components have certain simple relations of
pitch such as 1:2:3:4:5... and the lowest tone is
1 Hermantst, Phonophotographische Untersuchmgen, III., Arch. f. d. ges.
Physiol. (Pfluger), 1890 XL VII 386.
2 ZwAARDEMAKBR, Ueber IntermiitenztSne, Arch. f. Anat. u. Physiol. (Physiol.
Abth.), 1900, Supplementband, 60.
96 PERCEPTION OF SPEECH
much stronger than the others, the compound tone is said to
be ' complex,' or to be composed of a ' fundamental ' and its
' overtones.' To illustrate this, several series of holes in the
siren may be blown at once, the high tones being made quite
subordinate in intensity to the lowest one by using smaller jets.
Changes in the relative intensities of the series of overtones
produce changes in the character of the complex tone ; these
are said to be changes in timbre. A series of tones such as
1:2:3:4:5:6: — :8
10,
with the relative intensities indicated by the size of the
figures, will produce quite a different complex tone from
a series such as
1:2:3:4:5:6:7:8:9: 10.
The character of the tone differs also with the character
of its elementary puffs. When puffs of the same frequency
are compared, they will be found to differ in properties that
may be termed 'loudness,' 'suddenness,' 'smoothness,'
'sharpness,' etc. A single puff in any case is heard with
more or less complicated variations of intensity. Explosive
puffs like those from the siren and usually from the vocal
cords rise and fall suddenly in intensity. Very smootli
puffs are obtained from tuning forks.
Tones composed of puffs of different forms appear differ-
ent to the ear. These differences often resemble those pro-
duced by combining tones into compounds and complexes.
The puffs from some musical instruments whose tones differ
in timbre can be shown to be of forms that would arise from
adding harmonic series of sinusoids with regular systems of
decreasing amplitudes for the components of shorter periods.
The puffs from other instruments are of forms that would
not arise in this way. Tones :^rom the voice differ not only
in timbre but also in other ways ; the puffs have very
complicated forms.
PERCEPTION OF SOUNDS 97
When the puffs are of forms such as would arise from a
summation of puffs of the sinusoid form, with periods in a
harmonic series, the sound can often be heard as containing
a series of tones. This led Ohm to assert that each sensation
of tone corresponded to a sinusoid vibration and that complex
vibrations were analyzed by the organ of hearing into a
harmonic series of sinusoids resulting in a mental complex
of liarmonic tones.^ Although refuted by Seebeck, this
hypothesis was used by Helmholtz^ as the basis of his
theory of the action of the ear. The hypothesis is certainly
incorrect when applied to sensations. To the mind the tone
from a violin is just as simple as that from a tuning fork or
an organ, yet the analysis of the vibration into a series of sin-
usoids would give results differing greatly in complexity.
Moreover, the mind executes no such analysis ; the physical
differences represented by combinations of sinusoids appear
as the pi'operty of timbre. Finally, physical speech vibrations
cannot be treated as a series of sinusoids, and yet the ear
hears tones in speech. Moreover, the explosive tones from
the siren differ with the character of the explosion. When
elliptical or triangular holes and mouthpieces are used, the
resulting tones differ greatly in character with no apparent
presence of overtones.^
The range of pitch that can be heard by the ear is confined
between fairly definite limits ;* the lowest limit for a series
1 Seebeck, Beobachtungen iiber einige Bedingungen der Entstehung von Tonen,
Ann. d. Phys. u. Chem., 1841 LIII 417 ; Ohm, Ueber die Definition des Tones, nebst
daran gelcniipfter Theorie der Sirene und dhnlicher tonbildender Vorrichtungen,
Ann. d. Phys. u. Chem., 184.3 LIX 497 ; Seebeck, Ueber die Sirene, Ann. d.
Phys. u. Chem., 1843 LX 449 ; Ohm, Noch ein Paar Worte iiber die Definition
des Tones, Ann. d. Phys. u. Chem., 1844 LXII 1; Seebeck, Ueber Schwingungen
unter Einwirhuny veranderlicher Krafte, Ann. d. Phys. u. Chem., 1844 LXII 289 ;
Cjeber die Definition des Tones, Ann. d. Phys. u. Chem., 1844 LXIII 353 ; Ueber die
Erzeugung von Tonen durch getrennte Eindrilcke, mit Beziehung anf die Definition
<les Tones, Ann. d. Phys. u. Chem., 1844 LXIII 368.
^ Helmholtz, Lehre v. d. Tonempfindungen, 5. Aufl., 97, Leipzig, 1896.
s Seebeck, Ueber d. Erzeugung v. Tonen durch getrennte Eindriicke, Ann. d,
Phys. u. Chem., 1844 LXIII 37.5.
* Sauveuk, Me'm. de I'acad. roy. des sciences, Paris, 1700, p. 140 ; Chladni,
Akustik, 2, 36, 294, Leipzig, 1802 ; Savakt, Note sur la limite de la perception des
7
98 PERCEPTION OF SPEECH
of puffs from the siren depends on the intensity. A very-
low limit of 16 to 30 frequency can be reached by large and
powerful forks. The upper limit of pitch can be determined
by striking bars or forks of known frequencies ^ or by the
Galton whistle.^ The ordinary upper limit lies near a fre-
quency of 30,000 for moderately strong tones ; ^ it is lowered
sons graves, Ann. de chim. et de phys., 1831 XLVIII 69; Despketz, Observa-
tions sur la limiie des sons graves et aigus, C. r. Acad. Sci. Paris, 1845 XX 1214 ;
Moos, Beitrag zur Helmholtz'schen Theorie der Tonempjindungen, Archir. f. path.
Anat., 1864 XXXI 125; Patholog. Beobacht. ilber d. Tone, Arch. f. Augen- und
Ohrenhk., 1872 II (2) 139, Arch. f. Ophth. and Otol., 1873-74 III 113;
Magnus, Ein Fall von partieller Lahmung des Corti'schen Organs, Archiv fiir
Ohrenheilkuude, 1867 II 268; Pketer, Ueber die Grenzen der Tonwahrnehmung,
Jena, 1876 (also in Preter's Sammlung physiol. Abhandlungen, I 1, Jena,
1877) ; TuRNBULL, The limit of perception of musical tones by the human ear, Boston
Medical and Surg. Journ., 1879 C 741; Stef.i.nini, Dell' energia minima che e
necessaria a produrre la sensazione del suono, Atti deUa r. Ace. lucchese,
1888 XXV 239; Love, The limits of hearing, Journ. Anat. Physiol., 1888
XXIII 336 ; Appunn, Akusliscke Versuche ilber die Wahrnehmung tiefer Tone,
Jahresb. d. Wetterau'schen Gesellsch., 1889; Helmholtz, Die Lehre von den
Tonempfindungen, 5. Aufl., 290i, Leipzig, 1896 | Battblli, Sur la limite inferieure
des sons perceptibles. Arch. Ital. de Biol., 1897 XXVII 202 ; Schafer, Die
Bestimmung d. unteren Horgrenze, Zt. f. Psych, n. Phys. d. Sinn., 1899 XXI 161.
1 KcENiG, Catalog, Paris, 1889; Ann. d. Phys. u. Chem., 1899.
^ Galton, Inquiries into Human Faculty, 38, London, 1883 ; Stu-mpf iind
Meyer, Schwinungszahlbestimmungen bei sehr hohen Tonen, Ann. d. Phys. u.
Cliem., 1897 LXI 760; Sohvvendt, Exper. Bestimm. d. Wellenldnge u. Schiving-
ungszahl hochster horbarer Tone, Arch. f. d. ges. Physiol. (Pfliiger), 1899 LXXV
346, also in Verb. d. naturf. Ges. Basel, 1900 XII 149 ; Edelmann, Fortschrilte
in der Herstellung der Galtonpfeife, Zt. f. Ohrenheilkde., 1900 XXXVI 330;
Studien Uber d. Erzeugung sehr hoher Tone vermittelst d. Galtonpfeife, Ann. d.
Phys. u. Chem., 1900 II 469.
3 Chladni, Die Akustik, 24, Leipzig, 1802; Sauvedk, Me'm. de I'acad. roy.
des sciences, Paris, 1700, p. 140; Savart, Notes sur la sensibility de I'organe de
I'ouie, Ann. de chim. et de phys., 1830 XLIV 337 ; Despretz, Observations sur la
limite des sons graves et aigus, C. r. Acad. Sci. Paris, 1845 XX 1214; Schwaktze,
Totaler Verlust des Perceptionsvermogens f. hohe Tone nach hefligem Schalleindruck,
Archiv f. Ohrenh., 1864 I 136; Gottstein, Ueber den feineren Bau und die
Eutwickelnng der Gehorschnecke beim Menschen und den Saugethieren, Bonn,
1871; Blake, Summary of the results of experiments on the perception of high
musical tones, Trans. Amer. Otol. Soc, 1872-74; Diagnostic value of high musical
tones. Trans. Amer. Otol. Soc, 1873 118; Audibility of high musical tones, Amer.
Jonru. Oto]., 1879 I 274 ; Burnett, Ein Fall von verminderler Horbreite,
Archiv f. Augen- und Ohrenheilk., 1877 VI 238; Ratleigh, Acoustical
observations; very high notes, Philos. Mag., 1882 (5) XIII 344; Pauchon, Sur
la limite superieure de perceptibility des sons, C. r. Acad. Sci. Paris, 1883 XCVI
PERCEPTION OF SOUNDS 99
for weaker tones,^ by advancing age,^ etc. When the upper
limit is too low, the perception of certain consonants is
defective.^
When two sinusoid vibrations of the neighboring frequen-
cies n and n' unite, there is a rise and fall of the vibratory
movement with the frequency h = ±(n— n''). These fluctu-
ations, or 'beats,' can be readily heard by sounding two
neighboring piano strings.
When the beats have a sufficiently high frequency, they are
heard as ' difference ' tones * with a frequency equal to the
difference between the frequencies of the two primaries.^
Helmholtz's deduction of difference tones from the asym-
metrical vibration of the tympanum has been shown to be
inconsistent with the facts.^ Four difference tones can
arise from two simultaneous tones under favorable con-
ditions.'^ The first has a pitch of i)^ = ± (w — «'), where
1041 ; Stumpf, Tonpsychologie, I 414, Leipzig, 1883 ; Love, The limits of hear-
ing, Journ. Anat. Physiol., 1888 XXIII 336.
1 Blake, as before, 1872, 1873 ; Sckipture and Smith, Experiments on the
highest audible tone, Stud. Yale Psj'ch. Lab., 1894 II 105.
2 Blake, as before, 1872, 1873 ; Zwaakdemakek, Een Wet van ons Gehoor,
Ned. Tydschr. v. Geneeskunde, 1890 737 ; Der Ver/ust an hohen Tonen mit zuneh-
mendem Alter, Archiv f. Ohrenheilk., 1891 XXXII 53.
" Moos, Ueber das comblnirte Vorhommen mangelhafter Perception gewisser
Consonanten, sowie hoher musikaliscken Tone und deren physiologische Bedeutung,
Arch. f. Augen- u. Ohrenheilk., 1874 IV 165; Moos und Steibkugge, Ueber
Nervenatrophie in der ersten Schneckenwindung ; physiologische und pathologische
Bedeutung derselben, Zt. f. Ohrenheilk., 1881 X 1.
* SoRGE, Vorgemach musicalischer Composition, Thl. I, Cap. V, § 5, Hamburg,
1745 (this passage is repeated in Winkelmann, Physik, I, 779, Breslau, 1891);
Tartini, Tratto di musica secondo la vera scienza dell' armonia, Padova, 1754.
* Hallstrom, Von den Combinationstonen, Diss., 1819 ; also in Ann. d. Phys.
u. Chem., 1832 XXIV 438; Helmholtz, Ueber CombinationstSne, Monatsber.
d. Berliner Akad., 1856, 22 Mai, 279 (reprinted in Helmholtz, Wiss. Abhand.,
I 256, Leipzig, 1882); Ueber CombinationstSne, Ann. d. Phys. u. Chem., 1856
XCIX 497 (reprinted in Helmholtz, Wiss. Abhandl., I 263, Leipzig, 1882);
Tonempfindungen, 5. Aufl., 152 and Beilage XII, Leipzig, 1896.
^ VoiGT, Ueber den Zasammenklang zweier einfacher Tone, Nachrichten d.
kgl. Ges. d. Wiss. zu Gott., 1890 159; also in Ann. d. Phys. u. Chem., 1890
XL 652 ; Hermann, Zur Theorie der CombinationstSne, Arch. f. d. ges. Physiol.
(Pfluger), 1891 XLIX 499.
' KKiJGER, Beobachtungen an Zweikldngen, Philos. Stud. (Wundt), 1900
XVI 325.
100 PERCEPTION OF SPEECH
n and n' indicate the pitches of the two tones. The
second is D^ — ± (n — i>i), the third I)^=± (^D^ — 2>^},
and the fourth i>^ = ± (Z>3 — Dj). A fifth may even
occur. It is not yet known if these tones occur in vocal
sounds. A summation tone s = n + n' may possibly be
present.^
When two tones are heard in succession they may appear
to be alike ; vaguely unlike ; unlike in pitch, intensity or
quality ; or unlike in any two or all three of these properties.
Very small differences in one property may be mistaken for
differences in another one, as pitch for intensity.
The average amount by wliich the two tones must differ in
order to be generallj^ perceived as unlike is known as the just
perceptible difference. To determine the just perceptible dif-
ference in pitch, a tone of a certain pitch is first produced, and
then another of a slightly different pitch ; the hearer states
his judgment as to whether the two tones are the same or
different.^ An apparatus for this experiment consists of two
forks of the same pitch, with a small weight at the middle of
one prong of each fork. Starting with the weights at the
middle, whereby both forks give the same tone, one of the
weights is moved upward or downward by successive steps.
The forks are sounded alternately. When the difference is
large enough to be perceived, the two forks are sounded
simultaneously, the number of beats giving the number of
vibrations by which they dijBfer.
The dependence of the just perceptible difference on the
pitch of the tone follows the general rule that the just
perceptible difference, expressed in vibrations, is smallest
with low tones and largest with high tones without the
difference being very great, and that within the range of
1 Helmholtz, Lelire v. d. Tonempfindungen, 5. Aufl., 254, Leipzig, 1896;
Keugee, as before, 334.
2 Delezenne, Mem. sur les valeurs num^riques des notes de la gamine, Eecueil
des travaux de la Soc. des Sci. de Lille, 1826-27 1 ; Seebeck, Ueber d. Fdhig-
keit d. GehSrs, sehr kleine Unterschiede d. TonhOhe :u erkennen, Ann. d. Phys. u.
Chem,, 1846 LXVIII 462 ; Preyer, Akustische UntersucUungen, Jena, 1879.
PERCEPTION OF SOUNDS 101
the tones usually employed in speech it remains practically
constant.^
The smaller the difference the greater is the amount of
mental work required for its detection. The amount of this
work may be measured by 1. the minimum time required to
perceive the difference, 2. the time required to respond to
the difference, 3. the number of mistakes made in detecting
the difference. The first method has been applied to the dis-
tinction of visual objects ; the two that are to be distinguished
are placed behind a screen and exposed for a brief interval,
the interval being increased until the difference is perceived.
The second method has been applied to visual objects and to
sounds. For the latter, two sounds slightly differing in inten-
sity or quality are used. For one of them the person responds
in one way, for the other in another way. The number may
be indefinitely increased. The greater the number of distinc-
tions required and the smaller the differences, the longer the
time required.2 For the third method the sounds may be given
in pairs irregularly in a series with pairs of like sounds. The
smaller the difference the greater the number of mistakes.
The sense of hearing is able to distinguish continuous vari-
ations in pitch with an accuracy depending on the pitch of
the tone varied, its loudness and the rate of variation. Starting
with the finger at a certain place on a violin string, we can
change the pitch of the tone continuously by sliding it one
way or the other.
Following some preliminary observations,^ the least percep-
tible change was investigated by Stern.* The tone was
produced by a current of air blowing over the mouth of a
1 LuFT, Ueber die Vnterschiedsempjindlickkeit f. Tonhbhen, Philos. Stud., 1888
IV 511 ; Meyek, Ueber die Unterschiedsempjindlichkeit f. Tonhohen, Zt. i.Fsych.
n. Phys. (1. Sinn., 1898 XVI 352.
- Summary of methods and results in Wundt, Physiol. Psychol., 4. Aufl., II
362, Leipzig, 1893.
' ScRiFTDRE, On the least perceptible variation of pitch, Amer. Jour. Psych.,
1892 IV 580 ; Ueber die Aenderungsempjindlichkeit, Zt. f. Psych, u. Phys. d. Sinn.,
189+ VI 472.
* Stern, Die Wahrnehmunq von Tonmrdnderungen, Zt. f. Psych, u. Phys. d.
Sinn., 1896 XI 1, 1899 XXI 30, 1899 XXII 1.
102
PERCEPTION OF SPEECH
bottle QF, Fig. 63). The change in pitch was brought about
by mercury flowing in at the bottom of the bottle and thus,
by changing its capacity, raising its pitch. The mercury was
sent at a definite rate into the bottle andvariator F" together.
The variator had a carefully determined form, such that the
rise of the mercury in the bottle produced an even rise of pitch.
The various rates at which the tone was altered were produced
by different rates of movement of the piston in the reservoir.'
C. The tone was begun and gradually raised in pitch until
the change was detected. Two types of change were used; .
Fig. 63.
the continuous change of pitch _/~ and the intermitted
change (or discrete difference) of pitch _~. In the former the
change from the initial tone to the final one occurred by a
slide upward (or downward) in- pitch, in the latter there was
no sound in the interval. Different rates and extents of
change were used with the results : 1. continuous changes were
more accurately perceived than the corresponding differences ;
2. the clearness with which the change in pitch was perceived
increased with the actual extent of change from the initial to
the final tone ; 3. this increase in clearness was greater for
continuous changes ; 4. with continuous changes rise in pitch
was more clearly perceived than fall in pitch, while the fall in
pitch was remarkably well perceived with discrete differences ;
5. the accuracy of perception of the likeness of two discrete
tones was much less than that of the constancy of pitch of a
PERCEPTION OF SOUNDS 103
continuous tone ; 6. the accuracy of perception of the likeness
of two successive tones was less than that of the difference of
two successive tones ; 7. the accuracy of perception increased
as the change in pitch became slower.
The application of these facts to the changes in the cord
tone in speech is apparent. For example, the lack of the
ability to perceive the variations that may arise in a long
sound permits a long vowel to gradually change its nature
and finally to become a sound that has an ending quite differ-
ent from its beginning, that is, a diphthong. This is probably
the auditory factor that in Southern British English permits
most long vowels to become diphthongs although the spelling
remains the same and the diphthongization is unconsciously
done ; thus, ' so ' is pronounced as sou, ' fate ' as feit, etc.,
and even long i and long u show similar tendencies.'^
Every sound may be said to have its range of likeness or
its limits of imperceptible difference. Thus, a tone may varj"-
within certain limits without anjr perception of its variation.
Any one of the resonance tones constituting a vowel may vary
in pitch, intensity or duration without a perception that the
sound is different. Since each element may change imper-
ceptibly in any direction the timbre, that is, the vowel char-
acter, may gradually change to a quite different one.
The judgments of likeness of objects depend upon the coin-
cidence of similar elements. Two tones are judged to be alike
in pitch if the two sensations of pitch coincide.
The judgment of likeness is" made with a feeling of certainty
that varies from absolute certaint}'' to complete uncertainty.
One person will feel absolutely sure that two tones are ex-
actly alike in pitch, another will feel less certain about it,
still another will feel somewhat doubtful and a third will feel
quite uncertain, — all of them feeling the likeness and not
perceiving any difference whatever. Similar differences in
degree of certainty will be felt by one person on different
occasions. This judgment of likeness with different degrees
of certainty is quite different from the judgment of closeness
of approach to likeness or of the degree of unlikeness.
1 SoAMES, Introduction to Phonetics, 2d ed., 48, London, 1899.
104 PERCEPTION OF SPEECH
Investigations on these points are still lacking. They might
be readily carried out with tones according to the method of
equality judgments (or of the so-called method of right and
wrong cases).! ^he improved method of computation gives
definite results.^
Large differences of pitch can be judged with considerable
accuracy. Equal differences of frequency of vibration appear
as equal differences of pitch.^ For example, the tone 384 is
selected as halfway between 256 and 512, or 360 as half-
way between 296 and 424, giving the relations 2:3:4 and
37 : 45 : 53.
Certain special relations of pitch produce the musical
intervals. The simpler intervals include the unison (1 : 1),
octave (1 : 2), fifth (2 : 3), fourth (3 : 4), major sixth (3 : 5),
major third (4 : 5), minor third (5 : 6), minor sixth (5 : 8),
minor seventh (5 : 9), major second (8 : 9), major seventh
(8 : 15), minor second (15 : 16), also the duodecime (1 : 3),
double octave (1 : 4), etc.
The musical intervals can be estimated with a degree of
accuracy that depends on the individual and on various
conditions.*
The just perceptible difference from a simple interval
may be determined by comparing two tones, one fixed and
1 Fechner, Eleraente d. Psychophysik, 2. Aufl., 134, Leipzig, 1889 ; Mullek,
Zur Grundlegung d. Psychophysik, Berlin, 1878; Wundt, Grundziige d.
physiol. Psychol., 4. Aufl., '348, Leipzig, 1893.
2 Bkhns, Ueber d. Ausgleichung statistischer ZSMung in der Psychophysik,
Philos. Stud. (Wundt), 1893 IX 1 ; briefly indicated in Scriptuke, New Psy-
chology, 269, London, 1897 ; Mosch, Zur Methode d. richtigen u. falschen FiUle
im Gebiete d. Schallempfindungen, Philos. Stud. (Wundt), 1898 XIV 491.
^ LoKENz, Untersuchungen iiber d. Auffassung v. Tondistanzen, Philos. Stud.
(Wundt), 1890 VI 26 (discussed in Zt. f. Psych, u. Phys. d. Sinn., 1890 I
419, 1891 II 266; Philos. Stud., 1891 VI 604, 1892 VII 298); Wundt,
Grundz. d. physiol. Psychol., 4. Aufl., 463, Leipzig, 1893.
* Deleziennb, M^m. sur les valeurs num&iques des notes de la gamme, Eecueil
des travaux de la Soc. des Sci. de Lille, 1826-27 1 ; Cornu et Mercadiek, Sur
les intervalles musicaux, C. r. Acad. Sci. Paris, 1869 LXVIII 301, 424; Peetek,
Ueber d. Grenzen d. Tonwahrnehmung, 38, Jena, 1876; Schischmanow, Unter-
suchungen iiber die Empfindlichkeit des Intervailsinnes, Philos. Stud. (Wundt), 1889
V 558; Stumpf tjnd Meter, Maassbest immungen iiber d. Reinheit consonanter
Intervalle, Zt. f. Psych, u. Phys. d. Siun., 1898 XVIII 321.
PERCEPTION OF SOUNDS 105
the other varied, with pitch numbers corresponding to the
relations of frequency required by the interval. The just
perceptible difference increases generally in the order of uni-
son, octave, fifth, fourth, major sixth, major third, minor third,
major second, minor sixth, minor seventh, major seventh.^
Individual differences appear in the intervals of the fourth,
third and sixth.^
Related to the just perceptible difference is the accuracy
in judging the exactness of an interval. It may be deterr
mined by using one fixed and one varied tone and requir-
ing a judgment on each occasion concerning the correctness
or incorrectness of the relation. This accuracy is nearly
the same for all the usual musical intervals.^ An increase
in a consonant interval is accompanied by a feeling of ten-
sion, sharpness or irritation ; a decrease by one of depres-
sion, shallowness or dullness.* A slightly increased major
third (4 : 5 +) and a slightly decreased minor third (5 : 6 ")
and also a slightly increased octave (1:2+) and fifth (2 : 3 "•")
are preferred by the ear to the exact intervals.^
The presence of overtones diminishes the accuracy of judg-
ment of intervals ; this accuracy is the same for the third, the
fifth and the octave ; there is a general tendency to increase
the major third, the fifth and especially the octave ; the inter-
yals of simultaneous tones are much less accurately judged
than those of successive tones ; with simultaneous tones there
is a special tendency to increase the interval.^
Coinciding overtones in the consonant intervals prob-
ably have some effect on their character. Most musical in-
1 SCHISCHMANOW, as before, 596.
2 Pkeyee, as before, 38.
3 Stumpf nND Meyee, as before ; Buch, Veber die. Verschmelzung von Emp-
Jindungen, besonders bei Klangeindrildcen, Philos. Stud. (Wundt), 1900 XV 267;
Faist, Versuche iiber Tonverschmelzung, Zt. f. Psych, u. Phys. d. Sinn., 1897 XV
102 ; Meinong und Witasbk, Zur experimentellen Bestimmung d. Tonverschmel-
zungsgrade, same, 189.
* Planck, Die natilrliche Stimmung in der modernen Vocalmusik, Viertelj. f.
Musikwiss., 1893 IX 418 ; Stumpp ttnd Meter, as before, 392.
^ Stumpf dnd Meyer, as before, 396.
o Stumpf und Meyer, as before, 400,
106 PERCEPTION OF SPEECH
struments produce notes, or tone-complexes, consisting
of partials with frequencies in the relations 1, 2, 3, 4, 5,
6, . . Two notes with fundamentals in the relation 1 to
2 (octave) will have the partials f 2 ^ t ^ s ^ i Mo . ' :}•
Two notes with the relation 2 to 3 (fifth) will have the
partials \\^ % %" 1| V fs "':;J. The relation 3 to 4
rfr^^-,r.^\^\ ryi'Troo ( 3 6 9 12 15 18 21 24 27 30 ... I
(tourth) gives jig 12 le 20 24 28 . . . } •
The other consonant intervals less than an octave have less
and less coincidence of their partials, in the order mentioned
on p. 104. The coincidence of the overtones was assumed by
Helmholtz 1 as the basis of the feeling of gratification that
accompanies consonant intervals. The consonant intervals
may appear 1. in a succession of tones, whereby the coincid-
ing partials occur as repetitions, or 2. in simultaneous tones,
whereby the coinciding partials become stronger.
When two tones differing in pitch up to about a major
second are sounded together, the result appears as a tone
of intermediate pitch ^ with beats. With a great difference
in pitch, the result appears as a combination of two tones. ^
The shortest time during which a tone must be produced
depends on whether it is to be heard 1. as an indefinite sound,
2. as a tone, or 3. as a tone of a definite pitch. The three
degrees of recognition involved in these problems require
successively longer times ; only the second problem has been
investigated.
A sensation of tone can be produced by a small number of
tuning fork vibrations,* under favorable circumstances by 1.6,
for the higher tones, to 6, as the scale is descended ; ^ or by
1 Helmholtz, Lehre v. d. Tonempfindungen, 5. Aufl., 310, Leipzig, 1896.
2 Stumpf und Metek, as before, 321.
' Kruger, as before, 324.
* Maoh, Physikalische Notizen, Lotos, 1873, 23 ; Exner, Zu7- Lehre u. d. Ge-
hoTsempJindungen, Arch. f. d. ges. Physiol. (Pfliiger), 1876 XIII 228; Auekb.4.ch,
Ueher die absolute Anzahl von Schwingungen, welche zw Erzeugung eines Tones
erforderlich sind, Ann. d. Phys. u. Chem., 1879 VI 591 ; Gelle, De la diire'e de
Vexcitation sonore n^cesxaire a In perception, C. r. Soc. de Biologie, 1886 (8) III 38.
5 ScHULZE in WuNDT, Grundziige d. physiol. Psycliol., 4. Aufl., 451, Leipzig,
1893.
PERCEPTION OF SOUNDS 107
a small number of puffs of air,i the minimum being two
according to most experimenters. An electric fork in front
of a spherical resonator can be used to produce a simple tone,
which can be carried to the ear in a distant room by a rubber
tube. A stopcock in the tube can be made to turn on the
sound for a definite time.^ The shortest audible vowels have
never been determined. The experiment might be made
in a similar way or by closing the secondary circuit for definite
intervals while a vowel is being transmitted by a telephone
from a talking machine.
Sounds enter consciousness more or less gradually. The
tone rises suddenly to near its maximum and then increases
gradually for some time. A tone c increases in its intensity
for about 48 vibrations, a tone e~^ for about 44 vibrations.^
Weak sounds may require even V to 2° to reach the maximum.*
A sound persists in consciousness after the external vibra-
tion has ceased. If a tone is produced for a very short time
at stated intervals, the tone actually heard will be a little
longer in proportion to the silence than is the case physi-
cally. If the silences are now made successively smaller, the
moment will come when the persistence of the tone will
cover the physical silence and the sound will no longer
be heard as interrupted but as continuous. A fork giving the
desired tone is kept in vibration electrically before its coiTe-
1 Pfaundler, Ueher d. geringste Amahl v. Schallimpulsen, welche zur Hervor-
bringimg eines Tones nOthig ist, Sitzber. d. k. Akad. d. Wiss. Wien, math.-
naturw. Kl., 1879 LXXVI 2. Abth., 561; Kohlkausch, Ein Beitrag zur
Kenntniss der Empfindlichkeit des Gehorsinnes, Ann. d. Phys. u. Chem., 1879 "VII
335 ; Ueber Tone, die durch eitie begrenzte Aii-ahl von Impulsen erzeugt werden, Ann.
d. Phys. u. Chem., 1880 X 1 ; Cross and Maltby, On the least number of vibra-
tions necessary to determine pitch, Proc. Amer. Acad. Arts and Sci , 1891-92, 222 ;
Herroun and Gerald, Note on the audibility of single sound waves, and the num-
ber of vibrations necessary to produce u tone, Proc. Eoy. Soc. Lond., 1802 L 318 ;
Abraham und BRnuL, Wahrnehmung kiirzester Tone und Gerdusche, Zt. f. Psych.
u. Phys. d. Sinn., 1898 XVIII 176.
^ ScHiJLZE, as before.
' ExNEK, Zur Lehre v. d. GehSrsempJindungen, Arch. f. d. ges. Physiol.
(Pfluger), 1876 XIII 234.
* UreantschitSOH, Ueber d. An- und AbJclingen alcustischer Empfindungen,
Arch. f. d. ges. Physiol. (Pfluger), 1881 XXV 323.
108 PERCEPTION OF SPEECH
spending resonator (p. 14) ; a disc with openings in it at
regular intervals revolves between them. From the smaller
end of the resonator a tube leads to the ear, the other ear
being closed. The sound can reach the ear onlj' when one
of the openings is opposite the fork. The disc is revolved at
a constantly increasing rate till, instead of a succession of
sounds, the tone of the fork is heard as a continuous sound.^
When this occurs, each sound must persist mentally in nearly
full intensity long enough to fill the silent physical interval.
The time can be calculated from the speed of the disc. These
intervals as determined by Mayer were for c~^, 0.0395' ; c",
0.0222«; c\ 0.0142"; g\ 0.0098'; c\ 0.0076'; e^, 0.0065'; g"^,
0.0060' ; c^, 0.0055'. Starting with an interval so short that
the tone appeared unbroken and increasing it until it appeared
of irregular intensity, Urbantschitsch found this interval to
vary from 0.012' for low tones to 0.006' for high ones.^ This
would indicate the minimum time necessary in order to pro-
duce a tremolo effect. High tones are relatively louder
to the ear than low ones ; for equally loud tones the time
of persistence is probably constant.^
Measurements might profitably be made of the perceptibil-
ity of speech elements as judged by the minimum time they
must last in order to be recognized. The methods would be
analogous to those used for printed letters.*
When two tones are alternated as in a trill, they must each
last at least about 0.03' in order to be heard with the trill
effect ; the figure remains nearly the same for all regions of
1 Matek, Acoustical investigations : I. Determination of the law connecting the
pitch of a sound with the duration of its residual sensation, Amer. Jour. Sci., 1874
VIII 241 ; A redetermination of the law connecting the pitch of a sound with the
duration of its residual sensation, Amer. Jour. Sci., 187.5 IX 267.
2 Urbantschitsch, Ueber das An- und Ahklingen akustischer Empfndungen,
Arch. f. d. ges. Physiol. (PaUger), 1881 XXV .323.
^ Abraham, Ueber d. AbUingen v. Tonempjjndungen, Zt. f. Psych, u. Phys. d.
Sinn., 1899 XX 417.
* Cattell, Ueber d. Trdgheit d. Netzhaut u. d. Sehcentrums, Pliilos. Stud.
(Wundt), 1886 III 94; Sanfobd, The relative legibility of small letters, Amer.
Jour. Psychol., 1888 1 402; Javal, Rev. scientifique, 1881 XXVII 802; sum-
mary in Scripture, New Psychology, Ch. VI, London, 1897.
PERCEPTION OF SOUNDS 109
the scale and for all differences of pitch between the tones.^
A succession of notes can be heard when each tone lasts at
least 0.03^
The energy, or the physical intensity, of a sound wave is
defined as the work done by it in passing through a unit sur-
face in a unit time. It is directly proportional to the square
of the amplitude and inversely to the square of the period (or
directly to the square of the frequency).^
The relation between the energy and the mental intensity
is not a simple one. Under constant conditions a tone of
a given frequency will increase and decrease in apparent
loudness as its physical intensity increases and decreases.
The relation is fairly well expressed by saying that the mental
intensity varies as the logarithm of the physical intensity,^
a relation that has been established for certain tones.* The
exact expression of the relation is / = C Ig U, where I is the
intensity of the sensation, U the physical intensity of a tone
of a certain pitch, and O a personal constant, and where Ig
indicates the logarithm with the basis e (natural, not Beiggs's
logarithm). It is not known if this relation holds true when
tones of different pitches are compared.
The faintest audible degrees of tones, noises and speech
sounds can be determined with considerable accuracy by the
differential audiometer. This is a development from the in-
duction balance.® A secondary coil S (Fig. 64) connected to
a telephone T is placed between two oppositely wound pri-
mary coils PP in series with each other and a microphone M
(transmitter). A current is sent from a battery (the figure
.shows a lamp battery DU in connection with the dynamo cir-
^ Abraham und Schafer, Ueber d. maximah Geschwindigkeit v. Tonfolgen,
Zt. f. Psych, u. Phys. d. Sinn., 1899 XX 408.
2 Rayleigh, Theory of Sound, II 17, § 245, London, 1899.
^ Fechner, Elemente der Psychophysik, 2 Aufl., Leipzig, 1889.
' WiEN, Ueber die Messung der Tonstdrke, Diss., Berlin, 1888; also in Ann. d.
Phys. u. Chem., 1889 XXXVI 8.34.
* Hughes, On an induction-currents balance, and experimental researches made
therewith, Proc. Royal See. Lond., 1879, May \b; reported in Nature, 1879
-XX 77,
no
PERCEPTION OF SPEECH
cuit) through the primary coils PP. Sounds entering the
microphone M cause fluctuations and interruptions of the cur-
rent in the primary coils PP, whenever the circuit is closed
at K. If the secondary coil S is placed near one of them,
the sound will be heard in the telephone T. As the second-
ary is moved away from the primary, the sound decreases,
becoming zero at the middle where the effects of the pri-
maries neutralize each other. A scale showing the distance
of the secondary coil from the middle can be used to indicate
degrees of intensity of the sound ; the intensity is not pro-
Fis. 64.
portional to the distance. Pressure on the key K closes the
primary circuit and sets the audiometer in operation. With
a constant source of current the measurements on different
occasions may be made comparable. The figure shows an
ammeter G- for testing the current and a resistance R for
regulating it; the microphone should be short-circuited when
this is done. The ammeter can be short-circuited when the
audiometer is being used. By means of a phonograph or
gramophone record in fixed connection with the transmitter
the initial sounds may be kept constant. The telephone of
the audiometer may be placed in a distant room so that the
person tested is in no way disturbed. This instrument has
PERCEPTION OP SOUNDS 111
not yet been applied to the study of speech, although it would
furnish accurate measurements of the audibility of speech
sounds singly and connectedly under the most varied condi-
tions of enunciation. For ordinary tests of hearing the
microphone is omitted and clicks are produced in the tele-
phone by closures at K.
A less accurate method of determining the faintest audible
sound consists in removing the source of the sound to different
distances. Accurate but rather complicated methods of
measuring the intensities of tones have been devised by
WiEN 1 and Sharps. 2
The faintest audible tone of the frequency 240 was deter-
mined by WiEN with his special apparatus. Its energy
amounted to 0.068'"'°'^, which means that the energy of the
air vibration striking the tympanum was equal to the energy
represented in a weight of l"s falling through 0.068'"' (/tt/i =
millionth part of a millimeter, mg = milligram).
The just perceptible difference in intensity varies propor-
tionately with the loudness of the tone ^ or the noise.*
Among the other mental properties of sounds the objec-
tivity and the emotional tinge may be mentioned. The objec-
tivity of a tone is a property expressing the degree to which
we consider the tone not to belong to ourselves. A tone
produced in imagination, as in mentally singing a score, has
1 WiEN, Ueher d. Messung d. Tonstarke, Diss., Berlin, 1888; also in Ann. d.
Phys. u. Chem., 1889 n. F. XXXVI 834.
^ Shakpe, a double instrument and u. double method for the measurement of
sound, Science, 1899 N. S. IX 808.
2 WiEsr, as before.
* VoLKMANN in Pechner, Blemente d. Psychophysik, 2. Aufl., I 479,
Leipzig, 1889; Renz und Wolf, Versuche iiber d. Unterscheidung differenter
Schallstarken, Arch. f. physiol. Heilk., 1856 XV 185; Tischeb, Ueber die Un-
terscheidung von Schallstarken, Philos. Stud. (Wundt), 1883 I 495; Mekkel, Das
psychophysische Grundgesetz in Bezug auf Schallstarken, Philos. Stud. (Wundt),
1888 IV 117, 251 -.Die Abhangigkeitzw. Reizu.Empfindung,Fhilos. Stud. (Wuudt),
1888 IV 541, 1889 V 245, 499; Starke, Die Messung von Schallstarken, Philos.
Stud. (Wundt), 1886 III 264 ; ZumMassder Schallstdrke, Fhilos. Stud. (Wundt),
1889 V 157; Angeli,, Untersuchungen iiber d. Schdizung von Schallintensitdten,
Philos. Stud. (Wundt), 1892 VII 414; Kampfe, ^Be^Vr. zur exper. PrUfung d.
Methode d. richt. u.falschen Falle, Philos. Stud. (Wundt), 1893 VIII 511.
112 PERCEPTION OF SPEECH
almost no objectivity ; it is attributed to our innermost self.
Tones actually sung by ourselves are more objective ; those
sung by others are highly objective, though even here the
degree of objectivity increases with the strangeness of the
singer, the distance, etc. It is quite possible that the degree
of objectivity depends closely on the action of the muscles
connected with the voice and on the sounds heard by the ear.
Even in internal song and speech the vocal organs execute
minute movements, which can be registered.^ Weak move-
ments with no sounds heard would mean little objectivity.
Stronger movements as in actual singing and speaking with
sounds heard would have more objectivity. Sounds heard
when no movements are made by ourselves would have
more nearly complete objectivity.
The sound mass that we experience at any moment is gener-
ally resolved by us mentally into groups of sounds ; thus, in
the music of an opera we attribute certain portions to the
singer, others to the violins, others to the horns, etc. Much
of this analysis can be attributed to the results of past experi-
ences ; the analysis of tone complexes, however, seems to be
a fundamental process. The nature and laws of the mental
analjrsis are still experimentally uninvestigated and little can
be said beyond the facts obtainable by unaided observation.
References
For the psychology of tone and noise : Wundt, Grundziige d. physiol.
Psychol., 4. Aufl., Leipzig, 1893 ; Stumpf, Tonpsychologie, Leipzig, 1883.
For acoustical apparatus: Koenig, Paris. For magnetic markers:
Pf.tzold, Leipzig ; Zimmermann, Leipzig ; Vekdix, Paris. For the
Galton whistle : Edelmann, jMUnchen.
^ CnKTis, Automatic movements of the larynx, Amer. Jour. Psychol., 1900 XI
237.
CHAPTER IX
PERCEPTION OP SPEECH ELEMENTS
Each individual has a system of auditory habits in the
sense that the various speech sounds are more or less familiar
to his ear. These auditory habits are intimately connected
with his habits of speech. Just as the latter have been termed
the 'basis of articulation ' (Sievbrs ^), so the former may be
called the ' basis of aural perception ' (Oeetel^) ; the united
system of habits may be expressed by the term ' basis of
speech,' or ' phonetic basis.' Each individual has his own
phonetic basis. The bases of the members of a community
are closely alike, and we may speak of the phonetic basis of
a community. In like manner the various languages may
each be said to have its peculiar phonetic basis.
Just what vocal sound is perceived by the ear depends
largely on the sensitiveness to differences (p. 100) and on the
past sounds that are most familiar. A perfectly strange
language appears to a great extent as a murmur of indefinite
sounds ; it is only by familiarity with definite sound-groups
that the ear learns to recognize the separate sounds. Finer
discrimination usually occurs only when the vocal organs are
used to imita,te the sounds, and the results of a person's own
efforts are compared with the sounds imitated. The finest
discrimination occurs when special training is directed to it.
The experiments of Muller and Pilzecker ^ in learning,
by seeing and speaking, various series of syllables, each with a
1 SiEVERS, Grundz. d. Phonetik, 4. Aufl., 105, Leipzig, 18S3.
^ Oertel, Lectures on the Study of Language, 241, New York, 1901.
" MiJLLER vmD Vn^zECK.E'R, Experimentelle Beitrage zur Lehre vom Gedacktniss,
Zt. f. Psychol, u Physiol, d. Sinn., 1901, Ergiinzungsband I, 247.
8
1 14 , PERCEPTION OF SPEECH
vowel between two consonants, showed that the best remem-
bered (or most impressive) sounds were the vowels, then the
initial consonants, then the final consonants ; that for persons
who learned mainly by the ear the German vowels and diph-
thongs appeared in the following order of decreasing effec-
tiveness, oi, oe, i, ai, 62 («A), y, au, o, a, e^ {eh), a, u, depending
evidently on their acoustic impressiveness and on their impres-
siveness through relative lack of frequency in speech ; that
for persons not inclined to ear-learning the order of decreasing
effectiveness was a, Cj, ai, au, y, a, o, ce, oi, e^, i, u ; that for
the former class the order for the final consonants was s, p,
z, m, t, f, X or c, n, 1, r, k, s ; and for the latter class s, x or c,
z, p, n, 1, t, s, k, f, m, r ; that no regular arrangement of
impressiveness could be made for the initial consonants.
These seem to be the only experiments yet made on what we
may venture to call the ' acoustic impressiveness of vocal
sounds ; ' the investigation should be extended on account of
its important bearings on the methods of teaching language
and on the study of speech changes ; the methods of experi-
menting are described in Ch. XIV below.
Just as the intensity of a sound diminishes with the dis-
tance from its source, so does the amount of the just
perceptible difference change also for some reason when
weak speech sounds are heard at different distances. The
relative distances at which the average ear can perceive and
distinguish various German speech sounds under average
conditions have been approximately determined by Wolp.i
The vowel a is heard furthest ; then at decreasing distances
there follow o, ai, e, i, au, u, s, m, n, s, f, k, t, r, b and h.
"When whispering sounds so that at 8™ none could be
understood by a listener, Rotjsselot^ found by steadily de-
creasing the distance that at a little less than 8" i was heard ;
at 7.2"" ka was heard ; at 7"" ga was heard as ka but weaker ;
1 Wolf, Neue Untersuchungen ilber HSrprufung und HSrstSrungen, Arch. f.
Augen- u. Ohreuheilk., 1873 III (2) 35.
2 EoussELOT, Les modifications phon^tiques du lungage, 38, Rev. des pat.
gallo-rom., 1891 IV, V; also separate.
PERCEPTION OF SPEECH ELEMENTS 115
at 6"" u ; at G.SS™ za and sa as sa ; at 6.10™ ba and pa as ba ;
at 5" a ; at 5.62" ta ; at 5.46" da as weaker ta, ba and pa as
pa ; at 5.15" sa and za as sa ; at 5" za distinct from sa ; at
4.90" fa ; at 4.68" va as weaker fa ; after 3" e as i or e ; at
2" ma as pa or ba ; also at 2" e distinct from i ; also at 2"
o, oe, na, la, ra ; at 1" ma ; also at 1" pa always distinct ; at
0.50" u, va as fa or va ; at 0.25" ba and pa almost completely
distinct, fa always distinct, va nearly always correctly, da
distinct ; at 0.10" ga almost always correctly, ba, pa, fa, va,
ta, da always distinct ; at 0.05" ga with perfect clearness ;
even with the lips at the ear sa and za indistinct.
With an ordinary voice such that no sound was clear at
9.60" RousSELOT found that the hearer understood a, e, i,
o, y, at 9" ; pa, ka, ta at 8.55™ ; ba, sa, sometimes sa, rarely
fa at 7.10" ; sa and fa very distinctly at 7" ; da, ma, na at
6" ; za, ga at 5.70" ; za at 5.50" ; va, u, ce at 5".
The property of speech sounds illustrated by the experi-
ments of Wolf and Rousselot I venture to term their
' acoustic penetration.' These two isolated sets of results
for particular voices on particular occasions furnish sugges-
tions for systematic investigations of the penetrative power
of the various sounds of a given language, qf a given
speaker or singer, of given conditions of mind and body,
and of given methods of speaking; and of various national
habits of speech perception, of various habits of listening, etc.
In future investigations it may be found advisable to use a
phonograph or gramophone as a constant source of sound,
and also to use the audiometer method of weakening it
(p. 109).
The perception of a sound is greatly influenced by associa-
tive suggestions. Elements are unconsciously modified, sup-
pressed or created. Even hallucinations of weak tones sup-
posed to be physically present can be readily produced in
nearly all normal individuals by appropriate suggestions from
the surroundings.^ The suggestive influence of phonetic habits
1 Seashore, Meanurements of illusions and hallucinations in normal life, Stud.
Yale Psych. Lab., 1895 III 1.
116 PERCEPTION OF 'SPEECH
is marked in causing sounds to be heard differently. Rous-
selotI relates that his sister heard mo°povpjare 'mon
pauvre Pierret ' quite correctly, but heard popovpo as
popofpo, the V appearing as f before p when there was no
suggestion from the meaning. A record of the sounds noted in
an attempt by various Germans at recording some French words
at dictation showed^ effects of suggestion from the native
language such as to produce r9V3ne, revni and rgvne for rvani,
par for pa, etc. Innumerable examples have been reported
in studies of the speech of children and of the attempts of
persons, tribes and nations to acquire foreign words.^
Associative suggestion can be used to increase the accuracy
of perception of sounds. Rotjssblot,* after two months of
careful phonetic observation, perceived for the first time the
differences between the speech of his mother and that of him-
self, and noted that a regular progression in change of dialect
could- be heard in the speech of father, son, and grandson
every time he was able to get together the members of a family;
and also that steps of phonetic change occurred from village
to village. A specimen of this latter form of change is found
in the pronunciations of ' lapin ' in various neighboring vil-
lages in France : lape" (Dinan), jape" (St. Carn^), ]ape°
(Meillac), lapi" (Corseul), lapcs (Quessoy), lapa°5 (Mon-
contour), lapea (Pl^n^e-Jugon), lape" (Cesson).
The faults of perception may be illustrated by a series of
tests made on 530 pupils of presumably somewhat varied
nationalities in a Boston public school.^ The ages ranged from
8 to 14 ; only 5 were found to be somewhat deaf to the sound
of a tuning fork. The records for a series of spoken mono-
syllables included the following specimen results :
' fan ' was recorded as : fan (511 times), than (5), fair (4), thank
(3), fell (2), — (2), clams (1), fang (1), sam (1) ;
1 RoussELOT, Les modifications phon^tigues du langnge, 40, Rev. des. pat.
gallorom., 1891 IV, V; also separate.
2 RoossELOT, Principes de phonetique expe'rimentale, 37, Paris, 1897.
' Oeetel, Lectures on the Study of Language, 242, New York, 1901.
* RoussELOT, Principes de phondtique expe'i-imentale, 43, Paris, 1897.
^ WiLTSE, Experimental, Amer. Jour. Psych., 1888 I 702.
PERCEPTION OF SPEECH ELEMENTS 117
'log' was recorded as: log (434), love (66), — (10), flog (3),
dog (3), cock (2), long (1), lo (1), lack (1), lawl (1), lord (1),
lull (1), lock (1), lough (1), loud (1), lode (1), glove (1), bog
(1), bare (1) ;
'long' was recorded as: long (497), — (11), lawn (4), log (3),
loan (3), lamb (2), alarm (1), arm (1), kong (1), lung (l),lant
(1), length (1), lur(l), love (1), lone (1), laugh (1);
' pen ' was recorded as : pen (386), hen (48), pan (47), hand (13),
ham (5), pain (4), pine (3), pail (3), head (2), paper (1),
paint (1), pear (1), pland (1), can (1), han (1), land (1), ream
(1), ten (1), then (1) ;
' dog ' was recorded as : dog (519), dug (3), dove (3), dod (1),
dollie (1), god (1), dull (1) ;
In order to furnish any useful information such results
must be carefully analyzed. This' analysis cannot be based
upon the printed forms and the conventional orthography,
but upon the actual pronunciationSj It is sounds not letters
which must be classified. The record of ' fan ' would show
the following apparent representati^ of the three sounds f,
ae, n which make up the word:
f is reflected 518 times by f, 5 times by 3, 3 times by 6,
once by s, once by kl; ae is reflected -522 times by ae,
four times by ae, twice by e ; n is reflected 516 times by
n, 4 times by r or a, 3 times by T|k, twice by 1, and once
each by t], m, and mz. '■ '
But a careful consideration of these results will show that the
sounds which reflect f, ae and o, are by no means always due
to a faulty perception of these sounds. Many times the child
had evidently missed the sound altogether, and substituted
for it some other sound which, together with the other sounds
heard, would make a word. Such cases of sound-substitution
are entirely different from those where a sound is misheard.
In ' than ' for ' fan ' we have undoubtedly a lapsus auris (the
reverse of that which transformed ' Theodor ' in Russian into
' Fedor, ' and vulgar English ' nothing ' into ' nuffin '). In
' clams ' for ' fan ' f was not misheard as kl ; the child un-
118 PERCEPTION OF SPEECH
doubtedly heard only the meaningless sound aen which called
up the sound-memory of ' clams ' and this was written down.
Carefully gathered results of this kind would indicate ap-
proximately the relative likenesses of the sounds used to the
wrongly heard sounds. The errors in perception show some
curious resemblances to the errors of child-speech and to his-
torical phonetic changes. Further investigation of such phe-
nomena is highly desirable.
The perception of a speech sound depends on the ability to
produce it. Children and foreigners do not hear the inac-
curacies of their own pronunciations. Stammering children
often reach an advanced age without discovering their defect;
some come to the laryngologist for cure because they find their
speech unintelligible to others although they do not perceive
any peculiarities in their own sounds. A case is known of
a young man with a falsetto voice who had never perceived
the peculiarity.
A speech sound produced by an individual is the result of
a very large number of fine adjustments of the speaking
apparatus influenced by an infinitude of past and present ex-
periences in hearing, thinking and speaking. The sound
varies from moment to moment and from one occasion to an-
other. With sufficiently accurate methods of measurement
no two sounds would be found alike ; the variations are
limited by the ability to perceive the differences.
Owing to the inaccuracies of sensation, sounds in close suc-
cession may differ without the difference being perceived.
Owing to the inaccuracies of movement, sounds intended
to be alike will differ from each other. Owing to the
increase of variation in action and in the least perceptible
difference in hearing when sounds are repeated at increasing
intervals, they may on different occasions differ widely
although considered to be the same. Records of different
individuals show' that language elements supposed to be
identical are often notably different within the same dialect
1 JossELYN, Etiide sur la phon&iqm italienne, 172, These, Paris, 1900; also
in La Parole, 1901 III 251.
PERCEPTION OF SPEECH ELEMENTS 119
and that the same language element is often pronounced in
very different ways by the same individual.
With the formation of habits the vocal organism (mental,
nervous and muscular) acts within a steadily decreasing range
of variation. With the repetition of an auditory stimulus
the range of variations unnoticed by the sense of hearing
becomes more limited. With a source of comparison, such as
a phonograph record or the speech of a person or of a com-
munity, the individual keeps the more closely to an average
in reproduction the more carefully and repeatedly he makes
his comparisons. These factors tend to maintain a sound at
a constant average, and, in case of a steadily progressing
change, to keep like sounds changing likewise under the
same or different conditions.
The following cases may serve to illustrate different phases
of the interaction of sensory and motor variation.
The deficiencies in perceiving the details of words ami in
associating the proper movements show themselves in the use
of wrong sounds among those that can be produced correctly.
The records of my own child at 16 months showed : seek for
sup ' soup,' although p appeared in bapa and pun for papa
'papa' and spun 'spoon; ' baka for bata 'butter' and wa-
wa for WDta, although t" was frequently used. The failure to
perceive initial sounds appeared in aet for haet ' hat ; ' when
haet was spoken with an exaggerated h, the result was gaet.
An exaggerated perception of the explosives appeared in
kaetk for kset ' cat,' hotk for hot ' hot ; ' doka and doga for dog
" dog.' Mistaken perceptions of sonancy occurred in zuga for
suga ' sugar,' gam for kam ' come.' The use of the glottal
catch instead of t occurred many times in mu for mit ' meat ; '
the sound of > was distinct from that of t and the action of
the glottal catch could be felt by the finger over the larynx.
The child had never heard- a glottal catch. At later dates
some of these errors were perceived ; thus mit for mit was
used along with mi>, and wawa often corrected spontaneously
to wota. At 18 months he used oloba for olova ' all over '
and kuXit for kulit ' cool it ' (spoken with the Z-mouill^
120 PERCEPTION OF SPEECH
which he had never heard), although he could use v and 1 in
veigud 'very good' and labmama 'love mamma.'
When a group of fairly constant sounds is heard repeatedly,
there is a tendency to hear any slightly different sound as one
of the constant group. This is favored by the indefiniteness
of any sound owing to the size of the just perceptible differ-
ence for each of its elements and the indefiniteness of
remembered sounds. This phenomenon may be termed ' the
identification of similar sounds.' The processes of assimi-
lation of neighboring sounds that occur regularly in language
are permitted by the lack of distinction of difference in the
results. Thus in twenti ' twenty ' the w has become partly
devocalized on account of the greater ease in speaking it in
this way ; the change is permitted by the failure of the ear
ordinarily to detect it. A similar example is paedzs for paedz
' pads' ' before a pause or a voiceless sound.
JossELYN reports the case of a friend with a good musical
ear who insists that his own tlu tlgks tlaen does not differ
from the klu klaks klaen which he hears and that both begin
with k ; in these words he makes 1 merely a lateral surd ex-
plosion of the t and fails to notice the lack of the proper
k-explosion and the loss of sonancy in the 1.
The ear may fail to notice a gradual lengthening of the rush
of air at the end of an explosive, which arises from gradually
increasing slowness in moving the tongue or lips. The
occlusives thus become aspirated ^ ; and the aspirate may de-
velop into an independent sound. For example, t->t^->^th, as
seen in French to" 'ton,' German t'^on 'ton,' Danish thui^a
' tange.' If the channel is made farther forward in the cavity
of the mouth, the rush of air produces a fricative consonant
instead of an h. The character of the fricative varies accord-
ing to the place of stricture. Danish thuiig ' tunge ' may thus
become tsuiia,^ a development parallel to the second German
sound-shifting. The affricatee may further be simplified to
simple spirants, ts->^s . The whole development arises from
1 Storbi, Englische Philologie, 2. Aufl., 74, Leipzig, 1892.
^ Storm, as before, 74.
PERCEPTION OF SPEECH ELEMENTS 121
the tendency to an alteration in muscular action permitted
by the failure of the ear to distinguish the successive stages of
tlie change. In this way t->t^-^th^tB-+ts^ss-+s. Thus the
original word that appears in English as tel ' tell ' appears in
Danisli as thalg and with further development in German as
tselan. Likewise to the English hit 'heat' corresponds
German hitsg ' hitze ' and to hot ' hot ' the German hais
' heiss.'
The conscious, semi-conscious or unconscious distinction of
differences is a form of mental work, the repetition of the same
sound involving less perceptive stimulation than a series of
different sounds. The ' harmony of the vowels ' in some lan-
guages seems favored by auditory as well as by motor econ-
omy. In these languages the vowels in a word must belong
to the same group or must be the same, except in so far as
other influences are at work.^ For example, in Hungarian the
hard vowels form one group and the soft vowels another;
the language establishes a principle of vowel-harmonj^, ac-
cording to which in general only vowels of the same class
may occur in a word.^ This principle requires in many cases
two kinds of sufBx for the same meaning: 'hd,zn^l,' at the
house ; ' kertn^l,' at the garden ; ' irnak,' they write ; ' k^rnek,'
they ask. Similar requirements of vowel-harmony appear
in Finnish, Northern Turkish and other languages of the
Ural-Altaic group. Sporadic examples occur in the Arian
languages ; Heraclean x"^P"'^°'^ corresponds to Homeric
')^epaBo'i, in Latin nihil stands for *nehil. In alimentum
and monumentum the vowels i and u respectively are de-
termined by the color of the vowel in the preceding syl-
lable, a and o.^ In the French and Canadian dialectic
forms klerte = ' clart^ ' and serite = ' charitd ' the change
appears to be due to harmonic assimilation of the first vowel
1 References in Passy, Changements phon&iques, 186, Thfese, Paris, 1891.
2 Balassa, Phonetik d.hmgarischen Sprache, Inter. Zt. f. allg. Sprachw., 1889
IV I.M.
** Brugmann, Grundriss der vergleichenden Grammatikder indogermanischen
Sprachen, I 2. Halfte, 2. Aufl., 835 § 962 f., Strasburg, 1897.
122 PERCEPTION OF SPEECH
to the last one. Consonant harmony is not unusual. It
occurs in isolated cases in the colloquial forms of all lan-
guages,^ as well as constantly in infant speech.
It has been shown by Laclotte ^ that the tongue movements
used in producing a vowel affect not only the preceding con-
sonant but also the vowel before that consonant. The ten-
dency to assimilate the articulations for the former of two
vowels to those for the latter may be an impulse toward unity
of character, but such an impulse must, I believe, be favored
and developed by the ear in order to be effective.
The unconscious desire to avoid the labor of perceiving a
difference may be one reason why the two portions of a diph-
thong often show a tendency to assimilation. Sometimes the
second portion approaches the former in character and is ab-
sorbed by it : thus ai^a in Old English ' stan ; ' ei->^e in Swed-
ish and Danish ' sten.' Sometimes the latter portion prevails ;
thus ei, oi, ei and ui have all become i in Modern Greek
pronunciation. Sometimes the two elements of a diphthong
approach each other in character and form a vowel of inter-
mediate character ; thus au of Latin ' aurum ' has become o
in Spanish 'oro' and French 'or;' French ai is regularly
pronounced e.
To lighten the work of distinguishing among sounds that
resemble one another small differences may be exaggerated
or like sounds may be- made different. These phenomena
may take part in the development of diphthongs out of long
vowels. The diphthongization started by any cause per-
mitted to act on account of the inability to perceive varia-
tion is exaggerated as soon as the vowel is felt — even
dimly — to be a diphthong that must be pronounced in that
way. This usually affects the first portion most; often it
ultimately becomes a quite different vowel; thug hus ->
hous -* haus ' house ; ' rim -> reim -> raim ' rime.' These
changes are complicated by motor factors, but the motor
adjustments are governed largely by the ear. Such phenom-
1 Passt, as before, 189.
2 Laclotte, L'karmonie vocalique, La Parole, 1899 I 177.
PERCEPTION OF SPEECH ELEMENTS 123
ena as the tendency of English long vowels to become open
may be mainly or entirely due to auditory preferences and not
to motor habits; observations on the deaf might -be of value
here.
The result of the neglect of some differences and the
exaggeration of others has been stated by Passy to be that
'language tends constantly to bring into prominence what
is necessary.' 1 It sometimes favors, sometimes opposes an-
other principle, that of ease of speaking. Both are prin-
ciples of economy; from a psychological point of view the
former might be called that of 'perceptive economy,' the
other that of ' motor economy.' Perceptive economy re-
quires not only the suppression of needless distinctions but
also the emphasis of needful ones.
Closely connected with the variations in perception are the
changes in memory. Variations of the same sound fuse to
the same auditory-motor memorj'- image ; the memory images,
with their indefiniteness and their progressive changes, are
the important factors of many speech changes. Thus, the
general change of initial t to ts in Old High German prob-
ably arose from the change of initial t to ts when followed by
i or e ; this brought with it the change of initial t in tu and
to on account of the union of all t's in one memory-image.^
The gradual changes in the general speech of a commu-
nity, a family or an individual are brought about by internal or
external factors causing a shifting of the general average of
the sounds within the limits of the just perceptible difference.
The unnoticed variations in pronunciation are factors of
change in language. They are particularly effective in the
language of children whose peculiarities of speech are largely
due to deficient perception of sounds and their combinations,
to incomplete and erroneous associations with sounds in
memory, to imperfect control over the vocal organs and to
incorrect associations between the sounds heard from others,
1 Passy, as before, 227.
' 2 Kabsten, Sprecheinheiten u. deren Rolle in Lautwandel u. Laulgesetz, Trans.
Mod. Lang. Assoc. Amer., 1890 III 190; also in Phonet. Stud., 1890 III 5.
124 PERCEPTION OF SPEECH
the movements made by themselves and the sounds of their
own speech. Such variations cannot, however, furnish an
adequate explanation for the historical changes of speech
sounds, for the variability in articulation cannot be regarded
as the cause of a phonetic change. It must be regarded
rather as the condition under which certain changes are per-
mitted to take place. The ultimate causes for such changes
must be sought elsewhere. Variability of articulation can-
not work a change unless all deviations, or at least a large
majority of them, are in one direction. The force or forces
which turn them into the same direction are the real causes
for the change.^
Some of the causes of phonetic changes are certainly purely
auditory, others are purely motor, many are due to phe-
nomena of association and memory; on nearly every point,
however, experimental data are still lacking although it
would nofc be difficult to devise methods of investigation.
As a guide for future work we may assume the principle of
the unified nature of the human constitution. In its applica-
tion to an individual this principle would lead us to expect to
find in speech and language the more or less modified pro-
cesses that appear in other activities. Thus, as instinctive
economy shows itself in every form of perception (visual,
tactual), in every form of thought (scientific, literary), and
in every activity (writing, manual labor), we can with confi-
dence assume its action in speech and language. Again, the
modification of effort as the speed is increased, which is a
form of economy, must appear in speech because it appears
in all mental activities. Numerous other laws of human
activity are to be looked for in the study of speech for the
same reason. In regard to various individuals and communi-
ties our fundamental principle entitles us to expect as much
agreement in general with variation in particulars as we find
in anatomical, physiological and psychological work.
The facts summarized in this chapter suggest various ap-
plications to the methods of learning foreign languages:
1 Oertel, Lectures on the Study of Language, 103, 246, 269, New York, 1901.
PERCEPTION OF SPEECH ELEMENTS 125
1. the value of training of the sense of hearing to perceive (at
first consciously and then sub-consciously) the fiaer distinc-
tions among speech elements in spite of their associations
with other sounds ; this may be done by direct auditory train-
ing in distinguishing differences (p. 116), by aid from produc-
ing the sound (p. 118), by records of vocal movement (Part
III), etc. : 2. the training of the sense of hearing to correctly
perceive and distinguish complexes of sounds as wholes
without noticing the elements; this may be done by using
words, phrases, etc. in the ways just suggested.
References
For summary of the phenomena of sound change : Passy, Change-
ments phonetiques, Ch. Ill, Paris, 1891 ; Paul, Principien der Sprach-
gesohichte, Ch. Ill, Halle, 1898 ; Sweet, History of English Sounds,
Oxford, 1888; History of Language, Ch. Ill, New York, 1901; Oertel,
Lectures on the Study of Language, Lect. Ill, New York, 1901.
CHAPTER X
SPEECH IDEAS
The current of thought in consciousness varies in its
density from moment to moment. The regions of less
density may be used to divide off parts of greater density ; .
such portions of greater density are what we usually term
' ideas,' or ' thoughts.' Each denser portion of the speech
current in consciousness is an * auditory idea 'or — as a matter
of speech — a 'phonetic unit.' A word is sometimes said to
be the expression of one idea ; this is probably always true
for disconnected words. In connected speech, however, an
idea is usually expressed by several words. Thus, in the
phrase ' I, said the fly, with my little eye ' in a gramophone
record of a recitation of Cock Rohin'^ the idea ' I ' is followed
by the complex idea ' said the fly,' the subordinate idea ' with '
and the single idea ' my little eye.' In the last case the
thought in the mind of the speaker was evidently a particular
eye; the fact of its being 'mine,' and of its being 'little'
were only dimly present and really served only to make up
the details of the picture. The words of this group can
be heard to run together; in the tracing from the plate
there was no break in the vibrations of the speech curve.
The group ' my little eye ' was evidently a phonetic unit.
If by a ' word ' we understand the group of sounds or letters
usually considered as a linguistic unit, this phonetic unit
consisted of several words. The fusion of words into pho-
netic units appears often in the mistakes of foreigners and
1 Scripture, Researches in experimental phonetics (first series), Stud. Yale
Psych. Lab., 1889 VII 35.
SPEECH IDEAS 127
children. l^Its applications to the laws of phonetic change
have been pointed out by Jespeesen.^ i
It is to be noted that the term ' word ' is ordinarily used in
an inconsistent and arbitrary fashion. Thus ' another ' is one
word and ' a different ' two, ' city hall ' always two words,
' railway ' one word, 'street car ' two, ' Newhaven ' in England
one, ' New Haven ' in America two, etc., although in each of
these cases only one phonetic unit is present. The com-
pounding of words that occurs in German makes them more
closely represent phonetic units. ' Versicherungsgesellschaf t '
and ' insurance company ' both indicate one idea ; both are
spoken with no pause ; the English division into two words
is merely a matter of typography. ' Versicherungsgesell-
schaftsgebaude ' and ' building of the insurance company ' are
likewise psychologically identical though typographically
different. German compounds are in no way longer than or
different from English ones except in being written and
printed without spaces. In both languages the accent is the
unifying element, distinguishing in English, for instance, the
compound ' blackbird ' from the two independent words
' black bird.'
A ' phonetic unit ' in the meaning in which I have used
the term is to be distingviished from a ' phonetic element.'
QCabsten ^ considers that in each language we are to assume
/single sounds for such simplest sound-memories as we can
prove to exist ; thus the two Roumanian forms of t indicate
that at a previous time two memory-groups of t must have
arisen from the single Latin memory-group. SuchP single
sounds may properly be called 'phonetic elements.' ^Both
; terms, ' phonetic unit ' and ' phonetic element,' are psychologi-
' cal ones ; ) the ' phonetic element ' is the simplest (possible
' phonetic unit ' that can be proved to exist, while the
' phonetic unit ' is a synthesis of elements that varies on each
occasion. In like manner it may be possible to define words
1 Jespeksen, Zur Lautgesetzfrage, Int. Zt. f. allg. Sprachw., 1887 III 193.
2 ICaksten, Sprecheinheiten u. deren Rolle in Lautwandel u. Lautgesetz, Trans.
Mod. Lang. Assoc. Amer., 1890 III 195 ; also in Phonet. Stud., 1900 III 9.
128 PERCEPTION OF SPEECH
as such combinations of phonetic elements as can be proven
to have independent existence in memory.^
It may be considered as well established that printed words
are perceived in wholes as ideograms and not as combinations
of letters. In this respect English words are not to be dis-
tinguished from Chinese characters, and they are often not
inferior to them in complexity and useless adornment. We
may — I believe — even go further and say that groups of
words ordinarily form single ideograms, the undistinguished
elements acting to produce a combination of marks of suffi-
cient peculiarity to arouse a certain idea. Finally, it is
doubtless true that complexes of sounds in words or phrases
act also as ideograms. ^^This ideographic characteristic of a
word is indicated by several facts. It takes no more time to
recognize a short printed word than to recognize one of its
letters. 2 Among the phenomena of aphasia it has sometimes
been noticed that ' entire words are read more promptly than
the letters composing them ' and that ' words which were read
correctly were spelled wrongly,' the patient often attempting
from the sound of the perceived word to guess at the spelling
although he had the letters before him. Experiments by
GoLDSCHBiDER and MiJLLBE 3 showed that the perception
of certain determining letters was sufficient for the recogni-
tion of the word. Eedmann and Dodge* have shown that
words may be perceived under conditions that exclude any
perception of the single elements.
In experiments to determine the amount of change in a
printed word, used as an ideogram, that might be made with-
out a change in the perceived Avord, Pillsbtjry^ exposed
words containing various misprints. An omitted letter was
■ Kaksten, as before.
2 Oattell, Ueber d. Trdgheit d. Netzhaut und d. Sehcentrums, Philos. Stud.
(Wundt), 1886 III 97.
8 GoLDSCiiETDER UND MiJi.LEE, ZuT Psychologie u. Pathologic des Lesens,
Zt. f. klin. Med., 1894 XXIII 130.
« Ekdmanx und Dodge, Psychologische Untersuclrangen iiber das Lesen auf
exper. Grundlage, Halle, 1898.
<> PiLLSBUKY, The reading of words, Amer. Jour. Psychol., 1897 VIII 333.
SPEECH IDEAS 129
most often noticed, a wrong letter less and a blurred letter
least. The characterless blur of a mutilated letter furnished
more suggestive material for the mind to make into the
correct one than a wrong letter did ; an omitted letter altered
the picture of the syllable. Thus, ' shabbilw ' briefly exposed
was read as ' shabbily ' and associated with the word ' gen-
teelly,' the subject declaring that he saw the word spelled
' shabbily.' ' Eaxth ' was read as ' earth,' ' fashxon ' as
* fashion,' ' cotton ' as ' custard,' ' ordnary ' as ' ordinary,' etc.
The same phenomenon occurs even when the word is
left exposed indefinitely. I have found the following cases
in some experiments on the association of ideas. The word
shown (indicated by small capitals) aroused generally a
visual association (likewise indicated) or a motor word (in-
dicated by small letters). The records included :i beruf
— brief [before the word was fully recognized] — beruf-
stand — berufstiichtigkeit ; KARA vane — [at first read as]
KRAVATE — karavane — afrika, absicht — arbeit [because
only beginning and end noticed]. Closely related to this is
the involuntary completion of a word as in the case : the —
THEE — [a feeling that this is incorrect] — the door. Similar
cases have been observed by Cordes.'"' The word Scholk
appeared at once to the subject as schalk with the con-
sciousness that something else was present; the word im-
perative appeared as imperator before the word was read
to the end.
The effect of preceding suggestions on the perception of
words has been noticed by Munsterberg ^ and Pillsbury.*
For example, the word ' lid ' was called out just before the
letters ' xover ' were shown for an instant too brief for read-
ing them ; the subject supposed that he saw the word ' stove.'
1 Scripture, Ueher d, associativen Verlauf d. Vorstellungen, Philos. Stud.
<Wundt), 1891 VII 5-1.
2 CoKDES, Exper. Untersuchungen iiber Associationen, Philos. Stud. (Wundt),
1901 XVII 66.
s MiJNSTERBERG, Studien zur Assoc iationslehre, Beitrage zur experimen-
tellen Psychologie, 1892 IV 20.
4 PiLLSEURY, as before.
130 PERCEPTION OF SPEECH
The word ' small ' called out beforehand caused ' greal ' to
Be read as 'small,' the letters supposed to be seen being 'mal.'
The most striking characteristics of the whole word (or ideo-
gram) are perceived and begin to bring up some internal
word with the rest of the idea of which it is a component.
This internal word contains many visual elements that often
suffice to complete a defective visual word, or even to
modify it. ' If A, is seen as the second letter in a word, it is
associable with s or t, among others, for the first letter. If
the form of the word and some other letter suggest that the
word, as a whole, is should rather than though, the sh con-
nection will be effective.' ^ The condition of mind largely
determines how the perceived outline is completed.
That the perception of printed words does not occur by
letters or always even by single words is a fact clearly shown
by visual lapses, or misreadings. The phenomena are familiar
to everj' one ; a careful collection of cases has been made by
Meringee and Mayee.^ The following typical forms have
been noted : 1. exchanges of neighboring words (' zu viel '
for ' viel zu ') or of near sounds (zoka for koza) ; 2. anticipa-
tions of words (' des Brutus Casar Liebe zum Casar ' for ' des
Brutus Liebe . . .') or of sounds (' verspateter ' for 'verpes-
teter,' ' tanzen sahe ' for ' tanzen sahe') ; 3. postpositions of
words and syllables (seldom or never except in disease) or of
sounds (' er wiirschet euch zu sehn' for ' er wiinschet . . .') ;
4. rearrangements ( ' weiberfrucht ' for ' weiberfurcht ' ) ;
5. omissions of words [common] and syllables (' abhangen '
for ' abzuhangen ') or of sounds (' stet ' for ' stets) '. Many of
the other misreadings observed by Meeinger and Mayer
were due to other causes than union in visual perception.
Observations on the misunderstandings of auditory words
show that the vowel of the root syllable and the vowels in
general are most often heard correctly while the consonants
(especially the initial ones) are often heard wrongly.^ ' Feld
1 PiLLSBUKT, as before.
2 Mekingek nND Matek, Versprechen und Verlesen, 100, Stuttgart, 1895.
" Mekinger und Mater, as before, 157.
SPEECH IDEAS 131
im Meere ' was understood for ' Feld in Mahren,' ' Vetter
aus Kroke ' for ' Vetter aus Chikago,' ' Htihner isst ' for
' jiinger ist,' ' Bahnen ' for ' Vulkane.'
Bagley^ made phonograph records of mutilated words
(a) used without context, (5) used with one or two related
words, (c) used at the beginning of a complete sentence, (jT)
used in the middle of a complete sentence, (e) used at the end
of a complete sentence. The mutilations consisted in omis-
sion of an initial, medial or final consonant; for example,
' the book was put a-ide, ' ' the siege was interrupted by a
tru— . ' The observer was instructed to listen to the sentence
as reproduced and to repeat it to the operator, who recorded
it as given by the observer, noting the errors. Baglby
draws the following conclusions concerning the perception of
auditory words :
1. In monosyllabic words the elision of the initial conso-
nant affects perception more than the elision of the iinal
consonant.
2. When a word is given with one or two related words,
the chances for its correct perception (that is, perception of
the word in spite of the mutilation) are increased by 82% as
compared with the chances without context.
3. When the mutilated words are placed in a sentence
instead of being isolated, the chances for correct perception
are remarkably increased.
4. In the middle of a complete" sentence there is a signifi-
cant increase in the chances of correct perception as compared
with the chances at the beginning.
5. The position most favorable for correct perception is at
the end of the sentence.
6. Elision of p, t, k, b, d, g works the greatest injury to
the perception of a mutilated word ; elision of w, r, 1, j the
least (vowels not considered).
7. The mutilated word with context is not, as a rule, filled
out at once by aid of the sounds contained in it but by the
' Baglet, Apperception of the spoken sentence, Amer. Jour. Psychol., 1900
XII 80.
132 PERCEPTION OF SPEECH
idea of the correct word aroused by associations derived from
the context. Thus, with ' the matter is a function of ti —
and space ' one observer perceived ' ti — ' as ' tide ' but sub-
stituted ' time ' by association with ' space ; ' with ' wri-ing
in pain, he called for help ' one observer supplied ' writhing '
after perceiving ' pain. '
It may be suggested that auditory words and phrases form
' ideophones ' just as printed ones form ' ideograms.' The
further distinctions may be made of ideograms and ideophones
into sensory (visual words and auditory words) and motor
ones (written words and spoken words).
In all probability the most prominent features of a phonetic
unit are first perceived and the details are gradually filled in.
In the case of finger reading by the blind this double process
occurs through separate organs ; the right hand precedes in
traveling across the line of raised letters and furnishes a
vague idea of the words which is filled in by the left hand
following it.^
An idea is a more or less complicated union of elements.
These elements are derived from past and present experiences.
The idea ' milk ' as it occurs to me at the present moment is
the result of past and present experiences of sight, taste, touch,
etc., of the object itself and of associated experiences such
as visual, auditory and motor words. It is generally stated
that the idea of the object occurs first and the word follows ;
there seems to be no justification for this separation in time ;
experiences of every kind may arise at the same time.
The sum of the speech-elements in an idea is called the
' internal word ; ' visual, auditory, arm-motor, and voice-
motor groups, or factors, are distinguished. Some persons
see the printed word in memory most prominently, others
hear it, others speak it. When the motor factors are
strong, they can be detected in the action of the muscles.
Unconscious larynx movements in connection with internal
words have been observed in experiments by Hansen and
1 Hellek, Studien zur BUndenpsychologie, Philos. Stud. (Wundt), 1895
XI 460.
SPEECH IDEAS 133
Lehmann ^ and have been registered by Cuetis.^ The arm
movements have been observed by Cumbeelaxd.^
The most prominent elements of internal words are the
motor sensations of the vocal organs and the auditory sen-
sations. The idea is generally most closely connected with
these and to a less degree with the printed letters and the arm
sensations. The closer the connection between the various
factors and the greater the accurate familiarity with each, the
more complete the internal word.
The advantage of fusing the word with the object into one
idea appeared in an experiment by Bell * in teaching a deaf
child. The method has been extended to the instruction of
normal children. ^ Printed words like bed, dooe, window
are placed on the objects that the child sees and the words
are repeated as frequently as possible while the objects are
seen or handled. The result is a very intimate fusion of
the language elements with the other ones in the various
ideas of the child's experience.
In regard to acquiring a foreign language the facts men-
tioned in this chapter seem to indicate : 1. the desirability of
correctly associating groups of words to single ideas without
the necessity of thinking out the details (p. 126); 2. the
advantage of learning words and phrases as ideograms and
ideophones (p. 128); 3. the desirability of learning the details
of such ideo-units in the earlier lessons in order to form
the correct associations between them and their meanings
(p. 130); 4. the naturalness of first noting the most prominent
characteristics of ideograms and ideophones and gradually
distinguishing their details (p. 132); 5. the profitableness of
^ Hansen und Lehmann, Ueber unioillkurliches Flilstern, Philos. Stud.
(Wnndt), 1895 XI 471 ; Scripture, New Psychology, 63, 259, London, 1897.
2 CuKTis, Automatic movements of the larynx, Amer. Jour. Psych., 1900
XI 237.
3 Cumberland, A thought reader's experiences. Nineteenth Century, 1886
XX 867; Scripture, New Psychology, 255, London, 1897.
* Bell, Upon a method of teaching language to a veri/ young congenitally deaj
child, Amer. Annals of the Deaf and Dumb, 1883 XXVIII 124.
* Scripture, In the Japanese way. Outlook, 1897 LV 557.
134 PERCEPTION OF SPEECH
closely uniting the internal word to the idea of the object
itself (p. 132) ; the advantage of emphasizing the auditory and
motor elements in building up the internal word (p. 133).
Rbfeeences
For internal speech : Aubert, Die innerliche Sprache und ihr Verhalten
zu den Sinneswahrnehmungen und Bewegungen, Tit. f . Psych, u. Phys. d.
Sinn., 1890 I 52 ; Baldwin, Internal speech and song, Philos. Rev., 1897
II 385 ; Ballet, Le langage interieur et les diverses formes de I'aphasie,
2e edit., Paris, 1888 ; Egger, La parole intdrieure, Paris, 1881 ; Klein-
PAUL, Sprache ohne Worte, Leipzig, 1889 ; Laupts, Enquele sur le
langage interieur, Arch. Anth. Crim., 1895 X 128, 478, 609 ; 1896 XI 96,
307; Netteb, La parole interieure et I'Sme, Paris, 1892; Paulhan, Le
langage interieur, Rev. philos., 1886 XXI 34.
CHAPTER XI
ASSOCIATION OF IDEAS
An idea is generally followed by another whose content
stands to that of the former in the relation of previous con-
tiguity in space or time, of similarity, or of contrast.^ These
relations have been termed ' laws of association ; ' they have
proved useful in arranging objects for memorizing. It has
been pointed out ^ that these so-called ' laws ' are simply
classes in which pairs of associated ideas may be placed. It
is also true^ that the various classifications of associations
are merely classifications of objects as usually associated, or of
the relations of the meanings of words associated together. It
has been said that they have been made on logical and not
on psychological principles.*
Herbaet ^ treated ideas as definitely bounded groups of
sensations capable of indefinite existence outside of conscious-
ness ; they favored or opposed each other according to certain
formulas and thereby rose or fell in consciousness. Owing
to the lack of experimental data such a mechanics of ideas
was necessarily fanciful.
1 Aristoteles, De meinoria, Ch. II, 451 b 16f.
2 Wdndt, Grundziige d. phyaiol. Psychol., 4. Aufl., II 453, Leipzig, 1893.
8 Thumb und Maeee, Exper. Untersuchungen ilber d. psycholog. Grundlagea
d. sprachlichen Analogiebildung, 14, Leipzig, 1901.
* Okth, in a review contained in Zt. f. pad. Psychol, u, Path., 1901 III 222.
" Herbaet, Psychologie als Wissenschaft, Konigsberg, 1824; Lehrbuch zur
Psychologie, 2. Aufl., Konigsberg, 1834; Sammtliche Werke, herausg. von
Hartenstein, V, VI, VII; Deobisch, Empirische Psychologie, Leipzig, 1842;
Erste Griindlinien d. math. Psychol., Leipzig, 1850; Volkmann, Lehrbuch d.
Psychologie, Cothen, 1884; Ziehen, Das Verhaltniss d. Herbart'schen Psychol,
zur physiol. -exper. Psychol., Samm. v. Abhandl. aus d. Gebiete d. pad. Psychol, u.
Physiol., 1900 III Heft 5.
136 PERCEPTION OF SPEECH
One attempt to treat the results of an experimental investi-
gation showed that ideas could not be considered as definite
objects in any way, and resulted in the attempt to develop a
new theory.^ The technique of the experiment consisted
essentially in placing the subject in a dark compartment with
a ground-glass or tissue-paper screen on which the image of
a printed word or a picture was projected by a lens provided
with a photographic shutter. The subject observed and
stated the train of thought aroused by each image. This
arrangement in various modifications has proved useful in
further investigations.^ In stating the results of experiments
in the following pages visual images are indicated by small
capitals, motor images by lower case roman letters and auditory
images by italics.
The general characteristics of an association may be stated
in the following way. At any given instant t^ the roind con-
sists of an immense complexity of elements of various inten-
sities. One group of elements is of maximum intensity ;
others are in all degrees down to a faintness of intensity such
that their presence can be proved only by indirect means.*
For example, the mental condition at the present moment
may be that of looking at a book ; the book may be considered
as the idea present in mind at the moment, but all the other
things seen, heard, felt and thought at the same moment —
no matter how dimly — and all the unperceived elements of
mind foiTa part of it.
The main group in consciousness undergoes more or less
rapid change in its elements ; some remain fairly constant ;
some disappear ; some new ones appear. At a moment t-^ the
most intense group may be partly or wholly different from
1 Scripture, Ueber d. assoc. Verlauf d. Vorstetlungen, Diss., Leipzig, 1891;
also in Philos. Stud. (Wundt), 1891 VII 1; New Psychology, Ch. XIII,
London, 1897.
2 MuNSTERBERG, Studien zur Associationslehre, Beitr. z. exper. Psychol., 1892
IV 20; PiLLSBDRT, ^ studi/ in apperception, Amer. Jour. Psychol., 1897 VIII
313; CoRDES, Experimentelle Untersuchungen uber Associationen, Philos. Stud.
(Wundt), 1901 XVII 30.
3 ScEiPTDRE, as hefore, 136; New Psychology, 205, 391.
ASSOCIATION OF IDEAS 137
that of the moment t^. Thus, the thought of the book at the
moment ig was followed by the memory of a certain class-
room on a certain occasion at the moment t-^. This did not
occur through the substitution of one group by another, but
through gradual changes (at different rates) of the elements
of the fiist group and the gradual formation of the new
group. The total content of the mind at the moment t^ and
the changes during the time from t^ to t.^ determine the content
at the moment ij. An ' idea ' is not an object of unchange-
able form that appears and disappears but is a group of activ-
ities extended in time. For example, the term ' boat ' is given
to an object of fairly stable existence ; although the boat
may not be present to the senses, yet it is assumed to exist
and to be somewhere. Our idea of a boat is not of this
nature ; when we are not thinking of the boat, no idea of it
exists ; when we think of it, the idea forms, acts and passes
again out of existence.
An ' idea ' is a sum of conscious elements sufficiently distinct
from other elements to be more or less definitely marked off as a
group. An idea may be considered as a region of greater
density in the course of thought.
The persistence and disappearance of an element in an idea
depend on its intensity and on its connections with other ele-
ments of mental life. Memory elements have been shown ^ to
fade away at first rapidly and then more and more slowly, al-
ways approaching to but never quite reaching complete loss —
the curve of memory being an asymptotic one. Repetition of an
element adds to its intensity; with sufficient repetition its
strength even at a much later time will keep it ready for pro-
minence in consciousness by slight new additions. The de-
finiteness of an element, measured by its least perceptible change
(p. 101) or by the frequency of confusion with a slightly
different element (p. 103), has been shown to decrease with
time ; this renders it more and more likely to appear the same
as elements that originally differed somewhat from it.
1 Ebbinqhatjs, Ueber d. Gedachtniss, Leipzig, 1885 ; Wolfe, Untersuchungen
iiber das Tongeddchtniss, Philos. Stud. (Wandt), 1886 III 534.
138 PERCEPTION OF SPEECH
The new elements entering consciousness by the senses find
at hand in all degrees of intensity innumerable elements from
past experiences, and the resulting union of coinciding elements
with omission of disparate ones brings a new idea into
prominence and at the same time modifies it. The union of
certain elements of a new idea with familiar ones of previous
experience may be seen in the case where the printed word
ABTEi was at first almost read as aebeit and in the numerous
cases of mistaken perception of misspelled and misspoken
words (above). A word may be perceived as a word and then
followed by a thought of its meaning, but usually the per-
ception of the word and the thought of its meaning occur as
one act.i In either case the very first member in an associa-
tion is an assimilation of a group of sensations into a complex
of present and revived sensations ; in the former the sensory
elements are prominent, in the latter they are less so.
Each element in an idea undergoes changes in intensity
and in definiteness owing to the influences of other elements ;
these steadily modify the idea until in a short time it seems a
different one. Such changes are evident in the case where a
picture of a scorpion was followed 1. by a memory of a picture
of a scorpion, 2. in a class-room, 3. by a memory of a teacher
discoursing concerning it. The taste of tea on one occasion
suggested 1. a taste memory of the almost tasteless solution
of a homeopathic medicine, 2. a visual memory of such a solu-
tion in a glass, 3. a memory of the sensations of a throat sickness.
The steady development of the mental picture is apparent in
each case. The gradual grouping of mental elements is like-
wise seen in examples like that in which the printed word
SEWING was followed by a visual picture of a person sew-
ing, which ' gradually became that of a definite person in
the act of sewing.' The manner in which many elements are
strengthened so that a series of different groups appears as a
succession of ideas is seen in a case where a picture of a deer
was followed 1. by a visual picture of the land where deer are
1 CoRDES, Exper. Untersuchungen iiher Associationen, Philos. Stud. (Wundt),
1901 XVII 30.
ASSOCIATION OF IDEAS 139
used as draught animals, 2. by a memory of a deer seen in a
forest, and 3. by a memory of a picture of a deer seen long
before.
The processes that change one idea to another may be said
to result in the loss of some of the elements of the first and the
addition of new elements.^ Thus, the printed word fluc.h
suggested to an Englishman the printed word flush ; the four
letters flu h had been more often connected in a certain order
with s than with o ; consequently the c was suppressed and the
s added. Similar associations by the same person (English)
were eahm — eaum (the word ' Rahm ' was unknown
to him), SBD — SAID — SEED, LBEO — LEAP, KOT — COD,
GACOLUSIM — COLOSSAL — COLOSSEUM ; some of these indi-
cate that the visual and motor words were acting together. In
many cases the association consisted in simply adding new ele-
ments : HOHL — ■ HOHLBN, MON — MONTAG. Similar associations
by an American were : muhe — muhsam, kot — cotton,
LEFO — LBPEE, MASS — MASSACHUSETTS (with this Subject
English words were followed almost without exception by vis-
ual memories of objects and scenes, not by words). Another
American associated : this — that, goodness — goodness —
badness, from — from here — from this, sat — Saturday, is
— is not — be, of — of — of him, very — very — very true,
HOW — how — do you do, the — thee — (feeling of dissatis-
faction) — THE DOOR, SPURS — SUPURS. Some of the results
recorded for a German were : klug — und weise, fluch —
der bosen tat, mass — mass f iir mass, A — B, eaub — thier,
MUHE — los, etc., nearly all word associations being simply
additions. The records of a Japanese gave : hohl — hohb,
KLUG — heit, A — aal — B, beruf — brief — beruf sstand —
berufstiichtigkeit, etc.
It still remains to explain why some elements persist and
some disappear and why certain new elements arise. The
statement that ' elements that have been present together tend
to recall each other ' gives the principle on which objects are
to be used in order to form associations but does not indicate
1 Scripture, Verlauf d. VorstelL, as before, 18.
140 PERCEPTION OF SPEECH
the mental processes. The whole course of thought may be,
I believe, explained on the assumption of three principles : 1.
every element of thought fades more or less rapidly in inten-
sity in an asymptotic course (p. 137) ; 2. every element loses
its definiteness more or less rapidly also in an asymptotic way
(p. 137) ; 3. elements of the same kind are added. The first two
principles are famihar phenomena of memory. The definite-
ness of an idea is measured by its eonfusibility vsdth another
idea (p. 103). Ideas that are confusible with each other are to
be considered as the same. The addition of the same elements
is a familiar phenomenon in sensation; stimuh, as- of the skin,
too weak to be perceived finally become noticed and even
painful if frequently repeated ; similar summations of weak
stimuli have been noticed in experiments on muscles, nerves,
and the brain.
The whole course of thought at any time may, I believe, be
treated as consisting of all previous experiences, which are
fading in intensity and definiteness without ever being entirely
lost, which are being fused with each other whenever the
definiteness is so far diminished that they are practically the
same, and which rise into prominence according as the fusion
produces sufficient intensity. The entrance of a new sensation
adds new elements ; the resulting perception depends upon the
degrees of intensity and definiteness of the elements already
on hand at the moment of entrance; the course of thought
then follows on the usual principles with the result that at the
next moment a new combination reaches prominence.
The theory advocated was promised on a previous occasion.^
It differs from that of Hbrbart in denying definite boun-
daries to ideas and the principles of attraction and repul-
sion between them, biit resembles it in considering that ele-
ments of ideas have an existence in the mind although not
perceived.
This view is contrary to that of many contemporary psychol-
ogists. Their view is that when an idea fades away, it ceases
to exist, that it leaves brain adjustments behind it, and that,
' ScKiPTUKE, as before, 101.
Association of ideas 141
when these adjustments again affect brain action, the idea
reappears. Such a presentation of the case seems inade-
quate. Brain action is an uninterrupted sequence of
physiological activities. Mind action must, I believe, be
analogously treated as a continuity of mental processes. The
two sets of phenomena are, of course, closely related; for
convenience we may perhaps mix them in a discussion. But
the supposition that mental facts are accidental attachments
to members of the brain sequence is only a little less futile
than the one that they form a sequence of brain-mind-
brain-mind, etc. The former contradicts the fundamental
hypothesis of physiology ; the latter that — according to my
opinion — of psychology, namely, that the whole of mental
life must be explained by reference to elementary mental
processes. The habit among some writers of ' explaining '
any psychological difficulty by inserting ' sequences of neural
processes ' — generally inconsistent with the later discoveries
of neurology and often absurd — between mental ones has
been a great hindrance to the development of psychology ; its
appearance in linguistics would be deplorable.
Wundt's view ^ is a development of the principles of simi-
larity and contiguity formerly adduced as explanations of
association; it may be summarized as follows. The associa-
tive processes cannot occur between ideas but only between
the elements that compose them. A ' reproduced ' element
generally belongs to many previous ideas. Only two processes
are present in the association of ideas, namely, connection of
like elements, and connection of those that have entered into
a functional relation by occurrence together. These may be
called ' connection by likeness ' and ' connection by contact.'
Every association of ideas involves association of elements by
both connections. At the first moment an idea reawakens the
same elements of earlier impressions in a simultaneous associa-
tion with the result of greater prominence of familiar elements
and neglect of unfamiliar ones. As I have already pointed
■out, the principles of Hkeness and contact are classifications ;
^ WuNDT, Grundziige d. physiol. Psychol., 4. Aufl., II 467, Leipzig, 1893.
142 PERCEPTION OF SPEECH
the reawakening of the same elements involves a sleeping ex-
istence of elements in the mind similar to that for Hbrbaet's
complete ideas ; the neglect of unfamiliar elements also resem-
bles Herbaet's conflict of ideas.
According to Cordes,^ a single element or a complex of
elements in an idea becomes specially prominent on account of
favoring internal or external conditions; the other elements
fade away while the more prominent elements persist; new
elements assimilate themselves to the persisting ones and
form a new idea. This is a common form of association. In
other cases the new elements attract so much attention to
themselves that the persisting ones are neglected and rapidly
fade away ; in such a case the induced idea appears to contain
little or nothing of the preceding one. By special attention
certain elements of an idea can be made to persist through
a series of associations ; this principle is of great importance
in learning languages.
An idea may bring two or more ideas by association inde-
pendently of one another. The following cases occurred in
my records : A — aal — B ; GACOLUSIM — gahcia — gladstone,
is — is not — be, both the associated ideas being plainly
connected with the inducing idea and not with each other. In
like manner the word filz aroused^ the associations 1. of
brown felt, used in sound-boxes, 2. the economic term ' verfil-
zen ; ' the two associated ideas have no connection with each
other ; we may suppose that they were both aroused by filz
and that one of them developed more rapidly than the other.
The united effect of two ideas in producing a succession
can be seen in a case by Cordes ^ in which the simultaneous
presentation of a tone from a musical instrument and of the
word KLEIN was followed by the memory of a very small tun-
ing fork ; in another case a tone and the picture of Emperor
William II. was followed by the thought of the Hymn to
^Egir. In general we may say that two simultaneous ideas
have an effect that depends on their relative masses; if
1 Cordes, as before, 58. ^ Cordes, as before, 48.
" Cordes, as before, 61.
ASSOCIATION OP IDEAS 143
one of the ideas is overpoweringly weighty, the next idea
will be chiefly influenced by it ; but if two are nearly bal-
anced the next idea will be the result of the two. It has
been noted by Mayer and Oeth^ that when a person re-
sponds bj- an associated word to the word he hears^ the asso-
ciation may be more than a simple one. Thus, the word ' stif t '
was followed by the visual memory of a student friend of that
name and the person responded by the word ' student ; ' the
word ' lead ' was followed by a distinct visual image of a flat
piece of grayish-white lead, after which the response 'heavy'
was made ; ' soul ' was followed by the internal word ' body '
after which the response ' mind ' was made.
In many cases the course of thought seems to have no very
definite region of density. Such cases as those reported in some
association experiments by Coedes ^ are common ; upon see-
ing the word medici the observer gave no definite association
but said, ' I thought of the whole epoch in art that is repre-
sented by the name ; ' to the word donchj^ey he ' associated
the whole complex from Zola's Deb&cle, the word macmahon
came late and only because he wished something definite.'
In other cases there may be a sufficiently definite idea with a
whole region of semi-definite elements, as in the record by
Coedes where sulla was followed by an idea of Catiline
' together with a picture of the map of Italy and dimmer
thoughts of the wars that have occurred there.' On another
occasion the word stahl aroused ^ ' a lot of reproductions,
apparently simultaneous: an image of a fine piece of steel,
blue ; memory of a touch impression, that reminded of the steel
works of native place; Stahl as the name of a medical man
and of a philosopher ; the ideas seemed to be simultaneous ;
they came one after another into the focus of consciousness,
but with each it was clear that it was already present ; very
quick ; very fine phenomenon.'
1 Mater und Orth, Zur qual. Untersuchung d. Association, 7it. f. Psychol, u.
Physiol, d. Sinn., 1901 XXVI 1.
2 Coedes, as before, 51.
8 CoRDES, as before, 52.
144 PERCEPTION OF SPEECH
This generally overlooked existence of regions of not very-
definite density has been specially emphasized by Mayer and
Orth.^ The persons experimented upon often said that they
experienced certain mental conditions which they could not call
definite ideas or acts of will. Sometimes these vague condi-
tions acquired some definiteness ; for example, the word ' mus-
tard ' was followed by a pecuhar mental phenomenon that
seemed describable as a memory of a common proverb [without
the proverb actually being thought of] and the response ' seed '
was made.
On some occasions a word is followed by another appar-
ently entirely different; it is generally found, however, that
there is some common mental characteristic in the two words.
This common mental characteristic may be either their connec-
tion on some previous occasion, or their connection with other
words or objects that nevertheless do not present themselves.
The former characteristic seems to have no psychological
meaning except as a paraphrase of the latter. A connection
on some previous occasion is a statement of a fact of the
past ; adduced as an explanation of the association it implies
some present fact. In many cases of association the fact
present is the persistence of some elements from the inducing
idea into the induced ideas ; in cases where no such persist-
ence can be detected it may well be assumed that elements
have persisted without any notice of them. In my experi-
ments there were some associations of different words that
could be readily traced to a previous connection : zatjm —
ztigel (zaum und ztigel), sed — but, this — that, etc. They
were not common ; a word was most frequently followed by
words to complete a phrase or by memory pictures. Even
with foreign words the associations consisted generally in
adding elements of the same language or in bringing up
memory pictures, and not often in translating. Although
the similarity associations are among the most common visual
ones, they do not occur frequently between words. The few
examples found in the investigation mentioned were : truppe
1 Matek und Okth, as before, 5.
ASSOCIATION OF IDEAS 145
— mantiver [probably through memory images of scenes or
words], TKXTPPE — menge [likewise], zaum [confused with
ZATJisr] — hecke. AU these cases of ' previous connection '
are to be interpreted by the following principles.
An association where the word A is followed by a word 0
on account of relations of both A and O to an idea (or
group of ideas) B which does not fully or at all enter con-
sciousness has been called a 'mediate association.' In order to
investigate mediate association experimentally, the following
method was devised.^ On one card there was a German word
and some Japanese characters. On another card there was a
strange word (Japanese, in Roman letters), with the same
characters. A series of cards, with the same number of
each kind, was shown in irregular order. For example, in
one experiment the following series (the Japanese characters
being represented here by Greek letters) was shown in the
order here given : — (1) hana a/3, (2) hito 7S, (3) iukxt e?,
(4) KUEXJ 7)6, (5) MENSCH 7S, (6) GEHEN 69, (7) KOMMEN r)6, (8)
BLXTME a^. The subject was asked to state if he had noticed
any associations between the first four words and the second
four; he said he had not. Thereupon the words alone without
the characters were shown him, with the request to state the
first thing that entered his mind after each. The results were
as follows : — (1) hito — mensch, (2) ktjrtj — kommen,
(3) HANA — ? (4) ITJKU — GEGEN, (5) KOMMEN — HJKU, (6)
GEHEN — ? (7) MENSCH — HITO, (8) BLTJME — HANA. At
the end the subject declared that all the associations were
involuntary, that he could give no reason for them and
that the Japanese characters had not occurred to him at
all. Several of these associations were, however, correct:
it seems probable that they were brought about by the influ-
ence of the Japanese characters which, nevertheless, had not
entered into consciousness. This probability is increased by
other experiments in which the word-association was correctly
made, and was followed by the occurrence of the characters.
In still other cases the association was correctly formed with-
1 Scripture, as before, 81 ; New Psychology, 202, London, 1897.
10
146 PERCEPTION OF SPEECH
out thought of the characters, whereas the subject could re-
produce them when asked. Finally, the characters themselves
were found to be in all stages of indefiniteness and forgot-
tenness, even in correct associations. In experiments made by
AscHAFFENBUKG ^ a number of cases occurred in which the
connection between the two ideas was intelligible only on the
supposition of an intermediate idea. In the greatest number
of cases this intermediate idea appeared to have been a sound-
association with the inducing idea, while the relation of the
induced idea to the intermediate one was of any kind.
Several , excellent examples of mediate association were
observed by Coedes.^ In one case the word balde was fol-
lowed by a memory of ' Gothe's garden house at Weimar —
white, two-storied house on the side of a hill, clear — then
(that it was later, I can say with perfect certainty) the words :
" warte nur, balde ruhest du auch " — then the thought :
" ah, that explains it." '
In my experiments I pointed out that the mediate idea B
was found in all degrees of consciousness, from full conscious-
ness where the succession of ideas appeared as ^ — ^ — C, to
complete lack of consciousness where B was unnoticed and
the succession appeared as ^ — C with possible or even impos-
sible introspective recovery of B. The view of Cordes ^ is an
amplification of mine: mediate association is often to be
considered as a special case of the usual immediate association
where the second idea is a complex idea out of which certain
elements become specially prominent and form an idea whose
relation [through identical elements] to the first one can
readily be found ; and in other cases as an immediate associa-
tion where the usual formation of a definite idea out of the
complex of the second idea [through identical elements] was
hindered in some way, with the result that an unusual group
was formed.
1 AscHAFFENBUKO, Experimentelle Studien iiber Associationen, I. Theil,
Psychol. Arbeiten (Krapelin), 1895 I 40; Thomas, Ein weiteres Beispid von
Association durch eine Gerur.hempfindung, Zt. f. Psychol, u. Physiol, d. Sinn., 1895
XII 60.
2 CoBDES, before, 74. a Cordes as, as before, 73.
ASSOCIATION OF IDEAS
147
Related to mediate association is association through un-
noticed sensations or unconscious memories. In addition to
observations of such occurrences ^ they have been produced in
careful experiments.^ A series of four or five cards, each
containing a picture in the middle and a small letter or char-
acter in one corner, were shown in succession for so short a
time that, at the most, the subject was able to recognize only
the picture without the small letter. Thereafter the small
characters were exhibited alone, and the subject had to state
which of the pictures first occurred to him. The following
is a specimen series, the pictures being indicated by words.
Peacock.
*F
Shield.
A
Cat.
I
Flag.
Negro.
C
The results on one occasion were : I — cat, : : — flag, A — shield,
C — negro, F — ? . Upon being questioned, the observer stated
that he had not recognized any of the small characters in the
original series. Many similar results were obtained. In some
cases the subject would feel that a certain picture belonged
to a certain letter, although he had not seen the letter before
during the experiment, as far as his knowledge went.
Forgotten associations may also have their effects. On
seventeen successive days Kbapelin^ used the same series
of inducing ideas, and measured the quickness of association.
This increased during the first few days, and then remained at
a constant level of about one half the original time. After an
interval of one and three-quarter years the same ideas were
used among others in measurements of quickness ; for these
ideas the time was much shorter than for the others, although
1 Jerusalem, Ein Beispiel von Association durch unbewusste Mittelglieder,
Philos. Stud. (Wundt), 1894 X323; WnNDT, Sind die Mittelglieder einer mittel-
haren Association bewusst oder unbewussfi Philos. Stud., 1894X326.
^ ScRiPTnKE, as before, 136.
* Kbapelin, Ueber den Einfluss der Vebung auf die Dauer von Associationen,
St. Petersburg, med. "Wochenschrift, 1889, No. 1 (cited from Aschaffenburg).
148 PERCEPTION OF SPEECH
the' earlier associations had been forgotten. The greater ease
and rapidity of familiar associations as compared with un-
famihar ones is shown in the greater facility in learning
oriental languages by the use of Roman letters rather than
strange ones.^
Cases frequently occuried in my experiments in which the
train of thought had started before the word or object was
clearly perceived. Similar observations have been made by
MiJNSTEEBEEG, CoEDBS and others.
Mediate association is probably the source of many asso-
ciations said to occur by similarity and contrast, both the
inducing and induced ideas being connected with a group of
ideas more or less dimly in consciousness. Other similarity
and contrast associations contain identical elements that are
fully conscious. These phenomena of mediate association
have been used to explain the so-called ' free rise ' of ideas in
consciousness, that is, the appearance of ideas quite uncon-
nected with the preceding ideas. It does in fact account for
many of them. I do not believe, however, that such a ' free
rise ' is impossible ; it may occur spontaneously whenever
enough elements have become so indefinite (p. 140) as to
unite with others to form a sum of sufficient intensity.
The influence of mediate association appears clearly in
the memory experiments of Ebbinghaus,''' of Mullee and
Schumann^ and of Mullee and Pilzeckee.* In a series
of syllables successively read in trochaic rhythm associa-
tions are formed not only between neighboring syllables but
also between distant ones, especially between the emphasized
ones.
Associations may be formed between percepts of the same
sense (visual with visual, auditory with auditory, motor with
motor) or between those of different senses (visual with motor,
1 Sweet, Practical Study of Language, 36, London, 1900.
2 Ebbinghaus, Ueber d. Gediiohtniss, 139, Leipzig, 1885.
8 MuLLEK UND Sciiu.MANN, Exper. Bettrage zur Untersuch. d. Geddcknisses,
Zt. f. Psychol, u. Physiol, d. Sinn., 1893 VI 307.
* MtJLLEK UND PiLZECKEK, Exper. Beitrdge zur Lehre vom Gedachtniss, Zt. f.
Psychol, u. Physiol, d. Sinn., 1900, Erganzuugsband, I 216.
ASSOCIATION OF IDEAS 149
etc.). This is true not only of different forms of the same
word but also of different words.^ Thus, the two syllables
' bas ' and ' dut ' may be so connected that the sight of ' bas '
at once produces the motor response ' dut ' without inter-
mediate association of motor ' bas ' or visual ' dut.'
In the preceding account of the association of ideas no use
has been made of the term ' association by function.' The
meaning of this term may be illustrated by the following
quotation from Paul : ' We hear from time to time a num-
ber of sentences which are built up in the same way and which
therefore unite to a group. The memory of the special con-
tent of the individual sentences may fade more and more ;
the common element becomes strengthened by the constant
repetition; and thus the rule [for the construction of a sen-
tence] is unconsciously abstracted from the examples.' ^ It is
not necessary to assume a special form of mental activity for
such associations by function ; the cases can, I believe, all be
treated just as other associations. In speaking of ' me ' a
person has in mind a picture including not only the word but
a multitude of sensations and thoughts, past and present,
concerning himself among which are numerous elements of
past experiences in which he and others have been the object
of some action. To such past cases he has been taught to
associate the words ' me ' and ' him,' not ' I ' and ' he.' A
new experience suggested by ' me ' may bring ' him ' on
account of the apparently identical elements in the general
picture; the fading away of the definiteness of the two
objects in the original pictures brings an unconscious fusion
in memory of the previous objects of an action, and when a
distinction of persons is called for at a later time the objec-
tivity will still remain. The use of ' knowed ' as the past of
' know ' is due to the habit of using d when referring to the past
just as a word 'yesterday' or 'then' might be used. The
other functional associations and the formation of sentence-
habits may be reduced to the same principles. In going over
1 M&LLER UKD PiLZECKEK, as before, 12.
2 Paul, Principien der Sprachgeschichte, 3. Aufl., 100, Halle, 1898.
150 PERCEPTION OF SPEECH
Paul's examples ^ of speech changes due to association by
function I find that every one can be explained as I have indi-
cated. The mental process in using the Attic Greek ttoXltov
by association of the masculine genitive of the first declension
"with the genitive ov of the second,^ or ' Berthas ' for the
German genitive of ' Bertha,' is exactly as described for the
case of ' me' above. The associations may be considered as
' associations by function ' if this term is taken to refer only
to the use of the words associated ; the assumption of a mental
process of association by function is, I beheve, a needless one.
We must undoubtedly admit an association by function ^
just as we do an association of part to whole, an association
by contrast, etc., but as pointed out above (p. 135), these
terms are no more than convenient classifications of the
relations of the things associated. In this chapter an attempt
has been made to explain how the associations actually do
take place in the mind; what is philologically a functional
association is psychologically a process not essentially dif-
ferent from the other associations. The cases of ' agramma-
tism ' * are explainable as due to weakening of the fundamental
processes of memory and association.
For instruction in language the following conclusions are
suggested by the facts of this chapter: 1. a word-idea
should be learned as parts of various courses of thought
in order to form the necessary language associations in
addition to being learned separately in the earlier lessons;
2. the learner should make an effort to actively produce
the complete idea with its object-memories and internal
word and not rely on the so-called ' spontaneous ' rise of
the idea from memory; 3. constant repetition is neces-
' Paul, as before, 106.
2 Brugmann, Griechische Grammatik, 3. Aufl., 224, Munchen, 1900.
5 Paul, as before ; Thumb und Marbe, Experimentelle Untersuchungen ii. d.
psychol. Grundlagen d. sprachl. Analogiebildung, 61, Leipzig, 1901 ; Oertel,
Lectures on the Study of Language, 156, New York, 1901.
* Krapelin, Psychiatrie, 6. Aufl., II, 312, Leipzig, 1899; Heller, Ueber
Aphasie bei Idioten u. Imbecillen, Zt. f. Psychol, u. Physiol, d. Sinn., 1897 XIII
181.
ASSOCIATION OF IDEAS 151
sary; 4. both ear and eye should be carefully trained to
quick and accurate perception; 5. the learner should be
trained to associate the ideas and words that actually occur
in the language studied and to avoid other associations, that
is, as far as possible to learn exclusively in the language
studied with avoidance of translating into or thinking in the
native language; 6. the language instruction should be as-
sociated as far as practicable with surroundings that tend to
arouse and confine the thoughts peculiar to the language,
as occurs by residence in the region to which the language
belongs, by appropriate furnishings of the room, etc. ; 7.
words and phrases should be associated as much as possible
with objects or pictures; 8. the most frequently needed
words of a language should be learned in as many different
connections as may be practicable; 9. the associations of
language elements should be made so firm that they occur
even before the ideas reach full consciousness ; 10. owing to
the inability of speech to keep up with the rapidity of thought,
the formation of ideophones should be encouraged.
Repeeences
For a summary of data concerning association of ideas : Wundt,
Grundz, d. physiol. Psychol., 4. Aufl., Leipzig, 1893 ; Grundriss d. Psychol.,
4. Aufl., Leipzig, 1901 ; Jodl, Lehrbuch d. Psychol., Stuttgart, 1896;
Ebbinghaus, Grundz. d. Psychol. Leipzig, 1902 ; James, Principles of
Psychology, New York, 1890 ; Ladd, Outlines of Descriptive Psychology,
New York, 1898 ; and the various other works on psychology.
CHAPTER XII
HABITS OP ASSOCIATTON
The forms of association represent habits of thought and
speech. Some of these forms are easier than others for each
individual, community, language, etc.
The firmness with which ideas are associated may vary.
Two measures of firmness have been proposed: the recip-
rocal of the time required, and the relative frequency.
The time required for an association may be roughly meas-
ured by starting a stop watch as the inducing word or
picture is presented, and stopping it when the reply comes.
Finer measurements may be made by means of chronoscopes
with appropriate connections. Various chronoscopes have
been devised by Hipp, Ewald, D'Aesonval, and others ;
the pendulum chronoscope with its attachments has been
specially constructed for work on association and similar
problems.
The pendulum chronoscope (Fig. 65) contains in the first
place an accurately adjusted double-bob pendulum. This is
held by a catch at the right-hand side. In making an experi-
ment this catch is pressed noiselessly and the pendulum starts
on its swing. It carries on its lower bob a piece of magnet
iron that holds to itself a light indicator in an accurately but
automatically adjusted position. This indicator has two
branches, a point passing in front of the scale and a flat piece
passing behind it. This indicator replaces the older hanging
pointer shown in the figure. At a definite instant the pen-
dulum in its swing releases a catch and thereby drops a shut-
ter that covered an opening in a screen at the back of the
chronoscope. Behind this screen there may be a word which
HABITS OF ASSOCIATION
153
is exposed by the fall of the shutter. The person experi-
mented upon responds to this word by repeating it or by
calling out some association. A telegraph key held in a clamp
(fastened to the back of a chair) is adjusted so that its knob
¥lQ. 65.
is pressed Ughtly against the chin of the subject. The
moment this knob is pressed a current is sent through a
magnet under the chronoscope ; this releases a horizontal bar
moving behind the scale. Since the flat part of the pointer
swings between the scale and this bar, it is picked off the
pendulum when the bar snaps. In making the instrument a
zero-mark is placed on the scale plate at the instant the shutter
154 PERCEPTION OF SPEECH
starts to fall ; the other divisions of the scale in hundredths
are so carefuUy registered on the scale plate that a well-
made instrument gives readings accurate to less than half a
thousandth of a second. The finer divisions into thousandths
may be marked on the scale or, preferably, left for the
eye.
Behind the shutter there is a wheel carrying catches for
small cards with words or pictures on them. The subject is
seated so that he can see the shutter-opening comfortably.
When the subject is to be placed in a distant room, a break
contact attached to the shutter is used to operate a distant
magnetic shutter ; otherwise the manipulation is the same.
For response to a spoken word a magnetic release of the
pendulum is substituted for the catch that usually holds it,
the pendulum being now at zero ; this release is operated by
a key under the chin of the speaker. When the subject is
in a distant room, the inducing word is spoken into a tele-
phone or a speaking tube.
A hp key or a voice key may be used in any of these ex-
periments in place of the chin key. The lip key ^ is so ar-
ranged that the movement of the lips interrupts an electric
circuit and sends the current through the chronoscope mag-
net. The voice key is essentially a tube shaped at one end so
as to fit closely over the mouth, and closed at the other by a
diaphragm of leather, ^ metal, ^ or membrane,* which is agitated
by the vibrations from the mouth. A contact in the middle
of the diaphragm serves to interrupt an electric circuit with
1 CatteLL, Psychometrische Untersuchungen, Philos. Stud. (Wundt), 1886
III 312 ; Keapelin, Ueber d. Beeinflussung psychischer Vorgange durch einige
Arzueimittel, Jena, 1892; MiJLLER und Pilzeckee, as before, 7.
2 Cattell, as before, 313.
3 BouDET DE Paris, Des applications du telephone et du microphone a la
physiologic et k la clinique, Paris, 1880; Boddet de Paris, fitude de voix arti-
cule'e, Paris, 1880 ; Sceiptcke, So>ne new apparatus. Stud. Yale Psych. Lab., 1895
III 107 ; Thinking, Feeling, Doing, 53, New York, 1901.
* RoussELOT, Les modijications phone'tir/nes du langage €tudiees dans le paiois
d'une, famille de Cellefrouin, Revuedes patois gallo-romans, 1891 IV, V; also sepa-
rate ; RoussELOT, La methode graphique applique'e h la recherche des transfor-
mations inconscientes du langage, Paris, 1891.
HABITS OF ASSOCIATION
155
each vibration. The voice key shown in Fig. 66 has a metal
diaphragm D which touches the platinum point of the ad-
justable screw ;iS' at the slightest agitation ; the electric cir-
cuit B with poles at B and *S' is thus closed.
In using the chronoscope in this way the latent time of a
magnet should always be known. A simple change of wires
from the chin-key to the contact on the shutter causes the
chronoscope to pick off its own pointer ; the time registered
is that of the action of the magnet, the lost time of the other
mechanism having been included in the scale-graduation ; this
time is very constant at 0.007' to 0.009= according to the
strength of the current in relation to the adjustment of the
sensitiveness of the release. This adjustment of the release
Fig. 66.
is accomplished by a special screw at one side of the
apparatus.
To find the time required to read a word aloud a card with
that word is inserted behind the shutter ; the pendulum is re-
leased ; the shutter drops at zero ; the subject responds when
he sees the word; the indicator is caught off at the mo-
ment of response ; and the time elapsed is read from the scale.
This may be called the time of a visual-motor word-associa-
tion. To find the time of an association of ideas the person
is to call out not the word he sees but the first other word
that occurs to him ; the increase is called the association-time.
For each set of experiments the average and the im-
mediate probable error are computed. This latter quantity
— of great value as an indication of stability in the mental
operation — is calculated by finding the differences f i, v^.
156 PERCEPTION OF SPEECH
. . ., v„ between each separate result asj, x^, • • ., »„ and the
average a (that is, v-^ = ojj — «, v^ =^ x^ — a, . . ., v„ = x„ —
a), squaring the differences and adding the squares (obtaining
^1^ + ^2^ + ■ ■ ■ + ^3^)' dividing the sum by one less than the
number of experiments (w — 1), and taking two thirds of the
square root of the quotient. The complete process of finding
the average and the probable error from the individual results
a;j, x^, . . ., x^ is indicated by the equations
X-i -\- Xif -\- , , , -\- X ji^
a = —
n
^j= Xy — a, v^ = x^— a, . . ., v„ = Xn — «,
» V}. — 1
Among the measurements of association time that find their
application in speech those of Traittscholdt, Cattell and
AscHAPPENBTJEG may be mentioned. The experiments were
not undertaken vsdth any thought of such an application and
the data for our purposes are not numerous. Investigations
on the firmness of language associations, on the influence of
alliteration and rime, on language habits, and similar topics are
still to be made. The experiments by Teautscholdt ^ showed
that the quickest associations were those between words used
in common connections (gold — silver, hght — dark). Cat-
tell ^ has determined the time required by two persons for
associations of various kinds ; some of the averages are given
in the table opposite for B, a German, and C, an American.
The influence of language habits is marked. The shortest
associations are those between words of two languages. The
other short associations such as 'city — its country,' 'month
— its season,' ' month — following month ' ' prominent man
— his profession ' owe their facility to the numerous repe-
1 Tkautsoholdt, Exper, Untersuchiingen ilber d. Assoc, d. Vorstellungen,
Philos. Stud. (Wundt), 1883 I 24L
^ Cattell, Psychometrische Untersuchungen, Philos. Stud. (Wundt), 1886 III
452, 1888 IV 241.
HABITS OF ASSOCIATION
157
titions that have been heard, seen and made by the person.
Less past familiarity explains also the longer time for ' month
— preceding month ' and for translations of the longer
■words. Other longer times as for ' language — author,'
' author — one of his works,' ' class of objects — some
example,' seem to be explainable only on the supposition
that the first idea induces more than one association, and
that extra time is required for one of these to become more
prominent than the other.
B C
letter ...
group of letters
picture
picture
short English word
long English word .
short German word
long German word .
city
month . .
month
month .
author .
prominent man
country
season
language . .
author . .
class of objects
picture ...
adjective
verb • . . .
verb . . .
Experiments of somewhat different kind by Binet and
Hbnki 1 showed that the time necessary to repeat the heard
word ce" ' un' was 0.22^ when the word was known before-
hand, 0.54' when it was not, 0.78' when the word was to be
repeated not in the same accentuation ; that the time required
to cease producing a continued sound at a signal was 0.27';
and to stop counting a series of figures was 0.34'.
According to Aschapfenbueg^ the external associations —
1 BiNET ET Henki, Les actions d'arret dans les phe'nomenes de la parole,
Rev. phil., 1894, XXX VII 608.
2 AsCHAFFENEUEG, Exper. Studien iiber Associationen, Psychol. Arbeiten
(Kriipelin), 1896 I 209.
its name
0.43"
0,46'
the word .
40
40
its name ....
47
54
its name in a foreign Ian
guage
64
69
German word
22
25
German word
32
38
English word .
27
15
English word . . .
58
40
its country . . . .
34
43
its season
41
31
following months . .
34
38
preceding month
71
82
his language .
41
34
his profession . .
45
35
one of its cities .
38
34
one of its months . .
55
42
an author ...
68
52
one of his works
108
68
some example
69
50
a part of it .
38
43
substantive
85
33
subject .
81
51
object
61
35
158 PERCEPTION OF SPEECH
that is, of objects connected in space and time, or by speech
usages or by similarities of sound — are more common and take
less time than internal or logical associations. When a spoken
word is associated to an auditory word, the connection may oc-
cur directly or with one or more ideas between them.^ The
latter is the most common case ; it requires somewhat more
time.
According to experiments made by Beegstrom ^ the time
required for sorting a series of cards into piles was lengthened if
they had previously been sorted into piles differently arranged,
or if the different arrangement had been learned by the eye or
the ear alone. This showed the interference of a previous
habit with a present activity ; the effect decreased with the
time that had elapsed. It presumably increased with the firm-
ness of the previous habit. This interference of an asso-
ciation by a shortly preceding one is seen in the difficulty
of alternatmg s and s in similar associations as in ' Shall he
sell sea shells ? Shall she sell sea shells ? ' and in similar
confusions for other sounds where a group of associated
articulatory movements is to be sometimes varied by a small
change. This phenomenon may be called ' associative stam-
mering ; ' hke stammering it is due to improper coordination
of muscular movements, but the trouble arises from associative
interferences in thought and not from defects of muscular con-
trol. The term ' stuttering ' (Lautstottern) is hardly justifi-
able, as the mental and physiological processes utterly lack the
excessive innervations and muscular cramps that characterize
stuttering.
The interference of associations is assigned by Wheelee ^
as the compelling force that may carry a sound-change from
one word to another. For example, the later habit of pro-
nouncing u as ju in words like 'new' and 'Tuesday' ac-
1 Mayer dnd Orth, Zur qual. Untersuchung d. Association, Zt. f. Psych, u.
Phys, d. Sinn., 1901 XXVI 1.
2 Bergstrom, Exper. on physiol. memory by means of the interference of associa-
tions. Am. Jour. Psychol., 1893 V 356.
' Wheeler, The causes of uniformity in phonetic change, Trans. Amer. Philol.
Assoc, 1901 XXXII 6.
HABITS OF ASSOCIATION 159
quired in opposition to the earlier one of using u may lead a
person to extend the use of ju to most cases of u, as in ' tune,
due,' etc., and even in ' do ' and ' two.' The laws governing
such interfering associations might be investigated by time-
measurements according to the methods of this chapter or
by studies of the speech curves.
The firmness of an association may be judged bydts frequency.
No investigations have been conducted on the frequencies of
specific associations. Data concerning some such associations
have been obtained by statistics of favorite words, and of
phrases and rimes in written works.
The most frequent forms of associations that occurred with
six participants in my own investigations (p. 136) are shown
in the following table in which the figures give the number
of times each form occurred ; the ? in the table indicates the
inability to classify the association ; the indicates ' no
association.'
Subjects :
S. N. D. H. R. K. Total
Kind of association.
Picture — picture 13 61 17 50 70 0 211
— sound 0 0 10 0 0 1
■' —visual word 0 1 5 0 2 3 11
— motor word 37 0 10 .5 0 18 70
0 4 15 1 4 0 24
" _ ■? 23 6 26 2 2 0 59
Visual word — picture 6 6 9 1 5 0 27
— sound 0 0 0 0 0 0 0
-visual word 5 9 7 9 3 0 33
— motor word 18 33 6 0 5 10 72
2 0 3 0 10 6
— t 7 3 15 0 0 0 25
The table shows that pictures were followed most fre-
quently by memory pictures, less frequently by internal words
described as 'words as though spoken' (motor words, or,
perhaps, motor-auditory words) and still less frequently by
visual words. Visual words were followed most frequently
by motor words, less frequently by visual words (imagined or
remembered), and still less frequently by memory pictures.
In experiments in which the subject was required to associate
160 PERCEPTION OF SPEECH
a word to one that was called out, Krapelin^ found that
substantives produced associations of substantives in 90% of the
cases. MtJNSTEEBEEG^ found that substantives were an-
swered by 68% substantives, 14% adjectives and 18% verbs,
and that adjectives were answered mainly by adjectives, infini-
tives by infinitives. Aschafeenbtjeg ^ found that substan-
tives were followed by 81% substantives, 6% adjectives, and
10% verbs.
According to Bourdon * words form associations through
their meanings rather than through resemblance in sound.
Experiments made by Gtjicciaedi and Ferrari,^ in which
a paper with the five combinations He, eno, ago, ondo, olle was
placed before each subject with the instruction to write as
many rime-words as possible in ten minutes, showed for 54
persons 1347 rimes by spontaneous sound-association, 292 by
running over artificial combinations tdl a word was found, 163
by memory associations of past events, 30 presumably by a
mixture of sound- and writing-associations, 15 by visual asso-
ciations, — the kind of association being stated by each sub-
ject himself.
In the experiments of Thumb and Marbe^ a word was
called out and a stop-watch, marking fifths of a second, was set
going; upon the reply from the subject the watch was
stopped. Names of personal relations (father, brother, etc.)
. were answered 80 % of the time by names of relations (mother,
sister, etc.). Certain pairs of associations were favored in
frequency and quickness : father — mother, mother — father,
son — father, daughter — mother, brother — sister, sister —
1 Keapelin, Ber. iiber d. 56. Versamml. deutscher Naturforscher and
Aerzte, Freib. i/B, 1884 258.
^ MlJNSTEEBEKG, Studien ztir Associationslehre, Beitrage znr exp. Psychol.,
1892 IV 32.
5 Aschaffeneukg, as before, 206.
* BouEDON, Succession des ph^nomenes psychologiques, Eer. philos., 189.3
XXXV 238.
^ GuicciAEDi E Ferkaei, Di alcune associazioni verbale, Rivista Sperimen-
tale di Freniatria, 1897 XXIII No. 3.
^ Thomb und Maeee, Experimentelle Untersuchungen iiber d. psychol.
Gruudlagen d. sprachl. Analogiebildung, Leipzig, 1901.
HABITS OF ASSOCIATION 161
brother, etc., some of these (father — mother) being more
regular than others (mother — father). Adjectives were
answered mainly by adjectives of an opposed meaning : large —
small, old — young, etc. These pairs were favored also in
quickness. Pronouns were answered mainly by pronouns in
certain favored pairs : he — she, this — that, etc. Adverbs of
place and time were answered mainly by adverbs of the same
class : where — there, here — there, when — then, to-morrow
— to-day, etc., but with less tendency to favored pairs. A
numeral was generally answered by the next higher numeral :
one — two, nine — ten, etc. Verbs — interspersed with other
words — were answered by 52% substantives, 42% verbs, 2%
adjectives and 4% scattering. Some favored pairs appeared :
give — take, take — give, etc. "When the subject was limited
to a verb as his response to a finite form of a verb, it was
found that two persons associated most frequently the anal-
ogous number-person-tense form of another verb, while two
others associated other forms of the same verb. In the
latter case the favored associations were the next following
person, the same form of another tense, and the participle or
infinitive.
In some experiments by Oeetel ^ no such associations of
numerals were found.
On the pedagogical side it seems safe to conclude that
for the practical use of a language an effort should be made
to have the word-associations occur quickly; that even for
purposes of thought and general instruction the associations
between things shotild be accompanied by associations between
words; that one aim of instruction should be to form per-
manent habits of association; etc.
Rbfeebnces
For description and instructions for use of Hipp chronoscope : Wunbt,
(Jrundzuge d. physiol. Psychol., 4. Aufl., II 322, Leipzig, 1893. For tests
of its accuracy: Kulpe und Kirsohmann, Ein neuer Apparat zur Con-
irole zeitmessender Instrnmente, Philos. Stud. (Wundt), 1893 VIII 145;
1 Oertel, Note on the association of numerals, Amer. Jour. Philol. XXII 261.
11
162 PERCEPTION OF SPEECH
MiJLLEB DND PiLZECKER, Exper. Beitrdge zur Lehre vom Gedachtniss, Zt.
f. Psychol, u. Physiol, d. Sinn., 1900, Erganzungsb. I 289. For manipu-
lation of EwALD chronoscope : Gilbert, Mental and physical development
of school-children, Stud. Yale Psych. Lab., 189 411 47. For description
and manipulation of D'Arsonval chronoscope : Philippe, Technique du
chronomfetre de D'Arsonval, Paris, 1899. For description and manipula^
tion of the pendulum chronoscope: Scripture, Some new apparatus, Stud.
Yale Psych. Lab., 1895 III 98 ; Elem. course in psychol. measurements,
same, 1896 IV 133; New Psychology, 155, London, 1897; The Pendulum
Chronoscope (in preparation).
For the Hipp and Ewald chronoscopes : Peyer, Favarger & Cie.,
Neuchatel, Switzerland. For apparatus to test the Hipp chronoscope : Zim-
MKRMANN, Leipzig. For the D'Arsonval chronoscope : Verdin, Paris.
For the pendulum chronoscope : Psychological Laboratory op Yale
University (made to order).
CHAPTER XIII
SPECIAL ASSOCIATIONS IN SPEECH
Among the phenomena of speech that depend on the asso-
ciation of ideas no more easily investigated or more important
problems could be selected than those of syntax. Although
the methods readily suggest themselves, experimental work
can hardly be said to have been begun and the subject can
find no treatment on the present occasion. A short time ago
the same statements might have been made in regard to most
phonetic and linguistic phenomena involving associations of
ideas; some beginnings have, however, been recently made
by statistical and experimental methods.
The effect of more than one center of density (p. 137) at
the same moment may be seen in ' lapses ' of speech.^
According to Meeinger and Mayer the most frequent
lapses consist in exchanges between parts of a sentence, a
word or a sound appearing too soon (anticipation) or too late
(postponement) ; in general two parts are interchanged that
have similar or identical functions. Among the examples
are : ' wertlaut ' for ' lautwert ; ' ' malarium plasmodiae ' for
' Plasmodium malariae ; ' ' du leichst dir merk seinen namen '
for ' du merkst dir leicht seinen namen.' Words in antithesis
[or other favored associations, p. 142] are specially liable to
exchange : ' da steht der einsatz nicht ftir den gewinn.' In
most exchanges of words an adjective is exchanged for an
adjective, a substantive for a substantive, etc. [that is,
1 AvENAEius, Kritik d. reinen Erfahrung, II 472, Leipzig, 1890; Meringer
UND Mayer, Versprechen und Verlesen, Stuttgart, 1895; Bawden, A study of
lapses, Psychol. Rev., Mon. Suppl. Ill No. 4.
164 PERCEPTION OF SPEECH
according to the principle of favored associations]. The ex-
changes of syllables are not so common: ' gebrecherver-
hirne ' for ' verbrechergehirne,' ' musikatorisch-deklamatalisch.'
Vowels of nearly the same emphasis are specially subject to ex-
change: 'alabister-bachse' for ' alabaster-buchse,' ' reidflinsch '
for ' rindfleisch.' Further frequent changes are found in those
of initial consonants of syllables of approximately the same
emphasis : ' denile semenz,' for ' senile demenz ' [' Kelen
Heller ' for ' Helen Keller ' is a case of my own] ; and in final
sounds of differently or similarly emphasized syllables : ' stein-
beiss ' for ' steissbein,' ' ich verganz gass ' for ' ich vergass
ganz.' Such exchanges do not seem to occur at all frequently
between the vowels of differently emphasized syllables, as
' hendla ' for ' handle,' or between uiitial and final sounds, as
' tug ' for ' gut.' It is furthermore notable that the result of
such transposition is usually a well known and habitual sound-
sequence.^
The general rule seems to be that exchanges occur only
between sounds equally important in the particular case of
speech. 'The sounds of internal speech are not aU of the
same importance. With a sound that is just being spoken
there fuse traces of all those of equal importance that are in-
tended to be spoken and also traces of all past ones.' ^ This
principle is proposed as the explanation of verbal lapses, of
the action of a speech sound on those distant from it, of the
similar phenomena of assimilation and dissimilation in non-
contiguous sounds, of vowel harmony, of exchange, etc. In
associative effectiveness the various sounds could be arranged
in the descending order : 1. initial sound of the root syllable,
initial sound of the word ; 2. vowel of the root syllable, vowel
of a syllable with secondary emphasis ; 3. initial sound of an
unemphatic syllable ; 4. all other vowels, all other consonants.^
Other associations than those in the particular spoken
phrase are often active ; ' eine papstliche enklitika ' for ' eine
1 Oektel, Lectures on the Study of Language, 23], New York, 1901.
2 Meeinoer dnd Matee, as before, 164.
•' Meeingee tiND Matee, as before, 137.
SPECIAL ASSOCIATIONS IN SPEECH 165
papstliehe encyklika,' by a speaker who had written an essay
on the ' enchticae.'
The scheme of possible transpositions is indicated in an
example (Fig. 67) where + denotes the initial sound of an
emphatic syllable, X the initial sound of an unemphatic syl-
lable, 4 the vowel of an emphatic syllable, — the iinal sound
of an emphatic syllable, 0 the final sound of an unemphatic
syllable. At the time when the / of faul has become the
center of attention, the remainder of the sentence is also
present in mind. The most proniinent sounds st, D and m
become closely associated and may be interchanged. The
sounds of the second rank of importance are au, aa, a and
ar ; these also become associated and may be interchanged.
Like associations may occur in the third rank t, n, and also in
the fourth rank I, te, ne}
Etrvas iat faul im Staate D^nertiarlts.
»•» ) + 6- + 6 ite + 6 ■kp ■* 6 J
+ o- +oxe+oxe+OT-
Fig. 67.
Instead of a transposition a portion of speech may replace,
add itself to or unite with another without losing its own
place. This may occur by anticipation or postponement. Such
association occurs only between portions that are complemen-
tary to each other and not between portions of similar nature,
for example, between a root syllable and an ending but not
between two root syllables. Cases of anticipation occur in
the various forms, ' ungehallt verhallen ' for ' ungehort ver-
hallen,' ' funktion der geschnelligkeit . . . ' for ' funktion der
schnelligkeit des gefiihls,' ' schlecht liberlegt ' for ' schlecht
iiberlegt,' ' wie ich um die ecke gekommen bist, weist du ? '
for ' . . . gekommen bin. . . . ' The various types of postpone-
ment are to be seen in the following examples : ' er wiinscht
zu wunschen ' for ' . . . zu wissen,' ' warenheilkunde ' for
1 Meringek und Mater, as before, 28.
166 PERCEPTION OF SPEECH
' warenkunde ' in a conversation concerning ' heilkunde,' ' stoss
eines erdbobens ' for ' . . . erdbebens,' ' sozialistische zekten '
for ' . . . sekten,' ' wie ein botaniker blumen sa . . . ' for 'wie
ein botaniker blumen sammelt,' ' bessere leute als er sind '
for ' . . . er ist.' Anticipations are more frequent with the
young ; postponements occur mainly with the old and because
of fatigue, rarely in energetic speaking.^ Mbringer and
Mayer are hardly correct in placing here the cases like ' ein
rechter dummer mensch ' for ' ein recht dummer mensch.'
They are due rather to a lax giummatical coordination which
is often found in colloquial speech for the more logical sub-
ordinating structure. A parallel case in English is 'now you
are nice and dirty ' for ' . . . nicely dirty.' ^
Contamination^ is an alteration through simultaneous as-
sociation. In the former case there are two nearly equal
centers of density in the stream of thought which produce a
combined effect. Typical examples were found in : ' hangt
in zusammenhang ' from the familiar phrases ' hangt zusammen '
and ' ist in zusammenhang ; ' ' ich war bis 7 uhr zu haus habe
ich geschrieben ' from ' ich war bis 7 uhr zu haus ' and ' bis 7
uhr habe ich geschrieben ; ' ' ich habe eine empfohlung an sie '
for ' . . . empfehlung . . . ' just after hearing the words ' sie
sind mir empfohlen ; ' ' linsengericht ' for ' linsensystem.' Such
contaminations of two lines of association appear in ' mixed
metaphors.'
The mental processes that show themselves in lapses may
be assumed to be constantly active in speech and to have
contributed to the changes in language. It is possible that
some actual lapses frequently repeated may have been imitated
by the community and so passed into the language (Paul).
It is more probable that the lapses confused the memory
images of certain words so that the speakers became uncer-
tain as to the proper form and finally adopted the new one.
As examples of word forms that have arisen by contaminJi-
1 Meringee und Matek, as before, 52.
2 Storm, Eiiglische Philologie, 2. Aufl., I 690, Leipzig, 1896.
8 Meringer und Mayer, as before, .'JS ; Oertel, as before, 170.
SPECIAL ASSOCIATIONS IN SPEECH 167
tion we may select ^ German ' gewohnt ' from the Middle High
German adjective ' gewon ' and the participle ' gewent ' from
'wenen,' gewohnen; 'zu guter letzt ' from 'zu guter letz
(Middle High German 'letze,' departure) and ' zu letzt;'
'■ Fritzens ' for ' Fritzen,' under influence of the common geni-
tive ending ' s ; ' ' gegessen ' from ' gezzen ' from the feeling for
the usual extra prefix ' ge ' in the participle. These cases indi-
cate, I believe, the close connection between functional associ-
ation (p. 149) and the fusion of past associations. Examples
of phrase forms arising by contamination are : ^ ' das gehort
mein ' from ' das gehort mir ' and ' das ist mein ; ' ' I am friends
with him ' from ' I am friendly with him ' and ' we are friends ; '
' der selbe wie ' from ' der selbe der ' and ' der gleiche wie.'
The phenomenon of assimilatory condensation, as in 'ein
kleinernes ' for ' ein kleines schweinernes,' has been called
'contamination by successive association.' The whole
phrase is in mind while it is being uttered; the riming of
two of the prominent sounds causes an immediate transfer of
the center of density from one syllable to the other; in the
utterance the intervening sounds are slurred or omitted. I
have observed a related phenomenon in the case of a child two
years old who usually said dSteka for dadSteka, mazkot
for mamaizkot, Inbezrum for Inbeblzrum although the longer
forms were often used also. Careful observation led me to
believe that the omitted syllables were not simply dropped
out but were represented in utterance by some very brief
sounds at the ends of the preceding syllables ; thus the glide
from z to r in Inbezrum seemed to differ from the usual
one although I could not hear any very distinct sound or
detect any special b-movement of the lips. From the general
laws of mental life and from some experimental records of
speech sounds we would expect that even with such a con-
densation of thought the vocal organs will perform the usual
movements in a highly abbreviated way and produce what
might be called rudimentary sounds. The question can be
1 Paul, Principien der Sprachgeschichte, 3. Aufl., U5, Halle, 1898.
2 Paul, as before, 149.
168 PERCEPTION OF SPEECH
decided only by experiments. The phenomena seem related
to those of haplology found in the history of words, as in
' nutrix ' for ' nutritrix. '
It is to be noted that among the thousands of lapses in
German noted by Mbringer and Mayer certain forms were
rare : exchanges within a group of consonants (only one case,
' skenien ' for ' xenien ') ; omission of letters except in final
sj'^llables; omission of syllables that cannot be explained
by the principles of association (' erste katorie ' for ' erste
kategorie'). Omissions of letters and syllables occur, how-
ever, constantly in ordinary speech ; such phenomena of syn-
cope in the history of language have been explained as the
results of increased speed (Brxjgmann) but most of them are
rather to be considered as economy effected by condensation.
Such effects of visual and written lapses on a large scale
may have occurred ^ at the time of the great adoption of
French words into English ; thus the use of n for u in foreign
words may have produced two forms of which the incorrect
one finally prevailed. Thus the Old French has ' enhancer.'
In Anglo-Norman the forms ' enhancer ' and ' enhauncer '
are found which -^ ' enhaunce ' ->■ ' enliance,' the Anglo-Nor-
man forms probably (Koppel) arising from erroneous writing
of n for u in the Old French form.
In the history of language there has been at all times a
tendency to make commonly associated words alike in fonn.
Meringer and Mayer would explain such historical assimila-
tions in the same way as assimilations that occur in lapses
(above); it may be suggested that the principle of economy
by similarity may be applied here as well as to the explana-
tion of vowel harmony (p. 121).
A good illustration of the mental action in assimilation by
analogy is to be found in Browning's word ' gadge ' in the lines
' The dead back-weight of the beheading axe !
The glowing trip-hook, thumb-screws and the gadge I '
{A Soul's Tragedy, Act I.)
1 KopPEL, Spelling-Pronunciations, Quellen u. Forschungen zur Sprach- a.
Colturgeschichte, 1901 LXXXIX 1.
SPECIAL ASSOCIATIONS IN SPEECH 169
Upon being told that lie had probably meant to write ' gag '
he replied : ' Gadge is a real name, in Johnson, too, ior a
torturing iron.' The word is not given by Johnson. Brown-
ing probably ^ had in mind the form ' gagge ' and had changed
it to 'gadge' after analogy with 'egge,' 'wegge,' 'brigge'
which have become in Modern English 'edge,' 'wedge,'
' bridge.'
The assimilation of a somewhat different word to the form
of a group with which it is associated is seen in the change of
' October ' to ' octomber ' or ' octember ' to resemble. ' Septem-
ber,' 'november,' etc. in Vulgar Latin, Old French, Modern
Greek and Slavic. A like assimilation occurs in forming new
words as ' electrocute ' in America. The coromon ending is
felt by the popular mind as a suffix (e. g. ' mber,' ' cute ') to
be added to a class of words.
The process of assimilation as a principle of economy of
thought appears in the following remarks by Whitney.^
'When phonetic corruption has disguised too much, or has
swept away, the characteristics of a form, so that it becomes
an exceptional or anomalous case, there is an inclination to
remodel it on a prevailing norm. The greater mass of cases
exerts an assimilative influence upon the smaller. Or, we
may say, it is a case of mental economy : an avoidance of the
effort of memory involved in remembering exceptions and ob-
serving them accurately in practice. The formal distinction
of plural for singular was one which our language was never
minded to give up. Of all the plural signs, the one which
had the most distinctive character was s. The attention of
the language-users became centered upon this as an affix by
which the plural modification of sense was made, and then
proceeded to apply it in words where it had not before been
used ; and the movement, once started, gathered force in its
progress until it swept in nearly all the nouns of the lan-
guage. So with the verb. By the numerical predominance
of forms like loved from love, the addition of a rf got itself more
1 KoppEL, as before, 39.
2 Whitney, Life and Growth of Language, 75, New York, 1875.
170 PERCEPTION OF SPEECH
conspicuously associated with the designation of past time;
and men began to overlook the cases which by right of former
usage ought to be made exceptions.' This reaction against
unusual forms and purposeless differences is seen constantly
in the history of language.^
These phenomena of assimilation, which seem closely re-
lated to those of functional assimilation, are to be explained in
much the same way.
The principle of favored associations appears clearly in
various familiar examples of assimilation collected by Thumb
and Maebe.2 Vulgar Latin has ' grevis ' from ' gravis ' by
association with 'levis.' The influence of the word for one
numeral on that for the following one is seen in Greek ' hvai '
for ' hvolv ' influenced by ' rptcrl,' Gothic ' fidwor ' for ' *hwid-
wor ' by ' fimf,' Lithuanian ' septyni ' for ' *septimi ' by
' asztuni,' Latin ' novem ' for ' * noven ' by ' decem,' Lithuanian
' devyni ' for ' *navyni ' by ' deszimts,' Slavic ' deveti ' for
' noveti ' by ' deseti.' The influence of adverb-associations
is seen in the resemblances of forms in Gothic ' hwar, thar,
her ; ' Old High German ' wftr, thftr, hiar ; ' Anglo-Saxon
' thider ' for ' *th8eder ' by ' hider ; ' Middle High German,
' wannSn, dannftn, hinnfin ; ' Modern Greek rcopi (for rwpa
'now'), airo-^i for a7roi|fe, 'to-day,' etc. The influence of
associated pairs of pronouns is seen in Greek i^/^et? for *^ya6t9
by vfieK, Modern Greek ecrovvov for iaov by ' avrovvov,' etc.
Favored associations in verbs result in assimilative changes as
in Ital. ' rendere ' for ' reddere ' influenced by ' prendere,'
Port. ' bebesto ' from Lat. ' bibitus ' by ' comesto.' Assimila-
tions that correspond to such associations are found in ' ich
frug ' for ' ich fragte ' after ' ich trug.'
Assimilation by analogy may be made to include the
cases of change in pronunciation due to spelling. In re-
cent times, when the use of printed words has become as
' Paul, Principien d. Sprachgeschichte, Ch. X. Isolierung und reaction dage.gen,
3. Aufl., 170, Halle, 1898.
2 TunMB UND Marbe, Exper. Untersuchungen uber d. psychol. Grundlagfen
d. sprachl. Analogiebildung, 49, Leipzig, 1901.
SPECIAL ASSOCIATIONS IN SPEECH 171
frequent and as important as that of spoken words — for
some people even more frequent and important — there is
a tendency to assimilate spoken words to forms suggested by
the usual associations of sounds to spelling. It is to this influ-
ence that we may ascribe some of the variations of American
pronunciation in different parts of the country. A simi-
lar tendency appears in Northern British English. ' [The
North] is much less tolerant of pronunciations which go
against the normal force of the spelling, such as the z in
. . . discern, dishonour, sacrifice, abscission, transition.' ^ In
England the pronunciation of many words has been influenced
by the spelling ; a summary of the cases has been given by
KoPPEL ; 2 an illustrative case is that of h. ' Initial h, which
was preserved through First and Second Modern EngUsh,
began to be dropped at the end of the last century, but has
now been restored in Standard English by the combined in-
fluence of the spelling and of the speakers of Scotch and Irish
English.' ^ The restoration in Southern English has occurred
completely in a series of words* such as 'hereditary, hos-
pital, hostile, humility, habit, hebrew, hermit, homage, hori-
zon, hosanna, host, hostage.' In some words like 'hostler,
herb, humble, humor,' the h is sometimes omitted, although
this practice is decreasing. In ' heir, honest, honor, hour,' the
h is completely dropped. In America the restoration is com-
plete and the initial h of stressed words is never dropped
except in 'heir, honest, honor, hour' and some derivatives.
Even in unstressed syllables the American seldom drops the
initial h although he may weaken it. Such pronunciations as
anotel for 'a hotel,' 3nistorikl for 'a historical' are not used
in ordinary speech ; the spellings ' an hotel ' and ' an histori-
cal ' which are sometimes used in American books to conform
to British spellings, do not represent — as they do in England
— the actual pronunciation; an American would :fead 'an
1 Lloyd, Primer of Northern English, 31, Marburg, 1899.
2 KopPEL, Spetlinff-Pronunciations, Quellen u. Forsch. zur Sprach- u. Cul-
turges, 1901 LXXXIX 1.
" Sweet, New Engl. Grammar, 280, Oxford, 1892.
* KopPEL, as before, 4.
172 PERCEPTION OF SPEECH
hotel ' as an hotel or, rather, aen hotel and would feel it as a
foreign oddity. It is to be noted that the rough American h
is quite a different sound from the faint Southern English h
and that it cannot be dropped without making a much greater
change in the impression on the ear.^
When a sound or a group of sounds is perceptibly like
another, this other may be brought by association more or less
into mind, with possible interruption of the course of thought.
This effect is produced by alliteration and rime in verse for
just that purpose. It is offensive in prose when noticed
because it interrupts the succession of ideas. This is probably
the reason why the English language does not like the im-
mediate repetition of any emphatic word. It is possibly also
the cause of • such harmonic dissimilation of sounds as in
Latin 'Solaris' for ' solalis,' 'sepulcrum' for ' sepulculum,'
' meridies ' for ' medidies ; ' Greek eT€0r]v for edidrjv, Tidrjfjn
for dlOr^ixi ; Spanish ' arbol ' from ' arborem ; ' French ' ros-
signol' from ' lusciniolam.' The loss of consonants or of
syllables may occur also for the same reason. The peculiar-
ities of the cases in which these phenomena occur perhaps
depend on the degree with wliich the similarity attracts
attention.
In many of the lapses the tendency to dissimilation is
clear; it is to be noted that in the conversational speech of
to-day blunders are sometimes found when the conditions are
the same as those that in the history of language produced
dissimilations of sounds or syllables.^ It may be possible,
however, to explain some dissimilations on the principle of
mental economy (p. 122). Just what relation this principle
bears to lapses is not clear. Many of the phenomena of
dissimilation are still obscure.^
The explanations of Meringee and Mayek seem to apply
clearly to one form of metathesis, namely, that where two
1 Shaw, Three Plays for Puritans, 314; quoted in Oertel, Lectures on the
Study of Language, 239, New York, 1901.
2 IMeringer und Mayer, as before, v.
' Oertel, Lectures on tlie Study of Language, 208, 232, New York, 1901.
SPECIAL ASSOCIATIONS IN SPEECH 173
more or less distant sounds change their places, as Latin
' crocodilus ' -* Middle High German ' kokodrille,' Italian
' glorioso ' -* dialectic ' grolioso,' Latin ' periclum ' ->• Span-
ish 'peligro.' Metathesis of neighboring sounds ^ as Anglo-
Saxon ' fix ' = Old High German ' fisc,' ' first ' = ' frist,' seems
to find little analogy in lapses; Meeingbr and Mayer ^
found only one case in their records. I find one (' wist ' for
'wits ') in Bawden's list ■'in which two neighboring sounds
are exchanged. Michels's * explanation of Indger. ' *pot-
men ' -> ' ptomen ' as having arisen by widely spread lapses
seems difficult to accept. Under the influence of a printed word
where the elements are perceived simultaneously it might well
occur that two neighboring elements should exchange places
in the spoken word, as in tlie actual case of ' frith ' for ' firth ' ;
but the change of order in a succession of auditory and motor
elements, as ' fiks ' for ' fisk,' is a speech phenomenon whose
fundamental psychological principles have not yet been
investigated. A suggestion may be found in the fact that an
idea includes a group of elements extending over a region of
time ; considered in this way the cases do not differ funda-
mentally from those of metathesis of distant sounds.
The subject of associative interference in speech should be
attacked by experimental methods. Phenomena that show
themselves strikingly to cursory observation will be found —
according to a well-established psychological principle — to
occur regularly in minor degrees in all persons. By meas-
urements of the increase of time in speaking a phrase when
the subject hears or has just heard an interfering association
the nature of the fundamental process might be investigated.
The experimental conditions are not hard to arrange. In this
way the incipient stages of verbal lapses might be traced by
the differences in time.
Measurements of association-time may perhaps be used to
1 Paul, Principien der Sprachgeschichte, 59.
2 Mekinger und Mateb, as before, vii.
' Bawden, a study of lapses, 111, Paychol. Rev., Mon. Suppl., Ill, No. 4.
* MiCHELS, Metathesis im Indogermanischen, Indogerm. Forsclmngen, 1894 IV
•62.
174 PERCEPTION OF SPEECH
study the effect of the emotional tinge on the use and fate of
words, the stages of expansion and contraction in the meaning
of a word, the disturbance of the meaning of one word by a
change in that of another, and various other special associa^
tions in speech.
Repeeences
For effect of association of ideas in speech : Wundt, Vblkerpsychologie,
I, Leipzig, 1901 ; Paul, Principien d. Sprachgeschichte, 3. Aufl., Halle,
1898; Oeetel, Lectures on the Study of Language, New York, 1901.
For summary of the phenomena of analogy : Wheeler, Analogy and the
scope of its application in language, Cornell Studies in Classical Philology,
1887 I. For problems of syntax based on association of ideas (with
references to older literature) : Wundt, Vblkerpsychologie, I, Leipzig,
1900; Delbruck, Grundfragen d. Sprachforschung, Strassburg, 1901;
Wundt, Sprachges. u. Spraehpsychol., Leipzig, 1901 ; Morris, On Prin-
ciples and Methods in Syntax, New York, 1901.
CHAPTER XIV
rOEMATION OF SPEECH ASSOCIATIONS
Learning a language consists to a large extent in forming
associations among ideas and words. The determination of
the best methods of doing this should be one of the chief
objects of experimental phonetics.
When experiments are made to determine the influence of
one factor in a method, all other factors must be kept constant.
The skill of the experimenter is mainly involved in attempt-
ing this, and the accuracy of the results depends on the degree
of approximation to this condition. The disagreements of the
results of various investigators are generally due to lack of ac-
curacy in this respect ; the results of the less accurate experi-
ments are to be rejected in favor of those of more accurate
ones. With the progress of the science and the development
of its technique the accuracy steadily increases.
When it is desired to determine the facts of association
depending on the forms of speech and not on the meaning, ex-
periments may be made with symbols, such as meaningless
syllables, figures, or signs. Syllables come most closely to
the actual conditions of language ; their preparation requires
considerable care in order to make the conditions alike
in different experiments. The various series of syllables
should be built as equally as possible. The 'normal' series
of MuLLER and Schumann ^ consisted each of 12 syllables
made in the following way. The letters h, d, f, g, h, j, k, I, m,
n,p, r, s, t, w, z, sch representing 17 initial consonant sounds in
German were written on small pieces of cardboard. The
1 MuLLER UND ScHUMAiTN, Exper. Beitrdge zur Untersuchung d. Geddchtnisses,
Zt f. Psychol, u. Physiol, d. Sinn., 1893 VI 19.
176 PERCEPTION OF SPEECH
pieces were mixed in a box. Likewise a similar collection
aa, a, e, i, o, u, a, o, u, au, ei, eii, for the vowels and diph-
thongs and a third collection /, h, I, m, n, p, r, s, t, z, ch, sch
for the final consonants were arranged. To form a syllable
one card was drawn by chance from each box ; the cards
were not replaced in the box. In this way 12 totally different
syllables were obtained ; for example, baup, teir, sohos, mal, etc.
In case two successive syllables had similar contiguous con-
sonants or formed a word, the order of the syllables was
changed; certain objectionable combinations were avoided.
The method may be readily applied to any language.
The symbols, syllables, words or pictures to be learned may
be presented to the eye by holding up or turning over cards
successively at regular intervals,^ or by placing each behind
a shutter which opens for a definite time at definite intervals ^
(p. 136), or by placing them on the surface of a revolving
cylinder which shows each for a given time as it passes before
an opening in a screen.^ The first of these methods is very
convenient. Special care is required in keeping the condi-
tions constant. The cards must be shown evenly at a defi-
nite rate. The use of a revolving cylinder renders it possible
to obtain a constant rate of exposure ; the movement of the
syllable while it is seen is somewhat disturbing. The ideal
method would be a strip of syllables jerked forward and ex-
posed by a mechanism similar to that of a kinetoscope, or a
drum with syllables moved by an intermittent gear connected
witli a shutter over the exposure opening.
As tests for the formation of associations among series of
syllables, the following ones have been employed.
The test of first complete formation consists in going over
the syllables until the set can be repeated at a definite rate
1 Ebbinghaus, Ueber das Gedachtniss, Leipzig, 1885.
2 ScKiPTURE, Ueber d. qual. Verlauf d. Yorstellungen, Phllos. Stud. (Wuudt),
1891 VII 50.
' MiJLLER UND Schumann", Exper.Beitrdge :ur Untersuchung d. Gedachtnisses,
Zt. f. Psychol, u. Physiol, d. Sinn., 1900 VI 81 ; Muller cnd Pilzecker, Exper.
Beitrdge zur Lehre vom Gedachtniss, Zt. f. Psychol, u. Physiol, d. Sinn., 1900
Ergauzungsb. I 99.
FORMATION OF SPEECH ASSOCIATIONS 177
correctly from a given one for the first time without hesita-
tion and with a consciousness of correctness.^ They should
be read each time from beginning to end and not learned in
portions; the learner attempts as soon as possible to an-
nounce the next syllable before it is seen and in case of
success to continue the announcement from memory with-
out looking at the syllables ; upon any hesitation the remain-
der of the series is read through as usual ; when the whole
series is first repeated correctly from memory, it is considered
as having been learned; the number of repetitions (also the
length of time) required is taken as the measure of the work
involved.
Another test consists in recording the number of syllables
that can be reproduced. The subject may be left free during
a given time to recall in any way he can as many syllables as
possible, or a syllable may be shown him and the next one
required, or his method of recollection may be regulated in
some other way.
A third test may be made by measuring the time required
for a person to recall the syllable following one shown him,
that is, the time of association (p. 155). He may be instructed
to give the following syllable as quickly as possible, whether
he is perfectly sure or not, and to use the word ' no ' in case
he feels that he has forgotten the right one.^
The maintenance of constant internal conditions is impor-
tant. Maximum attention may be approximately obtained by
requiring the syllables to be learned as rapidly as possible;
rest is attained by pauses between series ; the same hour of
the day is employed for experiments that are to be compared ;
the manner of living is changed as little as possible; the in-
voluntary tendency to rhythmic emphasis is regulated by
adoption of some one form.^ Learning in trochaic rhythm
has been adopted in several investigations.*
1 Ebeinghaus, as before, 31.
^ MiJLLEE UND PiLZECKEK, as before, 8.
8 Ebbinghaus, as before, 34.
* MuLLEK UND ScHOMANN, as before ; Mulleb und Pilzecker, as before.
12
178 PERCEPTION OF SPEECH
In conducting experiments where the number of successes
is compared with the total number of the trials, it is necessary
to follow the established methods of statistics ; there must be
exact definition and treatment of the countable unit; there
must be careful investigation of the lavfs of probability in-
volved ; and accurate determination of the degrees of trust-
worthiness to be attached to the results.^ Many of the
mathematical methods that have led to discoveries in biology
will prove of value in the problems of memory.
Ebbinghaus's experiments have shown that the number
of repetitions required for learning different series of mean-
ingless syllables increases at first slowly, then rapidly, and
finally less rapidly, as the series are longer ; that each repetition
of a series saves the same amount of labor in relearning it at
a later time ; that the memory effect, as judged by the saving
in relearning a series at a later time, decreases as the logarithm
of the elapsed time ; that when many repetitions are necessary
it is more advantageous to scatter them over a considerable
time than to do them at once ; that members of a series of
syllables become associated not only with the adjacent ones
but also with the others.^
JoST ^ has shown that the repetition of an association adds
the more to its firmness the longer the time it occurs after
the association is first formed; the result is seen in the
greater effectiveness of spreading the repetitions of an asso-
ciation over long intervals as compared with bunching them.
This seems to indicate that in learning a language, if the
number of possible repetitions is limited, as by the number of
lessons, the repetitions should occur at the longest possible
intervals.
The experiments of Mulleb and Pilzeckek* showed
^ ScKiPTnRE, New Psychology, Ch. II, Statistics, London, 1897 ; EBBiNGHAns,
Ueber d. Gedachtniss, Leipzig, 1885.
^ Some of Ebbinghaus's tables and curves are reproduced in Scriptuke,
New Psychology, Ch. XII, London, 1897.
8 JosT, Die Associationsfestiglceit in Hirer Abhangigkeit v. d. Verteilung d.
Wiederholungen, Zt. f. Psychol, u. Physiol, d. Sinn., 1897 XIV 43S.
* MtJLLER UND PiLZKCKER, aS befoTC, 194.
FORMATION OF SPEECH ASSOCIATIONS 179
that the number of correctly reproduced syllables increased
with the number of repetitions ; that the time T^ for
associating the correct syllable decreased slightly ; that the
time Tf for associating the wrong syllable (when that
occurred) increased ; that the time T^ of saying ' no ' on not
being able to remember the syllable also increased ; that in all
cases y^ < 2} < T^; that under equal circumstances the time
of response increased with the time that had elapsed since
the formation of an association.
The associations in a series of syllables are weaker if the
learning of the series is immediately followed by active
attention to some other work. Syllables that have already
been associated with certain others are more difficult to bring
into new associations, but their use in forming the new asso-
ciations strengthens the older ones.^
The number of words that can be remembered in the
form of connected phrases is very many times that of dis-
connected words ; the most important words in phrases are
best remembered; in short phrases the number of replace-
ments of the original word by a synonym is greater than the
number of words forgotten ; in long phrases it is less ; for
phrases of more than 20 words there was, among more than
half the pupils tested, a slight alteration of the sense of the
phrase when reproduced.^
Arrangement of the material in rhythmic groups aids in
fixing associations. For meaningless syllables the trochaic
rhythm seems most favorable for Germans ; ^ individual
preferences are found.*
According to observations made by -Smith,® in which per-
sons were required to learn a series of figures in 20 seconds,
the slower the readings the better the memory, the percent-
age of mistakes being 3.3% for one reading lasting over the
1 MuLi.ER UND Schumann, as before, 177, 318.
2 BiNET ET Henki, M€moin des phrases, Annee psychologique, 1895 I 24.
3 MiTLLEB UND SCHUMANN, as before, IS, 257.
* Smith, Rhi/thmus und Arbeit, Philos. Stud. (Wundt), 1900 XVI 197.
^ Smith, On muscular memory, Amer. Jour. Psychol., 1896 VI 453.
180
PERCEPTION OF SPEECH
given time, 4.2% for two in the same time, 5.5% for three,
and 6.5% for four. Memorization was said to be aided by-
speaking in a loud tone.
The methods of learning quantities of language have been
investigated by Stbfpens. ^ The usual method of learning a
stanza of poetry, consists in repeating the lines in various
groups. The variations in the repetitions followed by differ-
ent persons in learning a strophe of Byron's Bon Juan
were indicated by Steffens in the following manner.
1. To horse ! to horse ! he quits forever quits
2. A scene of peace, though soothing to his soul ;
3. Again he rouses from his moping fits,
4. But seeks not now the harlot and the bowl.
5. Onward he flies, norjix'd as yet the yoal
6. Where he shall rest him on his pilgrimage ;
7. And o'er him many changing scenes must roll
8. Ere toil his thirst for travel can assuage,
9. Or he shall calm his breast, or learn experience sage.
I I
_L
II
I I
This indicates that the person read the first two lines in
succession, then repeated them, then read the first four lines,
then repeated them, then read the third and fourth, then
read the first six, but repeated the words 'and the bowl ' an
extra time, etc. The division of such a strophe into portions
differed with each individual, but the general principle of
learning by portions was always followed. Steffens then
had strophes of verse, series of syllables, etc., learned by the
same persons in two different ways : 1. as the person chose,
that is, by portions ; 2. by repeating the whole material each
time from beginning to end; the results without exception
showed a great economy in the second method. This advan-
tage of the totality method over the sectional method was
clearly and definitively established for material of practically
constant character, the common prejudice in favor of the sec-
1 Steffens, Exper. Beitrdge zur Lehre vom okonomischen Lernen, Zt.
Psychol, u. Physiol, d. Sinn., 1900 XXII 321.
FORMATION OF SPEECH ASSOCIATIONS 181
tional method being shown to be unfounded. In view of the
possible extensive application of this principle to the most
varied subjects of teaching, the experiments should be ex-
tended to such problems as the learning of a vocabulary, the
learning of phrases, etc. The principle seems to indicate, for
example, that, if 25 pages of a foreign language are to be com-
mitted to memory, the work should not be done in sections,
but that the whole should be gone over completely each
time until learned. If the principle holds good, the state-
ment 1 cannot be accepted that ' Economy teaches us to begin
with as small a vocabulary as possible, and to master that
vocabulary thoroughly before proceeding to learn new words. '
Experiments have shown ^ that under equal conditions a
firmer association is made between a picture and a printed
word than between two printed words, and that the learning
of foreign words is aided by placing pictures rather than
translations beside them. This principle has been used as
an aid in various methods of instruction (Combnius) and has
been adopted into most books for early lessons in the native
language. Its application to the teaching of modern foreign
languages, though comparatively recent, has been found to
be highly successful. No systematic attempt appears to have
yet been made to introduce it into the teaching of the ancient
languages.
Experiments by MtJNSTEEBEEG and Bigham ^ seem to indi-
cate that for figures the visual memory is superior to the
auditory, and that both combined give still better results.
In experiments by Ktrkpatrick,* ten short words were
pronounced to 379 school and college pupils, ten other words
on a blackboard were exposed, and finally ten objects were
shown, all at the same rate of one in two seconds. Immedi-
ately after each set was finished the pupils wrote as many as
they could remember ; three days later they again wrote all
1 Sweet, The Practical Study of Languages, 110, New York. 1900.
'^ Scripture, Education as a science, Pedagog. Sem., 1892 II 111.
8 MuNSTERBBRG AND BiGHAM, Memory, Psychol. Kev., 1894 I 34.
^ KiEKPATRiCK, An experimental study of memory, Psychol. Eev., 1894 I 602.
182 PERCEPTION OF SPEECH
that they could remember. The average numbers of items
remembered immediately were 6.9, 6.9, 8.3 respectively, indi-
cating the more efficient action of objects ; the average num-
bers of words correctly remembered after three days were 0.9,
1.9, 6.3, showing a surprisingly great superiority of the mem-
ory for objects. The value of objects for forming firm asso-
ciations probably depends on the much greater impression they
make. In another set of experiments on other pupils the
figures were 7.3, 7.8, 8.0 for immediate memory and 1.8, 0.5,
3.5 after an interval of three days. In order to determine to
what extent the power to recognize completely or partially
might remain when the limit of recalling had been reached,
the words for the thirty original items were, at the end of the
experiment on the third day, mixed with fifteen other words,
the pupil being required to pick out the correct ones. The
average results were 3.8, 2.7, 6.0 placed in the correct lists,
the ability to recognize being greater than the ability to
recall.
Series of meaningless syllables are generally more quickly
learned when presented to the eye than to the ear, but are
apparently not more firmly fixed in memory.^
Experiments on the methods of forming associations among
the elements of words have been made by Lay.^ According
to Lay every didactic theory or hypothesis has its import-
ance, but a feeling of insecurity is attached to it because it
presents deductions from deductions and is far removed from
the certainty of direct scientific experience; principles that
are to be valid for the practical arrangement of a subject for
teaching must have as high a degree of certainty as possible ;
the various methods proposed must be tested by accurate
scientific experiments. Lay's experiments in teaching the
spelling of meaningless words showed that the thoroughness
of learning a certain number of words depended on the method
1 Whitehead, A study of visual and aural memory processes, Psychol. Rev.,
1896 III 258.
' Lay, Fiihrer durch den Rechtschreibnnterricht, 2. Aufl., Wiesbaden, 1899 ;
Didahtisch-psychologisches Experiment, Rechtschreiben und Rechtschreibunterricht,
Zt. f. pad. Psychol, u. Pathol., 1900 II 95.
FORMATION OF SPEECH ASSOCIATIONS 183
employed ; simple hearing (dictation by the teacher) resulted
in an average of 3.04 errors per pupil; hearing with soft
verbal repetition by the pupil, 2.69; hearing with loud repeti-
tion, 2.25; simple seeing, 1.22; seeing with soft verbal
repetition, 1.02; seeing with loud repetition, 0.95; loud
spelling by the pupil, 1.02; copying with the hand, 0.54.
The most efficient method was thus that of copying; the
least efficient that of dictation. It may be suggested that
copying involves unusual concentration of attention to the
word shown and to the details of the word executed. The
words in these experiments were repeated the same number
of times ; the dictation method thus required less time than
the copying method. It was necessary, then, to determine
which method gives the best results for the same time
required in learning. Lay showed that, for the same time
spent in learning, the copying method was the best, the visual
reading with vocal repetition less good, the spelling and
dictation methods still less efficient. He showed by experi-
ment that the same order of efficiency prevailed in forming
the permanent spelling associations. His results also showed
that the vocal organs have far more influence in learning to
spell than did the ear (that speaking was better than hearing) ;
that the visual image was two to three times as effective as the
auditory image ; that copying was superior to seeing alone ;
that an understanding of the meaning of a word [greater
attention aroused by interest] was an enormous aid.
Schiller ^ reports two sets of experiments on learning the
spellings of words.
The first set consisted in learning to spell German words
by various combinations of the senses and the volitions. The
results showed the following relations in ascending order
among the errors in spelling made by the pupils in one
class : 1. copying the written word while softly pronouncing
it; 2. copying while loudly pronouncing it; 3. looking at it
and making movements of the hand in the air as if wi'iting
^ Schiller, Stiidien und Versuche iiber die ErUrnung d. Orthographie, Samml.
V. Abh. ans d. Gebiete d. pad. Psychol, u. Physiol., 1898 II 4. Heft.
184 PERCEPTION OF SPEECH
it; 4. spelling it aloud; 5. looking at it while pronouncing it
loudly; 6. looking at it while pronouncing it softly; 7. look-
ing at it with mouth closed ; 8. hearing the word and making
movements in air as if writing; 9. hearing it and pronounc-
ing it loudly; 10. hearing it and pronouncing it softly; 11.
hearing it only. The most accurate method was thus that of
copying a written (or printed) word, the least accurate was
that of hearing it. In general loud pronunciation at the same
time helped.
The second series of experiments was on the learning of
Latin words ; the results were similar except in the fact that
simultaneous loud pronunciation was disadvantageous in all
cases.
In investigations by Kemsies ^ ten dissyllabic Latin words
with their dissj-llabic German translations were presented live
times in succession to groups of school-children, who were
then required at once to write down all they had learned.
In one form of experiment the words were read off, in another
they were shown in print and in a third they were both read
and shown. The experiments in which the words were heard
showed great advantages over the others, both in the amount
learned and in its exactness. The visual method was least
successful. The combined learning by hearing and seeing
showed on an average no advantage over that by hearing
only ; the gain by using two senses seemed often overbalanced
by the distraction of attention thereby. This latter difficulty
might — I suggest — be due to the particular method em-
ployed. When the words were learned first by one method
and then by another, the result was not so good as Avhen
the same method was employed throughout. In learning the
spelling of words Kemsies found that visual learning was
not so good as the auditory or auditory-visual learning.
In later experiments Kemsies ^ used meaningless dissyl-
1 Kemsies, Geddchtnissuntersuchungen an SchiUern, Zt. f. pad. Psychol, u.
Pathol.. 1900 II 21, 84.
2 Kemsies, Gedachtnissuntersuchungen an SchiUern, HI., Zt. f. pad. Psychol, u.
Pathol,, 1901 III 171.
FORMATION OF SPEECH ASSOCIATIONS
185
labic words (to represent foreign words) with a dissyllabic
German word (native word) for each to represent its
translation. The following is a specimen set:
achtbar
Kutscher
lieblich
blasen
Idmsi
sipaf
eiibor
emok
tiigan
Flasche
T%ul
kogri
fedok
r^fus
geikul
neulich
scheinen
Nachte
Sclimiede
sonnig
These sets were spoken by the experimenter, or were shown
in large letters, or were both spoken and shown. For the
auditory presentation the pupils sat with closed eyes in a quiet
room; for the visual presentation the words appeared as
transparencies to the pupils in a dark room ; for the combined
presentation the experimenter spoke the words in connection
with the visual presentation. In one form of experiment
the words were presented at the rate of half a second for each
syllable, making 1' for each word, 20= for a set and 100
for five repetitions of a set. The test of learning con-
sisted in writing the words as soon as possible after the end
of the presentation. In another form of experiment the
words were repeated at the same rate until the entire set was
sufficiently learned so that the words were all recognizable
although not necessarily completely reproducible. The final
results by these methods have not yet been published.
The firmness of an association depends on the vividness of
the impressions. Various methods of producing vivid impres-
sions are used in forming language associations; they are
based on the pedagogical methods used to fix attention. ^
Motor (spoken) words are more firmly associated with each
other than with auditory words. Only rarely — as with the
deaf — are they directly associated with visual words. In
learning to speak a foreign language the words should be
constantly spoken in connection in order to establish direct
associations among the words in their motor forms. The
^ Summary of these in the chapters on attention and memory in Scripture,
Thinking, Feeling, Doing, 2. ed., New York (in press).
186 PERCEPTION OF SPEECH
attempt to form associations mainly with auditory words is
wasteful. With visual words it is quite impracticable to
form associations readily available in conversation. All these
sets of associations should be united so that the person
'thinks' a foreign language just as he does his own. The
formation of associations between the native and the foreign
lanofuaere is liable to hinder the associations within the
foreign language. The fundamental method to be followed
in learning a language may be said to be that of forming, as
firmly as possible, such associations among the language
elements as actually occur in the language itself. 'The
whole process of learning a language is one of forming
associations. ' ^
In closing this chapter I feel compelled to emphasize the
obligations of the science of experimental phonetics in respect
to the methods of teaching languages, native and foreign,
modern and classical. The present diversity of methods and
conflict of opinions can have no possible justification except
the lack of scientific data. The human mind acts according
to just as definite laws as the expansion of steam and the
transmission of motion. That more work can be gotten
out of a pound of coal by the proper scientific and technical
knowledge of applying it under the proper circumstances
is no more true than it is that a thorough understand-
ing of the mental processes involved in learning a language
will render the economy and progress far greater than at
present. These processes are still scientifically uninvesti-
gated, and will probably remain so except in so far as experi-
mental phonetics carries on the work. The results reported
in this chapter are no more than indications of what is to be
done ; each practical problem — the value of pictures in ele-
mentary instruction, the valpe of translations from the native
to the foreign language, etc., — must be solved by measure-
ments of such accuracy as to be conclusive.
1 Sweet, The Practical Study of Languages, 103, New York, 1900.
FORMATION OF SPEECH ASSOCIATIONS 187
Refeeences
For methods of instruction in language : Breymann, Die neuspraoh-
liche Reformliteratur von 1876-1893 (digest of the literature to 1893);
Berlitz, Methode Berlitz fiir den Utiterricht in den neueren Sprachen,
Berlin, 1890; Bilder zu den Lektionen, etc.; Bever, Der neue Sprach-
unterricht, Cothen, 1893 ; Bekbner, The INlethod of Teaching Modern
Languages in Germany, London, 1898 '(account of methods used in
various schools); Breal, Enseignement des langues vivantes, Paris,
1893; GouiN, L'art d'enseigner et d'etudier les langues, Paris, 1880;
GouiN, The Art of Teaching and Studying Languages (trans, by Swan
AND Btxis), London, 1892 ; IIengesbaoh, Die neusprachliche Reform
im Lichte der preussischen Dlrektoren-Versammlung, Neuere Sprachen, 1897
IV 346; Jagee, Au.<: d. Praxis d. franz. Unterriclits, Neuere Sprachen,
1894 I 65, 133; Jkcinac, Der deutsche u.der russisc/ie Berlitz in Russ-
land, Neuere Sprachen, 1897 IV 109 ; Jespeesen-Lundell-Western,
Quousque Tandem (series of publications) ; Junker, Lehroersuch im
Engliscken, Neuere Sprachen, 1894 I 105 (use of phonetics and phonetic
text for first instruction, use of pictures); Knorr, Ein Werj, der
icirklich zum Ziele fiihrt, Neuere Sprachen, 1898 V 483 ; Landenbach-
Passy-Delobeil, Methode directe, Paris, 1899 ; Meyer, Ueber franzosi-
schen Unterricht, Neuere Sprachen, 1894 I 5, 79, 143, 208, 258, 319, 400,
456 (with previous literature) ; Passy-Rambeau, Chrestomathie fran-
(jaise, 2nie gd., Paris and New York, 1901 (Introduction); Rambeau, On
the value nf phonetics in teaching modern languages, Neuere Sprachen, 1895
II 1; RoDEN, Inwiefern muss d. Sprachunterricht umkehren? Marburg,
1890; Sweet, The Practical Study of Languages, New York, 1900;
Tbaugott, Darstellung und Kritik d. Methode Gouin, Diss., Jena, 1898; ,
Kritik d. Methode Gouin, Neuere Sprachen, 1898 VI 345; Tupschewsky,
Die Verwerthung d. Phonetik f. d. grammatikalischen Unterricht auf d.
Oberstufe, Neuere Sprachen, 1895 II 501 ; Vietor, Der Sprachunterricht
muss umkehren, 2. Aufl., Leipzig, 1886; Viktor, A new method of language
teaching, Educat. Rev., 1893 VI 351 ; Walter, Englisch nach dem
Reformplan, Frankfurt, 1899 ; Thirteen Authors, Methods of Teaching
Modern Languages, Boston, 1893. For language lessons in connection
with pictures : Flemming, Hilfsmittel f. d. fremdsprachlichen Anschauungs-
unterricht, Neuere Sprachen, 1894 I 510, 558 (contains account of various
series of pictures with titles of books to be used with them) ; Hartmann,
Die Auschauungim neusprachlichen Unterricht, Wien, 1895 (literature of
the subject to 1895).
PART III
PKODUCTIOlSr OF SPEECH
CHAPTER XV
VOLUNTAEY ACTION AND THE GRAPHIC SIBTHOD
The production of vocal sounds results from the action of
muscles. A muscle is a contractile body of slight but very-
complete elasticity; it is stretched to a great extent by a
small pull but returns to its original length when released.
The voluntary muscles are composed of fine fibers that extend
the whole length of the muscle. The contraction of a muscle
consists of the contraction of its fibers ; this produces a de-
crease in length and an increase in thickness.
The laws of muscular contraction are best illustrated by
experiments with the gastrocnemius muscle removed from the
leg of a frog. To prepare the muscle, the top of a frog's head
is cut off by. inserting one blade of the scissors across the
mouth, and placing the other behind the skull ; the skin is
stripped from the leg, the heel-tendon is cut, the muscle of the
calf of the leg is separated from the rest, and cut from the
bone. The muscle is kept moist by a brush dipped in a
solution of -^^ of 1 % of salt in water. The muscle is sup-
ported by a hook ; another hook passing through the heel-
tendon is attached to a simple recording lever writing on a
moving surface. The scheme of the arrangement for record-
ing-a curve of contraction is shown in Fig. 68. The muscle
m, stretched by a weight g, is attached to the lever h whose
point r writes on a smoked plate o. This plate carries a pro-
jection d which, as it is moved, strikes the key k and breaks
VOLUNTARY ACTION AND THE GRAPHIC METHOD 189
the circuit containing the battery c and the primary coil p of
an inductorium sp. The breaking of the primary circuit epic
produces a momentary im-
pulse in the secondary coil s
which stimulates the muscle
m through the wires carried to
the hooks. Tlie plate is cov-
ered with soot in the usual
way (p. 7). It is first moved
slowly and stopped when it
just breaks the circuit; the
muscle contracts and draws
the vertical line seen at r in Fig.
69 ; this indicates the point on the plate at which the shock
occurs. The plate is now returned to its position and moved
rapidly along. The muscle draws its curve of contraction k on
the plate. The record shows that the contraction begins at
some time after the shock, rises to a maximum and then falls.
The ' myograph ' plate can be replaced by a recording surface
of any kind (Figs. 6, 7, 8, 71) ; a mechanism for breaking the
primary circuit is readily attached to the axle (p. 93).
Jig. 68.
i\a. 69.
By placing a common telegraph key or a contact wheel
(p. 91) or an electric fork (p. 15) in the primary circuit of
the inductorium (Fig. 68), a series of shocks may be sent to
the muscle. Each shock produces an effect; if the shocks
follow so rapidly that the relaxation is not complete, the
effects are added; when the shocks are sufficiently frequent,
the muscle is strongly contracted without visible relaxation,
the condition being called ' tetanus.' The tetanic contraction,
however, consists of contractile movements whose frequency
390 PRODUCTION OF SPEECH
is the same as that of the shocks ; when the muscles of the
cheek are tetaiiized by the indue torium, a person with
closed ears can hear a tone of the same frequency as that
of the interruption of the primary circuit. Shocks so weak
that no visible effect is produced when they are used singly
may produce contraction if they are repeated with sufficient
frequency ; thus an irritation leaves an increased irritability.
To demonstrate the action of the nerve on a muscle the
frog's gastrocnemius may be used with the sciatic nerve at-
tached. The preparation is made as before (p. 188), but the
sciatic nerve is carefully separated from the muscles along the
thigh and cut as near the spinal column as possible ; moreover,
the muscle is left attached to the bone at the knee and the
bones are cut just below the knee and half-way up the thigh.
The muscle suspended from the thigh-bone is clamped in a
standard and connected to the recording lever as before. The
vvdres from the secondary coil are brought close together in a
small handle and the nerve is laid across the ends by means of
a small brush dipped in the salt solution. A shock to the nerve
is followed by a contraction of the muscle. More time elapses
between the moment of the shock and that of contraction than
when the muscle is stimulated directly. This time increases
with the length of the portion of the nerve between the muscle
and the electrode. The irritation is conducted along the nerve
at the rate of about 27™ a second ; for the motor nerve of
the human arm the rate is from 34™ to 43™. Repeated shocks
are followed by repeated contractions, and by tetanus.
The voluntary contraction of a muscle is tetanic. When the
cheek muscles are contracted voluntarily, they can be heard to
give a tone that is always of the- same low pitch. The volun-
tary stimulation of a muscle requires the uninjured continuity
of the nerve from the muscle to the central nerve system.
The irritation from the central system proceeds along each
nerve iiber separately; there is never any transmission from one
fiber to another. The character of the irritation that proceeds
along the nerve is unknown ; it stimulates the end-plate at
the end- of the nerve and this plate irritates the muscle.
VOLUNTARY ACTION AND THE GRAPHIC METHOD 191
The curve of tetanic contraction, that is, its variation of
degree at each moment of time, depends on the strength of
the motor nerve-impulse at each moment. Thus, the lip con-
traction and closure may be sudden or gradual, weak or strong,
short or long, according to the degrees of irritation received
at each moment by the muscles. The variations among
speech sounds of the same type often depend not so much
on variations in the positions of the oigans as on variations in
the course of the movement through these positions.
The motor nerves come from groups of cells in the brain
and spinal cord. Sensory nerves from the tendons, joints,
skin, mucous membrane and the muscle substance carry to the
central nervous system irritations depending on the degree of
contraction of the muscle. Thus, the position of the tongue is
indicated at each moment by irritations from its surface and its
muscles. When a movement is repeated so often that definite
associations are established between the motor irritations of
the various muscles at each moment and the sensory irrita-
tions present at that moment, the sensory irritations serve
to regulate the motor ones and to govern the movement.
This regulation takes place in the reflex centers; when it
is once established, an impulse to the center is followed by
the complete movement properly coordinated. These sen-
sory impulses are of different degrees of fineness. Upon
them depends to a large degree the accuracy of intended
movements; their fineness can be increased by the proper
practice. In cases of great dullness special methods and
apparatus must be employed.^ The motor cells receive irri-
tations 1. from sensory fibers of the same or of another
level in the cord (direct reflex) ; 2. from intermediate nerve
cells that are irritated by sensory fibers (indirect reflex) ; 3.
from cerebral fibers, especially from the cortex of the brain
(voluntary movement).
The general scheme of reflex activity may be stated as
1 RonssELOT, Applications pratiques de la phone'tique expgrimentale, La Parole,
1899 I 401 ; Zund-Burguet, Applications pratiques de laphonitique experimentale,
La Parole, 1899 I 11, 138, 281 ; see also below, Ch. XXVIL
192 PRODUCTION OF SPEECH
follows. An immediate center, apparently without connection
with higher sense nerves, controls the simplest reflexes in
which mainly the organs of the stimulated portion are in-
volved. The sensory nerves, however, send communications
to many centers of other levels and to the highest centers.
The higher the center the greater its expanse of irritation
and control ; it controls not only the muscles of its own
level but also those of lower levels and thus brings, about
complicated activities. The highest centers act. only by stim-
ulating lower centers ; they are, however, in direct connec-
tion with the higher sense-centers, so that single irritations
from these may be followed by highly complicated activities.
The relations and connections of various portions of the
brain are shown in the schematic Fig. 70 after Auzoux's
model.
The spinal cord S at its upper end becomes the lulh B. In
the dorsal portion (to the left in Fig. 70) of the bulb, there
are groups of cells — ' centers ' — • that control various com-
plex muscular acti^dties. The bulb contains the centers of
coordination and reflex action for chewing, swallowing, action
of the vocal cords, coughing, etc. It also contains the auto-
matic center for breathing. The ventral portion of the bulb
is occupied by nerve fibers.
Just above the bulb lies the pons (Fig. 70, P) and behind
it the cerebellum (^OV). Overlapping the whole is the cere-
hruni ( CV), which consists of two hemispheres, a section through
one of them being shown in the figure. The various portions
known as the frontal, parietal, occipital and temporal lobes are
indicated by the letters FL, PL, OL, TL (see also Fig. 57).
Among other functions the pons has that of transmitting
the speech impulses from the voluntary centers to the lower
ones ; some diseases of the pons alter the length of syllables,
producing what is known as ' scanning speech.' The cere-
bellum has no known speech function. The higher portions
of the brain (^stalk, thalamus, corpus striatum, inner capsule,
etc.) are all to some degree involved in the muscular move-
ments of speech.
VOLUNTARY ACTION AND THE GRAPHIC METHOD 193
The various portions of the surface of the cerebrum — the
cortex — are connected with the lower portions of the nervous
system by ' longitudinal fibers ' (in Fig. 70 marked by — * as at
L) ; with other portions of the same hemisphere by ' associa-
FiG. 70.
tion fibers ' (as at A) and with portions of the other hemisphere
by ' transverse fibers ' (as at TR).
The cortex consists of immense numbers of nerve cells and
fibers that seem capable of entering into endless combinations.
Mental life is affected more directly and extensively by re-
moval of portions of the cortex or injury of it than by injury
of any other region of the body. The central portions of
13
194 PRODUCTION OF SPEECH
the cortex on each side are closely connected with the volun-
tary control of the muscles (see Fig. 57, p. 83). Injury of
definite portions' of these regions is followed by very de-
finite disturbance in the power to voluntarily perform certain
movements. This does not mean that the volitions are
located in these portions of the brain, but that the two are
closely connected in function.
A volition may be followed by a muscular act; thus the
decision to press the tongue against the teeth may be followed
by the actual performance of the act. The series of irritations
may be traced back from the muscles through the end-plates,
nerves and nerve centers in the central gray masses of the
cerebrum to cells in the cortex. The connection between the
vohtion, a mental fact, and the activity of the cortical cells, a
physical fact, has found no satisfactory explanation.
Some of the peripheral nerves arise directly from the brain.
Among these is the hypoglossus, which controls all the tongue
muscles and most of those connected with the hyoid bone.
The vagus contains motor fibers to the larynx and the bron-
chial muscles ; sensory fibers from the larynx, trachea, bronchi
and lungs ; motor and inhibitory fibers to the velum ; sensory
fibers from the pharynx. The glossopharyngeus contains motor
fibers to the velum (elevator of the velum), uvula, pharynx
(middle constrictor of the pharynx) and the stylopharyn-
geal muscle; sensory fibers from the tongue and velum. The
aousticus is the nerve of hearing. The facialis contains motor
fibers to the face muscles, the stylohyoid and the stapedius
muscles. The trigeminus contains motor fibers to the jaw
muscles, the tensor tympani and the tensor of the velum.
Voluntary impulses do not act on separate muscles, but
on the same complexes of cells as are involved in reflex
actions.
Owing to the lack of data concerning the action of the
nervous system and to the lack of methods for its investi-
gation in regard to the production of speech sounds, the
work of experimental phonetics lies at present almost entirely
in studying the muscular movements and the volitions that
VOLUNTARY ACTION AND THE GRAPHIC METHOD 195
give rise to them with little consideration of the nervous
mechanism involved.
The methods used for registering muscular movement in-
clude, among others, that of air transmission by Makby
tambours.^
Fio. 71.
The tambour is a metallic box with a rubber top and a side
tube. There are two tambours, the receiver and the recorder.
Any desired movement may be imparted to the straight
lever of the receiver (2, Fig. 71). This lever communicates
the movement to the air inside by varying the pressure on the
rubber top. The movement of the air is transmitted along
1 Makey, La methode grapliique daus les sciences experimentales. Paris, 1878;
2me tirage arec supplement, Paris, 1885.
196 PRODUCTION OF SPEECH
the rubber tube (5) to the recorder (5). The rubber top of
the recorder moves in response to the movements of the air,
and the light lever (P) resting on it repeats the movement.
The valve (^) is used to equalize the air pressure.
The details of the recording tambour of the most common
form are shown in Fig. 72. The tube N, to which the
rubber connecting tube (3 in
Fig. 71) is attached, opens into
the metal box M. The alu-
minum plate K, attached to the
rubber cover L, communicates
^'''- ^2- its movement by means of the
link J and the clamp H to the lever I. ■ A very light record-
ing arm T (only a portion is shown) is placed on I. As the
block F carrying the fulcrum G is movable by a hinge on
the fixed block E, the lever can be inclined and the record-
ing point raised or lowered by tipping F. The degree of
amplification can be adjusted by sliding H along i, the link
J being kept perpendicular to I by moving M by means of the
screw A. The screw 0 fastens the tambour to a rod through
R. A later form of the tambour is shown in Fig. 73; it is
smaller and more sensitive. The construction is in general
on the same pattern as the tambour in Fig. 72; the screw A,
however, moves the whole tam-
bour forward — a factor of great
importance in adjusting two tam-
bours so that their recording points
are in proper alignment ; the screw
B turns the tambour side wise and
adjusts the degree of pressure of the point against the drum.
The perpendicularity of J is maintained by fastening the
metal box at the proper point by means of a screw shown
near M; the lever D moves the block F on its hinge.
The receiving tambour resembles the recording one, but is
often modified to suit special requirements.
The care and repair of tambours require some technical
knowledge. The rubber membrane (heavier for the receiver,
VOLUNTARY ACTION AND THE GRAPHIC METHOD 197
lighter for the recorder) must be evenly stretched. The box
is made air-tight by inclosing in it a few bits of paraffine and
warming the edge so that this runs along the contact between
the rubber and the edge ; the truthfulness of the record de-
pends upon the prevention of any escape of air. The disc
that holds the connecting link from the lever to the rubber
is fastened to the latter by melted wax or paraffine.
The tambour ordinates are cui'ved and not straight. The
degree of curvature depends on the radius of the movement
of the recording point. Before a set of experiments the
tambour should be made to record a large excursion while
the drum is at rest. The curved ordinates may be drawn on
the drum before the record is varnished by turning it to each
desired point and recording an
excursion while it is at rest.
When this is not done, all com-
parisons must be made by a
curved line obtained from the
original excursion of the tam-
bour on the motionless driini.
In such cases all comparisons
by straight lines are false ; valu-
able and laborious researches Fig. 74.
have been ruined by the neglect
of this seK-evident fact. When comparisons are to be made
by perpendicular lines, a large excursion oabx (Fig. 74) of
the recording point is made while the drum is at rest, repre-
senting the perpendicular oo. The line oabx would be ' recti-
fied ' by moving its points to the left by the distances between
each one and the line oo. A tambour tracing is rectified like-
wise ; the tracing dVodVci' would thus be moved to the left as
indicated by the dotted line. The dotted line need not be
actually drawn. The most convenient method is to draw a
set of lines parallel to the horizontal axis as shown in Fig. 74
and subtract from all horizontal measurements the distances
from the line oo to a, 5, x, etc.
When the tambour is used to record on a drum, the lever
Fig. 75.
198 PRODUCTION OF SPEECH
h (Fig. 75) in its horizontal position should be parallel to the
plane tt of a tangent to the surface of the cylinder cy at the
point p touched by the flex-
ible recording point s. As
the lever rises or falls, the
point s would leave the sur-
face of the cylinder if it
kept in the plane of the tan-
gent, owing to its circular
movement which brings it
to one side of the vertical
line on the cylinder; considerable spring and flexibility are
thus required in the recording point in order to have it
remain on the cylinder.
Tambour records may be made on any suitable recording
surface. Drums for use with any motor power have already
been described (p. 7). The smoked drum may be replaced
by a glass wheel when the records are to be used in demon-
stration with a projecting lantern. The apparatus with a
tambour is shown above in Fig. 8. For travelling purposes
a clockwork drum may be made of aluminum.^ The special
clockwork drum shown in Fig. 71 is of a widely used tj^e ; it
is frequently called a ' kymograph.' Its speed is so carefully
regulated that, when its rate of revolution is once determined,
it can be depended upon to maintain that rate with a high
degree of accuracy, provided the spring is kept wound up to
about the same tension and the whole apparatus is in perfect
order.
The following instructions may be found of use. To re-
move the cylindrical drum grasp it at 0 and lift the arm F.
Place the drum on a separate horizontal support and smoke it
as usual (p. 7).
Lift the drum from the support, grasping it around the
ring 0 at the end. Raise the spring G of the kymograph by
the arm F till it catches. Let the end of the drum-axle drop
into the socket P. Bring the groove of the ring 0 up till it
1 JossELYN, Mtude sur la phonetique italienne, 3, These, Paris, 1900.
VOLUNTARY ACTION AND THE GRAPHIC METHOD 199
catches on the wheel at the end of the arm iV. Bring the
top of the axle just below the socket held by (?, and let F
snap. The drum is now in position ; it should be turned till
the projecting point at the bottom of the axle catches in a
notch of the spring P. If the kymograph is not firm upon
the table, adjust the leg M.
Wind up the clock-spring by the handle. Move the brake
Fl in order to release the governor D. When the screw B
is tight the drum will turn with the clockwork ; when it is
loose, the drum is disconnected. The connection of the
clockwork with the drum axle is established by the large fric-
tion disc which presses against the small friction roll X.
When handling the drum, always disconnect it by turning
-B ; this keeps the friction disc from being ground by acci-
dental movements of the roll X.
The speed of the friction disc is changed by different com-
binations of the gears in the clockwork. The case of the
clockwork can be readily opened by unscrewing the movable
side T. There are two gears that move sidewise on their
axles, a lower one and an upper one. When the upper
wheel is in the middle position the screw 0 should be turned
so as to bring the little wheel at the end of the arm into posi-
tion between the largest and smallest cog-wheels. There are
three sets of springs for the governor; that set should be
chosen which allows the wings of the governor to take a
medium position when in motion. The following table
gives the speeds approximately obtainable by the different
combinations.
SPEED
NAME.
POSITION OF
LOWBK WHEEL.
POSITION OF UPPER
WHEEL.
FRICTION ROLL
AT LOWEST
POINT.
FRICTION ROLL
AT HIGHEST
POINT.
I
II
III
IV
V
VI
Left.
Left.
Left.
Right.
Right.
Right.
Right (weak spring).
Left (medium spring).
Middle (strong spring).
Right (weak spring).
Left (medium spring).
Middle (strong spring).
6°
2»
12»
40"
16»
12»
45-
15"
5'
2- .
200 PRODUCTION OF SPEECH
The intermediate speeds between the figures in the table
are obtained by moving the roll X by means of the screw R.
An index connected with X moves over a scale so that a
speed once found can be reproduced by direct adjustment of
the index to the same point; to avoid back-lash the adjust-
ment should be made in the direction from zero upward.
For respiration records adjust the kymograph to about 20'
for one revolution.
To determine the speed of the drum a time-line may be
drawn on it by a vibrating fork (p. 15), a time-marker (p. 91)
connected to some regular interrupter such as a clock, or
directly by the graphic chronometer. ^
The following experiments on finger movements are de-
signed to illustrate some of the fundamental phenomena of
voluntary action. They are described as being performed by
the finger; they can be readily modified for application to the
lip or the jaw by placing the receiving tambour in the proper
position. The phenomena illustrated find their application to
speech in many ways.
To make the finger perform some work analogous to that
which occurs in stretching the vocal cords, a loop of tape over
the middle finger is attached to a string ending in a rubber
band. The arm may be laid on the table. The movement
is registered by attaching to the string the lever of the receiv-
ing tambour and having the recording tambour write on the
smoked drum.
The strength of the muscular contraction depends, within
limits of variation, on the strength of the effort intended.
The stronger the pull intended the greater is the actual pull.
A series of efforts, however, with intensities in the relations
1:2:3:4 results in muscular acts that generally bear quite
different and often changing relations^ (Fig. 76).
Separate pulls will vary although intended to be alike.
If the finger is first contracted with any desired force and
after relaxation is again contracted to what is supposed to be
1 Jacquet, Studien iiber graphische Zeitregistrirunq , Zt. f. Biol., 1891 XXVIII 1.
2 SCEIPTDRE, New Psychology, 216-218, London, 1897.
VOLUNTARY ACTION AND THE GRAPHIC METHOD 201
the same degree of force, a comparison of the two records will
indicate the amount of the error of execution.
To obtain the error of execution we first find
^^x^-\-x^ + ... + x„
n
where x^, x^, . . ., a;„ are the measurements of the separate
pulls, n the number of pulls and a the average pull ; and then,
calculating v^ = x^ — a, v^ = x^ — a,.. ., v„ = x„ — a, we have
P
~ 3 V n-1
as the probable error of execution. The size of this probable
error is used as the measure of the uncertainty of the mus-
FiG. 76.
cular movement. This error is never zero; all movements
vary around an average. The notion that the positions for
a speech sound are something fixed for an individual is quite
erroneous ; there is an average, for a certain intended position,
but every particular position varies from the average. Some
individuals have large ranges of variation, some have small
ones. If no variations are detected, it is because the methods
of measurement are not accurate enough.
202 PRODUCTION OF SPEECH
The error of execution is due partly to an error of percep-
tion and partly to an error of movement ; ^ and partly also to
an error in the intention.^ Such an analysis, though prob-
ably a lughly valuable undertaking, has not yet been made for
any speech movements. The regulation of muscular move-
ment occurs through adjustment of the impjilses sent to the
muscles in response to the sensations arising from them. The
error of perception can be indicated by having the person state
immediately after each contraction whether the pull appeared
to him to be greater than, equal to, or less thaii the intended
one.
The strength of a muscular contraction fluctuates although
it is intended to be constant ; the stronger or longer the effort,
the greater the fluctuation. Let the finger be kept contracted
to any desired degree ; the record on the drum will show con-
tinual fluctuations (Fig. 77). The fluctuations will be found
Fig. 77.
to be greater for a strong contraction ; they will become very
great when the effort is maintained for a long time.
The irregularities in the pitch of the voice in striking or
maintaining a note are due to defects in regulation of the
tension of the vocal muscles. In taking the pitch of a note
from an instrument the note is first perceived through the
ear, then the vocal muscles are adjusted to produce it, there-
upon the produced note is heard as coinciding with or differ-
ing from the desired one, and the vocal adjustment is corrected.
The maintenance of any given pitch depends also on the regu-
lation of the muscular tension by the sensations from the
1 FuLLERTON AND Cattell, On the perception of small differences, 65,
Philadelphia, 1892 (account in Sokipture, New Psychology, 224, London,
1897).
2 WooDWORTH, Accuracij of voluntary movement, Psychol. Rev., Monogr.
Suppl. m, No. 2, 71.
VOLUNTARY ACTION AND THE GRAPHIC METHOD 203
muscles. The error of execution in singing a note thus de-
pends directly on the size of the just perceptible difference
and on the size of the average error of movement. A large
range of unperceived difference in the organ of hearing per-
mits the voice to vary greatly but unconsciously. A large
error of movement causes the voice to vary greatly. With a
small range of unperceived auditory difference but a large
error of movement the variations in pitch are detected by the
singer but cannot be corrected.
A highly instructive experiment can be arranged by using
two sets of tambours with springs and cords connected to two
fingers. Records are to be made of the two when making con-
tractions and relaxations at the same instants, of one contract-
ing only with every second contraction of the other, of one
contracting only when the other pulls %vith a certain strength,
etc. By letting one finger (1) represent the action of the larynx
in producing a tone and the other (2) the action of the other
speech organs, we can produce greatly simplified representa-
tions of speech movements ; thus, aka would be the succession
of movements \.2.\ \ papa would be 2-2-2-2 ! ^^c. The records
will show that the movements of the two fingers are not made
accurately together, and that in such combinations as \.2-\ the
action of (1) tends to overlap its proper place, even produc-
ing 2-2-2' ^^- Exactly analogous results will be found to
such phonetic phenomena as the change of aka to aga, of
papa to baba, or as, in some cases, the partial desonation of
z in givzs ' gives,' and the desonation of final vowels as in
Fr. veky^ ' v&u,' etc. Oertel suggests that such experi-
ments may clear up the obscurity of dissimilatory loss and
substitution of certain sounds such as 1, n, r.^
Closer imitations of the action of the speech organs may be
made by using several tambours attached to different fingers
or parts of the body. On the principle that all activities
follow the same laws and differ only in complexity, these ex-
periments can be used not only to demonstrate the laws of
sound modification but also to suggest modifications that
1 Oertel, Lectures on the Study of Language, 234, New York, 1901.
204 PRODUCTION OF SPEECH
might otherwise be overlooked. Letting one finger repre-
sent the tongue and the other the velum, the tendency of the
latter to relax its contraction before the tongue movement
changes illustrates such a phenomenon as the nasahzation of
a vowel followed by a nasal consonant.
When several fingers are required to perform movements
in succession, it will be found that certain associations are
favored. Letting the separate fingers perform different move-
ments we find that certain successions are easier than others.
It is easier to repeat the movement of a finger than to change
to another, to move the fingers in the order 1, 2, 3, 4, 5 than
1, 3, 5, 2, 4, etc. A close analogy to this is found in the
vowel harmony that appears in isolated cases in nearly all
Fig. 78.
languages and is well developed in Hungarian and Fin-
nish (p. 121).
A movement of a definite kind requires a definite time for
its execution. When a voluntary movement is repeated
with great rapidity, there is a limit beyond which the move-
ment loses in definiteness. This may be illustrated by repeats
ing the contraction of the finger with increasing rapidity ; Fig.
78 shows a typical record. The hurriedness abbreviates the
extent both of the contraction and of the relaxation ; at some
points the finger is almost immovably cramped. The slurring
of sounds in rapid talking is an analogous example in speech.
Under each set of circumstances there is a rate of repetition of
a muscular movement which each person takes naturally ;
this natural rate is the least fatiguing and the most accurate
VOLUNTARY ACTION AND THE GRAPHIC METHOD 205
one.^ This natural rate varies with the individual, with the
nature of the act, with fatigue, etc.
Reflex movements are more precise than voluntary ones,
movements unattended to more so than those that receive
attention.^ These principles are of application in learning
new speech faiovements.
A voluntary movement involves a more or less conscious
volition, that is, a phenomenon of decision known to the
person performing the movement. The greater the con-
sciousness of the decision, the greater is the amount of mental
energy consumed. The first lessons in learning new speech
movements are liable to be very fatiguing. Through interest
or excitement the fatigue may not be noticed by the learner
till afterward, yet its effect in increasing the error of execu-
tion (p. 201) often becomes apparent to the teacher. As im-
pressions made on a fatigued person are not so accurate or
lasting, it is generally more economical to have frequent
short lessons than less frequent long ones, if the latter pro-
duce fatigue. The acquirement of ability to perform a new
activity does not increase in proportion to the number of
times it is practised. The gain is at first slow, then more
rapid, then very rapid, then slower, etc., until finally it becomes
very small when the person has about reached the limit of
his improvement.^ After this point the practice is required
in order to avoid loss.* The dependence of the energy and
precision of a movement on the degree of fatigue, though
1 ScKiPTUBE, Observations on rhythmic action, Science, 1899 n. s. X 807.
2 Bliss, Investigations in reaction-time and attention, Stud. Yale Paych. Lab.,
1893 I 4.'j ; SCKIPTTJKE, New Psychology, 127, London, 1897.
" Feohner, Ueher d. Gang. d. Muskeliibung, Ber. d. k.-sachs. Ges. d. Wiss.,
math.-phys. Kl., 1857 IX 113; Henky, Recherches exp&. sur I'entrainement 7nus-
culaire, C. i. Acad. Sci. Paris, 1891 CXII 1473 ; Lombard, Some of the influ-
ences which affect the power of voluntary muscular contraction. Jour. Physiol., 1 892
XIII 14; Bbtan and Hartee, Studies in the physiology and psychology of the
telegraphic language, Psychol. Rev., 1897 IV 27; Johnson, Researches in prac-
tice and habit. Stud. Yale Psych. Lab., 1898 VI 51 ; Manca, Studi suW allena-
mento, Atti della R. Accad. d. Sci. Torino, 1892 XXVII 564.
* AscHAFFENBUEG, Experimentelle Studien u. Associationen, Psychol. Arbeiten
<(Krapelin), 1896 16, 11.
206
PRODUCTION OF SPEECH
repeatedly investigated for finger and arm movements,i has
not yet been considered in the case of the vocal organs.
The degree of conscious intention may be so small as to be
unnoticed. Such movements may be called ' unintentional '
ones. They are being constantly made all over the body.
They are of special interest from the fact that speech ideas
have a tendency to express themselves in such movements.
It is hardly too much to say that the vocal muscles are
continually making unintentional speech movements in con-
nection with our thoughts. These movements have been
\\!A
[]x=3)====;wv\A ^wv
■J ■ "^
•A ^
6
Fig. 79.
«,
recorded by experimental means.^ Under proper circum-
stances the faint speech sounds from them can be heard and
understood.^
To respond to a stimulus by to act requires time. This time,
the ' reaction time,' is longer the greater the nervous or men-
tal action included in the process.
The time required to respond to a sound by a vocal move-
1 Literature given by .Tote yko, Revue (je'n&ale sur la fatigue muscidaire, Annee
psychologiqije, 1899 IV 1 ; Hirschlaff, Zut Methodik u. Kritik d. Ergographen-
Messungen, Zt. f. piid. Psychol, u. Pathol., 1901 III 185.
'^ Curtis, Automatic movements of the larynx, Amer. Jour. Psychol., 1900 XI
237.
3 Hansen und Lehmann, Oeber unwillkUrliches Flustern, Philos. Stud. (Wundt),
1895 IX 471 ; summary in ScRiPTnRE, New Psychology, 65, 259, London,
1897.
VOLUNTARY ACTION AND THE GRAPHIC METHOD 207
ment may be registered on a recording drum (p. 7). The fol-
lowing arrangement will do. A current from the battery B
(Fig. 79) is sent through the primary coil P of an inducto-
rium, then through a magnetic marker M (p. 91), an open
circuit key K^ and a closed circuit key K^. A telephone T
is connected to the secondary coil S of the inductorium.
Pressure on K^ closes the circuit, makes a cUck in the tele-
phone and deflects the point of the marker. While K^ is
held down and the marker deflected to one side, pressure on
K^ will break the circuit and remove the deflection. The
marker is arranged to write on a drum beside the time line
from an electric fork (Fig. 17) ; the time of the deflection of
the point can thus be determined. The person experimented
upon is to respond to a click in the telephone by actuating the
key K^ , the time of the deflection is his ' reaction time.'
For measurements in thousandths of a second the records
must be corrected by adding the excess of latent time of the
marker at the make over that at the break (p. 92). This
correction can be avoided through using K^ as a break-make
key by the additional contact at the back ; the sound occurs
at the break ; since it is registered by a break deflection the
reaction time is measured between two break deflections and
no correction is needed for latent time. Many variations may
be made in the apparatus. It is often convenient to connect
the reaction key K^ to a separate marker; to use a spark coil
(p. 12) instead of the marker; to use a chronoscope (p. 152)
instead of the fork and the smoked drum. The key K^ may
be a chin key consisting of a telegraph key so placed under
the chin as to break the circuit when the jaw begins to
move (p. 154), a voice key that breaks the circuit by the
action of the air against a metal plate (Fig. 66), or a lip
key that breaks the circuit when the lips are compressed.
When the subject responds by a simple movement such as
lowering the jaw, blowing, or pressing the lips, the action is
closely like that of a response with the finger ; all these forms
are termed ' simple reactions.'
Of this reaction time very httle is consumed by the trans-
208 PRODUCTION OF SPEECH
mission of the sound from the telephone to the brain and of the
impulse from the brain to the vocal organs ; nearly the whole
of it represents the time required by the processes in the
brain. These processes are known to us in consciousness as
perception and volition ; the ' simple reaction time ' thus
indicates very closely the time of these two. An average
simple reaction time to sound will lie in the neighborhood of
0.20% varying with the individual, with the character of the
sound, with the character of the movement, with fatigue, etc.
In a simple reaction the only mental processes involved are
perception and volition. Two others may be added by requir-
ing the subject to react to one stimulus and not to another.
In the arrangement already described, the sound in the tele-
phone may be weakened by moving the secondary coil S
further away from the primary F. The subject, not knowing
which will be heard, is to react when he hears the weak sound
and not when he hears the loud one. He is thus obhged
to discriminate between two sensations and choose between
action and non-action. The time is greatly lengthened.
When the subject responds to the telephone click by a
speech movement such as that of a vowel or a consonant, the
action may be distinguished as a ' semi-vocal reaction. ' A
' complete vocal reaction ' may be measured with two voice
keys, each in series with a battery and a marker. It may
also be done with one key having either a double mouth-
piece or a large trumpet. The repetition of a spoken word
by the subject is a vocal reaction with discrimination and
choice. A more complicated form of reaction occurs when
the response is to be another word than the one heard but
related to it in some way. Thus, it might be required to
respond to an adjective by giving some noun; this would
give an 'association time.' The methods of measuring asso-
ciation time with a chronoscope have already been described
(p. 152).
The following further technical details concerning apparatus
will be found useful not only in connection with the experi-
ments of this chapter but also throughout laboratory work.
VOLUNTARY ACTION AND THE GRAPHIC METHOD 209
The success of experimental work often depends on the
electrical facilities of the laboratory. Batteries of different
kinds are adapted to different forms of work. Nitric acid
batteries (Geove) give strong currents for a few hours
but are troublesome to handle ; chromic acid (dip) batteries
give very strong currents for a brief time ; ammonium chloride
batteries (LBCLANCHi:), giving momentary currents and re-
quiring renewal only at long intervals, are suitable for open
circuit work, as for signals, bells, etc.; copper sulphate bat-
teries (Daniell) give very steady but weak currents for a
long time; potash (Lalande) and soda (Edison) batteries
give very constant currents of any strength for a long time
with little trouble of renewal; storage batteries, which must
Fig. 80.
be charged from a d3aiamo, are universally available ; low
voltage direct currents may be obtained directly from a suil^
able dynamo or by a dynamotor from a high voltage direct
or from an alternating current ; low voltage currents can be
obtained directly from a high voltage circuit by means of
lamp batteries.
The lamp batteries are so convenient in a laboratory sup-
plied with the usual direct current that a description of them
seems advisable. In the three-socket battery (Fig. 80) the
current from the main line is brought to the posts E F; the
socket A contains a lamp that allows the desired amount of
current to pass ; the socket B contains a small lamp of such
resistance that the fall of potential between its poles gives
the voltage desired for the experimental work ; the socket 0
14
210 PRODUCTION OF SPEECH
contains a plug with wires to the apparatus ; i) is a switch.
On a 110-volt circuit a large lamp A called a 110-volt 32 c. p.
lamp (about 110 ohms) allows 1 ampere of current to pass
through it; a small lamp B of 10 ohms reduces the cur-
rent to a little less than 1 ampere ; the tension at the two
poles of the socket B as indicated by the volt meter is
about 8 volts. When connection is made through the appa-
ratus, the current will be divided between the small lamp
and the apparatus in inverse proportion to the resistance;
for an apparatus with small resistance, e. g. a key and low
Fig. 81.
resistance telegraph sounder, almost the entire ctirrent of 1
ampere mil pass through it ; the tension, however, is that at
the poles of the small lamp, namely, 8 volts. For apparatus
of greater resistance small lamps of higher resistance are used ;
for currents of greater intensity larger lamps are used in A.
Experience has shown that a 4:-ampere and a 1-ampere pair of
lamps suffice for most experimental work.
The four-socket batteiy is of great convenience. The socket
Q (Fig. 81) is placed in series witli that of the small lamp B;
a plug mth vidres r\ins from it to one piece of apparatus
while wres run from C to another piece. To run a high
VOLUNTARY ACTION AND THE GRAPHIC METHOD 211
resistance telegraph sounder it is connected to the wires of
Q while the key is connected to those of C. As long as the
key is closed, the current passes almost entirely through C;
when it is opened, it is forced at the tension of the main line
(consequently with greater effect) through the sounder. To
run a high resistance marker (Fig. 61) with an electric fork
(Fig. 17) having a magnet of low resistance, the fork is con-
nected to G while the marker is connected to G;va. this way
a strong current is obtained through each. Closure of the
circuit at G- changes the battery into a three-socket one.
Refebences
For a summary of the facts of muscular contraction : Hermann, Lehr-
buch d. Physiologie ; Howells, American Textbook of Physiology ;
ScHAEFER, Textbook of Physiology ; Foster, Handbook of Physiology.
For the technique of graphic records : Langendorff, Physiologische
Graphik, Leipzig-Wieu, 1891. For the technique of reaction time
experiments : Wundt, Grundz. d. physiol. Psychol., 4. Aufl., II 322,
Leipzig, 1893 ; Scripture, New Psychology, Ch. VIII, London, 1897 ;
Scripture, Elementary Course in psychol. measurements, Exercises
IX-X-XII-XIII, Stud. Yale Psych. Lab., 1896 IV. For various methods
of recording rapidly repeated movements : Scripture, New Psychology,
Ch. VII, London, 1897; Cross education, Pop. Sci. Monthly, 1900
LVI 589. For the mathematical treatment of muscle-curves : Hall-
8T)£n, Analys af muskelkurvor. Acta Societatis Scientiarum Fennicse
(Helsingfors), 1898 XXIV No. 1; 1900 XXIX No. 5.
For recording drums: Petzold, Leipzig ; Zimmermann, Leipzig;
Albrecht, Tubingen ; Verdin, Paris. For tambours : Verdin, Paris ;
Albrecht, Tubingen. For clamps, standards, scalpels, etc. : Petzold,
Leipzig ; Zimmbrmann, Leipzig ; MiJncke, Berlin. For Harvard
physiological apparatus : Prof. W. T. Porter, Boston. For Auzoux's
model of the brain (Fig. 70) : Montaudon, Paris.
CHAPTER XVI
BEEATHING
The diaphragm forms the bottom of the thorax. When at
rest it curves upward, part of it lying on the walls of the
thorax. When its muscles contract, it descends and the side
parts are removed from the walls. In this way the thorax is
lengthened. The pressure of the diaphragm on the organs
below pushes the abdomen outward.
Other muscles act to lift the ribs and thereby to deepen and
widen the thorax. According to the preponderance of the
diaphragm movement or the rib movement two types of
breathing are distinguished, the abdominal and the costal.
On account of atmospheric pressure the elastic walls of the
lungs are obliged to follow the walls of the thorax, and thus
air is drawn in. The lungs also descend slightly during
inspiration, drawing after them the trachea.
Expiration is mainly passive.^ When the inspiration-mus-
cles are relaxed, the elasticity of the lungs makes them con-
tract ; this draws the diaphragm upward and the walls of the
thorax inward, while gravity aids in lowering the ribs. The
air in the lungs is thus partly forced out again. To produce
forcible expiration the muscles of the abdomen may be con-
tracted ; the contents of the abdomen are pressed inward and
upward and the diaphragm and lungs are forcibly moved.
The ribs are also drawn downward by the same muscles, as
well as by sets of special muscles.
In ordinary breathing about 500'^°™ of air are inspired and
1 Perhaps slightly active ; Treves, Observations sur le mecanisme de la
respiration, Aroh. ital. de biol., 1899 XXXI 130.
BREATHING 213
expired.^ A forcible expiration can bring out about 1600"™
more, called the ' reserved air.' The lungs are not exhausted
by the deepest expiration, as there stiU. remains 'residual air ' to
the amount of 800"""°. With the deepest possible inspiration
an additional amount of 1600°™ can be added, called the
' complementary air.'
The movements of inspiration and expiration occur invol-
untarily with a definite rhythm and depth. The average
frequency with adults is 18 to 20 a minute. Both rhythm
and depth can be largely controlled by the will. The emo-
tions affect frequency, depth and form of the breathings and
sometimes cause characteristic noises and tones, as in sobbing,
sighing, laughing, etc.
The nerves controlling the breathing muscles come from the
spinal cord, where they have connection with the respiration
center in the bulb (p. 193). The activity of this center depends
on the nature of the blood reaching it. As the blood becomes
more or less venous the breathing becomes more or less violent.
Muscular activity causes an acceleration of breathing by a
chemical product that enters the blood. Heat also stunulates
the respiration center.
The movements of single points of the body during breathing
may be conveniently studied by the method of air-transmission
by means of the Marby tambours recording on a smoked
drum (p. 195). To register thoracic or abdominal movements
of a person sitting or lying down, a light projection at right
angles may be attached to the lever of the receiving tambour
of the arrangement shown in Fig. 71, and the tambour is so
placed on supporting rods that this projection rests upon the
desired point of the chest or abdomen. The movements are
registered in the way described on p. 198.
If preferred, the receiving tambour may have a small projec-
tion instead of a lever attached to its surface.^ Such a tambour
1 This is the usual statement ; according to Marcet, A contribution to
the history of the respiration of man, Croonian Lecture', London, 1897, the
average of 210 experiments gave 250""=™.
^ Bebt, Lefons snr la physiologie compar^e de la respiration, 290, Paris, 1870.
214
PRODUCTION OF SPEECH
is shown at k in Fig. 82 ; it may be used in any support on a
person lying down or otlierwise fixed so that the body will not
sway ; or it may be placed on one end of the adjustable calipers
rp, arranged to clasp the thorax on any diameter and to
register independently of a swaying motion.
The relations between costal and abdominal breathing can be
• studied by using several sets of tambours recording simulta-
neously on the same drum. The lengths of tubing and the
amplification adjustments should be
the same for all. The older doctrine of
a preponderance of abdominal or tho-
racic breathing in the male or female
sex seems without foundation except
in so far as brought about by clothing. '
The changes in the circumference of
the thorax or abdomen may be regis-
tered by means of a ' pneiunograph.'
In one form (Maeey) a hollow rub-
ber tube is kept expanded by a spiral
spring ; this tube is closed by metal
ends, one of which has a projecting
outlet connected to the small rubber
tube from the recording tambour. A
band is stretched around the body
from one end of the tube to the other.
An increase in the circumference of
the thorax stretches the tube and thus
draws air from the recorder. In a
better form of the pneumograph the
tube is of metal and the ends of rubber (Bert). In still
another form a tambour hke that in Fig. 72 is acted upon by
a band and a compound lever (Maret). A later form (Fig.
83) comprises two tambours with adjustable amplification
(Veedin) .
Some records with a pneumograph (of the first of the above
Fig. 82.
1 FiTZ, j1 studji of types of respiratory movement, Jour. Exper. Med., 1896 I 677.
BREA THING
215
Fig. 83.
kinds) aroiind the abdomen of a male person with chiefly
abdominal breathing are shown in Fig. 84. Ordinary breaths
followed by several deeper
ones are shown in the top
line of records ; it will be
noticed that the movements
are very small after the blood
has been refreshed by deep
breathing. A record of ordi-
nary breathing interrupted by
sniffing, sobs and a sigh-like
sob are shown in the second
record; the inspirations are
very sudden. The curves for a groan and a sigh are also
shown in the third record; the inspirations are not sudden,
and the expirations are more gradual than in the sigh, the
groan showing a specially long and irregular expiration.
All these sobs, groans and sighs were produced premedi-
tatedly. A series of premeditated laughs is also shown.
Each laugh consisted of 'ho-ho-ho-ho ' with falling pitch;
the laugh occupied the expiration-half of each curve. The
record marked '4 lines of song ' shows the breath expenditure
during the singing of
" Way down upon the Swanee River,
Far, far from home.
Oh, darkies, how my heart does quiver,
Far from the old folks at home."
The expiration of the breath not used during each line ap-
pears clearly each time at the end. The next to last record
shows the use of the breath in speaking the verses
" The Cities are full of pride.
Challenging each to each ;
This from her mountain-side.
That from her burthened beach " (Kipling).
The inspiration occurred just before the beginning of each
line. The last record shows the breath -expenditure when the
216
PRODUCTION OF SPEECH
stanza was spoken more rapidly; one deep inspiration with a
slight accession afterwards is made to do for each pair of
lines. The discharge of the air not used in speaking is indi-
FiG. 84.
cated by the sudden rise at the end of each line. Both rec-
ords were made with no intentional distribution of the
inspirations. The time-line with seconds is given for all
these records at the bottom.
BREA THING
217
The records made from the surface of the thorax or abdo-
men give indications concerning the suddenness of inspira-
tion and expiration and concerning the rate at which the
breath is expended. They give no direct measurements of
quantity or pressure. The relative amounts of breath used in
single sounds can, however, be indicated with this method by
repeating them a given number of times in a single expiration.
The expenditure of breath
was found in one case by
RoussELOT to stand in the
following relations: fa
> va > pa > ba.i
In taking records of
breathing during speech it
is necessary to guard'
against apparent small in-
spirations at the begin-
nings and endings of words ^
which are in reality due to
extraneous muscular con-
tractions.*
To register the variations in the breath from the mouth
or the nose, the receiving tambour (Fig. 71) is removed and
the end of the rubber tube is placed on a short glass tube ^
which is inserted loosely in the corner of the mouth or in one
nostril. The main body of air passes by, but the variations in
pressure will affect the recording tambour. The curves in
Fig. 85 were taken with the tube held at the corner of the
1 RonssEi.OT, Les modifications phon^tigues du langage, 62, Kev. d. pat. gallo-
rom., 1891 IV, V; also separate.
2 OussoF, Andes exp&imentales sur une prononciation russe. La Parole, 1899
I 785.
3 BiNET ET Henei, Les actions d'arret dans les phenomenes de la parole, Eev.
philos., 1894 XXXVII 608.
* Gregoike, Note sur faction du thorax dans la phonation, La Parole, 1899 I
718.
6 Glass tubing may be cut by first scratching it slightly with a triangular (hard)
file and then bending it with the fingers ; its edges are rounded by insertion in a
Bnnsen flame.
Fig. 85.
218
PRODUCTION OF SPEECH
open mouth. They show a series of equal breaths and several
attempts at blowing with forces in the relations 1:2:3:4.
The variations in mouth-pressure during the recitation of four
Fig. 86.
lines of The Cities (p. 215) and of the first four lines of Cock
Rohin are shown in Fig. 86. The expiration curves for the
word du spoken in various ways have been registered by
ViBTOE 1 mth a short tube held between the lips. The height
•^aj^
Fig. 87.
of a curve in Fig. 86 shows the force of expiration spoken in
an assertion, in a question, in irritation, in warning. The
full pressure from the mouth can be obtained by using a.
1 ViETOE, Kleine Beitrdge zur Experimentalphonetik, Neuere Sprachen, 1894
I, Suppl., 25; Elemente der Plionetik, 4. Aufl., 283, Leipzig, 1898.
BREATHING
219
mouth-piece fitting over the lips, that from the nose by a
nasal olive (Fig. 88) fitting the nostril.
Attached to a speaking tube whose end fits the mouth
rather closely, the Maeby tambour can be used to register the
expiratory impulses and thus to give
data concerning the lengths of sounds
and also some information concerning
their force. Since the Faxis in tam-
bour records is a peculiar curve, the
moments of time cannot be found by
lines perpendicular to the X axis but
must be referred to it by means of
curved lines. The exact curve of the
Y axis is readily found by using the tambour while the drum
is at rest (p. 197). The special vocal tambour of Rousselot i
(Fig. 89) comprises a bent metal tube ending in a membrane
of rubber which moves a very fight recording point. The
mouth-piece A passes into the tube D which ends in the rub-
ber membrane E. The vibrations of the membrane are recorded
Fig. 88.
Fig. 89.
by the lever F, adjustable by the screw (r. The combination
of clamps I H B afford great adjustabifity on the supporting
rod O. This tambour is adapted to registering small fluctua-
tions in pressure such as are found in various speech sounds
like r, the explosives, etc.
To give some idea of the total volume of air expended, a
1 RonssELOT, Laphonitique experimentale, La Parole, 1899 I 9.
220 PRODUCTION OF SPEECH
mouth-piece fitting closely over the lips may be attached to
a tube directly from a recording tambour for a short sound
and by way of a reservoir for longer sounds. The amount of
air corresponding to each degree of the excursion can be deter-
mined conveniently by attaching a graduated syringe to the
tube;i this is best accomphshed by inserting a T-tube in
the rubber tube connected to the tambour and attaching the
syringe to the long arm of the T- The tambour method, how-
ever, is not very accurate for this purpose. A tambour with a
mouth-piece communicating with the external air may be
used to indicate the rate of expenditure.
An accurate instrument that directly and proportionately
records the volume of air expired or inspired is found in the
breath recorder of Gad^ shoAvn
in Fig. 90. A box-like mica
cover d with its edges im-
mersed in a square trough of
water t is so balanced upon
the axle a by the weight g that
it is in equilibrium in any posi-
tion. The air entering through
the tube r in the bottom plate h raises the cover d and records
by the arm h. The graduation of the record is done by sending
in known quantities of air and marking the position of the
pointer. For this purpose a large syringe graduated in ccm.
may be attached to the tube before or after an experiment.
An apparatus like this in which the records are directly pro-
portional to the amount of air is often preferable to a tambour
with which the scale diminishes rapidly as the amount
increases. A special registering spirometer has been devised
by Makcbt.^ ■
In an ordinary expiration the air is driven out by the tension
of the lungs and the abdominal muscles ; the pressure begins
1 Lombard and PiLLSEURy, A new form of piston recorder, Amer. Jour.
Physiol., 1899 III 186.
2 Langendorff, Physiologische Graphik, 266, Leipzig und Wien, 1891.
5 M ARCET, Etudes des diff&entes formes de la respiration de I'homme, Rev. me'd.
de la Suisse romande, 1896 601.
Fig. 90.
BREATHING 221
with a maximum and falls to zero. For singing, a constant
rate of expenditure is often required. This must be obtained
by muscular resistance, such as that of the diaphragm and
of the muscles around the thorax. A tracing with the pneu-
mograph should show a steady movement of the chest or
abdomen and not one of varying rapidity or of irregular char-
acter. Some of the troubles of singers arise from letting out
the air too rapidly at first ; these can often be cured by proper
instruction in breathing, profitably aided by registering the
results with the pneumograph.^ The modulation of the voice
in song requires accurately coordinated and controlled move-
ments of the breathing muscles ; this is accomplished by modi-
fications of the usual respiratory action.
Using a Veedin spirometer — a rather rough instrument
constructed for other purposes but giving sufficiently accurate
results — RouDET^ found that under equal conditions the rate
of expenditure increased with the intensity of the sound. The
vowel a sung on the note df' during one second at three
degrees of intensity — feeble, medium, strong — showed, as
averages of 20 experiments, rates of expenditure of 10.6'=°™,
16.4°™ and 24.1°°" respectively. The increase in amphtude of
the vibration in a louder sound means that the glottis opens
more widely for each vibration and lets out more air.
When the vowel a was sung on the notes c", e°, g'^, c^ with
what appeared to be a constant intensity, the records showed
a decrease of the rate of expenditure with a rise in pitch.^
The physical intensity of a vibratory movement (p. 109) in-
creases as the square of the frequency and the square of the
amplitude ; two tones of the same amplitude in the relations
of 1 : 2 in frequency stand in the relations of 1 : 4 in physical
intensity. To maintain the relation 1 : 2 in frequency while
obtaining the relation 1 : 1 in physical intensity the relation of
1 Oliviek, Stiologie et traitement de certains troubles vocaux, La Parole, 1899 I
367.
2 EouDBT, De la depense d'air dans la parole et de ses consequences phonetiques,
La Parole, 1900 11 209.
8 KouDET, as before, 214.
222 PRODUCTION OF SPEECH
amplitude must be made 2:1; thus the higher the tone the
smaller the amplitude, and consequently the smaller the rate
of expenditure. The psychological relations of intensity — by
which RouDBT judges the equality of the sounds — for sounds
of different pitch presumably bear some relations resembling
the physical ones. The compensation is thus a nervous and
mental one. Rotjdet's first explanation by an increase in the
area of the glottis is unnecessary and probably erroneous ; his
second explanation is approximately the one I have just
given.
Experiments ^ with a sung on the note d'^ with apparently
the same intensity for l^ 2% 3^ showed rates of expenditure
that were 13.5"=™, 7.9'=°" and 6.2'=™ per second respectively.
This does not show, as Roxjdbt supposed, that there is uncon-
scious economy in the respiratory distribution, but that the
judgment was one of equal energy and not one of equal
intensity. In these experiments, as in many others, the sub-
ject presumably felt instinctively that he was to produce the
same total effect in each case. The energy of a sound varies,
both physically and mentally, in the same way — though not
proportionately — as the amplitude, pitch and length. Two
sounds are considered equal in energy (' intensity ' is the
usual term) when they produce the same total mental impres-
sion. A longer sound produces a more energetic effect and
is instinctively lowered ia pitch or amplitude when it is to
be made equal to a shorter one. These experiments confirm
the principle of compensation among the three factors.^
Measurements ^ of the rate of expenditure during different
vowels sung on the note # during 1^ showed for u^ 22.6,
03 21.7, Oj 16.5, ag 13.1, a^ 15.7, Cj 16.2, e^ 19.6, Cg 21.3,
ig 26.3°°" (1, 2>3 indicate 'open,' 'medium' and 'close' respec-
tively). As a general rule the rate of expenditure increased
as the passage above the tongue diminished. Rotjdet's ex-
1 RouDET, as before, 215.
'^ SoRiPTtTEE, Researches in experimental phonetics (first series), Stud. Yale
Psych. Lab., 1899 VII 100.
' RouDET, as before, 215.
BREATHING 223
planation that the friction of the air in the mouth passage
acts directly by back pressure, so as to produce a larger open-
ing of the glottis whereby more air can pass, indicates merely
a difference in the glottal action. The rate of expenditure is
unquestionably governed by the respiratory muscles and the
glottis; a particular rate is associated with each vowel for
reasons still unknown, possible in reference to auditory im-
pressions of the energy usually felt in each.
Measurements ^ of the average rate of expenditure for the
vowel a pronounced in several ways gave with the glottal catch
ggccm^ with a clear beginning (the cords closed for vibration
before breath action) 14.0°°", with a breathed beginning (cords
not fuUy closed before breath action) 24.2"°" and when
whispered 34.4"". Measurements ^ of six fricatives gave:
s 30.6, z 27.5, f 33.5, v 27.9, s 38.6, z 28.2. The expendi-
ture was much greater than in vowels ; it was greater for the
surds than for the corresponding sonants.
RouDET finds that an explosive has a smaller total expendi-
ture than a fricative, but the average rate of expenditure —
that is, the total expenditure divided by the duration — is
incomparably greater; among the explosives both the total
expenditure and the average rate are greater for a surd than
for a sonant, for an aspirated form than for a pure explosive.*
In Italian it has been shown* that the more closed the artic-
ulation the greater is the average rate of expenditure, for
example, i > o > a.
The expenditure of breath during vocal sounds is quite
variable but in ordinary speech certain constant relations are
found. The following facts in regard to French sounds of
Cellefrouin have been established by Rotjsselot ^ by means of
tambour records. The expenditure differs at different times
1 RoDDET, as before, 217.
2 RotJDET, as before, 220.
8 RocDET, as before, 222.
* JosSELTN, Stude sur la phone'tique italienne, Thfese, Pari-s, 1900 ; also in La
Parole, 1900 II.
5 RousSBLOT, Les modifications phon^tiques du langage, 65, Rev. d. pat. gallo-
rom , 1891 I ; also separate.
224
PRODUCTION OF SPEECH
nose
of the day, with different activities, with the position (seated
or standing), etc., although the person may consider the sound
to have been produced with the same force. Measurements
with a spirometer showed the following relations of breath
expenditure for consonants followed by a to be true without
exception for all subjects tested : p<f, t<s, k<ts< s<
X, b < V, d < z, g < dz < z < Y. This indicated that fricar
tives required more air than explosives ; nasals required less
air than the corresponding surd
explosives, ni<p,n<n<t;
fricative surds used more breath
than fricative sonants, v < f , z <
s, z < s ; finally r < 1. As final
sounds the fricatives required more
breath than explosives, nasals less
than the corresponding oral sounds,
all surds more than the sonants, 1
more than r and A about the same
as ]. Although some results were
conflicting, the general relations of
breath expenditure for the vowels
were : ig < I3, y^ < yg, o^ <
(for
Fig. 91.
meanings of the inferior numerals
see p. 222). The relaxation of artic-
ulation in one vowel as compared with another (from
which it may be historically derived) is accompanied by in-
creased rate of expenditure of breath but often also by a
shortening of duration. In nasal vowels less breath was ex-
pended through the mouth than in oral ones.
Tambour curves for the nose and mouth made by Gold-
SCHEIDER ^ show clcarly the final rush of breath as the cords
open after a vowel is finished (a, Fig. 91). This rush does
not appear when a vowel is purposely made to end softly.
The relations of the two currents of air in a nasal are shown
in the nose and mouth curves for ma.
^ GoLDSCiiEiDEE, Ueber Sprachstorungen, Berliner klin. Wochenschr., 1891
XXVIII 487.
BREA THING
225
The pressure of air may be obtained by connecting a tube
from the mouth to a water or mercury manometer. This is a
U-shaped glass tube with water or mercury at the bottom ;
when one arm is connected by a rubber tube to the mouth, the
water or mercury will rise in the other arm to a height depend-
ing on the air-pressure. Water records may be changed into
mercury records by dividing by 13.6. A convenient form of
mercury manometer is shown in Fig. 92. The mercury in the
U-tube m rises along the scale
so according to the pressure
transmitted through the tube e.
The apparatus can be con-
veniently used in connection
with a recording tambour.
The records of pressure are
taken in the usual way with
a tambour; they are then
graduated by attaching the
tambour tube to e. By means
of the bulb b, compressed by
the plate p by moving the
screw s, the apparatus attached
to e can be made to mark a
graduation on the record for
each unit of pressure.
After an inspiration the chambers of the thorax are filled
with air at the atmospheric pressure ; closure of the glottis and
relaxation of the inspiration muscles produce an increase of
pressure, measured through an external tracheal opening, vary-
ing from 3°" of water in whisper to 97°"" in a loud cry, from
13 to 16 in ordinary speech.^ To indicate the pressure
Avithin the mouth a small rubber tube from a U-shaped water
manometer is inserted into the mouth and passed behind the
place of closure.2 The effect of the glottal closure in v or b
1 CAGNlAKD-LATOtlK, Sur la pression a laquelle I'air contenu dans la trachge-
arlere se Irouve soumis pendant I'acte de la phonation, C. r. Acad. Sci. Paris, 1837
IV 201; Aun. d. Sci. Nat., 1837, 2me se'r., VII 180, VIII 319.
2 SiEVBRS, Grundziige d. Phouetik, 4. Aufl., 22, Leipzig, 1893.
15
Fig. 92.
226 PRODUCTION OF SPEECH
as compared with f or p in diminishing the pressure in the
mouth appears readily. It may be suggested that the stronger
articulation of the surds, even in whispering, arises from the
associated habit of resisting increased air pressure.
To measure the pressure of the air during speech, Weeks ^
used a small metal tube of 2°"° diameter inserted into the
corner of the mouth and bent around the teeth to the center
of the palate at the beginning of the arch where the tongue
rarely touches. Attached to a Makey tambour this indicated
the air pressure in the mouth at each moment. With this in-
strument Weeks showed that final h, d and g in the South
German pronunciation are not the same as p, t and k, as
they appear to the ear, but differ in being spoken with less
pressure.
In a sonant sound part of the lung pressure is used in set-
ting the cords in vibration. In a sonant explosive the
remaining pressure is borne by the mouth. In a surd explo-
sive, with open glottis, the entire lung pressure is borne by
the mouth. In whispering, a small part of the pressure is
borne by the glottis. With a constant lung pressure the
variations of mouth pressure in surd and whispered sounds
may arise from different degrees of closure of the glottis.
The lung pressure can hardly be supposed to vary from one
sound to another and we may well assume that the decreased
mouth pressure for b„, go, d^ (surd J, d, g) as compared with
p, t, k indicates more closure of the glottis. The essential
difference between bo, do, g„ and b, d, g (sonant b, d, </) thus
lies in a weakening of the glottal vibration to a whisper
action but not to the condition of rest found in surd sounds.
RosAPELLY 2 has, in fact, observed — with a laryngoscope and
an obstruction between the jaws — that in whispered aba the
glottis remains of constant width whereas in apa it opens
more widely during the p.
1 Weeks, Recherches expgrimentales de phonetiqne, Annee psychologiqne,
18931174; summary in Maitre phon^tique, 1894, juin.
''■ RosAPELLT, M^m. de la Socie'te de linguistique, IX 488 ; Rousselot, Prin-
cipes de phone'tique expe'r., 469, Paris, 1901.
BREA THING 227
With a water manometer consisting of a U-tube Roudet ^
registered the subglottal pressure in a person having an ex-
ternal opening into the trachea. In ordinary inspiration the
pressure became at first rapidly negative and then slowly
returned to zero; in expiration it became rapidly positive
and slowly returned to zero. In ordinary breathing the
minimum pressure usually reached —2"'" water while the
maximum reached +4°"" ; during speech the extreme figure
for expiratory pressure was 20°". When a (as in Fr. ' pas ')
was sung on the note e" with three degrees of intensity, the
pressures indicated were : feeble ll"", medium 14™, loud 19°".
When a was sung on the three notes c", e", ^ with apparently
the same intensity, the pressures were 19.5™, 17.5°'", 12.5°"', as
was to be expected from the fact that to be of equal physical
intensity sounds of different pitches must have amplitudes in-
versely proportional to their frequencies (p. 109). For vowels
sung on the same pitch with apparently the same intensity the
pressures were: for u 15.0, o 13.3, a 13.3, e 14.5, i 17.0°".
The pressure seemed to be slightly greater for the close
vowels. In whispered vowels the pressure varied greatly ;
for example, from 6°™ to 15°" for a. In a series kept at
the same intensity the pressures were : for u 10.5, o 9.0,
a 9.0, e 10.0, i 12.0™- The pressures for a series of sylla-
bles spoken on .e" were: for pa 15, ba 11, ta 12, da 11, ka
17, ga 15, fa 12, va 11, sa 14, za 13, sa 16, za 14, ma 15,
na 16, la 13, ra 16°™. The pressure was thus greater for the
surds than for the sonants.
The air pressure in millimeters of mercury during the
vowels in ordinary and ventriloquistic speech has been found
to be : ^
a
e
1
0
u
ordinary :
ventriloquistic :
10
60
40
70
30
55
30
70
50
60
1 Roudet, Recherches sur le role de la pression sousglottique dans la parole, La
Parole, 1900 II 599.
2 Flatau tjnd Gutzmann, Die Bauchredner-Kunst, Leipzig, 1894.
228 PRODUCTION OF SPEECH
The amount of air actually used, however, in ventriloquism
is much less than in ordinary speech ; in one experiment with
a spirometer 900""™ were used in saying ventriloquially what
required 1300""™ ordinarily. The curve of inspiration traced
from the abdomen rises more slowly in ventriloquism.
References
For breathing in connection with song and speech r Geutzner, Pliysio-
logie (1. Stimme u. Sprache, Hermann's Handbueh d. Physiol. I (2),
Leipzig, 1879; Joal, Respiration en chant, Paris, 1895; Curtis, Voice
Building and Tone Placing, New York, 1896. For anatomical plates :
SpALTEHOLTZ, Handatlas d. Anatomie d. Menschen, Leipzig, 1900 (Engl,
ed. by Barker); Testut, Trait6 d'anatomie humaine, 4me ^d., Paris,
1899 ; Quain, Elements of Anatomy, London, 1896 ; Gray, Anatomy,
Descriptive and Surgical, Philadelphia, 1901.
For model to illustrate ruction of lungs and diaphragm : Kohl, Chem-
nitz. For Auzoux's separable models of man : Montaudon, Paris. For
separable paper models : Witkowski, Anatomie iconoclastique (Le corps
huniain), Paris, 1879. For tambours and pneumographs : Verdin, Paris ;
ZiMMERMANN, Leipzig. For recording drums : see references to Chap. I.
CHAPTER XVII
VOCAL OEGANS
Ik the widest application of the term the ' vocal organs '
include all the organs directly concerned in speech and song ;
the nervous organs have been considered in Chapters VII and
XV, the breath organs in Chapter XVI ; the larynx will
receive separate treatment in the following chapters.
The general outline of the organs above the lungs that are
directly concerned in speech is given in Fig. 93.
The trachea (i^) is tlie outlet for the lungs. The rings of
cartilage (upper one at 2T) keep it distended. Behind it lies
the oesophagus (D) and the backbone (T). The trachea ends
in the larynx (i/). The cricoid (19') and the thyroid (18, T)
cartilages and the vocal cords (^ff) will be described in the
chapter on the larynx. In ordinary expiration the air passes
through the trachea and larynx past the epiglottis (E) into the
pharynx (C C) through the nasal cavity (A) and out the
nostril (n~) on each side.
The nasal cavity (A) on each side is of very complicated
form owing to the various processes projecting into it (2, 3,
4-}. The oral cavity (B} is roofed by the hard palate (/)
and the velum or soft palate (ii). The pharyngeal cavity
C'Cmay be divided by closure of the velum (11) across it
into two parts, the upper or nasal portion ( C) and the lower
or oral portion ( C")- In this case the entire mouth cavity from
lips to larynx is made up of the oral portion (B) and the
pharyngeal portion (C). The fall of the velum turns the
pharynx into a single cavity and separates it more or less
from the oral cavity.
The muscles controlling the movement of the lower jaw
230
PRODUCTION OF SPEECH
are: 1. th%temporal(l. Fig. 95), which raises it and, if it has
been projected, draws it back; 2. the masseter (i, Fig. 94;
W, Fig. 95), which raises it; 3. the internal pterygoidal, which
raises it and may give it slight side movement; 4. the ea;-
TiG. 93.
ternal pterygoidal, which projects it or twists it to one side.
The last two muscles are attached to the back part of the jaw.
The orbicularis oris (9, Fig. 94; 7, Fig. 95; 0, Fig. 96) con-
sists of a muscle layer formed of fibers radiating from the
corners of the mouth into the upper and lower lips, some of
the fibers running around the corners. When the fibers of
VOCAL ORGANS
231
the upper and lower lips act together, they constrict the
mouth. Physiologically each of these muscles can be divided
into an outer zone (away from the mouth) and a marginal zone.
Contraction of the outer zone alone compresses the lips and
projects them, that of the inner zone alone presses them back
against the teeth. Each lateral half of each muscle may act
independently. The buccinator (11, Fig. 95 ; b, Fig. 96) is a
flat muscle at the corner of the mouth, forming a large portion
of the cheek. It pulls back the corner of the mouth, closes
7
Fig. 95.
the lips and presses them and the cheeks against the teeth.
The mingling of the fibers of the buccinators and the orbicu-
laris oris is shown in Fig. 96. The triangularis (12, Fig. 94;
10, Fig. 95 ; t. Fig. 96) draws down the corner of the mouth.
The quadratus of the lower lip (10, Fig. 94; 8, Fig. 95; q-i.
Fig. 96) pulls the lower lip out and down. The quadratus of
the upper lip (q-s. Fig. 96) is a flat muscle dividing above
into the infraorbital branch (6, Fig. 94; 3, Fig. 95), the
angular branch (4, Fig. 94; 2, Fig. 95) and a zygomatic
branch whose fibers have the same insertion in the skull as
the zygomatic muscle (2, Fig. 94). The incisivus of the lower
232 PRODUCTION OF SPEECH
lip (^■-^,,Fig. 96) is a small muscle covered by the quadratus of
the lower lip aiad lying on the edge of the orbicularis from
the corner of the mouth to the jaw just below the lateral
incisor tooth; it pulls the corner of the mouth toward the
middle and downward. The incisivus of the upper lip (is,
Fig. 96) is a similar muscle covered by the upper quadratus;
it draws the corner of the mouth inward and upward. The
canine muscle (7, Fig. 94 ; 6, Fig. 95 ; c, Fig. 96) raises the
corner of the mouth. The zygomatic (S and 8, Fig. 94)
raises and retracts the corner of the mouth. The mentalis
(ii,Fig. 94; 9, Fig. 95; m, Fig. 96) raises the skin of the
Fig. 96.
chin and thus helps to protrude the lower lip. The thin
risorius (13, Fig. 94) pulls back the corner of the mouth.
The velum, or soft palate, (11, Fig. 93; c. Fig. 97) is a soft
fold just back of the hard palate. On looking into the mouth
it is seen to hang down in the back ; in the center it carries
the uvula (12, Fig. 93 ; F, Fig. 97) and on the sides it falls
in two arches, the front one being the glossopalatine and the
rear one the fharyngopalatine arch. The vertical portions of
the arches are known as the anterior and posterior pillars of
the velum. The velum is composed of a fibrous membrane
with muscles and a mucous covering. The palatine tonsil on
each side lies between the two pillars.
The elevator of the velum (b, Fig. 97) rising from the tem-
poral bone and the cartilaginous wall of the Eustachian tube
VOCAL ORGANS
233
and passing obliquely downward, serves to raise the middle
portion of the velum. The tensor of the velum (e) comes
from the skull, passes downward, turns over a hoop-like bony-
process at a right angle and enters as a thin broad layer into
the velum which it serves to stretch. The muscle (a) of the
uvula (J*) springs from the fibrous membrane of the soft
palate and passes into the uvula, spreading out its fibers
which insert themselves in the mucous membrane of its sur-
face ; it shortens the uvula and raises it up and back.
The glossopalatine muscle {R, Fig. 97; 7, Figs. 98 and 99)
comes from the tongue and passes up the anterior pillar to
the velum.
The pharyngopalatine muscle (Jf-, Fig. 97) passing from the
rear portion of the velum descends by the posterior pillar
partly to the central portion of the pharynx and partly to
the larynx. The muscle serves to narrow the nasopharyn-
geal opening and to raise the pharynx and larynx.
The hyoid bone (h. Fig. 93; B, Figs. 98, 99) is a
234
PRODUCTION OF SPEECH
U-sbaped bone lying just above tbe larynx. The mylohyoid
(^16, Fig. 93) is a muscular layer connecting the hyoid bone
to the front part of the lower jaw. Just above it near the
middle is tbe geniohyoid muscle (15, Fig. 93; ^ Figs.
98, 99). Also above it on each side is the digastricus
running from the front part of the jaw partly to the loop
(13, Fig. 98) and partly through it to the side of the skull.
All three approximate the hyoid bone and the jaw.
c
11' 13
■ Jcrc'J
B
Fig. 98.
The median septum (13, Fig. 93; 9, Fig. 97; s, Fig. 100)
extends along the middle of the tongue.
The styloglossus (1, Figs. 98, 99; 6, Fig. 97) runs from
the styloid process ((7) on the side of the skull to the side
parts of the tongue. The two muscles draw the tongue (par-
ticularly the rear portion) up and back and press it against
the soft palate. Some of the fibers also cross from one side
to the other in the tongue.
VOCAL ORGANS
235
The hyoglossus (2, ^', Figs. 98, 99 ; 7, Fig. 97) runs from
the hyoid hone upward and forward to the point of the tongue.
It pulls the tongue hack and down.
The genioglossus (14, Fig. 93 ; 5, Figs. 98, 99 ; g, Fig. 100),
the largest of the tongue muscles, runs from the inner surface
of the front of the lower jaw (A) into the tongue and radi-
n7T
Fig. 99.
ates on each side of the septum (13, Fig. 93) to all parts of it.
The lower fibers run to the hyoid bone and serve to raise it,
the larynx and the tongue; some fibers also go to the epi-
glottis. The median portions draw the tongue forward and
tend to push the point out of the mouth. The anterior fibers
draw the tip down and back.
The inferior longitudinal muscle (6, Figs. 98, 99; 5, Fig.
236
PRODUCTION OF SPEECH
97) is a band passing under the surface of tongue to the apex.
Tlie muscle shortens the tongue and makes it more convex
lengthwise.
The superior longitudinal muscle (5, Fig. 97; Is, Fig. 100)
forms a layer on the upper surface with fibers running from
the septum outward and forward to the free edge. It con-
caves the tongue upward longitudinally.
The chondroglossus (B", Figs. 98, 99) rises from the hyoid
bone and blends with the muscular substance of the tongue.
The muscle aids in drawing the tongue back and down.
KiG. 100.
The fibers of the transverse lingualis (t, Fig. 100), inter-
lacing with bundles of the genioglossus, pass from the sub-
mucous layer at the sides of the tongue to or through the
septum (13, Fig. 93); they are found in all parts of the
tongue. They compress the tongue transversely and make it
convex sidewise.
The fibers of the vertical lingualis (y, Fig. 100) are found
near the edges of the tongue; they connect points in the
upper and lower surfaces ; they flatten the tongue and push
out the edges.
The frenum linguae is a fold of mucous membrane con-
necting the under part of the tongue to the jaw; when too
short it interferes with the movements of speech.
VOCAL ORGANS
237
Three constrictor muscles, superior, middle and inferior
(1, 2, 3, Fig. 101), serve to diminish the cross section of the
pharynx. The latter two also shorten the pharynx by raising
the hyoid bone and the larynx.
The superior constrictor of the pharynx (1, Fig. 101; 8\
Figs. 98, 99) forms a band only about 2™ broad around the
nasal part of the pharynx. The contraction of this muscle
Fig. 101.
produces a ridge against which the velum presses to close
the nasal opening.^ One of its parts is the glossophart/ngeus
(4, Fig. 97; 8, Figs. 98 and 99) passing to the tongue.
The middle constrictor of the pharynx (2, Fig. 101 ; 9,
Figs. 98, 99) forms a band around the middle portion of
the pharynx from the points of insertion on the hyoid bone.
The inferior constrictor of the pharynx (3, Fig. 101) rises
from the larynx and passes around the lower portion of the
pharynx.
1 Passavant, Ueber die Verschliessung des Schlundes beim Sprechen, Frank-
furt a/M., 1863 ; Verschliessuhg d. Schlundes beim Sprechen, Arch. f. d. path.
Anat. u. Physiol. (Virchow), 1869 XLVI 1.
238 PRODUCTION OF SPEECH
Fig. 101 shows also the styloid process G, lower jaw R,
hyoid bone J, thyroid cartilage K, trachea M, oesophagus
L, and the following muscles : stylopharyngeus 5, stylohyoid
6, styloglossus 7, hyoglossus 8, mylohyoid 9, buccinator 13,
cricothyroid IJf.
The action of most of the organs of articulation can be
observed on a fluorescent screen when the side of the head is
illuminated by a very strong RoNTGBN apparatus.^ Owing
to the simultaneous demonstration of several organs and to
the freedom of the mouth from apparatus, this method will
probably be of great service. It requires a strong spark
coil (250™™ to 500™™ spark), an interrupter (motor, vibrator
with mercury contact, or Wehnelt's electrolytic break), a
RoNTGEN tube, a barium-platinum-cyanide screen, battery
or dynamo current, and connections.
References
For ao atomy : Testut, Anatomie humaine, 4"« ed., Paris, 1899;
Spalteholtz, Handatlas d. Anatomie d. Menschen, Leipzig, 1900
(Engl. ed. by Barker) ; Quain, Elements of anatomy, London, 1896;
Gray, Anatomy, Descriptive and Surgical, Philadelphia, 1901. For
Auzoux's separable models of man and of the larynx and tongue :
Montaudon, Paris. For separable paper charts : Witkowski, An-
atomie iconoclastique (Langue et larynx), Paris, 1879. For Rontgen
apparatus : Reiniger, Gebbeet & Schall, Eriangen.
^ ScHEiER, Die Yerwerthung d. Rontgenstrahlen f. d. Physiol, d. Sprache
u. Stimme, Arch. f. Laryngol., 1898 VII 116; Ueber d. Bedeutung d. Rontgen-
strahlenf. d. Physiol, d. Sprache u. Sfi'mme, Neuere Sprachen, 1897-98 V, Beiblatt,
40 ; Zur Anwendung d. Rontgenstrahlen f. d. Physiol, d. Gesanges, AUg. med.
Centralztg., 1898 No. 37 ; Zwaardemaker, Sur les sons dominantes des re'son-
nanles, Arch, ueerland. des sci., 1899 (2) II 241.
CHAPTER XVIII
STRUCTURE AND OBSERVATION OF THE LARYNX
The larynx (^, Fig. 93) is situated at the upper end of the
trachea (F, Fig. 93). It comprises a framework of cartilages,
partly united by joints and partly bound together by liga-
ments and membranes.
The cricoid cartilage (19, Fig. 93 ; M, Fig. 101 ; C, Figs.
104, 105, 107 to 111) is just above the top ring of the
trachea. It is a ring, narrower in front and much enlarged
in the rear.
The thyroid cartilage (7, Fig. 93; K, Fig. 101; T, Figs.
104 to 111) is a single large cartilage composed of two
Fig. 102. Fig. 103.
plates at the sides and front of the larynx. It rests upon the
cricoid at the cricothyroid joints in the rear. The thyroid
prominence (^', Fig. 107) may be felt on the front of the
neck (Adam's apple). The upper and lower horns of the
thyroid are indicated by b and h' in Figs. 107 and 109.
The two cartilages are shown in a front view in Fig. 102,
and in a side view in Fig. 103.
The arytenoid cartilages (A, Figs. 104, 105) are two
small pyramidal cartilages with triangular bases. They rest
240
PRODUCTION OF SPEECH
upon the upper edge of the posterior wide portion of the cricoid.
One projection of each arytenoid cartilage, the vocal process
(v, Fig. 106), carries one end of the vocal muscle ( V, Figs.
Fig. 104.
Fig. 105.
104, 105) ; the lateral projection is called the muscular pro-
cess (jn, Fig. 106). The upper corner carries a small pro-
jection called the corniculate (or Santorini) cartilage (6,
Figs. 109, 110).
The epiglottis (.£", Figs. 93, 109, 110, 111) is a thin, very flex-
ible, elastic cartilage just above the thyroid cartilage and just
behind the root of the tongue and the hyoid bone. The ten-
sion of the surrounding tissues keeps it erect.
The thyroarytenoid muscle (2!A, Fig. 106) comprises fibers
running from the arytenoid cartilage to the thyroid. It is
often treated as consisting of two parts.
The external thyroarytenoid comprises
the fibers lying nearer the thyroid carti-
lage ; the internal thyroarytenoid, or
'vocal muscle,' includes the fibers close
to the glottis. There is, however, no
distinct separation between the two
parts. The external fibers pull the
arytenoid cartilage forward as a whole. The internal fibers
pull the vocal process of the arytenoid cartilage directly
forward.
The glottis is the opening across the larynx (6r, Fig. 106).
The portion between the vocal muscles is called the ligamen-
tous glottis, that between the arytenoid cartilages the car-
tilaginous glottis.
Fig. 106.
STRUCTURE AND OBSERVATION OF THE LARYNX 241
The cricothyroid muscles (^OT, Figs. 107, 108) pull the
front portion of the cricoid cartilage upward around the cri-
cothyroid joints, whereby the upper rear portion of the cricoid
cartilage 0 with the arytenoid A moves backward and slightly
downward from the position indicated in Fig. 104 to that
in Fig. 105. The vocal muscle F'between the thyroid Tand
the arytenoid A is thus stretched and lengthened. Owing
to their slanting position, the cricothyroid muscles also com-
press the front of the thyroid cartilage and thus stretch the
Fig. 107.
Fig. 108.
vocal muscles by moving their front point of insertion forward.
The posterior cricoarytenoid muscle (CAP, Figs. 109, 110)
on each side pulls the arytenoid cartilage downward and
separates the vocal muscles.
Each lateral cricoarytenoid muscle (OAL, Fig. 110) runs
from the upper edge of the cricoid cartilage to the muscular
process and lateral edge of the arytenoid. It pulls the
muscular process of the arytenoid cartilage forward and
downward.
The tranverse arytenoid muscle (AA, Figs. 106, 109)
pulls together the two arytenoid cartilages. It moves the
16
242
PRODUCTION OF SPEECH
muscular processes upward and backward, whereby the vocal
processes are brought nearer together. Some of its fibers
run obliquely and are often represented as separate muscles,
the oblique arytenoids (^AAo, Figs. 109, 110).
The thyroepiglottic muscles (TE, Fig. 110) enlarge the
opening to the larynx. The aryepiglottic QAU, Fig. 110) and
the oblique arytenoid (^AAo, Figs. 109, 110) muscles narrow the
opening.
Fig. 109.
Fig. 110.
The thyrohyoid muscles (TR, Figs. 107, 108) from the
thyroid cartilage to the hyoid bone serve to raise the larynx.
The sternothyroid from the thyi-oid to the breast bone serves
to pull it down. The laryngeal portion of the stylopharyn-
geal (^10, Figs. 98, 99) muscle can also energetically raise the
larynx and tip it forward. The combined action of both sets
of muscles holds the larynx firmly in place. The larynx rises
and falls with different speech sounds, as can be felt when the
finger is placed on the projection in the neck and different
vowels are spoken or whispered.
Fig. Ill shows the front half of a longitudinal section of
the larynx, with the epiglottis E, the thyroid cartilage T^
STRUCTURE AND OBSERVATION OF THE LARYNX 243
the cricoid cartilage C, the thyroarytenoid muscles TA and the
ventricular bands VB. Fig. 112 gives an enlarged section
of one side of the larynx, showing the thyroid cartilage T,
the cricoid cartilage C, the external and internal thyroary-
tenoid muscles TAU and TAI, the vocal ligament VL, the
cricothyroid muscles CT and the ventricular band VB.
The ventricular band (^VB, Figs. Ill, 112) is a fold of
tissue extending from the thyroid cartilage in front just
above the vocal bands to the arytenoid cartilages. Muscular
fibers are found in these bands.^ The ven-
tricular band is often shorter and thicker than
that shown in the figure.
The laryngeal ventricles (ventricles of
MoEGAGNi) are small cavities, one on each
side, above the vocal bands and under the
ventricular bands. The opening between the
ventricular band and the vocal band is often
much greater than that shown in the figure.
The mucous membrane that lines the larynx
contains an unusual number of elastic fibers
that help to preserve its form amid the con-
stant stretching it undergoes. At the edges
of the thyroarytenoid muscles (F", Fig. 112)
the elastic tissue is thickened. The edges
are distinguished from the neighboring mucous covering by
a difference in the epithelial cells and by the lack of glands.
There is no special projecting ligament to form anything re-
sembling a cord or a membrane. The term ' vocal cord,'
' vocal band,' or ' vocal hp ' is sometimes applied to this edge,
sometimes to the entire muscle.
The mucous membrane lining the larynx is supplied with
glands except along the vocal ligament. These glands keep
the surface lubricated. The laryngeal ventricles, being
recesses or folds of the mucous membrane which have their
mouths wide below and directed downward toward the vocal
1 Steinlechner und Tittel, Der Musculus ventricularis d. Menschen, Sitzb.
d. k. Akad. Wiss. Wien, 1897 CVI 3. Abth. 157 (structure, earlier literature).
Fig. 111.
244
PllODUCTION OF SPEECH
u
%^^
cords, must empty tlie mucus which exudes from them upon
the upper surface of the cords. Somewhat of the shape of a
Phrygian cap with the apex above and wound around the
arj-tenoid cartilage, the ventricle presents an ample extent of
surface on which the mucous glands may open. Their
. , mouths may be seen extending in a few
cases to within a few millimeters of the
edges of the cords.
It is to be noted that many of the
muscles are more or less closely united
by common fibers. Thus, the thyroaryte-
noid and the lateral cricoarytenoid muscles
are, in man, parts of a general constrictor
muscle of the larynx, the cricothyroary-
tenoid ; again the transverse arytenoid is
the inner layer of an arytenoid muscle
whose superficial layers send off fibers to
several other muscles, particularly to the
aryepigiottic.
The muscular arrangement described
above is the typical one from which there
are great individual differences. These
individual differences probably have ef-
fects on vocal sounds, but nothing is
known in regard to the details.
The movement of the arytenoid carti-
lages may be resolved into three compo-
nents. The first is a rotation that brings
the vocal processes (p. 240) closer to-
gether and at the same time tenses the
cords. The second is a hinge-action that moves the vocal
processes upward-outward or downward-inward; the former
opens and raises the glottis, the latter closes and lowers it.
The third component is a sliding action sidewise whereby
the cartilages move toward each other.
The action of the muscles in producing the movements of
the ar^-tenoid cartilages may be indicated schematically by
Fir,
\\i.
STRUCTURE AND OBSERVATION OF THE LARYNX 245
means of Fig. 113. The position for respiration is shown at
A ; the muscles are all in a certain equilibrium of contraction.
The position at B shows increase in the rotation-component;
it might be due to stronger action of the lateral cricoary-
tenoid muscles, or to weaker action of the interarytenoid ; it
is a position used in light whispering. Closure of the liga-
mentous glottis with the cartilaginous glottis open is shown
at C; the rotation component is greater than at^; each car-
tilage has slid toward the other ; the action of the posterior
and lateral cricoarytenoids must be relatively larger or that
of the interarytenoid smaller. The entire glottis is closed at
D; it involves chiefly the action of the lateral cricoarytenoids
A B CD
Fig. 113.
and the interarytenoid with relatively little action of the pos-
terior cricoarytenoids.
Tensing of the cords by the cricothyroid would tend to
close the glottis. Contraction of the external portions of the
thyroarytenoid would relax the cords and open the glottis;
just what the contraction of the internal portion would do
can hardly be said.
The muscles in their action on the vocal bands and glottis
may be classed as 1. tensors: cricothyroid and transverse
arytenoid; 2. relaxers: thyroarytenoid; 3. closers (adduc-
tors) : thyroarytenoid, lateral cricoarytenoid, transverse ary-
tenoid ; 4. openers (abductors) : posterior cricoarytenoid.
It is to be noted that all tensors have also a closing action
and that all openers have also a tensing effect.
246 PRODUCTION OF SPEECH
The slant of the glottis is regulated by the interaction
of the thyroarytenoids and the lateral cricoarytenoids, the
former lowering the rear end, the latter raising it. The
effect of the change in slant is probably to alter the direction
of the air-blast; the higher position gives a more gradual
convergence of the walls below the glottis while the lower
position makes the convergence more sudden. The effect
on the character of the tone is not known.
In swallowing, the larynx is raised, the tongue is moved
backward and the epiglottis is depressed; in this way the
ventricular bands in the larynx are brought against the lower
part of the epiglottis and the opening is tightly closed. The
glottis is also closed. . The epiglottis has presumably an
important function in adjusting the tone of the laryngeal
cavity immediately above the vocal bands and in modifying
the tone of the pharynx. The rise and fall of the larj'nx
are due to the action of the thyrohyoid and stylopharyngeal
muscles.
The cricothyroid muscle is governed by the superior
laryngeal branch of the vagus nerve, the other muscles by
the inferior laryngeal branch. The movements of the mus-
cles are governed by sub-cerebral and cerebral centers. The
former originate the movements connected with organic
life, such as breathing, coughing, crying from pain, laugh-
ing, and various forms of intonation connected directly with
the emotions; the latter originate the specific movements
required in speech. For breathing the sub-cerebral center
lies in the bulb (5, Fig. 70). For intonation the sub-cere-
bral center seems also to be in the bulb. Cerebral centers
for closing and opening the glottis have been found in the
cortex, both sides of the larynx being actuated by stimula-
tion of either side of the brain. In each cerebral hem-
isphere there is a bilateral center for the adduction of the
vocal cords ; ^ a cortical center of abduction has not been
' Krause, Ueber d. Beziehungen d. Grosshirnrinde zu KeMkopf u. Rachen, Arch.
f. Anat. u. Physiol. (Physiol. Abth.), 1884 ; Semon and Horslet, An experimen-
tal investigation of the central motor innervation of the larynx, Phil. Trans. Eoy. Soc
Lend., 1890 CLXXXI (B) 187.
HTUUCTURE AND OBSEllVATION OF THE LARYXX 247
found ; separate cortical centers for the separate sides of
the larynx do not exist.
The opening of tlie glottis and the tension of the vocal
cords involve complicated muscular adjustments that have
become fairly well known only since the invention of the
laryngoscope.
The action of the vocal bands can be directly observed
by means of the laryngoscope. Although Sbnn, Bablngton
and CHAEEii;RE had made
mirrors for the purpose of
looking down the throat,
Gaucia, who liad bought
Chaeiueee's mirror, was
the first to successfully use
one.i The development of
the method was mainly due
to t'ZEEMAK.2
Tlie source of illumin-
ation must be strong and
concentrated. The most
convenient form is that of
a small electric lamp fast-
ened to the forehead of the
observer. The arrangement
shown in Fig. 114 has been
found adapted to the use
of students. The electric
lamp in the holder L can be
turned in any direction by the ball-joint J; its light is
focused by an adjustable lens. It may be connected to a
storage battery or through a suitable resistance to the city
wires. The apparatus shown in Fig. 114 comprises a lamp A
^ Rapport snr ' Garcia, M^moire sur la voix humaine' C. r. Acad. Sci. Paris,
1841 XII 638; Observations on the human voice, Philog. Mag, 1855 5 218;
Richard, Notice snr Finvention du Laryngoscope, Paris, 1861 ; Frankel, Unter-
suchtinqsmetkoden d. Kehlkopfes u. d. Luftrohre, Hej'mann's Handb. d. Laryng. u
Rhin., I 229, Wien, 1898,
- CzERMAK, Der Kehlkopfspiegel, Leipzig, 1860-1863.
Fig. 114.
248
PRODUCTION OF SPEECH
of about 100 ohms resistance on a llO-volt circuit (a 32-
candle-power lamp) ; the current of 1 ampere passes through
this and a small adjustable wire resistance R to the laryngo-
scope lamp. The degree of brightness in the latter is regu-
lated by adjusting the resistance. An extra lamp is shown
at 0. For observing others the spring B passes over the
head of the observer. For self-observation the spring is fixed
by an adjustable clamp D to the rod ^, and an adjustable
mirror F is attached to the same rod.
Oil and gas lamps with condensing lenses may also be
used. Physicians frequently employ an Argand burner with
a reflector fastened before the
eye which looks through a small
hole in the center.
A larjrngeal mirror (^M, Fig.
114) of the largest size that the
subject can conveniently bear is
sterilized by carefully cleaning
it with a brush in a hot solution
of sodium carbonate and by rin-
sing it before use in a 5% solu-
tion of carbolic acid. Persons
with tuberculosis or other infec-
tious diseases should have their
own mirrors. The mirror should
be warmed to about body temperature (tested by the hand)
against the large incandescent lamp or over a flame.
The subject should open the mouth as widely as possible,
the lips exposing the teeth. The tongue is stuck out (not
pulled out) and the point is wrapped in a cloth and held
down by the thumb and finger (Fig. 115). The subject must
continue his respiration quietly and without stopping, other-
wise the contact of the mirror is liable to cause retching;
he is to repeat whatever sound he hears.
The handle of the mirror is held like a pencil. "While
the vowel e is being sung, the mirror is inserted quickly
and evenly so that the uvula rests upon the back of the
Fig. 115.
STRUCTURE AND OBSERVATION OF THE LARYNX 249
mirror ; the handle is kept at the corner of the mouth.
A slight further movement and turning of the mirror suffice
to send the light down the larynx and to reflect its picture
to the eye. Considerable practice is required to attain a
ready and complete control of the adjustments. In the
laryngoscope mirror the picture appears reversed, the epi-
glottis being at the top. The Santorini cartilages appear
in the view; their movements serve to indicate the tilting
of the arytenoid cartilages.
Measurements can be made with considerable accuracy by
a graduated mirror or by glancing at a millimeter scale and
then estimating the distance in the laryngoscopic picture. An
instrument for directly measuring the width of the glottis
has been devised by Exnee.i
The character of the vibration of the vocal bands may
be observed by the stroboscopic method, which has been
rendered easy of use.'^
With a sufficiently bright light and a camera arrangement
the laryngeal view may be photographed.^ With a stereoscopic
1 ExNER, Das Laryngomeler, Zt. f. Instrumentenkunde, 1897 XVII 371.
2 Oertel, Ueber eine neue laryngostroboskopische Untersuchungsmethode, Cen-
tralbl. f. (1. med. Wiss., 1878 XVI 81 ; Laryngostroboskopische Beobachiungen iiber
die Bildung der Register bei d. menschl. Stimme, Centralbl. f. d. med. Wiss., 1878
XVI 99 ; Ueber d. Mechanismus d. Brust- und Falsettregist., Beitrage z. Biol., 25,
Stuttgart, 1882; Das Laryngostroboskop und seine Verwendung in d. Pkysik,
Physiol, u. Med., Archiv f. Laryngologie, III (also separate, Berlin, 1895);
Ueber d. Laryngostroboskop, Verli. d. Congr. f. inn. Med., 1895 3.31 ; Rethi,
Exper. Unlersuch. ilb. d. Schwingungstypus n. d. Mechanismus d. Stimmbander bei
d. Fa/settstimme, Sitzb. d. k. Akad. Wiss. Wien, math.-naturw. Kl., 1896 CV 3.
Abth. 197 ; Untersuch. ub. d. Schwingungsform d. Stimmbander bei d. verschied.
Ge.iangsregistern, .same, 1897 CVI 3. Abtli. 66; Mbsehold, Stroboskopische u.
photogr. Studien lib. d. Stellung d. Stimmlippen im Brust- und Falsettregister,
Arcli. f. Laryngol., 1898 VII 1; Spiess, Ein neues Laryngostroboskop, Arch. f.
Laryngol., 1898 VII 148.
8 French, On a perfected method of photographing the larynx, N. Y. Med. Jour.,
1884, Dec. 13; Lari/ngeal and postnasal photography with the aid of the arc light,
N. Y. Med. Jour., 1897, Jan. 23. (Accounts of earlier attempts and later metliods
bv KiLLiAN, Miinch. med. Wocheuschr., 1893, Feb. 7 ; Wagner, D. Photographic
d. Kehlkopfes, Heymann's Handb. d. Laryngol. n. Rhinol., I 1512, Wien, 1898;
Garel, Ann. des Mai. de I'OreiUe, 1899 XXV (2) 702.)
250 PRODUCTION OF SPEECH
camera ^ double views may be obtained that appear in relief
through a stereoscope.
The combination of a stereoscopic shutter and a photo-
graphic arrangement with a moving sensitive surface renders
it possible to get pictures of the cords in typical stages of
vibration.^
References
For the structure and action of the larynx : Spalteholz, Handatlas
d. Anat. d. Menschen, Leipzig, 1900 ; Engl, trans, by Barker ; Het-
MANN, Handb. d. Laryngol. u. Rhin., I, Wien, 1898 ; Grijtzner,
Physiologic d. Stimme u. Sprache, Hermann's Handbuch d. Physiol., I 2,
Leipzig, 1879 ; Hermann, Lehrbuch d. Physiologie, 338, 12. Aufl. Berlin,
1900. For the nerve centers and connections with the larynx : Semon,
Die Nervenkrankheiten des Kehlkopfes und der Luftrohre, Heymann's
Handb. d. Laryngol. u. Rhin., 1 587, Wien, 1898. For methods of laryn-
goscopy : Frankel, Untersuchungsmethoden d. Kehlkopfes u. d. Luftrohre,
Heymann's Handbuch d. Laryngologie u. Rhinologie, I 227, Wien, 1898.
For Auzoux's separable models of the larynx: Mo^taudon, Paris.
For separable paper charts : Witkowski, Anatomie Iconoclastique
(Langue et larynx), Paris, 1879. For apparatus for demonstrating
the action of the laryngeal muscles : Oertel, Ueber d. laryngolog.
Unterricht, 5, MUnchen ; Wagner, Schema d. hypokinetischen Mobi-
litatsneurosen d. Kehlkopfes z. laryngolog. Unterricht. For laryngo-
scopes : Kny-Scheerer Co., New York ; Sydow, Berlin. For Spiess's
laryngostroboscope : Reiniger, Gebbert & Schall, Erlangen.
1 Garel, La photographic stgrgoscopique du larynx, Ann. des Mai. de I'Oreille,
1899 XXV (2) 702.
2 SsiSMANOwsKY, Die Anwendung d. Photographic bei Untersuchung d. Stimm-
banderschwingungen. Arch. f. d. ges. Physiol. (Pfliiger), 1885 XXXVII 375;
MuSEHOLD, Stroboskopische «. photogr. Studien iib. d. Stellung d. Stimmlippen im
Brunt- u. Falsettregister, Arch. f. Laryngol., 1898 VII 1.
CHAPTER XIX
ACTION OF THE LAEYNX
A VIEW of the larynx from above during quiet breathing is
shown in the photograph (Fig. 116) made by French. The
epiglottis is seen as a fold curled upward at the top of
the picture. The arytenoid cartilages are rotated so that
their vocal processes are turned outward, thereby separating
the rear ends of the vocal cords. The whole glottis is thus
widely opened; the trachea can be seen below.
Speech sounds pro-
duced with this adjust-
ment of the larynx get
their acoustic character
entirely from the vocal
cavities and not at all
from the glottis ; they are
said to be ' surd ' or 'un-
voiced. '
Vocal sounds produced
when the cords are vi-
brating are said to be ' sung ' or ' intoned ' when they occur
in song, ' sonant ' or ' voiced ' when in speech.
In intonation the vocal cords are brought together. The
blast of air from the trachea sets them in vibration. The
rapidity of the vibration depends on the length, tension and
loading of the muscles constituting the vocal bands.
From photographs of the vocal bands during intonation
it is possible to make some deductions concerning the laryn-
geal action.
Fig. 116.
252
PRODUCTION OF SPEECH
im
Fig. 117.
The series of photographs shown in Figs. 117 to 121 were
taken by Feench^ from the larynx of a professional con-
tralto. The picture for her lowest note, Fig. 117, shows the
vocal bands quite short and wide, with the ligamentous glottis
(p. 240) open along three fourths of
its length and the cartilaginous glot-
tis partly open, the ligamentous glottis
being linear in shape. As the voice
ascended the scale the vocal bands
increased in length and decreased in
width, imtil at e^ the condition was
as shown in Fig. 118 : here the carti-
laginous glottis has opened further,
the vocal bands had increased in
length at least 3™™ in the seven notes,
and had become narrower. The rise
in pitch seems to have been brought
about by stretching the bands. With a further rise in pitch,
however, the condition suddenly changed. In the transition
from e^ to /^ (Fig. 119) the bands were shortened about 2°"°,
the cartilaginous glottis was closed and the ligamentous glot-
tis narrowed; the entire cavity of the
larynx was reduced and the epiglottis
lowered. The bands appeared to be
more tightly stretched, and to be ca-
pable of vibrating only along the liga-
mentous glottis. The change in action
was accompanied by a marked change
in the acoustic quality of the tone pro-
duced. Such a change in the manner
of producing tones is said to be a
' change of register,' a ' register ' in-
cluding all tones produced in the
same manner. As the scale was followed upward, the
vocal bands steadily increased in length and the ligamen-
tous glottis gradually opened. The condition at d? is shown
in Fig. 120 ; the bands had increased in length several mil-
' French, The action of the glottis in singing, N. Y. Med. Jour., 1891 Jan. 31.
$
T
Fig. 118.
ACTION OF THE LARYNX
253
limeters and the cartilaginous glottis was somewhat opened.
These facts apparently indicated a decrease in tension of the
bands. The epiglottis had risen slightly. As the voice passed
from d^ to e^ a distinct change in quality was again heard and
the laryngeal conditions were seen
to become readjusted (Fig. 121).
The vocal bands seemed slightly
shorter than before. There was
closure of the cartilaginous glottis
and a small portion of each end of
the ligamentous glottis. The epi-
glottis was depressed. There was
again a change of register. This
was the highest note that could be
sung with ease; the action in as-
cending this register could not be
studied.
Other carefully made photo-
graphs by Feench show both similarities and disagree-
ments. They seem to establish the following conclusions
regarding the action of the female
larynx.
Fig. us.
m.
f.1
'The larynx may act in a variety of
ways in the production of the same
tones or registers in different individ-
uals. The rule — which, liowever, has
many exceptions — is that the vocal
bands are short and wide and the
ligamentous and cartilaginous por-
tions of the glottis are open in the
production of the lower tones ; that,
as the voice ascends the scale, the
vocal bands increase in length and de-
crease in width, the aperture between
the posterior portions of the vocal bands increases in size, the
Santobini cartilages are tilted more and more forward, and the
epiglottis rises until a note in the neighborhood of e^ treble clef,
iirst line, is reached. The cartilaginous glottis is then closed.
Fig. 120.
m^
254 PRODUCTION OF SPEECH
The glottic chink becomes much narrower and linear in shape,
the capitula Santokini are tilted backward, and the epiglottis is
depressed.
' When the vocal bands are shortened in the change at the
lower break in the voice, it is mainly due to closure of the carti-
laginous portion of the glottis, the ligamentous portion not
usually being affected. If, therefore, the cartilaginous glottis is
not closed, there is usually no material change in the length of
the vocal bands.
' As the voice ascends from the lower break, the vocal bands
increase in length and diminish in width, the posterior portion
of the glottic chink opens more and more, the capitula Santorini
are tilted forward, and the epiglottis rises until, in the neighbor-
hood of e^, treble clef, fourth space,
another change occurs.
' The glottic chink is then reduced to
a very narrow slit, in some subjects ex-
tending the whole length of the glottis,
in others, closing in front or behind, or
both. Not only is the cartilaginous glot-
tis always closed, but the ligamentous
glottis is, I believe, invariably short-
ened. The arytenoid cartilages are til-
ted backward and the epiglottis is
depressed. As the voice ascends in the
J, ,., head register the cavity of the larynx
is reduced in size, the arytenoid carti-
lages are tilted forward and brought closer together, the epi-
glottis is depressed, and the vocal bands decreased in length and
breadth. If the posterior part of the ligamentous portion of the
glottis is not closed in the lower, it is likely to be in the upper
notes of the head register.' ^
Concerning the muscular action whereby the changes in
pitch are produced, we can hardly say more than that pre-
sumably they are mainly brought about by the interaction of
the cricothyroid and the thyroarytenoid muscles. The former
stretches the bands (p. 241). The latter can make them
more tense by contraction of its fibers (p. 240), whereby the
pitch would be raised, or it can shift the weight toward the
' French, as before.
■ACTION OF THE LARYNX 255
middle, whereby the pitch would be lowered. Changes in
load, in tension and in length of the vibrating portion can be
produced by the mUscles attached to the arytenoid cartilages
(p. 239). The intimate connections of all these muscles with
the other muscles of the larynx (p. 244) would suggest gen-
eral changes in any change of register.
The rise in pitch through several octaves evidently cannot
be produced by continuous stretching of the bands. For a
rise of seven tones the bands shown in Figs. 117 and 118 were
stretched about 3°"°. As this voice had a compass of about
2J octaves, a total lengthening of 10°"° to 12™°" would be
needed at the same rate. The impossibility of such a stretch-
ing indicates the necessity of a change of method in pro-
ducing the tones.
It was asserted by Galen that the edges of the glottis
were an essential factor in the production of voice ; ^ he
seems to have supposed them to act by forcing the breath
to form a jet whereby the air in the vocal cavity was aroused
to vibration as in a labial pipe. Dodaut recognized the fact
that the pitch of the voice depends on the tension of the vocal
bands. ^ According to Feerein the width of the glottis
has no influence on the pitch of the tone, and the bands
are essentially strings set in vibration by the air as the string
of a violin is aroused by a bow.^ He was the first to make
experiments on membrane strips stretched over a tube and
on the larynx removed from the human body.
Three views have been held concerning the manner in
which the vocal bands vibrate.
For some purposes a vocal band maj"^ be considered as
a string stretched between two points with its entire mass
concentrated in a material point at the center ; the laws
of its vibration can then be deduced mathematically.'' For
1 Galen, De usu partium, Lib. VI, cap. 2.
2 DoDAET, Mgm. sur Us causes de la voix, etc., Mem. de I'Acad. des Sc. de Paris,
1700, 1706, 1707.
3 Fekreis, De la formation de la voix de I'homme, M^m. de I'Acad. des Sc. de
Pari.s, 1741.
* Ratleigh, Theory of Sound, 2. ed., I § 52, London, 1894.
256 PRODUCTION OF SPEECH
vibrations so small that the additional stretching during
the elongation is negligible, we have
T= 2n/—
V 2s
where T is the period of vibration, s the tension, a the length
and m the mass. Thus the period increases with an increase
of length, an increase of mass, or a decrease of tension.
This last statement, more general than the equation, is valid
for the slowest mode of vibration of any stretched string ; the
fall in pitch with increase of length, increase of mass and de-
crease of tension can be readily illustrated on any stringed'
instrument.
For most purposes, however, we must consider the distri-
bution of the mass of the string along its length. The ideal
musical string is ' a perfectly uniform and flexible filament
of solid matter stretched between two ilxed points.' The
strings of most musical instruments approach the ideal closely.
The free vibrations of a string consist of the sum of a
series of harmonics whose periods are the natural periods of
the whole string, a half of the string, a third of the string, etc.
This can be easily shown by setting a violin string in vibra-
tion and then touching it exactly in the middle; the lower
tone ceases but tlie tone of half the string continues to be
heard. The higher tones may be similarly found. Such a
series of tones is often called a note ; the component tones are
called pa7'tials. The lowest partial is called the first partial
or fundamental. The higher partials are called second, third,
. . . loartials ot first, second, . . . overtones. For example, the
note of 100 vibrations a second from a piano string would
consist of the partials 100, 200, 300, 400, 500, 600, ... in
various intensities ; the first partial, 100, would be the funda-
mental and the others would be overtones (p. 72).
The vibrations of strings maintained by a blast of air as in
an .^olian harp are in a plane transverse to the direction of
the wind. ^ In the vocal bands such an action would tend to
compression sidewise, that is, to a cushion action.
1 Rayleigh, Acoustical observations, Phil. Mag., 1879 (5) VII 161.
ACTION OF THE LARYNX 257
The vocal bands are sometimes treated as membranes
stretched across a channel. J. Muller ' made experiments
on membranous lips, whereby he showed that the width of
the opening had little influence on the period, that the
period depended chiefly on the tension and on the damp-
ing and that small membranes could be made to give very
deep and powerful tones. He treated the larynx as a pipe
with membranous reeds whose periods followed the rules for
strings. That membranes can be made to give deep tones
only when stretched on soft supports that also vibrate,
that the cords can be treated as ordinary membranes only in
falsetto tones, and that in falsetto tones they have nodal
points — these are conclusions drawn by C. Mtjllbr.^
Membrane pipes have been made in various ways and have
been studied to some extent.^ They form convenient instru-
ments for illustrating the effect of tension on the pitch of the
membrane but are decidedly liable to mislead in implying
that the vocal bands vibrate like membranes and that the
tension is obtained wholly by bringing the points of support
further apart. A convenient model is that of Czekmak*
or that of LuDWiG.^ The simple form shown in Pig. 122 will
serve every purpose. Two opposite points of the thin-walled
rubber tube are each caught between the thumb and finger ;
the tube is then stretched till the sides come together. A
blast of air through the tube sets the edges in vibration. The
period of this vibration depends on the tension, which can be
regulated by the fingers.
The vocal bands may also be treated as elastic cushions
that yield by compression. The vibrations of cushions have,
as far as I am aware, received no extended treatment. The
1 MiJLLEE, Handbnch d. Physiol, d. Menschen, II 179, Coblenz, 1840.
'^ MiJLLEE, Untersuchungen iiher einseitig frei schwingende Membranen und deren
Beziehung sum menschlichen Stimmorgan, Schr. d. Ges. zur Beford. d. ges. Natur-
wiss. i.. Marburg, 1877 II 102, 166, 167.
' Hubert, Sur le mode de vibration des membranes, et le rdle du muscle thri/o-
aryUnoidien, C. r. Acad. Sci., Paris, 1891 CXII 715;
« CzEKMAK, Gesammelte Schriften, II 71, "Wien, 1879.
6 Ludwig's kiinstliche Kehlkopfe, in Petzold's Katalog.
17
258 PRODUCTION OF SPEECH
possibility of a cushion action of the vocal bands was sug-
gested by EwALD 1 and MtrsBHOLD.^ The blast of air pushes
the edges of the cushions to one side either through compres-
sion of the projecting cushions themselves or through yield-
ing of the walls. ' This view is favored by the fact that the
vocal bands are not of a nature and shape to readily vibrate
transversely. The true shape is given in Fig. 112; the
usual diagrams in works outside those specially pertaining to
laryngology give a quite erroneous idea of them. The vocal
bands suggest a pair of cushions suitable for compression,
Fig. 122.
and not a pair of membranes. When the bands are closed by
the action of the cartilages, the air is retained behind them
until the pressure is great enough to force them open, the
pressure being regulated by the tension of the vocal muscles
constituting the bands. When they have been forced apart
to emit the puff of air, they close again and remain closed
until the pressure is again sufficient to open them. ' ^
The character of the vibrations of the vocal bands seems
clearly indicated by the following additional facts.
1 EwALB, Physiologie des Kehlkopfes, Heymann's Handbuch d. Laryn". u.
Khin., I 181, Wien, 1898.
2 Mdsehold, Stroboskopische u. photogr. Studien iih. d. Stdlung d. Stimmlippen
im Brust- u. Falsettregister, Arch. f. Laryngol., 1898 VII 1.
^ ScEiPTUBE, On the nature of vowels, Amer. Jour. Sci., 1901 XI 309.
ACTION OF THE LARYNX 259
Observations ^ of the vocal cords of men singing in the
chest register showed that the cords touch along their
M^hole length ; that in loud tones they have a slightly-
rounded form, especially in the middle, indicating strong con-
tact in the middle with lighter contact at the ends ; that in
weaker tones the line of contact appears even and thin while
the top of the cords becomes flatter. These observations
made it clear that a vibration of the cords in the axial direc-
tion of the larynx did not occur, and established the fact of
cushion action. Observations on the head register showed
that the vocal cords did not touch but were separated by a
more or less weakly elliptical sht ; the exact method of vibra-
tion was not established.
Fig. 12.3.
Observations ^ by the stroboscopic method showed that in
singing in the chest register the glottis opens to a spindle-
like slit and then closes completely along its whole length ;
that the cords move sidewise, that is, outwards, and not cross-
wise in the direction of the axis of the larynx. Observations
on the head register showed that the edges of the cords did not
touch, but did not afford a decision concerning the direction
of their vibration. Fig. 123 A (Mtjsbhold) shows the edges
of the bands just touching (dotted line) at rest, and tightly
pressed together as in singing in the chest register (full line).
Fig. 123 B (Musehold) shows the directions in which the
bands yield to the pressure of air from the trachea.
1 Musehold, as before, 8.
'^ Musehold, as before, 12, 16.
260 PRODUCTION OF SPEECH
From the foregoing observations we may draw fairly relia-
ble conclusions concerning the manner in which the vocal
bands execute their vibrations.
In regard to the chest register the theory stated by
Gakcia 1 seems established. ' The vocal cords close the pas-
sage for the air and offer resistance to it. As soon as the air
attains a sufficient pressure, it separates the cords and pro-
duces an explosion, but at the same moment they come
together again on account of their elasticity and because the
pressure below has ceased, ready for a new explosion. A
series of these compressions and expansions, or explosions,
caused by the pressure of the accumulated air and the reac-
tion of the glottis, produces the voice. ... It is not neces-
sary for the glottis to close again completely after each
opening in order to produce the explosion; it is sufficient
that it should present a resistance to' the air sufficient to
develop its elasticity.' The action of the lips in blowing a
trumpet, as recently established by the stroboscopic method, ^
is closely like that of the cushion action of the cords.
In the chest register the vocal bands probably always
vibrate throughout their entire thickness, and never along the
edges only.^
In regard to the head register it seems probable that the
thyroarytenoids do not contract — at least strongly ; * that
the change in the tension of the cords is produced by stretch-
ing ; that the cords have somewhat sharper edges than in the
chest register; that the direction of vibration may deviate
more or less from the transversal direction. The action may
be similar to that of a stretched string with a mass of soft
material attached to it, or to that of a cushion with the main
tension along its edge. It is possible that the edge of the
band moves neither axially nor transversely and not even in
a straight line but in a more or less complicated curve.
1 Gaecia, Beobachtungen ilb d. menschl. Stimme, Monatsschr. f. Ohreuheilk.,
1878.
^ MnsEHOLD, as before, 19.
2 MusEHOLD, as before, 18. * Ewald, as before, 200.
ACTION OF THE LARYNX 261
In regard to the course of a vibration executed by the
bands no reliable data are at hand except those obtainable
from speech curves (Part I). These indicate regularly for
the chest register more or less sudden movements separated
by intervals of rest (p. 39). Concerning the head register
we have no published information.
Different adjustments of the weight of the muscle sub-
stance within the vocal bands would produce differences in
the character of the vibration and consequently puffs of
different forms (p. 96). These differences in adjustment may
be produced by differences in the groups of fibers contracted.
The character of the voice in singing or speaking doubtless
arises largely from these differences in the action of the bands.
The effect of different loads on the action of a vibrating string
may be readily illustrated. A string — for example, a violin
string — is stretched between two supports. By loading the
string with little blocks of paper or by pressing cotton wads
against it at different points the character of the tone may
be made to vary as a result of the modification of the over-
tones. The effect in vibrating cushions is not so readily
demonstrated.
It has been suggested (Swain) that the ventricular bands
may possibly descend and touch the top surfaces of the vocal
bands during intonation and thus modify the character of
the vibrations by acting as dampers, loads or nodal supports.
According to St6rk,i when the larynx is strongly illu-
minated below the cords by a light through the neck and
is observed with a laryngoscope, the light is seen in increas-
ing brightness through the cords as the pitch rises until in
the head register there seems to be only a thin membrane
in front of it. This shows merely that in high notes the time
of closure diminishes in comparison with the time of opening.
That the vocal muscle possesses the ability to contract
differently in its different parts is shown by an experiment
described by Ewald.^ Across the end of a tube two frog-
1 Stork, Klinik der Kehkopfkrankheiten, Stuttgart, 1876.
2 EwALD, Phi/siologie des KeUkopfes, Heymann's Handbuch d. Laryngologie
n. Rhin., I 202, Wien, 1898.
262 PRODUCTION OF SPEECH
muscles were placed to form an artificial larynx. The muscles
were stretched by causing them to contract by means of elec-
tric currents. A note was produced when a blast of air was
driven through the apparatus. By shifting the electrodes the
inner or the outer portions of the muscles could be made to
contract separately. Different distributions of the weight
in the vibrating muscles produced changes in pitch. In this
way considerable changes could be produced while the total
contractile force remained constant. This experiment of
Ewald's does not prove that the tension of the cords remains
constant and that the changes in pitch are entirely produced
by different degrees and distributions of the contraction. It is
much more probable that the differences in distribution aid in
great changes in pitch, but that the finer adjustments are de-
rived from differences in the tension of a portion already con-
tracted ; this would involve combined changes in the tensions
of the thyroarytenoid and the cricothyroid muscles.
According to Oebtel^ and Koschlakofp^ the vocal bands
in the head register vibrate in two longitudinal segments with
nodal lines running along not far from their edges. The nar-
row strip between the nodal line and the edge makes extensive
movements while the portion beyond the nodal line makes only
small ones. It may be suggested that the greater weight of
the outer portion would make its smaller vibrations equivalent
in energy to the more extended ones of the inner portion.
According to Rethi's observations^ on the head register
by means of the stroboscopic method, the edge of the vocal
cord rises for a vibration and then falls while a ridge-like
wave passes from the edge along the top surface outward, no
nodal lines being seen ; the phenomena were explained by the
contraction of the internal portion of the thyroarytenoid
1 Oeptel, as on p. 249.
'^ KosoHi/AKOFF, XJeber d. Schwinrjungstijpen d. Stimmbdnder, Arch. f. d. ges.
Physiol. (Pfliiger), 1886 XXXVIII 473.
' Rethi, Exper. Untersuch. iib. d. Schwingunr/stypus u. d. Mechanisnius d.
Stimmbdnder bei d. Falxettsdmme, Sitzb. d. k. Akad. Wiss. Wieu, math.-naturw.
Kl., 1896 CV 3. Abth. 197.
ACTION OF THE LARYNX 263
muscle while the external portion was relaxed. In the middle
register the action was similar to that in the head register.^
Studies of vowels sung and spoken in the chest register
seem to require the assumption of cushion action of the
bands, as indicated by the following facts. The movement
imparted to the air hj a freely vibrating membrane is neces-
sarily of the nature of a sinusoid (p. 2) or a sum of har-
monic sinusoids (p. 13). That the movement is not of such
a nature in sung and spoken vowels of the chest register has
been proven by recorded speech curves (p. 41). The vibra-
tions of cushions may be of any degree of sharpness or smooth-
ness from a practically instantaneous explosion to a move-
ment as regular as that of a fork. Such vibrations appear
in the various speech curves. Vibrations of a sinusoid char
acter can arouse only harmonic resonance vibrations, whereas
explosive vibrations require no such adjustment of the reso-
nating cavity. The evidence is conclusive (pp. 21, 39) that
in sung and spoken vowels a harmonic relation between the
resonance tones and the cord tone is not necessary. In con-
sidering the tone aroused by the cords it seems necessary to
treat it not as a note composed of a series of partials (p. 90)
but as a series of pufEs. These facts are conclusive in regard
to the cushion action in the chest register. Similar data for
the head register are not at hand.
When the vocal cords close to obstruct the air passage,
the greater the breath pressure the more energetic must be
the muscular action in order to maintain the closure. Even
when the cords are vibrating the tension and firmness of
closure must increase as the pressure increases in order to
prevent the cords from simply being forced apart and produc-
ing a breathy tone. That this is actually the case has been
shown by Mtjsehold (p. 259). According to Muller ^ an
1 R^THI, Unlersuch. ub. d. Schwingungsform d Stimmbander bei d. verschied.
Registern, Sitzb. d. k. Akad. Wiss. Wien, math.-naturw. Kl., 1897 CVI 3. Abth.
68.
2 MuLLEK, XJeber d. Compensation d. physischen Krafte am menschl. Stimm-
organ, 1839.
264 PRODUCTION OF SPEECH
increase in the pressure of the air produces an increased
' passive ' tension of the vocal membranes and consequently
a higher note ; this can be readily demonstrated by the instru-
ment shown in Fig. 122. To maintaiu a sound on the same
note this passive tension must be compensated by a relaxation
of the active tension. This theory would be apparently
incontestable if the cords were membranes, as Mullek sup-
posed. It has been shown to be inadequate by the observa-
tions of MusEHOLD ; the cords do not relax but contract more
firmly as the breath pressure rises. The maintenance of a
constant period of vibration in spite of the firmer closure may,
I suggest, be due to a redistribution of the contraction in the
different fibers, the fibers along the edge contracting more
strongly while those in the interior relax and act not only
to diminish the total tension of the cords but also as loads
to lengthen the period.
When the same note is sung by different persons or by the
same person in, different conditions, the ear will readily detect
differences in the character of the sound. Persons can be
distinguished by their voices when speaking and singing.
There are undoubtedly individual differences in the action
of the vocal bands and in the adjustments of the vocal cavity.
The changes in the character of a note coming directly from
the glottis are brought about by differences in the structure
and action of the larynx. On the theory that the cords exe-
cute vibratory movements like most musical instruments these
differences arise from the differences in the strengths of the
partial tones (p. 95). On the theory that the cords produce
puffs like a siren they arise from the shape of the puff
(p. 96).
The character of the voice depends also on the condition of
lubrifaction of the larynx. It must necessarily be the case
that a wet wall of whatever resisting quality must produce
sound waves of a different nature from a dry or nearly dry
surface, and this must especially apply to the vocal cords.
Any influence, whether physiological or pathological, which
tends to modify the amount and the consistency of the lubri-
ACTION OF THE LARYNX 265
facient mucus, must also influence the vocal tone. A large
part of the modification of the voice in attacks of laryngitis
must be due to this change in the mucus.
Changes in the tones from the cords are also brought about
by adjustments of the sizes, necks and apertures of the series of
cavities above them ; these are brought about by rise and fall
of the larynx, by changes in the position of the epiglottis,
tongue, velum, jaw, lips, by changes in the tension in the
walls of the cavities, etc. It is doubtful for the human sub-
ject, if the laryngeal ventricle, even during very strenuous
phonation, is dilated sufficiently to greatly modify the quality,
of the tone. We know that this occurs in some animals in
which these ventricles seem to serve the purpose of resonators.
The ventricular bands may perhaps in some way affect the
character of the voice tone.
Voices may be classed roughly as soft and sharp. The soft
voice has a character resembling that of a flute or a tuning
fork, that is, a tone with mainly low partials. The soft
voice is most readily produced when the head is slightly
inclined forward and the larynx lowered. The length of
the mouth cavity favors low tones and the softness of the
walls would hinder the development of high tones. The lips
are generally held rather close, to hinder the exit of high
tones. The tongue is drawn back and the soft palate raised
to close the nasal opening. On the vibration theory the bands
themselves may be supposed to swing in a way to develop
only low overtones. Geutzner^ supposes them to vibrate
without touching at the edges. It may be suggested that the
adjustment of the muscular load within the bands might be
supposed to dampen the higher overtones (p. 295). On the
explosion theory the puffs are of smooth shape ; they might
even approximate the sinusoid form (p. 2). In the production
of the sharp voice the larynx is high, the head is generally
tipped back, the muscular adjustments are firm, the mouth
1 Merkel, Die Fuuctioneii d. menschl. Schlnnd- n. Kehlkopfes, Leipzig,
1862; Gkutzner, Physiologie d. Stimme u. Sprache, Hermann's Handbnch d.
Physiol., I (2) 106, Leipzig, 1879.
266 PRODUCTION OF SPEECH
is open and the tongue depressed; the epiglottis is half
erected. On the vibration theory these adjustments fayor
the development of high partials and their exit from the
mouth ; the vocal bands are supposed to strike sharply against
each other and the load is not adjusted to avoid the over-
tones. The sound is like that of a membrane striking
against a solid edge (Gkutzner). A screeching tone is an
exaggeration of the sharp tone (Gabcia). The adoption of
the striking of the bands into the vibration theory practically
replaces it by the explosion theory. On the explosion theory
the glottis closes and opens in a way to produce puffs whose
shape is not smooth.
Experimental records from the larynx are generally con-
cerned with its vertical displacement or with the pitch of the
tone it produces.
The rise and fall of the larynx can be registered by tam-
bours with special projecting arms.^ The rise of the larynx
for high tones and its fall for low ones indicate activity of
the thyrohyoid and sternothyroid muscles. Movements dur-
ing internal speech have been similarly registered.^ Ob-
servations with RoNTGEN rays showed that for a the hyoid
bone is still while the larynx is somewhat raised. The
larynx is higher for a than for u, lower than for i. For
e it is somewhat lower than for i; for o somewhat higher
than for u. As a is made to pass into i both hyoid bone and
larjmx rise, maintaining their relative positions. As a is
changed to u, the larynx sinks and the hyoid bone is pushed
forward somewhat. For a the cavity bounded by the larynx,
the base of tongue, the rear wall of pharynx and the soft
palate is only moderately large ; it is large for e, still larger
for i and narrowest for u. For rise in pitch the epiglottis
rises, and likewise the reverse. Observations on over thirty
singers showed that with the falsetto voice the epiglottis is
1 T. Krztwicki, Ueber die graphiache Darstellung der Kehlkopfbewegungen
beim Sprechen n. Singen, Kiinigsberg, 1892 ; Eousselot, Principes de phonStique
experimentale, 98, Paris, 1897.
- Curtis, Automatic movemenin of the larynx, Amer. Jour. Psychol., 1900
XI 237.
ACTION OF THE LARYNX 267
nearly upright, and that the larynx is raised and brought
near the hyoid bone.i
The pitch of the tone from the larynx may be determined
in several ways.
To register from the outside of the larynx, a tube is fitted
tightly into the bottom of a small round box ^ (Fig. 124);
the edges may be cut to fit closely on the neck over the
larynx; it may be covered with a rubber membrane.^ A
rubber tube transmits the air vibrations from this box to a
Marby tambour (p. 195 ), whereby they may be made to
record themselves on a drum.
Rosapelly's electric interrupter for the larynx consists of
a small weight on a spring whose inertia closes an electric
circuit when its supporting frame is jarred.*
RousSELOT ^ has used a carbon microphone to interrupt an
electric current in accordance with the _
air movements of the sound spoken.
Instead of a telephone plate the fluctu-
ations in the magnet included in the
circuit were communicated to an arma- '°' ^"'*'
ture held by a membrane of varnished parchment; the move-
ments were registered by a recording arm.
The microphone known in America as the Blake trans-
mitter and a very light and carefully made time-marker may
be used for the same purpose, as may also a carefully adjusted
voice key (Fig. 66).
The methods of registering the vibrations in the speech
sounds issuing from the mouth aie mentioned in Part I; a
1 ScHEiER, Ueber d. Verwerthung d. RSntqen-Strahhn f. d. Physiol, d. Sprache
u. Stiinme, Arch. f. Laryngol., 1898 VII 126; Ueber d. Bedeutung d. Rontgen-
Strahlen fur d. Physiol, d. Sprache u. Stimme, Neuere Sprachen, 1897-98 V,
Beiblatt, 40.
2 RocsSELOT, Les modifications phon€t. du langage, 15, "Rev. des pat. gallo-
romans, 1891 IV", V (also separate) ; Principes de phone'tique expe'rimentale, 97,
Paris, 1897.
3 Meyek, Stimmhaftes H, Neaere Sprachen, 1900 VIII 261.
* RoSAPELLY, Essai d'inscription phon&iqm, Travaux du lab. de Marey, IT
117; RoussELOT, Les modifications, as before, 14; Principes, 106.
* RoussELOT, Les modifications, 3.s before, 16; Principes, 127.
268 PRODUCTION OF SPEECH
proper interpretation of the records gives the period of the
cord vibrations at each moment (p. 62).
The sounds from the larynx may be grouped under the
terms ' tone, ' ' whisper, ' ' breath ' and ' catch. '
In song and ordinary speech the sounds are mainly tones ;
it seems quite probable that the cords execute their vibrations
in the same general way for the tones in both eases. Accord-
ing to DoNDERS ^ and Helmholtz ^ the edges vibrate freely in
song, whereas in speech they strike. Hermann's^ results
for sung vowels would indicate that even in song the action
is not that of a freely swinging membranous reed; the vowels
show that the tone from the bands consists of a series of posi-
tive puffs of air separated by longer or shoi'ter intervals of
silence (p. 39). My own curves show conclusively that in
speech the puff from the bands is of an explosive nature giv-
ing a positive blow to the air in the mouth cavity and that the
sharpness of the explosion differs in different vowels. The
action in singing, as indicated by Herjlann's curves, is
probably of a similar nature, the explosion being a more
gradual one.
The difference between song and speech lies apparently not
in the kind of vibration executed by the cords but in the man-
ner in which the tone of the voice runs up and down in
pitch. The difference seems to have been correctly perceived
by Aristoxenus,* who in discussing KivrjaK ^wi'^? opposes
KivrjaK (Tvve')(fi<i to KLvrjo-f; Bia(TTT]ij,aTiKr]. The first term may
be translated as 'change in pitch of the voice,' the second
as 'continuous change,' and the last as 'change by steps.'
The continuous change he considers to be characteristic of
speech as opposed to song. 'Now the continuous move-
ment is, we assert, the movement of conversational speech,
1 DoNDERS, Over cle tongwerktuigen van hei stem- en spraakorqaan, n. p., u. d.
2 Helmholtz, Die Lehre v. d. Tonempfindungen, 5. Aufl., 169, Leipzig, 1896.
^ Hermann, Weitere Untersuch. il. d. Wesen d. Vokale, Arch. f. d. ges. Physiol.
(Pfluger), 1895 LXI 192.
4 ARiSTOXENna, Harmonica, I § 25, p. 8, Meib. The passages are collected
by Johnson, Musical pilch and the measurement of intervals, Thesis, Baltimore,
1896.
ACTION OF THE LARYNX
269
for when we converse the voice moves through a space in
such a manner as to seem to rest nowhere. ' ^
In song the intention is to maintain the voice on successive
notes at a constant pitch for each one, while in speech the
pitch is seldom constant. The effort at a constant pitch
necessarily fails as do all attempts at constant muscular con-
traction (p. 202) ; the result is an average pitch with a prob-
able error (p. 201) whose size depends on the accuracy of
muscular control (p. 202) under guidance of the ear.
The constancy with which a tone can be maintained is a
matter of considerable importance that has not yet been in-
FlG. 125.
vestigated. It may be demonstrated by singing in unison
with a fork or an organ pipe; the beats due to the differ-
ences are readily heard. It may also be observed optically by
an apparatus devised by Hbnsen.^ The flame of a mano-
metric capsule (Fig. 18) in front of a mirror on the end of a
prong of a vibrating fork (Fig. 125) is seen to have one point
(Fig. 125) for a tone sung with the ratio of frequency 1 : 1
1 Abistuxenus, Harmonica, I § 28, p. 8, Meib., quoted by Johnson, The
motion of the voice in the theory of ancient music, Trans. Amer. Philol. Assoc,
1899 XXX 47.
2 Hensen, ^i'n einfaches Verfahren sur Beobachtung der-Tonhohe eines gesunge-
nen Tons, Arch. f. Anat. u. Physiol. (Physiol. Abth.), 1879 155.
270
PRODUCTION OF SPEECH
(unison) to that of the fork, two points (Fig. 126) with the
ratio 2 : 1 (octave), three points (Fig. 127) with the ratio
3 : 1 (duodecime), etc. With the ratios 3 : 2 (fifth) there are
three points (Fig. 128) but the flames appear twisted to-
getlier, with 4 : 3 (fourth) four points (Fig. 129), and with
6 : 4 (major tliird) five points (Fig. 130). With the least
variation in pitch from the exact ratio the figure seen in the
mirror appears to rotate around a vertical axis. The greater
the variation the more rapid the rotation. When the tone is
Fig. 126.
Fig. 127.
Fig. 128.
Fig. 129.
Fig. 130.
too low, the flame appears to move in the direction in which
the points are directed; when too high, in the opposite direc-
tion. The mirror-fork may be provided with sliding weights
for altering the pitch.
The accuracy with which a tone can be reproduced by the
voice has been investigated by Klundbr.^ Two small pho-
nautographs, each consisting of a recording point attached to
a rubber membrane on the end of a tube, were arranged to
record simultaneously on a smoked drum. One of them re-
corded the vibrations of a tone from an organ pipe, the other
1 Klunder, Ueber die Genauigkeit der Stimme, Arch. f. Anat. u, Physiol.
(Physiol. Ahth.), 1879 119.
ACTION OF THE LARYNX 271
those from a voice attempting to repeat the same tone. Tlie
results showed under favorable circumstances an average error
in pitch of y^ of 1%, reaching sometimes li%.
A convenient method of measuring the accuracy of repeti-
tion is the following. In a darkened room a white siren disc
with series of holes from 90 to 110 in number is placed on
the axle of the siren motor (Fig. 59). The resistance of
the motor (p. 10) is adjusted until a blast from the tube
through the series of 100 holes produces a tone of the de-
sired pitch, found by comparison with a tuning fork or other
instrument of known periodicity. A manometric flame (Fig.
18) is held opposite the holes at one side. The standard
tone is produced for a moment by a blast on the siren tube ;
the person tested then reproduces it by singing into the trum-
pet of the manometric capsule. The vibrations of the flame
cause one of the series of holes to remain apparently still
while the others appear in progression or regression; the
number of this series from that used for the blast is
noted. The difference of this number from the row that
was used for the blast will indicate the error in pitch.
Let the number be n; then the tone sung was in error by
Y?^ X frequency of the tone produced by the blast. For
example, if the siren had been adjusted so that the tone of
the blast through the series of 100 holes was a}- = 435, and if
the third row from the blast-row toward the outer edge stood
still while the voice attempted to repeat this tone, then the
tone of the voice was zr-^ X 435 = 13 vibrations too high.
The method may be used to determine the accuracy in
striking unisons, fifths, octaves, etc., also the dependence of
accuracy on the interval that elapses between the tone heard
and the tone sung, also tone memory, etc. These problems
are still uninvestigated. Instead of the manometric capsule
a voice-key (Fig. 66) in the primary circuit of a spark coil
(like the contact wheel in Fig. 59) may be made to produce
sparks between two metallic points or flashes in a Geisslee or
272
PRODUCTION OF SPEECH
Pumj tube, whereby more brilliant illumination is obtained.
Various other modifications are possible.^
Tlie notes that can be produced by the voice are limited to
a small range. The ordinary range covers about an octave
and a half or more, as indicated in the accompanying diagram.
Soprano
Mezzo-soprano
_ Alto
Tenor
Baritone
Bass
A ' register ' is the range of the voice within which it pro-
duces tones of the same general acoustical quality. The
tones in such a register are presumably produced by similar
action in the larynx ; the two uses of the term " register '
probably coincide with these two closely united phenomena.
In most persons two typically distinct registers are present.
The ' chest register, ' with resonance vibrations generally felt
in the thorax, has a strong and smooth acoustic color and
requires little effort. The higher ' head register ' of a thin-
ner acoustic character has a resonance apparently in the
head; the tones are produced with more effort. The larynx is
raised for head tones and there seems to be a general extra
muscular effort. The head register is often called the ' fal-
setto register.'
Several other registers have been found. The ' middle
register ' includes notes higher than those for which the
chest register easily provides naturally, but for which the
head register is not required. The ' deep bass register ' lies
below the chest register; the arytenoid cartilages are not
brought together as in the chest register, whence results a
' breathy ' tone. The ' straw bass register ' lies below the
previous register, the muscles of the larynx are not stretched
1 ScKiPTURE, Ehnientary course in psychological measurements, Stud. Yale
Psych. Lab., 1896 IV 135.
ACTION OF THE LARYNX
273
SO much as usual, the larynx appears tipped backward, the
vocal bands are quite relaxed but are brought close together.
The tone is not only breathy but also somewhat rattling. ^
It is probably the case that each person may possess several
registers. The usual instruction in singing aims at an abil-
ity to pass from one register to another without a very notice-
able change in the character of the notes ; among the Alpine
jodelers the aim is to develop the registers separately and
distinctly.
Mackenzie has reported observations on four hundred
singers.^ The chest register was generally used throughout
by pure sopranos, among whom were Nilsson, Albani and
Valeria; the contraltos almost invariably used the head
register for the high notes; the mezzo-sopranos used both
registers ; the tenors generally used both, though a few used
the chest register only; the barytones and basses used the
chest register only.
For typical voices the notes may be assigned to the different
registers as indicated by the following diagrams. It will be
noticed that several tones may be made in either register.
Hend
Chest
Deep hass
I 5£raw bcLSS
Male.
Female.
The glottis may be closed sufficiently to vibrate but yet to
allow the continuous escape of breath. The result is known
in vocal music as the ' breathy tone ; ' pictures of the glottis in
1 For details concerning these two registers see Grutznee, Physiologie d.
Stlmme u. Sprache, Hermann's Handbuch d. Physiologifi, I (1) 89, Leipzig, 1879.
2 Mackenzie, Hygiene of the Vocal Organs, 35, London, 1888.
13
274
PnODUCTION OF SPEECH
two such cases are shown in Figs. 131 and 132 ;i they were
from pupils of a vocal instructor who taught a breathy tone.
Fig. 131.
Fig. 132.
In whispering, a current of air is blown through the
glottis. The laryngeal action differs somewhat in various
kinds of whisper.
The arytenoid cartilages may be brought toward each other
with the points tilted forward, while the vocal cords are
__^_ relaxed, and bulging in the mid-
"~~~^ die (Fig. 133).2 The epiglottis
may be considerably depressed
over the opening. The glottis in
these forms is used for soft whis-
FiG. 133.
pering.
Still rougher whisper sounds
may be produced by pressing the
base of the epiglottis against the ventricular bands and the
upper edges of the arytenoid cartilages, which are them-
selves brought close together. The three slits thus produced
meet at right angles. When these slits are tightly closed
and made to vibrate along their edges by strong breath
pressure, the peculiar Arabic ain is produced. ^ The vibra-
1 Curtis, Voice Biiildiiig autl Tone Placing, Figs. 41 and 42, New York, 1896.
^ CzEEMAK, Der Kclilkopfspiegel, Leipzig, 1863 ; Ges. Schriften, I 551, Leip-
zig, 1879.
^ CzEEMAK, as liefore, Ges. Schriften, 552, Sweet's supposition of a vibra-
tion occurring in tlie larynx helow the glottis involves an anatomical impossi-
ACTION OF THE LARYNX 275
tion is from the edges of the arytenoids and the ventricular
bands, not from the cords.
Records made by Rousselot ^ indicate that a vibration of
constant period — in one case -^-^ of a second — is present in
all the whispered vowels ; this would indicate that the vocal
cords vibrate at least to some extent in whispering. A
laryngoscopic observation in one case showed during whis-
pering the vibration of a fixed polyp on one of the cords,
indicating a vibration of the edge of the cord. Such a whis-
per might be termed a ' sonant whisper. '
The form of the glottis in whispering is variable. ^ In some
persons the entire glottis is open, forming an isosceles tri-
angle (5, Fig. 113) in others only the cartilaginous glottis is
open (C, Fig. 113). The former position gives a. soft whis-
per, the latter a strong one. Of 58 persons examined with
the laryngoscope 34 showed the former condition and 24 the
latter when they were told to whisper. In producing a soft
whisper the cords are approached hardly at all; with a
medium whisper the ligamentous glottis is closed and the
ventricular bands are somewhat approached; with a very
strong whisper the cartilaginous glottis is closed to a small
opening, the ventricular bands are in contact and the epi-
glottis is much . depressed. This typical action is not uni-
versal; some subjects keep the glottis widely open even in
loud whispering; others close the ligamentous glottis even in
soft whispering; with still others the ends of the glottis are
closed, leaving a small elliptical opening in the middle, for
a soft whisper.
It seems plausible to suppose that between a full tone used
in speech and song and a completely toneless whisper there
may be numerous gradations of breathy tone and sonant
whisper.
When the cord glottis is adjusted to vibrate while the car-
bility; Sweet, ISi ardbik 0rout saundz, Maitre phonetique, 1895 X 79; Passy,
le gijti/ral, Maitre phonetique, 1895 X 99.
1 Olivier, De la voix chuchote'e, La Parole, 1899 126.
2 Olivier, as before, 28.
276 PRODUCTION OF SPEECH
tilage glottis remains open, a combination of tone and whisper
may be produced, as occurs in groaning.
When the glottis is slightly narrowed, a rushing noise may
be produced ; there are many degrees of this ' glottal breath
sound,' or one form of h. That the glottis is somewhat
narrowed in the usual h seems to have been settled by
observations of Czeemak and Brijcke.^ Much narrowing
produces an h of heavier character than the ordinary h.
Many articulations are used to produce the sounds that may
be grouped as h-sounds. The friction may occur in the
glottis, just above it, with the velum near the pharynx wall,
between the tongue and velum, etc. Tones are imposed on
the h-sound by the resonance cavities ; the assertion that the
adjustment of these cavities is the same as that of the neigh-
boring vowel is probably erroneous for English speech.^ The
American h is apparently like the British h in not having
the same adjustment as the following vowel.
The h may be weakly or strongly sonant. The sonant h
was prescribed by the Sanskrit grammarians ; ^ it is used in
some modern languages.*
Several examples of sonant h are to be found in the records
at the end of this volume. In the words ' You had it ' on
Block V (Plate VII) of the Jefferson record the liquid j
changes to u in line 6, the u continues during the first half
of line 7, the ae of ' had ' appears in the latter half of line 7
and passes into d in the middle of hue 8. At no time from
the ] to the d do the cords cease to vibrate, as can be
1 Czeemak, Physiol. Untersuchungen mit Garcia's Kehlkopfspiegel , Sitzber. d.
k. Akad. Wiss. Wien, math.-nat. Kl., XXIX 557 ; Gesammelte Schriften, I 551,
Leipzig, 1879.
2 Lloyd, in Vietoe, El. d. Phon., 4. Aufl., 22, Leipzig, 1898.
8 Meyer, Stimmhaftes H, Neuere Sprachen, 1900 VIII 261 ; tsum ftimhaftn
ha, Maitre phonetique, 1901 XVI 87. Taittiriya Pratii;akliya, ii. 47, ed. by
Whitney, Journ. Amer. Oriental Soc, 1871 IX 77 ; Michaelis, Ueber das H
und die verwandten Laute, Arch. f. d. Studiiim d. neueren Sprachen (Herrig),
1887 LXXIX 49, 283. ■
* Meyek, as before; KlingHaedt, Stimmhaftes H, Neuere Sprachen, 1901
IX 85 ; Passy, Ii vocalique, Neuere Sprachen, 1901 IX 245.
ACTION OF THE LARYNX ' 277
plainly seen in tiie curve ; yet the record speaks a distinct h
in ' had.' Since the cord tone does not cease this must be a
sonant h. Two other cases appear in the Cock Rohin records.
The h is distinctly heard in ' saw him ' of ' I saw him die.'
The curve in Plate II shows the vowel o beginning at the
left of line 6 and increasing in amphtude during the first half
of the hne. The last 'quarter of the line shows the i of ' him.'
Between the o and the i there is a region of diminished amph-
tude, corresponding to the h. The record conclusively shows
that at no time during the three sounds does the cord tone
cease. The h is therefore sonant. A fainter h is heard in
' saw him ' of ' Who saw him die ? ' ; it is so faint that it
escaped my ear when listening to the gramophone record in
first studying the curves and I therefore recorded the sounds
as SDim. I have described them above (p. 63) in this way.
On carefully hstening again to the record I could distinctly
and certainly hear the h in this case also. The curve, how-
ever, in Plate I line 1 shows only a very shght weakening
between the o and the i just to the right of the middle of the
line. The sonant h in this case has vibrations as strong as
those of a large part of the o. The presence of the h in all
the above cases was apparent to other ears than mine. The
curves show that the cords do not relax their tension during
these cases of sonant h. The regularity of the period and its
agreement with the cord periods of the neighboring vowels
indicate that there is no very great readjustment in the larynx.
One view of the mechanism of sonant h is that the glottis
opens while the cords are vibrating and that this permits an
escape of air with a rushing noise (not appearing in my curves)
while the cords continue to vibrate. This involves fairly con-
stant tension in the vocal cords in spite of the opening of the
glottis. The action is like that of the sonant whisper. This
would probably be the vocal action corresponding to Seel-
mann's view ^ of the nature of the Greek spirifus asper and
lenis as being strong and weak breathy beginnings of vowels.
1 Seelmann, Die Ausprache des Latein, 255, 262, Heilbronn, 1885 ; Paul,
Ueber vokalische Aspiration und reinen Vokaleiusatz, Leipzig, 1888.
278 PRODUCTION OF SPEECH
The speech curves considered suggest another view. This
is that the cords continue to vibrate during the intervocalic
h with no disturbance, the glottis remaining closed as in the
adjacent vowels, and that the h is produced by narrowing the
air passage either by bringing the ventricular bands close
together or by partially closing the epiglottis down over the
larynx. I find that I can sing a breathy a, e, etc. indefinitely
long with some closure that is behind and below the tongue.
According to this view the sonant h is a sonant fricative of
the same class as j, 7, etc. with the passage narrowed above
the vocal cords. It thus differs radically from the ordinary
breathed h.
To the ear the surd and sonant h-sounds do not differ
very much. As forms of articulation they must differ. The
surd h is usually a glottal fricative. The sonant h may be a
glottal fricative produced by closing the ligamentous glottis
for the tone while the cartilaginous glottis remains open for
the friction,^ or by opening the glottis somewhat while the
cords are still vibrating,^ or by narrowing the air passage
above the cords without altering the glottal adjustments (as
explained above).
The speech sound known as the 'glottal catch' is made
by closure of the glottis and sudden opening; the time of
closure appears to the ear as silence ; the opening may appear
as an explosion unless it is masked by the following vocal
sound, or it may be so gradual that it is not heard.
The glottal catch may be used to replace an occlusive of
another kind. I have repeatedly observed this in my child of
12 to 18 months as a substitute for t and k, as in te>3WD' ' take
a walk,' mi' 'meat.' This child had never heard a glottal
catch used as a speech sound. A similar use of the glottal
catch to separate words occurs in the attempt to separate
with special distinctness such combinations of words as du'it
1 CzEEMAK, Ueher d. Spiritus asper und lenis, etc., Sitzb. d. k. Akad. d. Wiss.
Wien, math.-naturw. Kl., 1866 LII (2) 630, Anmerk. 1 ; also in Gesamm. Schriften,
I 756, Leipzig, 1879.
2 Meter, as before.
ACTION OF THE LARYNX 279
*do it,' slai.tiaz 'sly tears.' In the latter case there is
no glottal explosion because the t-closure occurs before the
glottis is opened. A frequent pronunciation of ' camp-
meeting ' is kaem'mititi ; the mouth adjustment for m is kept
between the two vowels, and the occlusive effect is xaroduced
by inserting a glottal catch in the middle. The explosion of
the catch is in this case nasal. The use of a glottal catch at
the beginning of a vowel often occurs in -a-a^a ' ah, ah, ah '
repeated in as an exclamation of warning. It is sometimes
used to end vowels, as occasionally ha> used for an exclama-
tion of surprise, or even in hwo' ' whoa.'
In the Scottish dialect of Glasgow the glottal catch accom-
panies intervocalic t, as in bat-gr ' butter,' or may replace it,
as in b3>gr.i
The glottal catch is the regular beginning of initial vowels
in North German; it disappears from a word whenever the
vowel in the union of speech is no longer felt to be an initial
one; thus, -ain far ain ' ein Verein, but herain 'herein'
instead of her 'ain. The failure of foreigners to use the
glottal catch in speaking German produces a strange im-
pression on the native ear.
The glottal catch appears as the Danish stdd, by which, for
example, ma-lar 'maler, [he] paints' is distinguished from
malar ' maler, painter. ' '^
The glottal catch seems to occur also as the Arabic hamza,
which is a regular consonant represented in the alphabet ; the
•closure, however, seems to be reinforced by pressing the base
of the epiglottis over the closed glottis. ^
The equivalence of a glottal catch to other consonant
articulations has been observed in some experiments by
MiYAKE.'* In beating time with the finger while a syllable
-was regularly repeated, the finger-beat occurred 1. at the
1 Passt, £tude sur les changements phonetiques, 155, These, Paris, 1891.
~ 2 ViETOE, Elemente d. Phonetik, 4. Aufl., 25, Leipzig, 1898.
2 CzERMAK, Phydol. Untersuchungen mit Garcia's Kehlkopfspiegel, Sitzb. d. k.
Akad. Wiss. Wien, math.-nat. Kl., 1858 XXIX 557; Ges. Schriften, I 555,
Leipzig, 1879.
4 MiYAKE, Researches on rhythmic action, Stud. Yale Psych. Lab., 1902 X 54.
280 PRODUCTION OF SPEECH
beginning of the consonant in a syllable composed of conso-
nant and vowel ; 2. at the beginning of the vowel in a syllable
composed of an English vowel with a smooth beginning ; 3.
ahead of the vowel by the regular consonant time in a syllable
composed of a vowel begun with a glottal catch.
The usual attempt to explain the glottal catch as a slight
cough is a blunder; the bulb centres (p. 193) for the two
activities are not the same and the muscular action is probably
quite different.
Repbkences
For the history of the theories of cord action : Wright, The nose and
throat in medical Jiistory, The Laryngoscope, 1901-02; also separate. For
the action and treatment of the cords in singing : Mackenzie, Hygiene
of the Vocal Organs, London, 1888; Curtis, Voice Building and Tone
Placing, New York, 1896 ; Stockhausen, Gesangsmethode, Leipzig.
CHAPTER XX
TONES OF THE VOCAL CAVITIES
When a blow is struck on the wall of a cavity or over an
opening, the molecules of air adjacent to the wall or the
opening are driven toward the neighboring molecules. This
produces a wave of condensation and rarefaction which is
propagated through the cavity. A negative blow, such as
that produced by a pull on the wall or the sudden removal
of an object from an opening, propagates a wave of rare-
faction similarly.
The excitation of the air of the cavity may arise from a
single momentary impulse, from a succession of such im-
pulses or from a series of impressions of a more or less wave-
like character.
The simplest conditions can be represented by supposing
a piston moving without friction in a rigid cylinder closed
at one end; for most purposes we can consider the air in
the cylinder to act as a spring in resisting the movement
of the piston. With a condition of constant density the
period of the vibration resulting from a blow will depend on
the mass of the piston and on the strength of the spring,
that is, on the size of the cavity. In moving through an
opening the air moves approximately as an incompressible
fluid. From the mass of the air and the elasticity of the
cavity, the period of vibration can be calculated.
The period of a cavity of the volume S communicating
with the external air by a long cylindrical neck of the length
L and the area A can be shown ^ to be
1 Ratleigh, Theory of Sound, 2d ed., II § 303, London, 1896.
282 PRODUCTION OF SPEECH
where a is the velocity of sound (p. 4). If the radius of the
neck is M, the area will be "ttB^ and the formula for the
period will be
y,_ 2 Vtt . VLS ^Q.
aR ■ ^^^
The period is lengthened by enlarging the cavity, or by
increasing the length of the cylindrical neck ; but it is short-
ened by enlarging the area of the neck.
The ease with which the air flows in and out through
the cylindrical channel, or the conductivity of the channel,
diminishes as the length is greater, and increases as the area
increases. The degree of conductivity can be expressed ^ by
. 0 = ^. (3)
where A is the area and L the length of the channel. The
period of the channel will be
T
J^JS^ (4)
n y 0.
where S is the volume of the channel and a the velocity of
sound. Less conductivity corresponds to increased mass of
the piston, with a resultant lengthening of period, and greater
conductivity to decreased mass with a resultant shortening
of period. Owing to the loss of movement at the open end
of the cylindrical tube, the conductivity c as calculated
above needs a correction which is found to be at least
h = \TrR for each end ; this must be added to L. The cor-
rection is exactly this for an infinitely long tube with an
infinite flange at the open end. For an unflanged end it is
equal to about 0.6i2, when the wave length is great in com-
parison with the diameter. Thus, instead of the formula (3)
above we should write ^
1 Rayleigh, as before, § 304.
2 Rayleigh, as before, §§ 307, 309, 3U.
TONES OF THE VOCAL CAVITIES 283
Some experimental determinations by Sondhauss ^ in the
case of resonators without necks showed that the influence of
the aperture depended mainly upon its area, although a very
elongated shape produced a rise in pitch; his empirical for-
mula gave for the pitch of the cavity
1 -^7
n = - = 52400 ~, (6)
T VS
the unit being the millimeter. For flasks with long necks he
found
n = 46705\/ -^, • (7)
The latter formula supposes that the neck is so long that
the correction for the open end may be neglected, the former
that it is so short that the length itself may be neglected. In
practice, formulas (4) and (5) will generally be required.
Various theoretical and experimental data have been given
by Helmholtz^ and Ellis. ^
When a cavity has more than one aperture, the separate
conductivities are to be added if the apertures are so far
apart that they can be considered as acting separately. This
is the case in the mouth, the labial aperture being at the end
opposite to the pharyngonasal aperture. In the case of a
vowel like i the front and the rear cavities on either side
of the elevation of the tongue have each two apertures.
The main vibration of the air in a cavity is frequently
accompanied by other vibrations of shorter periods. These
vibrations in cavities with narrow necks are relatively of
very much shorter period.
Owing to dissipative forces the vibration excited by the
blow of the piston will die away as explained on p. 5.
1 SoNDHAUSS, Ueber d. Brummkreisel u. d. Schwingunqsgesetz d. kubischen
Pfeifen, Ann. d. Phys. u. Chem., 1850 LXXXI 235, 347; Rayleigh, as before,
§ 309.
2 Helmholtz, Lehre v. d. Tonempfindung, 5. Aufl., 73, Beilage II, Braun-
schweig, 1896.
8 Ellis, notes to translation of Helmholtz, Sensations of Tone, London,
1885.
284 PRODUCTION OF SPEECH
The effect of repeated blows on a system with a period of
free vibration has been considered on p. 12.
The vibrations in cavities whose three dimensions are very-
small compared to the wave length and whose communication
with the external air is by small holes in the surface have
been investigated by Helmholtz.^ A later treatment is by
Rayleigh.2 The topics discussed include the cases where
a mass of air confined almost entirely by rigid walls com-
municates with the external atmosphere by one or more nar-
row passages, where there is a contraction making a double
cavity, where there is a long tube in connection with a
reservoir, where there are lateral openings, where the open-
ings are of different forms, where the necks are of different
shapes, where the cavities are not regular tubes, where the
cavities are not straight tubes, etc. It is evident that these
are to a great extent just the problems involved in the action
of the vocal cavities in producing speech sounds, and that
the same methods can be used to investigate their action.
For example, a mathematical treatment of the vibration
period of the mouth cavity, as affected by its size, its aper-
tures, its internal neck formed by the tongue, the labial tube,
the labial and nasal openings, etc., with such modifications
as may be needed to account for the lack of rigidity in the
walls and the special deviations from the general conditions,
would be of as great value to phonetics as similar treatments
of like problems have been to physics and other sciences.
These problems have not yet been attempted in spite of their
great importance.
The preceding general considerations find many practical
applications to the vocal cavities. As the walls of the
cavities are not very rigid, the formulas given above are not
strictly applicable; just how closely approximate they are,
it is impossible to say; a mathematical treatment including
1 Helmholtz, Theorie der Luftschwingungen in RBhren mit offenen Enden,
Journal f. reine u. angew. Math. (Crelle), 1859 LVII 1.
2 Eayleioii, On the theory of resonance, Phil. Trans. Roy. Soc. Lond., 1871
CLXI 77 ; Theory of Sound, 2d ed., II §§ 303-322, London, 1896.
TONES OF THE VOCAL CAVITIES 285
yielding walls seems to be still lacking. Yielding walls
lengthen the period of vibration and increase the factor of
friction.
The vocal cavities may be made to produce sound waves
by striking the hand against the open mouth, or by sud-
denly removing it from the mouth, or by snapping the finger
out of the mouth ; also by striking a blow on the cheek, and
by various speech movements.
For a blow on the cheek with the mouth closed the piston
(p. 281) is represented by the flesh of the cheeks. The air
within the mouth resists the blow. The period of vibration
that results is long owing to the large mass of the cheeks
and also to the weakening of the elastic force by the yielding
walls of the mouth. When the mouth is open, there is no
such resistance to the movement of the cheek as when the
mouth is closed. The blow on the cheek drives the air out
between the lips. The air between the lips now takes the
place of the piston and any effect on the cheeks is negligible
in comparison.
When the hand is struck against the open mouth the cavity
receives an impulse of condensation or a positive blow. When
the finger is snapped out of the mouth it receives one of rare-
faction, or a negative blow.
Blows may also occur by the release of compressed air.
This is the case in the explosive occlusives. In these there
is a closure of the air passage by the lips (p, b), by the tongue
against the teeth or palate (t, d, k, g) or by the glottis (>);
the air is compressed behind the closure and, on being re-
leased, strikes a positive blow on the cavity in front and a
negative blow on the one behind.
The puffs from the vocal bands are capable of arousing
tones in the cavities of the air passage. There are at least
two theories of the manner in which the complex note heard
in singing or in speech is derived from the vibrations of the
bands.
According to one theory the note produced directly by the
vibrations of the bands consists of a series of partial tones of.
286 PRODUCTION OF SPEECH
different intensities (pp. 72, 106, 256), and the vocal cavities
reinforce some of them by resonance. For illustration we may-
suppose the notes from three different pairs of vocal bands A,
B, 0 to be composed of partials having relations of intensity
as indicated by the sizes of the figures : —
^—123456789 10
J5 — 123456789 10
C — 123456789 10
Although the three notes will appear of the same pitch, their
characters will be different, the note of A being thin and
flute-like, that of B rich, and that of 0 piercing and sharp.
These relations are modified by the resonating cavities of the
chest, pharynx, mouth and nose, so that some of the partials
are reinforced. In this way the even partials might be rein-
forced in the case of A so that the voice obtains something of
the character of B. Likewise certain arbitrary partials might
be so reinforced in the case of 0 that it also becomes some-
what like B. To attain these results the system of cavities
must be carefully adjusted to resonate to the tones desired.
The second theory would suppose the bands not to vibrate
but to open momentarily and then close again in a series of
movements (p. 260), whereby the resulting air movement
is not a vibratory one of a sinusoidal (p. 3) or a harmonic
(p. 13) nature but is a series of brief puffs. The air move-
ment direct from the cords is thus not like that of a smooth
vibration but that of a series of explosions. The curve of
explosion, that is, the sharpness of the explosive rise and
fall, may be different in different cases. The effect on the
ear will be a tone of the pitch of the frequency of the explo-
sion, to which there may be added higher tones — not neces-
sarily harmonic in relation — arising from the character of
the curve of explosion (p. 94). Owing to the blows struck
by the explosions from the cords the cavities add tones
of free vibration (p. 2). That for the chest register this
TONES OF THE VOCAL CAVITIES 287
theory is certainly the correct one has been shown on pages
259-260; for the head register the matter has not been
finally settled.
The size of the cavity within the mouth can be varied by
movements of the tongue and jaw ; the size and shape of the
openings may be altered by adjustments of the lips, velum,
palatine arches, epiglottis and glottis. The tongue may also
divide the mouth into two or even three cavities with necks
between them. The pharyngeal cavity is subject to great
modifications by contraction of its muscular walls, by move-
ments of the tongue, and by the rise and fall of the larynx ;
its apertures are greatljf varied by the velum, tongue and
epiglottis. The tracheal cavity seems to be capable of little
variation. The entire system of cavities forms a compound
one. Its natural period depends on the sizes of the com-
ponent cavities, on the necks between them, and on the sizes
and shapes of the openings.
Several methods may be used to determine the natural
period of a cavity.
One method consists in holding vibrating bodies before the
opening and noticing which one is most loudly reinforced.
Forks of different pitch may be held before the mouth;
the period of the fork producing the loudest resonance
may be considered as that of the mouth cavity under the
given circumstances. This method has been used to deter-
mine the period of the mouth cavity adjusted for different
vowels, although the adjustment of the mouth under these
circumstances may differ considerably from the actual adjust-
ment in speaking. Helmholtz's ^ determinations for his own
voice (North German) were as follows : —
Vowel: ^^^ ''"^" o a e^(d) e^Ce) i ce y
Tones: /» /^ I* a' cP+g^ f^ + U" P+d^ P + <?^ P+g""
The ' bright u ' was like the French ou. In musical notation
the results are as indicated on next page.
1 Helmholtz, Lehre v. d. Tonempfindungen, 5. Aafl., 177, Braunschweig,
1896.
288
PRODUCTION OF SPEECH
1=
1:
^ ^ fe^ ^
W
Atjerbach's results (German) by the same method ^ were : ^
Vowel : Uj u
Tone : g^ c'
o
^2
/.2
or in musical notation
gi g3 ^2
4=-
ce
e^(a)
-t^
w=^-
=t:=t
U2
ce 62
The same method is used in a modified form to determine
the period of the chest cavity. When the voice . is made to
rise and fall in pitch, certain tones will be felt to be rein-
forced in the chest.
Another method consists in blowing across the opening
of the cavity; in this way the pitch of a bottle is readily
obtained. 3 'Although good results have been obtained in
this way, our ignorance as to the mode of action of the wind
renders the method unsatisfactory. ' * The pitch of the mouth
cavity may also be obtained by a similar method, whisper-
ing; the resultant tone is noted by comparison with some
musical instrument. ^ The whisper-method is inaccurate
^ AuERBACH, Untersnchungen ii. d. Natur d. Vokalklanges, Diss., Berlin, 1876;
also in Ann. d. Phys. u. Chem , 1876 Ergb. VIII 177; Zur Grassmann'schen
Vokaltheorie, Ann. d, Phys. u. Chem., 1878 IV 508.
^ AuEKBACH, Die phi/sikalischen Grundlarjen d. Phonetik, Zt. f. franz. Spr. u.
Lit., 1894 XVI 144.
' DoNDERS, according to Geutznek, Physiologie d. Stimme u. Sprache,
Hermann's Handbuch d. Physiol., I (2) 160, Leipzig, 1879.
* Rayleigh, Tlieory of Sound, § 314, London, 1896.
5 DoNDERS, Ueber d. Natur d. Vokale, Archiv f. d. holland. Beitrage z. Natur-
u. Heilkunde, 1 858 1 157 ; Kronig, Notiz iib. Vokallaute u. iib. eine mat. Slimmgabel,
TONES OF THE VOCAL CAVITIES 289
chiefly because the ear is incapable of correctly assigning the
pitch of the complex of tones in the whispered sound. The
results of different observers ^ are so completely discordant
with one another and with those of the later accurate methods
that they do not seem worth considering. The Tkautmann
vowel-system, 2 based on whisper-observations, asserts that
the resonances form two septime accords of the notes g"^ and
g^ respectively; this is contrary to the facts ascertained by
more accurate methods.
Still another method consists in striking a blow on the wall
of the cavity. Blows on the larynx have been tried. ^
None of these subjective methods gives any reliable results.
In the first place the judgment of the ear concerning the
pitch of a sound is largely influenced by the presence of
other sounds; a complex of resonance tones is not heard as a
strong lower tone with higher ones added, but as a tone of
a pitch that may be quite different from the actual pitch of
the lowest tone.* The resonance of a vowel-position is just
such a complex of tones and is inevitably heard of a pitch
not that of its lowest component. The results obtained by
the preceding methods give the pitch-impressions of various
mouth positions, and, in as far as these mouth positions
represent the positions for the vowels, they are of value as
phenomena of hearing. They are of no value for determin-
ing the resonance tones in a spoken vowel.
Another objection, that is fatal to any method in which
the cords are not used, lies in the inevitable difference in
muscular adjustment when any change is made. It can
be readily demonstrated by psychological apparatus that a
muscular adjustment or movement becomes immediately
altered with any change in attention or other muscular
Ann. d. Phys u. Chem., 1876 CLVII 339; Teautmann, Die Sprachlaute, 46,
Leipzig, 1884-86 ; Stokm, Englisehe Philologie, 97, 2. Aufl., Leipzig, 1892.
' Teautmann, Die Sprachlaute, 46, 48, Leipzig, 1884-86.
2 Tkadtmann, as before, 40.
^ AuEKBAOH, Bestimmung d. ResonanztSne d. Mundkohle durch Percussion,
Ann. d. Phys. u. Chemie, 1848 III 153.
* Stumpf, Tonpsychologie, II 406, Leipzig, 1 890.
19
290 PRODUCTION OF SPEECH
adjustments of the body. The passing of a thought through
the mind or the moving of a finger can be shown by physio-
logical and psychological methods to affect to a greater or
less extent the muscles of the walls of the blood vessels,
of respiration, of the larynx, of the sweat glands, etc. This
is due to the extremely delicate coordination of all the sen-
sitive and contractile parts of the body by means of the
nervous system. It is unquestionable that the removal of
laryngeal action changes the articulation in the mouth to
some extent.
At best the foregoing results apply only to sung vowels,
since in spoken vowels the cavities undergo constant change
and there is no means of knowing which part of the vowel is
represented.
Only the objective methods of determining the resonance
tones are to be trusted. These methods are two: 1. synthesis
of elements that produce an actual speech sound, 2. analysis
of an actual speech sound into its elements.
The synthetic method consists in manufacturing sounds
that approximate speech sounds. The closer the imitation
the greater the likelihood that the principles employed are
the same as those of the vocal organs.
The speaking machines of Kempblen^ and Fabee^ were
built on a study of the action of the vocal organs. The
vowel instruments of Willis, Helmholtz and Lloyd were
designed to determine the resonance tones of the voice.
With cylindrical resonators of known pitch acted upon by
vibrating reeds Willis^ found the tones necessary to pro-
duce sounds resembling the English vowels in the words
no, nought, paw, part, pad, pay, pet, see, to be c^, e'^, g\ dP, f^,
d*, <?, g^ respectively. Just what sounds occurred in these
cases cannot be accurately stated ; they appear to have been o,
1 Kempelen, Mechanismus d. menschl. Sprache, 1791.
2 Techmer, Phonetik, Fig. 7a, Leipzig, 1880 ; also in Int. Zt. f. allg. Sprachw.,
1884 I Fig. 11.
' Willis, On vowel sounds, and on reed organ-pipes, Trans. Carab. Philos.
Soc, 1830 III 231.
TONES OF THE VOCAL CAVITIES
291
Dj, Og, aj, ag, Cj, Cg, i (the numerals indicating varieties of a
sound). With spherical resonators used in a similar way
Helmholtz^ obtained the same results as Willis for o, Oj,
Dj and a, but different ones for the last three, namely, 6^ for
Cj, c* for e^, and d^ for i. The following musical notation
indicates the results: first four notes, Willis and Helm-
HOLTZ; fifth, Willis; last three upper, Willis; last three
lower, Helmholtz.
rfeiz
III
=1
et
In these two methods the action of the cords in emitting puffs
of air was imitated by reeds.
With a series of forks before resonators Helmholtz
obtained '^ a good o followed by u on the note h~^ by using
the tones indicated by the adjacent notes with intensities as
indicated by their sizes. The other vowels
seem not to have been successfully imitated.
The failure of the method seems mainly due
to the maintenance of the tones of the resona-
tors by a constant supply of energy, whereas
in the vocal organs they are intermittently
aroused in the chest register (p. 259) and
possibly also in the head register.
By sending a blast into bottles of different
sizes with different necks Lloyd ^ has imitated some of the
whispered vowels; calculations of the periodicities of the
body and neck gave the periods of the resonance tones.
Different relations between the resonance periods of body
1 Helmholtz, Lehre v. d. Tonempfindungen, 5. Aufl., 1 99, Braunschweig, 1896.
2 Helmholtz, as before, 200.
^ Lloyd, Some researches into the nature of vowel-sounds. Thesis, Liverpool,
1890; Speech sounds: their nature and causation, Phonet. Stud., 1890 III 251;
1891 IV 37, 183, 275; 1892 V 1, 129, 263; Neuere Sprachen, 1897-98 V,
Beiblatt, 1.
A-s
p^t:
U-.^
m
'i
292
PRODUCTION OF SPEECH
and neck gave sounds resembling different whispered rowels.
The ratio between the two resonances required to pro-
duce a vowel was termed its ' radical ratio.' The results
were as follows (the letters of the notation are to be
understood as arbitrarily indicating the vowels in the key
words) :
■ Sound resembling vowel whispered in : bean bin pity Welsh un Fr. d^
Radical ratio : 37 31 29 23 19
Notation: Ji ig ig 14 e;
Fr. maison
17
62
Fr. bete
13
men
11
64
man
7
father
5
foil foal
3 2
3 O
book
H
pool
1
"2
The vowel sound occurred whenever these two ratios were
present, regardless of what the actual notes were. For a
lower resonance tone of c" the vowel -like sounds would have
tones ^ as in the following notation :
tt^tt^^^t.
t t
^:
^
-i-^t-^^-l:— -t-— -t^-t
rH
The vowel character is said to depend primarily on the
' radical ratio, ' although the vowels actually produced have
normally certain definite ranges for their tones.
The results may have some application to the tones of whis-
pered vowels, depending on the closeness of the imitation.
They give no information concerning the tones of spoken and
sung vowels, as such sounds were not produced. It is
probably true that in each vowel a certain relation of cavity
' ViETOR, Elemente d. Phonetik, 4. Aufl., 34, Leipzig, 1898.
TONES OF THE VOCAL CAVITIES 293
tones occurs; this is the supposition of the Wheatstonb-
Helmholtz overtone theory. Some of these relations have
been determined by Pipping and Hermann (pp. 21, 23, 48);
they are utterly different from Lloyd's ratios. It is not true
that the cavity tones may be of any pitch, as has been abun-
dantly shown by the results described in Part I.
Another method consists in using puffs of air to arouse
adjustable cavities. A carefully trued siren disc (p. 90) runs
in a slit in a coupling connecting tlie two ends of opposite
portions of a blast pipe; the adjustments must be so accurate
that little air escapes. The blast is brought to some musical
instrument, such as a flageolet, acting as a cavity giving
musical tones. The puffs of air, like those from the vocal
cushions (p. 257), blow the pipe intermittently, producing
resonance effects like those of magnetic impulses on a damped
spring (p. 7) and thus imitating the vocal action (p. 260).
The blast may be divided into two portions by a Y-tube and
two instruments used at the same time; or two independ-
ent blasts may be arranged at different points on the siren.
The investigations with this instrument have not yet been
completed.
The analytic method has been applied in two ways.
An analysis has been attempted by singing a tone before
a series of resonators in the manner mentioned on p. 73.
Some approximate success might be attained for a tone
sung with perfect constancy on a given note, a condition
that can at best be satisfied with only a fair degree of accu-
racy (p. 269). The resonators, however, respond somewhat
to other tones than their own (p. 73).
The analysis by means of registered curves of speech has
been described in Part I ; at the present time it is, in some
of its forms, the only method whose results can be trusted.
The tones for the sung vowels, in as far as determined by this
method, have been given above (Swedish, p. 21; Finnish,
p. 22; Russian, p. 25; American, pp. 28, 50; German, p. 48).
These tones represent the lowest ones of the system of cavi-
ties ; the higher tones are still undiscovered. Most of these
294 PRODUCTION OF SPEECH
lowest tones are probably mouth tones in the sense that they
are specially sensitive to modifications of the oropharyngeal
cavity with its various apertures and necks. Some, however,
may be trachea tones. Pipping^ considered as chest tones
the low ones found in the neighborhood of the note 250 in a
series of Finnish vowels (p. 22). The lower resonance tone
of constant pitch found in a number of cases of a in ai (' I,
eye,' etc.) may possibly arise from the chest instead of the
mouth and pharynx ; in seven recorded cases ^ the frequencies
were as follows: 'I,' 286; 'I,' 286; 'I,' 286; 'I,' 286; 'I,'
360 ; ' eye,' 435 ; ' fly,' 256. The resonating chamber for these
tones can hardly include the lungs, as the lung capacity is
undergoing continual change during respiration. The trachea
and the bronchial tubes with their hard walls seem better
suited for resonance and are more appropriate in size ; their
capacity remains approximately constant ; their accordion-like
structure permits minor adjustments. The size of the chest
cavity has been shown to vary as the musical scale is sung.^
This was supposed to occur for the purpose of reinforcing
the cord tone by resonance (p. 13).
The attempt has been made to calculate the resonance tones
for the vowels from maps of the mouth-pharynx cavity.
Lloyd gives* the following frequencies: marme 2816, pit
2500, rein 2112, there 1508, man 1431, father 1082, law 834,
note 623-444, pwt 528, brwte 314-287. So little is known
concerning the resonances of a compound cavity and the
methods of mouth-mapping are still so crude that at present
no reliance can be placed on such calculations.
The cavity tones of the consonants have been studied only
by Hermann (p. 43). The acoustic characters of conso-
1 Pipping, Zur Phonetik d. finnischen Sprache, Mem. de la Soc. finno-ougri-
enne, XIV, Helsingfors, 1899.
2 Scripture, Researches in experimental phonetics (first series), Stud. Yale Psych.
Lab., 1899 VII 6i.
^ Sewall and Pollard, On the relations of diaphragmatic and costal respira-
tion, with particular reference to phonation, Jour. Physiol., 1890 XI 159.
■* Lloyd, The interpretation of the phonograms of vowels. Jour. Anat. Physiol.,
1897 XXXI 251; On consonant sounds, Proc. Roy. Soc. Ediu., 1897-98 XXII 241.
TONES OF THE VOCAL CAVITIES 295
nants depend largely on the tones in their explosions or
noises; the relations of these tones to their modes of articu-
lation have been observed but not experimentally recorded.
The cavity tones in a vocal sound probably always in-
clude more than two tones. These tones change with the
constantly changing shapes of the cavities and probably
never remain constant. The cavity tones with the cord
tone form a more or less musical harmony which changes by
more or less sudden gradations at each instant. The simul-
taneous and successive relations of harmony among these
tones determine the character of the sounds spoken.
Certain differences in acoustic character run through
the speech of a person; other differences run through a
family; others through a community, a dialect, a region, or
a country. These differences arise partly from the different
forms of the cord vibrations (p. 264), from the different sys-
tems of tones from the vocal cavities, and from the courses of
each of these tones in the vocal harmony.
References
For the phenomena of resonance : see References to Chap. I. For
summary of data on vowel-resonance : Vietor, Elemente d. Phonetik,
4. Aufl., 27, Leipzig, 1898.
CHAPTER XXI
TONGUE CONTACTS: METHODS OF PALATOGEAPHY ; AMEEI-
CAN, lEISH AND HTJNGAEIAN EECOEDS
Foe describing the positions of the tongue different terms
for the various regions have come into use. A convenient
arrangement is that of Lenz^ given in Figs. 134, 135. The
letters on the diagram in Fig.
135 are those used by Jes-
PEESEN.^ A diagram showing
the vertical section through
the vocal cavities maybe called
a ' sagittal diagram. ' The dia-
gram recording the contact of
the tongue with the palate is
called a 'palatogram.'
The contacts of the tongue
with the roof of the mouth
are stated in compound terms,
in which the first part indi-
cates the main portion of the
tongue that touches, and the
second the main portion touched. The tongue articulations are
classed as ' dorsal ' (referring to the top of the body of the
tongue) and ' marginal, ' with subdivision of the latter into
' frontal ' and ' lateral. ' An articulation of the extreme point
is often termed ' apical.' The roof articulations are indicated
^ Lenz, Zur Physiol, u. Gesch. d. Palatalen, Diss., Bonn, 1887 ; also in Zt. f.
vergl. Sprach., 1888 XXIX 1.
^ Jespeesen, The Articulations of Speecli Sounds represented by Analpha-
betic Symbols, Marburg, 1889.
Fig. 134.
PALATOGRAPHY
297
by the names, 'dental, alveolar, pre-, medio-, postpalatal,
pre-, postvelar, uvular and pharyngeal.' Fig. 135 indicates a
dorsal-postpalatal articulation. The term ' cacuminal ' (or
' cerebral, ' or ' inverted ') is applied to a frontal articula-
tion, in which the point of the tongue is turned up and back.
The frontal-prepalatal articulation is thus a cacuminal one.
Cacuminal articulations are found in various Dravidian and
Sanskrit sounds, in some pronunciations of the English r,
in several French and Swedish dialects, etc. The terms
' lateral ' and ' central ' refer to the openings between the
Fig. 135.
tongue and the surfaces of the mouth. A lateral opening
occurs, for example, in 1.
For describing the ■ positions of the tongue in forming
vowels several general terms have come into use. ' Front,
mixed, back ' or ' anterior, neutral, posterior ' indicate that
the tongue is raised toward the front, middle or back part of
the palate. ' High, mid, low ' refer to the degree of eleva-
tion. ' Narrow ' or ' close ' indicates that ' the tongue and
flexible parts of the mouth are made tense and convex in
shape; ' ' wide ' or ' open ' indicates that they are flattened. ^
1 Sweet, The History of Language, 17, London, 1900.
298
PRODUCTION OF SPEECH
The points at which the tongue touches the palate (and, to
some degree, the velum) in forming sounds can be registered
by a mixture of meal and mucilage ^ or by carmine water color
or Chinese ink ^ spread over the previously dried tongue. The
sound is spoken naturally; the mouth is at once opened and
the marks on the palate are observed with a small dental or
laryngeal mirror in the mouth and a larger mirror in front.
The results obtained are called ' palatograms. '
This method has developed into the use of a thin shell-like
' artificial palate. ' ^ It is covered with chalk and placed in
the mouth; after the speech sound
is made, it is removed and examined
at leisure. Kingsley's palate is
shown in Fig. 136. Owing to the
cutting away of some of the sides of
the posterior velar portion parts of
the articulation are lost in some
peech sounds ; usually the artificial
1 lalate is still further limited by be-
ag cut off at the last teeth. The
results of an experiment may be
marked on a plaster cast* (Fig-
137), drawn on a diagram, or photo-
graphed. °
A cast of the palate may be made
either with dental modeling com-
pound or with plaster of Paris.
A portion of the modeling compound is held in hot water till
softened; it is then placed in a dentist's mouth tray, or, if
1 Coles, Trans. Odontolog. Soc. Grt. Britain, 1871 n. s, IS' 110.
'^ Gkutznek, Phi/siologie d. Stimme u. Sprache, Hermann's Handbuch d.
Physiol., I (2) 204, Leipzig, 1879.
2 KiNGSLEY, Illustrations of the articulations of the tongue, On Oral Deformi-
ties,London, 1880; also in Internal. Zt. f. allg. Sprachwissenscliaft, 1887 III 225 ;
Balassa, Phonetik d. ungarischen Sprache, Internal. Zt. f. allg. Sprachwiss., 1889
IV 130.
* KiNGSLEY, as before.
^ Hagelin, Stomatoskopiska undersokningar af franska sprukljud, Stock-
holm, 1889.
Fig. 136.
PALATOGRAPHY
299
that is lacking, on the end of a wide flat stick. The operator,
standing behind, bends back the head of the subject and
inserts the soft compound into his open mouth, pushing it up
firmly against the palate. It is kept against the palate until
fairly hard, then removed by loosening first at the back, and
dipped in cold water to completely harden it. A plaster
cast is now to be made from the form thus obtained. This is
rubbed with soapy water and surrounded by a wall of clay
or wax. The cup-like dish thus obtained is filled with
water. Fresh plaster of Paris, mixed with water to a rather
thin paste, is poured in this dish ;
it is allowed to remain till it
hardens, which it should do in
about 20 minutes. On removing
the wax by softening in warm
water, a model of the palate will
be found.
If plaster of Paris is used in-
stead of modeling wax for the
original impression, it is mixed
to the consistency of batter. The
tray is filled. The subject bends
the head forward. Standing on
one side, the operator pushes the
tray into the mouth and against
the palate. It is held in place
till the plaster feels hard in the mouth. The tray is then
rocked slightly till the plaster is loosened from the teeth.
The surface of the negative cast thus obtained is colored
with ink or dye and then covered with sandarac varnish.
The positive cast is then made from this just as from that in
modeling compound. The negative cast is removed from the
positive by chipping it off till the colored surface appears.
The surface of the model thus obtained is rendered ad-
herent by plunging it into a bath of stearine or wax ; it is
then rubbed with a stiff brush dipped in powdered graphite
until it is entirely coated. A saturated solution of copper
Fig. 137.
300 PRODUCTION OF SPEECH
sulphate in water with 10% of sulphuric acid is poured into a
jar. In this a porous battery cup is placed; the cup con-
tains a piece of zinc in pure water. A wire from the zinc
supports the plaster model in the copper sulphate solution;
this wire must make electric contact with the graphite coat-
ing. After the copper begins to deposit on the surface a few
drops of sulphuric acid are added to the water around the
zinc. The thickness of the deposit of copper is examined
from time to time; it should be sufficiently thick in about
twelve hours. It is then detached. It is blackened by
boiling it in a solution of sodium sulphite or by covering it
with black Japan varnish.
It is generally preferable to have a dentist make a cast and
a thin plate of metaP or celluloid^ in his own way.
An artificial palate may be made in a simpler fashion
by using thin tough filter paper. ^ A drop of oil is poured
on the mold<; a sheet of filter paper wet with water is
applied to it carefully, — tearing rather than folding it if
necessary. A paste is made of chalk powder and strong
liquid glue or cement without a bad taste ; a thin layer is
spread over the paper. Another sheet of paper (wet, if pos-
sible ; dry, if time presses) is put over the first and pressed
into the depressions carefully with the fingers and a small
blunt stick. The whole is set aside to dry. When half dry,
it is well to press it again carefully into the mold. When
fully dry (half a day or more) it is coated with black enamel
or varnish. These paper palates cannot be used long with-
out being dried over a fire or a flame. They can be rendered
waterproof by pouring oil on them when half dry.*
The quickest method ^ — often the only one possible in
traveling — is to use a sheet of tinfoil of 0.2™™ thickness or
1 Hagelin, as before.
2 ViETOR, Kleine Beitrage zur ExperimeidalphoneUk, Neuere Sprachen, 1894
I Suppl. 35.
3 RoussELOT, Principes de phoue'tique experimentale, 57, Paris, 1897.
* JossELYN, £t'ide sar la phonetique itaiienne, 2, Thfese, Paris, 1900; also in
La Parole, 1900 II 422.
' RoussELOT, as before, 58.
PALATOGRAPHY 301
two sheets united by a flexible rubber varnish. If one side
has previously been covered with this varnish, it suffices to
press it against the palate at the time of the experiment with
the thumb or a slightly pointed stick ; the . sheet is then
removed, and the edges cut off around the teeth. To make
it adhere more strongly to the palate, the sheet may be coated
on its upper side with strong paste.
For an experiment the inner surface of the artificial palate
is oiled, and sprinkled with powdered chalk or some similar
substance ; it is inserted with chalked fingers ; the sound is
spoken ; and the palate is at once removed. The parts
touched by the tongue appear black, the chalk having been
removed where the tongue pressed heavily and moistened
where it touched lightly. The result may be sketched on
a plaster cast made for the purpose, or may be photographed.
A convenient method of recording results is to sketch them
on a diagram of the person's palate printed from a zinc block
made from a drawing showing the outline of the artificial
palate.
It must be constantly borne in mind that contacts may
occur outside the limits of the artificial palate. The plate of
KiNGSLEY loses at the sides of the velum and those of most
other investigators lose everything back of the hard palate.
Techmbk's plate included the sides of the velum.
To obtain records for a sound in the interior of a word,
such words should be selected as give no other records or no
records which can be confused with the one desired.^ Thus
the 1 gives clear traces in the front of the palate for ' blanc '
(tonic), ' blanchir ' (initial atonic), ' blanchisseuse en gros '
{initial tonic distinct from accent of phrase), 'va chez M.
Blanc ' (tonic preceded by a number of syllables).
To compare analogous sounds they are spoken in analogous
words and the tracings are superimposed. For example, the
record for the vowel in ' long ' is compared with those for the
vowels in log, log, lag in order to 'determine its character.
Sounds may be tested by their effects on known articula-
1 RonsSELOT, Etudes de prononciations Parisiennes, La Parole, 1899 I 547.
302 PRODUCTION OF SPEECH
tions of other sounds. Thus anterior vowels tend to make
the k mouill^ ; tracings for various forms of kav show ^
that aj makes the k mouill^ and is distinct from a^ and
Eg (the inferior numerals indicate contacts of different
backwardness).
Palatograms of some of his American sounds have been
given by Kingslby.
The contact surfaces were recorded for the vowels e in
ken 'cane' (Fig. 138), and i in si 'see' (Fig. 139). The
tongue evidently divides the mouth cavity into two por-
tions connected by a neck. For i the anterior portion is
smaller than for e and the neck is narrower and longer; the
posterior portion is apparently not greatly changed. Con-
cerning the tones of the cavities under such conditions we
can hardly say more than that at least one of the tones for i
will be higher than the corresponding one for e ; this is actu-
ally the case in all the experimental determinations (pp. 21,
23, 25, 26, 48, 287, 288). The contact for t (Fig. 140) shows
complete alveolar contact of the tongue. If this contact was
marginal, the release was probably quick with a sharp explo-
sion ; if predorsal, slower with a slight following aspiration.
The mouth cavity was apparently large ; the tone of the ex-
plosion of t has been registered only for a German example
(p. 48) ; no comparison can be made with this t as the contact
for Hermann's t was not recorded. The n (Fig. 141) shows
complete alveolar contact ; although the mouth cavitj' ap-
pears to be nearly the same in size as that of the t, the open-
ing of the nasal aperture must lower its tone. The k is
postvelar (Fig. 142) ; the g is also postvelar but slightly
further forward (Fig. 143) ; the projection to the rear in the
middle of Kingsley's artificial palate (Fig. 136) rendered it
possible to record such backward contacts ; the contacts prob-
ably extended further to the sides than in Figs. 142 and 143
but could not be recorded on account of the narrowness of
the projection ; concerning the nature of the release and the
tones of the explosions of the k and g nothing is known. A
' RoDSSELOT, as before, 548.
AMERICAN PALATOGRAMS
303
somewhat more forward velar contact occurs in r\ (Fig. 144) ;
pharyngeal and tracheal cavity tones were presumably present.
The backwardness of the contacts for k, g, -q is remarkable
when the palatograms are compared with those of other
languages. For f and v the tongue seems to rise slightly
Fig. 138. Pig. 139. Fig. 140. Fig. 141. Fig. 142.
Fig. 143. Fig. 144. Fig. 145. Fig. 146.
Jii| jfc H^ /^
Fig. 147.
Fig. 148.
Fig. 149.
Fig. 150.
in the postpalatal and prevelar regions (Fig. 145) ; similar
records may be occasionally found in the work of other ob-
servers. This rise presumably influences the cavity tones
aroused by the friction at the aperture of the cavity and
heard in the fricative noise ; in the case of v it must also
influence the cavity tone aroused by the cord tone.
304 PRODUCTION OF SPEECH
Kingsley's s (Fig. 146) shows a very short narrow nozzle-
like opening with a gradual approach and a very free exit ;
the laws governing the action of such apertures have been
treated in works in hydrodynamics, but their acoustic appli-
cations have not been made. The record for Kingsley's
s and z was so nearly identical with' that for e (Fig. 138)
that the same diagram was used for both ; the nozzle is broad
and the cavities large, indicating perhaps a soft rushing
sound with low tones.
The record for 9 (Fig. 147) shows firm contact at the sides
and loose contact in front. The record for c, J (Fig. 148)
is that for the sounds heard in ' church ' and ' judge. ' These
consist of an occlusive t- or d-sound with a fricative release
producing a rushing sound instead of the explosive release of
an ordinary t or d. It is customary to assume that these are
consonant diphthongs and to indicate them by ts and dz. It
is quite possible, however, that the fricative release may not
be of the character supposed; moreover, the occlusive and
fricative elements may be too closely fused to permit us
to consider the sounds as diphthongs. Similar cases will be
found in the following chapters.
The 1 (Fig. 149) shows frontal-alveolar contact; the rear
portion of the tongue is generally supposed to be raised but
no record appears. In r (Fig. 150) the front portion of the
tongue is turned up against the palate.
It must be added that the identity of the contacts for t and
d, f and v, s and z, s and z, c and J, is due to the failure of
KiNGSLEY to distinguish the finer differences. In general
the surd has a more extended contact than the corresponding
sonant. This indicates stronger muscular action, as would be
expected from the fact that part of the lung pressure is used
to make the cords vibrate in a sonant (p. 244). This rela-
tion of corresponding surd and sonant as strong and weak
is a general one.
Some palatogralns by Rousselot ^ of the sounds of an
^ Rousselot, L'enseignement de la prononciation par la vue, La Parole, 1901
III 587.
IRISH PALATOGRAMS 305
American showed results in general like those by Kingsley
but with several exceptions. The relations of contact were
6 > ?5, c > J as usual, but t<d, s<z, s<z contrary to
the general rule. Rousselot believes that in such a case the
surds t, s, s are made as whispered sounds and not surd ones.
He apparently implies that the closure of the glottis for
whispering uses some of the lung pressure and requires less
closure in the mouth. This is a valid explanation for less
pressure in whispered t than in surd t ; it is inadequate for
less pressure in whispered t than in sonant d, for the vibrat-
ing cords in d let less air pass than the narrowed glottis in
whispered t.
Palatograms of a native of Neale, County Mayo, Ire-
land, ^ showed many ' subpalatal ' vowels, that is, vowels in
which the elevation of the tongue was too feeble to reach the
artificial palate or too far back to reach its rear boundary;
such were a, e (generally), u, o, a. The consonants p, b, m,
f, V were likewise subpalatal. The two classes of vowels are
much more clearly separated in Irish than, for example, in
French, where all vowels leave traces on the artificial palate.
The action of the broad Irish vowels a, o, u is quite distinct
from that of the thin ones e, i in palatalizing the preced-
ing consonants; the phenomenon occurs for e and i with a
regularity that does not appear in most other languages. In
addition to the data of special interest in regard to Irish,
Rousselot points out some that are of general importance :
1. reciprocal influence of vowels and consonants ; 2. influence
of syntactic groupings or of morphology on the articulations ;
3. great variations in the articulations of sound without loss
of auditory identity; 4. differences of force between initial
and final consonants, and between a final consonant followed
by a vowel and one followed by a consonant in the next word ;
5. the existence of k, g, t, d and s mouilM in Irish also.
Using Kingslby's method, Balassa gives diagrams of
his own Hungarian sounds. ^
1 Rousselot, Les articulations irlandaises, La Parole, 1899 I 241.
2 Balassa, Phonetik d. ungarischen Sprache, Internat. Zt. f. allg. Sprachwiss,
1889 IV 130.
20
306
PRODUCTION OF SPEECH
u 62
Fig. 151. Fig. l.'J2.
1 oe
Fig. 153. Fig. 154.
y
Fig. 155.
Fig. 157.
n
Fig. 159.
J S S 1
Fig. 160. Fig. 161. Fig. 162. Fig. 163.
Fig. 165.
HUNGARIAN PALATOGRAMS 307
No contacts appeared for aj ^ydvnd; ' a^ 'alma ' [aj and a^
having the same tongue position, with lip retraction for aj and
lip projection for ag] ; Oj ' tJta; ' O2 ' okos; ' Cj ' este; ' f '/a; '
V ' v^t: ; ' p ' joiros ; ' b ' bov ; ' m ' ?MOSt, ' ' hamvad ; ' t|
' hang; ' r ' va?-. '
Figures 151 to 167 give the contacts for the other long
vowels, liquids and surds: u ' iit; ' e^^ elet; ' i ' tiz; ' ce
' szollo; ' y ' hw,' kj ' a^aro/c; ' k^ ' ^ev^s; ' t " ty\x\; ' medio-
palatal n ' nyvl ; ' ] ' ha/o ; ' s ' sz&iaz ; ' s ' sas ; ' 1 ' /6 ; ' n
' n^p ;' t ' ie ; ' ts ' apacza ; ' ts ' csdszar. ' The short vowels
have slightly different contacts ; the sonants were assumed to
have nearly the same contacts as the corresponding surds.
The palatograms clearly indicate a classification into rear
and front vowels, usually called ' deep ' and ' high ' on ac-
count of the difference in pitch. These two classes, a, o,
u, and e, i, ce, y, show themselves in the phenomena of 'vowel
harmony ' (p. 121). The middle vowels are lacking entirely
in modern Hungarian.
The rearwardness of the 1 is noteworthy. After t, d the
1-contact is made at the same place as for those sounds. The
ancient Z-mouill^ has become j or z in most dialects, 1 in the
others.
In the ' consonant diphthongs ' ts ts the first portion is
not the same as the ordinary t, but an occlusion at the
places for the second portion. They are affricates, that is,
occlusives with fricative releases. According to the palato-
grams the former seems to be an affricated t, the latter an
affricated t. The ts is said by Balassa to be the same as
Kingsley's c, but Fig. 148 has hardly any resemblance to
Fig. 167.
References
For tongue action in English : Grandgent, German and English
Sounds, Boston, 1892 ; Soamks, Introduction to the Study of Phonetics
(English, French, German), 2d ed., London, 1899; Sweet, Primer of
Phonetics, Oxfox-d, 1890 ; Vietor, Elemente d. Phonetik, 4. Aufl., Leip-
zig, 1898.
For dental materials : S. S. White Dental Mfg. Co., New York.
CHAPTER XXII
TONGUE CONTACTS : GERMAN RECORDS
Diagrams of their own palatal records for German sounds
have been given by GrtjtznerI (jn,_ 1347 at Festenberg in
Kreis Polnisch-Wartenberg, Schlesien), Techmer,^ Vietor^
(a native of Nassau), and Lenz.*
Geutzner's diagrams (painted tongue) of continuously
produced sounds are given in Figs. 168 to 172. For 1 (Fig.
168) the tongue is pressed against the palate just above the
front and side teeth, with a small opening opposite the first
molar on each side. The velum closes the nasal cavity.
The auditory character of the 1 changes only slowly as the
AAAAA
1 r t s s
Fig. 168. Fig. 169. Fig. 170. Fig. 171. Fig. 172.
region of articulation is advanced or retracted. The opening
in 1 may be on one side only. The curves of the German
1 have been given by Wendbler (p. 19) and Hermann
(p. 43). Grutzner's articulation for r is given in Fig.
1 Grutzner, Physiologic der Stimme und Sprache, Hermann's Handbuch der
Physiologie, I (2), 204, 207, 219, 221, Leipzig, 1879.
2 Techmer, Phonetik, 30, Tafeln III-IV, Leipzig, 1880 ; Naturwiss. Analyse
u. Synthese d. horbaren Sprache, Interiiat. Zt. f. allg. Sprachwissensehaft, 1884 I,
140, Tafeln III-IV.
2 ViETOR, Elemente der Phonetik, 4. Aufl., 307, 308, Leipzig, 1898.
* Lenz, Zur Physioloi/ie nnd Geschtchte der Palatalen, Diss., Bonn, 1887 ; also
in Zt. f. vergl. Sprachf., 1888 XXIX 1.
GERMAN PALATOGRAMS 309
169. The curves for r have been recorded by Bonders,
Wendelee (p. 19) and Hermann (p. 44). Grutzner's t
(Fig. 170) is alveolar and dental -at its moment of closure.
His s (Fig. 171) shows the touching of the tongue against the
teeth and alveolae with the narrow opening in front. The
channel is very small. Getjtzner's s (Fig. 172) is one of a
possible series of rush sounds that begins with s and changes
as the tongue articulates further backward. The lower pitch
of s as compared with s was observed by Kempelen. Curves
of the vibrations in t, s, s and their corresponding sonants
have been obtained by Hermann (p. 42).
Vietor's diagrams (artificial palate) of continuously pro-
duced sounds are given in Figs. 173 to 183, which contain
also lines indicating the positions for the related American
and French sounds as obtained by Kingsley and Rousselot.
Vietor's close i (Fig. 173), as in libn ' lieben ' is
formed with the tongue so high in the middle that the
breath makes a rushing noise; the lips are not generally
drawn back. His y (-it) (Fig. 174), as in ybri§ ' iibrig, ' has
a tongue position not quite identical with that of i, with
the lips not in the neutral i position but in the projected
and rounded u position. For his Cj (Fig. 175), as in re ' reh, '
the tongue does not rise so high as in the previous cases. His
oe (6) (Fig. 176), as in seen ' schon, ' combines an approxi-
mate Cj position of the tongue with an o position of the lips.
The 62 (Fig. 177), as in ber 'bar' in his pronunciation,
has a still larger opening. Figs. 178 to 183 give the conso-
nant articulations for c as in ig ' ich, ' s as in ist ' ist, ' s as
in seen 'schon, ' t as in ton 'ton, ' r (lingual) as in rabQ 'rabe, '
and 1 as in libg 'liebe.'
Vietor's g (Fig. 178) may be considered as a development
in the series s — s — 9 by regression and lengthening of the
tongue articulation. The transversal extent of the articula-
tion in 9 varies with the preceding vowel. ^
In Vietor's s (Fig. 179) the narrow stream of air passes
out through a partially dorsal contact of the tongue, differing
1 ViETOK, Elemente d. Phonetik, 4. Aufl., 174, Leipzig, 1898.
310
PRODUCTION OF SPEECH
r
Fig. 182.
1
Fig. 18.3.
GERMAN PALATOGRAMS 311
from the apical contact of Geutznee (Fig. 171). The hiss-
ing noise of the s is probably made by the friction in the
passage and not by the impact against the teeth as Vietok
supposes. In Vietoe's s (Fig. 180) a broad stream of air
passes between the tongue and the gums ; the articulation is
dorsal-alveolar or dorsal-postdental. It differs little from
that of Geutznee (Fig. 172).
In Vietoe's t (Fig. 181) the articulation closely resembles
that for s without the opening; the pressure is strongest
in front of and around the notch shown in the figure.
Vietoe's r (Fig. 182) has an alveolar contact with the
front of the tongue further back than Geutznee 's (Fig.
169). It is generally sonant with no rushing noise; before
or after surds the cord tone is often partly or entirely lost.
The number of beats is variable.
\ For 1 the contact is alveolar and along the front teeth
(Fig. 183); it is firmest along the molars, lighter in front;
the chief opening is at the eye teeth and the first molars. ^
Refeeences
For tongue action in German : Bremer, Deutsche Phonetik, Leipzig;
1893 ; Brucke, Grundziige d. Physiol, u. Systematik d. Sprachlaute,
Wien, 1855; 2. Aufl., 1876; Grandgent, German and English Sounds,
Boston, 1892 ; Klinghardt, Artikulations- und Horiibungen, Kothen,
1897 ; Merkel, Physiol, d. menschl. Sprache, Leipzig, 1866 ; Sievers,
Grundziige d. Phonetik, 5. Aufl., Leipzig, 1901 ; Vietor, Elemente d.
Phonetik, 4. Aufl., Leipzig, 1898.
1 ViETOE, as before, 214.
CHAPTER XXIIl
TONGUE CONTACTS : FRENCH AND ITALIAN RECORDS
The Parisian dialect is being subjected to a careful study-
by RousSELOT ; ^ the published palatograms show the follow-
ing facts. The results are interesting in view of the growing
claim of Parisian to be considered as standard French.
In Paris three forms of a are clearly distinguished. The
' medium a ' as in ' patte ' is pronounced with the muscles of
the tongue and hps completely relaxed; the tongue is left
at rest in the mouth, the jaw is lowered, the larynx emits a
sound that resonates softly and hollowly in the cavity ; this
a is indicated in Roxjsselot's notation by a, in ours by
a^. The ' close a,' as in ' pSte ' or ' ah ! ' is pronounced
with tense muscles, the tongue drawn back and the lips
slightly contracted , it tends toward o ; it is indicated by a
(Rousselot) or bj' ag. The 'open a,' as in 'cave,' is pro-
nounced with the tongue slightly raised in the front portion
and slightly depressed in the rear, while the Ups are slightly
separated; it is indicated by a or by a^. The artificial palate
may show no records of these forms of a, but ordinarily they
appear at the back portions with the contacts increasing in
the order aj, ag, a^, as shown in Fig. 184. When a^ and a^
occur in atonic syllables they become a^ the relaxation in
force bringing an associated relaxation in articulation. When
aj receives an oratorical accent, it becomes ag, the extra force
bringing an extra effort in articulation.
Three very distinct forms of e may be recognized in Paris :
Cj (e), 'open e,' as in 'fait'; e^ (e), 'medium e,' as in 'Eh
^ Rousselot, J^tudes de prononciations parisiennes, I Les articulations ^tudi^es
a t'aide du palais artijiciel, La Parole, 1899 I 481.
FRENCH PALATOGRAMS
Fig. 184.
Fig. 185.
Fig. 186.
Fig. 188. Fig. 189.
Fig. 190.
Q^iS,
Fig. 192.
Fig. 193.
Fig. 194.
A/!\/i
Fig. 196.
Fig. 197.
Fig. 198.
313
Fig. 187.
Fig. 191.
Fig. 195.
Fig. 199.
^
Fig. 200
Fig. 201.
Fig. 202.
/^ ^
Fig. 203.
Fig. 204.
Fig. 205.
814 PRODUCTION OF SPEECH
bien ! ' ; and Cg (e) ' close e.' The ej and eg are readily isolated.
As Cg occurs only in an accented syllable before a consonant
or in an unaccented syllable, the Parisian isolates it with
difficulty ; in certain provinces Cg has an independent value.
The same conditions are found in general for the other medium
vowels ij, 72, Oj, U2, 062- The palatograms (Fig. 185) show
increasing contacts in the order e^, e^, eg. Changes occur as
the results of differences in stress just as for a. The explana-
tion for Cg -»^ 62 when unstressed is the same as for ag -^ aj.
The change ej -* e^ (open to medium) when unstressed re-
sults from the fact that the relaxation of the internal tongue
muscles along the septum produces a \videning of the tongue
and more contact along the palate. The same reasons explain
certain changes of long Cj to short Cg in speech, that is, of a
tense open vowel to a relaxed close one.
Two very different Parisian i's were found, namely, \^ (i),
a medium i, and ig (I), a close i. The ij is followed hj a con-
sonant or is unstressed, as in mi2ni2St deg bOgZ a^r^ ' ministre
des beaux arts.' Both i's appear in i2si3 ' ici,' fi^ni^ ' fini '
and in 12! 'il' and igl 'ile.' The palatograms show less con-
tact for i2 than for ig (Fig. 186). The changes ig -> 12 and
ig -^ ig are analogous to those for e (above).
Parisian speech has three forms of oe : ce^ ((^), ' open cje,'
as in ' heure ' ; cEj (ce), ' medium ce,' as in ' je parle ' ; and cEg
(cb), ' close ce,' as in ' eux.' The palatograms show increasing
contacts in the order cej, oCg, cCg (Fig. 187). The palato-
gram for cej corresponds closely to that for a^ ; those of ce.^
and CBg are quite different from that of Cj. Accented ocj regu-
larly becomes oeg when the syllable loses the accent.
The so-called ' mute e ' in French is, if sounded, oCg when
named, ceg in a phrase, and cjCj when accented. In other
investigations Roussblot indicates this sound by a special
symbol corresponding to a of this book. As a final, the
' mute e ' does not entirely disappear ; it at least modifies
the preceding consonant.
The two forms of y are y^ (m), ' medium y,' as in ' pudeur ' ;
and yg (i<), ' close y,' as in ' pur ' (Fig. 188). The y contacts
FRENCH PALATOGRAMS 315
occur within the region for those of e (Fig. 185), just as
do those for oe within the region between a and e.
There are the three forms of o : Oj (o), ' open o,' as in
'or'; 02 (0), 'medium o,' as in 'botte '; O3 (<?), -close o,' as
in 'beau.' Their contacts do not differ greatly. The ex-
tremes Oj and 03 are shown in Fig. 189; O2 is seldom distin-
guishable- from the others except in combinations.
There are two fonns of u : Ug ' medium u,' as in ' boule ' ;
Ug ' close u,' as in ' cou.' The palatograms (Fig. 190) — not
from the same subject as for Figs. 184 to 189, owing to an
anomaly for u — show more contact for Ug than for Ug.
The nasal vowels rarely have the same contacts as the
corresponding oral ones. The palatograms for the same
subject that furnished Figs. 184 to 189 showed the contacts
for oe", o", a", e" (Fig. 191) all to lie in the oe-region (Fig.
187). The records seem to indicate correspondences of
tongue position between ce" and oBj, o" and Oj, a" and Eg, e"
and an intermediary between a^ and ej.
The contacts for k and g show many varieties of a typical
form (Figs. 192, 193), depending on the adjacent sounds.
The K and \ sounds are those known as ' k- and ^-mouill^.'
Their contacts are further forward than those for k and g
even in the combination kj and gj ; comparisons are shown in
Figs. 192 and 193. These k and \ sounds are distinct from
the k and g sounds both in place of articulation and in the
nature of the explosive release.^ The treatment of k and \
as ' soft ' or ' mouiU^ ' forms of k and g arises from their
interchangeability in French speech and from the use of the
same letters to indicate them. The ' mouillure of k and g,''
that is, the change of k and g to k and \, occurred with one
subject before i^, e^, ce,^, aj but not constantly ; it happened
generally only in a moment of negligence, and in rapid rather
than slow speech. Two examples of constant occurrence
were loCg majr^Kig d ka^rgajbag ' le marquis de Carabas '■ and
loe^ sOjkOglag majfgKig 'le chocolat Marquis ' (r^ = uvula r).
The contact for j (consonant i as in ' yeux ') is shown in
^ Lenz, Zur Physiotoyie u. Geschichte d. Patatalen, Diss., Bonn, 1887 ; also in
Zt. f. vergl. Spraehf., 1888 XXIX 1.
316 PRODUCTION OF SPEECH
Fig. 194. The sound which has replaced the Z-mouilM has
contacts as shown in Fig. 195 in baje 'Miller' (i) and br2i3
' brille ' (^). Its practical identity with j of Fig. 194 is evident.
The sounds s, z, s, z (Figs. 196, 197, 198) have contact
surfaces that are small in comparison with those of the corres-
ponding American and German records.
The contact surfaces for t and d in ta and da are shown in
Fig. 199 (t > d). They are somewhat more extended in te
and de and still more in ti and di, ty and dy. The sounds
i and y incite to the still greater contact that characterizes the
t- or c?-mouill^. These have the next distinguishable con-
tacts in front of k and \ with a softer release than those of
t and d ; they may be indicated by t and 6. This mouillure
is heard ordinarily in familiar words, as ' turguet,' ' naturel ' ;
it occurs constantly in the interjection ' natureknent ! '
The n contact involves an anterior occlusion (Fig. 200) ;
it depends somewhat on the following vowel (ni > ne). In
the combination nji (Fig. 201) the occlusion for n is a very
small one along the front teeth and undoubtedly along the
sides while the contact for ] Occupies considerable space at
the sides as usual.
The n, so-called 7?.-mouilld, has a contact utterly different
(Fig. 202) from that of n.
The contact of 1 varies considerably. An ordinary initial 1
is illustrated in Fig. 203, a mecUal 1 in Fig. 204. The in-
fluence of final ' mute e ' on the contact for 1 is shown in the
records (Fig. 205) for ' bal ' (/) and ' balle ' (f ).
Hagelin's excellent photographs of palatograms by several
French speakers ^ are not available for reproduction.
Rousselot's^ diagrams for the contacts of the tongue in
his native dialect (Cellefrouin in Charente) were obtained from
himself by means of observation with a mirror, by the use of
an artificial palate and by employing small rubber bulbs in
the mouth.
1 Hagelin, Stomatoskopiska undersokningar af franska sprakljud, Stock-
holm, 1889.
^ RousSELOT, Les modifications pkonit. du langage, 23, Revue des patois
gallo-romans, 1891 IV, V; also separate.
FRENCH PALATOGRAMS 317
In the fricatives z and s the tongue scarcely touches the
edges of the palat& (the outside lines in Fig. 206), less for z
than for s ; the passage for the air current is not indicated.
Rousselot's s and z are quite different from any of the
others yet given; several of Hagelix's figures show openings
more like those of the German and English diagrams. For
n, d and t the tongue touches the palate on all sides, cover-
ing more of the central portion as the contact rises from z
and s through n, d, t. French t and d are regularly dorsal -
alveolar or else frontal-postdental as given by Hagelin (see
also Fig. 199); the dorsal-palatal t and d of Rousselot
seem unusual and hard to understand when we consider the
nature of the release of such backward contacts; I am in-
clined to consider them as palatalized t and d, or the so-called
t- and c?-mouill^.
In the pairs s, z and t, d the surd has the greater contact
surface, indicating stronger articulation (p. 304).
In Rousselot's labials p, b, f, v, m, the tongue is in repose,
touching the palate with its edges at the rear teeth. The line
across each corner in Fig. 207 marks the front limit of contact
for p, b, f, V spoken with a following r or 1, and also for m.
For m there is a slight rise of the lower jaw. The differences
in articulation for the'se sounds are produced at the lips.
Rousselot's palatals j (consonant i as in ' yeux '), s (as in
'cache'), and z (as in ' je ') have side contacts as shown in Fig.
207. In the French s the tongue is placed along the teeth,
sometimes far to the front, and even all around (HiiGELiN);
the opening is quite different from that of German s (Figs.
172, 180) and the air passage is much thinner and wider.
For k and g the tongue is raised across the palate, the
marked variations seem to show the influence of the follow-
ing sounds, as may be seen in Fig. 208 from k(o), g(o), k(i),
g(i), k(]), g(]), g(l) and k(l), the line in each case marks
the front limit of contact. In general the surds have more
extensive contacts than the sonants. In kj, gj, the occlu-
sions seem rather to have become k and \. Rousselot's k
before i is quite different from the cases given by Hagelin.
318 PRODUCTION OF SPEECH
The w (Fig. 208) has about the same contact as g(o) and
the lip position of u.
For 1 (Fig. 209) and r (Fig. 210) the tongue touches the
palate in about the same regions, for 1 more than for r.
When consonants are grouped, the first consonant some-
times tends to accommodate itself to the following one. This
is not shown in Rousselot's palatograms of p(l), p(r), b(l),
b(r),.etc. (Fig. 207), but does appear in the records for p(j),
bQ), f(]), v(]) (Fig. 211) and in those for s(j), t(]) (Fig.
212); it is quite marked also for k(j), g(j), k(l) and g(l)
(Fig. 208), the k and the g contacts being much advanced
toward the lips and strongly palatalized.
The record for A, or ' Z-mouill^,' (I, Fig. 209) shows that it is
a very different sound from 1. When followed by j as in ' Ueux,'
the contact (Ij, Fig. 209) indicates an even greater contact than
for A alone although it resembles it more than that for 1.
The record for n ( w, Fig. 212), or 'w-mouilld,' shows a
contact utterly different from that for n (Fig. 206). When
followed by ], the contact (w^, Fig. 212) indicates a sound
differing somewhat from n and still more so from n, biit lying
between ii and n. The conclusion is evident that A and n
are not in any way to be considered as composed of 1 + j
and n + ], although in this dialect n was always derived
historically from nj and A partly from 1]. The records seem
to indicate, moreover, that the combinations are A] and nj
rather than 1] and nj.
The anterior limits of the regions of contact in Rousse-
lot's forms of a are given in Fig. 213; ag is the vowel a
produced by Roxjsselot with the least effort (as in ' Paris ' ) ;
a, (as in ' partir ') requires a slightly greater opening of the
mouth and retraction of the tongue; a^ (as in ' pftte ') in-
volves a still further retraction of the tongue to leave a large
cavity in front. Fig. 213 shows successively larger contacts
for the neutral and anterior vowels in the order a.^ ' pdte,'
a° ' ewfawt,' aj ' partir,' a^ ' Paris,' Cj ' f^te ' and e" ' vwi,' e^
' eglise,' Cg ' maison,' i^ ' Rivoli,' ] ' yeux,' ig ' ici.' A similar
series of contacts is seen in Fig. 214 ; the contacts increase
FRENCH PALATOGRAMS
319
F»Tr'-^d\
z, B, n, d, t
Fig. 206.
p, b, f, V, a, z, ]
Fig. '207.
k(j),S(3),k(i),s(i),s(l),
k(l),k(o),g(o),w
Fig. 208.
Fig. 209.
i -<*•
r,r(j)
Fig. 210.
P(]),b(]),fO),v(j)
Fig. 211.
Fig. 212.
as, a", ai, ^, ej, e", e^, 63, is, ], i.
Fig. 213.
oei, 0621 °e3. U> y2i 73
Fig. 214.
as, Oi, O2, O3, o", U2, Us
Fig. 215.
320 PRODUCTION OF SPEECH
in the order cEj 'hewre,' ocj ' heitreux,' oCg 'heurewx, few,' q
(w) 'liti,' 72 (i^a) 'Mtile,' yg (u^ 'fendw.' The back vowels
ag ' p<ite,' Oj ' or,' o^ ' ehocala,' O3 ' chapeait ' and 0° ' on,' Ug
'boMche,' Ug 'OM,' form another series of steadily increasing
contacts (Fig. 215).
Comparison of Rousselot's own records with the Parisian
ones shows instructive resemblances and differences. The order
of increasing contact is ag, aj, a^ (Fig. 213) instead of a^, ag,
ag (Fig. 184). For e (Fig. 213) the order is the same (Fig.
185) but the contacts are all further back; his e sounds are
evidently all more open and nearer to the a sounds than the
Parisian ones are. For i (Fig. 213) the records are closely
similar (Fig. 186). The same is true for ce (Fig. 214 and Fig.
187). The contacts for y (Fig. 214, u^, Wg) occur in the
region for i (Fig. 213) and not in that for e (Fig. 213) as in
the Parisian records (Figs. 188, 185, 186). The contacts for o
and u (Fig. 215) are practically the same as the Parisian ones
(Figs. 189, 190). The nasal vowels show the following rela-
tions of tongue contact (js' indicates ' approximates ') : a°^ ag
(Fig. 213), e^cK" Bj (Fig. 21 3) ; ce" is not given ; the palatograms
give o" = Og but the graphic records of tongue elevation and
pressure (Figs. 256, 259) indicate o°JK'Oj. The records for
n, s, z, s, z, ], k, g, 1, show general agreement. The record for
' ?-mouilM ' (Fig. 209) shows a true X and not its usual sub-
stitute ] (Fig. 195). In the combination ' Ij ' the contact shows
that the first sound is A rather than 1. There is likewise a
true n (Fig. 212), with indication that ' nj ' is ii] rather than
nj. Rousselot's r (Fig. 210) is evidently the lingual r and
not the uvular one, or ' r grassey^.'
Palatograms taken in the province of Nivernais, France,
showed a gradual unperceived geographical and phonetic
progression from s and z at the Loire boundary to s and z
at the opposite side ; the Latin words ' capellum ' and
' gambam,' for example, having become clearly sagpjo and
zaginb at Chaulgnes and se^pjo and z&^mb at Chaumard.^
1 Meunier, Emploi de la methode qraphique, etc., La Parole, 1900 II 67.
ITALIAN PALATOGRAMS 321
Palatograms of Italian sounds i show that there are two
distinct forms of a as in SL-^ma^ (Fig. 216), three forms of e as
in kreide2re3 (Fig. 217), two of i as in iini2 (Fig. 218), three
of o as in pOipOglOg (Fig. 219), and two of u as in virtUj
and rujinore (Fig. 220).
These palatograms and all the following ones, unless
otherwise stated, are from a physician, a native of Terni
(Perugia), who, after living in many of the large cities of
Italy, had settled at Siena (Tuscany).
The t and d are regularly dental (Figs. 221, 222). The
palatograms for ka, ki, ko (Fig. 223) and ga, gi, go (Fig. 224)
show the different forms of k and g depending on the follow-
ing vowel; in all these forms the point of the tongue was
against the lower teeth. The records for ca and ci (Fig.
225) show that c includes the contacts for t (backward t)
and J (consonant i) and not those for t and s ; the fact seems
also to have been established that the entire contact for t -(- j
was made at the same time. According to Jossblyn it is
quite wrong to consider c as composed of the articulations t
and s, or even as composed of a succession of articulations.
The contacts for ja and ji were, for this subject, practically
the same as for ca and ci. With other subjects the ja and ji
showed a tendency toward a fricative form (Figs. 226, 227).
The contact for ts as in ' zio ' (Fig. 228) resembles that of t
(Fig. 221) but covers a smaller surface ; that for dz as in
' dozzina' is like that for ts with the tongue less firmly against
the palate in the rear. Josselyn seems to consider ts, dz to
be nearly as closely unified as c, J. For s (Fig. 229) the
contact is against the alveoles with a short opening near the
middle; for z (Fig. 230) the contact surface is slightly
less. For s (Fig. 231) the channel is very wide. The 1
(Fig. 232) involved a frontal-prepalatal contact but in an-
other subject was frontal-dental (Fig. 233). The rolled r and
fricative r did not differ in contact (Fig. 234). The almost
complete closure in the prepalatal region for the fricative r
1 JossELYN, Ji'tude sur la phanHique italienne, These, Paris, 1 900 ; also in La
Parole, 1900 II 422, 449, 673, 739; 1901 III 41.
21
322
PRODUCTION OF SPEECH
Fig. 216.
Fig. 218.
k
Fig. 223.
ITALIAN PALATOGRAMS
323
Fig. 233.
m
Fis. 235.
324
PRODUCTION OF SPEECH
may be made complete by a slight movement; in such a
case if the sides of the tongue are not sufficiently firm the
lateral escape of the air will produce an 1, or if they are firm
the velum can descend and produce an n; such phonetic
changes are common in the Romance languages. In the
change from fricative r to s the closure may readily become
complete and produce an intermediate t, as in pwo dartsi
for pwo darsi. The articulation for m (Fig. 235) is post-
palatal with occasional alveolar contact of the tongue tip.
The n is frontal-prepalatal (Fig. 236) or dental. The record
of ng in ' vengo ' shows a postpalatal ti (Fig. 237), quite
different from the prepalatal n (Fig. 236). The dorsal con-
tact for X or Z-mouill6 (1 in Fig. 238) is very different from
the frontal contact in 1] (^ in Fig. 238). There does not
appear to be so much difference in the case of n and nj (Fig.
239). The ] in jeri ' ieri ' (^ in Fig. 240) is clearly distinct
from the vowel i in io (7 in Fig. 240) although somewhat
resembling it.
Rbfeeences
For the tongue action in French : Passy, Les sons du franpais, 5™^
^d., Paris, 1899; Passy, Le fran9ais parle, 4. ed., Leipzig, 1897; Beyer,
Franzosische Phonetik, 2. Aufl., Kothen, 1897; Beyer dnd Passy, Ele-
mentarbuch des gesprochenen Franzdsisoh, Kothen, 1893 ; Borxer-
ScHMiTZ, Lehrbuch d. franz. Sprache, Leipzig, 1901 ; Vietor, Elements
der Phonetik, 4. Auii., Leipzig, 1898.
CHAPTER XXIV
TONGUE POSITIONS AND MOVEMENTS
The regulative sensations (p. 191) coming to conscious-
ness from tlie tongue are ratiier indefinite; they seem to be
derived mainly from the contacts of the mucous surfaces.
The main guidance for tongue movements is found in the
sounds heard.
In the production of a speech sound the tongue makes
more or less complicated, movements. As in the case of all
muscular action the tongue is never still and never occupies
exactly the same position for any period of time. If a cer-
tain range of variation of position is or must be considered
negligible, then it can be said to remain in a given position
for the time its movements are confined within that range.
Thus in the production of i the tongue rises and then falls;
there is no moment at which it is perfectly still. Even in
producing i for a considerable period of time the tongue is
constantly fluctuating in its position. In a spoken i it main-
tains no such position for any great length of time but
passes from the position for the previous speech element
through all the positions involved in producing i and then to
the position for the following element. The manner in which
the tongue goes through the series of changes is certainly as
characteristic of a speech movement as its position at any
moment; acoustically the changes in the rush of air and in
the cavity tones are as important as the conditions at any
one moment. The usual custom of assigning some one posi-
tion as the characteristic of a speech movement is often mis-
leading. Thus a diagram showing the point of the tongue
326 PRODUCTION OF SPEECH
pressed against the gums is described as a frontal-alveolar
articulation, and a t produced in this way is said to be frontal-
alveolar, whereas the chief characteristics of this t may lie
in the manner in which the closure is made and released.
The use of the term ' articulation ' has sometimes resulted
in a misconception of the nature of speech movements.
' Articulation ' is a term applied to the joints, whose bones
are said to articulate ; it is not applied to the movement of
the bones ; an articulation is a relatively fixed thing. The
tongue, however, does not ' articulate ' with the palate but
touches it at various points in various ways. The movement
of the tongue is the characteristic of the speech action; its
contact with the palate — • even if this be called an ' articula-
tion ' — is in both time and extent only a small portion of
the whole speech action. Phonetic writers seem to have
been confused by a quite different use of the word ' articu-
late.' In such phrases as ' he articulates distinctly,' it refers
to the intelligibility of the sounds by the ear ; this has no
reference whatever to precision of the vocal movements, but
to the likeness of the sounds produced to those we are accus-
tomed to hear. Some writers also seem to have had in mind
the incorrect theory that a spoken word consists of a series
of distinct sounds united by glides ; supposing that the dis-
tinctness of articulation (that is, for the ear) would depend
on the precision with which the separate sounds were made
and marked off, they would naturally think of some connec-
tion between the auditory articulativeness and the motor pre-
cision of movement. In this book I have, for want of a
better word, often used ' articulation ' in the usual way, but
the reader should not forget that it refers merely to vocal
movements, that it has no connection with the distinctness
of speech and that it implies no action of the organs in any
way resembling articulation in the joints.
In studying the diagrams of the positions of the tongue it
Biust be constantly borne in mind that they give only phases
of the movements in speech. These phases are usually ob-
tained by producing the sound continuously as in singing.
TONGUE POSITIONS AND MOVEMENTS 327
In song the tongue assumes fairly constant positions for
considerable lengths of time and these positions are approxi-
mately the same on different occasions. It is thus possible
to map out the positions with considerable accuracy, although
the work requires a long time.
A careful education of the sense of touch in the mouth
renders it possible to feel the movements of the tongue with
greatly increased accuracy. By repeatedly touching the sur-
faces inside the mouth with the finger the sense of location
can be made more definite. Observation in a mirror is aided
by inserting into the mouth a small incandescent lamp on
a handle.
Several observers have given diagrams of what they con-
sidered to be the positions of the tongue during speech
sounds. Among the early sets that by Mbrkel i was care-
fully obtained. His sagittal diagrams show the positions of
the tongue while he emitted various sounds continuously.
In Grandgent's ^ determinations of the mouth positions
the sound was spoken (or imagined), a ruler was inserted
to measure the distances from the upper front teeth to 1.
the rear pharyngeal arch, 2. the front pharyngeal arch, and
3. a point half-way between the latter and the rear edge
of the palate. This method was used to obtain a series
of sagittal diagrams ^ of Geandgent's sounds and of the
German sounds of Hochdoefee, a native of Magdeburg.
Geandgent's speech is a good approximation* to the Boston
dialect; but it differs considerably from other American and
English dialects. In some respects the Boston dialect has
English characteristics, using pa9, ba6, for 'path,' 'bath,'
etc. The form of Geandgent's palate is a fairly typical
one.
1 Meekel, Physiologie der menschlichen Sprache (physiologische Laletik),
Leipzig, 1866 ; this is a thorough revision of the last section of Meekel, Anatomie
und -Physiologie des menschlichen Stimm- und Sprachorgans (Anthropophonik),
Leipzig, 1. Aufl., 1857, 2. Aufl., 1863.
2 Gkandgent, Vowel measurements. Pub. Mod. Lang. Assoc, 1890 V 148.
3 Geandgent, German and English Sounds, Boston, 1892.
* Kambeau, Bemerkungen, Neuere Spracheu, 1895 II 528.
328 PRODUCTION OF SPEECH
The diagrams for Hochdorfer's sounds are given in
Plates XVII to XXII at the end of this volume, those for
Grandgent's sounds in Plates XXIII to XXVI. Each
figure for a sound includes a sagittal diagram of the mouth
cavity, a transverse diagram showing the opening between the
tongue and the roof of the mouth at its narrowest part, and
a front diagram of the position of the lips. The sounds are
indicated by key words. The following account resembles the
original in general, though differing at a few points.
The upper figure in Plate XVII shows Hochdorfer's
uvula r. In producing this sound a deep channel is formed
in the back part of the tongue, in which the uvula lies. The
breath-pressure raises it, a puff of air occurs in the mouth,
and it falls. As the vocal cords are vibrating at the same
time this produces a series of puffs of tone (p. 19). Often
only one such puff is used in speech ; the rate at which the
puffs come depends on the muscular adjustment and the
breath-pressure. The tongue position resembles that for
Hochdorfer's a in mala ' male ' (Plate XIX) to which
final r in words like bia 'bier,' vasa 'wasser, ' is actually
reduced by many Germans in ordinary conversation.
Hochdorfer's x is a hiss produced in the back of the
mouth by a broad stream of air escaping between the inner
part of the tongue and the lower edge of the velum. The
uvula rests on the tongue without vibrating. In some dia-
lects, however, it is caused to vibrate ; this makes it like the
uvula r without a tone from the cords. His 5 is a hiss
from a narrow passage between the alveolae and the fore-part
of the tongue ; the point of the tongue is against the lower
teeth. His j {j as in ja) is a sonant buzz with the tongue
closer than in 9 and with the lips less open ; the point of
the tongue is raised. His s is a dull hiss made by a broad
stream of air between the tongue and the palate, and modified
by the position of the lips and teeth. His s is a sharp hiss with
the tongue close to the teeth. The auditory characteristics
of these sounds lie in the hiss mixed with groups of tones of
different pitch. The hiss is produced mainly between the
TONGUE POSITIONS AND MOVEMENTS 329
tongue and the palate; the teeth have little effect. The
tones depend on the combinations of resonance chambers
formed in the oral cavity by the positions of the lips, tongue,
velum, jaw and larynx. The chief resonance tone can be
varied by rise and fall of the larynx without destroying the
character of the sound ; thus s can be whispered through a
range of an octave, s through a somewhat smaller range and
9 and x through a still more limited one. The acoustic curves
for these sounds from Hekmank's voice have been given on
p. 42.
Grandgbnt's r (Plate XXIII) is a slight buzz with the up-
turned tongue-point near the palate. His w position closely
resembles that of u (Plate XXIV). His ] position differs little
from that for i (Plate XXVI) ; the tongue and lip passages are
opener than for Hochdokfee's German j. Geandgbnt's s
and s have the tongue-apex raised higher than Hochdokfee's
German s and s. In pitch his s seems to the ear higher than
his s and also than Hochdoepbe's s. In Geandgbnt's 6 the
air escapes through the spaces between the upper front teeth
and between the notches of these teeth *and the edge of the
tongue. The resonance tones differ from those of f (with
the lower lip against the upper teeth) on account of the differ-
ent cavities formed by the positions of the tongue and lips.
In the back vowels u, o, o (Plates XVIII and XXIV) the
tongue rises in the back and leaves a large cavity in the front
of the mouth. Where the palate is low, this space is gained
by drawing the tongue more strongly back (Plate XVIII) than
otherwise (Plate XXIV). The rounding of the lips includes
greater projection for Hochdoe&ee than for Geandgent.
The palate and tongue positions form a regulai,iy descending
series in the order shown in these two Plates.
The vowel positions in Plates XIX and XXV have un-
rounded lips, except a in hat ' hurt ' and a in mala ' male. '
When the Germ, a is used for er, the lower jaw is somewhat
higher. These two vowels are much alike for Hochdoefee
and Geandgent both in sound and in position. It should be
noticed that the pronunciation of o in ' hot ' by Geaotdgent
330
PRODUCTION OF SPEECH
is rather like a than o. Geandgbnt's vowels in Plate XXV
differ in sound from any of Hochdoefee's.
The front vowels in Plates XX and XXVI are unrounded.
The short e and i seemed to be the same in sound for the
two speakers ; the long e and i differed slightly for the two ;
the a in ' bar ' appeared quite unlike the ae in ' bat ' and
noticeably different from the ae in 'fairy.'
The front rounded vowels of Hochdoefee did not corre-
spond to anything in English. They were produced in two
ways; the commoner method is shown in Plate XXI, the
other in Plate XXII.
To obtain the curve of the surface of the tongue several
methods have been devised by Atkinson. In one of them ^
a strip of vulcanized rub-
ber about 1"™ thick and
7mm tQ gmm ^i(jp^ softened
in boiling water, is in-
serted as far as possible
in the mouth and fixed to
the upper front teeth (Fig.
241). The desired sound
is produced ; the strip
is bent into position by
the tongue and is allowed
to harden. The cooling
may be hastened by a jet of cold water. It then shows the
curvature of the tongue and its relation to the front teeth.
The exact shape of the palate is obtained by a dental mold;
this is sawed in half to give the sagittal section; the rubber
tongue curve attached to this aids in completing the sagittal
diagram of the phase of greatest movement. Such a tongue
curve has been used in teaching vowel positions to the deaf 2
and in investigating phonetic changes. ^ An accurate instru-
1 RonsSELOT, Principes de phone'tique expe'rimentale, 278, Paris, 1897 ;
JjACLotte, AlTr6\os-BovK6xos, La Parole, 1899 I 349.
2 Meunier, Emploi de la melhode graphique pour Vgducation des sourds-
muets, La Parole, 1900 11 65.
^ Laclotte, as before.
Fig. 241.
TONGUE POSITIONS AND MOVEMENTS 331
menti for obtaining the position of a point on the surface of
the tongue is shown in Figs. 242, 243. A fine wire C slides
in a tube A, its other end being caught in the coil 3 through
a slot in A extending from E to F. When D is -dt F the
wire O is completely within A; when D is at F, it projects
Fig. 242.
4.5™". The tooth -stop Cr slides freely on A when its project-
ing end is down, but is fixed when its end is up as shown in
the figure. The index, middle and ring fingers are placed in
the handle B; the thumb is used to move D. The tooth-
stop Gr is placed in the depression between the two front teeth.
. Fig. 243.
The tube A is rested against the edge of the teeth (Fig.
243, a) and the hard palate (Fig. 243, b). The wire is pushed
out until it touches the tongue, which is held in the desired
position. The instrument is then applied to a plaster mold
1 Atkinson, Methods uf mouth-mapping , Neuere Sprachen, 1899 VI 494.
382
PRODUCTION OF SPEECH
of the mouth cut in half sagittally, and the position of the
point of 0 is marked on the diagram. The position of the
tooth-stop shown at (? in Fig. 243 slightly lowers the direc-
tion of the tube in the mouth ; that at F in Fig. 243 directs it
as shown at BB so as to measure the front cavities in back
vowels. The dotted outline G in Fig. 243 shows a modifica-
tion for measuring the position of the velum. A series of
sagittal diagrams of the sung vowels obtained with this instru-
ment is shown in Fig. 244. The
diagrams indicate the tongue posi-
tions for the following vowels :
1. i as hi bit ' beat ' ( )
and an attempt at i as in German
bitn ' bieten ' ( ) ; 2. e as
in belt ' bait ' ( ) and an
attempt at e as in French bet
' bete ' ( ) ; 3. a as in ba9
'bath'; 4. 6 as bout 'boat'
( ) and an attempt at 6 as
in German b5t ' bot ' ( ) ;
5. u as in but ' boot ' ( )
I ' ^'^ I ) /^' ^^'^ ^'^ attempt at u as in French
bu ' bout.' A comparison of these
diagrams with those of Plates
XVIII to XXVI shows in gen-
eral a closer resemblance of these
British positions to HocH-
dorfeb's German positions than
to Grandgent's American ones.
Little hollow rubber bulbs of any desired form ^ (Fig. 245)
can be introduced into the mouth to record by air transmission
(p. 195) the pressure or tenseness of the tongue at a given
point. They may be called ' exploratory bulbs. '
In Italian the relations of pressure of the tongue below the
•rear part of the palate were found ^ for one person to be
^ RoossELOT, Principes ile la phonetique experimentale, 86, Paris, 1897.
2 JosSELTN, Etude sur la phon&. ital., 40, 91, These, Paris, 1900; also in La
Parole, 1901 III 41.
J
Fig. 244,
TONGUE POSITIONS AND MOVEMENTS
333
p>b>m; also postpalatal, t>d>n>l; mediopalatal, the
same; alveolar, n>t>d. For another they were: post-
palatal, t > n > 1 =: d ; mediopalatal,
d > n > t > 1 ; alveolar, t > d > 1 > n ;
postpalatal, k>g; mediopalatal, c
slightly > J ; postpalatal, s > z ;
labial, v > f ; postpalatal, v > f ; me-
diopalatal, s>z>s; postpalatal,
p>b>m; labial, p >m>b. The
differences in the position of the
tongue and its pressure against the teeth in d and t and
their variations in different positions, as shown ^ in Fig.
246 for the words dido, tito, were recorded by a bulb behind
the teeth. The records would seem to indicate differences
among consonants usually considered the same that are not
inferior to those found among the varieties of a vowel. Such
Fig. 245.
Fig. 246.
records indicate not only differences in pressure but also in
the character and extent of the movement.
Records for the Italian vowels ^ of eight subjects showed
that the tongue did not rise so high in the niiddle of the
mouth for the first vowel as for the second in ama and
similar words, indicating the existence of
two forms of a (Fig. 247). Similar records
showed that there were three forms of e,
two of i, three of o and two of u (Figs.
248 to 250). These vowels are found
in ajinaj, krcjdegreg, i2nij, \n\^ of one person, fiini^ of
another, pOjpOglOg (oj = o?), virtUj, ru2more ; their palat-
FlG. 247.
1 JosSELYN, as before, 3.5.
^ JosSELTN, as before, 13.
334 PRODUCTION OF SPEECH
ograms have been given above (Figs. 216 to 220). The
elevation of the tongue throughout the word ijnij is shown in
Fig. 248.
Fig. 251. With an exploratory bulb in the rear of the mouth
the word popolo gave the tracing shown in Fig. 252. The
first p hardly appears at all; this is due to the fact that while
Fig. 249.
the articulation for p is made at the lips, the tongue has the
o position. The rise of the tongue during the second p ap-
pears clearly, and that for 1 strongly. The three o's have
Fig. 250. Fig. 251.
successively higher tongue positions. The o-l glide shows a
considerable rise of the tongue ; it is heard as a kind of u ;
the record shows it almost as a distinct sound.
The total elevation of the tongue may be conveniently
registered by turning the artificial palate into a tambour by
TONGUE POSITIONS AND MOVEMENTS
335
covering it with a sheet of thin rubber and attaching an out-
let tube.i It may be called a tongue-tambour.
Fio. 252.
The elevation of the tongue may be indicated by a receiv-
ing tambour with a knob attached so that it rests under the
soft portion of the chin ; ^ the mylohyoid, geniohyoid and
digastricus (p. 234) muscles contract to support the tongue
in action. The apparatus is shown in Fig. 253. The tam-
bour M is held in place by the set
of adjustable rods and clamps P S K
attached to the band J at J" and
by the set A B attached at C. The
knob T is moved by the contraction
of the geniohyoid and transmits the
motion in the usual way by a lever
with the fulcrum F to the rubber
top L. The screw 0 affords finer
adjustment. This apparatus does
not interfere with the resonance of
the mouth as the exploratory bulb '
(Fig. 245) does.
For the French vowels spoken
by RoussELOT^ the relative de-
grees of elevation of the tongue as indicated by the tongue-
tambour are given in Figs. 254, 255, 256, and those by the
geniohyoid tambour in Figs. 257, 258, 259.
1 RoussELOT, Les modifications phonetiques du lanqage, 11, Rev. pat. gallo-rom.
1891 IV, V; also separate.
2 RousSELOT, Les mod., as before, 1 1 ; Principes de la phonetique experi-
mentale, 95, Paris, 1897.
8 RoussEr.OT, Les mod., as before, 29.
Fig. 253.
336
PRODUCTION OF SPEECH
A series of records of Dutch sounds showed ^ for a almost
no increase of tension in the mouth floor, for u a strong
increase, for o somewhat less, for i a very great increase, for
Fig. 254
Fig. 2.55.
e somewhat less — the order of increase being thus a, o,
u, e, i; for t, d, n strong increase; for velar sounds a
relaxation.
The flapping of the tongue in producing a rolled r inter-
FlG. 256.
Fig. 257.
rupts the breath as it passes through to the mouth. A closely
fitting mouth-piece connected with a tambour (p. 219) may
be used to record the rolls or pseudobeats (pp. 19, 44).
The German r was found in one case ^ to have usually from
Fig. 258.
m^Kl^K^^E^n
Fig. 259.
20 to 35 beats a second ; in initial positions usually 8 (4 or 5
in especially distinct speech) ; medially 2 after a long vowel,
3 after a short one ; before a consonant or as a final sound
1 Gallee und Zwaaedemaker, Ueber Graphik d. Sprachlaute, Neuere
Sprachen, 1900 VIII 17.
2 Vietor, Elemente d. Phoiietik, 4 Aufl., 208, Leipzig, 1898.
TONGUE POSITIONS AND MOVEMENTS 337
only 1. Tambour records by Heemaxn ^ showed that the
frequency of the rolls varied greatly. The uvula r had often
a higher frequency than the tongue r. Wendeler ^ found
in his speech curves that for his German tongue r the
pitch of the cord note during the r had no influence on
the frequency of the roll unless the note was made loud,
and that loud cord notes raised the frequency of the roll.
Zwaakdemakee's tambour records^ led him to conclude
that for his Dutch r the period of the roll was regularly a
multiple of the period of the cord tone. Such a harmonic
relation points to an almost inconceivable preference of the
ear for a simple relation between the period of the beat,
which is distinctly heard, and the period of the physical tone-
vibrations, which are not heard separately but as a simple
sensation of tone. We cannot, however, always trust the
tambour records of the cord tone.
The tongue in action may be observed by means of the
RoNTGEN rays.*
References
For studies of tongue action : see References to the preceding
chapters.
1 Hermann, Forlgeset^te Untersurhunqen iiber d. Konsonanten, Arch. f. d. ges.
Physiol. (Pfluger), 1900 LXXXIII 12.
2 Wendeler, Ein Versuch, d. Sckallbewegunq eimqer Konsonanten u. anderer
Gemusche wit d. Hensen'schm Sprachzeichner qraphsch darzusteUen, Diss., Kiel,
1886; also in Zt. f. Biol., 1887 XXIII 303.
3 ZwAARDEMAKER, Le legistre de I'li, Archives neerlandaises des sci. exactes
etnat., 1899, (2) II 257. „, . , , c
4 SCHEIER, Die Verwerthung d. Rontgenstrahlen f. d. Physiol, d. Spraehe u.
Stimme Arch, f. Laryngol., 1898 VII 116; Ueber d. Bedeutunq d. Rbntqenstrahlen
f. d Physiol d. Spraehe u. Stimme, Neuere Sprachen, 1898 V Phonet. Stud. 40.
22
CHAPTER XXV
PHARYNX, NOSE, VELUM, LIPS AND JAW
The pharynx {OC, Fig. 93) acts as a resonating cavity in
communication with the oral and nasal cavities. Its main
period of free vibration (p. 2) depends on its capacity and on
the sizes and shapes of its laryngeal, oral and nasal apertures
(p. 281). The condition of the pharyngeal walls influences
the factor of friction (p. 5) and thus produces changes of
auditory timbre (p. 96). Owing to its irregular shape the
main free vibration may be accompanied by accessory ones ;
their relations in period and intensity to the main vibration
also produce effects of timbre. Changes in shape without
change in capacity may thus affect the timbre. Observations
on the changes in timbre in song and speech brought about
by diseases of the pharnyx have been frequently made ; the
results, have always been stated in vague terms ; the attempts
at explanation have been unsuccessful ; no experimental data
have been collected.
The separation of the upper (nasal) from the lower (oral)
portion of the pharynx is a comphcated act, requiring accu-
rate muscular adjustment. The contraction of the superior
constrictor {1, Fig. 101) forms a ridge ; the velum rises
through contraction of the elevators of the velum (5, Fig. 97).
The pharyngopalatine (8, Fig. 97) and the glossopalatine
(1?) muscles act as antagonists to the elevators and serve to
give an angular form to the velum during speech, whereby the
front part is more nearly horizontal and the rear part more
nearly vertical; they pull the velum down on relaxation of
the elevators, and the pharyngopalatine s pull it forward. Con-
PHARYNX, NOSE, VELUM, LIPS AND JAW 339
traction of the pharyngopalatines raises the larynx and the
pharynx waU and also narrows the pharyngopalatine arch.
The regulation of the action of these muscles in song and
speech is mainly, or wholly, auditory, that is, by the sound
produced.
The naml cavity on each side {A, Fig. 93) communicates
with the upper part of the pharynx by an opening of fixed
size called the nasopharyngeal meatus or choana (just in front
of 9 in Fig. 93). The nasal cavity on each side is to a large
extent filled with three turhinal bodies (2, 3, 4, Fig. 93) of
very irregular form ; it opens in front at the nostril (w, Fig.
93). The entire cavity is lined with mucous membrane.
Several accessory nasal sinuses (two of them shown at a and
d in Fig. 93) have small openings into the nasal cavity.
The bone and cartilage walls of the nasal cavities adapt
them well to act as resonators in connection with the pharynx.
Since the capacity and the apertures are practically fixed at
any moment, the effect on the vocal sounds is a constant
factor that enters into the adjustments.
Typical groups of sounds are produced by changing the
connections of the pharynx with the nasal and oral cavities,
and by altering their apertures.
With the lips open and all cavities connected, the nasal
vowels (such as a°, o", ce", e° in French) are produced.
With the upper part of the pharynx and the nasal cavity
cut off by closure of the velum across the pharynx and with
the lips open the pure vowels (such as a, e, i, o, u, g, o) are
formed in the oropharyngeal cavity.
When the oral cavity is cut off from the pharynx by the
velum or tongue, the nose acts with the pharynx as a com-
plex cavity. This is the case in the groups of sounds char-
acterized by n, n, and t|. The sound n is a regular one in
French and Italian (Ch. XXIII). As it is usually lacking in
in English and German it will be omitted in the following
discussion. In respect to articulation it can be considered as
intermediate between n and y\.
With the lips closed, the oropharyngeal cavity open, and
340 PRODUCTION OF SPEECH
pharyngonasal cavity free, the sounds produced belong to the
m group.
When one nostril is closed during the pronunciation of
m, n, T], hardly any difference is noticed in m, more in n and
most in t| without loss of distinctive character. When both
nostrils are closed, these sounds come to an end owing to the
stoppage of breath, but without becoming b, d, g (for which
the velum cuts off the upper pharyngeal and nasal cavities).
When the entire nasal cavity is filled with cotton ^ or when
one of the choanse is closed (no observations yet reported on
both) by a membranous growth '^ the m, n, ii characteristics
are still retained.
When the oropharyngeal cavity is closed by the lips or
tongue and by the velum across the pharynx, the sounds of
the b, d, g class are produced. They come to an end owing
to the lack of an opening for the escape of the breath, whereby
the blast that operates the vocal cords is gradually reduced
to zero.
With large adenoid growths in the nose the speech changes
greatly ; the result is known as the ' dead voice ' of the ade-
noid patients ; ^ m, n, t^ approach but do not merge into b, d, g.
The filhng of the nasal cavities in ' colds ' changes m, n, 11
toward b, d, g. On the other hand the removal of a large
single nasal polyp which leaves a cavity behind gives to the
voice for a time a hollow, or ' amphoric,' character. When
the accessory nasal sinuses become filled, the voice acquires a
' dead ' character.
With a tube inserted through the corner of the mouth and
passed behind the place of closure, m and n are produced in-
stead of b and d when the velum is closed across the pharynx.*
1 Sanger, Ahistische Wirkung der NasenhShlen, Arcli. f. d. ges. Phvsiol.
(Pfluger), 1896 LXIII 301.
^ ZwAAKDEMAKER, .Shi' hs sons dominants (tes resonnantes, avec qudques observa-
tions sur la voix morte des ad^noldiens, ■ ArcMves neerland. des sci. exactes et nat.,
J899 (2) 112.53.
3 Mbyee, Ueber adenoide Vegetationen in d. Nasenrachenhblde, Arch. i. Ohren-
heilk., 1873 VII 2-H, 1874 VIII 129.
* Sanger, as before.
PHARYNX, NOSE, VELUM, LIPS AND JAW 341
The following view of the action of the nasal cavity in vocal
sounds seemed justified by the foregoing observations.
The lower resonance tones of m, n, r\ depend on the size of
oropharyngeal cavity and on its apertures. On account of
the decreasing size of the cavity behind the closure we should
expect the lowest tone for m (labial closure), a somewhat
higher one for n (alveolar or palatal closure) and a still higher
one for r[ (velar closure). Kcenig's flames (p. 27) and Her-
mann's curves (p. 44) give the same tone for m and n ; no
data are at hand for -ry. For the lowest tone we may suppose
the nasal cavities to act as two necks of fixed size, shape
and conductivity to the pharyngeal resonator. The size and
shape can probably be taken as those at the choantB ; the
conductivity depends mainly on the amount of free space in
the nasal cavities. The differences between the lower cavity
tone in m, n, r\ and that in b, d, g cannot yet be definitely
explained. The addition of the upper portion of the pharyn-
geal cavity in the former case would give a lower tone, the
addition of the nasal necks a higher one ; the final effect is
probably a higher tone. Increasing stoppage of the nasal
cavities would lower the tone of m, n, -q ; with complete
stoppage it would be lower than that of b, d, g unless other
adjustments were made.
The higher resonance tones of m, n, r\ are influenced by
the various cavities into which the turbinals divide the nasal
passages, by the condition of the mucous membrane, by addi-
tional cavities present in the accessory sinuses and by the
character of the walls. These higher tones are markedly
characteristic of different voices and different conditions, but
experimental data concerning them are entirely lacking.
Similar considerations are probably valid for the nasal vowels.
The rise and fall of the velum affects the character of the
vocal sound.
Observations with Rontgen rays ^ showed that the velum
1 ScHBiEK, Die Verwerthunff d. RSntgenstrahlen f. d. Physiol, d. Sprache u.
Stimme, Arch. f. Laryngol., 1897 VII 125; Ueber d. Bedeutung d. Rontgenstrahlen
f. d. Phijsiol. d. Sprache u. Stimme, Neure Sprachen, 1 898 V Phonet. Stud. 40.
342 PRODUCTION OF SPEECH
rises for the vowels in the order a, e, o, u, i, that for a it
does not rise to the line of the hard palate, and that for u
and i it forms an arch up into the nasal cavity. For conso-
nants, except the liquids, it rises higher than for i. For
occlusives like b and k it flies up and falls at once. For
fricatives like f and v it does not rise as high as for the
occlusives. For m, n and i], it rises only a little from the
position of rest. For nasal vowels there is little or no move-
ment. Rise in pitch or increase in intensity of the cord tone
is accompanied by rise of the velum.
The tightness of the closure between the velum and the
pharynx wall can be tested by putting water into the nose ;
this is best done by inserting a thin elastic rubber tube far
into the nose and, the head being bent back, injecting
water at the moment of producing a vowel. According to
ScHUH 1 the closure is complete for i but not for a. Czee-
MAK 2 found that in speaking a the water ran down into the
throat, that in speaking i the water collected and was easily
retained for a considerable time, and that u and o resembled
i in this respect, but e less so. The difference in tightness is
presumably due to the difference in the angle which the palate
makes with the pharynx wall.
In order to determine if air issues from the nose in speak-
ing a sound, a cold polished surface — a mirror or knife blade
— may be placed at the proper moment under the nose ; the
faintest trace of breath is indicated by moisture on the sur-
face.^ In the production of the pure vowels no air issues
from the nose ; any trace of a nasal tone is accompanied by
emission of air.
To detect any passage of the air through the nose it can be
1 ScHDH, Die Bewegung d. weichen Gaumens b. Sprechen u. Schlucken, Wiener
med. Wochenschr., 1858 VIII 33.
2 CzEKMAK, Ueber das Verhalten des weichen Gaumens beim Hervorbringen der
reinen Vokale, Sitzb. d. k. Akad. d. Wiss. Wien, math.-naturwiss. Kl., 1857; also
in Czekmak's Gesammelte Schriften, I 425, Leipzig, 1879.
• ^ CzEKMAK, Ueber reine u. nasalirte Vokale, Sitzber. d. k. Akad. d. Wiss.
Wien, math.-naturwiss. Kl., 1858 XXVIII 575; also in Czekmak's Gesammelte
Schriften, I 464, Leipzig, 1879.
PHARYNX, NOSE, VELUM, LIPS AND JAW 343
made to act upon a small flame ^ by inserting into the nostril
a nipple connected to a rubber tube ending in a glass tube
with a small opening.^ The slightest trace of air produces
a fluttering of the flame.
A person with the velum grown to the pharynx wall so
that no air could pass through the nose was able to produce
the pure vowels but not the nasal ones.^ The sounds m,
n and n] could not be pronounced, but were replaced by some-
what similar sounds produced by keeping the cords in vibra-
tion while using mouth movements similar to those for b, d
and g, but with the least possible noise in the closure and
opening.
Gentzen's* direct observations through a cavity agreed
essentially with those of Czbrmak. A really considerable
pressure with a rod was required to press down the velum from
above during speech. A stylus resting on the velum was
made to record on a smoked plate.
According to records by Gtjtzmann ^ in a case where the
top of the velum was accessible, owing to removal of the
upper jaw, etc., the velum was pressed more or less firmly,
but always tightly, against the rear wall of the pharynx in
all vowels and consonants except m, n and r\, while a cross
ridge was plainly apparent just above the closure ; the rise of
the velum was least for a, greater for o and o, still greater
for e and u, greatest for i; for consonants the velum was
raised at least as high as for i and generally higher, except
for m, n and ii, for which it remained quiet; high or loud
tones were accompanied by greater rise. The natural action
1 Bbucke, Grundziige d. Physiol, u. Systematik d. Sprachlaute, 2. Aufl., 37,
Wien, 1876.
2 SiEVEKS, Grundziige d. Phonetik, 5. Aufl., 53, Leipzig, 1901.
3 CzEEMAK, Einige Beobachtungen ii. d. Sprache bei vollstandiger Verwach-
sung d. Gaumensegeh mit der hinteren Sr.hlundwand, Sitzber. d. k. Akad. d.
Wiss. Wien, math.-naturwiss. Kl., 1858 ; also in Czermak's Gesammelte Schriften,
I 468, Leipzig, 1879. . ^ , ,
4 Gentzen, Beobachtungen am weichen Gaumen nach Entfernung etner Geschwulst
in der Augenhohle, Diss., Konigsberg, 1876.
5 Gptzmann, Die geschichtl. Entwick. d. Lehre v. d. Gatimensegelbewegung
beim Sprechen u. s. w., Monatsschr. f. d. ges. Sprachheilk., 1893 III 217.
344 PRODUCTION OF SPEECH
may have been somewhat disturbed owing to the extensive
surgical operation.
The rise and fall of the velum can be indicated by its action
on a lever inserted through the nose.^ A straight iron wire
about 200°"" in length and 1.8™™ in diameter has a loop of 12™™
formed on the side at one end and covered with wax. A piece
of 40™™ in length at the other end of the wire is bent at right
angles to the staff but in the plane of the loop. The wire is
inserted through the nose so that the edge of the loop rests
upon the rear edge of the velum. Any rise of the velum
raises the edge and turns the wire ; the amount of turning can
be seen in the movement of the projecting arm in front. With
this indicator Czermak showed that the velum occupied a
different position for each vowel ; the rise of the velum (from
higher articulation with the back of the pharynx or else from
greater curving) was greatest for i, somewhat less for u, con-
siderably less for o, much less for e and nothing or almost
nothing for a. With the consonants the rise was the greatest
for the surd occlusives (p, t, k), the velum evidently being
raised passively by the air pressure. With sonant occlusives
(b, d, g) the rise was a trifle less, the pressure being evidently
not so great. With the surd and sonant fricatives (f, s, s,
9; X ; V, z, z, j, -y) the velum acted in the same way as for the
stops, the rise, however, being in all cases less than for the
corresponding occlusives. For the 1-sounds the rise was less
than for the fricatives, a distinct movement being observable,
for example, in passing from 1 to s. For tongue r the rise
was greater than for uvula r.
A simple light rod inserted through the nose and pivoted
by a thread from a band on the forehead may be made to
record directly on a smoked drum the rise and fall of the
velum. 2
1 CzEKMAK, Ueher das Verhalten des loeichen Gaumes beim Hervorhringen der
reinen Vocale, Sitzber. d. k. Akad. d. Wiss. Wieu, math.-naturwiss. Kl., 1857;
al|o in Czekmak's Gesammelte Schrifteu, I 423, Leipzig, 1879.
^ Allen, On a new method of recording the motions of the soft palate. Trans.
Coll. Phys. Philadelphia, 1884 (3) III; summarized in Internat. Zt. f. allg. Spr.,
1885 11 287.
PHARYNX, NOSE, VELUM, LIPS AND JAW 345
A small plaster knob on a light wire from a Marey tam-
bour attached to the forehead may be inserted into the mouth
and attached to the velum, i A record of the rise of the
velum during the word ' confabulate ' spoken with three
degrees of rapidity is given in Fig. 260. It shows the sudden
rise for k, the partial relaxation during a, the fall for n, the
rise for t, the slight relaxation
for ae, and the rise maintained Kr~\ (\l — \ (\l — \
during b]ulet. Fig. 261 shows
jijv_^^^-jyfv_j^^^
_J^/l_/UVLJVlA.
Fig. 262.
■A_
Fig. 263.
JJL
M^
a record for ' con — tra — ^'o- 2^''-
vene — contra — contravene ; '
the fall during a is nearly as
great as that for n, indicat- Fig. 261.
ing a decidedly nasal a. Fig.
262 shows a record for ' pant
— banana — blanch — branch
— can't.' In 'pant' the
velum makes two distinct
movements. In ' banana ' the
last two vowels involve only
a slight rise above the position
for n ; they are strongly nasal.
In ' blanch' and ' branch ' the fig. 264.
relaxation for n is complete.
The accented vowel in these
five words was se. Fig. 263 Fig. 265.
shows " ' hand ' spoken with
extreme slowness; the rise ^— ^/-^.
for h and that for d are clear; Fio. 266.
the semiclosure between them
indicates as much nasalization for k as for n. Fig. 264 shows
the phrase ' into Mount ^tna ' whispered and then spoken;
the action of the palate is practically the same in both cases.
Fig. 265 gives the record for the French oe° 'un' spoken three
times by a native and then for a° ' an 'likewise three times;
1 V^EEKS, .4 method of recording the soft palate movements in speech, Studies
and Notes in Philol. and Lit., Harvard, 1893 [I 213.
346 PRODUCTION OF SPEECH
Fig. 266 gives similar records for o° 'on ' and e° ' iu' ; in all
cases there was a movement of the palate forward from its
place of rest.
The relaxation of velar closure across the pharynx allows
the upper pharyngeal and the nasal cavities to influence the
sound. This gives the sound an auditory characteristic
called — not quite accurately ^ — ' nasalization. ' This char-
acteristic consists in changes in the cavity tones of the
speech sounds. For a given position of the vocal organs
the opening of the upper pharyngeal cavity will add tones
due to the size of the cavity and the nature of the aper-
tures; it will also necessarily modify some of the tones of
the mouth on account of the additional aperture and the
reciprocal action of connected cavities. Thus, a" will
differ from a formed by the same positions of the tongue,
lips, etc., not only in having additional cavity tones, but
also in changing the mouth tones. To retain the mouth
tones unchanged to the ear, the tongue, lips, etc., must make
readjustments; this is perhaps the explanation of the fact
that for the sound heard as o" Rousselot finds the tongue
position varying from Oj (p. 315) to O3 (p. 320).
The action of the velum may be conveniently studied by
comparing the record of the breath curve from the nose with
that from the mouth. A nasal olive of convenient size
(Fig. 88) is connected with a small tambour, and a mouth
trumpet (p. 219) with another one; the two tambour points are
synchronically registered. When the velum is closed com-
pletely across the nasal passage, the recording point is at its
position of rest and usually no vibrations appear in the line
it draws. As the velum falls, air passes through the nose,
the recording point rises and, if delicately adjusted, registers
the vibrations from the larynx that may be present; the
extent to which the point rises depends on the amount of air
issuing from the nose, that is, on the size of the opening be-
tjveen the velum and the pharynx-wall. "When vibrations do
1 Sanger, Ueber d. Entstehung d. Ndselns, Arch. f. d. ges. Phvsiol. (Pfliiger),
1897 LXVI 467.
PHARYNX, NOSE, VELUM, LIPS AND JAW 347
appear in the nasal tracing without any rise of the line to in-
dicate exit of air, they are often due, I believe, to the fact
that the vibrations in the mouth set the velum itself (and
consequently the air above it) in vibration without any re-
laxation of the closure.
In interpreting the records, it must ever be borne in mind
that the inertia of the recording levers distorts them; thus
a constant emission of air from the nose during n will show
itself as a somewhat gradual rise of the lever, while a sud-
den strong one from the mouth, as in p, will give a sudden
rise. It must never be forgotten that the scale of rise is not
proportional, but that a considerable rise of the lever from
jest indicates only a small emission, while a small increase in
rise beyond this indicates a great increase in emission.
Fig. 267.
Ordinarily the vocal cords cease to sound when the vowel
position is relaxed before a pause; the surplus air is then
expelled noiselessly (p. 224). In some dialects the cords are
still sounding when the position is relaxed ; the opening of
the nasal cavity then gives a nasal twang to the sound. ^
Records showing this fact have been made by Rousselot.
Investigations by the above method ^ have been made on
Italian sounds. In Figs. 267 to 273 the upper record is from
the nose, the lower one from the mouth. In ma and na
the nasalization includes not only the liquid but, to a less
degree, the entire vowel, especially its close (Fig. 267).
The vibrations in the tracing indicate that the velar move-
1 Rousselot, Principes de phone'tique experimentale, 240, Paris, 1897.
2 JOSSELYN, De la nasality en italien, La Parole, 1899 I 602.
348
PRODUCTION OF SPEECH
Fig. 268.
ment finished before the cord action did, thus giving a
special nasal twang to the end of the vowel. This relaxation
of the velar action before that of the cord action is analo-
gous to that of the tongue action in English long vowels
whereby they frequently acquire a ' diphthongal ' character.
The record for mano (Fig. 268)
shows the nasal current of air
during m, the mouth current and
the smaller nasal one during a,
the increased nasal current dur-
ing n, the cessation of this cur-
rent during o, and the expulsion
of surd air from both nose and mouth after the laryngeal
vibrations have ceased.
The explosion of t in ta (Fig. 269) seems to have been ac-
companied by a fall of the velum in Josselyn's records; this
does not, I believe, indicate any nasalization, as Josselyist
asserts, but only a sudden relaxation of the tense curvature
produced by the pressure of air and a rebound of the record-
ing lever; the bend in the line at this point in Josselyn's
figure is just what Avould be expected from such action. For
p in pa a considerable emission of air is indicated at the
moment of the explosion of p ; this is due, I believe, to the
fact that the vowel a (which is somewhat nasalized) begins ap-
FiG. 269.
parently during the rush of air in the explosion of p (p. 45).
The k of ka shows no disturbance in the velar record.
During 1, r, g, d, b, z of la, ra, ga, da, ba, za the nasal
line shows vibrations without any deflection from the point
of rest; as stated above, these vibrations indicate, I believe,
PHARYNX, NOSE, VELUM, LIPS AND JAW 349
the transmission of vibratory movements and the roll of the
r to the velum and thus to the nasal cavity, but no relaxation
of velar closure; Josselyn's deduction of nasalization for
these sounds is, I believe, incorrect. During f of fa a small
rise was noted in the nasal line, indicating a slight escape of
air through the velar closure.
Fig. 270.
In the record for ' nessuno ' (Fig. 271) the full opening of
the nasal cavity is seen for n ; the e is considerably nasalized
throughout, but most strongly near the end ; the glide from e
to s close's the nasal cavity entirely and the mouth somewhat;
the glide from s to u is marked by a slight explosive puff from
the mouth and then an almost complete closure ; the u shows
a steadily increasing stream from the mouth and the glide
from u to the n a steadily decreasing one ; the u is without
nasalization; the u-n glide is strongly nasalized; the n
is nasal as before ; the o is slightly nasalized. Records of
Fig. 271.
' menare,' ' mentire ' and ' mandare ' are given in Fig. 272.
The m is of nearly constant length. The first n is very
short, the second more than 3 times, and the last about 3
times as long ; similar results show a long n in ' infante, '
' imperare, ' mm.
mente, ' and a short n in ' onore, ' ' nes-
B50 PRODUCTION OF SPEECH
suno, ' ' mano, ' ' tenere. ' The results show that n at
the end of a syllable is stronger and longer, while at the be-
ginning and after an open syllable it is feebler and shorter.
Rejecting the usual view of the nature of a word as made up
of series of separate pieces called 'syllables,' I would give
the explanation that in the flow of speech the n forms part
of the vowel material that with the consonant movement goes
to make up the content between centroids of stress, and that
for intervals of equal stress-effect its length is varied just as
those of the vowels to produce equal amounts of auditory
and motor work. The tendency of Italian to velar relaxa-
FiG. 272.
tion, that is, to nasality, has been conclusively proved by
JossELYN, whose records show many nasalized vowels, oc-
clusives and fricatives.
JoasELYN found ^ that in Italian the nasalization required
for a ' nasal ' regularly extended to the neighboring sounds
and sometimes to the entire word, and that after a vowel the
m and n might take a form intermediate between the vowel
and the usual m or n.
An interesting and not uncommon pronunciation ^ is indi-
1 JossELTN, Etude sur la phone'tique italienne, 139, Th&se, Paris, 1900; also in
La Parole, 1900 II 179.
2 JossELYN, Etude, as before, 139.
PHARYNX, NOSE, VELUM, LIPS AND JAW 351
cated in Fig. 273. Here the first n has disappeared, as is
proved by the fact that there is no oral occlusion, the vowel
being strongly nasalized and merging immediately into the f.
This is an approach to the character of a French nasal vowel.
As already pointed out (p. 346), ' nasality ' is an auditory
term indicating the presence of tones from the nasal cavity.
These tones regularly arise when the rear nasal passage is
more or less open.
In a nasal breath record any such opening indicates itself
by a rise in the course of the curve. Just how great an
effect on the speech sound may arisB from the transmission of
laryngeal vibrations through the velum when entirely closed
across the nasal passage, it is impossible to say. It may per-
FlG. 273.
haps be tested by stopping the front nasal opening while_ a
vowel is sung, and b}' having other persons listen for the
presence of any change. In an absolutely unnasalized vowel
I am unable to observe any difference.
The vibrations in the nasal line indicate vibrations of the
air due to the cord action ; their appearance indicates either
transmission through the velum (p. 347), while still closed
against the pharynx wall, or transmission through a velar-
nasal opening. In the former case they might perhaps give
rise to a faint nasal tone, although this is doubtful. In the
latter there is a velar-nasal opening whose size would bear
some relation — but no simple one — to the amplitude of the
registered vibrations. The auditory effect would be ' nasal-
ization, ' properly so termed. In a tight tambour system, how-
352 PRODUCTION OF SPEECH
ever, any such nasalization must be accompanied by a rise in
the nasal line ; the presence of vibrations without such a rise
indicates either the transmission of vibrations through the
velum without opening (as indicated above), or a leak in the
apparatus.
The absence of vibrations from the nasal tracing does not
indicate absence of nasal cavity tones; no cavity tones of
an}^ kind ever appear in tambour tracings. As long as the
cords vibrate, any opening of the velar-nasal passage will be
accompanied by nasal resonance.
When there is no cord tone, the opening of the nasal pas-
sage is accompanied only by the noise of the escaping breath,
which at most can make only a soft h-like sound during
speech and cannot ordinarily be heard. The relative weak-
o ag 32 e
Fig. 274. Tig. 273. Fig. 276. Fig. 277.
ness of such a nasal explosion may be made apparent by pro-
nouncing ' cat ' and ' cap ' first with the regular mouth
explosion and then with a nasal explosion, the mouth being
kept closed. An explosion by dropping the velum cannot
be made strong enough to be heard when the mouth passage
is already opened, as in vowels. This ' surd nasality ' occurs
as a nasal modification of explosives and fricatives.
The positions of the pillars of the velum (glossopalatine
and glossopharyngeal arches) have been carefully observed
and drawn by Thudichum ^ for his Swiss-French vowels : o
as in 'or, fort, sotte, forte ' (Fig. 274); ag (close) as in 'pfite,
tasse, cas' (Fig. 275); aj (medium) as in 'page, part, papa'
(Fig. 276) ; and e as in ' pere, tete, perte, nette ' (Fig. 277).
The positions and movements of the lips can be observed
1 Thidichum, La prononciatioti de I'a/ranfais, Neuere Sprachen, 1897 IV
Phonet. Stud. 22.
PHARYNX, NOSE, VELUM, LIPS AND JAW 353
directly and drawn ; or they may be photographed by making
an exposure at the proper moment. Lip-positions for various
American and German sounds are given in Geandgent's
Plates XVII to XXVI at the end of this volume ; they are
freehand drawings of rather schematic character, often show-
ing angles instead of curves.
The activity of the lips is frequently overlooked, although
important results depend on the various degrees of length-
ening the mouth-opening by retraction of the corners;
of rounding by closing the lips except for an opening in the
middle, or by pulling the corners toward the middle ; and of
projection with rounding, or without rounding. These varia-
tions in the opening and neck of the vocal resonator change
the sounds greatly (p. 281). One special character of Eng-
lish sounds as compared ,with the corresponding German and
French ones arises largely from the small movements of the
lips.
The movements of the lips have been recorded by Dbmeny,
in a series of photographs taken with a kinetocamera ; ^ one
set of his views for ' Je vous aime ' is reproduced by
Jespersen.2
The kinetographic method has not yet been systematically
applied to the study of lip movements in speech although
it will presumably replace all others, because it leaves the
speaker unhindered by apparatus attachments, and because it
can be made with any required accuracy. A good register-
ing kinetograph (or cinematographe) is focused on the lips of
the speaker and the pictures are made at the rate of 40 or
more a second. Before ending the experiment a millimeter
scale is laid on the lips and also recorded. The film is
developed by means of a special equipment, or is sent to a
photographic company. The measurements may be made
directly on the negative film, or on a blue print. A positive
film may be made for projection. For finer measurements or
1 Demeny, Analyse des mouvements de la parole par la chronophotographie.
C. r. de I'Acad. des Sci. Paris, 1891 CXIII 216 ; Journal de physique, 1893 328.
2 Jespeksen, Fonetik, Tavle I, Kobenhavn, 1897-99.
2.3
354
PRODUCTION OF SPEECH
for demonstration on a large scale the pictures can be greatly
enlarged. In demonstrating the movements by projection
the details can be studied by running the film slowly, after
the manner employed for explaining surgical operations in
medical schools.
The movements of the lips may be registered by a pair of
light arms inserted into the mouth and attached to a Maeey
tambour;^ an open tube may be placed before the lips to
register the breath pressure. Each lever may be made to
register separately.^ A complete instrument^ is shown in
Fig. 278.
Fig. 278. The lips rest on the arms at L' L." Any com-
pression pulls out the tops of the tambours J' J" by the arms
K' K " ; the movement may be transmitted to two separate
registering tambours by the tubes iV' iV" or to one tambour
Mhy the joint tube iV. The breath current is caught in the
mouth-piece U and registered by the tambour F. The re-
ceiving apparatus is held on the rod R by the adjustable arm
' RosAPELLT, Inscriptions des mouvements phonetlques, Travaux du laboratoire
de M. Marey, 1876 II 119.
* ^ HonsSELOT, Les modifications phonitiques du langage, Rev. pat. gallo-rom.,
1891 IV, V; also separate.
" RousSELOT, Principes de phone'tique expe'rimeutale, 92, Paris, 1897.
PHARYNX, NOSE, VELUM, LIPS AND JAW 355
S. The recording tambours M and F are placed on the
Maeey support V, by which the contact of the recording
points may be adjusted. This is likewise fastened to the
rod E by the screw 0. The two recording levers are adjusted
to register synchronously on the drum.
The pressure of the lips may also be measured by a small
rubber bulb (Fig. 245) between them. In Italian the relations
of lip pressure have been found ^ to be p > m > b or m > p > b.
To register the lip projection a small tambour or an ex-
ploratory bulb may be rested lightly against the upper or the
lower lip; any contraction of the muscle of the lip presses
the air from the tambour. ^ As indicated by such a tambour,
the lips in Rousselot's labials are less completely and firmly
closed for v than for f, and less firmly closed for b than for p,
while for m the closure is like that for b. A bulb against the
lips indicates ^ a greater projection for u than for w in Italian,
as in the words duo 'duo ' and dwomo 'duomo.' A cylin-
drical tambour or three small rubber balls may be used.*
The movements of the lower jaw may be registered by a
small rubber bulb placed in the ear and attached to a tambour
in the usual way. The arrangement is very convenient and
fairly accurate, though not sensitive to very small movements.^
To register the movements of the jaw Gall^e and Zwaae-
DEMAKER ^ used a movable arc connected to the lower jaw
and supported on a framework attached to the head. The
movements of the arc were recorded by air transmission with
tambours.
In ventriloquism the movements of the lips and lower jaw
are made as small as possible. The lower lip is slightly
1 JossBLYN, Etude sur la phonet. ital., 91, Thfese, Paris, 1900; also in La
Parole, 1901 11141.
2 EoussELOT, Principes de phonetique expe'rimentale, 93, Paris, 1897.
8 JossBLYN, as before, 115.
4 Gallee und Zwaakdemaker, Ueber Graphik d. Sprachlaute, Neuere
Sprachen, 1900 VIII 16.
5 Gallee und Zwaaedemakek, as before, 11.
•5 Gallee und Zwaabdemaker, as before, 17.
356 PRODUCTION OF SPEECH
drawn back and rested against the upper teeth. The tongue
articulations are greatly altered.'
References
For structure of pharynx : Disse, Anatomie d. Rachens, Heymann's
Handb. d. Laryngologie u. Rhinologie, II 1, Wieu, 1899. For physiol-
ogy of pharynx : Einthoven, Physioiogie d. Rachens, Heymann's Handb.
as before, II 46. For anatomy of nasal cavities : Frankel, Gefrierdurch-
schnitte zur Anat. d. Nasenhbhle, Berlin, 1891. For the use of Rontgen
rays : Flatau, Die Anwendung des Rontgen' schen Verfahrens in d.
Rhinologie u. Laryngologie, Heymann's Handb. d. Laryngologie u.
Rhinologie, III 1245, Wein, 1900.
For graphic apparatus for phonetics : Verdin, Paris. For tambours
and manometers: Albrecht, Tubingen; Petzold, Leipzig; Zimmer-
MANN, Leipzig. For cinematographe camera ; Lumiere et ses Fils,
Lyon-Montplaisir.
1 Flatau und Gutzmann, Die Bauchredner-Kunst, Leipzig, 1894.
CHAPTER XXVI
SIMULTANEOUS AND SUCCESSIVE SPEECH MOVEMENTS
By records of the breath curve from the mouth (p. 219)
and the vibration of the larynx (p. 267) Rousselot ^ has shown
that in the German pronunciation of pa and ba the larynx
does not begin to vibrate till much later than in the French
pronunciation. This is evident in the record shown in Fig.
279; the two upper lines are French pronunciations, the two
trii^tk
i>Ttd,tk
iofds
ir^ei^tk
Fig. 279.
lower ones German. In French the p is surd with a sonant
explosion ; in German the explosion is also surd. In French
the b is sonant during its last portion and during the explo-
sion ; in German only the explosion is sonant.
Similar records have been made^ on an American. An ex-
ample is given in Fig. 280. The breath line of paet shows
the explosion of p at ^, a rush of surd air from 2 to S con-
stituting the surd explosion of p, the vibrations of ae, and the
1 "KoDSSELOT, Appiications pratiques de la phone'tique exp&imentale. La Parole,
1899 I 401.
^ Rousselot, U Enseignement de la prononciatton par la me. La Parole, 1901
III 577.
358
PRODUCTION OF SPEECH
occlusion for t; the cord line shows quiescence during the
occlusion of p, the advance of the thyroid cartilage from B to
S, the beginning of the cord vibrations of ae at 3, and their
cessation for t. The p thus ends with an aspiration consisting
in a rush of surd breath. The record for bsed shows sonancy
Fig. 280.
throughout the occlusion of b from 1 to ^, a sonant explo-
sion at ^, followed by the curves for ae and d. The explosive
rush of air is probably much weaker than for p. This set of
records showed that p and t were regularly aspirated, but k
not, and that b and d were sonant throughout their occlu-
sions while g was either wholly or partially sonant.
^mvxWWWWUVW^MtfArt'V****^*^
t*irtt>Hn^H***H^t**,Vv*iMmft'^
aba
Fig. 282.
ama
Fig. 283.
RosAPELLY registered simultaneously the lip closure with
the apparatus mentioned on page 354, the cord vibrations by
an electrical vibrating apparatus on the larynx, and the passage
of the air through the nose by a tube inserted into it (Fig. 88).
Records of French apa, aba and ama are given in Figs. 281 to
283 (upper line, nose ; middle line, larynx ; bottom line, lips).
SIMULTANEOUS AND SUCCESSIVE MOVEMENTS 359
Using simultaneous registration for the larynx by RosA-
PELLY'S electrical vibrator, for the nose by a tambour (p.
219) and for the lips by levers connected to a tambour
(Fig. 278, without the mouthpiece), Rousselot^ obtained
diagrams like that shown in Fig. 284. The curve from the
nose indicates the issue of the air through velar relaxation
(p. 346) ; upward movement in the larynx registration can be
taken to show vibration of the cords, the single vibrations
seldom appearing ; in the lip curve upward movement shows
increase in lip pressure.
^"' ^ ^.---^
"
V -V
Ootids — . lj-
-
t -
^^\ M
-' V/^^^"^ ^ —
—
7n ^ c. vi t e I b j e-
Fig. 284.
Among the results obtained by Roussblot for his own
dialect of French — without consciousness of the peculiarities
— the following may be noted (I have not specified the
varieties of the vowels, a^, a^, ag, etc., or their lengths);
I have ventured to add physiological and psychological
explanations for some of the phenomena.
The nasalization of a nasal vowel varies with the nature of
the preceding sound, being complete for the initial position
(as a°ta") and after s, s and probably all the continuants (as
in so°, se°, sa°trie) but lacking in the first portions after
p, b, t, d, k, g (as in pa°s, po", ta", ka"). In the case of
the occlusives mentioned, the velum is closed. The explo-
1 RoussELOT, Les modifications phongtiques du langage, Rev. des pat. gallo-
rom., 1891 IV, V; also separate.
360 PRODUCTION OF SPEECH
sion usually occurs through the mouth; if the velum is
opened instead of the mouth, the explosion is nasal ; if both
are opened, it is orinasal. If the cords do not sound during
the explosion, there is a surd breath explosion. To produce
a surd oral explosion the velum cannot open till the explosion
is over ; if it then opens before the cords begin to vibrate for
the following vowel, a silent interval occurs. If the cords
begin to vibrate during the oral explosion and before the
velum opens, the result is a glide of the character of an un-
nasalized vowel. In the case of the sonant explosives the
cords are already vibrating during the explosion and the
occlusion must be followed by a non-nasal vowel glide unless
the explosion is to be nasal. In the cases of s, s, etc. the
vowel glide may be very short.
The consonants z and z are nearly always surd in the
middle though sonant at both ends (as in e"zur ' un jour ; '
aza, an artificial group), while v is nearly alwaj'S completely
sonant (as in sivavforsave ' si vous vous forciez'). This
may possibly be due to the elimination of the breath .pressure
in the larynx due to the obstruction in the mouth ; the pala-
tograms for z and z (Figs. 206, 207) indicate rather consider-
able closure, but d and g have still more closure and are yet
sonant. The cause is not to be sought in some association
between articulation and lung pressure, as the lung action is
not jerky.
Final sonants often become surd before they end, on
account of the long pause for which the larynx prepares;
thus in e°zurko]aviynomeynfcem 'un jour 9a [= il]y avait un
homme et une femme ' the first m is entirely sonant while the
second one, occurring just before a short pause, is half surd.
Surds between vowels may become sonants (as in to"-
p.o^po" 'ton pompon,' lap.yzen ' le plus jeune,' sap.alav
' s'appelait'). This is due to the greater ease in maintaining
the glottal position as compared with the double act of relaxing
ai^d then tensing the cords. This gain is mental ; it saves
two changes of volition with the necessity of auditorily
verifying the results.
SIMULTANEOUS AND SUCCESSIVE MOVEMENTS '361
Simultaneous sonation and nasalization of a surd were
found in ' diable ton happeur,' which appeared in the records
as ' djabXtuna^pSur ' (Fig. 285), the p actually becoming a
nasalized sonant labial somewhat resembling m. The phe-
nomenon is probably due in the first place to the avoidance
of glottal readjustments, for the reason just stated, and in the
second place to the greater ease of
gradually closing the velum through
p? to u rather than very suddenly after
a". The difficulty of the rapid velar
action may lie not only in the move-
ment itself but also in the dissolution
of its association with cord action
through the preceding sounds. In the
following examples the " attached to
the letter for an occlusive indicates
velar explosion.
In groups of occlusives (p, b, t, d, k, g) with fricatives (f,
V, s, z, s, z) an initial surd remains such but an initial sonant
often becomes surd (as in kvuty ' que veux-tu ? ' but dofyra
' refuser '). In the latter case there is a partial loss of cord
action. In such groups between vowels there is most often
an assimilation of the first consonant to the second in respect
to sonancy, very rarely of the second to the first ; sometimes
Fig. 285.
Fig. 286.
the two retain their values. Examples were found in : aboka
for abka, pip.zi ' pipes- y,' upik.babje" ' il pique bien bien,'
kopustboje" ' §a pousse bien,' pus.bje" ' pouche [tousse]
bien ! , ' ebokvuty ' eh bien ! que veux-tu ? , ' fob.opra°dr
' il faut bien prendre, ' padokitpa" ' pas de quitte pain, 'e"-
ledtopyl ' un lait de poule, ' zyg.oP3ti ' joue, petit, ' mo"-
362 PRODUCTION OF SPEECH
povopati ' mon pauvre petit!,' mapovtofcem.,, 'ma pauvre
femme!' as in Fig. 286, koekin3buz„paso''so ' celui-ci ne
bouse pas son sol,' kunpuzopa ' qu'il ne puisse pas.' The
surd forms of the sonants retain their mouth action and do
not become the same as the corresponding surds, and con-
versely ; thus bo remains distinct from p, p« from b, k. from
g, etc. (pp. 304, 317).
When a surd is followed by 1 or m it sometimes becomes
sonant. In many cases of tr and pr the t and p are sonant
(as in uprfatana", ' ils pr^tendent').
In liquids the presence or absence of the cord tone during
the whole or part of the sound varies from case to case.
Fig. 287.
In general, initial or final liquids remain sonant or become
half surd. Between a vowel and a surd the nasals m and
n and also 1, A, r and ] are always sonant. Between a surd
and a vowel they are almost always sonant, though j has a
marked tendency and r, 1 and A a less tendency to become
surd during part of the length.
In an investigation by Rotjsselot ^ tracings of lip pressure
(p. 354), larynx action (electrical vibrator, p. 267) and nasal
breath (p. 219) were made on a native of Bonn, a native of
Biitow (Further Pomerania) and a native of Hamburg.
Fig. 287 gives a record of the Pomeranian dialect for ' wir
woUen nach Eldena laufen.' The dotted lines aid in com-
*
1 RoossELOT, Recherches de phone'tique exp&imentale sur la marche des evolu-
tions phonetiques d'apres quelqnes dialectes bas-allemands, La Parole, 1899 I 769.
SIMULTANEOUS AND SUCCESSIVE MOVEMENTS 363
paring corresponding points of the records. The approxi-
mate boundaries between the sounds I have tried to indicate
by the vertical lines at the bottom. For v the beginning
IS indicated by the closing of the lips (rise in the lower
line) and of the velum (fall in the upper line) ; the lack
of vibrations in the middle line indicates that it is surd or
that the cord vibrations are y&tj weak; it is a very long
sound. The lips open for i and the cords vibrate. The lips
again close for v, which is this time sonant. The lips sepa-
rate vertically again for y, still more for 1, slightly less for n
and most for a. The middle line indicates that y, 1, n and
the first part of a are sonant, the latter part of a being surd.
The top line indicates complete closure of the velum during
Vo, i, V, y and the first part o, 1, and nasalization of the
parts of 1 and a adjacent to the fully nasalized n. A pause
with slight lip movement occurs between a and e. The lips
come gradually nearer throughout e, 1 and the first part of
n, open during a, close during 1, open again during o, close
suddenly for the implosion of p, relax slightly but do not
open for the explosion of p, and remain closed during the
last sound, which is thus m and not n as written. The
nose line shows that the portions of 1 and a adjacent to n are
slightly nasalized, and that the explosion of the p occurs
entirely through the nose. The p is thus not the ordinary
sound but the nasal explosive p". The cord line indicates
sonancy up to the implosion of p. The p is surd. In spite
of the lack of vibrations in the cord line their presence in
the nose line shows that the m is sonant. The m is a
long sound like v ; all the others are short ones of about the
same duration. The phrase may be expressed by Vgivylnal-
elnalop°ni. The record makes it very clear that each part
of the phrase is a synthesis of continuous movements and
any separation into distinct sounds or into sounds and glides
must be a rather arbitrary one. Other records of the same
phrase showed like results.
Records of the Bonn pronunciation of ' ich ging meines
Ganges und dachte ' showed that ' ging ' occurred twice (in
364 PRODUCTION OF SPEECH
slightly slower speech) as girik and five times (in slightly
faster speech) as giii, the k being lost between r\ and m,
although the speaker supposed it always present. In pro-
nouncing ' Ganges ' the speaker supposed himself to have
always said gai^s but the records often indicated ganks.
The i and a were often nasalized, producing gi°ti, mi^nas,
ga"Tiks etc.
The supposedly lost k of juii ' jung ' in the Pomeranian
dialect often appeared in the records (though not heard) of
' dat jung Pirt ' which was spoken as dat juTik pirt. In the
Bonn dialect a supposedly lost final n of ' vekofe 'was found
to reappear (though not noticed) in one record in three of
' de al Hemde vekofe' 'die alten Hemden verkaufen, ' the
record on this one occasion indicating fskofan and on the
other two fakofa. Such appearances of sounds supposed to
have been lost in the history of the dialect seem to indicate
the transmission of unperceived elements of a language.
Records of the Pomeranian pint and bint showed the in-
tonation of the larynx starting after the explosion in p and
near the beginning of the explosion in b. The difference
between German p and b often seems small; a difference
in articulation is probably always present (p. 304). Medial
b between vowels was found to be sonant throughout in
Pomeranian but often surd in the Bonn dialect. Not only
b but also gy V and d were often found to be surd in the
Bonn dialect.
A labial maj' labialize the following sound, lopn ->■ lopm
in the Pomeranian (above) and Hamburg dialects; a vowel
between two nasals or followed by a nasal may become nasal-
ized in the Bonn dialect, vai-\\e ->■ mi'hie, man -> ma"n ;
a consonant may be exploded nasally on account of a pre-
ceding nasal (Bonn), enda -> end"3, embda -* emb"d"a ; a
surd may become a sonant when followed by a sonant, but
not necessarily, lopm may-+lop.m ; a surd may or may
not be changed to a sonant between two vowels in the Bonn
dialect, hatyranox may-> hat.yranox, opemol -> op.emol.
There may exist, totally unsuspected by the ear, a tendency
SIMULTANEOUS AND SUCCESSIVE MOVEMENTS 365
to change occlusives to fricatives. In one case of the Bonn
dialect the lip pressure for p was found to be less firm than
that for m in lopm ; in one of the Pomeranian dialects it was
found to be no stronger than that for v. In both dialects
there appeared a tendency to nasalize the vowels ; in records
of the Hamburg dialect this tendency was very strong.
RoTJSSELOT 1 considers it to be proven that phonetic trans-
formations are accomplished by degrees and that if they are
extended over considerable territory they leave traces of
their various stages; that they are to be considered as the
products of physiological tendencies which can he detected
even before they are noticed in speech ; that they show sur-
vivals of past forms which are unnoticed by the speaker or
the hearer.
Records of the Italian pronunciations ^ of several "natives
showed variations of the moment at which the cord tone begins
in combinations of surds with
sonant consonants and vowels.
In the following typical records
the upper line is that of cord
vibrations registered by an ex-
ternal capsule (Fig. 124), the
lower line that from a mouth-
piece connected to a tambour -pia. 288.
(p. 219). In making compari-
sons it should be remembered that the tambour Y axis is
curved (p. 197). The tambour curve should be referred to
its X axis by a curved line. The vertical lines indicate
corresponding points of the two X axes.
In the record for totale (Fig. 288) the vibrations for o do
not begin till a moment after the t has exploded, indicating a
rush of ' surd air ' between the t closure and the o. The
vibrations for a, however, begin at the moment of the explo-
sion of the t. The record shows that the recording lever began
1 RoussELOT, as before, La Parole, 1899 I 790.
2 JossELYN, itude sur la phonetique italienne, These, Paris, 1900 ; also in La
Parole, 1899 1 ; 1900 IL
366
PRODUCTION OF SPEECH
its fall at the end of o but, owing to its friction and inertia,
did not have time to descend completely before the explosion
occurred; the moment of explosion can be found by com-
pleting the curve as indicated by the dotted line. These two
forms of t are of frequent occurrence in Italian. The
difference between a t with a
surd explosion and one with a
sonant explosion must have its
effect on the ear. That the
sonant explosion may or may
not possess the cavity tones —
and therefore the mouth con-
figuration — of the following
vowel has been shown by Hermann (p. 45).
The record for slita ' slitta ' (Fig. 289) shows an almost
entirely sonant s (or s»), a very long t closure, a long surd -
explosion and the vowel a. A synchronic dotted curve
shows where the intonation of a occurs on the mouth line.
The curve along which the recording point falls when the
breath ceases to act is shown at the end of the i where the
breath is cut off by the t closure ; at the end of the t the point
does not fall in this way but sinks gradually, indicating a
more gradual cessation of the breath, that is, simply the
Fio. 289.
Fig. 290.
fading away of the explosion of the t. All records of Italian
' double consonants ' show, just as in this case, that they are
never double movements, but single consonants lengthened
^nd intensified.
In the record of riordinare (Fig. 290) the cord tone started
before the tongue began the roll for the first r. The first r
SIMULTANEOUS AND SUCCESSIVE MOVEMENTS 367
showed four flaps, the second r apparently one somewhat
complicated one, the last r one flap.
In the record of at]ene ' attiene ' (Fig. 291) a long t ap-
pears as in slita. Its explosion does not die away like most
explosions but passes into a surd breath involved in the
' consonant i '.
The records also showed that in the occlusives the dis-
tinctions of sonancy between surd p, t and sonant b, d
differ in each individual; that in a true sonant the cord
tone begins during the occlusion ('iozzina'); that in some
initial and double consonants (as in ' acZc^entro '), where the
muscular wall of the cavity is more firmly contracted, the
sonancy ceases before the explosion owing to the equalization
of the air pressure above and below the glottis, in one case
ceasing at the end of the
glide-movement from a to
J in ' aggetivo ; ' that in
one form of surd (aspirated
surd) the vibrations of the
following vowel begin after
the explosion is finished
-1 • , i r„ ,s Fig. 291.
(as in caino shtta, );
that in surds the vibrations of the vowel may begin before
or during the explosion (as in ' avvocailo '); that even in such
a surd the larynx vibrations may begin during the explosion
but nevertheless may be marked by a rush of air later in the
explosion (as in ' fotale' ). Similar conditions seem to occur
in German according to the curves published by Hekmank.^
The character of ' consonant i and u ' in Italian has been
investigated by Josselyn.^ A tambour recording the breath
current (p. 219) and another recording from the exterior of
the larynx (p. 267) gave tracings similar to those just dis-
cussed. In ' piena ' the laryngeal vibrations begin some time
1 Hermann, Fortgesetzte Untersuchun<jen iiber d. Konsonanten, Arch. f. d. ges.
Physiol. (Pfluger), 1900 LXXXIII 1 ; also above, Fig. 33.
2 JossELTN, Note sur i et u consonnes, c[e] et g[e] en italien, La Parole, 1899
I 833; iStude sur la phonitique italienne, 104, The.se, Paris, 1900; also in La
Parole, 1901 III 85.
/ Xwtf^JmlUV*.
368 PRODUCTION OF SPEECH
after the p has exploded while in ' pena ' they begin almost at
the same time as the explosion. There is thus a current
of surd breath between the p and the e of ' plena ' while
there is no such current in ' pena. ' That the sound is not
simplj' a surd i is shown by the different palatograms for ' pi '
of ' pieno ' and ' pi ' of ' pia.' The word is thus phoneti-
cally p]eno. Similar results were obtained in comparing
fjatiko ' fianco' with fiXo ' figiio,' kjama ' chiama' with kilo
' chilo,' tjene ' tiene ' with tenero ' tenero,' kjaro ' chiaro '
with karo ' caro.' The ] was surd throughout its length,
or became sonant at its end under influence of the follow-
ing vowel. When a sonant consonant was followed by j
as in bjeko 'bieco,' vjeto ' vieto, ' djetro ' dietro,' gjaco
' ghiaccio, ' the explosion of the consonant was weak ; the ]
was sonant in these cases. It was evident throughout that
' unsyllabic i ' preceded by a consonant and followed by a
vowel was a palatal consonant j which tended to unite with
the preceding consonant and modify its articulation. The
union of j with, and the modification of, the consonant were
greater as the places of articulation were more nearly alike ;
the labials p and b showed little influence from j ; t and d
showed more; for k the change was marked, the explosive
in kjaro ' chiaro ' losing much of its force and its breath
record resembling rather that of the fricative in fja'qko
' fianco,' while the g in gjaco 'ghiaccio ' gave a breath record
scarcely differing from and even weaker than that of v. A
palatogram showed a consonant with a prepalatal articulation
between t and k to exist in ' c ' of ' cece. ' This articulation
coincides closely with that of ]. The union of the con-
sonant and 3 was probably made with a single movement
of the tongue that caused an occlusion and then a gradual
opening. This sound may be indicated by c, its sonant by
J. A similar result was foi;nd in the records of ' cielo.'
These words would thus be phoneticallj^ cece and celo.
Likewise ' gente ' appeared as Jente. This hinders us from
considering Italian c, J as the consonant diphthongs ts, dz
(p. 321). A palatogram of the first sound in ' ieri ' showed
SIMULTANEOUS AND SUCCESSIVE MOVEMENTS 369
that it was the consonant ], the word being ]eri. It is worth
noting that in general ' i ' has a vowel character where it cor-
responds to Latin vowels ('pia, via ') and becomes ] where
it replaces a consonant ('clarum-chiaro,' ' flammam-fiamma ')
and where it arises from diphthongization (' tenet-tiene ').
In a similar manner it was shown that in ' puoi, fuoco,
quinto ' there was a wholly or partly surd w after the initial
consonant. The words were thus phonetically pwoi, fwoko,
kwinto. This consonant w was often very weak or lacking
in some pronunciations ; ' puoi, ' for example , being the same
as 'poi.' The distinction in labial action was clearly shown
between w and u in records of the projection of the lips
(p. 355). This consonant w is a common Italian develop-
ment from a Latin labial vowel (o -> uo -* wo) or consonant
(5'M-* kw).
Simultaneous records of the air issuing from the mouth
and of the vibrations of the larynx (by an external capsule)
have been used for the study of the dialects derived from the
ancient Armenian. i In the popular speech of Constantinople
there are three ways of pronouncing the sonants b, g, d,
dz, dz, namely: 1. with the beginning of larynx tone preced-
ing the explosion by about 0.08'; 2. with the tone begin-
ning at the moment of explosion ; 3. with the tone beginning
slightly after the explosion, by 0.01= or 0.02^ There are
thus in fact three kinds of sonant explosives ; the first, corre-
sponding exactly to the French sonants, is employed in
cases of emphasis ; the second, corresponding to the German
sonants, may be called the standard sonants, since the great
majority belong to this class; the third class, appearing only
occasionally, resembles the corresponding surds but differs in
having less force and in having the larynx tone earlier in
the explosion. In the literary language of Constantinople
and in the dialect of Aslanbeg these sonants are regularly
pronounced like the aspirated surds (see below) but have less
force; sometimes, however, they are pronounced as in the
popular speech. In the dialects of Nouxa and Choucha
1 Adjarian, Les explosives de I'ancien arm^nien, La Parole, 1899 I 119.
24
370 PRODUCTION OF SPEECH
these sounds are fully sonant; the larynx vibrations begin
before the explosion, sometimes even as long as 0.10^ before.
In the dialects of Mouch and Sivas they have formed two
distinct classes ; the first is like the pronunciation at Nouxa
and Choucha; in the second the consonant is pronounced
with more force than in the first class and the mass of air
emitted is larger; in both classes the vibrations generally
begin 0.02^ to 0.03' after the explosion. The unaspirated
surds p, k, t, ts, ts are considered in both the popular and
the literary speech of Constantinople to be the same as the
sonants and are, as the records showed, pronounced in exactly
the same ways. At Aslanbeg and Sivas they have also
become sonants, the vibrations beginning at 0.015* to 0.08°
before the explosion. In the dialects of Nouxa, Mouch and
Choucha they have remained surd and are perfectly distinct
from the sonants and the aspirated surds. The aspirated
surds p", k*", t"", ts*", tS'^ fall into three classes in both the
popular and literary speech of Constantinople; in the first
the vibrations begin at 0.01* after the explosion; in the
second they begin just as the explosion begins ; in the third
they begin after explosive emission of air is completed (or
nearly so) — the first and third classes being rare. At Mouch
the pronunciation is like that of the second class (standard)
at Constantinople. In the dialects of Nouxa, Choucha,
Sivas and Aslanbeg these consonants are completely surd.
The bearings of these facts on the phonetic development of
Armenian have been discussed ; ^ the resemblance of some
of the Armenian sounds to the aspirates of the Sanskrit gram-
marians had already been noticed. ^
Simultaneous registrations of the movements of the jaws
and lips and of the tension of mouth-floor (p. 355) have
been made by Gallee and Zwaardemakee.^ Some of their
1 Rousselot-Meillbt, Note sur les evolutions phongtiques. La Parole,
1899 I 127.
* 2 SiEvERS, Grundziige d. Phonetik, 5. Aufl., 171, Leipzig, 1901.
3 GallSk xind Zwaardemaker, Veber Graphik d. Sprachlaute namentlich
der explosivae, Neuere Sprachen, 1900 VIII 8.
SIMULTANEOUS AND SUCCESSIVE MOVEMENTS 371
js/\/\/\f\M\/\tH'j\t^M^N^N^W^N^
Fig. 292.
records are given in Figs. 292 and 293. Eacli wave in the
time line at the bottom indicates ^^ of a second. The ver-
tical checks on the
curves indicate syn-
chronous positions of
the recording levers.
The upper line in each
case shows the jaw
movement; depression
indicates lowering of
the jaw. The upward
curves in the second
line indicate tension of
the upper lip. The
rise in the third line
indicates increase of
tension of the floor of
the mouth cavity.
The curves in Fig. 292 for mowoder ' moeder ' spoken in
the Deventer dialect show lowering of jaw for o and e and
tension of the lip for m and w.
The tensing of the mouth-floor be-
gins during m, increases strongly
during the rise of the tongue for o,
relaxes somewhat during w, in-
creases again for o, maintains
itself during the energetic tongue
articulation of d, increases during
the raising of the tongue for e and
falls for r.
For mudar ' moeder ' in the
North Holland dialect the records
are given in Fig. 293. The jaw
movements are analogous but
much more energetic; the lip is
tensed with much more energy
for m; the mouth-floor follows in general the same course
as before.
-V-
7n
rr.
*f"— ■"^'
r-r
Fig. 293.
372
PRODUCTION OF SPEECH
That a vowel may have an influence not only on the pre-
ceding consonant but also on the vowel before this con-
sonant has been shown by Laclotte.^ An exploratory bulb
(p. 333) was placed between the tongue and the palate and
a breath receiver (p. 219) placed before the mouth; they were
attached to two tambours. The records showed that the tongue
position during the consonant is lower in ba than in bi, in.za
than in zi, in za than in zi. Laclottb considers the records
to show that the tongue takes, for the beginning of the work
of articulation of the syllable,
the position necessary for the
vowel and maintains it through-
out the consonant and its
explosion. In da the d-artic-
ulation is frontal, in di it is
rather dorsal. Records of ela
and eli showed that the artic-
ulations differed for e, the
tongue being higher and nearer
to the i position in eli. Similar
results were found for eba and
ebi, venta and venti. The in-
fluence of a vowel on the artic-
ulation of the vowel of the
preceding syllable shows itself
in the tendency to vowel har-
mony and renders it possible to
explain why ' illi ' -^ French ' il ' but ' ilia ' -^ French ' elle, '
' viginti ' -* ' vingt ' but ' triginta ' -^ ' trente, ' and similar
cases. The auditory factor in vowel harmony and the ten-
dency to similarity in movements have been considered above
(p. 121).
The method of attacking historical problems by experi-
mental methods may be effectively illustrated by an investi-
gation by Laclotte.^ The words aliroKo^ (goat-herd) and
^ovk6\o<; (ox-herd) are evidently composed of two portions ;
1 Laclotte, L' Harmonie vocalique. La Parole, 1899 I 177.
Fig. 294.
SIMULTANEOUS AND SUCCESSIVE MOVEMENTS 373
the first designates the animal, al^ (goat) and /SoO? (ox) ; the
second is related to the Inao-European root qel which is found
in Latin as *quel, whence inquilinus, oolo, etc. In the occi-
dental European languages the evolution of q into k or p
has been independent of the neighboring vowels, whereas in
Greek it became regularly t before t or e ; « before o and after
ov or ib; and tt before o and after e, i, etc. Why should the"
labialization that occurs after the other vowels, not occur
after ov?
Using the vulcanite strip (p. 380) Laclotte recorded the
tongue positions for his pronunciations of e, i, o, u and ko ;
the typical sagittal diagrams are given in Fig. 294. For
u the tongue is drawn back, raised and held at 5°™ from
the palate, touching the edges with
the point of articulation at the last
molar. The action for ko is very
similar to that for u. For e and i
the action is quite different. It is
evident that it is much easier to
pass from the u than from the e or i
to the ko position. In the case of a j-j^, 295.
general tendency to change ko to po
this tendency would be favored after e and i but resisted after
u ; the p would afford opportunity for the tongue adjustments
after e and i while k is easiest between u and o.
With an exploratory bulb (p. 333) between the lips,
another at the point of articulation of k and a breath mouth-
piece (p. 219) pierced to allow the lip tube to pass, Laclotte
obtained the records shown in Figs. 295 to 297. For bukolos
(Fig. 295) the lip line shows the closure for b; the tongue
line shows the small posterior rise for b, the larger one for u,
and the still larger one for k, followed by the very small (pos-
terior) one for s; the breath line shows the explosion for b,
the air current for u, the closure and explosion for k and the
air current for the following sounds. The set of movements
in uk is continuous with easy transitions. For bupolos
(Fig. 296) the lip line shows the two closures for b and
374 PRODUCTION OF SPEECH
p ; the tongue line shows the posterior action for u with less
posterior action for p and none for o (but with some anterior
action for o, Fig. 294) ; the breath curve shows the closure
for p. There is more than double the amount of lip work
and more tongue work owing to
the changes between different arti-
culations. The form */Sou7roXo?
would be more difficult than
/3ovKo'Xo9. In aipolos (Fig. 297)
the lip line shows the pressure for
p, the posterior tongue action is
"'' ^ ■ small, the breath curve is marked;
the tongue at i is far from its position for k but is readily
relaxed to the p position. The tendency of g to p is thus
favored in al-n-o\o<s.
Simultaneous records may be made of the curve of speech
by the methods described in Part I, and of any of the mus-
cular activities by the methods described in this Part.
RoussELOT^ has made simultaneous records of the curve
of speech and of a muscular movement by using a phonauto-
graph (Ch. II) and an exploratory bulb (Fig. 245), and also
of the curve of speech and the breath pressure by attaching
the phonautograph to one arm of a Y-tube, a tambour to the
other and a mouthpiece to the stem. Some of his results will
be briefly mentioned. Characteristic records are given in
Figs. 298 and 299; they were taken on different occasions
and do not refer to identical sounds.
Fig. 298 shows the records of lip pres-
sure and voice vibrations in the middle
portion of apa. The gradual closing and
opening of the lips is seen in the upper
line ; the closure begins before the cords
cease to vibrate ; the a-p glide is thus
sonant. The opening, p-a glide, is likewise sonant. Fig.
299 shows the curve of breath pressure from the mouth and the
speech curve. The a-p glide is sonant; the explosion of the p is
likewise sonant. Tracings of the tongue pressure against the
1 EoussELOT, Principes de phonetique experimentale, 353, Paris^ 1901.
Fig. 297.
SIMULTANEOUS AND SUCCESSIVE MOVEMENTS 375
rear part of the palate, and of the speech curve during the vowel
a, showed the following formation. From its position of rest
against the palate the tongue descended to a minimum position ;
the tongue then gradually rose to a position of maximum articu-
FiG. 298.
lation and gradually fell again to a position lower than the
first minimum. This movement was probably compounded
of a gradual fall of the jaw and a fall-rise-fall of the tongue.
The voice-tone began during the fall to the first minimum ;
its intensity rapidly rose to a fairly maintained maximum and
fell with the relaxation of articulation at the close ; its pitch
followed about the same course. Analogous results were
found for other vowels. Records of the speech curve and
the breath pressure at the mouth for aj a and aia showed for
i little diminution in the intensity of either during i as com-
FlG. 299.
pared with a, but for j a gradual diminution of both during
the first and last portions [on- and off-glides] and practical
loss of the vibrations during the middle portion. In the com-
bination aga— a surd between two vowels — the change
376 PRODUCTION OF SPEECH
from a to 5 [the a-g glide] and from 5 to a [the 5-a glide]
appears in the records to be a sonant g, which must be
nearly identical with ]. Some resemblance was found be-
tween the speech curve during 1 or r and that during the
neighboring vowels. One example of lingual r showed about
32 flaps a second, a cord tone of about 140 frequency, and a
lower resonance tone of about 450. In asa the changes from
a to s and from s to a were gradual; similar results were
found for afa, aza, and ava. In aba, ama, ata, ada, aka,
aga the speech vibrations recorded from the mouth quickly
lost their amplitude as the occlusion was made and did not
regain them until it was almost completely over. The passage
from the vowel to the occlusion seemed to be more sudden
in French than in German. In apa, ata, aka two types of
explosion were found. In one the speech vibrations appeared
with the first portion of air at the explosion ; in the other
there was a considerable interval between the beginning of the
explosion and the beginning of the vibrations. The latter
type occurred in German and in emphatic French, the former
in ordinary French. In aba, ada, aga the speech vibrations
appeared at the first moment of the explosion, as would be
expected from the continued vibration of the cords during
the occlusion.
In concluding this chapter I may point out 1. that the
intention to perform two movements simultaneously is never
perfectly executed; 2. that the character of any movement
depends on other movements occurring at the same time;
3. that it depends on the immediately preceding movements ;
4. that simultaneous movements become associated with a
single intention; 5. that a course of movement is to be con-
sidered as a fluctuation in muscular action rather than as a
succession of sharply defined separate movements.
The adjustments of large groups of simultaneous and suc-
cessive movements in the flow of speech depend on auditory
, and inotor habits.
The speech experiences of an individual under given cir-
cumstances associate themselves together and with the cir-
SIMULTANEOUS AND SUCCESSIVE MOVEMENTS 377
cumstances. Two or more different sets of habits are
formed, and these may be associated with different circum-
stances. Cases are frequent of children with a foreign nurse
speaking the foreign tongue whenever addressed by the
nurse, even though she may use the native tongue.
In the same speaker the somewhat different, though corre-
sponding, sounds of two languages will usually be associated
readily in each lauguage singly but less readily from one
language to another. German words are spoken with more
difficulty when cited in the middle of an English- sentence
than when used alone. This is not due to any physiological
difficulty, but to interruption in the natural associations of
voluntary movements (p. 158). The so-called 'basis of
articulation ' (p. 113) is a system of habits of simultaneous
and successive motor impulses. The mental habits are best
formed by confining the instruction exclusively to the sounds
of the language being taught; alternate reading and trans-
lation of a foreign tongue makes the acquirement more
difficult (p. 151).
The importance of early training in pronouncing various
languages correctly is deducible from the general law that
habits formed in childhood are the most permanent ones.
In spite of later correction the old habits may often appear
in cases of fatigue ^ or of excitement. Vietor relates ^ an
observation of an actress who in emotional scenes used her
native "y, j in various German words instead of the g required
by the rules of the stage. I have heard a German ' Gymna-
siallehrer ' drop back into his native South German dialect in
the process of relating an exciting experience.
In forming habits of speech the movements should be made
as correctly as possible from the start. Every incorrect or
imperfect movement tends to create a habit of its own, which
must be overcome (p. 158) if improvement is to be made. In
the formation of a habit the immediately noticeable result
1 HoPFNER, Ueher d. qeistige Ermudung d. Schulkmder, Zt. f. Psych, u. Physiol.
d Sinn, 1893 VI 191
^ Vietor, Elemente d. Phonetik, 4. Aufl., 169, Leipzig, 1898.
378 PRODUCTION OF SPEECH
does not always bear a fixed relation to the amount of prac-
tice. It has been established by experiments on practice and
habit that the strength ^ and precision of control,^ after increas-
ing slowly for the initial stage, then show a stage of rapid
gain after which the increase is very slow. The highly im-
portant law ^ that ' it is intense effort which educates ' has
been estabUshed for the telegraphic language ; it is probably
valid for all habits.
References
For the physiology of simultaneous movements : see works on physi-
ology as in References to Ch. XV. For the psychology of simultaneous
and successive action : Wundt, GrundzUge d- physiol. Psychologie, 4.
Aufl.., Leipzig, 1893; Vorlesungen ii. Menschen- u. Thierseele, 3. Aufl.,
Leipzig, 1897 ; trans. London and New York, 1894. For the habits of
articulation : Sibvers, Grundz. d. Phonetik, 5. Aufl., 114, Leipzig, 1901;
Sweet, Primer of Phonetics, 69, Oxford, 1890 ; Beyer, Franzbsische
Phonetik, 2. Aufl., 59, Kothen, 1897; Vietor, Elemente d. Phonetik, 4.
Aufl., 262, Leipzig, 1898; Passy, Changements phonetiques, 245, Thfese,
Paris, 1891.
1 Fechnee, Ueber d. Gang d. Muskeliibung, Ber. d. k. sachs. Ges. d. Wiss.,
math.-phys. Kl., 1857, IX 113.
2 Bryan and Harter, Studies in the physio!, and psychol. of the telegraphic
language, Psychol. Rev. 1897, IV 27; Johnson, Researches in practice and habit.
Stud. Yale Psych. Lab., 1898 VI .51.
" Bryan and Haetee, as before, 50.
CHAPTER XXVII
VOCAL CONTROL
The degree of contraction of a muscle is governed by the
amount of stimulation from its nerve (p. 191). When a con-
traction of a definite nature has to be performed, the amount
of this stimulation must be regulated by reference to the con-
dition of the muscle at each moment of its action. When
several muscles act together, the subordinate centers for
separate muscles are controlled by higher centers for group-
action (p. 192).
The combined action of groups of muscles requires the co-
ordination of several centers of regulation by a higher center.
Thus, the production of a tone of constant character re-
quires regulation not only of the thyroarytenoid, cricoaryte-
noid and other laryngeal muscles, but also of the breathing
muscles. The production of any concrete sound requires
regulation not only of these muscles but also of the cavity
muscles and — for expression, gesture, posture, etc. — of
most of the other muscles of the body.
Vocal movements are controlled by systems of centers.
The center for the control of each single muscle lies in the
spinal cord or in the basal portion of the brain. Centers for
automatic control of separate activities, as of breathing, laryn-
geal action, etc., lie at higher levels, namely, in the bulb (p.l93).
Centers for voluntary control of the separate activities are
found in the cortex of the cerebrum in the anterior and pos-
terior central convolutions (for the head muscles of the right
side in the region between ' Arm ' and ' Speech ' in Fig. 57, for
the trunk muscles in the region above ' Arm.') The centers
380 PRODUCTION OF SPEECH
for voluntary control of combined activities are also in the
cortex (for vocal movements at ' Motor words ' in Fig. 57.)
They are all subordinated to the centers for ideas of speech (in
front of ' Motor words ' in Fig. 57) and these in turn to the
centers of thought supposed to be located in portions of the
frontal, parietal and temporal lobes (^F, P, T, association
centers of FlechsigI). The scheme of subordination is
partly shown in Fig. 58.
Any center may act to a great degree independently but its
intimate connection with the others makes it highly probable
that even such independent ■ action is influenced by their con-
dition. For example, we may safely say that although the
contraction of the thyroarytenoid muscle is brought about
directly by its special center it is nevertheless influenced at
each instant not only by the activities of the centers for the
other laryngeal muscles, of those for the pharynx, tongue, lip,
thorax, etc., and of the higher ones for song and speech, but
also by the activities of the still higher centers of mental life.
The intimate connection of all parts of the nervous system
leads us to suppose that the action of such a single center is
influenced to a greater or less degree by the activities of every
nerve center and that thus it stands in connection with every
part of the body. Experiments on the nervous and mental
reactions of the vaso-motor system, of the heart, of the
muscles of the sweat glands, bladder, anus, etc., make it prob-
ably safe to say that the production of any vocal sound is
accompanied by nerve impulses to and from every organ of
the body. Vocal sounds of a certain character, such as a
clear, smooth, energetic phrase in song, become associated
with the regulation not only of the vocal muscles but also
of those of the arms and hands, and, in fact, the entire bod}'.
The disturbance of any of these by restraint or unnatural
posture interferes — to a greater or less degree, depending
on the individual and on circumstances — with the vocal
action. To produce the proper modulation the singer or
1 Flechsig, Gehirn und Seele, Leipzig, 1896 ; Die Lokalisation d. geist.
Vorgange, Leipzig, 1896.
VOCAL CONTROL 381
speaker should put his entire body into the appropriate
condition.
On the psychological side we can draw analogous con-
clusions.
In first attempting to make a new sound or in' attempting
to notice the details of a speech movement, we are specially
' conscious ' of the movement or group of movements. As the
movement is repeated, it occurs with less and less attention
until it is made with no distinct knowledge of the perform-
ance (' automatically ' in one sense). In the last case it is said
to be quite ' unconscious, ' although when reminded of it we
may often remember a conscious fact that passed unnoticed at
the time. Even movements of which we can obtain no definite
consciousness, such as those of the muscles of the diaphragm
or the larynx, are probably represented in consciousness
by faint elements, for their influence on other elements can
be proved. With a pathological degree of attention they
may come distinctly into consciousness, as in hypochondria.
If the term ' consciousness ' is not limited to what is distinctly
present in mind, it is not too much, I believe, to assert that
all muscular movements are accompanied by some degree of
consciousness.
If objection is made to the use of the word ' consciousness '
in such a broad sense, we may say that all centrally originated
movements represent mental phenomena, of some of which
we are distinctly conscious, of others less conscious, and of
still others ' unconscious ' in the usual meaning of the word.
' Conscious ' might be used as a term representing a phenom-
enon varying between a maximum and zero ; ordinarily it is
used to represent it between a maximum and an undefined
lower limit, beyond which the phenomenon is said to be " un-
conscious.' To conform to ordinary usage I shall use ' con-
scious ' in the meaning of fully conscious, and shall speak of
' semi-conscious ' and ' unconscious ' as usual. ' Mental '
refers to any phenomenon that can be proved to affect any
element that may be ' conscious. ' The whole motor produc-
tion of speech is thus to be treated not only as a physiological
mechanism but also as a psychological process.
382 PRODUCTION OF SPEECH
The differences in tlie degree of consciousness of the sensa-
tions of movement are evident in the case of the vocal organs.
We are, under ordinary conditions, clearly conscious of the
positions and movements of the lips, much less so of those of
the tongue, and completely unconscious of those of the velum
and interior of the larynx. Moreover, we cannot in any way
acquire consciousness of some of them. No one can feel the
contact between the velum and the rear wall of the pharynx
or gain even the remotest notion of the action of the muscles
within the larynx. As already pointed out (p. 247), even
the simplest facts of laryngeal action were learned only by
aid of the laryngoscope, and our uncertainty to-day concerning
the interaction of the laryngeal muscles is defined by the
limits of clinical and experimental methods. I know of only
one case in which the internal action of the larynx was sup-
posed to be observable in consciousness; the result was the
supposition of a tone produced by the anatomically impossible
contraction of the larynx below the glottis. The notion of
many phonetists that any very definite knowledge can be
gained of tongue action by attending to its sensations is a
delusion of a kind familiar to psychologists.
Several factors of vocal control are now to be considered :
1. reflex-tonus ; 2. force of movement ; 3. accuracy of
movement; 4. precision of movement; 5. accuracy of co-
ordination ; 6. quickness of response ; 7. quickness of move-
ment ; 8. forms of sensory-motor control ; 9. ideo-motor
control ; 10. general voluntary control. Another important
topic, the adjustment of simultaneous and successive move-
ments, has been considered in detail in the preceding chapter.
Even the muscles apparently at rest in the body are
contracted to some extent. The utmost voluntary relaxation
of the hand hung over the edge of a table is not complete ;
during sleep it relaxes still more. The muscles of the face
relax in fatigue and sleep. This condition of faint continual
contraction has been named tonus; it is due to continuous
mild nerve stimulations sent from the spinal cord and brain in
response to sensations from the skin and elsewhere (Beond-
VOCAL CONTROL 383
GEEST reflex). The view that it is due to centrally originated
nerve action is erroneous; it has been clearly proved that
there is no tonal action of the central nervous system.
The effect of the degree of tonus on song and speech has
not been experimentally investigated. It may be suggested
that flabby muscles in the resonance cavities would diminish
the duration of the free vibrations on account of the loss of
energy at the soft walls ; this may be expressed as an increase
of the factor of friction k in the formula on p. 5. The effect
on the ear would be a change in the ' color ' of the vocal
sound in a way still undefined and yet readily recognizable
in the depressing voices of weak or.sick persons in compari-
son with a stimulating healthy voice. Such changes in
color appear as the result of fatigue, ill-health, and other
devitalizing conditions; in smaller degr^ .they result from
any disturbance — mental or bodily, such as'grief, disappoint-
ment, colds, the missing of a meal, etc. — thp^ diminishes
the vitality of the nerve centers. '*<V j
The attempt is instinctively made by the speaker or^^inger
to correct such a fault Iw voluntary innervation of the &f»g-T
cles; this cannot succeed perfectly because an increase M t±^,
innervation brings about- cdffti^ctions of associated and ap-^p*,
tagonist muscles with the result of changed conditions and
changed sounds. Such extra muscular effort is, moreover,
very fatiguing.
The lacking tonus can often be temporarily replaced by
drugs that act upon the nervous system; among them are
tea, coffee, strychnine and other tonics. There is evidence
to show that it can be temporarily or permanently improved
by influence — success, encouragement — that stimulates
mental activity.
The amount of tonus can often be measured by a tonal
dynamometer (or 'tonometer') consisting essentially of a
spring that registers the amount of pressure required to
impress the muscle.
The force of the movement depends on the amount of stimu- .^
lus sent to the muscle. The muscles contracted to perform a
384 PRODUCTION OF SPEECH
movement include not only those directly involved but also
their antagonists. This requires an excess of eifort over
what might be expected, but when the innervations are
properly coordinated this excess is not necessarily large.
In learning a new movement the contraction of both favoring
and antagonist muscles is unnecessarily large and fatiguing.
The presence of some contraction in all the muscles con-
nected with a group of movements is highly favorable to
quick and accurate control of the movement and its varia-
tions. This principle is involved in the contraction of the
entire abdominal wall or part of it during inspiration in
order to accurately control the expiration in singing. ^ In
singing the scale the chest and the abdomen often make
movements of expansion in opposition to the general expira-
tory contraction. This may be due to an attempt to adjust
the thorax to resonate to the cord tone,^ or, perhaps, to
adjustments giving more control over the expiration.
Changes of the force of movement in the organs of articula-
tion show themselves in various ways. The properties of song
and speech thact depend on the energy of action have not yet
been determined. Whatever they may be, they are instinc-
tively felt by the hearer and affect his general mental attitude
strongly. An energetic (not necessarily loud) voice in oratory
or in song commands attention and approval — other things
being equal — by appeal to some of the fundamental instincts
of the hearer. The first words of a born leader are often suffi-
cient to move an assembly. That this ability need not be
connected with a high grade of intellect has often been shown.
The results arise, I believe, from differences in motor en-
ergy, but their details must be left to a future experimental
analysis of speech curves, muscular activity and mental
impressiveness.
The distinction of ' tense ' and ' lax ' action, as in the for-
1 Maokenzik, Hygiene of the Vocal Organs, London, 1888; Cdktis, Voice
Building and Tone Placing, 68, New York, 1896.
'' Sewall and Pollard, On the relations of diaphragmatic and costal respira-
tion, with particular reference to phonation. Jour. Physiol., 1890 XI 159.
VOCAL CONTROL 885
mation of vowels, has to do with the energy of muscular
movement, which is associated with the degree of contraction,
and has nothing to do with the reflex tonus. Most of the
philological speculations in respect to tense and lax sounds
are probably erroneous as they do not agree with what is
known concerning muscular action; experimental data are,
however, lacking.
The force of effort may be measured by dynamometers of
various forms. The most usual measurement is that of the
pressure required to compress a spring or to raise a column
of mercury or water. Special dynamometers have been de-
vised for the lips, tongue and respiration muscles. A con-
venient dynamometer for the mouth can be made by attaching
an exploratory bulb (Fig. 245) to a water or mercury mano-
meter (p. 225).
The accuracy of movement may show itself in various
ways. The curve of movement described by a given point of
the moving body may not be the same as that intended.
Thus, a certain speech sound requires an elevation of the
point of the tongue along a certain definite line till it strikes
the prepalatal region ; any change from this line will produce
a change in the speech sound. With the other appropriate
speech adjustments this speech sound may be made the implo-
sion of a frontal-prepalatal t; any inaccuracy of movement
will produce a different implosion.
The inaccuracy of movement is a fundamental source of
inaccurate and wrong sounds.
The inaccuracy of the action of the cricothyroid muscle
produces inaccuracies in the cord tone (p. 269) which the
singer may be able to hear but powerless to correct. A tone
may be out of pitch, or may fluctuate in pitch , instead of
being constant; a succession of notes may be united by
glides instead of sudden jumps.
Inaccuracy in the various laryngeal muscles may produce
harsh tones instead of the smooth ones intended. Inac-
curacy in the breathing muscles may produce the fluctuations
heard in one form of tremolo, or a too rapid expenditure of
25
386 PRODUCTION OF SPEECH
breath which makes it impossible to sing properly, etc. Inac-
curacy of velar action produces nasality, modifications of
vowels, etc. Inaccuracies of tongue and lip action modify
speech essentially.
The methods of studying the accuracy of various vocal
movements have been described in the preceding chapters.
The movements can frequeatly be registered on a smoked
drum ; the defects are studied by the eye at leisure and the
improvement followed in the successive records. This method
has proved highly effective in correcting the defects of
singers and speakers.^
The precision of movefnent refers to the regularity and
evenness with which it is executed. It depends mainly or
entirely on the nervous control.
Accuracy and precision of coordination represent the ner-
vous control over simultaneous muscular adjustments. The
defects of one form of stammering arise from defective co-
ordination. A typical case, due to altered nerve activity,
may be found in the ' thickened ' speech of alcoholic intoxi-
cation. The typical form of defect due to excessive nerve
activity may be seen in the incorrect adjustments that arise
during excitement.
The accuracy and delicacy of the coordination of the
regulative centers vary greatly with individuals. Just as
in the case of painting or violin-playing the coordination
and regulation are naturally good in some persons and poor
in others; the degree to which they can be improved by
training is also variable.
The quickness of response in a movement depends mainly
on the rapidity of the action in the nervous centers and on
the number of centers involved in the reaction to the sensa-
tion. When full consciousness is involved, as in the reactions
discussed in Ch. XV, the time required is considerable. As
the reactions become less conscious (more ' automatic ') the
1 Natier et Rousselot, Les applications de la phonetique expe'rimentale \
la raedecine, Paris (in press).
VOCAL CONTROL 387
time is reduced. One object of vocal training should be to
render song or speech as automatic as possible.
The quickness of movement depends on both muscular and
nervous quickness. They must be properly balanced. Un-
usual slowness of nervous (mental) action renders speech ap-
parently labored and pedantic ; unusual quickness slurs it.
Hurried movements readily become inaccurate. The diffi-
cult rolled r often becomes unrolled. The effect of hurry
shows itself in careless utterances, like gmoin for ' guten
Morgen ' or sple for ' s'il vous plait.' A defective relation
between the speed of thought and the control of the vocal
organs results in inaccuracy in the exact formation of sounds.
The effect of extreme nervous haste can be seen in the defect
of speech known as ' Poltern ' in German.^ In German cases
i often sounds like e, u like o, oi like ai ; f and v often appear
as p and b, s and 2 as t and d ; p, t, k are often hardly dis-
tinguishable from b, d, g. Only rarely does a large change
of articulation take place ; in a few cases k, g -► t, d or t, d -+
k, g ; somewhat more often m -> n, n -> m, 1 -> r, r -> 1. The
errors are not constant as in stammering.
Quickness of movement can be studied by the graphic
methods described in the preceding chapters; the necessary
allowance must be made for the friction and inertia of the
apparatus.
Rapid, precise speech seems to have a stimulating effect
on the hearer. Its physiological and psychological character-
istics are still uninvestigated. The ability to speak rapidly
and clearly is associated with great activity of the nervous
system and of the train of thought. Owing to this associa-
tion such speech has a stimulating effect on the hearer and is
used by skiliful speakers for this purpose.
The sensory-motor control is generally muscular and
auditory.
The action of the vocal muscles occurs under guidance of
the sensations of movement obtained from them (p. 191).
1 LiEEMANN, Poltern (Paraphmsia praeceps), Vorlesungen iib. Sprachstorun-
gen, 4. Heft, BerUn, 1900.
388 PRODUCTION OF SPEECH
The association of the correct movement-sensation ordinarily
occurs with the aid of hearing the sounds produced. In the
deaf it occurs without this aid; the usual teaching of vocal
articulations to the deaf is done through touch and sight;
special teaching of the muscle sensations directly has been
shown to be of use.
The character of the direct control of single muscles with-
out the aid of control through other senses may be tested in
various ways. Labial, lingual and velar action may be
studied by graphic methods while sound movements are
silently made, the results being compared with those ob-
tained in the usual way; I am not aware of any experi-
mental work along this line.
Some factors of auditory- motor control have already been
discussed: the uncertainty and indefiniteness of auditory
sensations in Ch. VII I; the inaccuracy and lack of precision of
muscular movement in Ch. XV. In addition to these the in-
accuracy of the connection between sensation and movement
is of importance.
The amount of stimulation sent to the muscles at each
movement is governed by the sensations. Too much stimulus
at one instant produces too much contraction, and consequently
a change in the complex of sensations ; this is followed by a
reduction in the amount of nerve impulse. The reduction
is generally too great ; the sensations then vary in the reverse
direction; and renewed correction is attempted. For a con-
traction intended to be constant, as of the cricothyroid in
singing a tone of constant pitch, the continually fluctuating
and erroneously changing motor impulses produce changes
in the sensations from the tendons and in the pitch of the
tone heard (this last is not a factor of control in the deaf).
The intention to keep a constant pitch results in an adjust-
ment of the vocal centers to receive constant sensations and
to impart motor impulses standing in definite relations to
tfcem. The fluctuating sensations actually received are used
to regulate the impulses. An analogy may be drawn to an
engine with its governor; too great speed causes the governor
VOCAL CONTROL 389
to reduce the steam supply, and conversely; without a fly-
wheel to make the changes slow the engine would require
rapid readjustments by the governor. The vocal mechan-
ism is light and delicate ; its small inertia renders its action
very fluctuating (p. 269); it thus requires continual regu-
lative action. When a rising tone is desired, the governing
center is adjusted so that each degree of intensity of the
sensations is answered by an increased motor impulse. Fall-
ing tones are regulated by the opposite relation. A rise or
fall that seems steady to the ear requires a complicated —
probably not proportional, perhaps logarithmic (p. 109) —
relation.
The learning of speech sounds consists largely in forming
connections between the motor sensations and the auditory
ones. When such associations already exist, new sounds
are liable to confusion with familiar ones. The sounds of a
foreign language may be heard to resemble familiar ones
and the motor associations of the familiar ones become
attached to the new ones. The incorrectness of the associa-
tion is discovered to some extent by the speaker's hearing of
his own sounds, but remains also to some extent undetected.
The formation of ideo-motor associations has received little
attention from phonetists.
Sounds occurring simultaneously with sights, touches,
tastes, smells, emotions, acts of will, etc., tend to be con-
nected with them so that when any one of a complex group
occurs again the others are revived more or less clearly in
consciousness. It is in this way that speech movements
become associated with printed letters. The introduction of
new letters requires the formation of new associations ; the
use of letters to represent sounds in an unfamiliar manner is
resisted by the associations already formed. The neglect of
these evident facts is one reason why the phonetic alphabets
hitherto devised have all failed to find general acceptance.
The close interconnection of all the nerve centers indicates
that the action of any one may influence all others. Among
others the intellectual and emotional centers influence the
390 PRODUCTION OF SPEECH
X-
vocal centers and consequently the vocal mechanism. E
perimental data are stiU entirely lacking except in regard to
internal speech in its effect on nasal whispering (p. 132) and
on the action of the larynx (p. 112). Yet from what is
known of other activities it can be safely asserted that the
character of the vocal movements in song and speech depends
most intimately on the mental condition. From this we can
readily deduce a conclusion used in musical and oratorical
instruction that the singer or speaker must feel what is to be
said if he wishes to say it properly.
The ideo-motor associations affect the contractions of the
various vocal muscles and consequently the character of
the air-vibration produced. At present the most promising
method of investigation seems to lie in analyzing speech
curves (Part I) obtained in connection with various modes
of thought (conversation, declamation, etc.) and emotion (ex-
citement, anger, etc.). The effect of the mental condition
consists exclusively in modifications of the factors of vocal
control just considered. That the ' mind ' can affect the air-
vibrations of the singer or speaker, or directly communicate
with the ' mind ' of the hearer is a superstition born of igno-
rance and credulity.
The yet-uninvestigated minute auditory variations in the
sounds of speech and song have great effects on the mind
of the hearer. Although no one can say in what the vocal
difference consists, there is an intimate connection between
the mental attitude of the hearer and the voice-character of
the speaker. This relation is well known to speakers and
singers; fatigue, worry or embarrassment often seriously
affects the voice.
A highly important problem — still hardly investigated
experimentally — lies in the relation of vocal control to the
emotions. Every change in the emotional condition results
in changes not only of the action of the involuntary muscles
(heart, blood vessels) but also in changes in breathing and
vocal action. The resulting changes in the voice have power-
ful emotional effects on the hearer. These emotional changes
VOCAL CONTROL 391
in vocal action cannot be perfectly reproduced by the speaker
voluntarily in the absence of the emotion. It is a familiar
principle with orators and singers that to produce the full
vocal effect they must first arouse the emotion itself and
then allow it to find its natural expression.
For singers and speakers we may safely say that vocal
training should include not only a development of the vari-
bus other factors of control but also a thorough practice in
voluntarily bringing up the typical mental conditions and in
properly expressing them.
To phonetists we may point out that the object of speech
is the attainment of a certain result, that this may often be
done by very different muscular adjustments, and that mus-
cular adjustments do not by any means go all the way in elu-
cidating phonetics. ' It is well to remember that ridiculous
old paradox quoted by Galen from the Stoics: "It is evi-
dent the voice cometh from the mind ; it is evident also it
cometli from the larynx ; hence the mind is not in the brain."
Galen splutters over this a good deal, and fails entirely to
see its bearing; but Galen had very little esprit. If he had
seen and heard what can be done without a larynx at all, he
would not have considered without reservation the larynx as
the " principalissimum organ um vocis," as the Medievals put
it in their Hog-Latin. No two larynxes are alike, and doubt-
less if we could get down to very fine points it would be
found that more or less identical results are reached by very
different dynamics and neural discharges.' (Wright.)
It is now the place to consider the dependence of vocal
control on the general voluntary control, as in changes of
nutrition, fatigue, emotion, and general habits.
The laws governing the amount of energy used in action,
its rate and accuracy of expenditure, its curve of fatigue, its
dependence on emotions and motives, etc., have been studied
with great success in the case of arm and finger movements
with results which — on the principle of similarity in all actions
of an individual — are directly applicable to vocal move-
ments. Experimental records have not yet been made directly
392 PRODUCTION OF SPEECH
on the vocal organs, but we can assume that each individual
has his own peculiar forms of vocal movement; that he
expends his vocal energy in a fashion peculiar to himself ;
that the forms of vocal movement and of vocal expenditure
change in their details with every mental condition ; that they
are directly expressive of conditions of thought, emotion and
motive ; that they cannot be completely changed from one ex-
pression to another except by changes in the mental conditions.
When the nervous system is well nourished, its elements
accumulate quantities of higlily complicated substances.
These represent potential energy which may be turned into
kinetic energy by the breaking down of the more highly com-
pounded substances into simpler ones. This kinetic energy
shows itself in the forms of mental and muscular work. The
discharged energy can bfi replaced by re-formation of the com-
plex substances through nutrition.
A large store of potential energy shows itself mentally —
other things being equal — in a feeling of good nourishment.
Whether this feeling arises directly from the nervous system
or indirectly from the nature of the reflex-tonus and action
of the internal organs and muscles, may be left undecided.
A small store of potential energy is accompanied by a feel-
ing of weakness.
The condition of good nourishment shows itself in 1. in-
creased energy of the reflex-tonus ; 2. increased force, quick-
ness and precision of movement; 3. increased accuracy of
coordination and association.
These are the fundamental factors in the production of good
vocal sounds. Flabby muscles, that is, with poor tonus, do
not have the firm configurations necessary for forming cavi-
ties with firm walls. Weak muscles cannot hold out against
the work required of them. Slow muscles cannot perform
to perfection the rapidly changing adjustments. Inaccurate
control produces inaccurate results. Poor coordination pro-
duces defective results. The typical forms of these results
hftve just been considered; their variations in conditions of
poor nourishment have not been experimentally investigated.
VOCAL CONTROL 393
The defective results are often perceived by the speaker
or singer; he instinctively tries to correct them by extra
muscular exertion. The over-exertion of some muscles re-
quires over-exertion of their antagonists in order to obtain
the proper positions. This brings about an abnormal condi-
tion of the vocal organs and difficulty in movement and con-
trol. The larynx tone, for example, becomes a strained and
fatiguing falsetto with little flexibility instead of a readily
modulated chest tone ; there is often huskiness of the voice
owing to lack of precision in the vocal muscles ; the weakened
breath action has to be strengthened by a special effort.
The effects of fatigue are analogous to those of poor nour-
ishment. The activity of a day's work or of direct electrical
stimulation has been shown (Hodge) to result in a decrease
in the active elements of the nerve cells. Fatigue, from
this point of view, might be said to be the exhaustion of
nourishment. Another element in a condition of fatigue
arises from the presence of toxic products in the blood result-
ing from organic activity.^
The vocal effects of fatigue are marked; fatigue can
appear in the voice even before it does in the face. The
effects resemble those of poor nutrition but are not quite
the same. Just what their elements are has not been ex-
perimentally determined; they probably appear, as in all
voluntary action, in lack of steadiness, precision, quickness,
endurance, etc.^
The voice in singing depends not only on the structure
of the vocal organs and the ear but also on all the factors
of vocal control that we have considered. These should be
taken into account at the outset of special vocal instruction
in order to avoid a mistaken career, and also for the possi-
bilities of improvement. The troubles of singers arise not
merely from larj'ngeal and aural defects but largely from
those of control.
1 Mosso, Ueber d. Gesetze d. Ermudung, Arch. f. Anat. u. Physiol. (Physiol.
Abth.), 1890 Supplem.-Bd. 89.
2 Scripture, New Psychology, Ch. XVI, London, 1897.
394
PRODUCTION OF SPEECH
Laryngoscopic and otoscopic examinations should be sup-
plemented by accurate tests of vocal action and auditory
perception, by examination of their association, by various
tests of control, and by as thorough a study of mental and
bodily peculiarities as may be practicable.
We have now to consider some typical methods of altering
the vocal control in defective speech ; no general treatment
of the subject wiU be attempted.
Defects of articulation often arise from lack of sensitive-
ness in regard to the motor organs, combined probably with
a lack of acoustic sensitiveness which is usually developed
Fig. 300.
Fig. 301.
only in connection with correct articulation. They can
often be corrected by special methods and devices. Two
methods may be employed, education of the sensations of
the motor organs directlj^, and education by appeal to an-
other sense.
The direct education of the motor organs maj^ be illus-
trated by the following cases. In producing s the tongue is
usually slightly curved over the lower alveolae,^ the anterior
dorsal portion being raised toward the upper incisor teeth and
the palate without touching them, while the sides are pressed
strongly against the molars and the upper lateral alveolae.
T^ this way a small narrow channel is formed through which
the air rushes (Fig. 300). This may be called the ' interior
1 ZuND-BuKGUET, De, la prononciation de I's et du s, La Parole, 1899 I 281.
VOCAL CONTROL
395
lingual-alveolar s.' The s may also be formed by placing
the point of the tongue behind the superior incisors so that
it does not touch either the teeth or the alveolae (Fig. 301).
The rush of air then occurs between the point of the tongue
and the palate. This may be called the "anterior linguo-palatal
s. ' It is frequent in the speech of Roumanians. It is never
so intense or clear as the other s; it readily becomes
a kind of soft s if the point of the tongue is slightly pushed
backward, or a lisped M if it is advanced a trifle too much
toward the teeth. To transform the lisped s into the nor-
mal s ZuND-BuRGUET uscs a Uttle wire loop (Fig. 302) that
catches the point of the tongue and directs it to the proper
Fig. 302.
Fig. 303.
position. Great numbers of lispers have been cured with this.
In producing s the sides of the tongue usually rest along the
alveolse and the superior molars (see preceding Chapters) ;
the point of the tongue lies quite free between the anterior
part of the jaw and the palate; the medio-dorsal portion rises
toward the palate and forms with the point a little depres-
sion whose depth varies (Fig. 303). This little medio-
lingual cavity forms a sort of resonance chamber that is of
great importance for the sound. A second cavity is formed
between the teeth and the lips by projecting the latter. The
most frequent fault arises from touching the point or the
dorsum of the tongue to the palate, whereby the sides of
the tongue leave the teeth, and the air finds issue at one or
both sides instead of through the medio-lingual cavity ; this
396 PRODUCTION OF SPEECH
fault has been called ' chlintfement ' by Roussblot on ac-
count of the resemblance of the sound to a kind of mixed s
and 1. The defect is readily cured by practice with a wire
that directs the tongue down.^ It is to be noted that this
description of the formation of s differs considerably from
those given by most other investigators, who do not specify
the existence of any such medio-lingual cavity.
The ■ auditory-motor associations may often be advan-
tageously replaced by visual-motor ones.
The appeal to the eye is regularly made in teaching the
deaf to speak. It is also used to advantage in the cases of
persons whose defective movements cannot be corrected by
auditory teaching. The manner oi doing this for improv-
ing a foreigner's pronunciation has been illustrated by
RODSSELOT.2
In the case of a child of 11 years who used t for k, s
and s, d for g, and 1 for z, without yielding to any auditory
correction and without evidence of any auditory or vocal defect,
RoussELOT succeeded in producing the correct articulation
by an appeal to the eye. He placed an exploratory bulb (Fig.
245) in his own mouth at the point of articulation for the t
and pronounced successively ta and ka ; the long lever of a
tambour (p. 195) connected to the bulb made large movements
for ta and none for ka. The child understood the difference,
and after practice with a bulb in his own mouth was able to
produce the sounds correctly. For s, z, s and z the lever
was arranged to pass over a paper with the proper position
for each sound marked on it. Two lessons were enough to
correct the fault in this case.
In teaching the distinctions of articulation among the
palatal consonants, Meuniee^ has used two thin rubber
chambers attached to an artificial palate (Fig. 304), each
attached to a tambour (Fig. 305); as the correct articulations
1 ZuND-BoEGUET, as before, 285.
» ^ RoussELOT, Applications pratiques de ta phon^tique experimentaie. La Parole,
1899 1401.
' Meunier, Emploi de ta mAliode grapliique pour t'£ducation des sourds-mueti
La Parole, 1900 II 82.
VOCAL CONTROL
397
are made, the lever of one of the tambours attached to the
chambers will point to the proper letter on the cardboard
diagram.
A similar method is of use ^ in
teaching the proper pressure to be
used in a certain articulation ; a small
bulb is used in teaching the precise
place of articulation, a medium bulb
in teaching the lip-pressure and a large
bulb in teaching certain vowels. A
mouth-piece inclosing a large bulb
is useful in teaching the proper lip
action, as, for example, in the French y (m). A bulb
applied beneath the chin indicates the proper degree of
Fig. 304.
Fig. 305.
1 ZUND-B0RGUET, Applications pratiques de la phonetigue expmmentale, La
Parole, 1899 I 18,45, 149,
398 PRODUCTION OF SPEECH
tongue retraction for vowels like a, o, u. A bell-like alarm
placed over the larynx serves to indicate sonancy. Zund-
BuEGUET Hses an indicating alarm tambour to impress on
the eye and ear the results obtained by air-transmission;
the movement of the arm can be seen ; an adjustable bell,
placed at the point which should be reached^ by the move-
ment, indicates to the ear the success of the articula-
tion. The points to be reached in pronouncing various
vowels may be indicated on a scale above the pointer; the
movements necessary to produce the differences between
closely related vowels can thus be taught. The indicator may
likewise be used with a lip-bulb to teach such differences as
those between e, i and y. Attached to a breath mouth- ,
piece, the indicator shows impressively the difference between
the greater expense of breath in English or German p, t, k
and the smaller expense in French p, t, k, in which the
glottis is closed during the explosion (Figs. 279, 280).
Numerous other uses can be made of the Zund-Btjeguet
apparatus.
Refere^tces
For hygiene of the voice in singing and the troubles of singers:
Mackenzie, Hygiene of the Vocal Organs, London, 1888; Flataq,
Hygiene d. Kehlkopfes u. d. Stimme, Stimmatorungen d. Sanger, Hey-
niann's Handb. d. Laryngol. u. Rhinol., I 1448, Wien, 1898, (full lit-
erature) ; Krause, D. Erkrank. d. Singstimme, Berlin, 1898 ; Browne
AND Behnke, Voice, Song and Speech, London, 1895. For vocal training :
Mackenzie, as before ; Curtis, Voice Building and Tone Placing, New
York, 1896; Stockhausen, Gesangsmethode, Leipzig. For diseases
of speech : see References, p. 88. For the literature of fatigue :
JoTEYKo, Revue generale sur la fatigue musculaire, Ann^e psychologique,
1899 V 1 (many important references lacking) ; Mosso, La fatigue
intellectuelle et physique, Paris, 1894; Binet et Henri, La fatigue
intellectuelle, Paris, 1898.
PART IV
FACTOES OF SPEECH
CHAPTER XXVIII
VOWELS
The necessity of a study of the physical nature of the-^
vowels was emphasized by Willis. ' The mouth and its
apparatus were constructed for other purposes besides the
production of vowels, which appears to be merely an inci-
dental use of it, every part of its structure being adapted to
further the first great want of the creature, his nourishment.
Besides, the vowels are mere affections of sound, which are
not at all beyond the reach of human imitation in many ways,
and not inseparably connected with the human organs, al-
though they are most perfectly produced by them ; just so,
musical notes are formed in the larynx in the highest possi-
ble purity and perfection, and our best musical instruments
ofter mere humble imitations of them ; but who ever dreamed
ot seeking from the larynx an explanation of the laws by
which musical notes are governed? These considerations in-
duced me, upon entering on this investigation, to lay down a
different plan of operation; namely, neglecting entirely the
organs of speech, to determine, if possible, by experiments
upon the usual acoustical instruments, what .forms of cavities
or other conditions are essential to the production of these
sounds, after which, by comparing these with the various posi-
tions of the human organs, it might be possible, not only to
deduce the explanation and reason of their various positions,
but to separate those parts and motions which are destined
400 FACTORS OF SPEECH
for the performance of their other functions, from those which
are immediately peculiar to speech (if such exist). ' i
Willis's idea of studying the physical characteristics of the
vowels has been developed by a series of later observers,
finding its full expression in the study of curves of speech by
the investigators referred to in Part I. In its perfection the
' physical definition of a vowel ' will consist of a mathematical
expression for the course of the molecular vibration which
it involves.
The nature of the vibrations in spoken vowels can perhaps
be made clear by a study of the records in Plate II.
The curves^ shown in the Plate are from a record con-
taining the nursery rhyme of Cock Robin, spoken by an
American. The words in the Plate occur in the following
phrases: ' I ' in ' I, said the beetle,' ' my ' in ' With my bow
and arrow,' ' parson ' in ' I '11 be the parson,' 'saw him ' in
' I sajv hjm die, ' ' caught ' in ' Who caught his blood ? ' and
'said' in ' I, said the rook.' The record was traced off as
described in Ch. IV. The equation beneath the Plate in-
dicates the relation between length and time.
The curve for ' I ' shows a series of vibrations in which each
group resembles the neighboring one, while there is a gradual
*T;hange in character from a typical form for the a in the first
part to a tj'pical form for the i in the second part of the
diphthong ai of which the pronoun ' I ' is composed. In the
first portion there appears a succession of strong vibrations,
each followed by a series of weaker ones. These strong
vibrations recur at periods of steadily decreasing length.
If we consider separately each group of vibrations beginning
with a strong one, we find that it is, aside from minor details,
the typical curve (Fig. 4) of a vibration initiated by a blow
and dying away by friction, for which the equation is
z/ = a . e"*' . sin 27r^ ,
' Willis, On vowel sounds, and on reed-organ pipes. Trans. Camb. Phil. Soc,
1830 III 231 ; also in Ann. d. Phys. u. Chem., 1832 XXIV 397.
2 ScKiPTnRE, On the nature of vowels, Amer. Jonr. Sci., 1901 XI 302.
VOWELS 401
where y is the elongation at the moment t, a the amplitude,
e the basis of the natural series of logarithms, k a factor repre-
senting friction and T the periodic time' (p. 6).
The succeeding groups of vibrations following the first '
group are of the same form but of steadily increasing ampli-
tude. They recur at steadily decreasing intervals. The for-
mula for each group is approximately the same except for the
difference in amplitude. The vibrations are evidently
aroused by a series of blows (p. 11) of steadily increasing |
strength at steadily decreasing intervals.
It seems clear that these vibrations represent the free vibra-
tions of the air in the mouth cavity aroused by a series of
sudden blows and that these sudden blows are due to explo-
sive openings of the vocal cords (p. 260).
The tone from the cords results from the succession of
groups of vibrations ; it is a tone of intermittence (p. 94).
The period of the tone from the cords is represented by the
distance from the strong vibration at the beginning of each
group to the strong one at the beginning of the following
grotip (p. 65).
The method of studying the details of such curves has
been given in Ch. V.
The complexities of the small vibrations indicate the pres- v
ence of several partial tones. These complexities change
steadily from the beginning of the vowel onward as the pitch
rises, in a way to indicate the presence of at least the follow-
ing partials: 1. the fundamental cord tone consisting of a
series of explosions rising from a period of 0.0170' (fre-
quency, 59) to one of 0.0052' (frequency, 192); 2. a con-
stant cavity tone of 0.0034' period (frequency, 294); 3. a
constant cavity tone of 0.0013' period (frequency, 769) and
4. higher cavity tones undergoing change.
The minor complexities in the vibrations disappear at ■
about one-quarter of the distance from the left on the second
line in the figure. At the same time the amplitude is
strongly increased. Shortly afterward the amplitude de-
creases and finally reaches zero. Throughout the whole
402 FACTORS OF SPEECH
latter portion the curve has an entirely different character
from that of the first half ; we are probably quite safe in con-
sidering it the curve of i in the diphthong ai. Throughv.
out the i the groups consist of two vibrations, one slightly-
stronger than the other. The period for the group 0.0052^
(frequency, 192) remains constant till near the end, where it
lengthens to about 0.0122" (frequency, 82). The cavity
vibration forming half of each group remains constant
at 0.0026' (frequency, 384) through nearly all of the i.
Toward the close it still apparently remains at the same
period, producing phenomena of interference as the group
period is lengthened.
From the curve for i it seems justifiable to conclude that \
the vocal cords emit explosions instead of sinusoid puffs of
air here as well as in the a. The explosion produces a
strong free vibration in the mouth cavity which is followed
by another of diminished amplitude. This would be fol-
lowed by a third of still less amplitude, just as in a, but a
new explosion from the cords occurs at just that moment.
The coincidence of double the period of the cavity tone
with the period of the cord explosions explains the rapid gain
in amplitude when the cord tone rises sufficiently to produce
the coincidence (p. 13). The maximum is followed by a
relaxation in the force of breath, but the two tones main-
tain the same relation for a considerable time. As the
sound finally dies away, the cords also relax, both breath and
pitch falling together. The explosions from the cords seem
much less sharp in i than in a.
In ' my ' the m vibrations are too faint for accurate measure-
ment. The a resembles somewhat, but not closely, the a of
" I. ' The period of the cord explosions remains constant at
0.0074= (frequency, 135) instead of decreasing. The lower
resonance tone has a period in the neighborhood of 0.0022'
(frequency, 455); it apparently undergoes a slow change
from the beginning of the a to the i.
The last third of the curve somewhat resembles the i por-
tion of ' I. ' There is, however, only a faint rise in ampli-
VOWELS 403
tude, and the i portion is very brief. The vibrations in this
portion are in groups of three; the groups have a period of
0.0074= (frequency, 135) constant to the end. The vibra-
tions virithin the group have a period one-third that of the
group itself, indicating a constant cavity tone of 0.0025'
(frequency, 400).
In the a of ' parson ' the cord tone rises from a period of
0.0090» (frequency, 111) to one of 0.0072= (frequency, 139)
and falls again to the pitch from which it started. There
are indications of a constant cavity tone of 0.0022= (fre-
quency, 455) and of higher tones vrith changing periods.
In respect to the pitch of the lowest cavity tone there, is
close agreement of this a with that of ' my,' yet the form of
the curve resembles that of a in ' I ' more closely than that
in ' my. ' The peculiarity of ' my ' seems to lie chiefly in the
suddenness with which the vibrations within a group fall in
amplitude after the initial strong vibration. In both ' parson '
and ■ I ' the cavity vibrations within each group during a die
away less quickly. Such differences may perhaps find their
explanation either in the greater friction in the free vibratory
movement in the mouth (less rigidity of the walls ?) or in
the sharper character of the cord explosions in the case of
'my.'
The curve for o in ' saw him ' indicates a quite different
vocal action from that present in a. Instead of a strong
initial vibration followed by decreasing ones the earlier por-
tion of the vowel shows groups that contain at least two
strong vibrations. It is presumably the case that the cord
explosions are of a more gradual character or else that the
action of friction is much less. Even later in the vowel
where there is apparently only one very strong vibration in a
group, this probably occurs because the lower portion of the
second one is cut off by interference with another partial tone.
The cord tone, starting with a period of 0.0072= (frequency,
139), remains at this pitch for a time and then falls to 0.0080=
in period (frequency, 125). A lower cavity tone with a
period of 0.0026= (frequency, 385) is apparently present.
404 FACTORS OF SPEECH
The last part of the line shows the vibrations for i, resem-
bling those for i in ai of ' I ' and ' my. ' The middle portion,
where there is a weakening in amplitude, belongs to the
sonant h (p. 277). The m is just begun where the record
is cut off. The grouping in the i is by threes. The cord
tone of i starts with a period of 0.0083= (frequency, 121) and
steadily rises to one of 0.0072^ (frequency, 139) in the m.
The lower cavity tone has a period of about 0.0025= (fre-
quency, 400).
Another example of ' saw him ' from the same record was
partially discussed in Ch. V and on page 277.
The curve for the o of ' caught ' exhibits a decided differ-
ence from that for the o of ' saw, ' although both vowels
are generally considered to be the same. The o of ' caught '
shows a quick and strong increase in amplitude followed by
a rather sudden decrease. Its pitch is approximately con-
stant. The initial strong vibration of a group is followed by
very much weaker vibrations; the vocal action resembles
that in a rather than in the o of ' saw. ' In the last few
groups there is a marked change as the o alters to t.
The cord tone rises from a period of 0.0074' (frequency,
135) to one of 0.0064= (frequency, 156) but falls again in
the last few periods. The lower cavity tone seems to have
a period of about 0.0024= (frequency, 417). Other tones
of higher pitch are present.
In the e of ' said ' the vocal action is seen to differ essen-
tially from that in a or o, and to resemble somewhat that in
i. There is much less indication of the explosive character of
the cord tone. There are three cavity vibrations to each
group. The pitch of the cord tone is nearly constant at
0.0072= period (frequency, 189); the lower resonance tone
has a period of 0.0024= (frequency, 417). There are minor
fluctuations in the curve that indicate higher cavity tones.
The amplitude increases steadily until the vowel is ended
rather abruptly by the change to d.
The preceding account gives in general the pitch of only
the lower cavity tones in each vowel. A determination for
VOWELS 405
the higher tones would require more elaborate methods. It
IS probable that the higher tones are quite as important
for the vowel characters as the lowest ones. The disagree-
ment in the accounts of various investigators in regard to
the tones found in the vowels may have arisen partly from
finding different ones.
The results seem to justify the conclusion that the move^
ment of the air in the mouth cavity is a free vibration and not
a forced one. The 'curves of spoken vowels given in Plate
II all show that the mouth tone is constant even while the
cord tone is steadily changing. It follows from these facts
that the period of the mouth tone is independent of the period
of the cord tone and that there is no necessary relation be-
tween the adjustment of the size of the mouth cavity and
the tension of the vocal cords. If the period of the mouth
vibration is independent, it must be the period of the free or
natural vibration. j.
Two theories have been held concerning the relation be-
tween the cord tone and the cavity tone ; these may be
termed the Willis-Hermann and the Helmholtz theories ;
the supporters of each have tried to prove them in various
ways.
Willis^ fitted a reed to the bottom of a funnel-shaped
cavity and obtained sounds resembling vowels by modifying
the opening of the cavity. He then tried closed cylindrical
tubes of different lengths and found that different vowel-like
sounds were produced by different lengths of the tube (p. 290).
His experiments led him to the conclusion that the vowel-like
sounds are produced by the repetition of one musical note in
such rapid succession as to produce another. ' It has long
been established, however, that any noise whatever, repeated
in such rapid succession at equidistant intervals as to make
its individual impulses insensible, will produce a musical
note. For instance, let the musical note of the pipe be g^
and that of the reed c^ which is 256 beats a second, then their
1 Willis, On vowel sounds, and on reed-organ pipes, Trans. Canib. Phil. Soc,
1830 m 231 ; also in Ann. d. Phys. o. Chem,, 1832 XXIV 397.
406 FACTORS OF SPEECH
combined effect is y"^ ■ • • g'^ ■ • • g"^ ■ ■ • g'^ • • • (256 in
a second) in such rapid equidistant succession as to produce
ci, g'^ in this case producing the same effect as any other noise,
so that we might expect a priori, that one idea suggested by
this compound sound would be the musical note c^.
' Experiment shows us that the series of effects produced
are characterized and distinguished from each other by that
quality we call the vowel, and it shows us more, it shows us
not only that the pitch of the sound produced is always that
of the reed or the primary impulse, but that the vowel pro-
duced is always identical for the same value of s [the length
of the pipe]. Thus in the example just adduced, g"^ is pecu-
liar to the vowel o [as in 'all']: when this is repeated
256 times in a second the pitch of the sound is c^, and the
vowel is o : if by means of another reed applied to the same
pipe it were repeated 171 times in a second, the pitch would
be ,P, but the vowel still o. Hence it would appear that
the ear in losing consciousness of the pitch of s [the length
of the pipe] is yet able to identify it by this vowel quality.
But this vowel quality may be detected to a certain degree in
simple musical sounds; the high squeaking notes of the
organ or violin speak plainly i, the deep bass notes u, and
in running rapidly backwards and forwards through the
intermediate notes, we seem to hear the series u, o, a, e, i,
i, e, a, o, u, etc., so that it would appear as if in simple
sounds, that each vowel was inseparable from a peculiar
pitch, and that in the compound system of pulses, although
its pitch be lost, its vowel quality is strengthened. . . .
Having shown the probability that a given vowel is merely
the rapid repetition of its peculiar note, it should follow that
if we can produce this rapid repetition in any other way, we
may expect to hear vowels. Robinson and others had
shown that a quill held against a toothed wheel would pro-
duce a musical note by the rapid equidistant repetition of the
snaps of the quill upon the teeth. For the quill I substituted
a piece of watch-spring pressed lightly against the teeth of
the wheel, so that each snap became the musical note of the
VOWELS 407
spring, the spring being ab the same time grasped in a pair
of pincers, so as to admit of any alteration in length of the
vibrating portion. This system evidently produces a com-
pound sound similar to that of the pipe and the reed, and an
alteration in the length of the spring ought therefore to pro-
duce the same effect as that of the pipe. In effect the sound
produced retains the same pitch as long as the wheel revolves
uniformly, but puts on in succession all the vowel qualities
as the effective length of the spring is altered, and that with
considerable distinctness, when due allowance is made for the
harsh and disagreeable quality of the sound itself.'
Thus Willis maintains two theses: 1. that a vowel con-
sists of [at least] two tones, a cord tone and a mouth tone ; 2.
that the mouth tone is independent of the cord tone in regard
to pitch.
~ The theory of Willis was adversely criticised by W heat-
stone, ^ who supposed that the vowels arose from the vibra-
tions of the vocal cords through the strengthening of certain
overtones by the resonance of the mouth. Whbatstonk's
view was expounded as a general hypothesis by Grassmann ^
and developed into a theory by Helmholtz.^
' We may well suppose that in tones of the human larynx,
as in those of other reed instruments, the overtones would
continuously diminish in intensity as their pitch is higher, if
we could observe them without the resonance of the mouth.
In fact they correspond to this assumption fairly well in those
vowels that are spoken with widely-opened, funnel-like
mouth-cavities, as in sharp a or e. This relation is, however,
very materially changed by the resonance in the mouth. The
more the mouth-cavity is narrowed by the lips, teeth, or
1 Wheatstone, London and Westminster Review, 1837, p. 27.
2 Grassmann, Leitfaden d. Akustik, Program d. Stettiner Gymnasiums, 1854 ;
Ueber d. phi/sik. Natur d. Sprachlaute, Ann. d. Phys. u, Chem., 1877 I 606.
3 Helmholtz, Deber d. Vokale, Arch, f, d. holl. Beitr. z. Natur- u. Heilk., 1857 I
354; Ueber d. Klangfarbe d. Vokale, Gel. Anz. d. k. bayr. Akad. d. Wiss., 1859
537 ; also in Ann. d. Phys. u. Chem., 1859 CVIII 280, and in Ges. wiss. Abhandl.,
I 395, 397, Leipzig, 1882; Die Lehre y. d. Tonempfindungen, 5. Aufl., 168,
Braunschweig, 1896.
408 FACTORS OF SPEECH
tongue, the more prominently its resonance appears for tones
of very definite pitch, and by just so much more it thus
strengthens those overtones in the tone of the vocal cords
which approximate the favored degrees of pitch; and by just
so much more the others are weakened. ' ^
' The pitch of the strongest resonance of the mouth depends
only on the vowel for whose production it has been arranged,
and changes essentially even for small changes in the charac-
ter of the vowel, as, for example, in various dialects of the
same language. On the other hand, the resonances of the
mouth are almost independent of age and sex. I have found
in general the same resonances for men, women, and children.
What is lacking to the childish and female mouth in capacity
can be easily replaced by narrower closure of the opening, so
that the resonance can still be as deep as in the larger male
1 mouth.' 2
According to Helmholtz, ' the vowel sounds are different
from the sounds of most musical instruments, essentially in
the fact that the strength of their overtones depends not only
on the ordinal number of the overtone but above all on its actual
pitch. For example, when I sing the vowel a on the note
e—^ '', the reinforced tone is h^, or the 12th partial, and when
I sing the same vowel on the note b"^ it is the second one. ' ^
The theory of Helmholtz necessitates the assumption of an
accommodation of the resonance tone to the voice tone within
quite a range ; thus as the voice tone rises or falls the mouth i
must also change its tone or be able to extend its resonance '
to a considerable degree. This assumption was made by
Helmholtz, the range of accommodation being supposed to
extend over as much as an interval of a fifth in music each
way from the tone of best resonance. This view has been-
, called the ' accommodation theory.' According to this theory
^ the mouth must accommodate itself to one overtone of the
cord tone, and when this rises or falls to a considerable degree
' Helmholtz, Die Lehre v. d. Topempfindungen, 5. Aufl., 170, Braunschweig,
1896.
2 Hklmholtz, as before, 171.
^ Hei,.\[ii(iltz, as before, 191.
VOWELS 409
it must readjust itself to some other one in order to keep the
resonance tone within a limited range.
The difference between the theories of Willis and Helji^
HOLTZ lies chiefly in the relation between the mouth tone and
the cord tone; for the former there is no relation, for the
latter the resonance tone. is one of the overtones of the.
cord tone.
Helmholtz devised an apparatus^ of electric tuning forks
and with some success produced Yowel-like sounds by combin-
ing different sets of tones (p. 291).
'Willis's description of the acoustic movement in the
vowels doubtless coincides closely with the truth; but it gives
only the manner in which the motion occurs in the air, and
not the corresponding reaction of the ear to this motion.
That even such a motion is analyzed by the ear according to
the laws of resonance into a series of overtones is shown by
the agreement in the analysis of the vocal sound when it is
executed and by the resonators. ' ^
Helmholtz was greatly influenced in his theory by his •
views of the action of the ear.
Tl'e hypothesis that all regular vibratory movements reach-
ing the ear are analyzed by it into a series of harmonics of the
fundamental period is an assumption that seems to lead natu-
rally to the Helmholtz theory. This assumption, however,
we must disregard at the present time ; the problem concerns
the nature of the vibratory movement characterizing a vowel
and the solution must be found in an unbiased analysis of the
vowel curve ; the question of how the ear acts is a separate
one.
• The Helmholtz theory was for a long time accepted in the
main by later writers. It was made the basis of Pipping 's
first analysis of vowel curves.
Pipping's^ work with Hensen's instrument (p. 20) led 1
him to the following conclusions.
1 Helmholtz, as before, 191.
2 Pipping, Zur Klangfarbe der gesmgenen Vohile, Zt. f. Biologie, 1890
XXVII 77.
%
\J
410 FACTORS OF SPEECH
f ' In agreement with Helmholtz I have found that each
I vowel is distinguished by one or more regions of reinforce-
ment of constant pitch. The intensity of its partial tone is,
caeteris paribus, greater as it coincides more accurately with
the range of reinforcement.
' In regard to the range of the reinforcement I cannot agree
with Helmholtz. Helmholtz does indeed state that the
range can be different according to the opening of the mouth,
the firmness of walls of the oral cavity, etc. But he lays so
little weight on this difference that he does not attempt to
use it in the characterization of the different vowels. To
judge from page 183 of the " Lehre von den Tonempfin-
dungen, " Helmholtz thinks that the range of reinforcement
must extend in general at least a musical fifth above and
below, and this is certainly not the case.
' Sung vowels contain only harmonic partial tones. ' That
is, a vowel produced by singing consists of a series of tones
whose vibrations stand in the relations of 1 : 2 : 3 : 4 : etc.
' The intensities of the various partial tones do not depend
to any essential degree on their ordinal numbers. ' That is,
in distinction from most musical instruments it is not a fact
that the first partial is much the stronger and that the higher
partials are in general weaker.
' The various vowels differ from each other in ranges of
reinforcement which are- of different numbers, width, and
position in the scale of pitch. ' That is, one vowel may have
two ranges of reinforcement, another three, etc., and these
ranges may differ.
On a later occasion i Pipping believes that the range of \
accommodation may exceed even the limits allowed by \
Helmholtz.
.— - The first point at issue between the two theories may be
thus stated: is a cavity tone found in a vowel necessarily
an overtone of the cord tone?
• .Among the results that support the view of Willis we
1 Pipping, Zur Lehre von den Vokalkldngeu, Zt. f. Biologie, 1895 XXXI 573,
583.
VOWELS 411 -
may notice those obtained by Donders with the Scott
phonautograph (p. 17).
'Each of the fourteen vowels when sung on a constant tone
produces a constant curve! . . . For each vowel the form V
of the curve changes with the pitch. This result is connected
with the peculiarity of the vowels, that their timbre is deter-
mined not by overtones with a certain relation to the funda-
mental, but rather by overtones of a nearly constant pitch. ' ^
This last statement rests on the fact that if the tones of
the mouth are overtones of the cord tone bearing a definite
relation to it, such as 1st, 2d, etc., the curve will remain i/.
the same in form no matter what the pitch,- just as the curve -^
of vibration for a violin string has a typical form which per-
sists in spite of changes in the pitch of the string. On the
other hand, if the tone of the mouth is a constant one, as
Willis assumes, the combined vibration produced, by the
cord tone and the cavity tone will change for any change
in the pitch of the cord tone. Donders's conclusion seems
indisputable.
Hermann's investigations were carried out by transcribing
the curves of song from the phonograph (p. 39). He finds
that the essential fact in a vowel is the intermittent or oscilla-
tory blowing of the cavity tone by the cords. Under such
circumstances it makes no difference whether the cavity
period coincides with any fraction of the cord period or not.^
Hermann thus supports the theory of Willis in asserting\
that the cavity tone is completely independent of the cordj
tone.
Hermann has objected to the overtone theory of the cavity y^
tone that in many voices it is so high above the cord tone ^
that it cannot be supposed that an overtone of that pitch can
possibly be present. Thus with the cord tone g "^ the vowel
i has a strong cavity tone that would correspond to the 28th
1 Donders, Zur Klangfarbe der Vokale, Ann. d. Physik u. Chemie, 1864
CXXXIII 528.
2 Hermann, Phonophotographische Untersuchungen, Arch. f. d. ges. Physiol.
(Pfliiger), 1890 LXXIV 380, 38L
\
412 FACTORS OF SPEECH
or 29th partial of the cord tone, whereas such a high partial,
if present at all, would be too weak to be heard. ^
In his latest investigations Pipping ^ finds that the cavity v
tones are independent of the cord tone and abandons the J^
Helmholtz theory.
Similar independence of the cavity tone in some of the
vowels when sung appears in the work of Merritt- (p. 27),
Nichols and Merritt (p. 28), Bevier (p. 49), and others.
In the analysis ^ of the curves of many cases of the diphthong
ai (App. II) in words spoken in the chest register by a male
voice, I find that the cavity tone in a is quite independent
of the cord tone in pitch. The cavity vibration can be seen A^
to remain of constant period while the cord tone rises through J
a distance of several octaves within a single vowel. *-/
Hermann's researches * on the modifications of tones by%^
the telephone show that the partial tones of a complex are
weakened by the transmission according to a definite rule: "Js^
the amplitudes are transmitted in relative amounts directly
proportional to the frequencies.^ Thus, the three partials
100, 200, and 300 with the original amplitudes a, h,c would
have after transmission the relative amplitudes f, |, ^. The
lower the original tone the weaker is the result. Bass
music is greatly weakened in comparison with soprano. In
spite of this difference between the original and the trans-
mitted sound the vowels retain their specific characters ; the
relations of intensity between the tones of a vowel are there-
fore not essential characteristics of the vowel. / Thus, the~\
mouth tone in a certain vowel may differ greatly in intensity )
while that of the cord tone remains constant and yet they
vowel will be heard as the same one.
1 Hermann, Phonophotographische Untersuchungen, Arch. f. d. ges. Physiol.
(Pfluger), 1894 LVIII 274.
^ Pipping, Zur Phonetik d. finnischen Sprache, Mem. de la Socie'te finno-
ougrienne, XIV, Helsingfors, 1899.
^ Scripture, Researches in experimental phonetics {first series). Stud. Yale
Ksych. Lab., 1899 VII 1 ; On the nature of vowels. Am. Jour. Sci., 1901 XI 302.
* Hermann, Die Uebertragunq der Vokale durch das Telephon und das Micro-
phon. Arch. f. d. ges. Physiol. (Pfliiger), 1891 XLVIII 543.
* Hermann, as before, 561.
VOWELS 413
In the same research Hermann also shows that the rela-
tions of phase have no influence on the character of the musi-
cal sound heard. 1
These results of Hermakn's appear irreconcilable with any—,
vowel theory that regards the mouth as a resonator strength-
ening one or more of the partial tones of the cord tone.^
The failure to produce vowel curves by adding harmonics
(p. 69) and to produce vowel sounds by compounding tones
(p. 291) have already been mentioned. To these may be
added the failure of wave discs representing sums of har-
monics to give vowel sounds when used with a slit blast in
a siren. (p. 89). Curves for the wave siren produced by
summing harmonic sinusoids according to the data given
by AUERBACH for the vowels gave no satisfactory results.^
Similar curves according to a Fourier analysis (p. 71) of
the vowels made by Lahr * gave ^ a and e well, o and u
poorly, i not at all, and u for y.
In the face of such conclusive evidence it is hard to see
any point in which a decision in favor of the theory pro-
posed by Willis and developed by Hermakn can possibly
be attacked. It is natural to assume that a theory found to
be valid for one vowel will be valid for all; it is, of course,
possible that other laws may hold good in other vowels, but
until this possibility is proven we can treat all vowels on
the independent-tone theory; at any rate, the cavity tone is
not necessarily an overtone of the cord tone.
Several problems concerning the physical nature of vowels
still remain: 1. the 'nature of the cord tone; 2. its method of
arousing the cavity tones; 3. the nature of the cavity tones.
These will now be considered. ' "^
Willis supposed the cord tone to be produced like a reed
1 Hermann, as before, 560.
2 Hermann, as before, 566. ,oo, ytv
B KcENiG, Bemerkungen iib. d. Khnqfarbe, Ann d, Phys. u. Chem , 1881 XIV
369 • also in Quelques exp&iences, 234, Paris, 1882.
4 Lahr, Die Grassmann'sche Vokaltheone im Lichte des Expermentes, Ann. d.
Phvs u Chem. 1886 XXVII 94.
fE^CHHORN, P.. Vokats>rene, Ann. d. Phys. u. Chem., 1890 XXXIX 148.
I
414 FACTORS OF SPEECH
tone, namely, by a series of explosive puffs. That this may
be the case has been shown by the curves of Hekmann
(p. 39); he says that an explosive puff followed by an
interval of ' cord silence ' (if I may use the term) is one of
the essential characteristics of a vowel. Such a series would
be similar to that emitted by a siren with holes passing before
an air jet (p. 89).
This view is also supported by the following facts. A
vibratory body, whatever its natural period, when acted upon
by a force varying harmonically, must itself vibrate with the
period of the impressed force. If the variations of the act-
ing force are of the nature of a sum of harmonics, the period
impressed will be that of one of them. If the cords acted
like most musical instruments, their vibrations could be
properly treated as the sum of a series of harmonics and the
mouth tone would necessarily be one of them. The forced
vibrations of the mouth cavity can include only harmonic
partials of the larynx note.^
It has been shown above that the mouth tone is inharmonic
to the cord tone and that it is a free vibration. It follows
that the cord vibrations are not of the nature of the sum of
a series of harmonics. Hermann draws the conclusion that
the vibrations of the cords must be of an explosive nature, to
which a harmonic analysis is not applicable. To this it has
been answered that when the cavity tone is high in relation
to the cord tone, the treatment by analysis into a series of
harmonics may not be applicable and that this may not dis-
turb the usual views of resonance, but that, when the natural
period of the vocal cavity is not distant from the cord
period, the cavity must vibrate with a period that is har-
monic to the cord period.^ Rayleigh apparently does not
regard the deductions of Hermann as conclusive. The issue
seems clearly presented in the curves of the nature of those
in Plates I and II. In the first part of the vowel a in ' I ' the
cord tone rises steadily till it is only about a duodecime below
1 Ratleioh, Theory of Sound, 2d ed., II § 397, London, 1896.
" Rayleigh, as before, § 397.
VOWELS 415
the cavity tone, and yet the latter remains constant with
no tendency to become one of the harmonics of the cord
tone. Continuing along the curve, we find that beyond the
middle the period of the cavity vibration is somewhat length-
ened while that of the cord vibration continues to become
shorter. In the latter third of the curve the vibrations ar^^
clearly in groups of twos, alternate ones being stronger. As
the change from the a portion to the i portion is continuous
without anything like a break that might indicate a sudden
readjustment of the cords, each pair of waves in the i portion
must belong to one cord vibration and each single wave must
represent a cavity vibration. The cavity period is slightly -~~
less than half the cord period. Thus, even when the two
tones used in forming the vowel i are nearly in the relation
of a simple musical interval, there is no accommodation of
one to the other in respect to period. It is to be noticed
that the first vibration of each pair in the i is stronger than
the second, just as in the a portion the first vibration is
stronger than the following ones. It is worthy of remark
that the relative strengths are not the same in the two cases,
and that the character of the explosion from the cords must
differ to some extent in the two halves of ai. Similar condi-
tions are found in the other vowels.
Such relations between the cord vibrations and the cavity
vibrations are incompatible with the theoretical requirements
of the supposition of the sinusoidal nature of the cord vibra-
tion. The conclusion seems quite justifiable that the cords i
emit a series of puffs, or explosions of air, instead of vibrat- 1
ing regularly back and forth. The same conclusion was \
reached above (p. 260) from a study of cord action.
The sharpness of explosion is, I believe, a matter of degree. \
Hermann seems to consider all explosions as very sharp ; the
intermittence — or period of cord silence — he has even as- y
serted to be an essential of the vowel character. That the /
intermittence is not necessary can be seen in Hermann's
curves for i, 1, etc. (p. 39). In my curves I find vowels with
all degrees of sharpness of explosion.
X
416 FACTORS OF SPEECH
I would amend the Willis-Hermann view by saying that
the cords emit a series of puffs whose nature may vary from
the sharpest of explosions to a perfectly smooth sinusoid. I
would also add, as already stated (p. 94), that the character
of the sound emitted by the cords depends essentially on the
na^re of the puff.
tx-^xhe manner in which the cord tone arouses the cavity
tones can now be definitely stated.
\ Willis considered the mouth and cords to be analogous
to a reed pipe. Each vibration of the reed sends a wave of
condensation and rarefaction along the pipe. When the pipe
is of such a length that this wave is reflected back in such a
way as to reinforce the vibration of the reed, the cavity
tone is a loud one. Thus, when a properly adjusted reso-
nator is placed behind a vibrating fork the tone of the fork
is strongly reinforced (p. 14). The reinforcement is also
strong when the resonator period coincides with a sub-multiple
of the reed period.
Such a coincidence between the periods of the cavity tone
I and the reed tone is not necessary. Each impulse from the
reed may be considered as striking the pipe with something
of the nature of a blow, whereby the proper tone of the cavity
itself may be aroused for an instant (p. 285). The pipe may
thus have its own pitch and be heard, no matter what rela-
tion there may be between it and the pitch of the reed.
When the blow from the reed is rapidly repeated, both the
reed tone and the pipe tone will be heard (p. 94).
Such a method of producing cavity tones has been
declared to be impossible by Hensen,^ who remarks that air
from a reed pipe cannot arouse a resonance tone. The ex-
periment on which he bases this statement consisted in placing
a resonator at the end of a reed pipe. At a certain pressure
of air the pipe sounded its own tone, at a different pressure
it was silent. The resonator sounded only when the pipe
was silent. Nevertheless there were occasions when both the
pipe tone and the resonance tone appeared together.
1 Hensen, Die Barmonie in den Vokaten, Zt. f. Biol., 1891 XXVIII 39.
VOWELS 417
To these experiments and deductions Hermann replied
that a labial pipe can be used to sound a reed pipe, and some
experiments were made to demonstrate the fact.i
I have attempted in another way to show that a series of
puffs of air of any periodicity may be used to sound a labial
pipe of any pitch.
A disc with its edge cut into waves forming approximately
a sine-curve (Fig. 2) was rotated by an electric motor at any
desired speed. Its edges passed between the ends of two
pieces of rubber tubing so arranged that the air blown into
one of them passed directly into the other one if the waves of
the disc permitted ; the position was so chosen that the waves
of the disc regularly interrupted the air current completely.
The farther end of the rubber tubing was flattened and placed
so as to blow against the edge of a piece of brass pipe stopped
at the other end. In this manner a series of puffs from the
disc was used as the blast of a pipe. The experiment began
with the disc at rest and the air passage free. A current of
air was blown through the tubing ; the pipe gave forth a tone.
The disc was then set in rotation ; the tone of the pipe was
regularly intermitted. As the disc moved faster, this inter-
mittence became more rapid. Finally, the intermittence itself
was heard as a tone in addition to the pipe tone (p. 94).
Thus an intermittent air current, such as is employed for pro-
ducing tones directly (p. 89), can be used to produce a pipe
tone in addition. The development of this apparatus into a
vowel producer has been mentioned on p. 293.
I have succeeded in arousing the tone of a closed tube
by blowing through a membrane pipe (p. 258). The pipe
was made by binding a piece of thin soft rubber around
the end of a glass tube. Two opposite points of the thin-
walled rubber tube thus made were each caught between the
thumb and finger ; the membrane was then stretched till the
sides came together. A blast of air through the tube set
these edges in vibration and produced a tone. By placing
1 Hermann, Weitere Untersuchmgen u. d. Wesen d. Vokale, Arch. f. d. ges.
Physiol. (Pfluger), 1895 LXI 195.
27
418 FACTORS OF SPEECH
the edges at the right spot over the mouth of a bottle or a
test-tube or a key (Fig. 74) the tone of the latter could be
distinctly heard in addition to that of the pipe. The pitch of
the membrane tone could be altered at will. It was not so
easy to arouse a tube of low pitch, such as a bottle, because
the volume of air passing through the pipe was not large.
It can thus be regarded as definitely settled that the cur-
rent of air from a reed can be used to arouse a tone in a
cavity properly adjusted to receive the air. To this state-
ment we may add that the reed tone and the cavity tone may
vary independently of each other, but that the cavity tone is
loudest when its pitch is higher than that of the reed tone.
Fig. 306.
Willis's view of the way in which the cavity tone was
superimposed on the reed tone is very explicit. ' According
to EuLEE,! if a single pulsation be excited at the bottom of a
tube closed at one end, it will travel to the mouth of this tube
with the velocity of sound. Here an echo of the pulsation
will be formed which will run back again, be reflected from
the bottom of the tube, and again present itself at the mouth,
where a new echo will be produced, and so on in succession
till the motion is destroyed by friction and imperfect reflection.
«
1 EuLER, Conjectura pht/sica circa propagationem soni, Opuscula II ; and in
Tentamen novae theoriae musicae, Petropoli, 1739.
VOWELS 419
. . . The effect, therefore, will be a propagation from the
mouth of the tube of a succession of equidistant pulsations
alternately condensed and rarefied, at intervals corresponding
to the time required for the pulse to travel down the tube and
back again; that is to say, a short burst of the musical note
corresponding to a stopped pipe of the length in question
will be produced. ' ^
The true view of the action of the cords in producing a
cavity tone seems to be the following one. The sudden
puff of air from an explosive opening of the cords may be
considered to act as a piston compressing the air before it in
the vocal cavity. The air acts as a spring by its resistance
to compression and drives the piston back beyond its position
of equilibrium ; the resistance to dilatation draws it back, and
so a vibratory movement is set up. Under these circum-
stances the air acts merely as a spring; the form of the cavity
is immaterial and the period of vibration remains the same,
provided the capacity is not varied. The single impulse
of the piston thus makes the cavity a source of vibration,
whose period remains practically constant but whose ampli-
tude steadily diminishes from loss of energy mainly by com-
munication to the external air (Fig. 4). Such vibrations are
seen in the curves for a in Plates I and II. This view is an
adaptation of that given by Rayleigh for resonators in
general (p. 28). «—
The question still remains as to the nature of the tone thus
toduced in the cavity.
There are cases in which the Helmholtz view of the
action of the vocal cavity might seem to have a possibility
of correctness. If we assume (1) that a uniform condition
has been attained, (2) that the natural period of the cavity
does not differ greatly from that of the cord period, and (3)
that the cord vibrations are not of an explosive nature, it
follows that the effect of the cavity can only be to modify the
intensity and phase of the partials of the cord note.. The
partial or partials nearest to the natural periods of the mouth
1 Willis, as before, 243.
V-
420 FACTORS OF SPEECH
cavity will be reinforced, and they can be found from the
speech curve by the Foueiee analysis. The cavity tone
must thus have a period of one of the overtones in the cord
tone.
Under the assumptions made above, the vibration of the
cavity is a ' forced ' one, and the conclusion concerning the
action of the vocal cavity is necessarily correct. ^ The first
and second assumptions made above have been explicitly
stated by Rayleigh, who concludes that both the Willis-
Hbkmann and the Helmholtz ways of treating the action
of the cavity are legitimate and not inconsistent. ' When
the relative pitch of the mouth tone is low, so that, for
example, the partial of the larynx note most reinforced is the
second or the third, the analysis by Foueiee's series is the
^ proper treatment. But when the pitch of the mouth tone is
high, and each succession of vibrations occupies only a small
fraction of the complete period, we may agree with Hermann
that the resolution by Foueiee's series is unnatural, and that
we may do better to concentrate our attention upon the actual
form of the curve by which the complete vibration is
expressed. ' ^ The two forms of treatment imply that the
\ cavity tone is to be considered in the one case as a free
\ vibration of the air in the cavity, and in the other case as a
forced vibration. Some cases of i in my study ^ of ai may be
■ reconciled with the Helmholtz view, the resonance tone
being an overtone of the cord tone and changing with it.
AH cases of a and most of those of i are decidedly inconsis-
tent with the overtone theory. Possibly the variation from
the overtone theory arises from the explosive manner in which
the cords open. The general description of their action for
ai probably holds good even when the cavity tone is only
about an octave above the cord tone; each puff of air is
stronger at the start and fades away, setting the air in the
vocal cavity into free instead of forced vibration. This gen-
eral characteristic can be traced in each ai even to the point
1 Ratleigh, Theory of Sound, §§ 48, 66, 322k, 397, London, 1894, 1896.
2 Rayleioh, as before, § 397. s Appendix II.
VOWELS 421
where the resonance tone is slightly lower than the octave of
the cord tone.
In my opinion the explosive blow theory (free vibration of
the vocal cavity) and the overtone theory (forced vibration)
express the conditions in the extreme cases. When the
puffs have infinitely sharp forms the former is necessarily cor-
rect ; when they are sinusoidal the latter is also necessarily
correct. Puffs of forms between these extremes will modify
the waves from the vocal cavity according to their forms.
For a very sharp puff the vibration in the speech curve will
have the period of the cavity with an initial amplitude de-
pending on the energy of the puff; the vibration will fade
away rapidly. For a puff of the same total energy the initial
amplitude will be less as the puff becomes less sharp, the
period remaining that of the free vibration of the cavity;
the vibration will fade away less rapidly. For sinusoid puffs
the cavity vibration will have an amplitude depending on the
relation between the natural period of the cavity and the
period of the puff; for harmonic relations it will be greater, for
inharmonic ones smaller; there will be no fading away.
We are thus justified in defining a vowel physically as a
vibratory movement, consisting of a series of puffs of more or
less explosive form and of one or more free frictional sinusoids J
(aroused by the action of each puff) whose periods are those/
of the natural tones of the cavities.
Free vibrations (p. 2) are frictional sinusoids (p. 5) when
there is no constant supply of energy (p. 14). Since the soft
walls of the mouth cause great damping or loss of energy, the
value of the frictional factor k (p. 6) is large and the vibra-
tions die away quickly. The rapidity of decrease may be
clearly seen in the resonance yibrations in the first portion
of ' I ' in'' Plate II. As the ^uffs from the cords become
stronger, the free vibration lasts longer. As the puffs come
at shorter intervals, the last vibration does not have time
to die away before the impulse comes from the next puff.
The phenomenon of resonance thus appears, as shown in
Fig. 14.
422 FACTORS OF SPEECH
^We will now consider the auditory nature of a vowel.
A sung vowel can be heard to include at least-. one tone,
namely, the voice, or cord, tone. Even a spoken vowel has
some general pitch character to it. This cord tone is prac-
tically lacking in whispered vowels (p. 274). Since the
whispered vowels can be distinguished from other sounds
and from each other, and since they seem to vary in pitch,
it is evident that vowels possess other tones than the cord
tone. Since these vary with the adjustments of the vocal
cavities, we may call them ' cavity tones.'
It is clear that a sung vowel consists of at least the cord
tone and one cavity tone. Does the vowel character depend
on the relation between these two tones ?
A direct proof that the vowel characteristic cannot lie in
such a relation has been obtained by singing vowels into the
phonograph going at the usual speed and then reproducing
'lihem at quite different speeds.^ A change in the speed of
the record not only changes the pitch of the cord tone but
also changes the vowel. In Hermann's experiments,^ with
increase of speed e approached the sound of i, u that of o,
and finally all vowels approached a sound between e (ct) and
ce ; with decrease of speed all the vowels approached a bleat-
ing sound resembling ce. In running a celluloid phono-
graph at different speeds I have found that the French a
in ' pas ' changes to e as the speed is increased, and to o as
it is decreased. Experiments of the same kind by RoussE-
LOT^ show that changes in speed produce systematic modi-
fications in the vowels, the amount of change required to
reach a given modification being different for different
speakers. With decreasing speed the vowels appear to be-
come as indicated in the following list: Hg -^ o -♦ Oj ^ oe^ ; aj
1 Blake and Cross, Helmhoitz's vowel theory and the phonograph. Nature,
1878 XVIII 93; Cross and Wendell, On some experiments with the phonograph
relating to the vowel theory of Helmholtz, Proc. Amer. Acad., 1892 XXVII 271.
2 Hermann, Oeber das Verhalten der Vokale am neuen Edison'schen Phono-
graphen, Arch. f. d. ges. Physiol. (Pfliiger), 1890 LXXIV 42.
8 RoussELOT, Principes de phonetique expe'rimeutale, 226, Paris, 1897.
VOWELS
423
^ Eg ->
oi -* «! ; ei
-^ 63 -» D -* CEi -> oe, ; Co -> i„
-y'-
<^i ~* CEg ; ig -> gradually weaker ig or -+ y ; ccj -> gradually
weaker
OBj, or ^ cCg
5 "^3 ^ gradually weaker ocg, or
-^y, or
->u; y
-^ gradually
weaker y =* oe^, or -^ gradually
weaker
y-^u;
03 ^u; Oi -
•^ O3 -* u ; u ->• gradually weaker v
1. The
inferior
numerals
indicate varieties of a vowel
as in
Ch. XXIII.
These phonograph experiments show not only that an im- ^
portant essential of the vowel character does not lie in the V\
relation of the cavity tone to the cord tone, but also that
it does lie in the presence of a tone of a limited range of
pitch.
Similar evidence is furnished by the attempts to manufacture
vowels. When a band of metal is carefully cut so that its
edge reproduces the curve of a vowel, and is made to pass in
front of a narrow slit from which a blast of air issues, the
vowel itself is distinctly heard when the frequency with
which the waves of the edge pass the slit is the same as
that of the original cord vibrations, and therefore when the
cavity tone is of the original pitch. At other frequencies the
vowel sound appears modified. An essential characteristicY ^
of many vowels is thus a fixed cavity tone.^
The supposition that a vowel requires a cavity tone of a
certain pitch can be tested by removing that tone. This
has been done ^ by the use of interference tubes adjusted to
kill any tone with its even-numbered overtones or with its
odd-numbered overtones. Most vowels (especially a and o)
became nasalized by extinction of the cavity tone and its ^
odd overtones, with a greater or less change of the special
vowel character. All vowels with high cavity tones (e, i,
ce, y, ae,) became a deep indefinite murmur when the chief
cavity tone was removed. With the extinction of its higher
cavity tone a ^ o" ; with extinction of its lower one a -> ae°
1 Hermann, Ueher die Priifung von Vokalkurven mittels der Komg'schen Wetlen-
sirene. Arch. f. d. ges. Physiol. (Pfliiger), 1891 XLVIII .'574.
2 Sadbebschwarz, Interffrem-Versuche mit Vokalkldngen, Arch. f. d. ges.
Physiol. (Pfliiger), 1895 LXl 1.
^
424 FACTORS OF SPEECH
though rather indefinitely. The vowel ae was completely anni-
hilated by loss of its cavity tone. From the agreements and
differences in his results Saubeeschwarz concludes that
the vowel-characteristic lies in tones of certain relations of
pitch, in some vowels a fixed tone being the most prominent
element, in others a certain overtone of the cord tone.
The supposition i that the essential of vowel character lies
solely in the relations of two or more cavity tones to each
other and that these tones may be of any pitch is devoid
of the slightest foundation (p. 292). A relation between
cavity tones of fixed pitches may be suggested ^ as requisite
for the vowel character, but this seems hard to reconcile with
Hbemann's telephone experiments (p. 412).
The conclusive proof of the relations of the cavity tones in
the vowels is given by exact determinations of the vibrations
necessary to produce the proper effect on the ear. The con-
siderations presented in this Chapter have shown that the
vowel-character — that is, its distinctiveness to the ear —
depends at least partly on fixed cavity tones.
The *■ essential characteristic ' means that characteristic
which is most effective in enabling the ear to distinguish
certain groups of sounds that we designate as a, e, i, etc.
There are other possible groupings of vowel sounds and
other characteristics of them all and singly. One character-
istic that varies among them is the character of the cord puff.
This seems to differ in the various vowels. It has already
been noted (p. 291) that u and o can be approximated by
using tones with sinusoidal puffs, while the others cannot.
Generalizing from the individual peculiarities of different
speakers, I would say that a sung vowel consists of a voice
tone with its various overtones, and of various cavity tones,
these tones being in pitch independent of the voice tone.
' Lloyd, Speech sounds ; their nature and causation, Phonetische Studien, 1890
III 275, 278 ; 1890 IV 39 ; 1891 V 125 ; The genesis of vowels, Jour. Anat. Phys.,
1^97 XXXI 233.
^ M'Kendkick, Observations on the theories of vowel sounds, Proc. Roy. See.
Edin., 1897-1899 XXII 71, 87.
VOWELS 425
The different vowels are distinguished!, by a. fixed region
of pitch for eacb vowel, and 2. by certain relations among
the cavity tones;. It should be added that these distinctions
are not perceivable by the ear directly, but go to make up
the characteristics that distinguish unit-sounds from each ^
other. ^^/^
Turning now to the motor nature of vowels, we may distin-
guish between whispered, sonant, and surd vowels.
A whispered vowel implies 1. a contracted passage, gener-
ally at the glottis (p. 274), to produce a fricative noise ; and 2.
a cavity or series of cavities in front of this passage through
which the fricative noise passes. A slight degree of sonancy
often seems to be present (p. 275). The character of the
whisper noise varies greatly. . The cavities probably resonate
(p. 13) to periodic impulses picked out of the irregular
fluctuations in the rush of air.
A sonant vowel implies a vibration of the vocal cords and
a fairly constant open adjustment of the cavities. In cases
of extirpation of the larynx an artificial larynx may some-
times be inserted. A vibrating reed takes the place of the
vocal cords.
A surd vowel consists typically of a cavity or series of
cavities through which an unobstructed current of air passes.
There is none of the glottal friction which is present in the
whispered vowels. The cavities are aroused to resonance.
Surd vowels are very weak sounds.
We thus have as physiological definitions: whispered
vowel = laryngeal friction + faint cavity resonance ; sonant
vowel = cord vibration -|- cavity vibration; surd vowel =
breath -I- cavity resonance. These are typical forms; the
actual vowels often combine them in succession.
Speech is sometimes possible when the laryngeal passage is
entirely closed. The voice of Hickey^ in speech and in
1 Cohen, Trans. Phila. County Med. Soc, 1892 XIII 302 ; Trans. Coll. Phys.
Phila., 3d series, 1893 XV 131 ; Ein Fall von gut modutationsfahiger Stimme
u. s. w. Arch. f. Laryngol. u. Rhinol., 1894 I 276; Allen, Speech without a
larynx, Med. News, 1894 Mar. 17.
426 FACTORS OF SPEECH
song is audible at a distance of 12 meters ; it appears to be
a true voice and not a whisper, perhaps due to vibratory
movement of some edge within the cavity. Such a vowel
consists of edge vibration + cavity resonance.
In such cases where the laryngeal passage is entirely
obstructed a peculiar kind of whispered vowel may be pro-
duced by gathering air in the mouth or pharyngeal cavities
and emitting it. Such a vowel consists of mouth or pharyn-
geal friction -1- cavity resonance.
In cases with a closed larynx and an external tracheal
aperture a metallic reed may be inserted in the aperture
whereby a musical tone- may be produced while a vowel is
whispered by the mouth. Such an abnormal vowel would be
defined as reed vibration + mouth or pharyngeal friction +
cavity resonance.
The supposition that spoken and sung vowels consist of
whispered vowels + a cord tone is an absurdity. ' The con-
comitant resonances [mouth tones] which create or constitute
vowel qualitj' are animated, primarily and essentially, by the
irregular noises which issue, "together" with the vocal tone
from a speaking or singing glottis, but "without " it from a
whispering one. Some of these are always found capable of
affording just the appropriate impulse, and of kindling the
resonances of the configuration [mouth cavity].' ^ A whis-
pered vowel produced at the same time with a violin note
does not become a sung vowel by the addition. Moreover,
the addition of the cord tone necessarily produces a vibration
of the resonance cavity far stronger than any obtainable by
whispering — one that would utterly overpower a whisper ele-
ment. Finally, in ordinary speech there is no whisper action
added to the cord vibration; even a small whisper action
in the cords while vibrating produces a breathy tone (p. 273)
readily noticed by the ear as abnormal.
The different vowels have been usually defined according to
tlje relation between the maximum elevation of the tongue
1 Lloyd, Speech sounds: their nature and causation, Plioriet. Stud., 1890 III
277.
VOWELS 427
and the roof of the mouth + certain types of lip action + the
condition of the nasal opening. The place of the maximum
elevation of the tongue — its so-called ' articulation ' (p. 325)
— is named ' velar ' (' guttural ' ) or ' palatal, ' according as it
occurs in the region of the velum or the hard palate. The
size of the opening may be typified as ' high, mid, low '
(referring to the degree of elevation of the tongue).^ In
regard to another distinction,^ the division into ' narrow '
and ' wide ' (or ' primary '), there is much uncertainty and
dispute ; even the existence of such a difference is denied.^
The firmness of contraction of the tongue muscles is added
in the terms ' tense ' and ' lax.' The lip action is typified as
'neutral,' 'rounded' or 'spread.' Nasal modification is
indicated by ' nasality. '
The motor relationships of the vowels have been indi-
cated in various systems with more or less accuracy. These
systems are useful for various purposes, but have led to the
misconception of the maximum tongue movement as the
essential of the vowel, whereas the whole course of the ever-
changing movement must be considered (p. 325).
Sweet's system* of the vowels is given in the following
list; his phonetic characters (Romic) are enclosed in quota-
tion marks : unrounded, back: high, 'a,' Gaehc laogh; mid,
' a,' b«t, father ; low, ' v,' French pas ; mixed : high, ' i,'
Welsh wn; mid, 'e,' eye, better; low, 'a,' how, sir; front:
high, ' i,' hit, see ; mid, ' e,' men, say ; low ' aj,' care, man ;
BOUNDED, back: high, 'u,' pitt, too; mid, 'o,' boy, sow; low,
'0,' not, law; mixed: high, 'ii,' Norwegian hws ; mid, '6,'
fellow ; (no example for low) ; front : high, ' y,' Fr. htne ;
mid, ' a,' Fr. pew ; low, ' oe,,' Swedish for. It should be
noted that the English key-words refer to the sounds used
1 Sweet, Primer of Phonetics, § 35, Oxford, 1890. _
2 Bell, Visible Speech, 40, London, 1867 ; Science of Speech, 14, Washing-
ton 1897 ■' Sweet, Primer of Phonetics, 18, Oxford, 1890.
3 Evans On the Beli vowel system, Phonet. Stud., 1889 II 1, 113; Soames,
Introduction to Phonetics, § 96, London, 1899; references in Bretmakn, Die
phonetische Literatur, 41, Leipzig, 1897.
i Sweet, as before, 21.
428 FACTORS OF SPEECH
in the London, pronunciation ; some of them differ considei'-
ably from those commonly heard in America.
ViETOR 1 classifies the chief vowel types on mixed auditory
and motor principles : A. pure vowels : a. guttural vowels,
1. the u sounds, 2. the o (o) sounds, 3. the a sounds; b. pal-
atal vowels: I. unrounded, 1. the e (ae) sounds, 2. the i
sounds; II. rounded, 1. the oe sounds, 2. the y sounds; c.
guttural-palatal vowels ; B. nasal vowels.
The system of the Association Phonetique Internationale^
is the following:
Pa
LATAL
Vela
close
I Y
i li
UI u
u
half-close
e 0
e 0
V o
medium
3
€ ce
a D
A
J
half-open
open
A further development of this system has been proposed
as follows :
Mouth
Tongue
articulation:
PASSAGE : Lips :
PALATAL
VELAR
-y^l-luZu^ded
y
i
u
i
u
m
, ( rounded
"'"'^ [unrounded
V
I
u
UI
, „ , rounded
half close |uj,ro^„ded
0
e
6
0
V
i_ ij. rounded
half open ^ ^^^^^^^^^
8
0
n
D
A
open
se
0
very open
a
Q
All the vowel systems ' suffer from the defect that they
rest mostly on inaccurate observation and subjective estimates
— especially in the case of the forms of articulation. It is to
be hoped that . . . experimental phonetics will lead also to
*i ViETOK, Elemente d. Phonetik, 4, Aufl., Leipzig, 1898.
" Used in Le maitre phonetique (edited by Passy) and a number of books.
2 Ideophonic Texts (edited by Pierce), New York.
VOWELS 429
an exact vowel-system of cultured German, English and
French.' 1
The nasal vowels have been mentioned above (p. 339) ;
they have been discussed at length by Rotjsselot.^
Great differences are found in the ' attack,' or ' on-glide ' (I
would prefer the term ' entrance ') of a vowel. In German
an initial vowel regularly begins with the glottal catch (p.
278) ; the glottis is firmly closed, the cords are stretched to
nearly the pitch required for the vowel, and the vibration
begins suddenly with considerable amplitude. In American
speech an initial vowel begins regularly with small tension of
the cords and with small amplitude. • This appears clearly in
the curves for ai (Plates I and II), a (Plate VIII), an (Plate
X, line 5), and in the cases of ai discussed in Appendix II.
In general the American initial vowel begins with a low pitch
and rises more or less rapidly ; in exceptional cases it is con-
stant or falling, as in the interjection of satisfaction a (Plate
VIII), and in ai ' eye,' at the end of a phrase (Appendix
II). In connected speech an ' initial vowel ' means, of course,
a vowel at the beginning of a phrase after a pause. French
attacks resemble the American ones ; Hungarian attacks
resemble the German ones ; in German Swiss (St. Gall) they
seem to vary.^ The coup de glotte of singers is the German
attack ; for Americans it is an artificial action that must be
learned ; it sometimes produces nodules on the cords in
Americans, probably due to inaccurate and too energetic
action.
The exit of final vowels (in a phrase) seems among Amer-
icans to consist regularly in a fading of intensity often with a
fall in pitch. This is clearly seen in the final o of ' sparrow '
(Plate I), i in ai of ' I,' i in ai of ' my ' (Plate II), a of ' ha,'
e (a"^) of ' eh ' (Plate VII), a of ' ah ' (Plate VIII), a of ' ah '
{Plate XI). In French and Hungarian a single record of
1 ViETOK, Elemente d. Phonetik, 4. Aufl., 64, Leipzig, 1898.
2 RoussELOT, Principes de phou^ique expe'rimentale, S""' partie, 532, 582,
Paris, 1901.
3 RonssELOT, as before, 484.
430 FACTORS OF SPEECH
each 1 seems to indicate a somewhat less gentle exit for the
isolated vowel ce.
The term ' vowel ' is often used in contrast to ' diphthong.'
The supposition that a vowel is of constant character through-
out its duration is, however, quite erroneous, many of the
vowels being as thoroughly diphthongized as the usually
recognized diphthongs. The term 'vowel' is properly used
to indicate the class of open sounds. These sounds may be
of all degrees of constancy. A perfectly constant vowel may
be termed a monophthong, one with two clearly distinguish-
able parts a diphthong, one with three a triphthong, etc.
The diphthongal character of a vowel may be primarily
settled by its effect on the ear ; it may be due to changes in
the pitch of the cord tone or of any of the cavity tones, or to
changes in intensity.
' Diphthongization ' may be used to indicate a difference
between the beginning and the end of a sound. A sound
beginning like an a might change till it ended like an i;
such a changing vowel might be said to be diphthongized.
The change might not be distributed evenly throughout the
vowel; it might change at first slowly, then more rapidly,
and then again more slowly. A case might occur where
the sound beginning like an a did not pass evenly into i,
but changed more rapidly in its interior ; such a sound might
be considered as a 4- glide + i. This latter case is gener-
ally the one in mind in discussions of diphthongs. Again,
a vowel might undergo little change during most of its
length but a rapid change just as it ended. Thus we might
have a + glide. Finally, a rapid change might occur at the
start, after this the vowel being as constant as any vowel
ever is. Thus we might have glide -I- a.
These four types represent cases arbitrarily selected but of
all the possibilities within the extremes of 1. a perfectly
constant sound ; 2. a sound beginning in one way and ending
in another; 3. a sound with continuous change; 4. a sound
With an abrupt change.
1 RonssELOT, as before, 485.
VOWELS 431
If we add that the beginning may be any one of the infinite
number of possible vowels and the ending any other one,
and if we consider that the four factors of change are each
infinitely variable, it becomes evident that the number of pos-
sible ' diphthongs ' is limited only by the possibility or prac-
ticability of distinguishing among them.
Among the phenomena of diphthongization we may note
the following:
Absolutely constant vowels never occur in speech except
in the sense that their changes are unnoticed. When the
changes become distinctly perceptible, we have the ' on-glide, '
the interior change, and the ' off-glide ' in a vowel. A
development of the ' on-glide ' produces the rising diph-
thong; of the ' off -glide ' a falling diphthong; of the inte-
rior change a diphthong of two more or less nearly equal
elements.
Just what forms of diphthongization actually appear in a
language must be settled by experiment. Aside from Mar-
TBNS's study (p. 20) of au and ai, and my own work on ai,
I know of no experimental study of diphthongs. An account
of the latter condensed from a previously published mono-
graph 1 is given in Appendix II.
Curves of various vowels, diphthongs and triphthongs are
to be found in the Plates at the end of this volume.
References
For the history of vowel systems: Michaelis, Ueber d. Anordnung d.
Vokale, Archiv f. d. Stud. d. n. Spraohen u. Lit. (Herrig), LXV, LXVI ;
also separate Berlin, 1881. For summary and discussion of various
systems : Vietor, Elemente d. Phonetik, 4. Aufl., 39, Leipzig, 1898.
1 Scripture, Researches in experimental phonetics (first series). Stud. Yale
Psych. Lab., 1899 "VII 1.
CHAPTER XXIX
LIQUIDS AND CONSONANTS
' Consonant ' is a term loosely applied to sounds that are
not distinctly vowels. The sounds m, n, r\, r, 1 are often
classed as ' liquids ' or even as ' semi-vowels.'
"Vowels, liquids and consonants may be distinguished by
the degree of openness of the vocal cavities ; thus the three
sounds i in biova 'be over,' j in ju 'you' and j in lejn
Germ. ' legen ' have successively narrower passages above
the tongue. The diminution in the opening is accompanied
by increased fricative noise. On this principle m, n, ii are
liquids and j, 1 may be vowels, liquids or consonants. The
three typical degrees of j and 1 may be denoted by j„, j„ j^,
!») ^11 ^c-
■ The unfortunate classification of liquids solely as conso-
nants has led various writers on verse to speak of a syllable
containing a vowel followed by a liquid as short, whereas the
total vowel quantity is really long. In fact, in speech the
specific vowel may be omitted, leaving the liquid as a vowel
between two consonants.
The liquids are probably also to be distinguished from
related vowels and consonants by the fact that they undergo
greater changes.
The vowel i of bit ' bit ' undergoes presumably no more
change between its beginning and end than the other vowels
do; the liquid j of ]u 'you' undoubtedly passes rapidly
through a considerable range of change ; the consonant j of
lejn (Germ, 'legen') is again of a more constant character
^han the liquid. The character of the liquid ] appears in
' draw your ' (Plate I, next to last line) ; instead of the
LIQUIDS AND CONSONANTS 433
Strong vibrations found in all cases of i and in the neigh-
boring vowels 0 and u, it shows faint ones of a changing
character.
The sound w in wil ' will ' is the liquid form of u ; it is
sometimes called ' consonant u.' The final sound of ' bow '
in Plate I seems to end rapidly by some articulatory action
in the mouth rather than in the usual fashion for vowels
(p. 429) ; it is the consonant w rather than the vowel u; the
word is thus phonetically bow or bouw. This addition of w
is due to the fact that a vowel follows ; before a consonant
the word would presumably be bou or bo. The development
of such a hiatus-filler is a common linguistic phenomenon.
French q (' consonant y ') as in ' lui ' also seems to be a
liquid. Perhaps other vowels have liquid forms.
This view of the liquids as involving more movement than
the vowels I had arrived at in my own phonetic work. It is
strikingly confirmed by Roussblot's ^ simultaneous tracings
(p. 874) of lip movement, breath record and speech curve for
y and q, and of breath pressure and speech curve for i and j,
and by tongue and breath curves for ei, ej, ia and ]a.
The essential factors in consonants seem to be occlusion,
explosion, friction and roll. These are combined in the
most varied ways. Occlusion and explosion appear in
p, b, t, d, c, J, k, g, ' . Occlusion without explosion ap-
pears very often when the corresponding sounds occur in
certain combinations ; if separate letters are needed for these
non-explosive occlusives, they can perhaps be obtained by
breaking off unessential parts of the type, for example,
o, D, t, a, c, ;, K, g. The occlusives may have nasal ex-
plosions ^ instead of oral ones ; this may be indicated by
w-modifiers, thus p", b", etc. Friction appears as the most
essential factor in <t>, p, f, v, s, z, s, z, 6, S, 9, j, x^ Y. h. The
roll appears in the various forms of r. Fricative explosions
of the occlusives occur in various forms and degrees.
1 RonsSELOT, Principes de phonetique expe'rimentale, 2"'= partie, 404, 637,
Paris, 1901.
2 RousSELOT, as before, 527.
28
434 FACTORS OF SPEECH
The attack in consonants differs in different cases. For
pa and sa the attack seems to be more energetic in German
than in French ; ^ for ba it is less energetic in German be-
cause the explosion is weakened by the probably complete
closure of the glottis during the surd portion of the b
(Fig. 279).
We have now to consider the phenomenon known as
' mouillure.' Its general character has been outlined by
SiEVBES.^ It is produced by the adaptation of a consonant
or a group of consonants to the articulation of a palatal
sound (generally i or j), that is, by a rise of the forward part
of the dorsum of the tongue. Such a rise is an addition to
the articulation of a labial sound ; it is difficult for a velar,
a postpalatal, or a frontal one, but is easy for a prepalatal,
and unavoidable for an alveolar-prepalatal. The degree of
rise may vary. In some cases the mouillure appears to the
ear throughout the duration of the sound without any addi-
tion at the end, as in X, n ; in others it appears as a differ-
ence only at the end ; in others it modifies the sound and
appears also in the release, as in A-mouill^. A mouill^ sound
is not the same as the corresponding consonant followed by
] ; thus ^ is not the same as Ij, n as nj, etc. All the mouill^
sounds are characterized by the formation of a narrow passage
between the predorsal part of the tongue and the prepalatal
region.
The mechanism of the mouillure has been established by
the experiments of Lbnz ("Figs. 307 to 323).^
For aka, eke, iki Lenz obtained by means of an artificial
palate (p. 298) the contacts shown in Figs. 307, 309, 310. They
show for k successively more anterior contact and the forma-
tion of a channel under the influence of the adjacent vowels.
For k as in Fr. ' qui, ' Ital. ' chi, ' ' chiesa, ' the contact
was still more advanced and the channel diminished (Fig.
1 RoussELOT, as before, 487.
• 2 SiEVERS, Grundzuge d. Phonetik, 5. Aufl., 185, Leipzig, 1901.
8 Lenz, Zur Phi/siologie u. Geschchte d. Palatalen, Diss., Bonn, 1887; also in
Zt. f. vergl. Sprachf., 1888 XXIX 1.
LIQUIDS AND CONSONANTS
435
k in aka
Fig. 307.
k in aka
Fig. 308.
k in eke
Fig. 309.
k in iki
Fig. 310.
C
Fig. 312.
K
Fig. 314.
T
Fig. 319.
436 FACTORS OF SPEECH
311); the sound may be indicated by c. It is evident that
a slight advance of the tongue would close the entire pre-
palatal region. This occurs in the medio-prepalatal contact
shown in Fig. 313; the anterior untouched region is very
small. The direction of pressure is, as in the preceding
cases, mainly upward. The contact is released by drawing
the tongue down and back, that is, by starting with a small
groove in the prepalatal region and proceeding backward
until the release is complete. The breath presses through
with a relatively weak explosion and a following short rush;
the slower the movement the weaker the explosion and the
stronger the rush. This k is the so-called ^-mouill^; it may
be indicated by k. The fricative portion, when spoken as
an independent sound, may have its opening at the front or
at the rear of this prepalatal position ; in the former case the
sound resembles a <r (specifically a backward or dorsal -palatal
s), in the latter a | (specifically a forward 5). Owing to the
fact that the curvature of the tongue is greater than that of
the palate at this point, it results that the upward pressure
produces the firmest contact not in the middle of the pre-
palatal region but at its front or rear edge. The release of
contact would thus begin either at the rear or at the front of
the region of contact; in the one case the fricative addition
to the sound would be cr-like, in the other |-like. The
former seems to be the usual one as described above. When
the palatal cr is spoken alone, the region of contact is that
bounded by the white lines in Fig. 315 ; when it is followed
by K the contact covers the whole portion shaded in
Fig. 315; the region of contact is practically the same as for
isolated k (Fig. 313), though slightly greater, owing to the
fact that the previous position of the tongue for <r renders
only a slight movement necessary to complete the closure.
This shows clearly the nature of the release of k.
The alveolar-prepalatal contact (Fig. 316) marks a sudden
ohange from the previous series. The pressure of the tongue
is directed forward against the alveolae. The release of the
large contact occurs from the rear by formation of a small
LIQUIDS AND CONSONANTS 437
groove through which the air mshes after the weak explosion;
the sound resembles a combination of t and |, the t portion
being weaker as the opening is slower. The sound is the so-
called f-mouill^ ; we may indicate it by t. On account of the
convexity of the alveolar region its closure is firmer than that
of K. For the group §t the closure is limited to a small supra-
alveolar region (Fig. 317). The sound ras usually produced
has a contact region as shown in Fig. 318; the likeness
of Figs. 817 and 318 indicates that the release of t is like |.
The sounds of Fig. 316 and Fig. 318 are absolutely identical
to the ear. Fig. 318 shows the fricative opening already
partly formed. It also shows that a prepalatal closure is not
necessary for t; nevertheless the closure next forward from
K IS necessarily t, as the tongue cannot make the prepalatal
closure without the alveolar one.
Lenz's investigation shows that the passage of the surface
of the k contact forward through c and k prodi^ces a change
in the character of the explosion, whereby it receives a frica-
tive addition of ct or |. Such a change in the pronunciation
of a word has usually been erroneously considered as the
introduction of a parasitic ] (consonant i).
The sounds k and t are in respect to movement neither
pure k sounds nor pure t sounds, but are somewhere between
them. We must now consider: 1. the nature of the differ-
ences between the group k, c, k (indicated by K) and the
group t, T (indicated by T) ; 2. the difference of k from k
and c, and of t from t; 3. the mechanism of the fricative
addition; 4. the essential characteristics of k and t.
The chief difference to the ear between a K and a T cannot
lie in the tone of the rear mouth cavity; as Lenz points
out, this would produce a change in pitch ; a steady fall in
pitch is observed in the series k, c, k without their becoming
the same as T. The direction of the expiratory rush of air
cannot make the difference, because a T can be formed by the
tongue-point against the same portion of the palate as is used
in K without sounding the same as K. A frontal-postpalatal
T may have the same pitch as a dorsal-postpalatal K, but they
438 FACTORS OF SPEECH
are quite different in sound ; even dorsal-alveolar T does not
resemble a K, while dorsal-alveolar and frontal-alveolar T are
hardly distinguishable. Physiologically the difference be-
tween T and K seems to lie in the use of the front and pre-
dorsal portions of the tongue for T and that of the medio- and
postdorsal portions for the K, with the result of a different
action in the two cases. As Lbnz points out, the anterior
portion of the tongue is much more movable than the posterior
portion. The mediodorsal portion is hindered in its move-
ment, especially along its middle line, by the frenum,
while the front portion has (ordinarily) free movement.
When the mediodorsal portion is pressed against the palate,
the chief pressure is at the sides. In all K-closures the con-
tact is looser in the middle, and the release naturally occurs
first along the middle line. The K-explosions have thus all a
scratchy character, due to the gradualness of the release and
the consequent rush of air over the middle before the tongue
is fully away from the palate. When the anterior portion of
the tongue is closed against the palate, the contact is firm at
all points and the release may occur evenly ; when this hap-
pens quickly the explosion is sudden and there is no noticeable
rush sound. In certain contacts the gradual release is un-
avoidable, as explained for k. The gradual release for t is
not a necessity of the contact but is derived from an addi-
tional muscular action. Just behind the region of closure the
tongue makes contact along the sides, tending to form a chan-
nel in the middle ; the release is naturally fricative.
The relation between T and K may be illustrated by that
between the bilabial explosive p and the labio-dental explo-
sive p formed by the lip against the teeth. The sound of p is
a clear explosive, owing to the sudden relaxation of the lips,
while that of p is a scratchy explosive, owing to the slower
completion of the explosion due to the irregularities of the
teeth.
« The difference between the explosions of the K and T
sounds lies in their dullness and sharpness. The k and t
differ from the others in having fricative additions. The
LIQUIDS AND CONSONANTS 439
groove is in both formed in the prepalatal region ; this occurs
in K in the chief contact surface, in t just behind this surface.
The sounds k and t are each produced by a single con-
tinuous movement of the tongue; they may be said to be
simple speech movements. To the ear they may appear as
simple as the explosives if the fricative addition is weak.
Increase in the fricative addition makes the consonant diph-
thongs Kcr and t|. Still further development into independ-
ent sounds requires change of contact and produces eg or
kx and ts.
By raising the lower jaw and lifting the predorsal portion
of the tongue against the alveolar and prepalatal regions
(Figs. 320, 321) a t^ can be produced with a somewhat (r-like
fricative addition. (The numerals j, ^, g, ^ indicate succes-
sive degrees of backward contact.) Still further progres-
sion of the region of contact may bring about the clear
dorsal-alveolar tg. Thus the progression of the contact sur-
face forward gives k-»c^K->T-*t. Such a development
is relatively seldom in the history of sounds, the most usual
occurrence being an independent development of the fricative
addition to t by a more gradual release of the contact; this
produces a K-like movement in place of the T-movement,
whereby the predorsal-alveolar contact is first opened only
along the middle line, sq that instead of tg we have the
combination tgSg. The region of contact for tgSg is shown in
Fig. 322.
Further progress of the closure forward and downward to
the teeth occurs by raising the jaw and lowering the apex of
the tongue through a supradental t^ till finally a bidental t^
is reached. The gradual release of the supradental t^ pro-
duces a s addition, that of tj a 6 addition; the compound
sound becomes tgS or tj6.
The unified sound c, corresponding to the compound sound
ts, differs, according to Lbnz, from t in having the medio-
dorsal portion of the tongue less raised, whence it results
that the gradual release of contact covers a larger space,
producing a fricative release that passes rapidly from | to <r
440 FACTORS OF SPEECH
and (T, with cr usually the most prominent element. A c sound
does occur in which the elements | and o- are more prominent
than cr. It is quite erroneous according to Lbnz, to consider
c = t + s (or more accurately t + d") as such an analysis
separates the portions artificially. It must be said, however,
that the degree of fusion in many Yowel diphthongs is just
as complete as in c, and that a notation such as ai cannot
usually be considered to indicate distinctly separated sounds
(p. 430). The contact surface of c is shown in Fig. 323;
the white line incloses the surface for & alone. A com-
parison of Figs. 318, 320, 323 shows that there is a closer
similarity between t^ and c than between t and tgSg.
As a development of t the c has an advantage over tgSg
in being spoken with nearly the same portion of the tongue,
that is, mediodorsal, while tgSg uses the predorsal portion.
Whether a dialect develops t to 6 or to tgSg depends upon
circumstances; only one of the developments occurs at the
same time. Lenz suggests that the choice depends on the
preference for similarity of the acoustic impressions or for
that of the sensations of movement. If the former prevails,
T -^ tgSg ; if the latter, t -> c.
The so-called ' mouill^ ' explosives k, \, t, 8 are com-
posed of an occlusion with a fricative release; they thus
differ from the explosives, properly so-called, which have an
occlusion with an explosive release. With nasal openings
there arise ji and n, the latter being the w-mouill^ of the
Romance and Slavic languages, and the former a 77-mouill^.
A relaxation of the articulation for t, 6 by a side contraction
gives surd and sonant A, or Z-mouill^.
The audibility due to the explosion as in t, d, c, j, k, g,
etc., is lacking in m, n and j\ on account of the passage of the
air through the nose, and in 1 through the side openings; the
fricative form of release, which is distinctly audible in t, 8
and K, \, is scarcely heard in ji, ii and A.
Roussblot's palatograms ^ of k, \, t, 8 for various French
dialects are closely analogous to those of Lenz.
1 RoussELOT, Principes de phonc'tique expe'rimentale, 2'"e partie, 607, Paris,
1901.
LIQUIDS AND CONSONANTS 441
The mechanism of the mouillure is illustrated by Oussof's
palatograms ^ of the Russian mouill^ labials. They all show
that during the articulation for p, b, f, v the tongue jjlaces
itself in the j position ; the release is thus j -modified.
Experiments with exploratory bulbs (p. 333) show^ that
the r-mouill6 differs from the ordinary r by a considerably
greater rise of the tongue in the dorsal region, while the
point of the tongue rises slowly instead of abruptly. The
w- and Z-mouill^ seem^ to be produced by more extensive
tongue contacts. It is to be remembered that these two
sounds may be continuously produced.
These explanations of the so-called ' mouill^ ' sounds may be
completed in some particulars. The term ' mouiller ' means
' to make wet,' 'to give a liquid sound to.' It refers primarily
to the auditory impression. Such an impression of softness
can be produced in occlusives by retarding the explosive re-
lease ; this occurs naturally in those whose articulations are
dorsal-prepalatal ; it may be produced by retarding the tongue
in the release of other articulations. The fact that forward
K and backward T are naturally ' mouill^ ' sounds has led to
the identification of the motor-phenomenon vfith. the auditory
one. A sound like 1 or n, produced with a more backward
region of articulation, is said to be ' mouilM.' The nature of
the release in these sounds is, in my opinion, a minor matter ;
the continuous ' I- and w-mouiU6 ' are rather to be considered
as members of the 1 and n series that to the ear differ merely
in their cavity tones ; a connection with the occlusive mouill^-
sounds arises only from the auditory impression of softness.
Sounds like c and J, s and z, etc., Rousselot calls* ' semi-
occlusives.' He asserts that they have developed from the
mouilld consonants by further relaxation of the muscles con-
stituting the ' vocal hindrance.' If the point of the tongue
remains in the alveolar region with the dentals t, 8, or if it is
placed there with the palatals k, x, the release is modified, and
the mouill^ sound becomes transformed into an articulation
1 Rousselot, as before, 604. ^ RonssELOT, as before, 607.
3 Rousselot, as before, 610. * Rousselot, as before, 618.
442 FACTORS OF SPEECH
that appears double to the ears of those who cannot produce
it. Persons who naturally employ the semi-occlusives cannot
tolerate ^ the confusion with the sound groups ts, dz, ts, dz.
Eousselot's palatograms and records with exploratory bulbs
at different points in the mouth show conclusively that the
region of occlusion is much further back for c, J, s, z, than
for t, d (being rather that for t, 8) and that the occlusion is
weaker. The relative proportions of occlusion and friction
vary from complete occlusion in t, d, to complete friction in
O^ Zf O) Zm
Experimental records are still needed to show just where
the separation is to be made between c, j, s, z, and the cor-
responding diphthongs in each language. For Italian c, j
JossELYN has shown (p. 321) that the fricative element is
like ] and not like s, z. He seems to have proven that the
sounds in c, j are too closely unified to be treated as diph-
thongs ; perhaps also we should say the same of the sounds
represented here by s, z but on p. 321 by ts, dz.
Josselyn's^ experimental records of Italian articulations
show varieties of ' soft c ' extending from a purely fricative form
to nearly a k (^-mouill^). Indications of k -> k were even found
in some cases. The change of Latin k before e or i to k prob-
ably occurred in the same manner as the similiar change occur-
ring at present in Parisian French (p. 315). The further
change of k to c is illustrated by Josselyn's records.
Physical definitions of the consonants have been given
by Heemann ^ on the basis of his speech curves (p. 43).
A. Consonants with cord action (^phonic consonants).
1. Smooth semi-vowels: 1, m, n, t]. These are sounds,
without noises, having one or more fixed characteristic reso-
nance tones (formants) like the vowels ; they differ from them
in not having their force, openness and distinctly musical
character.
1 Dauzat, Contributions a I'gtude des articulations consonanttques, La Parole,
1899 1619.
• ^ JossELTN, jStnde sur la phonetique italienne, 67, These, Paris, 1900.
3 Hermann, Fortrjesetzte Untersuchimgen ii. d. Konsonanten, Arch. f. d. ges.
Physiol. (Pfluger), 1900 LXXXIII 8.
LIQUIDS AND CONSONANTS
443
2. Remittent semi- vowels : r. The various forms of r
have the same properties as the smooth semi-vowels, and the
additional one of relatively slow periodic changes in intensity.
3. Phonic noise-continuants: p, v, g, z, z, j, -y. These con-
sist of a noise accompanied by a cord tone, the characteristic
tones (formants) in the noise having apparently no relations
to the cord tone.
4. Phonic explosives: b, d, j, g.
B. Consonants without cord action {aphonic consonants).
1. Aphonic noise-continuants: <j>, f, 6, s, s, 9, x- These are
noises of very variable nature, containing certain characteristic
tones.
2. Aphonic explosives: p, t, c, k. These have a silent
period of occlusion, usually with a following explosion.
Brtjcke's motor definitions ^ may be summarized in the fol-
lowing classification. Where two letters are given, the for-
mer indicates the surd, the latter the sonant. The laryngeal
sounds (h, Arabic hha, Arabic ain, laryngeal r) are placed in a
separate class.
Simple Consonants
1st Series {lip)
2d Series {front
of tongue)
Sd Series (middle and
back of tongue)
Explosives
, ( pi, Ji bilabial
P' " \ p\ 62 labiodental
f t\ rfi alveolar
, , it^, d^ cerebral
*• " "i t\ d^ dorsal
«*, d* dental
k,g
( fci, g^ velar
\ W, g"^ palato-velar
{ B, g^ palatal
Fricatives
{ /■!, wi bilabial
/, ''^/2,y2 labiodental
■ si, zi alveolar
J s2, z2 cerebral
°'' ^ "< sS, z3 dorsal
,s*,z* dental (9,
x,y
5)
i X^> y^ velar
s X^t 'iP' palato-velar
{ X3, y3 palatal
L-sounds
\i, /I alveolar
J , , A.2, p cerebral
'*■' ' 1 \3, /3 dorsal
A.4, li dental
Vibrants
Resonants
( ml bilabial
™ 1 irfl labiodental
' rji alveolar
I n^ cerebral
" ] nS dorsal
)!* dental
ttI velar
•ir2 palato-velar
ifi palatal
1 Bkucke, Grundzuge d. Physiologie u. Systematik d. Sprachlaute fiir Lin-
^uisten u. Taubstummenlehrer, 1. Aufl., Wien, 1856; 2. Anfl., Wien, 1876.
444 FACTORS OF SPEECH
Complex Consonants
sx, ^>J in various combinations of the forms of s, x, ^t V-
MOUILLE SOUNDB
ly, nij with various forms of I and n.
Beucke's treatment was a great achievement for the pre-
experimental time. Later work has partially remedied its
incompleteness and corrected its faults. s%, zy (= s, z) are
no longer supposed to be complex sounds. The mouillure
has been extended to other consonants than I and n and is
known to be different from the addition of a parasitic y (j^.
p and 6, t and d, are known to differ in other ways than in
regard to sonancy. We can no longer accept Beucke's
view that for the consonants as for the vowels — with the
exception of the diphthongs — ■ the letters are not to be
considered as signs for active movements of the organs of
speech but as indications of definite conditions and definite
adjustments of the mouth organs and glottis, in which they
are found while the expiratory muscles seek to press out
the air.
Vietoe's classification is also motor. ^ To show the rela-
tions between the groups of sounds, I give his complete outline.
Laryngeal Articulation. I. Sounds with laryngeal opening.
II. Sounds with laryngeal narrowing or closure. Mouth Ar-
ticulation: I. Sounds with opening of mouth : 1. sonants :
vowels (p. 428) ; 2. surds : h. II. Sounds with narrowing
or closure of the mouth: 1. fricatives: A. gutturals and
palatals : (1) uvular r; (2) gutturals: j and o; (3) palatals:
j and f ; B. dentals: (1) sibilant sounds: z and S, z and
s; (2) 3 and p; (3) liquids: r and I; C. labials: v and /;
2. occlusives : A. without nasal resonance : a. gutturals and
palatals: g, k; b. dentals: d, t; c. labials: 5, p; B. with
nasal resonance : a. gutturals and palatals: »y, w; b. dentals:
n; c. labials: m. Vietoe's j corresponds to y of this book,
c to x» J* to 9, dotted n to n.
1 ViETOR, Elemente d. Phonetik, 4 Anil., Leipzig, 1898.
LIQUIDS AND CONSONANTS
445
The Association Phonetique Internationale classifies and
represents tlie consonants in tlie following way. (The mis-
leading " g " has recently been replaced by a crossed g.)
Laryn-
Gdt-
Uvu-
Velar
Palatal
Lingual
Labial
geal
TURAL
lar
Plosive
?
qo
k9
cj
td
pb
Nasal
1
J
u
m
liateral
I
l
1
Rolled
Q
aK
r
Fricative
h
hR
HB
(WAl)xg
(q)?j
•I, 63,73 ,
zs
£v ru
wAi a
The steady progress in our knowledge of the consonants
has lately been aided by experimental researches. The
details, as far as summarized in this book, can be found by
consulting the Index.
CHAPTER XXX
SOUND FUSION
The person producing a vocal sound is not distinctly con-
scious of the separate muscular movements involved. He
has a more or less definite idea — derived from past auditory
and motor experiences — of what he wishes to say, and he wills
to do it. The auditory and motor experiences from the past
and those of the actual production of the sound are, in ordi-
nary speaking, fused in his mind into a single experience. It
is only by attending to some of these groups of elements more
than to others — that is, by introducing artificial conditions
— that they can be made to appear specially prominent to
him.
This fusion occurs not only among the elements at the
same instant of time, but extends over intervals of time.
Speech actions are fused into ' experiences ' that may cover a
whole discourse, a paragraph, a sentence, a phrase, a word
or a phonetic element.
To the ear speech is one continuous flow ; even the pauses
are just as effective mental elements as the sounds; an
attempt to pick out elements of speech by the ear modifies
and alters them from the sounds actually occurring. On the
motor side the fusion is just as complete ; there are no dis-
tinctly marked volitions for successive sounds, but a course
of volition resulting in a course of movement.
Speech cannot be considered as made up of separate ele-
ments placed side by side like letters. In the flow of speech
it is just as arbitrary a matter to consider certain portions to
be separate sounds as to mark off by a line where a hill
begins and the plain ends. Moreover, an assignment of char-
SOUND FUSION 447
acteristic portions that would be allowable for disconnected
sounds or for some occasions would not be allowable for others.
Thus, an independent surd occlusive (p, for example) can
not be distinctly made without the movement of closing and
the movement of explosion. The closing may be faintly
audible, the time of complete occlusion is one of silence, the
explosion is audible. In a word^ however, not only may
the time, movement and sound of the closing be fused with
the preceding vowel, ^ but even the period of occlusion
itself is to be considered — Gk^goieb^ points out — as the
final portion of the preceding syllable, while the explosion
begins the following one. In ordinary speech a division
into syllables of the French word papa would not be pa-pa
but pa I *-<''a where the 1 1" indicates the closing and occlu-
sion of the p, and <p its explosion. Even in such a case, we
must add, there is no sharp boundary between the portions
of the p, owing to the varying relations of larynx and mouth
action.
Having set aside the view of speech as an agglomeration
of elements, we must attempt a consistent treatment on some
other principle. >
In all speech there is constant variation in the quantity of
auditory and vocal energy from moment to moment. This
quantity consists of the sum of all deviations from the medium
conditions of pitch, intensity, duration and difficulty of enun-
ciation. A rise or fall of pitch beyond the general tone of
the discourse, a lengthening of a sound, an increase of inten-
sity, a change to silence, and any increase in the difficulty
of fixing the mouth to produce sounds — these are all elements
that tend to make a sound more energetic. These elements
in their varying degrees and combinations produce a total of
energy that varies at each, instant. The effect is directly
related to the rapidity of change from the medium condition.
The variations in energy may be utterly irregular, as in some
1 ViETOR, Elemente d. Phonetik, 4. Aufl., 297, Leipzig, 1898.
2 Gkbgoire, Variations de dure'e de la syllabe frangaise, La Parole, 1899 I 161,
263, 418,
448 FACTORS OF SPEECH
cases of prose, or may show a great degree of regularity, as
in verse.
We may try to treat speech as a phenomenon whose
energy varies with the time. Such a curve of auditory
impressiveness, or of vocal effort, might, for some phrase, be
the imaginary one given in Fig. 324. The curve of energy
rises and falls in a more or less regular manner; limits
between sounds must be more or less arbitrary. Limits be-
tween syllables must be also very arbitrary; points of lowest
energy might be assigned as limits, but there may be two
neighboring points nearly alike. Points of maximum energy
are frequently assigned as the most important parts of
syllables ; but in the case of two neighboring maxima a slight
A\odtLal5etic9j[*s
'j'
Fig. 324.
increase of one over the other would hardly justify the
utter neglect of the latter. The difficulties and the solutions
for the curve of energy are the same as for all empirical
curves of like nature. The points of maxima, minima and
flexion all have their values, but detailed treatment is be-
yond practicability. Just as in the case of an irregular solid
body, we are driven to pick out points at which we can
consider the whole mass to be located without altering the
result under discussion. This is the centroid theory of the
auditory and motor nature of speech that corresponds to
the centroid theory of the course of thought. Thus, for
a certain purpose, such as beating time, we can consider
the whole flow of speech energj^ in Fig. 324 to be concen-
trated into three material points whose positions are so de-
termined according to the actual distribution of the energy
that for the purpose in view the result is exactly the same ;
SOUND FUSION 449
tlie points might occupy the positions indicated by the
dots.
The curves of volition-energy, of muscular work, of vibra-
tion-energy in the air, of auditory impressiveness (sonority),
etc. for a given portion of will differ from each other,'
although closely related. , No experimental determinations
I of any of these curves have yet been made, although some
' experiments indicate the possibility of determining the audi-
; tory and motor curves of total energy with fair accuracy.
Other experiments in connection with the rhythmic effect
render it possible to locate with accuracy the position of the
rhythm centroids (Ch. XXXV).
A full treatment of the flow of speech would take into
account various factors of the total energy, such as energy of
sensation, energy of emotional effect, energy of association,
! etc. For three such factors considered in combination the
centroids would be like those of an irregular mass passing by
a given point, or, to illustrate more concretely, of such a
mass of iron passing by a coil of wire connected to a galva-
nometer whose needle indicates the effect^
It is perhaps necessary to say a few words concerning the
concept of the syllable. The assignment of the limits of
syl-lables as the moments \where during continuous expira-
tion there is a passage through a sound of less sound quan-
tity ' shows some hints of the conception of speech as a flow
of auditory and motor energy, but misses the essential point.
For example, the word ' manly' is said to contain two sylla-
hles ; it can be spoken with two breath-puffs, but nobody ever
does so in speech ; its curve of energy may show various max-
ima and minima, but it can hardly be said that a minimum
occurs between n and 1. The word ' manly ' represents con-
tinuous action of the breath organs, continuous action of the
vocal cords with a smooth rise and fall of pitch, continuous
movements of the lips, tongue and velum through various posi-
tions. In fact the word is a continuous sound-change with
no limits or minima of any noticeable kind. The word is
indicated by letters as a sum of separate things, but the ex-
29
460 FACTORS OF SPEECH
perimental results in similar words show that this is not the
case. The word is to be considered as a fusion of a series of
continuous changes, certain stages of which may be charac-
terized as m, ae", se, ae", n, 1, i. As far as the vocal move-
ments are concerned, the word is just as continuous as
' manned.'
According to Meyer ^ the answer to the question of how
many syllables a word has is given by our feeling concerning
the number of syllables into which it can be made to fall,
not the number into which it actually does fall in a given
case. Meyer's thought that a word is monosyllabic or
polysyllabic according to whether it can be spoken in the
rhythm of speech with only one impulse or with more,
seems to have some relation to the centroid view of the
flow of speech.
I do not believe, however, that a division of the flow of
speech into separate blocks (termed ' syllables ') has the
slightest justification or the slightest phonetic meaning. As
long as speech is indicated by letters, it is easy to divide the
letters into groups, but, as has been pointed out, speech
really does not consist of any such elementary blocks of sound
as the letters are supposed to indicate, or of any such large
groups as syllables. The attempt to divide speech into
syllables might be compared to a division of a landscape into
hill-blocks and valley-blocks ; there can be no dividing lines
drawn. The terms ' hill ' and ' valley ' apply to character-
istics of the land and not to separate divisions.
According to Sibvers ^ the ear divides the flow of speech
into certain portions, that is, into masses of sound that form
relatively close unities, which we call syllables; the division
rests upon discontinuities (minima) in the strength of the
flow of sound. If instead of confining the first statement to
the ear we make it read the ■ ear, the vocal impulses, and
all their associations ' etc., and if we add to the second that
«
1 Meter, Die Silbe, Neuere Sprachen, 1898 VI 479.
2 SiEVEKS, Grundziiged. Phonetik, 5. Aufl., 198, Leipzig, 1901.
SOUND FUSION 451
the division rests upon all factors that tend toward the effect
of a series of unities in the flow of speech, we have prac-
tically the centroid theory that I have proposed. A word
would be said to have as many syllables as it was felt to have
centroids. A syllable would then be defined as a portion of
speech within which a centroid is located, the boundaries
of syllables being of all degrees of indefiniteness.
Just as in the case of any irregular body, we can find
different grades of centroids by limiting the consideration to
larger or smaller portions. There are thus phrase-centroids,
syllable -centroids, sound-centroids, etc.
(Systematic treatments of the syllable are to be found in\
the works of Sievees, Vietok, etc. ; so little experimental I
work has yet been done that a discussion is hardly in place/
here. To what has been said I will merely add that the cen-
troid treatment of speech differs from what may be called the
' maxima-minima, ' or ' apex-depression, ' treatment mainly
in considering the whole mass of speech — physically, physi-
ologically or psychologically, as may be desired — in locating
the critical moment, instead of locating it at the moment of
greatest energy. The centroid will rarely coincide with the
maximum of energy.
The continuity of speech action is strongly impressed on
any one who attempts to study the curves of connected
speech (Part I). There is often no possibility of assigning
boundaries to phonetic sounds, words or phrases. Thus, in
the record for mailitlai ' my little eye ' in Cock Rohin (p.
126) the speech curve is continuous from the beginning of m
to the end of the last i ; at various places along the record
typical m, a, i, 1, vibrations are to be found, but the change
from one type to the next is a gradual one; this condition
shows itself through the whole record. It cannot be said
that the record is fully represented by m,a,i,l,i,t,l^a,i,, where ^
indicates a glide, unless each of the symbols is understood to
represent a typical stage of a continuous process.
Speech records of all kinds seem to indicate the applicabil-
ity of the concept of the centroid, or center of unification, of
452 FACTORS OF SPEECH
speech action. For example, the speaker considers something
he wishes to say and executes a complication of movements
which we indicate by ' my little eye.' That the idea he
wished to express required a certain complication ' my little
eye ' and not ' mit meinem kleinen Auge ' or ba or a was
due solely to his past teaching ; for him there would have been
no essential difference between ' my little eye ' and a except
in the amount of labor involved. For him and for all pur-
poses of speech this complication may be treated as having its
whole effect at some point of time; in ' my little eye ' it
might perhaps lie at the moment of the closure of the t.
Such a centroid of speech movement is the point in the
flow of speech at which the total speech movement may be
considered to have occurred without changing the result.
Centroids may be found for portions of speech. If it is
found practicable to^ cut off certain subordinate portions with-
out introducing too much error into the results, the centroids
may be found for such portions. Thus, a portion may be
taken from the speech record of ' my little eye ' that may he
considered to cover the ' i ' ; the limits must be arbitrarily
assigned, but if, for the purposes of the investigation, this
does not introduce too much error, then the i can be consid-
ered as definitely limited. This i will be found to be a phe-
nomenon changing in its nature from beginning to end. For
many purposes its acoustic effect and its motor production
may be considered to be located entirely at a definite point of
time, or at a definite centroid. Similar centroids may be
found for other portions of the phrase.
The number and character of the centroids is really what is
indicated by a phonetic spelling. Thus ' mailitlai ' indicates
that /or the purpose in view — instruction in pronunciation, or
discussion of phonetic change, or indication of parts of a record,
etc., etc. — the entire speech quantity of ' my little eye ' on
the particular occasion discussed may without appreciable
error be considered as occurring at nine different points of time
and as consisting of nine speech elements whose characters are
indicated by m, a, i, 1, i, t, 1, a, i. That a phonetic notation
SOUND FUSION 453
or an ordinary spelling indicates a construction of speech out
of separate fixed elements, like a house out of bricks, is a
notion that. has arisen from a study of spellings instead of
speech itself. A person speaking from printed characters or
indicating what he says by written ones has no such notion ;
to him the letters singly or in groups are schematic signs,
or ideograms (p. 128), that roughly indicate a desired result.
This view — which is a necessary result of a considera-
tion of the general phenomena of voluntary action — seems to
have been what Paul had in mind. ' An actual separation
of a word into its elements is not only very difficult, it is
actually impossible. A word is not a placing side by side of
a definite number of independent sounds, each of which can
be expressed by a letter of the alphabet, but is nearly always
a continuous series of an infinite number of sounds, and the
letters indicate, in an incomplete fashion, nothing more than
certain characteristic points of this series. ' ^
From the linguistic side it has been pointed out^ that
speech movements occur as unities; that there may be a unit
volition for the movements required for a certain speech
sound and also unit volitions for complex groups of move-
ments such as are required for words [and phrases] ; that the
unit volitions for the complex groups have an independent
existence and are not the sums of volitions for the elements.
The unification of volitions is, according to Kaesten, still
more elaborate ; the suffixes, prefixes and other parts of words
become merely parts of a unit. Such volitional units become
associated with corresponding auditory units. It is interest-
ing to note that recent work on the cerebral cortex (p. 86)
indicates separate anatomical centres for the various group
movements.
Certain relations of similarity become established among
sounds used in different word-units in the speech of a person,
a community, a period, etc. The properties that remain
1 Patti,, Principien d. Sprachgeschichte, 3. Aufl., 48, Halle a/S., 1898.
' Kaksten, Sprecheinheiten u. deren Rolle in Lautwandel u. Laittgesetz, Trans.
Mod. Lang. Assoc. Amer., 1887 III 186 ; also in Phonetische Stud., 1890 III 1.
454 FACTORS OF SPEECH
approximately unchanged in different units may be called
the stable ones. The average properties that appear in the
various word-units may be said to constitute the typical
sounds. The properties of an independent sound generally
differ from those of a typical sound in connected speech.
The sounds used in actually spoken words may be called
specifio ones.
The relations between the typical sounds and the spe-
cific ones depend on sensory and motor laws.
The first topic now to be considered will be the differences
between typical sounds and the specific sounds in connected
speech. I shall attempt nothing more than to point out
some of the more important facts to be looked for in experi-
mental records. (Fuller and more systematic treatments of
speech fusion are to be found in the works mentioned in the
References at the end of this chapter.
The articulations of a sound are undoubtedly very much
modified by the neighboring sounds in a phrase, by the idea
expressed in the sentence, by the emotional condition of the
speaker, etc. According to Sievees ^ phonetic investigation
should begin with the sentence and proceed to the subdivi-
sions, the artificially separated sounds being only abstrac-
tions. / The physiological methods of registration, however,
have not yet been sufficiently developed for long records,
although the analysis by means of speech curves proceeds in
just this way (p. 61).)
One result of fusion is the tendency to ease of innerva-
tion, motor impulses being spared whenever possible. One
form of this smoothing is found in the general law that con-
tiguous articulations maybe modified to produce easy passage
from one to the other. This adaptation is, of course, in-
stinctive, or ' unconscious ' in the usual meaning of the
word (p. 381).
An effect of this unification is seen in the neglect of differ-
ences in speech and in the approximation to easj' curves
of motion. This may go so far that long and complex move-
1 Sievees, Grundziige d. Phonetik, 5. Aufl., 8, Leipzig, 1901.
SOUND FUSION 455
ments are reduced to short, simple ones. The process repre-
sents economy in physical movement and in the number of
changes of innervation. Examples are furnished by every
spoken word; experimental results may be found through-
out the preceding chapters.
This adaptation shows itself in avoiding changes. One of
them is the omission of change in respect to sonancy. Surds
may become sonants between two vowels, as in soh.im
' saw him ' (pp. 24, 276) and in the numerous examples
given on pages 360 to 362. Sonants may likewise become
surds where a retention of sonancy would hinder unification
(p. 361).
Illustrations of the smoothing off of tongue movements
may be seen in ' *anakts ' -> aj/af, ' *nokts ' ^ Lat. 'nox,'
Lat. ' octo ' -* Ital. ' otto. ' This phenomenon is constant
in the ' smoothing ' of the two elements of a diphthong, as
ai ->^ a in O. Eng. ' stan, ' ei ->• e in Swed. and Dan. ' sten. ' ^
A smoothing of the tongue action also changes intervocalic
explosives to continuants as in Lat. ' paganum ' -^ Fr. ' paien '
(see also p. 372).
Velar smoothing may be assumed in the nasalization of
consonants and vowels when adjacent to nasals. Many
experimental illustrations have been given on pages 345 to
351, and 359 to 365. ,
Lip smoothing may be illustrated by Lat. ' obstinatum ' -*
Ital. 'ostinato. ' It appears also in the communication of
lip action to preceding consonants by the Russian rounded
vowels 2 (see also p. 364). Smoothing of the tongue-lip
combination appears in ' *entfangen ' -> Germ. ' empfangen. '
General smoothing off of all vocal innervations and move-
ments may be said to occur constantly. It finally goes so far
that various elements disappear altogether, as in gmoin
' guten Morgen ' of the Germans, sivple or siple ' s'il vous
plait ' of the French, daete,3 for *daed3te 3 ' Dada take him '
in a case of child speech. Numerous examples are given by
1 Sweet, History of English Sounds, 22, Oxford, 1888.
^ Sweet, as before, 38.
456 FACTORS OF SPEECH
Passy:' tsepa ' je ne sais pas,' keseksa ' qu'est-ce que
c'est que ga?', sa"temnami ' je suis enchants, mon ami,'
etc. Experimental data concerning such extreme fusion and
condensation are not yet numerous, owing to the fact that
most records have been made from careful speech and not
from rapid conversation. The latest form of zonophone
(or gramophone) apparatus (p. 53) is able to record con-
versation without the knowledge of the person speaking;
it may be useful for this problem.
Curiously enough the effects of these condensations often
remain unnoticed. Unconscious movements are present in
speech as the remains of sounds that have been consciously
made in the past history of the language but have disap-
peared. RotJSSELOT's^ tracings for a case of the Lorraine
dialect show a difference between ap ' arbre ' and simple ap ;
the larynx ceased to vibrate for a before the lips formed the
p in the former case and not in the latter, the historical r
being represented by a condition of silence, or unrecorded
movement (see also p. 365).
Connected with and yet often opposed to the process of
condensation, is the process of distinction which also is a
result of fusion. To mark off unities the mind requires dis-
tinctions; the subordination of some elements implies the
elevation of others. Illustrations of this fundamental law
of mental action will probably be found in future, experi-
mental work. The lengthening of initial consonants for the
purpose of accent, the use of the glottal catch before strong
initial vowels, the retention of surd fricatives in tonic syl-
lables, may occur for this purpose. It is perhaps an instinc-
tive desire for more force in a word that leads to the surdation
of intervocalic sonants, as in O. E. ' etan ' from the same root
as Lat. ' edere, ' or to the pronunciation ekzsekt instead of
egzaekt ' exact.'
In general it may be said that the less the importance of
' Passy, Changements phonetiques, 123, Thfese, Paris, 1891.
' RonssELOT, Les modifications phonetiques du langage, 143, Rev. des pat. gallo-
romans, 1891 IV, V; also separate.
SOUND FUSION 457
a speech element in reference to its function in expressing
an idea, the less the amount of volition-energy given to it.
Decrease in volition-energy is accompanied by decreased
energy of movement, by decreased time given to it, by de-
creased accuracy of coordination, etc. But, on the other
hand, the greater the mental importance of a sound, the
greater the amount of energy given to it. To subordinate
in speech what is subordinate in thought is but one example
of a universal principle of human activity. It is also true
that neglected or unimportant elements of all organisms and
activities generally show greater tendency to variation than
the important ones. In speech, for example, tonic syllables
will resist changes that influence atonic ones.
The close relation between the density of ideas and the
distinctness of speech has been asserted by Jespeesen to
' account alike for most of the gradual sound changes in lan-
guages, and for . . . violent curtailings. ' ^
Another effect of fusion is found in changes in the coordi-
nations of the various simultaneous movements. Examples
of the almost innumerable experimental illustrations of this
principle have been given in Ch. XXVI ; only a brief sum-
mary of the typical cases will be given here.
A most frequent case of this kind occurs in the lack of
completion of some element in a complex set of movements.
Another case occurs in beginning some element too soon.
Tailure to properly coordinate breath action to mouth
action may result in the defect noted on p. 221. Failure to
maintain breath pressure sufficiently long during final occlu-
sives deprives them of their explosions, consequently of their
full audibility and ultimately of their existence,^ as in Fr.
' tout ' -* tu, ' trop' -* tro. Failure to begin the breath action
soon enough enfeebles initial fricatives and, if sufficiently
extended, causes them to disappear, as in the common loss of
initial h in Cockney English and in modern Spanish.
Failure to coordinate cord action to mouth action results
1 Jespbksen, Progress in Language, 55, London, 1894.
^ Passt, Changements phon^tiques, 164, These, Paris, 1891.
458 FACTORS OF SPEECH
in the partial surdation of intended sonants and in the partial
sonation of intended surds.
The cord tone frequently begins too soon or lasts too long
in a vowel bounded by a surd, with the result that the surd
becomes partly or wholly sonant. A combination of both
extensions gives sonant h in soh.im (p. 276).
A low cord tone is more readily lost than a high one.
The great fall in pitch at the end of a phrase in French
results often in making a final vowel surd, especially in the
case of i, u and y (as in veky^ ' v^cu '). These are said to
be still very audible when surd ; but for u this does not agree
with Rotjsselot's experiments on the whispered vowels (p.
114). In Portuguese the final vowels are frequently surd, as
in ka^mphUo ' campo. ' Many languages have only surd con-
sonants as finals. In German and Russian all final explosives
and fricatives are regularly surd ; in Icelandic the final ex-
plosives are surd, while the fricatives are more or less sonant ;
in French the Latin explosives and fricatives have all
become surd at the ends of words except those that have been
followed until recently by a, as in ' vif, vive ; ' in Sanskrit
the final explosives and fricatives are always surd unless
the following word begins with a vowel, as ' tdt ' but ' tdd
dnnam ; ' in Dutch the same rule holds good in unaccented
words, as in ' is, was, ' but in others the final consonant is
surd unless followed by a sonant consonant.^
Progression of cord action may add sonancy to a preceding
surd. Tardiness in beginning the cord tone would cause
initial surdation in initial sonants (p. 361) ; increasing tardi-
ness would finally make them entirely surd (Fig. 287).
This may be the reason for such changes as ■)(^eicro/j,at from
the root ' *ghend, ' ' f ero ' from ' *bhero. ' The late begin-
ning of the cord tone regularly produces ' aspiration ' in
German after the surd explosives.
The alteration of the mouth articulation before the cords
cease to sound gives the diphthongal character to English
^ Passt, as before, 160.
SOUND FUSION 459
vowels (pp. 103, 122). It occurs also in the Dutch vowels '
e, o, cEi, 062, ^' which in emphasized final syllables become ei,
ou, ce^i, ce2U, bI [the b is a sound difficult to define precisely].
Anticipation of lip action may labialize the preceding sound,
as in Fr. t'^wa for 'toi.' In this word the cord action is,
moreover, tardy and the w is partially or wholly surd, pro-
ducing t*^/Aa in the extreme case.
Initial and medial consonants often differ greatly. Initial
b is often partly surd in German while medial b in same word
may be entirely sonant. Semi-surdation and semi-sonation in
French have been illustrated above (p. 360) ; the frequent
surdation of final sonants has also been discussed; numerous
further examples have been collected by Rotjsselot.^
Any change in the tongue-lip relation will produce a vowel
not exactly like the one intended. A relaxation of the lip-
rounding for u with maintenance of the tongue position makes
the vowel change in the direction of the vowel in the Gaelic
' laogh ' ; the change can be imitated by speaking u while the
fingers pull out the corners of the mouth.
Examples of widely spread tendencies to change of lip action
while the tongue action is preserved have been collected by
Passy.3 Among children and the illiterate there is a marked
tendency to change the rounded palatals to neutral or wide
ones, oe, y, ce° to e, i, e", as in kivet for ' une cuvette,'
e" menje for ' un meunier. ' This change is a common one
in many German dialects, as in South German git for ' gUte '
and kena for ' konnen. '
Differences in motor coordinations intended to be the
same may be found in all experimental records.
Sounds in fused speech differ from the typical ones on
account of their auditory characters also (Ch. IX).
When a sound ceases to impress the ear, there is a ten-
dency to neglect it. The unflapped j is auditorily so weak
1 LoGEMAN, Darstellung des niederldndischen Lautsy stems, Phonet. Stud., 1890
III 283.
2 RousSELOT, Principes de plion^tique expe'rimentale, 2™e partie, 498, Paris,
1901.
3 Passy, as before, 134.
460 FACTORS OF SPEECH
that it has disappeared in a large number of English words
where its loss makes no noteworthy difference in the auditory
physiognomy of the word, as in pal for pajl. This can be
observed constantly in American children,* who say biii for
'bring,' toll for 'trolley' etc., because the faint unflapped
J escapes their notice.
It is generally the case that sounds in atonic syllables
tend to become weakened in articulation and to partly or
wholly lose their sonancy, and finally to disappear. This
slurring of atonic vowels may be illustrated by ' glaedlice ->
gladly, ' ' cnihtas -^ knights, ' ' populum -^ peuple, ' ' facere -*
faire. ' Such phenomena are perhaps due to the fact that
the stronger portion of the word is sufficient to arouse the
entire auditory image in the mind ; in such a case the speaker
or hearer might unconsciously not care what happened to the
weaker portion and would instinctively condense it as much
as possible. This view is supported by the experiments of
Bagley (p. 131).
Whenever a sound element catches the ear — more or less
consciously — with a force greater than what would usuallj'
come to it in its condition at the time, the speaker has an
impulse to give it more prominence by increasing some of
its factors of energy, such as length, loudness, abruptness,
etc. Thus, wild -^ wild -* waild probably because for some
reason the I before 1 caught the ear and was prolonged un-
consciously (in the usual sense, p. 381) and unintentionally.
The development of glides into vowel elements is probably
partly due to similar reasons. As it is difficult to keep the
movements properly coordinated in a long vowel, the ending
is liable to be different from the beginning; if this difference
catches the ear the difference readily becomes exaggerated.
The replacement of one vocal movement by another that
produces approximately the same sound is a constantly recur-
ring phenomenon. It rests upon the auditory resemblances of
vocal sounds. These have, unfortunately, received little
attention and almost no experimental study, although very
many phonetic changes can be attributed directly to them.
SOUND. FUSION 461
The attempt to assign some of the auditory reasons for the
motor changes has been made by Passy. His conclusions
concerning the vowels are based on a mistaken theory (p.
289) ; the few reasons assigned for consonant changes are
deserving of notice as attempts in the right direction.
The replacement of one movement by another that produces
a sufficiently similar sound is seen in^ Northern French fiXa
' fille ' -> fi], bri^e, ' briller ' -> brije ; Lat. ' plenum ' -^ Ital.
pjeno; Cretan avicd, Attic oKKri; French ' chevaux ' from
'chevals;' New Zealand pronunciations of English, as in
rawiri for ' David, ' the d being lacking and the r somewhat
resembling d. The close resemblance of r and 1 favors the
use of one for the other, as in O. French ' orme ' from
' ulmum, ' the Chinese pronunciation ' Melican ' for ' Ameri-
can, ' etc. The resemblance between -y and uvula r in Ger-
man often leads to interchange ; the resemblance of tongue r
to uvula r may even lead to the use of the former for -y.^
The close resemblance of x and -y to uvula r is suggested by
Hermann's curves of x and uvula r in Fig. 33; the flapping
of the uvula is plain in the curve for x-
The continual occurrence of individual differences of speech
movement held within the limits of ' sameness ' of auditory
effect has been described as ' sound compensation. ' ^ ^_,
Some of the phenomena of fusion will be illustrated in the
study of speech curves in Appendix II.
Refekences
For fusion in speech : Sievers, Grundzuge d. Phonetik, 5. Aufl., III.
Abschnitt, Leipzig, 1901; Vietok, Elemente d. Phonetik, 4. Aufl.,
Liepzig, 1898; Passy, Changements phonetiques, These, Paris, 1891;
Paul, Principien d. Sprachgesohiohte, 3. Aufl., Halle a/S., 1898.
1 Passy, as before, 145.
2 ViETOR Elemente d. Phonetik, 4. Aufl., 165, Leipzig, 1898.
8 Sheldon and Geandgent, Sound compensations, Mod. Lang. Notes, 1888
III 358.
CHAPTER XXXI
PEOGEESSITE CHANGE
The specific sounds in speech at any moment vary around
an average in each of their properties ; the average sound
has been called the ' tj^pical' one (p. 454). Any change that
takes place in the typical sound, resulting from a gradual
change in the conditions, may be termed a ' progressive ' one.
The typical sounds of an individual may change gradually
throughout life under the influence of internal and exter-
nal conditions. Such changes we may term ' personal ' ones.
There are also average, or typical, sounds of a community
speaking the same dialect; the progressive changes in such
types may be termed ' dialectal ' ones. In a similar way
we may speak of ' national ' changes. The typical sounds
may be studied in respect to their differences as the locality
differs ; we then have progressive ' geographical ' changes.
The work of the experimental phonetist in regard to
the problem of sound change consists in establishing what
changes actually do take place under any variation of con-
ditions, in correlating them with similar changes of various
kinds, and in deducing them from more fundamental changes
in the human organism and its activities.
The historical data concerning the changes that have actu-
ally taken place in speech sounds have been collected with
great care and completeness, but experimental data concern-
ing the action of the vocal mechanism (in the widest sense
of the terra) that can bring about such changes are almost
entirely lacking. All hypotheses concerning the historical
changes proceed on the assumption that human beings were
PROGRESSIVE CHANGE 463
always subject to internal and external influences in much
the same way as they are to-day. If changes are produced in
speech to-day like those that occurred historically, the causes
that produce the present changes can be assumed for the
historical ones, with a probability increasing with the re-
semblance of all the conditions.
Some of the various causes which have been assigned for
the small differences and gradual changes in sounds will now
be considered briefly in order to point out what the experi-
menter must be on the watch for in all his records. For
the various principles discussed I have used historical cases
merely as examples.
The dependence of differences in speech sounds on differ-
ences in the structure of the vocal organs and consequently
in their activity and control has been pointed out (Lotze,
Benfey, Merkel, Scheeer, Paul) but the data are not
yet numerous enough for generalizations.^ It is unquestion-
ably true that the vocal organs and their methods of control
differ in individuals and that there are just about the same
differences and likenesses among members of a family, of a
community, and of a nation in these respects as in respect to
their faces. The family, communal and national similarities
and differences in speech depend to some degree on these
factors. Moreover, although an infant of one race may learn
to speak to apparent perfection the language of another race,
we may expect that careful experimental records wiU show
the differences that are unnoticed bj'^ the ear.
No extended study seems to have been made of individual
differences in the auditory perception of speech (Ch. IX).
They are perhaps just as efficient causes of change as the
motor ones.
The results of motor weakening are familiar phenomena.
Persons to whom occlusions or narrow passages are more dif-
ficult on account of the structure of the tongue frequently
relax them in case of fatigue, excitement or timidity. In
the case of a person with a predorsal-alveolar t due to short-
1 Obrtel, Lectures on the Study of Language, 193, New York, 1901.
464 FACTORS OF SPEECH
ness of the frenum linguae (p. 237) I have repeatedly observed
t->^9 in such conditions. In the case of an overworked, ex-'
cited German in a foreign seaport I have observed si -^ si in
his pronunciation of Italian.
The use of a mouillure instead of a sharp explosion I have
recorded in the case of my own child at 16 months in the
words kaer ' cat, ' hoT ' hot, ' the | explosion being clearly
noticeable (p. 437). I originally recorded the words as kaetk
and hotk (p. 119), not having considered the possibility of
their containing the mouill^ sound t.
Passy's statement 1 that children never change explosives
into fricatives is certainly incorrect for some cases. His
supposition that it is easier to form an occlusion than a
fricative opening is refuted by observations on speech during
fatigue, alcoholism, etc. A weakening of the occlusion in
explosives seems to have been tlie characteristic of many of
the historic changes; thus, Spanish b, d, g are often pro-
nounced as the fricatives p, 8, -y.
The muscular action during an occlusive may become less
energetic during the latter part of the occlusion. When the
sound continues to occupy the same time, this leads to short-
ening of the occlusion and lengthening of the explosive re-
lease into a fricative release. If the process goes far enough,
the occlusion entirely disappears, leaving a fricative in place
of the original occlusive. This is perhaps the process that
produced the changes of p ^ <j) -* f , t -> 6, k ^ X or h, as illus-
trated by Lat. ' pallidus ' — Engl. ' fallow ' — Germ. ' f ahl, '
Lat. ' capio ' — Goth. ' haf jan ' — Germ. ' heben, ' Goth.
* sliupan' — O. H. Germ, 'sliofan,' Lat. 'tres ' — Eng.
'three,' Goth, 'bok' O. H. Germ, 'buoh.'
Lenz^ considers the historical development or -* S -* X
to be the result of a decrease of energy. The laxity of
contact for 6- results in greater fricative surface, including
the space between tongue and gums, and produces | -+ 9 — > X-
1 Passt, Changements phcmetiques, 144, Thfese, Paris, 1891.
' Lexz, Zur Physiologie v. Geschichte d. Palatalen, Diss., Bonn, 1887 ; also in
Zt. f. vergl. Sprachforsclmng. 1888 XXIX 51, 55.
PROGRESSIVE CHANGE 465
Still further laxity changes the fricative sound entirely; it
may be replaced by the glottal fricative h or by a vowel.
Lenz asserts that in the last case a substitution is required
for the loss of friction because there must be a balance
betvreen the breath pressure and the resistance to it ; this he
considers the fundamental law of all formation of sounds.
This change s ->• a is to be added, although Lejstz does not
consider s ->^ o- ^ o- to be the result of vi^eakening. The
entire series is then s ->■ cr (s) -» | ->^ 5 -*■ x ->■ h or vowel.
For the sonants the development is not the same ; as the
articulation is diminished in force the opening becomes
larger as well as more backward, and the glottis must open
in order to furnish an equal amount of breath ; the fricative
gradually becomes surd, or the friction disappears, leaving
only the cord tone to lengthen the neighboring vowel. Thus
we have z ^\ (z) -^ . . . ^ \ (instead of 7) ->^ h or vowel.
As examples, Lenz gives : Fr. mezo" ' maison, ' Lorraine
mozo°, Remilly moho" ; Fr. plezir ' plaisir, ' Lorr. piaezi
-^pigehi; Fr. bus 'bouche,' Remilly box; Fr. mus 'mouche,'
Remilly mox-
The fact that the sonants b, d, g more readily become frica-
tive than the surds p, t, k may be due to their less energetic
contact (p. 317). In German the g has often become frica-
tive, (as in vavn ' wagen,' gejn ' gegen ') while the k in
corresponding positions has been retained.
The continued weakening of consonants leaves finally h
for the surds and a vowel for the sonants, as in Sanskrit
' 45 was ' -> ' 49wah,' ' *sweks ' ^ Greek e^, Lat. 'filium ' ->
Span. ' hi jo.'
The weakening of the rolled r produces the fricative or
liquid J. In Southern England and in parts of America
the r has disappeared to a great extent, having become j in
some words, a in others and entirely omitted in still others.
In Parisian French the uvula r tends to become -y. Her-
mann's curve for the fricative x in Fig. 33 (p. 42) looks like
that of a weakened surd uvula r.
An excess of energy in articulation often occurs. This
30
jim FACTORS OF SPEECH
may be due to increased vitality of the organism, whereby all
movements are affected, or to diminished accuracy of control.
Increase in the energy of speech movement has been as-
signed by LenzI as the cause of the spontaneous develop-
ment of palatal fricatives into occlusives, illustrated by
the historical change of j to 5 and its later stages. The
more seldom i -* 8, 1 ^ X, and n ->■ n are likewise results of
increasing energy of movement. The process may be illus-
trated by examples. It is a familiar fact that cotis. + e
+ vowel ->- cons. + j [or 5] + vowel as the result of the
narrowing of the e-passage between the tongue and palate
so that e^i ->]-*] [or 5]. A further narrowing produces
the occlusion with fricative release, and j [9] -* 8 [t].
Such a development of Lat. ] has occurred in nearly all
Romance languages.
The development Lat. nn, II -* Span, n, A results from
the more energetic dorsal-alveolar closure (in Lenz's opinion),
whereby an increase of contact occurs further back in the
mediodorsal-prepalatal region. Palatograms ^ of an energetic
dorsal-alveolar 1 show contacts between those for ordinary 1
and A.
The air from the glottis passes in the occlusives into a
closed chamber; the pressure within the chamber must finally
become equal to that in the thorax and the vibration of the
cords must cease if the occlusion is sufficiently prolonged,
unless the energy of action of both breath and occlusive
muscles is increased. Double — that is, long — occlusives
tend to become surd (p. 367), as in Italian 'addentro,
aggettivo. ' ^
Economy of effort — motor, sensory and associational — is a
fundamental principle of the human organism ; it shows itself
in varying degrees in every individual and in every activity.
1 Lenz, Zur Physiologie u. Geschichte d. Palatalen, 49, Diss., Bonn, 1887;
also in Zt. f. vergl. Sprachforschung, 1888 XXIX 55.
^ Lenz, as before, 56.
2 JosSBLYN, iStude sur la phonitiqm italienne, 157, Thfese, Paris, 1900; also in
La Parole, 1900 11451.
PROGRESSIVE CHANGE 467
The principle of economy has been proposed in explanation
of most of the historical changes that have occurred by
weakening or increasing the articulatory effort. It undoubt-
edly does explain many of them, but many others are just as
certainly due to diminished or increased vitality of the ner-
vous system. The slurring of articulations due to diminished
vitality — due to laziness, as Sweet puts it^ — is certainly
quite different in its nervous origin from that due to
economy ; both cases result in a saving of labor, but the
former implies a poor condition of nervous activity, while
the latter is evidence of a good one. The objections to the
explanation of sound changes as being the results of econ-
omy have generally been due to the misconception that the
economy is entirely one of muscular work, to the expecta-
tion of large and sudden results, and to a confusion between
economy and weakness. That an adaptation of neighboring
articulations for economical production actually does occur
constantly has been abundantly shown in the preceding
chapter. It occurs as the result of the usual instinct to
save energy quite independently of an increase in speed.
The economy may also appear as an auditory one. Distin-
guishing differences involves labor of perception (p. 121).
When a difference is felt to be too sm'all for the necessary
clearness of perception, the tendency to economy will require
its exaggeration (p. 122). This may occur even at the ex-
pense of more articulatory effort ; the principle of economy
will strike a balance between the two. Economy, in its true
sense, implies efficient activity of the adjusting organism;
greater economy means a better activity of the nervous
system.
Increased energy of articulation may result from economy,
but, like all other human activities, it may be due directly to
increased vitality of the nervous system.
A tendency toward extravagance is just as fundamental a
principle of organisms as economy. Activities are constantly
being exaggerated and these variations may be seized upon
1 Sweet, History of Language, 32, London, 1900.
468 FACTORS OF SPEECH
and developed. We may expect to find many sound changes
that arise from extravagance.
The nature of the vocal movements depends on the speed
of utterance. A speaker ordinarily makes about the same
effort during a conversation unless he wishes to produce a
difference in effect. An increase in speed requires either an
increase in effort or a decrease in the muscular work per-
formed. For an expenditure of the same mental energy in
a given time an increase of speed involves assimilation of
contiguous movements and diminution in the energy, ex-
tent, accuracy and duration of the separate ones.
Speech depends directly on the energy of the idea to be
expressed by it. Various observations have been reported
concerning the details. Increase in energy of expiration as
well as in emphasis maj^ change sonants to surds because of
a wider opening of the glottis,^ which is due — I believe —
to an association with the increased breath action and not
to any necessity for letting the air escape at the moment ;
thus, emphatic ' dead ' may -^ d^edp or doCdoh.
Sensory or motor habits in the succession of sounds lead
to preferences for one form instead of another ; for example,
of a vowel within an unfamiliar group of consonants, of a
familiar succession of consonants for an unfamiliar one, etc.^
The influence of general habits of expression^ has been
suggested by Wtjndt ; for example, the habit of speaking
with open mouth among the Iroquois is given as the cause
of the absence of labials, that of subdued expression among
the Chinese, English, Americans, etc. as the reason for some
of their speech peculiarities.
The differences in the organism of the child (sensory,
motor, associational) lead him to attain understandable
speech in somewhat different ways from those of adults.
These habits may remain to some degree as he becomes
, ^ SiETERS, Gruncjzuge d. Phonetik, 5. Aufl., 290, Leipzig, 1901 ; Vietor,
Elemente d. Phonetik, 4. Aufl., 278, 1898.
'' Oertel, Lectures on the Study of Language, 220, New York, 1901.
8 WuNDT, Volkerpsychologie, 1900 I 359, 403 ; Oertel, as before, 198.
PROGRESSIVE CHANGE 469
older and thus make his speech slightly different. The
transmission of sounds is generally practically perfect ; ^ but
it is safe to assume that finer methods of observation will
show that no sound is ever made in exactly the same manner
by succeeding generations.^
The results of mental and bodily processes at work in pro-
gressive changes may be summarized by 'phonetic laws.'
These are based on the principle that the human organism
acts according to laws as precise and vahd as those of the rest
of nature ; exceptions to a phonetic law (in this sense) indicate
merely that the law has not been properly formulated. The
term ' phonetic law ' has been used to include not only the
mental and bodily factors but also inferences concerning
the history of sounds ; such ' laws,' as frequently pointed
out, are not the same as natural laws.
The changes among individuals and communities that are
occurring under the influence of surroundings can be regis-
tered by experimental means. The accumulation of phono-
grams, palatograms, breath records, tongue curves, etc., etc.,
from an individual year by year under the influence of a cer-
tain environment or internal condition would indicate the
results of such conditions. Agreement of results from differ-
ent cases would lead to definite general conclusions.
The influences at work are generally very complex and
very slow ; experimental methods may often be devised that
accurately determine the action of single factors or of many
factors in a brief time. Such typical investigations of con-
crete sounds have been made by Lenz, Rotjsselot, Laclotte,
and others.
Investigations of more general problems have hardly been
attempted. The various hypotheses that have been put forth
as explanations of phonetic changes might be directly tested
by reproducing the conditions and recording the speech
results. Thus the hypothesis that the changes known as
1 Sweet, History of Language, 20, London, 1900.
'' RocssELOT, Les modifications phon^tiques du langage, 349, Rev. des pat. gallo-
rom., 1 893 IV, V ; also separate.
470 FACTORS OF SPEECH
Grimm's law are the results of increasing rapidity of speech
might be tested by recording language spoken at different
speeds ; the finer details that cannot be detected by the ear
could be measured in tracings made as described in Part I. If
increasing rapidity was the cause of the historical changes, we
can confidently expect to find indications of such changes in the
phonograph records. Thus, if the Indogermanic ' *ghortus '
changed to ' chortus ' and ' hortus '_ or to ' garto ' and ' garten '
as the result of increased rapidity, exactly similar changes
of a less degree should be found to-day in similar words
spoken with increased rapidity by persons ignorant of the
object of the experiment. The methods of registration are
now delicate enough to exhibit the details. The absence
of such results in carefully executed experiments would tell
heavily against the hypothesis, while their presence could be
accepted as final proof.
Even if increased rapidity has brought about some of the
changes, other changes have probably arisen from other
mental factors. It is unquestionably true that mental and
bodily vigor or weakness show themselves clearly in the
speech of an individual. If speech records on individuals in
various conditions of excitement, health, depression, fatigue,
etc., show regular relations between these conditions and
certain sound changes, and if similar relations are found
between the vigor and the speech of communities now exist-
ing, there will be strong presumption of similar relations in
the past history of speech.
The experiments may be extended to unusual, abnormal
or pathological ponditions of the individual. Thus the pro-
gressive defects in coordination of muscular action whereby
thickness of speech is caused may be carefully observed in
the progress of alcoholic or etheric intoxication. The effects
of rapidity can be observed at each degree of speed in the
slowly increasing rapidity of speech in some cases of mania.
^ Various other conditions may be obtained by administering
drugs.
It is quite possible that experimental data may definitely
PROGRESSIVE CHANGE 471
confirm any one of the hypothetical causes assigned, but the
well-known facts of mental life make it quite as possible
that all these causes and others also are involved. I may
add here that arrangements have already been made for
making gramophone plates of speech under various con-
ditions of excitement, emotion and fatigue so that the curves
can be traced off by the machine described in Chapter IV.
Plans are also being developed for the systematic preparation
and preservation of speech records in phonetic libraries.
References
For general works on phonetic change with references to monograph
literature: Paul, Principien d. Spraohgeschiohte, 3. Aufl., Halle a/S.,
1898; Passy, Changements phonetiques, These, Paris, 1891; Wundt,
Volkerpsychologie, I, Leipzig, 1900 ; Oertel, Lectures on the Study of
Language, New York, 1901; Sievers, Grundziige d. Phonetik, 5. Aufl.,
IV. Abschnitt, Leipzig, 1901.
CHAPTER XXXII
MELODY
A DISCUSSION of the melody of speech should, perhaps,
include a treatment of melody in song, particularly the
primitive song of uncultured peoples and the spontaneous
song of children. The unconscious modifications of a musi-
cal melody made by a singer should also be treated in their
dependence on different conditions of activity of intellect
and emotion. In spite of the great importance of these
topics very little experimental work has been done. Collec-
tions of phonograms of the voices of singers have been made
but have not been studied. Several collections of the songs
of the American Indians have been made ; one of these was
carefully studied by Fillmore. ^ The present chapter will
be confined to a study of pitch in speech.
It was pointed out by Aristoxenus ^ that the difference
between the tones in song and those in speech lies in the fact
that the voice in singing proceeds by jumps from one note to
another, while in speech it continually slides up and down.
The fact has been thus stated by Storm : ' Characteristic for
the voice in speech is its continual gliding through several
tones whereby these do not impress the ear as clearly differ-
ent musical tones but as an impure mixed unmusical noise.
. . . The character of speech music can hardly be completely
1 FiLLMOBE, The harmonic structure of Indian music, Amer. Anthropol., 1899
1297.
2 Aristoxbnus, Harmonica, I § 25, p. 8, Meib. (the passages are collected
in Johnson, Musical pitch and the measurement of intervals. Thesis, Baltimore,
1896); ARiSTOXENns, Harmonica, I § 28, p. 8, Meib. (quoted in Johnson, The
motion of the voice in the theory of ancient music, Trans. Araer. Philol. Assoc, 1899
XXX 47); GooDELL, Chapters on Greek Metric, New York, 1901.
MELODY 473
investigated without the help of a phonograph or a similar
instrument which records and measures the musical glidings
completely. ' i
The pitch of short speech sounds is hard to catch by the
ear not only because each sound contains many tones that
influence the total impression (p. 95), but especially because
the pitch is always changing. Even from a long sound the ear
receives only a vague impression of pitch when it is a chang-
ing one. These difficulties render it impossible to obtain by
the ear any reliable data concerning the melody of speech.
Experiments by Maktens,^ in which the tone of a siren was
made to fall continuously at different rapidities, showed that
the ear heard two successive tones when the fall of pitch was
very rapid, three when less rapid, etc.
A problem by itself is that of th^ impression the ear
receives of the melody; in all except the vaguest generalities
such as high and low, rising and falling, the impression
differs from the actual melody produced. Many interesting
and valuable observations have bfeen recorded of the impres-
sions of melody. 2
An impression of a kind of average pitch may be obtained
by disconnecting the side feed of the reproducer and placing
it at any -desired point on a phonograph record. It thus
repeats continuously the sound contained in one turn of the
groove.*
Experimental determinations of melody have been made by
the methods described in Part I.
The interesting results obtained by Maktens (p. 19), are
summarized in the following Table. In the Figures (redrawn
from his chart) the horizontal distance indicates the serial
number of the vibration; it may be taken roughly to repre-
1 Storm, Englische Philologie, 2, Aufl., I 205, Leipzig, 1892.
2 Martens, Ueber d. Verhalten v. Vokalen u. Diphthongen in gesprochenen
Worten, Zt. f. Biol., 1889 XXV 297.
3 Summary by Stokm, as before, 188.
* Makage, Les phonographes et I'etude des vof/elles, Anne'e psychologique,
1899 V 226; Thierrv, Le tonal de la parole, IV"' Congr. Interuat. de Psychol.,
Paris, 1900.
474
FACTORS OF SPEECH
sent time, the unit decreasing with a rise in pitch. Martens
leaves no spaces for the intervals occupied by the consonants.
The proper method of plotting the curve of melody is that
described in Chapter V. These curves, however, may be
considered as indicating approximately the melody of the
words.
Table
Phrase
Voice Sound Length Frequency
IN SEC. Lowest Average Highest
' Vater und Mutter '
Male voice,
a
0.231
139
165
178
(Fig. 325)
71 years
3
0.331
168
185
201
U
0.257
164
183
201
u
0.123
158
187
201
3
0.136
139
174
189
' Der Donner roUt '
Male voice,
e
0.162
160
173
189
(Fig. 326)
29 years
o
0.126
170
183
208
3
0.189
201
211
221
O
0.226
146
160
178
' Vater und Mutter '
Male voice.
a
0.181
362
394
453
(Fig. 327)
13 years
3
0.249
342
425
447
U
0.118
345
377
388
u
0.140
428
441
453
3
0.262
296
338
362
' Oh neln ' (high) , ' oh nein '
Male voice,
O
0.226
283
334
394
(reproachfully)
29 years
ai
0.283
156
242
302
(Fig. 328)
o
0.261
133
178
197
ai
0.417
103
133
158
' Lauf , mein Kind '
Male voice,
au
0.303
144
212
273
(Fig. 329)
29 years
ai
0.239
173
195
245
i
0.106
197
228
238
' Back sUsses Brot '
Female voice,
a
0.104
307
325
342
(Fig. 330)
30 years
Back
0.366
0.205
y
0.153
312
347
370
sii
0.172
e
0.095
288
296
324
sses
0.411
T
0.184
o
0.229
273
324
348
' Bau hiibsche Hanser '
Female voice.
au
0.328
302
326
355
(Fig. 331)
13 years
Bau
"1
0.342
0.099
y
0.077
318
325
336
hiib
0.194
3
0.105
312
323
336
sche
0.215
3i
0.216
292
328
351
Hau
0.152
3
0.083
229
252
283
ser
0.203
MELODY
475
a /u
Fig. 325.
A /r^°
Fig. 326,
Fig. 327.
ai
Fig- 328.
aa
'V
Fig. 331.
476
FACTORS OF SPEECH
PHEiSE
Voice
Sound
Length
Fkequenc
Y
IN Sec.
Lowest Highest Average
' Was giebt 's dort f '
Female voice,
a
0.155
329
371
431
(Fig. 332)
33 years
was
0.201
0.517
i
0.145
377
463
518
giebt 's
i 0.609
T
0.311
o
0.178
273
309
342
dort
0.325
' Mein kleines Kind '
Female voice,
ai
0.267
292
313
336
(Fig. 333)
13 years
mein
1
0.473
0.064
ai
0.157
307
321
336
klein
0.283
e
0.051
304
311
318
es
0.073
1
0.073
i
0.116
283
309
345
Kind
0.328
Martens made no application of his results to speech
melody. They seem to show clearly that the cord tone
never stops on any note but is always rising or falling; even
the shortest vowels show just as rapid changes as the long
ones. The affirmative phrases end in a vowel of a lower aver-
age pitch than the others but of a ' circumflex melody. ' The
last vowel in the commands has a higher average pitch ; the
circumflex form of melody also appears in them.
In regard to the charts of the course of the cord tone given
by Martens I may point out that in two independent words
' Mokka ' and ' Mutter ' the first vowel shows a rising pitch
and the second a falling one ; each word as a whole may be
said to be of ' circumflex pitch.' The phrases also have
usually a circumflex form to conform with which the word
circumflexes are modified; some phrases, however, show
other forms.
Schwann and Pringsheim^ have made phonautograph
records (p. 17) of French words and phrases. Measure-
ments of the speech curves showed that as a rule both vowels
in a two-syllable word were spoken at the same pitch, with
the same intensity and with the same duration. A word at
1 Schwann und Pringsheim, Der franzSsische Accent, Archiv f. d. Stadium
d. neueren Spr. u. Lit., 1890 LXXXV 203.
MELODY
477
Tig. 332.
Fig. 333.
the end of a sentence differed from that within a sentence by
having a lower pitch and a less intensity. An isolated word
was pronounced like a word at the end of a sentence. One
.of the curves of melody is shown in Fig. 334; the hori-
zontal axis indicates time in seconds.
300-
250-
ZOO-
i
-
'*'"*''■**-_
C /
ISO-
100
SO-
il est mi di \
0
0 0.1
1
0.2
0.4 0.S 0.6
/./ 1.2 1.3 l.t AS I.e 1.1 1.6 1.9 2.0
Fig. 334.
PiPPiNa (p. 121) gives melody curves for a number of
independent Finnish words ; they show without exception a
circumflex pitch. When the changes in pitch are not of a
complicated character, we can consider a vowel in which the
average is higher than its beginning as a ' vowel of rising
pitch ' and one in which the average is lower than its begin-
ning as a ' vowel of falling pitch. ' Pipping's dissyllabic and
trisyllabic Finnish words showed without exception a vowel
of rising pitch in the first syllable, a vowel of falling pitch
in the last syllable, a vowel of steady or falling pitch in the
478
FACTORS OF SPEECH
middle syllable of a trisyllabic word, the average pitch being
lower for each succeeding syllable.
The curves of the vowels of German words spoken iso-
latedly, recorded on a phonograph and measured by the aid of
a corneal microscope, have been studied by Meyee.^ In a
long German vowel, under the conditions given, the cord tone
begins low and rises (region of tone-rise), remains a time at
a maximum (region of tone-maximum), and falls (region of
tone-fall). The course of pitch is greatly influenced by the
neighboring consonants; the more emphatic the consonant,
the greater is its influence on the pitch-curve; the following
consonant often cuts the vowel oft' at or near the maximum.
With short emphasized vowels the influence of the conso-
nants is even greater than with long vowels. With short
unemphasized vowels the cord -tone has only the region of
tone-fall. Every vowel has a favorite pitch for its tone-
maximum in this descending order: u, i, o, e, a. Emphatic
consonants raise the tone -maximum. Greater loudness seems
to be regularly accompanied by higher pitch, on the prin-
ciple — I may suggest — of greater muscular effort in one
direction being accompanied by greater effort in others.
The course of pitch in the cord tone can be registered with
some accuracy on the breath-curve by a short tube from the
lips to a tambour. Experiments by Vietor ^ gave results
for the u of ' du ' which are roughly indicated in the following
notation :
questioD
warning
The breath-curves for these forms of ' du ' are given in
Fig. 87 (p. 218). Records of sentences showed that em-
1 Meter, Zur Tonbemegung des Vokals im gesprochenen und gesungenen Einze-
woTt, Neuere Sprachen, 1897 IV Phonet. Stud. 1.
2 Vietor, Elemente d. Phonetik, 4. Aufl., 293, Leipzig, 1898.
MELODY
479
phasis on any portion of the subject or the predicate raised
its pitch.
Using a phonograph, Marichblle found ^ melody curves
as follows (the staves are of the treble clef) :
Qui est Ik . ? C'est Paul . . C'est Paul . t Tiens . . ! Voim . . Jeau
^^^-^^F^
Ah
! Jean . Re . . ne . . . . ! Fer . di . nand! As . sez . . . . !
Com ment ! tu n'as pas tra va illd
(female voice)
P
^
=5z:
Tres bien . . ! Com ment ! tu n'as pas tra vaille . . ! (male voice)
Com (1) (■?) n'as pas tra . . va. . . ilM . . . . !
(voice of deaf person).
Fig. 335.
The curve of melody (or curve of pitch) for the tracing of
Jefferson's voice in Rip Van WinMe's Toast (p. 61), as
reproduced in Plates III to XI at the end of this volume, is
given in Plates XII and XIII.
The measurements of the successive periods of the cord
tone were made and computed according to the methods
explained in Ch. V. The plot was made as described on
p. 65. The dots were, however, joined by straight lines, and
a smooth curve was not drawn through them as in Fig. 44 ;
the successive dots are more readily indicated in this way but
the general course of the cord tone would be more truthfully
represented by a smooth curve.
Horizontal distance in Plates XII and XIII indicates time
1 Marichblle, La parole d'aprfes le trac^ du phonographe, Planche U, Paris,
1897.
480 FACTORS OF SPEECH
at the rate of 1°™ — 0.0035% or a reduction to one fifth the
size shown in Plates III to XI. The vertical scales indicate
frequency, or the number of vibrations a second. Each
group of words refers to a portion of the melody-curve
extending from its beginning to a group of large figures on
the horizontal line ; during each portion the horizontal line
remains unbroken. The large figures indicate, as in Plates
III to XI, the portions of straight line in the original tracings
that were omitted in preparing the Plates ; they are turned
into time by the equation 1"™ = 0.0007^
The interruptions in the melody-curve indicate surds, or
very weak sonants, or pauses.
The curves in Plates XII and XIII indicate a very low and
even melody of speech that is varied at times for emotional
expression. In general each sentence begins low, rises
gradually, and then falls, but variations occur. The changes
in the tone are generally continuous.
' Come, Rip ' shows a rise at the end, which is a common
inflection for a cheerful, animated invitation. ' What do you
say to a glass ? ' shows a low vowel, then a rise to the u of
' you ' ; this u, however, begins to fall just before the follow-
ing word. ' Say ' is of high pitch, as is frequently the case
for the verb of a question ; the fall at the end of ' you ' may
have been a kind of preparation by contrast for the high pitch
of ' saJ^' The highest pitch for the phrase is found in ' glass ' ;
it is even higher than in 'say,' probably because of the
greater emphasis given to the word ' glass.' The pitch falls
toward the end of ae in 'glass'; such a fall is usual in a
sentence beginning with an interrogative word (or phrase)
that is not specially emphatic. These words were spoken by
Jeffeeson" as introductory to the Toast itself. The invita-
tion is followed by a long pause of 2.86' before the reply
comes.
The Toast begins with a repetition of the question of invi-
.tation. It is spoken in a rather soft manner, as appears not
only to the ear but also in the small amplitude of the waves
in Plate IV. The pitch curve is fairly level, with some rise
MELODY 481
at the end instead of a fall. This rise is the usual ending of
a repeated interrogative sentence. The general pitch is lower
than that of the invitation. A pause of 0.41° follows.
The exclamation 'huh' is a kind of chuckle. It is of a
very high pitch but small intensity and short duration. It is
followed by a pause of 1.05\
' Now what do I generally say to a glass ? ' shows a very
even rise and a very gradual fall ; its general pitch is low.
It is a kind of bantering statement. The long pause of 2.16'
seems to express a simulated expectation of a reply.
' I say it is a fine thing ' is a decided statement with em-
phasis on ' fine thing.' It has the usual circumflex form as
far as ' a.' If the sentence had been completed with no
further emphasis, the pitch would probably have continued to
fall. The rise in pitch for the specially emphatic ' fine thing '
adds an accessory circumflex. The pause of 1.78' and the
fall in pitch lead the hearer to suppose the sentence finished.
' When there 's plenty in it ' is muttered as a joke. Its
pitch is not lower than usual. The emphasis on ' plenty in
it ' gives a higher pitch to the latter portion. The whole state-
ment has the usual circumflex form. The long pause of 2.90°
is presumably occupied by the first sip of the toast.
The soft exclamation of satisfaction 'ha' has a falling
pitch. It is followed by a pause of 1.79°. The 'so' ex-
presses deep satisfaction. It begins moderately high and
falls steadily in pitch. To the ear it has a peculiar rattle of
a low pitch as if some particles of liquid had lodged on the
edge of the epiglottis, as is sometimes the case after di'inking.
This peculiar effect shows itself in the alternately louder
and weaker character of the groups of vibrations shown in
Plate VII. Such a curve could be produced by the cord
explosions striking against a mass of liquid that would
vibrate readily at a sub-multiple of the cord period ; the por-
tion of liquid would rise and fall, weakening the cord tone
on alternate periods. It is quite probable that when speaking
into the gramophone recorder Jeffeeson produced this effect
by some muscular adjustment (epiglottis, ventricular bands)
31
482 FACTORS OF SPEECH
and not by an actual sip of some liquid. ' So ' is followed by
a pause of 2.00^
' You had it ten years ago, eh ? ' is spoken as a continuous
sentence ; there is complete fusion of the vowels at the end.
The first part rises rapidly to a high pitch. The circumflex
form is marked, the fall beginning in the o of 'ago.' The
' eh ' has a circumflex form joined to the o curve. In spite
of the complete fusion of these vowels we may perhaps con-
sider ' eh ' as a stressed tag with a pitch-curve of its owu.^
The sound indicated here by ' eh ' begins with a very weak
breathing and seems slightly nasalized ; it thus inclines some-
what toward h9°. The long pause of 2.45° indicates perhaps
the time of another sip.
' Ah ' is an expression of satisfaction ; it appears to the ear
much lower and smoother than the 'ha.' The following
pause is very short, 0.13'.
' That 's fine schnapps ' is not an emphatic statement but
expresses a decided conviction after a satisfactory trial. It
shows the usual initial rise for a declarative sentence, but
instead of falling at the end, it rises slightly. This peculiar
rise seems to express conviction after a doubt. The figures
333™™ after ' that 's ' indicate the portion of tracing (tsf)
left out in the original record, and not a pause. The sen-
tence is followed by a pause of 1.92^
' I would n't keep it as long as that ' has the usual circum-
flex form ; it is followed by a brief pause of 0.29'.
' Would I ' is used to turn the declaration into a question.
It is very brief. A short pause, 0.25% follows.
The verj' brief and faint chuckle ' huh, huh ' is followed
by a pause of 1.20'.
The introductory 'well' — presumably spoken as the glass
is lifted — rises steadily to a high pitch. It is followed by
a long pause of 3.43'-
' Here 's your good health ' rises steadily to a very high
^itch. The speaker makes a rather long pause, 0.94', perhaps
for emphasis. He then completes the thought in his mind by
1 Sweet, New English Grammar, II 39, London, 1S98.
MELODY ' 483
' and your family's.' This tag-phrase has, however, some-
what the character of a separate sentence; its pitch-curve
is circumflex. It is followed by a pause of 1.54^
The invocation ' and may they all live long and prosper '
appears to have the solemn steady intonation of a somewhat
religious utterance. The pitch-curve shows great evenness ;
there is a rise at the beginning and a fall at the end. The
fall appears during the first part of 'prosper'; during the
last part the cords are so relaxed that they produce only
a few rather irregular vibrations (Plate XI) ; the last
syllable appears to the ear almost as a surd or whispered one.
It is followed by a pause of 1.74', during which the toast
is presumably drunk.
The ' ah ' is a low, soft exclamation of gustatory satisfac-
tion after the toast. The peculiar rattle is heard as in ' so '
above ; the same alternation in the character of the groups
of cord vibrations appears in Plate XI. The pitch-curve
shows a steady fall. The last vibrations are of a very low
pitch ; they appear clearly in the tracing but are probably too
low in pitch for the ear to catch (p. 98).
Several series of experiments have been made on the
melody of speech sounds, the average pitch of each sound
being noted.
Rotjssblot's records ^ of the kind shown in Figs. 284 to
286 enabled him to give certain average frequencies, as, for
example,
mo" po V pj a r e
240 220 220 210 210
The melodies of two selections of his dialect were worked out
in this way and expressed in musical notation. From these
results RoussELOT^ concludes for French: 1. In ordinary
voice an isolated vowel has no fixed pitch for the cord
tone. 2. Consonants are generally higher than vowels.
3. Generally the approximation of a vowel and a consonant
' ]loussELOT, Le.s modifications phonetiques du langage, 109, Rev. des pat.
gallo-rom., 1893, IV, V; also separate.
2 RoussELOT, as before, 40.
484 FACTORS OF SPEECH
raises the pitch of the consonant and lowers that of the
vowel. 4. The voice often varies gradually in pitch in a
single syllable. 5. There is a musical rhythm which is less
influenced by the physiological conditions of speech than the
rhythm of intensity or that of duration, and which is conse-
quently better adapted to render the shades of thought. 6.
In words of two syllables the second one is raised in pitch in
Roxtssblot's dialect, but this pitch accent is less firmly fixed
than the duration accent. 7. In words of three and four
syllables the musical accent of rise in pitch coincides gener-
ally with the accents of duration and intensity, with differ-
ences sometimes in the unstres"sed syllables. 8. Groups
ending in an atonic have the pitch accent in the same place as
the historical accent even when the atonic has become longer
and louder than the tonic (as in kokote ; frequencies : 480,
600, 520; periods in hundredths of a second: 6, 11, 9 ;
the e still has its ancient accent shown by a rise in pitch).
9. The phrase is a song whose measure follows the intensity
or the duration of the syllables and whose melody chiefly
follows the changes in thought. 10. The pitch accent falls
readily but not necessarily on the more intense or the longer
syllables.
In the record of the first stanza of Cock Rohin (p. 58),
I find the average periods of various sonants as follows (in
thousandths of a second) : —
Who ki lied Cock R obi n?
3.3 5.1 4.2 1.8 5.3 5.6 8.4
I, s ai d th e sp a rr ow,
18-4 5.3 5.3 5.3 2.8 5.2
W i th m y b ow a nd a rr ow
5.3 2.1 5.3 5.6-3.6 7.0 5.3 4.2 2.5 7.0
I k i lied C o ck R o b i n.
12-4 5.6 7.0-5.3 . 3.9 3.9 4.2 5.6 8.8
In studying the melody of a vowel we may for some pur-
poses disregard cavity tones, overtones and the peculiar char-
acteristics of the cord puffs (p. 94, 260), and can conven-
MELODY 485
iently consider the tone as a simple sinusoid (p. 2); the
equation of the vibrating particle will then be
y = F(t-) sin ^"^^
/(O
where y is the elongation, Fif) the amplitude, and f(t) the
period. If the amplitude is constant as in the curves of
T'igs. 19 and 31 (though this is rarely the case in a spoken
vowel), we have F^f) = a.
A vowel during whose course the pitch remains constant
can be said to be of ' sustained ' pitch. If ^is the period of
vibration of the cords, we have in the ideal case
27r«
y = F(f)siry~.
Vowels of sustained or constant pitch are not very com-
mon in the cases I have studied. Most vowels seem to rise
or fall, yet some of them are approximately constant. The
vowel i as found in ' see,' 'needle,' 'I,' etc., is approximately
a sustained vowel although it generally falls slightly. The
following measurements of i in ' see ' are typical (" = 0.001") :
2.3, 2.3, 2.4, 2.4, 2.8"^ ... to the 22d vibration, 2.4- to the
42d vibration, 2. 1'^ to the end at the 64th vibration. (^Cock
Rohiii, Series I, p. 58; see also Appendix II.)
The rather unusual case of two vowels of sustained pitch
forming a diphthong is found in the word ' my ' of the phrase
' With my bow and arrow. ' The a has a constant period of
5.6^ and the i that of 3.6"^. The a has also a constant ampli-
tude of 0.4™"; the i, beginning with 0.5™ falls to 0 as usual
in ai in an independent word (^Gock Rohin, as before). The
diphthong ai is of nearly constant pitch throughout most of
its length in two cases of ' thy ' in Lord's Prayer, Series J
(p. 58).
Nearly all vowels in the earlier parts of words in the
records studied (p. 58), whether preceded by a consonant or
not, are characterized by a rising pitch. In such a case the
period is not a constant, T, but a function of. the elapsed time,
fif). A typical example of this kind of vowel is regularly
486 FACTORS OF SPEECH
found in the a of ai. A determination of the particular
form of /(O for various vowels is a highly important matter,
as different vowels and different manners of speaking are
possibly characterized by different forms of this rise in pitch.
Some of the cases of a suggest the form /(O = ke^', a
formula which expresses many of the phenomena found in
nature (k and m are empirical constants, and e the constant
2.71828).
When the rise in pitch (decrease in period) is proportional
to the elapsed time, we have
2-77^
where T^ is the period of the first vibration and m the factor
of proportionality. Such a vowel is found in the a of the
fourth example of ' I ' in Coch Mobin, Series I. During an
interval of 180'^ its period is shortened by 5.5", or at the rate
of 0.03^. Its cord equation, on the suppositions made above,
would be (in seconds)
y = i-(0smg_^_Q3^-
In the latter portions of words the vowels in the records I
have examined are generally nearly constant in pitch, with
often a slight fall as the intensity decreases. Typical ex-
amples are found in the cases of i in ai (Appendix II). This
slight fall in pitch does not necessarily indicate a relaxation
in the tension of the vocal cords ; as the force of the expired
current of air decreases, the frictional forces involved in the
cord vibration may gradually lengthen the period. Yet the
amount of fall is generally too great to be due to anything but
a relaxation of the cords. ^
Melody is used as a speech factor in different ways. As a
method of distinction among speech sounds, it may be used
like any other speech factor, such as an explosion or a reso-
nance. In Chinese this use has reached a high degree of de-
1 SOEIPTDKE, Rfsearches in experimental phonetics (first series], Stud. Yale
Psych. Lab., 1899 VII 93.
MELODY 487
velopment, the pitch of the cord tone being used phonetically
just as all languages use the pitch of the cavity tones in
the mouth to make different vowels. The words that to us
appear, for example, as ba spoken on four different tones
are for the Chinese perhaps more distinctly different than
ba, bo, bi, be would be to us. I have observed this percep-
tion of the cord tone as a characteristic part of a speech
sound by an infant who used the degrees and modulations of
intonation for each sound as it occurred in the first speaker
from whom the sound was learned. In most European lan-
guages this use of melody to distinguish special sounds is
subordinate to its use as a means of expressing more complex
mental states, as in assertion, question, etc.
The preceding account has treated the cord tone alone as
the basis of melody ; a fuller treatment would include all the
relations of harmony among the various resonance tones and
the cord tone ; a vowel is in fact not a melody alone but a
harmonized piece of music. Experimental data on this
subject are almost entirely lacking ; the harmony in vowels
is illustrated in Appendix II.
References
For observations and literature on speech melody: Sievers, Grund-
ziige d. Phonetik, 5. Aufl., 242, Leipzig, 1901 ; Storm, Englische Philolo-
gie, 2. Aufl., 1, 203, Leipzig, 1892; Viktor, Elemente d. Phonetik, 4.
xVufl., 290, Liepzig, 1898 ; Swbet, New English Grammar, II 37, Oxford,
1898 ; Hempl, German Orthography and Phonology, 167, Boston, 1897.
For literature on and views of the nature of melody: Hblmholtz,
Lehre v. d. Tonempflndungen, 5. Aufl., Braunschweig, 1896; Stumpf,
Tonpsychologie, Leipzig, 1883-1890; Wundt, Grundziige d. physiol.
Psychologie, 4. Aufl., Leipzig, 1893; and the large monograph litera-
ture in the bibliographies of Zt. f. Psychol, u. Physiol, d. Sinn., Ann^e
Psychol., and Psych. Rev.
CHAPTER XXXIII
DURATION
As in the case of melody the treatment of the duration of
speech sounds should include — perhaps begin with — a study
of these sounds in song. The conditions in song are simpler
than in speech, owing to the fact that a norm is imposed upon
the singer, from which his variations are to be treated as
measurable expressions of his personal rendering. Songs are
never sung — or intended to be sung — exactly as written.
Even the most mechanical popular tune is rendered differently
by each individual, the differences lying mainly in the duration
of the elements, in the stress assigned to them, and, above
all, in the attack by the voice and the utterance of each
sound. In artistic performances all these sources of varia-
tion are employed — mainly unconsciously — to express the
thought or emotion of the singer. Concerning just how they
are varied and how they are employed there are at present no
experimental data. Curves of various songs have been traced
off by the machine described in Ch. IV, but have not yet
been studied.
The duration of a portion of speech may be registered 1.
automatically, 2. by a special movement of the speaker, 3.
by a movement of a different person. The results of investi-
gations will be considered in connection with the method
employed.
The automatic methods consist in direct registration of the
voice vibrations (Part I) or of the vocal movements (Part III).
The time occupied by the sound may be determined by any of
the methods used in such cases.
DURATION 489
From the Tables (pp. 474, 476) of the results of measure-
ments by Martens it seems clear that the length of a vowel
in a given word varies greatly even for the same speaker.
Thus, for ' Vater und Mutter ' Maetens found the follow-
ing durations in thousandths of a second :
M
Male voice
71 years
231
331
257
'* **
52 "
269
120
K tt
29 «
183
133
121
tt tt
It (t
145
122
97
tt ((
It tt
205
127
150
i( it
28 "
166
213
296
tt U
26 "
318
222
117
tt u
13 "
181
249
118
iTemale "
6 "
177
100
56
123
136
148
94
67
179
64
98
151
116
140
262
84
We see here all sorts of variations. Even the so-called long
a in ' Vater ' on three occasions was found to be shorter than
e. The so-called short u of ' und ' was often longer than
a of the ' Vater, ' but was often also extremely short. Other
similar cases may be found in the Table on pp. 474, 476.
Measurements of the lengths of the speech sounds in the
Jeppbeson records were made by an assistant under my guid-
ance. The completeness of the fusion of sounds in connected
speech (p. 451) made it impossible to assign any very definite
limits to most of the sounds. When a sound was next to a
pause or a surd, its limit was placed at the extreme vibration.
Thus the first vibration of a in ' Come ' (Plate III, hue 1) and
the last distinct one of i (hne 4) gave fairly definite hmits. Yet
the curve shows quite clearly that the i-vibrations began to be
weakened by closure for the p somewhere about 60°"° from the
right end of line 4 ; faint vibrations can, however, be detected
at about 15°™ from the end ; thus even in a case like this it
is impossible to mark off the limits of i, i-p glide, and p. In
other cases there is no possibility of assigning any limits,
because the sounds are fused into gradually changing ones ;
thus in line 13 the u of 'to ' changes to a ' a,' but the change
is a gradual one beginning far back in the u and extending
throughout the a.' In fact, there are not two sounds u and a
united by a glide ; there is a changing sound which at some
490
FACTORS OF SPEECH
one instant may be an u and at a later one may be an a, and
which to the ear (trained to various associations) gives an
impression resembling a sequence of u and 3. In spite of
these facts I venture to give figures for the duration of
sounds in the Jefferson records in order to furnish some
approximate data; the figures are subject to the limitations
just explained ; where I have been utterly unable to decide on
a limit I have indicated the fusion by a brace.
The phonetic notation is used in the following list merely
m
I
i
P
■1
h-w
a
d
9 )
] [
U )
B
e
t
SI
fl
se
h-w
3
n
t)
s
e
t
fl
s
-\
h
■5!
au
1
a
0.22«
a
.11
u
.15
a
.21
i
J
.33
e
n
.19
I
.06
a
1
.66
i
.24
.24
.10
.31
.14
.57
3.94
.19
.43
.16
.21
.10
.30
.10
.45
.46
.16
1.28
0 22
.04
i
z
a
f
I'l
i
"1 i
hw J
e
3
.22
.02
1.33
1
8
.14
e
.22
t
.07
SI
.25
g
.02
1
.11
se
.35
s
2.23
a
i
.25
s
.11
e 1
i i
.25
.05
.16
.04
.13
.10
.33
.10
.29
.10
1.90
.16
.13
.14
.22
.45
3.07
.49
2.06
.39
.70
.80
.26
.23
fl
■w
u
d
n
LI
hshs
;i
e
1
i
i|
.14
.12
.09
.24
1.97
.30
.07
.11
.03
.16
.11
.13
.06
.11
.06
.24
.34
.07
.93
.32 (?)
.09(?)
.30
.13
1.14
.32
3.56
.29
z
.09
u
.24
g
.02
u
.16
.03
e
.30
1
.07
?l
.97
3
.16
n
.16
]
U3
.14
f
.12
ee
.23
m
.07
3
.06
1
.08
i
.13
-1
1.64
ae
.01
m)
.12
e
.16
5
.05
e
.31
0 )
U [
1.76
i )
V
.10
1
.18
o
.62
'' )
I'
.34
n
p
.10
r
a
.30
s
1
p
i
0
1
DURATION 491
to indicate the sounds in order to aid in marking off their
duration ; it is not intended as an accurate phonetic analysis.
For example, the use of a for the stressed vowel in ' what '
does not necessarily mean that the sound is the same as the
a in 'come'-; to the ear the brief vowel in this case seems
related to a, o, and a, but it is hardly possible to decide on
the degrees of likeness.
General averages of the lengths of the vowels have been
made for Swedish and Finnish by Pipping (references on p.
20).
The graphic methods described in Part III are also used for
measurements of duration. A voice-key (Fig. 66) attached
to a Depeez marker (Fig. 61) can be used to register the
lengths of vowels.
By means of records from the larynx (p. 267) and the nose
(p. 219) RoussELOT was able to establish the following facts
for his own speech ^ :
1. In isolated words the explosive consonants are slightly
shorter than the fricatives (a pap a, afafa, atata,
8 9 12 15 9 13
asasa, asasa); the sonants are often shorter than the surds
15 15 15 16
(o°fo°, o°vo", o°to°, oMo°, etc.); the lengths of the consonants
18 15 14 13
generally diminish with increased length of the word; the
last consonant in a word is generally lengthened ; ' double
consonants ' are really long and strong single consonants ;
consonants in groups tend to occupy less time than the sum
for single ones. The numerals in the above examples in-
dicate hundredths of a second.
2. In records such as d, 6, o, dp, op, op, po, p6, po, pdp,
p6p, pop for the vowels a, e, i, o, u, y, oe, the open ( ' ) and
close (') vowels were regularly somewhat longer than the
medium ones (unmarked).
' The first lessons I received in grammar taught me to con-
fuse quantity with timbre, and this confusion persists in my
1 RonssELOT, Les modifications phonitiques du langage, 81, Rev. des pat. gallo-
rora., 1891 IV, V; also separate.
492 FACTORS OF SPEECH
present appreciation of the vowels in my dialect. I perceive
as long all open or close vowels, as short all medium vowels.
That the appreciation contains exaggeration is shown by my
measurements. But it is correct for isolated vowels.' ^
3. In groups of two syllables the last vowel of the group
is almost always the longer one; the difference between
long and short vowels is often rendered almost imperceptible
by an approximation of both to a medium length.
4. In groups of three or four syllables with one vowel
checked (papatpa) as compared with all vowels free (pa-
papa), the checking tends to abbreviate the vowel.
5. Groups of syllables show a decidedly iambic character,
the principal forms being u_,u , _u , w u ,
,_,_u ,w u ,_u w ) where ^ indicates very
short, ^J short, _ long, very long. The end of a group of
syllables is always an iambus ; the first part varies between
iambic and trochaic.
6. In groups of two syllables with the same vowel, both
stressed and unstressed vowels are long when open or close,
and short when medium.
7. In groups of syllables with the same vowel, the vowels
perceived as short are often longer than those perceived as
long.
8. A vowel naturally short is strongly abbreviated when
followed by a long vowel (kute, ' couteaux,' 12, 19); when
both vowels are naturally short or long the stressed vowel is
the longer (mati ' matin,' 7, 13; otur, ' autour,' 6, 14).
9. The vowels, except a, are longer after f, v, s, m than
after b, p ; after z than after g ; after v, z or z than after f, s
or s ; after p or k than after b or g ; after k or g than after p
or b ; after n than after n. The vowel a is shorter after m
than after p, b ; after v than after f ; after p than after b ;
after g, k than after b, p. The vowel a is longer before
sonants than before surds; before f, v than before p, b; be-
fore gutturals than before dentals ; before dentals than before
labials.
' PoussELOT, as before, 88.
DURATION 493
10. Diphthongs are shorter than the sum of the two sep-
arate vowels supposed to compose them ; a stressed diphthong
equals the two unstressed separate component vowels.
The above relations appear clearly in connected speech.
The durations in hundredths of a second were calculated
from experimental records (p. 359). In the following
illustration of connected speech- the lengths are marked
beneath the vowels in hundredths of a second.
a"ta"tyT sa"t akoe kuk y
13 6 9 11 10 U 14 6 15 6 18 6 9 8 10
ta ekut sa"t oe so" su
8 9 6 10 15 16 15 14 15 16 15
' Entends-tu chanter ce coucou ?
Ta! dcoute! chante-t-il son saoul! '
RoussELOT finds in the records of connected speech that
final atonies scarcely exist in his case except in plural nouns
and in verbs in the second person. In only two cases did a
final 9 appear; these were under exceptional conditions.
In respect to the duration of different sounds in language
the ear gives little correct information, as is shown by the
graphic records. ' Vowels which I believed always long were
often short; others that I regarded only as short often sur-
passed in length those I considered as long. ' ^
Experiments by Binet and Hbnei^ with Rousselot's
microphone registering apparatus (p. 267) showed that in
speaking a series of numerals the sound before a pause is
lengthened, this being supposed to indicate that it is easier
to change from one motor condition of articulation to an-
other than to end a motor condition. Increase in velocity of
speech shortened the pauses rather than the sounds. Stress
lengthened a sound.
Using a mouthpiece connected to a Maeey tambour to
1 RocssELOT, as before, 76.
2 BiNET ET Henri, Les actions d'arret dans les phenomines de la parole, Rev.
philos., 1894 XXXVII 608.
494 FACTORS OF SPEECH
register breath impulses in French sounds, Wagner ^ found
that long vowels occur only in stressed syllables that end
with a consonant ; that all such vowels are long when followed
by a sonant fricative or ] ; that close o, velar a, and all the
nasal vowels are long before a consonant; that the other
vowels, i, y, u, open e, ce, o, and palatal a, are generally
short when stressed and followed by a consonant that is not
a sonant fricative or j, but are generally half-short within a
sound group ; that consonants before a pause are the longest
ones ; that such consonants are not influenced by the length
of the preceding vowel.
The variations in the duration of French syllables in dif-
ferent phonetic combinations have been investigated by
Geegoire.2 a mouthpiece connected to a Maeby tambour
recorded the breath explosions and the cord vibrations.
Records were made on comparative groups of the follow-
ing types : ' pate : pSteuse, pSteuse : p&te, pate : pateuse : pate,
pateuse : p§.te : pateuse ; ' they showed that the monosyllable
with a long vowel is always longer when pronounced alone
than when it enters into a dissyllabic compound ending
in another long syllable. Examples may be seen in the
records of
pate
33
: pateuse
2!
: pate
35
pateuse
: pate :
pateuse
20
33
20
tgtard
: tete :
tetard
21
36
20
tete
: tetard :
tete
37
23
39
il tate :
ils taterent
: il tate
34
iT
34
The figures give the number of hundredths of a second occu-
pied in each case by the syllable pd,, U or tS,. The ordinary
1 Wagner, Franzosische Quantitdt, Phonet. Stud., 1892 "VI 1.
'^ Gregoike, Variations de dur€e de la syltabe frangaise, La Parole, 1899 I
161, 263,418.
DURATION 495
lelation of length between the monosyllabic fragment and
the same fragment in a dissyllable ending with a long syllable
is about 1:0.6 with some considerable variations. Thus a
comparison of x^- : x^-y- shows that x^ < x^.
When the fragment appears in a dissyllabic compound
ending in a short syllable, the relation is approximately the
same. Examples are
pate :
pateux
: pate
36
27
38
tete
: t8tu :
tete
38 24 36
Thus x^- : x^-y gives also x^ <x.^^.
The duration of a monosyllable remains in general the
same when it is made the final syllable of a polysyllabic
word ; for example,
prgte :
interprete
: prgte
^
40
"sf
interprete
; prete :
interprete
40
40
40
Thus ajj : V . . . yx^j gives x^ = Xy Before a suffix added to
the polysyllable the fragment is shortened just as in the case
of a monosyllable.
In comparing similar words of the types - - and - " it was
found that in the majority of cases there was a slight length-
ening of the first syllable in the trochee. In the following
example the first of each set of figures gives the time of the
syllable and the second the time of the occlusion of the t.
p& teuse : pS, teux : pa teuse
21 15 26 19 23 17
Thus x^-y^ : x^-y^' gives x^ > x^.
Ge^GOIEE calls attention to the relatively long durations
for the occlusions (not the explosions) of the occlusives. In
a dissyllable the occlusion (not the explosion) comprised in
the second consonant is often nearly as long as the whole
496 FACTORS OF SPEECH
first syllable. The shortest occlusions measured from 0.10'
to 0.18=; many reached 0.18' to 0.20' — rather large figures
in comparison with 0.30' to 0.32' for the preceding vowels.
The occlusions are brief moments of silence that are not
noticed as such. Even to the ear a word like ' p8.teux '
seems to be one continuous flow of sound, probably because
the motor activity of the speech organs goes on just as vigor-
ously during the occlusion as before and after, and by associa-
tion unifies the auditory parts in the hearer. If we insist
upon dividing words into syllables, to which syllable does
the occlusion belong? For the ear Geegoiee believes the
following syllable to begin with the explosion while the
occlusion belongs to the preceding one. In speaking at un-
intentionally different rates the duration of a syllable is
increased by lengthening sometimes the vowel alone, some-
times the occlusion alone and sometimes both.
When used in a polysyllable, a long syllable is shortened
as the number of following syllables increases. Thus, the
syllable ' pon ' in ' pontife ' becomes shorter in ' pontificat '
and still shorter in ' pontificalement. '
For a short syllable analogous results were found in all the
preceding cases that have been .discussed.
. In groups of words in ordinary speech a monosyllabic
fragment alters the proportion of its vowel to the following
occlusion according as it is followed by a vowel or by a con-
sonant, although the total of vowel -f- occlusion may remain
the same. A typical result was
p8. te sucree pa te et de creme
18 14 25 9
In this case the complication of the group ts for some reason
shortens the vowel and lengthens the occlusion.
In discourse the union by contiguity shows its effect chiefly
in shortening the monosyllables.
Geegoiee also shows that a final syllable of a word
becomes longer before a pause.
Geeigoiee's explanation for the decrease in length of a
DUPiA TION 497
syllabic fragment according as it occurs in a monosyllable, a
dissyllable or a polysyllable is that a portion of the accent of
the monosyllable is transferred to the other syllables in com-
binations and that the ' language, ' foreseeing the effort re-
quired at the end of the word, shortens the earlier portion.
Records of the breath curve (p. 219) by Vietoe ^ for his
own speech (Nassau, Germany) gave as typical lengths:
extra-short vowels in Kamm, wart' (0.08'); short in Pappe,
Gwttenberg, da! (0.15); half-short in Sckretar (0.19, 0.18),
mit, mitteilen, Packkorb, TawfEeier (0.20); half-long in
Unthaten, M«tra (0.25); long in Thaten, GMtenberg, thwst,
Unthat, bi'eten, Bawmmeise, Mitra, Pape (0.30); extra-long
in Sekretar (0.35), That (0.87), war't, thwt, Fraw, Maid,
frew', thw, ivei, kam (to 0.42), do (0.45). From these and
similar results the following conclusions were drawn. A.
The so-called ' long ' vowels are (I) extra-long when tonic in
the last syllable and followed by no consonant, a single conso-
nant, or a liquid and a consonant; (II) long when tonic in
the last syllable before several consonants, .or in the next-to-
last or some preceding open syllable (before a single con-
sonant), and also long when having post-tonic secondary
stress when the tonic vowel is extra-long; (III) half-long
when tonic in the next-to-last syllable followed by several
consonants, or when having post-tonic secondary stress
when the tonic vowel is long; (IV) half -short when hav-
ing pre-tonic secondary stress. B. The so-called 'short'
vowels are (I) half-short when tonic in the last syllable
before a single occlusive, and in the next-to-last syllable
before a double or long occlusive ; (II) short when tonic in
the last syllable with no following consonant, and in the
next-to-last or some previous closed syllable; likewise un-
stressed e = 9 ; (III) extra-short when stressed in the last
syllable before a liquid or before a liquid and a consonant.
O. The diphthongs show the same, relations as the long
vowels.
1 ViETOK, Elemente d. Phonetik, 4. Aufl., 269, Leipzig, 1898.
32
498 FACTORS OF SPEECH
Similar records ^ for an Englishman from Sydney (Aus-
tralia) living in Marburg (Germany) showed on one occa-
sion half-long vowels in bite, pate (0.25^); long in pat
(0.30); extra-long in part (0.41), paid (0.45), bide (0.49),
pad (0.55), pard (0.61); on another occasion short in beat
(0.15^); half-short in pat, goddess (0.20); half-long in bead,
gawdy (0.25); long in pate, bite (0.30); extra -long in pad
(0.35), part (0.86), god, gawk, gawd (0.42), paid (0.45), pard,
bide (0.51); on a third occasion: short in pate, bite (0.15),
half -long in paid, pad (0.25) ; extra-long in bide (0.86). The
relations between the lengths of the vowels before the sonant
and surd explosives were pad : pat :: 1.7:1; pard : part ::
1.5 : 1, or 1.4 : 1; bead : beat :: 1.4 : 1; paid : pate :: 1.7
: 1, or 1.4 : 1, or 1.9 : 1; bide : bite :: 1.9 : 1, or 1.6 : 1, or
2.1 : 1. The influence of a sufQx was shown in god : god-
dess :: 1.9 : 1; gaud. : gawdy :: 1.7 : 1. Great variations
occurred in the relations of such pairs as paid : pad :: 1 :
1.2, or 1.3 : 1, or 1 : 1; pate : pat :: 1 : 1.2, or 1.6 : 1; bead
: bid :: 1 : 1.5, the ' long' vowel being sometimes shorter
than the ' short ' vowel and likewise the reverse.
Records ^ from a native of Lidge (Belgium) living in Mar-
burg gave: diwde (0.50=), towt, doitx, Aind.on (0.27 to 0.31),
viens-tM? (0.19, 0.20), Veaucoule (0.16, 0.30).
Records by Vietor also gave: mit (0.10=), Ka.mm (0.30),
kawi (0.29), Hohenheim (0.12), Kawwer (0.12), kamen
(0.14), Baumeister (0.15), Bau??i??ieise (0.33), Tau^eier
(0.38), Bau/eier (0.24), Schi/e (0.30), Scha/e (0.23),
thusf (0.18), Baumeis^er (0.05), wart', war't (0.14), ^S'ekretar
(0.17), Baummeise (0.17). Consonants may be short or
long ; a consonant need not be longer after a short vowel.
Meyer's measurements ^ of Wagner's records * of the
Swabian dialect showed that a consonant was lengthened
1 Vietor, as before, 271.
2 Vietor, as before, 273.
8 Meyer, konzo'nantndaoer "^im doiftfn, Maitre phone'tique, 1901 XVI 114.
* Wagner, Der gegenwdrtige Lauthestand dts Schwabischen in d. Mundart
von Reutlirujen, Program, Reutlingen, 1889-91.
DURATION 499
after a short vowel ; the ratio of consonantal length after a
long vowel to that after a short one was for final surd
occlusives 1 : 1.36, for final surd fricatives 1 : 1.20, for final
nasals 1 : 1.13; and for medial surd occlusives 1 : 1.15.
Records i of the Italian dialect of Bari (Apulia) showed
that the length of t after a long vowel (as in fata) bore
the relation to that after a short one (as in fata) of 1 : 2.07.
One similar case in Finnish showed a relation of 1 : 3. 39
for k.
Kkal and Mares, employing a telephone whose vibra-
tions electrically stimulated a nerve-muscle preparation (p.
188) arranged to write on a smoked drum, were able to
record the lengths of various sounds.^
In Bohemian the first syllable of a word always receives
the stress ; ' long ' and ' short ' sounds and syllables are dis-
tinguished; the rhythm of verse is produced by the word-
stress and not by the length of the syllables. The results of
the investigation showed the following facts :
1. Even for the same person a vowel receives different
lengths according as it is spoken with more or less stress ;
2. The 'long' vowels are on the average a little — ^but
only a little — longer than the average short vowels, short
and long often occupying nearly the same time. The differ-
ence between ' long ' and ' short ' Bohemian vowels seems to
lie rather in the mode of expending the breath; in a short
vowel the beginning is strong and the decrease sudden; in
a long vowel the stress is more nearly even; 'legato' and
' staccato ' more appropriately express the two forms than
' long ' and ' short. '
3. Diphthongs have about the same length as the vowels,
ou being a little longer.
4. Consonants are very short. With a group of conso-
nants the length of a syllable does not necessarily increase;
1 Meyer, as before, 115.
2 KrAl a Mares, Trvdni hldsek a slahik die ohjehtivng miry, Listy Filolo-
gicke', 1893 IV 17; the explanations and summary I owe to Prof. Frantisek
Czada of the Bohemian University in Frag.
500 FACTORS OF SPEECH
syllables ending in t are usually shorter than those with the t
omitted, thus akt may be shorter than ak.
5. In scanning verses with the greatest possible evenness
the various ' feet ' continually vary ; the thesis and arsis por-
tions never have the relation 1 : 1 but approximately 31 : 30,
33 : 32, etc.
Josselyn's records 1 of Italian speech showed that a
double consonant was simply a single consonant strength-
ened and lengthened (p. 332); its effect on the preceding
vowel was to deprive it of about a third or a half of its
length. Examples, in hundredths of a second, are
c
a
P 0
fata
c
a
d e
1 e
g a
26
16
24 26
20
20
20
19
c
a
PP 0
f a tt a
c
a
dd i
1 6
gg a
14
29
10 37
9
36
12
32
When the portion of speech to be measured is a long one,
less accurate but less laborious methods may be employed.
To obtain the duration of a sentence or a larger unit the
method analogous to one familiar to navigators and astron-
omers may be conveniently employed. The recorder watches
the face of the speaker who sits beside a clock having
a seconds-hand. At the first word he turns his gaze as
quickly as possible to the clock and notices the position of
the hand; at the last word he again notices the position.
By practice this method can be made accurate to ±l^
With almost as great accuracy the recorder can observe the
face of a watch while listening to the sounds.
Experiments by Bourdon,^ presumably made by observ-
ing a watch, show the average ordinary length of a syllable
in reading French to vary from 0.184' to 0.234'. The
length varies somewhat with the emotion contained in the
selection, with the temperament of the reader, etc.
1 JossELYN, £tude sur la phonet. ital., 146, These, Paris, 1900; also in La
Parole, 1901 III 226.
2 Bourdon, L'Expression des emotions et des tendances dans le langage,
Paris, 1892.
DURATION 501
The reaction method is an improvement on the preceding
one. The recorder makes a movement of the finger at the
moment the speech unit begins and another at the moment it
ends. His time of response, or reaction time (p. 206), can
generally be relied upon to be constant within 0.1» at both
beginning and end ; a record made in this way thus gives the
true time. The time between the two movements may be
registered by a stop-watch, by a chronoscope, or by some form
of the graphic method (pp. 152, 206).
The phonograph has also been used to measure the inter-
vals between points of stress (centroids) in prose and verse of
different kinds. This is accomplished by adding a contact
wheel to the axle and recording electrically the number of
contacts passed over between one point and another in
the speech record.
Measurements, of feet, lines, etc., in verse by means of
these methods will be considered in the chapter on rhythm.
Among the phenomena found in investigations of mental
time-estimates we may notice 1. that judgments of the
equality of two successive tones or empty intervals are more
irregular with some persons than with others; 2. that with
some persons or under some circumstances the first tone
seems longer, with or under others shorter; 3. that by mak-
ing a tone louder it can be made shorter without any appar-
ent decrease in length; 4. that similar results are found
for empty intervals marked off by sharp clicks; 5. that the
apparent length of an interval bounded by clicks is made
greater by inserting intermediate clicks. These results have
important phonetic applications. Further investigation is
needed of the manner in which loudness or change in pitch
can be used to replace duration.
The experiments on speech sounds have made it clear that
at best the terms ' long ' and ' short ' for the vowels or syl-
lables of a word can mean no more than that they are on an
average long and short. The assignment of definite relations
of average length as two moras or one mora likewise means
only that on a general average the relation will be more like
502 FACTORS OF SPEECH
2 : 1 than any other simple expression. The actual averages
for any speaker or for any language will be found to differ
from such a relation. The relations of duration in any par-
ticular word may be entirely reversed, the ' long ' vowel occu-
pying perhaps less time than the ' short ' one. These terms
express really a total mental effect that the hearer or speaker
feels — or is taught to feel — to be one of duration. A
•stronger or more difficult sound may produce a greater
impression ; a person is readily taught to lump the whole as
a matter of duration and consequently supposes himself to
hear the sound as actually longer while it may be only stronger
or higher. The illusion is a familiar one to psychologists.
Such an illusion of considering staccato vowels as short and
legato ones as long, while they are nearly equal in duration,
has been shown (p. 499) to be the regular thing for Bohe-
mian. In fact, the terms ' long ' and ' short ' are really terms
for certain mental impressions that might more safely be
expressed by some such general forms as ' of the first class '
and ' of the second class, ' or ' prime ' and ' double prime, ' etc.
' Strong ' and ' weak ' (referring to the auditory impressive-
ness and not to loudness) might be good substitutes. The
amount of error that has entered into the teaching of lan-
guages and the discussion of verse by supposing that ' long '
vowels are anything more than auditorily strong ones is
Tery great.
Refeeekces
For a general treatment of the length of speech sounds: Sievers,
•Grundziige d. Phonetik, 5. Aufl., 254, Leipzig, 1901. For duration in
Terse : see Ch. XXXV (below). For a, sketch of experiments and
literature on time-estimates: Wundt, Grundziige d. physiol. Psychol., 4.
Aufl., II 408, Leipzig, 1893; Scripture, New Psychology, Ch. X,
London, 1897.
CHAPTER XXXIV
LOUDNESS
Under ' loudness ' we may understand that property in
which a sound (p. 109) may vary while its pitch and dura-
tion remain unchanged. The term ' stress ' may be used to
express the motor characteristic.
Direct measurements of the loudness of speech sounds are
still impracticable. The loudness of a sound increases —
other properties being constant — with the amplitude of the
vibration, but not proportionately ; the physical energy in-
creases as the square of the amplitude, but the mental loud-
ness follows some quite different law (p. 109).
The chief factor of the motor stress, the breath pressure,
may be registered by any of the ways described in Ch. XVI.
In Rousselot's ^ experiments the rather weak French stress
was found not to have an absolutely fixed place ; the stress-
accent may occupy the last syllable of a group, and it
scarcely leaves this place in energetic pronunciation, although
it tends to change to the next to the last syllable in phrases
spoken in a soft or caressing tone or in phrases forming a
conclusion.
The lungs practically always contain more than enough air
for a phrase of speech ; the breath is not renewed before each
■stressed syllable, but only for whole groups of them. These
facts are sufficient to dispose of the supposition ^ that the
weakening of unaccented final syllables in French is due to
1 RoussELOT, Les modifications phonetiques du langage, 71, Rev. d. pat. gallo-
Tom., 1 893 IV, V ; also separate.
2 Marcou, Influence of the weakness of accent-stress on phonetic change in
French, Publ. Mod. Lang. Assoc, 1890 V. 47.
504 FACTORS OF SPEECH
the failure of breath; this weakening is rather to be referred
to some mental preference.^
In. the records studied I have rarely found a vowel with a
constant amplitude. Vowels at the beginnings of words show
invariably a rise in amplitude. This rise may continue until
the vowel ends in some other sound. Such is the case in a
of ai, and in ae of ' and ' in ' thread and needle. ' Most
independent vowels rise to a maximum and then fall. Such
vowels might possibly be called vowels of circumflex stress.
Even in the middle of the word the vowel has a tendency to
the circumflex form, as is well shown in most cases of the i
of ai. The rise and fall may be quite elaborate as in the
case of the triple-circumflex vowel o of ' bow ' (Plate I);
this long o, however, might with propriety be considered a
molecular union of three o's in succession. These changes in
amplitude in the vowels appear clearly in an inspection of the
Plates of speech curves at the end of this volume ; measure-
ments of some are given in Appendix II.
In a vowel of constant amplitude represented by the sinu-
soidal vibration (p. 2) we would have F(f) = a and
. 2Trt
,^«sm^^.
In a rising vowel -F(i) might take some such form as mt,
which expresses a proportional increase. We would then
have
In a circumflex vowel we may assume the amplitude to be
of sinusoid form whereby
F(_t) ^ E sin
2-ITt
and
„ . 27ri . i-jrt
y — L sm. . . sin
/(O
1 WoNDT, Volkerpsychologie, I ii 278, Leipzig, 1900; Oertel, Lectures on
the Study of Language, 318, New York, 1901.
LOUDNESS 505
where E would be the maximum amplitude and s the length
of the vowel. When the pitch is constant the curve will
have the form
„ . 27ri . 27r(
y = E sin . sin
T
I have found one vowel, e in ' said ' in the line ' I, said the
sparrow,' that can be considered as a circumflex vowel of
approximately constant pitch. The equation of the curve
traced from its record (p. 58) is (in seconds and millimeters)
y-^-^ ^^^-0l08--^^^():0053-
It does not fill a complete period of circumflexion as' it is
suddenly cut short by the d.
Among the hundred or so English vowels that I have in-
spected, I have been unable to find one that can with any
close approximation be considered as steady in intensity and
constant in pitch. A vowel of the form y = a sin —^
must be a rare one. Some vowels during part of their
course are of this form, but a change of some kind seems
characteristic at some moment. Even such approximations
have been found only in the interior of words, that is, with
boundaries of consonants or of vowels with the vocal organs
already in action. It seems to be the rule in English that a
vowel following a pause shall be a rising or crescendo one,
and one preceding a pause shall be a falling or diminuendo
one.
CHAPTER XXXV
ACCENT
Great unclearness prevails in the usual discussions of
accent on account of the confusion among physical, psycho-
logical and physiological terms.
Physically the three properties of a tone to be considered
should be the duration, the pitch and the energy CP- 89) ;
the last depends on the pitch and the amplitude (p. 109),
and for many comparisons where a proportional scale of
energy is not well obtainable the amplitude may be used.
Mentally the case is utterly different; there are two asso-
ciated factors of accent: auditory and motor. The one prop-
erty that characterizes auditory accent is ' impressiveness ; '
this may arise from increase in loudness but also from de-
crease, from rise in pitch but also from fall, from lengthening
of the duration but also from diminution — in short from any
change that produces a mental effect. Motor accent repre-
sents volitional work. More work may arise 1. from a longer
time of work ; 2. from greater effort, as in increasing the breath
pressure, or as in intentionally decreasing it, or in increasing
the tension of the vocal cords, or in stronger muscular
movements of any of the other speech organs; 3. from in-
creased complexity of effort as in less familiar sounds, etc.
Physiologically the variable properties are in expenditure
of muscle and nerve energy. The work of a muscle is some-
thing quite different from the work it performs in displac-
ing bodies; a large amount of the energy is expended in
heat, etc. The phenomenon of the breaking down of nerve
compounds is still little understood. A sudden relaxation
ACCENT 507
in breath pressure reduces the physiological work of the
muscles, but it may represent increased volitional work and
correspond to a very emphatic mental effort. A rise of
the larynx tone represents a steady increase of muscu-
lar work and innervation, but it may — on account of its
gradual character — fail to produce any auditory or motor
emphasis, or may even produce the impression of decreasing
emphasis on account of lack of the change that is required
by the mind.
The term ' accent ' may profitably be restricted to its
psychological meaning; an accented sound is thus one that
impresses the hearer more strongly or that requires more
mental effort on the part of the speaker.
The first duty of a study of accent is to determine what
mental elements are involved.
The main features of accent have been outlined by SiE-
VEES,^ who has discussed the pitch of a sound, the pitch
intervals between successive sounds, the rise and fall of pitch
within a sound, the tremolo, the increase in loudness, the
change in loudness, etc. Duration is considered by Sievees
in a special section on ' Quantity ' separate from ' Accent.'
It will be noticed that the point of view is mainly auditory
and that as an outline of auditory accent the sketch is practi-
cally complete. Motor accent requires a consideration of the
factors of volition involved ; these would also find expression
in changes of pitch, intensity, etc., but in relations different
from those of the auditory factors; the factor of volitional
work would need to be added as a special one. The various
treatments of accent rest upon judgments by the unaided ear.
It is unquestionably the fact that here, as in all the senses
without exception, attempts to specify anything beyond the
general outline just given can result only in a statement of
illusions. In a judgment of impressiveness the ear is unable
to distinguish with any accuracy, except in extreme cases, the
factors of pitch, loudness and length; accents stated to be
<iue to increased stress may often be due to changes in pitch
1 SlEVEKS, Grundzuge d. Phonetik, 5. Aufl., 215, Leipzig, 1901.
608 FACTORS OF SPEECH
without any possibility of a detection of the fact by the ear.
This diificulty of separating the two factors of loudness and
change in pitch has been strongly emphasized by Sievbrs.
The inadequateness of a treatment of accent on the basis of
a classification into accented and unaccented syllables has
been emphasized by various writers.^ Accent, we may say,
is a continuous property that runs with the flow of speech.
So little experimental work has been done on accent that a
general treatment of the subject beyond the outline already
given would be out of place here ; I shall confine the remain-
der of the chapter to a summary of the disconnected experi-
mental results with no attempt to work them into a theory.
Pipping ^ distinguishes three factors of accent ' in addition
to psychological phenomena ' : the energy of articulation, the
energy of the sound waves, the intensity of the auditory sen-
sations. His calculations of the average physical energy of
the sound waves (intensity, p. 109) give the following
relations :
s a
5660
The average pitch of each vowel is also indicated in five cases ;
in mylyn the first vowel was the higher. The general re-
lation between musical accent and physical intensity appears
clearly.
In an investigation by Squike,^ records were made by
a RoussELOT voice-key (p. 154) of the repetition of the
syllable mi by school-children. When the children were
asked to emphasize the first syllable and then every alternate
one (trochaic rhythm), the records showed the emphatic
syllable usually longer than the unemphatic one, the relation
varying from approximately 87 : 31 to 1 : 1 ; the pauses varied
from 6 : 35 to 41 : 44.
1 HiKT, Der indogermanische Akzent, 11, Strassburg, 1895.
^ Pipping, Zur Phonetik d. finnischen Sprache ; Untersuchungen mit Hensen's
Sprachzeichner, 227, Mem. de la Societe finno-ougrienne, XIV, Helsingfors, 1899.
8 Squire, A genetic study of rhythm, Amer. Jour. Psychol., 1901 XIII 492.
t a m
a
sad
a n
m
y
1
y
4011
1460
15649
2243
672
474
/
<P
gOt
cO«
■ACCENT 509
When the second syllable and then every alternate one were
emphasized (iambic rhythm), the relations of length varied
from 61 : 69 to 56 : 36 ; the pauses varied from 27 : 74 to
41 : 35. When the first syllable and every third following one
were emphasized (anapestic rhythm), the general type of rela-
tions of lengths of the syllables was for one subject 67 : 61 : 44,
for another 37 : 39 : 81 and for a third 41 : 30 : 83, with pauses
in the typical relations of 17 : 16 : 84, 17 : 18 : 29 and 44 : 50 :
52 respectively ; the variations were, however, considerable.
When the third and every third following syllable were
emphasized (anapestic rhythm), characteristic relations of
length were for one subject 83 : 28 : 46, for another 39 : 33 :
28, and for a third 31 : 29 : 47, with pause relations of 10 : 11 :
35, 13 : 13 : 52, and 32 : 41 : 58 respectively.
Fig. 336.
Experiments on the relations between stress and duration
and between stress and pitch have been made by Miyakb.i
The voice key (p. 154) was connected to a Dbpeez marker
(p. 92). This might have been done by any appropriate
battery arrangement; it seemed most convenient to connect
the key to one of the sockets of a four-socket lamp battery,
while the marker was connected to the other so that the
key made a high-tension shunt around the marker (p. 210,
Fig. 81, key to socket 0, marker to socket (7). Any other
method of recording voice vibrations might have been used.
1 MiTAKE, Researches on rhythmic action. Stud. Yale Psych. Lab., 1902 X 22.
510 FACTORS OF SPEECH
The subject repeated the sound a continuously in what he
felt to be a trochaic rhythm (thus, a' a a' a a' a . . .J, or an
iambic (a a' a a' a a' . . .), or a dactylic (a' a a a' a a . . .),
or an amphibrachic one (a a' a a a' a . . .). The marker
registered the vibrations on the smoked drum; a specimen
record with time line in 0.01° at the bottom is shown in
Fig. 336.
The following summary is from Miyake's tables; the let-
ters indicate different persons; each separate record is the
average of a separate experiment of ten measurements.
Subject. Ratio a' : a. Ratio a : a'. Ratios a' : a : a. Ratios a : a' : a,
0.92.1.00 1.00.0.82,0.81 0.83:1.00:0.79
0.87 : 1.00 1.00 0.87 0.77 0.99 : 1.00 ; 0.93
0.92 : 1.00 : 0.92
0.85:1.00 1.00 : 0.76 • 0.63 0.90:1.00:0.92
0.90:1.00 1.00 0.89:0.78 0.92.1.00.0.91
„S 1.00:0 61 0 69:1.00
°l 1.00 : 0.68 0.80 : 1.00
From the table we may generalize as follows: 1. a
stressed syllable was always longer than an unstressed one;
2. the relation between the two varied with different individ-
uals and on different occasions; 3. in a dactyl the first
unstressed syllable was slightly longer than the last one.
Since each vibration was indicated, the records used in the
preceding case were also available for determining the pitch
of the cord tone in the stressed and unstressed syllables.
For example, in an experiment on a' a with subject K, the
successive periods of a were : 7, 7, 7, 7, 7, 6, 6, 6, 6, 6, 6, 6,
6, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 4, 4, 4, 4, 4". The successive periods of a were :
8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7-. Or
• more briefly a': 1" (5 times), 6" (8 times), 5" (16 times), 4" (38
times); a: 8"^ (11 times), 1" (82 times). [" = 0.001'.]
ACCENT 511
The changes in the lengths of the periods were not sudden
but very gradual in this record as well as in all other records.
The accented syllable began with a period of 1" and changed
gradually through 6°- and 5" to 4°-, which was reached at the
30th vibration and maintained to the end. That is, the
pitch rose from 143 complete vibrations through 170 and 200
to 250 per second.
The unaccented syllable, on the other hand, began with a
period of 8"^, and reached 7<^ at the 12th vibration, which was
kept to the end. That is, the pitch changed from 125
upward to 143 complete vibrations a second, and remained
fixed thereafter.
The results of various experiments are shown in Plate XIV
at the end of this volume; the specific measurements are
given in the original monograph. The horizontal axis indi-
cates time, while the vertical ordinates give the numbers of
cord vibrations a second. The space between the curves
corresponds to the empty interval between the syllables.
An inspection of the Plate is sufiScient to show 1. that
the accented syllable had a higher pitch than the unaccented
syllable; 2. that the accented syllable began in general with
a higher pitch than the unaccented syllable ; 3. that even in
the cases where both accented and unaccented syllables began
with the same pitch, the former glided upward higher than
the latter ; 4. that the pitch of the accented syllable under-
went greater changes than that of the unaccented one; 5. that
the pitch of the accented syllable always glided upward; 6.
that the pitch of the unaccented syllable also glided upward
in the majority of cases, but sometimes glided downward.
MiYAKB makes the following observations :
' According to Mitford,i the strengthened syllables in
English have an acuter tone or a higher note. The fact can
be abundantly proved, he supposed, if we find or coin a word
which is composed of syllables without variety of vowel sound
and pronounce it with a strong accent on either syllable.
1 MiTFORD, Inquiry into the Principle of Harmony in Language and of the
Mechanism of Verse, Modern and Ancient, 57, London, 1804.
512 FACTORS OF SPEECH
' MiTFORD supposed that when we pronounce an accented
syllable, we raise the tongue near to the palate, with the con-
sequence of the rise of the height of tone. "To produce the
proper English intonation," he says, "the tongue must be
raised up in pronouncing the strengthened syllable, the vibra-
tion will be felt more about the palate and the tone will be
acuter, it will be a higher note." The change of the position
of tongue in the mouth cavity would only affect the resonance
tone and not the cord vibration. It thus gives no explana-
tion of the fact.
' MuLTiER 1 noticed that in a larynx separated from the
body the pitch of the tone might be raised by an increase of
the force of blast. He thought that one of the modes of pro-
ducing high notes without increasing the tension of the
vocal ligament is to blow with greater force, by which means
the notes may without difficulty be raised through a series of
semi -tones to the extent of a " fifth. "
' Brucke ^ supposed that in strong accentuation the vocal
cords are more stretched on account of the strong pressure of
the air and come closer to each other and that, as a conse-
quence of the increase of the tension of the cords, the pitch
of the tone is raised. Scripture ^ thinks that the relation
between the rise of pitch of the cord tone and the increase
in the force of the puff would naturally result from a gradual
tightening of the vocal muscles which is due to associated
habits of innervation and not to the physical effect of the
air pressure in stretching the cords. '
Muller's supposition concerning the relation between
pitch and intensity arose from the fact that a thin membrane
under a constant tension gives a higher tone as the pressure
of the blast is increased. The action of the cords as now
known (p. 263) indicates that the two properties are not
1 MoLLER, The Physiology of the Senses, Voice and Muscular Motion, with
the Mental Faculty, trans, by Baily, London, 1848.
^ Bkucke, Die physiologischen Grundlagen der neuhochdeutschen Verskunst,
3, Wien, 1871.
' Scripture, Nature of vowels, Amer. Jour. Sci., 1901 XI 302.
ACCENT 513
physically interdependent to any notable extent. A rise in
pitch is the result of increased tension in the vocal cords,
which is probably brought about entirely by contraction of
the cricothyroid muscle and not by breath -pressure on the
vocal bands. A loud, high tone represents thus not only
increased respiration work but also increased larynx work.
It is safe to say that, in English at least, increase in dura-
tion and rise in pitch are ordinarily associated with increased
stress, and that these associations are essentially mental ones
and not interdependent physical or physiological phenomena.
This relation seems to hold good in general for initial
vowels ; it may be seen in many of the vowels in the Jeffer-
son records (Plates III to XI). In other cases a change
in pitch does not go with a change in intensity ; the intensity
may even decrease while the pitch remains constant. To
keep a constant pitch with a changing air pressure there
must be continual readjustment of the vocal bands. As the
pressure rises the tension of the bands must decrease or their
weight must increase (p. 263). As these adjustments are of
considerable difficulty, it naturally results that loudness and
rise in pitch readily occur together. In song a fixed relation
between pitch and intensity is carefully avoided.
The utter inability of the ear to distinguish the loudness,
pitch and duration factors in accent has been strikingly
illustrated in the discussions of Lithuanian accent.^ The
discussion has stimulated experimental work.
The Lithuanian and Lettic accent is mainly one of stress,
two forms of accented syllables being distinguished: the
broken, or rough (' gestossen '), and the slurred, or soft
■('schleifend '). The factors of loudness (as indicated by
breath pressure) and duration appear in records made by
Schmidt- Wartenberg.2
1 Literature and summary in Hirt, as belore, 102.
^ Soiimidt-Wartenberg, Zur Physiol, d. litauischen Akzents, Indogerm.
Forschungen, 1897 VII 211; Phonetische Untersuchungen zum lettischen Akzent,
Indogerm. Forschungen, 1900 X 117; Further contributions to the Lithuanian
■accent question. Trans. Amer. Philol. A.ssoc., 1901 XXXII xxiv.
3.3
514
FACTORS OF SPEECH
Records of the dialect of Mariampol show that the medium
long quantity does not exist there, and that Kueschat's con-
tested description of the two accent varieties, the slurred and
the broken, is true with regard to that dialect at least.'
In a curve of the Lithuanian but! (Fig. 337) the 'broken
tone ' u shows greater intensity at the beginning followed by
a sudden fall and then by a small recovery near the end of
the u due to a slight aspiration, after which the vowel stops
abruptly; the t, indicated by the
low curve, is followed by the
breath-rise for the i. A curve
for budas shows a similar con-
dition for the ' slurred tone ' u
without the final aspiration and
with a more gradual ending.
The expiratory curves of vari-
ous Lettic words show that there
exists — in the district of Wol-
mar — a third variety, the ' fall-
ing ' accent,* by the side of the
lengthened intonation and the
' Stosston. ' It is this falling
accent that corresponds histori-
cally to the slurred accent (' ge-
schleifter Akzent'} of Lithuanian
speech, not the ' Stosston ' as hith-
erto supposed. As to the quan-
tity of the three accents there is
no difference. The lengthened tone is rising or level as to
expiration, the falling tone is gradually decreasing in inten-
sity, the ' Stosston ' breaks the vowel or diphthong into two
parts by means of an energetic explosive utterance of the
second part, which may be preceded by a closure of the glottis.
Fig. 337.
1 Schmidt- Warteneekg, Further contributions to the Lithuanian accent question.
Trans. Amer. Philol. Assoc, 1 901 XXXII xxiv.
2 Schmidt-Wartenberg, Plionetische Untersuchungen zuni lettischen Akzent,
Indogerm. Forschungen, 1900 X 117.
ACCENT 515
In the case of diphthongs this second expiration lies within
the glide. The curve of the Lettic put (Fig. 337) shows the
u aspirated at the end, followed by the closure for the t and
its explosion. The curve for put (Fig. 337) shows the
decrescendo or * falling ' accent for the u followed by the
closure for t and its explosion as before. The curve for
ka-ut shows a rise for the a followed by a small decrease for
the so-called ' Stosston ' and a rise for u, the t being indi-
cated as before; the 'Stosston ' between the two parts of the
diphthong seems to occur without complete cessation of
breath. This " Stosston ' is probably an incomplete glottal
catch (p. 278).
Gauthiot has studied ^ the relations among the ' accent of
loudness,' the ' accent of pitch ' and the duration. The first
two do not necessarily coincide and, in fact, do so only
in certain definite cases. Records of breath pressure show
that the " rough ' accent is one of strong initial stress suddenly
decreasing and that this form is invariable. The ' soft '
accent has two summits of stress, one at the beginning and
one at the end of a medial or final vowel ,■ in an initial vowel
the first summit is lost; the accent of pitch and the duration
remain the same for initial, medial and final vowel.
Gauthiot's records show that there are two main classes
of vowels in respect to duration — long and short, whose
average lengths bear the remarkably constant relation of 4 : 2.
The influence of the accent shows itself in the abridgment
of ' short ' vowels to ' extra-short ' ones in a tonic medial
syllables; these vowels may perhaps be grouped as a third
class, making in all three groups with the relations of
duration 4 : 2 : 1. These extra-short vowels are of modern
origin.
Applications of these facts to the study of phonetic changes
in Lithuanian and Greek have been made by Gauthiot and
Meillet.^
' Gadthiot, De I'accent et de la quantity en hiuamen, La Parole, 1900 II 143.
2 Meillet, J propos de t'article de M. R. Gauthiot sur les intonations litu-
amennes. La Parole, 1900 II 193.
516 FACTORS OF SPEECH
Researches by Wallin ^ indicate that in both prose and
verse the average emphatic syllable is invariably longer than
the average unemphatic one, that the ratios between the two
vary with different persons and that the ratios are never ex-
pressible by simple numbers.
Rbfeebnces
For complete account of Indogermanic accent and references : Hiet,
Der indogermanische Akzent, Strassburg, 1895. For the individual lan-
guages : Hempl, German Orthography and Phonology, 167, Boston,
1898 ; SiEVERS, Grundzuge d. Phonetik, 5. Aufl., 215, Leipzig, 1901 ;
also the works mentioned on pages 307, 311, 314.
^ Wallin, Researches on the rhythm, of speech, Stad. Yale Psych. Lab., 1901
IX 1.
CHAPTER XXXVI
AUDITORY AND MOTOK EHTTHM
In the flow of speech there appears a phenomenon that is
recognized as extending through all the mental experiences
of the individual; this is known as ' rhythm.'
As an auditory phenomenon rhythm appears in certain
relations of sounds. These may be illustrated by the follow-
ing experiments. In describing them I shall merely indicate
the general plan, leaving the details of technique to the skill
of the experimenter.^
A tone may be sounded for a definite time at definite inter-
vals. The result is a 'rhythm of sound and pause,' or, more
briefly, a ' pause rhythm. ' .
Two tones of the same quality Avith equal energies and
durations but of different pitches are alternated with no
silence between them. The result is a 'rhythm of pitch.'
The experiment can be performed by a contact apparatus,
forks, telephones, resistances, etc. It can be illustrated by
using two organ tones or two voice tones, but the higher
tone will have a greater energy for the saihe blast and the
mental relations of duration and energy impel the per-
former to falsify the equality of duration.
Two tones of the same quality, duration and pitch, but of
different energies, can be used to produce a 'rhythm of inten-
sity.' When the duration, pitch and energy are the same
but the quality different (as in tones from two different musi-
cal instruments, or from two different voices), the result is
a ' rhythm of quality. ' A ' rhythm of duration ' is also a
1 Apparatus suggestions may be found in Scriptuee, Elementary course in
psychological measurements, Ex. XIII, Stud. Yale Psych. Lab., 1896 IV 127,
and in Scripture, New Psychology, Ch. X, London, 1897.
618 FACTORS OF SPEECH
fundamental form arising when the tones are of different
lengths; it cannot be readily illustrated by simple tones xDf
the same quality, pitch and energy without introducing a
break or a click to mark off the first from the second ; this
break is a new factor that modifies the results. It can be
readily illustrated in combination with some other factor.
Similar experiments may be made with three or more tones.
These experiments illustrate the fundamental law that
rhythm in tones is dependent on changes in quality, pitch,
intensity and duration, or, if by r we indicate the rhythmic
effect (or rhythtaic feeling), r =f {x, i/, z, w) where x, «/, z,
w indicate the factors just mentioned. Investigations should
be made on the strength of the rhythmic feeling as depend-
ing on each one of the factors by varying one while the
others are kept constant in the usual way. For example,
a series of rhythms graded by equal steps of effective-
ness might be established by adjusting the intensity alone;
the law connecting the two factors would be the law of in-
tensity rhythm. Likewise the just perceptible difference
(p. 100), the just perceptible change (p. 101), etc., might
be similarly determined. If the laws of rhythm prove to be
linear functions (e. g., r oc log x, r oc x^, or similar rela-
tions), the ultimate expression oi r — f (x, y, z, w) ought
not to be looked upon as impossible. It must be borne in
mind that a;, y, 2, iv represent changes from the average (or,
perhaps, expected) quality, pitch, intensity and duration, and
not the actual amounts of these quantities themselves.
On the assumptions 1. that the rhythmic feeling increases
or decreases with increase or decrease of each one of these
factors (or ? a a;, r a «/, ?• a s, r oc w) and 2. that in their
effect on the rhythmic feeling the factors x, y, z, w act inde-
pendently, we can deduce the following laws :
1. The total rhythmic effect varies with the total change
in the factors of rhythm ;
2. Any factor may be used to replace another in producing
the rhythmic effect.
These laws are of fundamental importance for the theory
AUDITORY AND MOTOR RHYTHM 619
of verse ; their proof, extension or refutation forms one of the
most important problems that can be undertaken by experi-
mental phonetics. The first of the above assumptions is a
highly probable one. The second is still unsubstantiated;
it may, perhaps, be tested by establishing equalities between
rhythmic feelings produced by different factors ; for example,
if, in a rhythm produced by a combined change in pitch
and intensitj', a decrease in one of the factors can be com-
pensated by an increase in the other without any diminution
in the rhythmic feeling (perhaps without our perception of a
change), the assumption would be verified for the particular
case. Some experiments on tones ^ have shown that a change
in pitch can be used as a substitute for or as an intensifica-
tion of intensity.
Stronger forms of rhythm may be produced by varying
two or more factors. One tone may be made higher and
louder, or higher and longer, or louder and longer, or higher,
louder and longer than the other one.
Complex forms of rhythm are produced by using several de-
grees of one factor, as in •••••• or ••••••• • etc.
In the preceding experiments each element of rhythm
has been supposed to be of constant character throughout its
duration, and the change from one element to the other to be
sudden ; thus, a pause, quality, pitch or intensity rhythm of
Fig. 338. Fio. 339.
Fig. 340, i'lG- 341.
this kind would be of the character indicated by Fig. 338.
Another typical form of rhythm would be one in which one
factor rises and falls evenly, with sudden changes, as indi-
cated by Fig. 339. Still other forms involve less easily de-
1 Squire, A genetic study of rhythm, Amer. .Tour. Psychol., 1901 XIII 560.
520 FACTORS OF SPEECH
scribed changes, such as in Fig. 340 or Fig. 341 or Fig.
342. The rhythmic effect depends not only on the amount
and direction of the change but also on its rapidity and
regularity.
A L
Fie. 342. Fig. 343.
Very sudden changes of strong rhythmic effect are pro-
duced by alternations of sharp noises and much longer
silences; the rhythm would be as indicated in Fig. 343.
Most of the experiments on auditory rhythm have been made
on such a scheme by use of sharp clicks. The conditions do
not closely resemble those of the flow of speech, but the
results show some of the fundamental facts of auditory
rhythm.
In some experiments by Bolton ^ an apparatus was so con-
structed that a sequence of sounds was produced which did
not vary in pitch, intensity, or quality. The rate at
which the sounds were made to succeed one another could be
varied from one in two seconds to ten in one second. When
a person listened to the series and gave his attention closely
to it, the series seemed to break up into groups of sounds ; it
did not appear uniformly continuous. These groups con-
tained a larger or smaller number of sounds according as the
rate was fast or slow. If the rate was slow, groups of two
sounds seemed the more natural. At a faster rate grouping
by threes or fours was more easy and pleasing, and with the
fastest rates the sounds seemed to form groups either by sixes
or eights, and sometimes the sequence seemed to rise and fall
in intensity at regular intervals of one second or more. The
grouping was not distinct. Whatever the rate, the sounds
might be made to group by suggesting to the subject a pen-
dulum or some other rhythmical instrument. Groupings
1 Bolton, Rhythm, Amer. Jour. Psychol., 1893 VI 214; the following account
is condensed from a summary in my New Psychology, Ch. XI.
AUDITORY AND MOTOR RHYTHM 521
might be suggested by counting 2's, 3's, 4's, 6's, or 8's,
accenting the first sound. It was difficult and even impos-
sible with most persons to group by 5's or 7's. The group-
ing was usually accompanied by some muscular movement.
Frequently it was tapping with the foot or fingers; some-
times it was beating time with the hand or the thumb.
Some subjects nodded the head, others counted inaudibly,
and still others felt indefinitely localized muscular contrac-
tions in the larynx, diaphragm, viscera, scalp, eyelids, etc.
Muscular twitchings were to be seen in the muscles of the
face and limbs at times when the subject declared he felt
nothing of the kind.
The grouping was accomplished by placing a stress or
accent upon the first sound in a group. In groups of three
the first and second sounds were accented, the first more
strongly than the second. In groups of four the first and
third were accented, the first again being the stronger.
Groups of four seemed at times to break into two groups of
two sounds each, groups of six into two groups of three, and
groups of eight into two groups of four. In groups of six
the accents came alwa3^s upon the first and fourth, and in
groups of eight upon the first and fifth. The accents were
apparently at the basis of the splitting up of the longer
groups, and, when they did so break up, the subject felt a
tendency to swing forward or backward or from side to side.
This invariably suggested the pendulum. Many persons
pictured or visualized some moving object which seemed to
swing or revolve as the sounds were grouped.
When the various rates at which the different groupings
were felt to be most pleasing and natural were compared and
the average times for each taken, it was found that the time
limit of each group was nearly the same — a little more than
a second. The explanation of this was based upon the
rhythmical character of the attention. Attention is periodic,
and, when it is concentrated upon a continuous series,
becomes quite regular in its period. An object that does not
change cannot be attended to for more than a few seconds.
622 FACTORS OF SPEECH
The attention will pass involuntarily from the object to some
one of its parts or to one of its associates.
The experiments showed that an even series of clicks is
subjectively transformed into groups by accenting some of
them. Thus, in one case the even series
would be subjectively made into
• ••••••••
The supposition of BOLTOX is that this accentuation is one
of intensity; it apparently did not occur to him that it might
contain other factors.
There is no doubt but that the accentuation includes factors
of pitch and quality also. The accented sounds receive
different characters; a series of clicks from a telegraph
sounder is heard not as tick-tick-tick-tick but as tick-tock-tick-tock
(the sounds from a clock differ physically).
WxJNDT notices 1 an apparent lengthening of the interval
before each loud click in a series of alternate loud and soft
ones ; thus the even series
• ••••■•••
appears as
^ •••••••••
A short interval filled with clicks or mental work of any
kind appears longer than an equal empty one ; ^ thus, in such
a series of clicks as
a h c
the time from a to b appears longer than that from h to c.
The relation is reversed for long intervals. Louder series of
clicks appear to have shorter intervals. A stronger click
inserted in the middle of an interval between two short ones
makes the first half appear shorter. A louder click at the
beginning of an interval makes it seem longer than the
following intervals with weaker clicks.^
1 WuNDT, Volkerpsychologie, I ii380, Leipzig, 1900.
2 Hall and Jastrow, Studies of rhythm. Mind, 1886 XI 62; Meumann,
Beitragezur Psychologic des Zeitbewusstseins, Philos. Stud. (Wundt), 1896 XII 127.
' ME0MANN, Beitrdge zur Psychologic des Zeitsinns, Philos. Stud. (Wundt),
1893 IX 306.
AUDITORY AND MOTOR RHYTHM 523
As a motor phenomenon rhythm appears in repeated
actions; it is a mental activity utterly different from and,
except by association, entirely unconnected with auditory
rhythm. It is convenient to use the term ' rhythmic action '
or ' motor rhythm ' in contrast to ' auditory rhythm. '
The laws of rhythmic action may be illustrated by the fol-
lowing experiments.
Eepeated actions tend to be regularly repeated. This
observation has been carefully tested by Smith, i Squire^
and MiYAKE.3
In a series of experiments in which children repeated the •
syllable mi six times in succession, Squire* usually heard
an involuntary rhythm produced by variations in duration,
intensity or pitch, but found no rhythm when the effort of
articulation was so great that attention was necessarily
directed on the separate syllables and when two syllables
occupied more time than the normal pulse of attention (time
between two maxima of attention, usually 2' to 3').
In Miyake's experiments two Marey tambours were
arranged so that the recording point of one of the tambours
drew a line on the smoked surface of a drum.^ The subject
was required to hold the lever connected with the other
tambour between his thumb and index linger, and, his eyes
being closed, to move it up and down at intentionally irregular
intervals at a rather rapid rate.
The experiments were made on three subjects ; a specimen
record is shown in Fig. 344. The height of the curve is
related to the amplitude of the movement, and therefore to
the intensity of the exerted muscular energy, while the hori-
zontal distance indicates the length of the time between the
successive movements. The line at the bottom indicates
1 Smith, Rhythmus und Arbeit, Philos. Stud. (Wundt), 1900 XVI 282.
2 Squire, A genetic study of rhythm, Amer. Jour. Psychol., 1901 XIII 497.
8 MiYAKE, Researches on rhythmic action. Stud. Yale Psych. Lab., 1902 X 1.
' Squike, as before, 516.
5 Details of the arrangement are given in Scripture, Elementary course in
psychological measurements, Ex. VIII, Stud. Yale Psych. Lab., 1896 IV 108,
109, and are shown in Fig. 71 above.
624
FACTORS OF SPEECH
fifths of a second. It was observed in this record as well as
in the others (1) that there is a constantly recurring tendency
to repeat equal intervals in succession, (2) that the same
JW\11j^||JIjIJIiJWvJ^-.'J
ftl^irw||JUilH^fv^^ xtmB
[\M\imr^'^^'^-Awm '
'l|Ljiu4uiy
44Hi^4iMM(;^tM-*
Fig. 344.
intensity of the nauscular energy is also often repeated, and
(3) that the weak and strong intensities often alternate. The
attempt at irregular action thus shows a persistent tendency
to revert to action regular in time and intensity.
/ 10 20 30 40 50 60 70 80 90 100 110 120 130 MO
Fig. 345.
The records obtained in the above experiments show the
ojiaracteristics of arhythmic action under the various circum-
stances for the various subjects. Accurate measurements of
the lengths of the intervals could be better obtained by another
AUDITORY AND MOTOR RHYTHM 525
method. A Dbpeez marker (p. 92) and a key with a break
contact were put in series in a 1="° current. The pointer of
the marker was rested lightly against the smoked surface of
a drum. The subject was asked to tap the key at intervals
as irregular as possible, the longest time between two succes-
sive beats being limited to about one second. He was seated
comfortably before the apparatus and his eyes were closed
during the experiments. The results of many experiments are
averaged in the diagram (Fig. 345), in which the figures
on the horizontal axis indicate the serial number of the tap
and those on the vertical axis the length of the successive
periods, S = 0.01^ being the unit.
The diagram shows the following facts : (1) there are repeti-
tions of equal or about equal periods ; (2) the unequal periods,
which occur after or in the middle of a group of the repeated
equal periods, are, in many cases, simple multiples of the
latter; (3) the periods from 12 to 17 are most frequent; (4)
rhythmic alternations of long and short intervals also occur.
These facts seem to indicate that arhythmic movements
have a constant tendency to become rhythmic, notwithstand-
ing the voluntary effort of the subject to execute the move-
ments at irregular intervals. The subjects of the experiments
invariably agreed in confessing that the arhythmic tappings
required strenuous effort and that the performance was very
fatiguing.
The involuntary tendency to rhythmize sounds appears
in the attempt to repeat the same sound monotonously. It
can be very strikingly shown by having the sounds regis-
tered in a phonograph and studied at leisure. In some
experiments of this kind the sounds a a a a a a a a lo lo lo
lo lo lo lo lo were spoken in what was intended to be a per-
fect monotone. Each record was made by one of three per-
sons and the judgment of the character of its rhythm was
thereafter made by the other two while listening to the phono-
graph. The judgment by J and ilf on a record made by S
was that it was ' spoken in trochee with emphasis pro-
duced by increased intensity and rise in pitch ;^' on listening
526 FACTORS OF SPEECH
to the record the speaker S made the same judgment. A
record by J" was judged by ilf and S to have the same charac-
teristics as the preceding one. The record by M (Japanese)
was judged by S to begin with syllables of qqual emphasis
and to end with trochaic rhythm produced as in the preceding
cases, but by J to have an iambic character.
Two entirely different forms of regularly repeated action
are to be distinguished, i In one form the subject is left free
to repeat the movement at any interval he may choose. This
includes such activities as walking, running, rowing, beating
time, and so on. A typical experiment is performed by tak-
ing the lever of a Maeey tambour between the thumb and
index finger and moving the arm repeatedly up and down;
the recording tambour writes on the drum the curve of move-
ment (Fig. 71). Another experiment consists in having the
subject tap on a telegraph key or on a noiseless key and
recording the time on the drum by sparks or markers. Other
experiments may be made with an orchestra leader's baton
having a contact at the extreme end, with a heel contact on
a shoe, with dumb-bells in an electric circuit, and so on.^
For this form of action I have been able to devise no better
name than ' free rhythmic action. '
In contrast with this there is what may be called ' regu-
lated rhythmic action.' This is found in such activities as
marching in time to drum-beats, dancing to music, playing
in time to a metronome, and so on. A typical experiment is
that of tapping on a key in time to a sounder-click, the
moment of the click and that of the movement of the finger
being registered on a drum.
Regulated rhythmic action differs, I believe, from free
rhythmic action mainly in the manner of judging the coin-
cidence of the movements with the sound heard (or light seen,
etc.). This view puts aside all purely physiological theories
1 Scripture, Observations on rhythmic action, Science, 1899 X 807; also in
Stud. Yale Psych. Lab., 1899 VII 102.
^ Scripture, Thinking, Feeling, Doing, 2d ed., Ch. on Rhythm (in press).
AUDITORY AND MOTOR RHYTHM 527
of regulated rhythmic action. One of these theories ^ is based
on the assumption that the labyrinth of the ear contains the
tonus -organ for the muscles of the body. It asserts that
vibrations arriving at the internal ear affect the whole con-
tent, including the organ for the perception of sound and
the tonus-organ. Thus, sudden sounds like drum-beats or
emphasized notes would stimulate the tonus-organ in unison,
whereby corresponding impulses would be sent to the muscles.
This theory has very much in its favor. It is undoubtedly
true that such impulses are sent to the muscles. Thus, at
every loud stroke of a pencil on the desk I can feel a result-
ing contraction in the ear which I am inclined to attribute to
the tensor tympani muscle (p. 78). Likewise a series of
drum-beats or the emphasized tones in martial or dance music
seem to produce twitchings in the legs. Fere has observed
that, in the case of a hysterical person exerting the maximum
pressure on a dynamometer, the strokes of a gong are regu-
larly followed by sudden increased exertions.^ Nevertheless,
these twitchings are not the origin of the movements in regu-
lated rhythmic action. For many years I have observed that
most persons regularly beat time just before the signal
occurs ; that is, the act is executed before the sound is pro-
duced. Records of such action have been published,^ but
their application to the invalidation of the tonus-theory was
first suggested by Miyakb. This does not exclude the use
of muscle sensations, derived from tonus -twitches, in cor-
recting movements in regulated rhythmic action, although
they presumably play a small or negligible part as compared
with sounds.
Another argument in favor of the subjective nature of
regulated rhythmic action is found in the beginning of -each
experiment on a rhythm with a new period; the subject is
quite at loss for a few beats and can tap only spasmodically
1 EwALD, Untersuchungen iiber d. Endorgan d. Nervus octavus, 294, Wiesbaden,
1892.
^ Fere, Sensation et mouvement, 35, Paris, 1887.
3 ScEiPTUBE, Thinking, Feeling, Doing, as before ; New Psychology, 182,
London, 1897.
528 FACTORS OF SPEECH
until he obtains a subjective judgment of the period. If the
tonus-theory were correct, he should tap just as regularly at
the start as afterward.
The conclusion seems justified that regulated rhythmic
action is a modified free rhythmic action, whereby the subject
repeats an act at what he considers regular intervals, and
constantly changes these intervals to coincide with objective
sounds which he accepts as perfectly regular.
In free rhythmic action there is one interval which on a
given occasion is easiest of execution by the subject. This
interval is continually changing with practice, fatigue, time
of day, general health, external conditions of resistance, and
so on.
' It has long been known that in such rhythmic movements
as walking, running, etc., a certain frequency in the repeti-
tion of the movement is most favorable to the accomplish-
ment of the most work. Thus, to go the greatest distance
in steady traveling day by day the horse or the bicyclist must
move his limbs with a certain frequency ; not too fast, other-
wise fatigue cuts short the journey, and not too slow, other-
wise the distance traveled is unnecessarily short. This fre-
quency is a particular one for each individual and for each
condition in which he is found. Any deviation from this
particular frequency diminishes the final result. '
It is also a well-known fact that one rate of work in nearly
every line is peculiar to each person for each occasion, and
that each person has his peculiar range within which he
varies. Too short or too long a period between movements
is more tiring than the natural one. This law appears also
in some experiments made by Smith. ^
It is highly desirable to get some definite measurement of
the difficulty of a free rhythmical action. This cannot well
be done by any of the methods applicable to the force or
quickness of action, but it may be accomplished in the manner
described in Appendix III.
' Smith, as before, 302.
AUDITORY AND MOTOR RHYTHM 529
The function of the ear in free rhythmic action has been
investigated by Miyake i by a method and with apparatus
that seem specially adapted to the determination of the fun-
damental laws of motor rhythm. Analogous experiments
should be made for song and speech.
The experiment of Miyake consisted in tapping on a
noiseless key. The small ' strap ' key used in this experiment
was made of an elastic brass strip
B, 46°™ long and 9"™ wide,
mounted on a wooden block E;
a brass stop C kept the free end
of the spring from rising more
than 4"" from the block. A
slight pressure on the button A at the free end of the
strap forced it nearer the block and broke its contact
with the brass stop. Platinum points I> were used to
ensure good contact between the strap and the stop. The
key was put in a rubber bag and packed in felt so that the
sound was rendered absolutely inaudible; the key was thus
a perfectly noiseless one. The wires F G- projected from
the bag. A spot on the surface of the rubber bag under
which the button of the key was situated was marked with
a sign. It indicated the. point where the tapping was to
be done. The adjustment of the key was such that the
slightest touch broke the circuit. This highly successful
instrument is the first solution of the problem of an abso-
lutely noiseless key breaking contact exactly at the moment
touched.
For producing the auditory stimuli a pair of discharging
points were connected to a spark coil. The two brass rods
were put in a horizontal line with a distance of about 2™™
between them and connected to the poles of the secondary
circuit. When the primary circuit was broken, a sound was
produced by the spark. The points were put behind a black
screen, so that the spark could not be seen by the subject.
The general plan of the arrangement is shown in the ac-
1 MiTAKE, as before, 8.
34
530
FACTORS OF SPEECH
companying diagram (Fig. 347). The noiseless key K, with
condenser 0 around the break, was placed in the circuit with
the battery B and the primary circuit P of a spark coil.
The secondary coil S was connected in series with the
metallic registering point of a Pfeil marker (p. 91) M,
the recording-drum D, and the discharging points J, so
that a break in the primary circuit would produce a dot
(p. 12) on the time-line at the point of the marker and also
between the discharging points at the same moment. Thus
the movement of the finger on the key, breaking the primary
Fig. 347.
circuit, resulted in a sound of the spark between the discharg-
ing points and a record on the smoked drum D simultaneously.
A switch H was put in the secondary circuit around the
discharging points. When the switch was closed, the short-
circuit prevented sparks at the discharging points, and the
tapping on the key was not accompanied by the sound of the
spark, although still recorded on the drum. A time-line
was drawn on the drum by the marker M run by a fork
(pp. 15, 91).
The key and the discharging points were placed in a
special quiet room ; the rest of the apparatus was in another
room.
AUDITORY AND MOTOR RHYTHM 531
The subject with closed eyes beat time with the index
finger of his right hand at what he considered to be a con-
stant interval. The rate of the movement was left entirely
to his own choice.
The average time chosen by the subject, his immediate
probable error p (calculated as on p. 201), and his relative
probable error r =p/n expressed as a percentage, are given
in the following table in thousandths of a second.
With Sound.
Without Sound,
bject.
Average
Average
'
time.
P
T
time.
P
r
M
519
10.8
2.1%
575
17.1
2.9%
H
37.5
13.6
3.6%
379
20.5
5.+ %
Y
718
23.4
3.3%
838
28.0
3.2%
C
747
12.1
1.5%
729
21.3
2.9%
A comparison of the corresponding probable errors in the
same horizontal line will show that those of the free rhythmic
movement with the sound are in general smaller than those of
the movement without the sound. This holds true for both
the simple and the relative probable errors. The general
conclusion may be drawn that free rhythmic movement with
the sound is more regular than that without the sound.
It can be also noticed in the table that the length of the
period is in general shorter in the movements with the
sound than in those without the sound. This is especially
clear in the cases of the subjects M and Y, in which the
periods with the sound are always shorter than those without
the sound. The apparently contrary case with O is due to
one erratic set of results differing entirely from the rest.
This general difference is probably due to the fact that the
interval which is marked off by the muscle, joint and skin
sensations and the auditory sensations appears longer to the
subjects than the equal interval whicli is marked off by the
former group alone, and that thus the subject is uncon-
sciously influenced by the relative degrees of fulness or
emptiness (p. 522).
Another problem of fundamental importance in the con-
632 FACTORS OF SPEECH
sideration of speech and song as motor phenomena lies in the
relation between intensity and interval in rhythmic action.
Ebhaedt ^ made two series of experiments on this prob-
lem. In the first series the tapping was done on an electric
key, and in the second on a piano with electric connec-
tions. The records were taken in both cases on a kymo-
graph (p. 198). The results showed that the interval
following the emphasized beat was lengthened as compared .
with that which followed the unemphasized beat. In
Ebhaedt's experiments the tapping was accompanied by
the noise of the instrument.
Miyake's work 2 includes a study of movements without
noise. The noiseless key (p. 529) was put with the Ppeil
marker (p. 91) in series in a 1™ current. The metallic
point of the marker was connected with one pole of the
secondary coil of a spark coil, the other pole being connected
to the base of the recording drum. The current from a 100-
fork (p. 15) was sent through the primary coil. In this way
the beats were recorded by checks in the line on the drum;
these were divided by the sparks into equal spaces, each of
which corresponded to ^-J-j of a second. The subject tapped
with his -finger (generally with the index finger of the right
hand) on the noiseless key, according to the following
schemes :
(a) ■l'-2, V-2, l'-2, . . .
lb) 1-2', 1-2', 1-2', . . .
(c) l'-2-3, l'-2-3, . . .
(d) l-2'-3, l-2'-3, . . .
where the beat to be emphasized is marked with the sign '.
In the scheme l'-2, for instance, the subject was asked to
emphasize every first beat of the rhythmic group, but he had,
at the same time, to try to keep always a uniform interval
between two successive beats, not only between 1' and 2, but
1 Ebhardt, Zwei Beitrage zur Psychohgie des lihijthmus und des Tempo, Zt. f.
Psych, u. Physiol, d. Sinn., 1898 XVIII 99.
2 MiTAKE, as before, 13.
AUDITORY AND MOTOR RHYTHM 533
also between 2 and 1' although he was to think of the groups
as in pairs l'-2, not 2-1'. The speed of the movements was
left to the choice of the subject.
The results showed that the lengthening of the interval
following the emphasized beat was more marked with the
scheme 1-2' than l'-2. The average ratios of the two inter-
vals in the two different rhythmic schemes were
l'-2 1-2'
1' to 2 : 2 to 1' 1 to 2' : 2' to 1
C. W. 1.00 : 0.94 0.82 : 1.00
M. M. 1.00 : 0.93 0.90 : 1.00
J. K. 1.00 : 0.91 0.90 : 1.00
The expression ' 1' to 2 : 2 to 1' ' means the ratio between
the average time from the emphasized first beat to the second
beat and the average time from the second beat to the empha-
sized first one.
The relative lengths of the long and short intervals are not
the same in the two different schemes; the interval which
comes after the emphasized beat is comparatively longer
in 1-2' than in l'-2. The same fact was observed by
Ebhaedt.^
Why is the interval following the emphasized beat length-
ened more in one rhythmic scheme than in the other? This
can be accounted for by assuming another factor, besides
emphasis, that lengthens the period of the movements. It is
due, as already pointed out by Ebhabdt, to the formation of
the rhythmic group. Rhythmic movements with grouping
differ in their nature from those without grouping. The
latter is merely a series of repeated movements at a uniform
interval, in which every single movement is regarded as a co-
ordinate unit. In the former, a series of the movements is
divided into groups containing a certain number of move-
ments as their content, and each of such groups is regarded
as a unit.
Ebhaedt supposed that at the end of the rhythmic group
1 Ebhardt, as before, 99.
534 FACTORS OF SPEECH
a suspension of attention takes place and that the moment of
suspension can be considered as a dead time, which is to be
added to the length of the foregoing group. We are not
certain whether such suspension of the attention takes place
or not. But it seems to be more probable that we have a
tendency to insert some ' pause ' between two successive
rhythmic groups, in order to mark off the groups distinctly
from each other. The ' pause ' is to facilitate the formation
of the groups.
We may suppose then that a certain length of ' pause ' was
inserted between the groups in the scheme l'-2 as well as in
1-2', and that because the interval from 2' to 1 of the scheme
1-2' is lengthened both by the ' pause ' and the emphasis,
it is made considerably longer than the time from 1 to 2',
whereas in the scheme l'-2 the time from 1' to 2 is length-
ened only by the emphasis, while the time from 2' to 1 is
lengthened by the ' pause,' whereby the difference between
1-2' and 2'-l is not so great.
The averages of the ratios for the schemes l'-2-3 and
l-2'-3 were
l'-2-3 l-2'-3
1' to 2 : 2 to 3 : 3 to 1' 1 to 2' : 2' to 3 : 3 tol
M. M. 1.00 : 0.60 : 0.94 0.60 : 1.00 : 1.00
J. K. 1.00 : 0.94 : 0.92 0.92 : 1.00 : 0.99
C. W. 1.00 : 0.98 : 0.95 0.99 : 1.00 : 0.97
The scheme l'-2-8 shows again that the interval following
the emphasized beat was the longest. The scheme l-2''-3
shows likewise the same tendency. The interval 3 to 1 is
longer than 1 to 2', evidently including the ' pause. '
If l-2'-3 is compared to l'-2-3, we find that there is a
remarkable difference between the two rhythmic schemes in
regard to lengthening of the intervals between the groups.
The interval 3 to 1' of the scheme l'-2-3 is not so much
lengthened as 3 to 1 of l-2'-8. In other words the ' pause '
between the groups is longer in l-2'-3 than in l'-2-8. This
AUDITORY AND MOTOR RHYTHM 535
fact indicates that the length of the ' pause ' is not the same
in all rhythmic forms. It depends, probably, on the amount
of difficulty in the formation of the rhythmic groups. The
more difficult the formation of the groups, the longer is the
pause. In the case l'-2-3 with the first beat of a group
emphasized, the group can be easily marked off from the
preceding or the following groups, and the rhythmic group
can be formed, without lengthening very much the interval
between them. But the case is different with the scheme
l-2'-3, where neither the first nor the last beat of a group is
emphasized. Of the two similar beats one comes at the end
of a group and the other at the beginning of the next group ;
the two successive groups can be marked off distinctly only
by lengthening the interval between them.
From this series of experiments the conclusions can be
drawn 1. that the interval which follows an emphasized
beat is lengthened, 2. that the interval which comeg between
rhythmic groups is lengthened, 3. that the lengthening of
the interval between rhythmic groups is not equally great
in all the rhythmic schemes.
It will be noticed that the forms of motor rhythm investi-
gated were the trochee, iambus, dactyl and amphibrach; it
is to be regretted that the spondee and anapest were not
included.
Using a similar method Miyakb extended the experiments
to the sounds of speech ; the results have been summarized in
the preceding chapter in their bearings on accent.
The grouping of rhythmic impressions and movements by
twos is easier than that by threes. ^ The tendency shows itself
in the preponderance of iambic and trochaic verse over the
anapestic and dactylic. The trochaic and dactylic forms seem
easier than the iambic and anapestic ones.
Researches of the kind described in this chapter have as
their object the determination of the fundmental laws of
1 Bolton, Rhythm, Amer. Jour. Psychol,, 1893 VI 216; Smith, Rhythmus
und Arbeit, Philos. Stud. (Wundt), 1900 XVI 217; Sqciee, A genetic study of
rhythm, Amer. Jour. Psychol, 1901 XII 535.
536 FACTORS OF SPEECH
rhythm. These laws appear also in the rhythm of speech and
help to an understanding of its complexities.
References
For auditory and motor rhythm : Wundt, Physiologische Psychologie,
4. Aufl., II 83, Leipzig, 1893; Vblkerpsychologie, 1375, Leipzig, 1900 ;
ScKiPTURE, New Psychology, Ch. X-XI, London, 1897; Meumann,
Untenuchungen zur Psychologie und Aesthetik des Rhyihmus, Philos. Stud.
(Wundt), 1894 X 249 (full references) ; Riemann, Katechismus d.
Musik, Leipzig, 1888; Musiklexikon, 4. Aufl., Leipzig, 1894; Wagner,
Gesammelte Schriften und Diohtungen, 2. Aufl., VIII, Leipzig, 1888 ;
Westphal, AUgemeine Theorie d. musikal. Rhythmik, Leipzig, 1880.
CHAPTER XXXVII
SPEECH RHYTHM
The earliest experiments on the rhythm of connected speech
were by BRtJCKB,^ who recorded, with a marker on a smoked
drum, the movements of a finger in beating time while he
recited verses in iambic hexameter, and in alcaic and sapphic
meters, in a scanning fashion ; he also made records of the
movements of the lips. He found that the lengths of the
successive feet were equal, as far as his apparatus indicated ;
this was, however, not fine enough to detect small differences.
The records of Keal and Mares ^ (p. 499) for various lines
of verse gave results like the following (F denotes length of
foot in different records, T length of thesis, A length of arsis,
all in hundredths of a second) :
F^
T^
^1
F^
F^ F,
Louka ko
61
30
31
92
84 62
sou sece
65
32
33
87
91 62
n^ voiia
67
37
30
80
83 65
vd tady
61
40
21
77
78 66
z^ pachy
63
38
25
80
73 65
ddvd
—
—
—
—
— —
(ViNAKICKT,
'The sowed field sends forth perfume.'
Nase Labe vody vali proudem mocnym v ddlnon zem.
F, 51 52 47 59 66 67 57 —
Fa 52 61 49 70 64 66 54 —
(The mighty waves of our river Elbe flow far into the world.)
1 Brucke, Die physiologischen Grnudlagen d. neuhochdentschen Verskunst,
Wien, 1871.
2 KrAl a Mares, Trvdnt hldseh aslabik die objehtivngmiry, Listy Filologicke,
1893 XX 2.57.
538 FACTORS OF SPEECH
Their results indicate that even with the same person the
same vowel has a different length according as the emphasis
is greater or less when a verse is recited in a scanning fashion ;
and that neither in intensity-verse nor time-verse arc the
lengths of feet ever exactly equal, the ratio of the emphasized
part to the unemphasized part of a foot not keeping a relation
like 1 : 1, but rather like 30 : 31, or 32 : 33.
In HtJEST and McKay's^ experiments on the time relation
of poetical meters the subject recited poems representing
each of the four usual meters, iambus, trochee, dactyl and
anapest, while he beat in unison with the finger on a pointer
which registered the lengths of the beats on a smoked drum.
It was found that in iambic meter the syllables had a ratio
of about 1:2, iu trochaic of a little less than 1.5 : 1, in
anapestic of about 1 : 1 : 1. 2, and in dactylic of 1.6 : 1.1 : 1.
In these experiments the investigators did not take any
records of the spoken sounds, but only of the rhythmic
strokes of the hand.
Since even in scanning the syllables do not have simple
relations of length, it is justifiable to conclude that in nat-
urally spoken verse the relations differ even more widely
from the theoretical ones. Indeed, what is known of the
psychology of human action makes it quite incredible that
any such simple relations as 1 : 2, etc. ever occur (or have
occurred) regularly in actually spoken verse.
The problem of where the stroke of the hand occurs in beat-
ing time to verse was investigated by Mbyee,^ with the
purpose of determining the position of the thesis in rhythmic
articulation. For recording the voice he used a mouth
trumpet ending in a Maeey tambour (p. 219) covered with
a fine rubber membrane to which a small straw lever ending
in a light pointer was attached. The beat of the finger was
made on an apparatus comprising a plate of hard rubber con-
nected by a string to a time marker (p. 91). The subject
1 Hurst and McKat, Experiments on the time relation of poetical meters, Univ.
of Toronto Stud., Psychol. Series, No. 3, 1899.
2 Meyek, Beitrage zur deutschen Metrik, Neuere Sprachen, 1898 VI 1, 121.
SPEECH RHYTHM ■ 539
recited some syllables into the tambour through the trumpet,
while he beat time on the rubber plate. Thus the breath
curve and the moment of beating could be recorded simul-
taneously on the smoked drum. In all cases, except where
the syllable began with a sonant explosive (b, d, g), the beat
came before the vowel. Both the tambour and the beating
apparatus used in the experiment had considerable latent
times which could be only roughly estimated to be about
0.008^ for the former and 0.025' for the latter.
More accurate experiments have been made by Miyake.^
The subject spoke into the voice key described above (p.
154). As the light diaphragm of platinum vibrated very
easily at a short distance from the mouth, it recorded the first
vibration of the voice with a latent time of not over half a
thousandth of a second. The voice key was put in one of
the circuits of a lamp battery (p. 210) and a Deprez marker
(p. 92) in the other. The latent time of the marker was
less than I'', as had been previously determined by frequent
tests (p. 92). For the beating apparatus the noiseless
key in a rubber bag (p. 529) was used. The tension of the
key was very small and the slightest touch was enough to
overcome the resistance for breaking the contact; the time
lost in compression of the finger before the key acted was
infinitesimal. The key was connected to the primary circuit
of a spark coil (p. 12) while the metallic point of the Deprez
marker was attached to one pole of the secondary circuit
(p. 630). The arrangement for drawing the time line was
the usual one of a 100-fork (p. 15). The drum was run by
a motor with a storage battery ; a very constant speed was
attained.
The subject held the voice key in his hand and, putting its
mouth-piece close to his lips, recited a syllable in a scanning
manner, while he beat time on the noiseless key with a
finger of his right hand (generally the index finger), the rate
of the recital being left to his choice.
1 MiTAKE, Researches on rhythmic action, Stud. Yale Psych. Lab., 1902 X 39.
540 FACTORS OF SPEECH
The following syllables were used by different subjects:
(1) a, (2) ,a, (3) ma, (4) ha, (5) pa, (6) ap, (7) ap, (8)
mam, (9) mam. In these the a was pronounced like a in
' father. ' The a had the usual smooth English entrance (p.
429). The 'a was the same as a, but with a slight glottal
catch at the beginning (p. 278). Both a and a were the
same in quality, but a was shorter than a, as the sign indicates.
All the consonants were pronounced as in English.
A summary of the results of the experiments is given in
the following tables. The positive signs indicate the devi-
ations when the beats of the finger came before the vowel,
and the negative ones those when the beats came after the
beginning of the vowel. The fourth and fifth columns in
the first table give the number of the cases in which the
positive and negative deviations occurred.
Sdmmakt for Sounds
Average time of beat Number of -.t u i i xt u n
before vowel. measurements. Number of +. Number of-.
ma +132 210 209 1
pa
+ 143
'206
205
1
ha
+ 118
190
187
2
>a
+ 131
170
90
0
a
+ 52
20
107
12
ap
+ 59
100
92
8
ap
+ 52
90
65
25
mam
4 57
80
80
1
mam
+ 62
80
78
1
Unit of measurement, o- = 0.001'
Summary
FOR Individuals
Subject.
ma pa ha
'a
a
ap
ap
mam mam
K
+ 76 + 96 + 56
+ 19
+ 30 +10
+ 54 +60
E
+ 120 + 140 + 103
+ 132
+ 88 +94
+ 61 +64
T
+ 169 + 181 + 133
+ 86
M
+ 163 + 157 + 172
+ 130
Unit of measurement, a = 0 001'
The tables show that the beat of the finger comes before
the beginning of the vowel under the following conditions :
,(1) when the vowel is preceded by a consonant and is not fol-
lowed by any other sound; (2) when the vowel has the glottal
catch at the beginning; (3) when the vowel is neither pre-
SPEECH RHYTHM 541
ceded nor followed by any sound; (4) when a short vowel is
followed by a consonant; (6) when a long vowel is followed
by a consonant; (6) when the short vowel is preceded and
followed by consonants; (7) when the long vowel is preceded
and followed by consonants. The conclusions, of course, are
valid only for independent syllables, but they probably apply
— with some modification — to those in connected speech.
It will be observed also that the amount of time by which
the beat occurs before the beginning of the vowel is not the
same for the different combinations in which the vowel stands.
The results for the subjects ^and ^show that the length
of time by which the beat occurs before a, when not preceded
by a consonant, is considerably shorter than that before the
vowel when preceded by a consonant. This fact indicates
that the consonant lengthens the time between the beat and
the beginning of the vowel.
The amount of time between the beat and the beginning of
the vowel differs with the different consonants which precede
it. The subjects K, E, T all agree in making this differ-
ence greatest in pa, next greatest in ma, and least in ha.
The amount of time by which the beat is ahead in 'a is not
very different from that in ma, pa and ha. It is probably
due to the fact that the glottal catch at the beginning of the
vowel is of the same nature as a consonant in so far as the
complexity of action of the vocal organs is concerned.
The results for mam and mam seem to indicate, if not in
a very conclusive manner, that when a vowel is preceded as
well as followed by a consonant, the beat tends to come
nearer the beginning of the vowel than when the vowel is
preceded by a consonant but not followed by another.
The preceding observations show that the finger beat occurs
before the vowel. But where does it come in respect to a
consonant which precedes the vowel?
Among the three consonants m, p and h which formed the
objects of the experiments in combination with the vowel a,
the last two (p and h) could be found in the records. The
curve for the consonant in the records did not consist of
542
FACTORS OF SPEECH
vibrations like those of the vowel, but of a smooth deviation
from the record line, due to the air pressure. The lengths of
the consonants could thus be measured. The results of the
measurements may be summarized as follows:
Finger beat with ha
bject.
Average
length
oJh.
Immediate
absolute
probable
error.
•Average time
of beat
before h.
Immediate
absolute
probable
error.
Number
of meas-
urements.
Number
of+.
Number
of—.
E
39
6.7
+ 51
9.6
30
30
0
K
116
15.5
-62
23.8
50
1
48
Finger beat
WITH pa
bject.
Immediate
Average absolute Average time
length probable of beat be-
ef p. error. fore p.
Immediate
absolute
probable
error.
Number of
measure-
ments.
Num-
ber
of+.
Num^
ber
of— .
E
48
6.2
+ 69
153
50
50
0
K
58
15.3
+ 15
16.5
60
42
17
Summary
bject.
Average
length of p.
Average time
of beat before p.
Average
length of h.
Average time
of beat before h.
E
K
49
59
+ 69
+ 15
39
117
+ 51
— 61
Unit of measurement, (r = 0.001'.
The following points may be observed in the tables : 1. for
the syllable pa two subjects agree in beating time before the
beginning of the consonant; 2. for the syllable ha, with the
subject U the beats come constantly before the beginning of
the consonant, but with jfiTthey come in most cases after the
beginning of the consonant, about midway between the conso-
nant and the vowel which follows it.
From the observations reported in this investigation the
final conclusion can be drawn that the beat of the finger in
connection with the rhythm of speech comes before the vowel
and before or in the course of the consonant which precedes
the vowel.
MiYAKE adds the following observations concerning the
point of emphasis in rhythmic articulation. The question is
first raised as to the relation of the beat of the finger to the
point of greatest emphasis. Our experience seems to show
that when we recite a verse while we beat time with the
SPEECH RHYTHM 543
hand, the point of the highest emphasis in the rhythm comes
at the same moment with the beat.
Although it is not certain whether the innervations of the
movements of the hand and vocal organs proceed from their
nervous centers at exactly the same moment, still we may
suppose that the two movements are so closely associated
that the innervations take place almost simultaneously.
But when we attempt to determine the position of the point
of emphasis from the beat of the finger, we find that it cannot
be easily done. It does not follow that the movements
themselves are executed at the same time from the mere sup-
position that innervations of the movements of hand and vocal
organs take place simultaneously.
Meyer ^ supposed that the movements of hand and vocal
organs would take place at the same moment, provided the
nerve fibers which transmit the impulses are equal in length.
He calculated from the rate of nervous transmission that the
impulse reaches hand aboat 0.015' later than vocal organ.
Adding the latent time of the apparatus to this lost time of
nerve transmission he arrived at the final conclusion that the
point of emphasis lies in the course of a sonant consonant or
shortly before it when it follows an explosive.
The diiference in the length of the nerve fibers is not the
only factor which disturbs the simultaneity of the two move-
ments. KiJLPE's experiments ^ showed that we have difficulty
in moving the hands at the same time to react to a single
stimulus. If even the two hands — ■ alike in construction and
symmetrically arranged — are not moved simultaneously, it
must be still more difficult to execute the movements of two
disparate organs like the hand and vocal organs at the same
moment.
Besides these differences there may be several other factors
which cause the deviation of the two movements. The differ-
ence in the complexity of the construction in the two organs
1 Meyer, Beitrage zur deutschen Metrilc, Neuere Sprachen, 1898 VI 121.
■' KtJLPE, Ueber die Gleichzeitigkeit von Bewegungen, Philos. Stud. (Wundt),
891 VI 514.
5i4 FACTORS OF SPEECH
might be one of such factors. The condition of attention
during the movements might be another. Therefore, until
we know all the conditions on which the simultaneity of the
two movements depends, we cannot state exactly the relation
of the point of emphasis to the finger-beat although we may
conclude that they do not differ by more than a few hun-
dredths of a second. ' Simultaneous,' like other terms, de-
pends for its meaning on the degree of accuracy involved ;
for most purposes we may regard the finger beat as simulta-
neous with the point of emphasis (centroid of speech effort,
p. 428) although the relation may vary in different indi-
viduals and in varying circumstances.
If we assume, then, that the movements of the hand and
vocal organs are executed simultaneously, we can conclude
from the foregoing experiments that the point of emphasis in
rhythmic speech comes before the vowel and before or in the
course of the consonant which precedes the vowel. In other
words, the point of emphasis in rhythmic articulation lies
at the beginning of the movement of the vocal organs for the
production of the sound.
The length of time between two points that are felt to be
emphatic may be studied by the following method. The
speech is recorded on a phonograph in the usual way (p. 32).
A contact wheel is then prepared by under-cutting the teeth
of a gear wheel so that the bases of the spaces are larger than
the tops, filling the spaces with rubber, vulcanizing it, and
turning it true. This wheel is placed on the axle of the
phonograph and a contact brush is applied to its edge. When
an electric current is sent through it, the circuit is closed
as each tooth passes under the brush. A magnetic counter
(EwALD chronoscope, p. 162) placed in the circuit will indi-
cate the passage of each tooth as long as the current is sent
through it. Two shunts are placed across the circuit, one
around the counter and the other around the contact wheel.
Each shunt contains a key held down in contact by the
finger. On releasing the former key the counter begins to
register the teeth of the contact wheel as they pass the
SPEECH RHYTHM 645
brush; on releasing the second key it stops. A simple calcu-
lation from the number of teeth in the wheel and the speed
of the phonograph gives the time between the action of the
two keys.
This method I have used to measure the intervals between
what I felt to be the points of emphasis in a specimen of
French verse. By measuring the so-called ' feet ' separately,
in twos, threes, etc., an idea could be obtained of the con-
stancy of the method ; there was rarely any difference between
the total time for two or three feet and the sum of the times
for the separate feet.
For the first stanza of Le sonnet d'Arvers spoken by M.
Maitrb I obtained the results indicated by the figures over
the lines in the following scheme. By measuring the time
from the last point of emphasis of one line to the first of the
next the length of the final pause and the following initial
syllable was obtained; by special measurements the length
of this initial syllable was obtained separately ; the pause was
found by subtraction.
I < 255 > I < 117
< . . 3 . . > I < . . . 135 . . . > I < . . . . 50 . . . . > I < . . . . 70 > I < ... 73 ... >
He'las! j'aurai passe prea d'elle inapercju
>|< 212 >|< 110
< . . 28 . . > I < . . . . 87 . . . . > I < . . . . 53 > I < .... 72 > | < . . . . 85 . . . >
Sans cesse a ses cotes et toujours solitaire
> I < 217 > I < 112
<,.25..> I <...75...> I < 68 > I <. ..58...> I < 51 >
Et j'aurai jusqu'au bout fait men temps sur la terre
■ > < •
183 >
< . . . . 50 . . . . > I < 57 > I < 60 > I < 67 >
N'osant rien tlemander et n'ayantrien reyu.
Unit of measurement, 0.01^.
Let US call a point at which the emphasis is felt to be
located a ' centroid ' ; and the time from one centroid to the
next a ' centroid interval.' The centroid intervals within a
line give the averages a^ = 0.85% a^ = 0.71% as = 0.72% a* =
0.61' for the respective lines, and a = 0.72' for all four lines.
The intervals from the first centroid to the last one in the
same line are 5. = 2.55% h = 2.12% b, = 2.17% h = 183% and
b = 2.17' for all four. The initial intervals for each line are
35
546 FACTORS OF SPEECH
ci =: 0.02% c^ = 0.28S (73 = 0.25% c^ = 0.50% and for all four
c = 0.34^ The intervals from the last centroid in one line to
the first in the next are d-^ = 1.17% d^ = 1.10% dz = 1.12% and
for the three cases d — \ .13'. The pauses between the lines
measure ei = 0.73% e,, = 0.85% es = 0.51% and for all three
e = 0.70^ The average intervals for the line including the
internal centroid intervals and the final pause are /i = 0.82%
/, = 0.73% /a = 0.62% and for the three /=0.72=. The
lengths of the lines up to the last centroid are gi = 2.58%
g2 = 2.40% ^3 = 2.42% ^, = 2.13% or for all four g = 2.38^
The lengths of the lines including the final pauses are (A, =
g, + e, [i = 1, 2, 3]) h, = 3.31% h, = 3.25% h = 2.91% and
for all three h = 3.15'. The intervals between the last cen-
troid in one line and the last in the next are k^ — 3.29% Jc^ —
3.28% ^3 = 1.95% and for the three cases k = 2.13'.
When one phenomenon occurs more regularly than another,
it may be assumed to be more characteristic of subject investi-
gated. On this principle we may draw several important
conclusions concerning this specimen of verse. Since c?, =
«i + <^i (j- = li 2, 3) and since di is very constant, the effective
element in the passage from line to line is the time between
final and initial centroids ; the filling of this passage-interval
is made up of pauses (which are effective elements both in
movement and to the ear) and sounds adjusted to each other.
The facts that d > a, that a often shows considerable varia-
bility within a line, and that the length of the line (^ or A) is
quite constant, seem to indicate that the line is itself felt as a
unit in the composition. The fact that g and h are almost
equally constant (that is, have the same probable errors) indi-
cates that a line may be considered as ending either with the
last centroid or with the completion of the pause.
The rhythm in> speech may be studied by an analysis of the
speech curve.
A short portion of the speech curve of the Cock Robin
record (p. 58) has been studied in reference to the elements
of rhythm; the results for the first stanza are given in the
following tables.
SPEECH RHYTHM
547
The first column gives the sounds in the phonetic transcrip-
tion used in this book. The second column gives the dura-
tion of each sound as determined by measurements of the
Line 1 : • Who killed Cock Bobin ?
■9 So
ijf
•S 0)
1
1
Remarks.
0
u o n
Cm"
W
a
^
h
>I0
Very short sound, not distinguishable in
the record, not over XO" in length.
u
189
3.3
0.4
strong
Forcible vowel, large amplitude in earlier
portion, rises somewhat in pitch, average
period 3.3.
k
119
Appears in the record as a straight line.
i
154
1.8
0.6
strong
Long vowel, large amplitude throughout,
double circumflex in amplitude. The high
pitch of this i is in contrast with that of
' killed ' in the 4th Line (below).
1
74
1 8
0.1
(d)
0
No sound of d can be heard in this record;
the record plate speaks ' Who kiE Cock
Robin ^ '
k
53
Appears in the record as a straight line.
a
126
42
0.5
weak
Rises somewhat in pitch to 4.2 in the
main portion, weak on account of lowness in
pitch.
The vibrations of the a are suddenly cut
k
101
short by a few vibrations of a different form
that rapidly decrease in amplitude. In listen-
ing to the record plate the ear hears no glide
between a and k ; the word seems to be sim-
ply and distinctly kak and not kaak.
r
74
1.8
0.3
"Very distinctly and heavily rolled r ;
pseudobeats are apparent in the tracing.
a
140
5.3
0.5
strong
Of very low but constant pitch ; steady
rise in intensity till the vowel is cut short
by b ; forcible on account of length and
amplitude.
b
49
Straight line from a to i.
i
56
5.6
0.3
weak
Short but distinctly heard ; weak on ac-
count of shortness, lowness and faintness.
n
74
8.4
0.2
Falls in pitch and amplitude.
1
770
curves in the records. The third column gives the periods
of the cord tone, and the fourth gives the amplitudes of the
vibration in the tracing, not the amplitudes of the vibration
548
FACTORS OF SPEECH
on the gramophone plate or of the movement of the vocal
cords. The fifth column gives what I consider to be . the
character of the syllable in respect to being ' strong ' or
' weak ' ; the judgment is based on the sound of the gramo-
phone record, aided by the tables.
The analysis of the first stanza of Cock Rohin as given
Line 2 : /, said the sparrow.
a
-Si
5S
O
03
O
1
*>>
Remarks.
ai
452
18 to 4
0,7
strong
Strong by length, pitch of i and ampli-
"1
210
tude ; tracing given in Plate II.
s
1
Very brief sound, no trace in record.
e
105
5.3
0.5
weak
Rather long and loud, but low in pitch.
d
81
5.3
0.1
Pitch falls from 5.3.
■s
32
1
>0.1
Very weak vibrations.
3
84
5.3
0.2
weak
sp
273
Impossible to distinguish between the two
sounds in the tracing ; the s is heard as a
brief sound.
-?
18
1.9
0.4
Distinct sound different from the follow-
ing «e, exaggerated explosion of p ?
ae
170
5.3
0.5
strong
Constant very low pitch but steadily in-
creasing amplitude ; falls suddenly in inten-
sity during 5" to r; strong on account of
length and amplitude.
r
11
2.8
0.2
Clearly marked vibrations ; the rolling
of the r can be distinctly heard.
Very long vowel of constant pitch, but
o
294
5.2
0.6
strong
of rismg and then falling intensity ; strong
by length and amplitude ; followed without
pause by w of next Line.
Tracing of ' sparrow ' given in Plate I.
in these tables shows that it contains not only an intensity
rhythm but also a pitch rhythm and a duration rhythm.
The three elements — length, pitch and intensity — are all
used to produce strength. Thus the forcible vowel u in
Line 1 is short but moderately high and loud.
The strength of a syllable may be kept the same by increas-
ing one of the factors as another one decreases (p. 549). The
SPEECH RHYTHM
549
vowel a of ' Robin ' in Line 1 is strong on account of its
length and intensity, although its pitch is low. A syllable
necessarily short may be made as strong as a longer one by
making it louder or higher; or a syllable necessarily of small
Line 3 : With my how and arrow.
•S "a
11'
laf
all
"3 a)
|]
c
la
in mm.),
yllabic effect.
Remarks.
1
(5 =
&B
OQ
w
108
5.3
o.s
Amplitude rises from 0.
i
60
2.1
0.4
strong
Circumflex sustained vowel; strong by-
«
56
■!
0.1
pitch and amplitude.
m
74
5.3
O.I
a
179
5.6
0.4
1
> strong
Both parts of this diphthong are nearly
constant in pitch and amplitude ; strong by
i
112
3.6
0.5
)
length and amplitude.
S!
140
' My ' is followed by a brief rest in order
to bring out the b distinctly. The b makes
no curves in the record.
5
490
7.0
0.4
strong
Extremely long vowel of very low pitch
with two maxima of intensity ; it might
be considered as a close succession of those
o's ; strong by length and amplitude ; trac-
ing given in Plate I.
"1
11
382 1
7.7-5.3
5.3
The ae begins at a very low pitch 7.7 and
0.2
0.1
weak
rises steadily to 5.3, which is maintained
throughout the n. The form of the curve
for ae ditf'ers from that for n, yet the change
is so gradual that it is impossible to assign
any dividing line.
d
18
Straight line in the record.
9
102
5.3
0.4
'Jhis extra vowel arises from the attempt
at extra distinctness iu speaking.
ae
189
5.2
0.3
strong
Strong by length and pitch.
r
39
2.5 (?)
0,1
Kolled r, brief
0
331
7.0
0.6
strong
A single vowel of circumflex intensity,
strong by length and amplitude.
1
420
intensity may be strengthened by lengthening it or raising its
pitch. Thus, the short i of ' with ' in Line 3 is strong on
account of its high pitch and large amplitude ; and the weak
ee of ' arrow ' in Line 3 is strong on account of its high
pitch and its length. This might be called the principle of
substitution.
550
FACTORS OF SPEECH
It appears necessary to attempt some interpretation of the
foregoing experiments on speech rhythm. As previously
Line 4 : / killed Cook Robin.
■a
11-
S 3'=
1
Remarks.
ai
334
12-4
0.6
strODg
Strong by length, pitch of i and ampli-
tude.
k
125
Straight line in the record.
i ;
It is impossible to assign any definite point
i
324
5.6
0,2
weak
as the limit between these two sounds ; weak,
1^
low i in contrast to the i in the first Line
above.
d
33
This d is distinctly heard ; compare d in
first Line above
9
81
4.9
0.2
Additional vowel due to the extra dis-
tinctness in speaking the d ; it arises from
the explosive opening of the mouth ; the pro-
nunciation of the word ' killed ' is different
from that in the first Line chiefly in the great
difference in pitch and in the greater distinct-
ness of the d.
k
133
Straight line in the record.
a
147
7.0-5.3
0.3
weak
Pitch rises from beginning to end.
k
122
See the same word in the first Line above.
J
60
3.9
0.6
The r is more vowel-like than the corre-
sponding r in the first Line ; the strong roll
is not heard ; the curve of xa. very much re-
sembles in period and amplitude the curve
of an ai in ' thy ' turned backward ; the
strong
period of the cord tone is practically con-
stant ; the resonance tone of the mouth un-
dergoes a continuous change ; any assignment
of a limit between the two sounds must be
somewhat arbitrary ; Ja is apparently a
rising diphthong; the sound ja is strong
a
103
3.9
0.5
by length, pitch and amplitude.
b
53
4.2
0.1
The b cuts off suddenly the sound of a.
i
82
5.6
«...
0.0
The i is heard, but not so distinctly as in
weak
the first Line above.
n
74
8.8
Weak, low, diminuendo.
-1
955
indicated (p. 447), speech is a flow of auditory and motor
energy with no possibility of division into separate blocks
such as letters, syllables, words, feet, etc., except in a
SPEECH RHYTHM 551
purely arbitrary manner that does not represent the actual
case. To the speaker and the hearer this flow may be treated
in its rhythmic effect as a series of centroids (p. 451). This
is the basis of all comparisons of verse with rhythmic clicks
and with rhythmic movements. In prose the centroid is the
place at which the whole effect of accent can be placed;
the factors that make accent are those that locate the
centroid. In verse the centroids are located by the prose
effect, and also by the rhythmic swing of the movement of
the verse form itself. Experiments with beating time to
verse are attempts to locate the centroids.
Spoken language is usually classified as prose or verse. It
might, perhaps, be better to say that in spoken language there
are certain elements that may be present in greater or lesser
degrees; the forms with little of these elements are termed
' prose ' while those with more of them are termed ' verse, '
without the possibility of always making a sharp distinction.
These elements include rhythm, melody, and probably also
agreeableness of quality, etc. The speech of some persons
appears to be totally lacking in melodiousness ; the voice has
no tunefulness, the words come out without inflection and
there is no regular distribution of the emphatic elements in
a sentence. Other persons naturally speak even the simplest
sentences in a melodious manner; the pitch of the successive
syllables rises and falls pleasantly and the points of emphasis
are evenly distributed. This may go so far that the sponta-
neous utterances of a speaker possess all the charm of blank
verse, or may even be indistinguishable from it. An exam-
ple of such speech can be found in Jefferson's rendering of
Rip Van Winkle's Reverie ; it is contained on the gramophone
plate numbered 699. This prose speech possesses all the
beauty of verse without rime. The plate has not yet been
traced off and the elements that produce the melodiousness
are still undetermined.
As a model of vocal rhythm we may assume that the series
of vocal sounds is divided into relatively large portions of
equal lengths, or ' measures ' ; these measures are divided either
552 FACTORS OF SPEECH
into two portions, thesis and arsis, bearing simple relations
of length, or into a number of small equal portions termed
' morae ' or XP°^°'' "^pSnoi. The relations of length between
thesis and arsis will be as 2 : 1, 1 : 1, etc. Such a mathe-
matical relation was called by the Greeks prjTO^, which has
been translated as 'rational.' A rhythm with such simple
relations of length may be called a ' rational rhythm,' ^ or,
perhaps preferably, an abstract rhythm. In vocal music we
would expect to find the nearest approach to the relations of
abstract rhythm, although even here measurements will show
that the actual relations are not exact or constant.
The relation 2 : 1 for thesis and arsis does not occur, ex-
cept by chance, in spoken verse ; verse rhythm is ' irrational.' ^
The traditional doctrine that a long syllable has twice the
length of a short syllable rests upon a not clearly understood
application of musical expressions to actual speech.^
In prose the various syllables have lengths that depend on
general usage, on accent, emotion, etc. In verse these ' natu-
ral ' lengths may be more or less modified. The actual con-
crete rhythm of a particular piece of verse is a compromise
between the natural lengths and those required by abstract
rhythm.*
According to SrEVBES,^ modern spoken verse has properly
only one kind of time-division, the foot. In respect to time
it cannot properly be divided into smaller units, or morae ; and
there is no definite relation between thesis and arsis. The
syllables are lengthened or shortened so that the desired time
is occupied by a foot. An attempt to speak modern verse
with regard to the lengths of the syllables at once destroys
its character as verse and turns it into a hybrid thing known
as ' scanned verse.'
By foot SiBVEES appears ^ to mean the time from one mini-
1 SiEVEES, Metrische Studien, I., Abhandl. d. k. sachs. Ges. d. Wiss., philol.-
hist. Kl., 1901 XXI 34.
2 Aristoxenus, see Goodell, Chapters on Greek Metric, Ch. II, New York,
1901 ; SiEVERS, as before, 41.
3 SiEVERS, as before, 41. i Sievers, as before, 42.
■'■' SiEVERS, as before, 43. 6 Sievees, as before, 49.
SPEECH RHYTHM 653
mum of speech energy to the next ; for reasons apparent in
Ch. XXX it would seem preferable to define the foot as the
time between two centroids of speech energy.
In place of the unit-times of abstract rhythm we have to
consider the number of syllables per foot in verse. Modern
so-called iambic and trochaic forms of verse have two syl-
lables per foot, anapestic and dactylic three. The lengths of
the syllables in no wise enter into consideration.
The time of a foot is approximately constant. When a
two syllable foot occurs in the midst of three syllable feet, it
takes approximately the time of the others ; and contrariwise.
The simplest English poetical line seems to consist of a
quantity of speech-sound distributed so as to produce an
effect equivalent to that of a certain number of points of
emphasis at definite intervals.
The location of a point of emphasis is determined by the
strength of the neighboring sounds. It is like the centroid
of a system of forces or the center of gravity of a body in
being the point at which we can consider all the forces to be
concentrated and yet have the same effect. The point of
emphasis may lie even in some weak sound or in a surd con-
sonant if the distribution of the neighboring sounds produces
an effect equivalent to a strong sound occurring at that point.
Thus the first point of emphasis in the third line lies some-
where in the group of sounds ' mybow,' probably in 'b'
between ' y ' and ' o. '
With this view of the nature of English verse all the
stanzas of Oock Rohin can be readily and naturally scanned
as composed of two-beat (or two-point) verses. The proper
scansion of the first stanza would be :
Who killed Cock Eobin?
« •
I, said the sparrow,
With, my bow and arrow
I killed Cock Eobin.
554 FACTORS OF SPEECH
The study of this and other specimens of verse has made
it quite clear that the usual concept of the nature of a
poetical foot is erroneous in at least one respect. Lines in
verse are generally distinct units, separated by pauses and
having definite limits. A single line, however, is not made
up of smaller units that can be marked off from each other.
It would be quite erroneous to divide the first stanza of Cock
Rohin into feet. No such divisions occur in the actually
spoken sounds and no dividing points can be assigned in the
tracing. In fact there does not seem to be any system of
feet that can be assigned to it or any form of such rhythm
under which it can be classified. Yet it is felt distinctly as
verse and has been a factor in forming the rhythmic feeling
for verse in the history of our language and in individuals.
Treated on the centroid method, it is perfect verse with two
centroids to the line.
In verse the centroids are grouped. One group may form
the line, another the stanza, etc., or, in blank verse, the
various phrases may be made into more or less complicated
groups.
As in all forms of rhythm (p. 523) there is a tendency
to arrange the centroids regularly. Thus, in the Sonnet
d'Arvers (p. 545) the internal centroids come at about 0.72^
apart, the line centroids at about 2.13^
A centroid analysis of this stanza would be approximately
Helas! j'aurai passe pres d'elle inaperQu,
• • • •
X X
Sans cesse a ses cQfces et toujours solitaire,
• • e •
X X
Et j'aurai jusqu'au bout fait mon temps sur la terre,
" • • •
X X
N'osant rien demander et n'ayant rien recu.
• • • •
X X
SPEECH RHYTHM 555
The small dots • indicate the primary centroids (or ' foot
centroids '), the crosses x indicate the phrase centroids, and
the large dots • the line centroids. The positions of these
centroids might be determined by having the listener beat
on a telegraph key at the moments he feels them to occur. I
have indicated them only approximately without making
measurements. Influenced by the rhyme and the long pause,
I feel the line centroid to occur during the last word in each.
The distribution of energy around the centroids differs in
different cases, but there are certain typical forms of distri-
bution to which definite names have been given, as iambic,
trochaic, etc.
In much English verse there is little or no regularity of
distribution ; just so many centroids are grouped into a line
and there is no feeling for the distribution of energy between
centroids. In a stanza like
The Cities are full of pride,
• • •
Challenging each to each —
• • •
This from her mountain-side,
• • •
That from her burthen ed beach
• •■ •
(Kipmstg)
it would be quite a mistake to say that the meter is iambic,
anapestic, trochaic or dactylic. Even if this particular stanza
might be said to be mainly dactylic and trochaic, the follow-
ing ones change constantly. In the minds of both the speaker
and the hearer the only rhythmic essential lies in the presence
of three beats to a line. Such verse should be called 2-beat,
3-beat, 4-beat verse, etc. ; an attempt to force on verse of
this kind the classical schemes cannot have the slightest
justification.
In other cases of English verse there is a careful (generally
unconscious) regularity in the distribution of energy. Thus,
556 FACTORS OF SPEECH
And the stream will not flow and the hill will not rise,
And the colors have all passed away from her eyes
(Mooee)
shows clearly a regular distribution of energy with a slow
rise and a sudden fall. Again,
Merrily swinging on brier and weed
(Bryant)
appears to rise suddenly and fall slowly. A moderately slow
rise with quick fall appears in
The way was long, the wind was cold
(Scott)
and a quick rise with moderately slow fall in
Fifty times the rose has flowered and faded.
(Tenn-yson).
These are specimens of the types known as anapest, dactyl,
iambus and trochee. They are distinct types, although the
poet may sometimes mix his forms and often lapse to the
form of simple beat verse. In some specimens of English
verse the distribution of energy around the centroids is very
carefully elaborated so as to produce effects analogous to
those of classical verse. In still more highly developed
forms of verse the poet may use relations of duration among
the syllables to produce a truly quantitative verse.
Concerning the influence of rime and alliteration in
establishing centroids, and concerning the relations between
the natural prose rhythm of a portion of speech and the verse
rhythm into which it is fitted, we have no experimental data.
References
For verse in general : Aristotle, Treatise on Poetry; Mitford, In-
quiry into Principles of Harmony in Language and of the Mechanism of
Verse, Modern and Ancient, London, 1804; Pierson, Metrique naturelle
SPEECH RHYTHM 657
du langage, Paris, 1884 ; Sylvester, Laws of Verse, London, 1870 ;
Grimm, Zur Geschichte des Reims, Kleiners Sohriften, IV, Berlin, 1887;
Meumann, Untersuchungen zur Psychologie mid Aesthetik des Rhylhmus,
Philos. Stud. (Wundt), 1894 X 249, 393 (full references) ; Kawczynski,
Essai oomparatif sur I'origine et I'histoire des rhythmes, Paris, 1889;
BiJCHEK, Arbeit u. Rhythmus, 3. Aufl., Leipzig, 1902. For classical
verse : Goodell, Chapters on Greek Metric, New York, 1901 (references
to work of classical and modern authors will be found here) ; West-
PHAL, Griechische Khythmik, 3. Aufl., Leipzig, 1885. For English verse :
Guest, History of English Rhythms, new ed., London, 1882; Schipper,
Englisohe Metrik, Bonn, 1882-89 ; Lanibr, Science of English Verse,
New York, 1880; Poe, Rationale of English Verse, Works, VI, 84, Chi-
cago, 1875 ; Goodell, Quantity in English Verse, Trans. Amer. Philol.
Assoc, 1885, XVI, 78. For German verse : Bruoke, Die physiologischen
Grundlagen d. neuhochdeutschen Verskunst, Wien, 1871 ; Minor, Neu-
hochdeutsche Metrik, 1893; Saran, Ueber Hartmann von Aue, Beitr. zur
Geschichte d. deutschen Sprache u. Literatur, 1898 XXIII 42 ; Sievers,
Zur Rhythmik u. Melodih d. neuhochdeutschen Sprechverses, Verhandl. d.
XLII. Versammlung deutscher Philologen u. Schulmanner, 370, Leipzig,
1894; Sievers, Metrische Studien, I., Abhandl. d. k. sachs.Ges, d. Wiss.,
philol.-histor. Kl., 1901 XXI No. 1 ; Westphal, Theorie der neuhoch-
deutschen Metrik, 2. Aufl., Jena, 1877. For French verse : Lubarsch,
Franzosische Verslehre, Berlin, 1879; Kressner, Leitfaden d. franz.
Metrik, Leipzig, 1880; Passy, Les Sons du Franqais, 5™ ed., Paris, 1899.
APPENDICES
APPEIS^DIX I
FOURIER ANALYSIS
An understanding of the underlying principles is desirable but
not necessary for the use of the Foubibr analysis. The follow-
ing exposition to the words ' In practice,' on page 566 may be
omitted if desired.
The theory and practice of the analysis have been well pre-
sented by Hermann.*
According to Fourier's theorem any periodic function of t
may be represented as the sum of a series of harmonic sine and
cosine functions as follows :
o o o
y = i^o + «! • cos -^t + a^. cos -^,t + as . cos ~t + ■ ■ •
+ bi . sin Y^ + ^2- S"i rjf^ + *8 • sin — ,« + • • •,
where a and & are the amplitudes of the harmonics, T the period
of the lowest harmonic, and ^a^a. constant expressing the dis-
tance of the curve above the t axis.
By putting Vot.^ + *i^ = "n '^'^2^ +*2^ = c^, . . .,
and T — *^'^ S'l' T — *^° S'^' • • •'
we get
2/ z= Jao -f ci . sin r -^< + ffi j 4- C2 . sin (^t + qA + ■ ■ ■
We thus have the curve expressed as the sura of a series of sine
harmonics, that is, of sinusoids with periods that are sub-mul-
tiples of the longest one. The quantities qi, q^, . ■ ■ indicate
the different phases of the sinusoids.
1 Hekmann, Phonophotographische Untersuchungen, IT., Arch. f. d. ges. Physiol.
(Pfluger), 1890 XL VII 45.
562 APPENDIX I
Such a series can well be used to find the partials in the curve
of a violin tone or of any sound of a similar nature. When an
empirical curve of this nature is given, the periods and ampli-
tudes of the sinusoids may be found in the way described here.
After the ordinates have been measured (p. 74), the computa-
tion may be carried on as follows. The length of one period
(group) is divided into n parts of h units each ; the length T of
the period is thus nh. In the Fouriee series as expressed
above we substitute distance along X for time, whereby nh ^= T
and x=t. Thus
2ir Itt 1ir
y = ^a„ H- a-i . cos —x + a^ . cos -J-ya; + Sg . cos o— r^; + • • •
Itlt IVth Ihfh
2ir , . _27r , . „2ir
+ 6, . sm —rx + 02 . sm 2—^x + 0, . sin 3— =-x -|- • • •>
nh nh, nh
wherein the constants ao) <ht ^ii o-^t 621 ■• • are determined by
,+ !i
the equations
2 r^'T 277 , 2 /""" 2,r ,
a_ = ^ I ^ cos r— ra; ax = — - I V cos r^x ax,
nhj „j ^lA nhj a nh
~T
2
, y sin r— ;^a; <ia;
nhJo nh
(r = 0, 1, 2, . . . .).
These constants are to be approximately determined from a
limited number of measurements of y for a number of values
of x. We possess the set of values Xo, yo ; Xi, yx; ^21 Vi'i ■ ■ ■
from which we must calculate «o. Oi, 61, a^, b^, • • • The
results will come nearer the truth the greater the number of
ordinates included, being absolutely true only for an infinite
number. For a finite number the integral equations are changed
to sums in which the successive values of x run through the
series Xo = 0, x^ = h, x^ — 2h, . . ., whereby the successive
values differ by h, or dx — h. We substitute h for dx as the
expression is changed from an integral to a summation, and
cancel h in both numerator and denominator. Another simpli-
fication arises by considering the whole period measured to be
equal to 2ir. For each value of x the summation will reach from
a; = 0 to x= {n — l)h, and the expression vh is substituted
for X, whereby we mean that the summation is to be extended
from v = Otov = ?i — 1; the upper limit of summation is thus
FOURIER ANALYSIS 563
,. = M — 1. We likewise indicate by y^ the corresponding values
of y. We have finally
v = n — 1
2'
ar = - ^ 2/„ . cos rvh,
id likewise
1
' = 0
V
n •<
= re-l
1^ y- ■ sin
= 0
0, 1, 2, . .
•)
The successive
coefficients
are thus
«o = -( 2/o-cosO + 2/1. cos 0 + 2/2. cos 0 + (-y„-i.cosoY
2/
«! = - ( 2/o • cos 0 + 2/1 . cos A + 2/2 . cos 2A + • • • ■>
** \ + y„-i.cos (n — l)h\
2/ '^
Ji = - ( 2/o • sm 0 + 2/i . sin A + 2/2 . sin 2A + • ■ • s
""^ +2/„-i.sin(«-l)AV
2/
«2 = - ( 2/o • cos 0 + 2/1 . cos 2/i + 2/2 . cos 4A + • • • v
'^^ +2/„-i.cos2(»i-1)aV
2 /
62 = - I 2/0 • sin 0 + 2/1 . sin 2A + 3/2 . sin 4/i + • ■ • ^
'*^ -4- « .-sin2(w-l)A ],
+ y„_i.sin;
2 /
ftj = - ( 2/0 ■ cos 0 + 2^1 . cos ill + 2/2 • cos 2ih + • ■ • \
'*\ +2/„-i.cos(n-l>'Aj,
2/
6,- = - ( 2/0 • sin 0 + 2/1 • sin iA + 2/2 • sin 2iA + • • • \
'*\ + 2/„_i.sin(?i-l)iA V
By inserting in the original equation the constants of ampli-
tude thus obtained the original curve is resolved into a series
of cosine and sine harmonics. By uniting the two series as
indicated above, the resolution into a single set of harmonics is
obtained.
Such an elaborate computation would be ordinarily quite im-
564 APPENDIX I
practicable without various shortenings, of which Hermann has
suggested a number.
Let the number of ordinates be w = 40, whence, since 2:7 = nh,
h = 9°. Then
tti = ^ (2/0 • cos 0° + 2/1 . cos 9° + 2/2 . cos 18° H h 2/39 ■ cos 351°)
h + jV (2/0 • sin 0° + 2/i , sin 9° + 2/2 . sin 18° + • ■ ■ + 2/39 • sin 351°)
«2 = A (yo • cos 0° + 2/1 . cos 18° + 2/2 • cos 36° H h 2/39 • cos 702°)
*2 = A (2/0 • sin 0° + 2/1 . sin 18° + 2/2 • sin 36° + • • • + 2/39 • sin 702°)
«i = ^ (2/0 . cos 0° + 2/1 . cos [i X 9]° -I- 2/2 ■ cos [2i x 30]° + . . .)
*i = jV (yo • sin 0° + yi . sin [i X 9]° + ya . sin [2i X 30]° + • . •)
There are thus 40 multiplications for each a and b, or 1600 mul-
tiplications for the first 20 harmonics. But when n is a factor
of 360, the siaes and cosines repeat themselves (thus, cos 702° =
cos 342°), the values cos 18°, cos 27°, . . . are repeated a
number of times, and there are simple relations between sines
and cosines (thus, cos 9° = — cos 181° = — cos 189° = cos 351° =
sin 81° = sin 99° = - sin 261° = - sin 279°). The result is
that with w = 40 only nine trigonometric values are needed in
addition to 0 and 1, making a total of eleven.
These eleven values are then to be properly multiplied with
the forty values of y. To do this the products are arranged in a
table as indicated on this page.
Angle 0°
9°
18°
27°
36°
45°
54°
63°
72°
81°
90°
Cosine 1
0.99
0.95
0.89
0.81
0.71
0.59
0.45
0.31
0.16
0
y„ 0.99y„ 0.95)/„ 0 89y„ O.Sly„ 0.71 1/„ 0.592/„ 0,45l/„ 0.31 y, 0.16si„ 0
y, 0.99t/i 0.95?/i 0.89i/i 0.81 j/i 0.712/, 0.59l/i O.tSy^ 0 31 !/, 0.16j/i 0
j/2 0.991/2 0.95^2 0.892/2 OMy^ O.ny^ 0.592/2 0-452/2 0.31^2 O.I63/2 0
Vm 0.99 J/3S 0.95 2/39 0,895/39 0.81 j/39 0.71 yj,, 0.59^39 0.45 1/39 0.31 2/39 0.165(39 0
The smaller figures indicate those that are to be multiplied
and written by the experimenter. In the column 0° all the values
of the forty ordinates are written; in the column 9° all these
values multiplied by 0.99 are written ; etc. A multiplying table
(p. 70) is of great use.
To obtain ao the numbers in the column 0° would be added ;
this is not necessary for selecting the harmonics and is omitted.
FOURIER ANALYSIS 565
For cti add the first main diagonal thus + {y^ + 0.99«/i + OMy^
+ • • • + 0.16yo + 0) ; then the diagonal - {O.l&y^ + O.Slyi^ +
• ■ ■ + 2/20 ) ; then the diagonal — (0.9%2i + 0.95y22 + • • • 4-
O.I62/29 + 0) ; then + (0.16^31 + O.3I2/32 + • • • + Q-hy^^). The
diagonal sums are to be taken + and — as indicated ; the sign
changes always at the right-hand edge.
For a^ add the series + {y^ + 0.95?/i + O.8I2/2 + OMy^ + O.Sli/,
+ 0) - (O.3I2/6 + 0.592/7 + • • • + 2/i„) - (0.952/n + O.Slyi^ + ■ • •
+ 0) + (0.31yi, + 0.59yi, + • • • + y,,) + (0.952/21 + O.ily,^ +
• • • + 0) — etc. The successive jumps are like the knight's
move in chess.
For 0,3 add the series + (2/0 + 0.892/i + 0.592/2 + O.I62/3) —
(0.31?/, + O.7I2/5 + 0.952/6) - (0.99?/, + O.8I2/8 + QA5y, + 0) +
etc. The successive jumps omit two columns.
For a^, aj, . . . add similar series, with three omitted columns
for tti, four for a^, etc.
For bx add the series along the diagonal + (0 + O.I62/1 + 0.31^2
+ 0-.452/S + . . . + 2/i„) + (0.992/n + 0.952/^2 + • • • + Q.l&y^, + 0)
- (O.I62/21 + O.3I2/22 + • ■ • + 2/30) - (0.992/31 + ■ • • 0.16^3,) ; the
sign changes always at the right-hand edge.
For 62 the numbers are taken by skipping one column as for
«2 but on diagonals downward from right to left. For b^, bi,
. . . the procedures are as for a^, a^, . . . but along diagonals
in the direction for bi and b^.
Each of the values thus obtained would need to be multiplied
by ^g to give ai, a^, . . . b^, b^, . . ■ but, since only the relations
between the ordinates are desired, the constant multiplier is
omitted.
o
For 12 ordinates, w = 12 and A = — = 30°. Then
n
m = J(2/o . cos 0° + 2/1 . cos 30° + . . . )
b, = ^(2/0 . sin 0° + 2/1 . sin 30° + . . . )
«, = ^(2/0 . cos 0° + 2/1 . 00s (30t)° + . . .)
5, = J(2/„ . sin 0° + 2/1 . sin (30i)° + . . .)
For cos 0, cos 30°, cos 60°, cos 90° we have' 1, 0.87, 0.60, 0;
from which the values of all the cosines and sines are obtained
by prefixing the proper signs.
566 APPENDIX I
We thus have the scheme :
Angle
0°
30°
60°
90°
Cosine
1
0.87
0.50
0
y„ 0.87y„ O.BOyo 0
2/1 0.8Tj/i 0.501/1 0
j/2 0.871/2 0.501/2 0
j/u 0.87yu 0.50yu
For «! take the first main diagonal thus + (i/o + 0.87yi
+ O.5O2/2 + 0), then the diagonal — (0.502/4 + 0.87yj + 2/5),
then the diagonal — (0.872/7 + 0.502/s + 0), then the diagonal
+ (O.SOyio + 0.87yn), and add the results. For a^ take + (2/0
+ O.5O2/1) - (O.5O2/2 + 2/3 + O.6O2/4) + (O.5O2/5 + 2/a + O.5O2/7) -
(O.5O2/8 + 2/9 + O.5O2/10) + O.6O2/11. Proceed likewise for the other
values.
In practice the computation proceeds according to schedules
prepared once for all beforehand according to the preceding
principles. The use of such schedules can be illustrated by a
case with 12 ordinates.
The values of the 12 ordinates are written in the first column
of a table ; each is then multiplied by 0.87 to fill the second
column and then by 0.50 to fill the third.
The curve shown in Fig. 49 gives ordinates with values as in
column 1 of the following table. Multiplication by 0.87 gives
— after the last decimal is dropped — the values in column 0.87,
and by 0.50 those in the last column.
The computation proceeds according to the following schedule.
1 0.87 0.50
yn
0
0
0
y\
10
8.7
5.0
ya
31
27.0
15.5
Vs.
36
31.3
18.0
Si
30
26.1
15.0
ys
18
15.7
9.0
y<s
8
7.0
4.0
Vi
11
9.6
5.5
Se
26
22.6
13 0
ys>
35
30.5
17.5
yio
30
26.1
15.0
i/11
8
70
4.0
FOURIER ANALYSIS 567
SCHEDnLE FOE 12 Ordinatbs
(Multipliers : 1, 0.87, 0.50.)
"1 "2 «s "4 "6 Ob 6i ia 63 h h ^e
I'D + ■ • + ■ ■ + . . + . . + . . +
y^ ■ ■ ■ - + -• ■ + - +
Vi ■■- ■■- +•• ■■- ••- +•• • + • .- + . ._. ...
2"= -■ ■■+ - • + • -• ••+ -■ +-. .-. ..+ ...
ye -. . +.. _. . 4... _. . ^ y ___
S'7 ■-• ■ •+ - . + . _. . . ._ . + . _. . ... ...
ys . ._ . ._ ^... . ._ . ._ +.. ._. ,^, . . ._. ._!_. ...
3/9 ■•• - + -.. - + _
S'lO + •■— — •• ••— ••+ +■• • — + . . + . ...
m ■ + ■ • •+ — — -• • . — •-• -. . . — . ._ ...
The squares with + and — in the schedule for Oj indicate the
figures of the table that are to be added in order to give the
value 0.1. Thus, for a^ we have 0 -f- 8.7 + 15.5 + 15.0 + 7.0 —
16.0 - 15.7 - 8.0 - 9.6 - 13.0 = - 16.1. For b^ we have 5.0 +
27.0+ 36 + 26.1 + 9.0 -6.6 -22. 6 -35.0 -26. 1-4.0 = + 9.9:
The schedules for a^, a^, . . ., 61, b^, 63 . . . are used in like
manner. The values c^ = Va^^ + b^^ c^ — Va^~+T^, . . . give
the relative amplitudes of the partials.
With these schedules the values obtained from the table are,
for the curve under consideration :
Squares
gg + fr^
The nearest whole number is taken as the value of \/ «^ + ^'^■
The indices attached to the values for c give the serial numbers
of the partials. The second partial is thus the main tone present
in this curve.
To save the labor of consulting the schedules they may be re-
produced on a large scale as diagrams with black and white
squares, the white squares bearing the signs + and — as re-
quired. A piece of tracing paper is placed over the first diagram
and the values of the ordinates are written in the first column y.
The following column is then filled by multiplying (with Zim-
meemann's tables) each value of y by the number written above
it. The following columns are likewise filled. The result is a set
of figures written in squares. The transparent paper is now
placed over the diagram %. The numbers seen over the white
«1 *1
a^ *2
"3 *3
°i h
"6 6s
"a h
15.1 +9.9
-98.0- -7.0
-13.0 +8.0
—3.0 —1.8
+4.1 -1.1
+7.0 0
228 98
9604 49
169 64
9 3
17 1
49 0
' V ■
■ — . — '
^ y ^
•^ — y — -*
' — , — '
^' 1 '
326
9653
233
12
18
49
18
98
15
3
4
7
568
APPENDIX I
squares
and—.
are to be added, with careful regard to the signs +
The paper is then placed over the diagrams a^, a^, . . . iu
Schedule for 24 Okdinates
(Multipliers : 1, 0.97, 0.87, 0.71, 0.50, 0.26.)
3/1
y«
ys
Va
yio
yii
2/12
yiB
yii
2/15
^16
2/17
2/18
2/18
5/20
y%i
^22
y23
"1
flj
0-3
a-i
05
as
+
+
■ +
• • +■
+
+
.+....
■ + ■ ■
■ ■ +
■ • +
• +
• • + •
■ • • • +
• •-
.
—
... + ..
—
. . .
....+.
. —
—
—
■ • + •
+
+
— .
■ —
+ •
■ ■ +
■ + ■ • •
_
_
. .
■ +
■ ■ +
. . .
. . . ._
+
.—
...
+
• +
. _.
• ■ +■ ■
....-(-
—
..+...
—
...
• ■ + • ■
.-
■ ■ +
—
_
+ ■ ■ • ■
—
• +•
— . . ■ -
+
...
. . + . .
- —
• +
—
■ • ■ ■ +
■ —
■ + •
—
.
■ +
— .
...+..
.—
+ ■
.—
...
+
-
— .
■ +
■ +•
• • +
— ....
_
+
. .
.—
■ +
.+....
. . . + .
. . . ._
_
. . -
• + •
+
. . + . .
—
. . .
.. + ...
• ■ ■ • +
. . + . .
• +
■ • +
+
7
a, Og
ag
"12
2/0
yi
2/2
2/3
2/4
2/5
2/8
2/10
2/11
2/12
2/18
yii
2/15
Via
2/17
2/18
2/18
2/20
2/21
2/22
+ • ■ • •
. . . + .
• ■ • +
. + . . .
. . + . .
^
H
. . + . .
• + ■ • •
■ ■ ■ ■ +
■ • ■ + •
+
+
+
+
+
+
+
+
-
+
+
+
+ •
+ ■
+ ■
+ ■
+ •
+ ■
+. +• ■ ■ +
.... + . . + . . . +
-. ■ • + ■ +
• +• . ._ —
. — +
■ + ■ — —
-■ ...+. +
+ • .. + ■•. +
■ + + +
.+• ■+■ +
— + +
- + ■ +
+• ..+■.. +
succession ; the numbers over the white squares are likewise to
be added in each case. The values for a^, a^, a,, . . . are written
FOURIER ANALYSIS
569
on the margin of the paper. In like manner the values for
h, *2) h, - ■ ■ are obtained.
Schedule fok 24 Okdinates
(Multipliers: 1, 0.97, 0.87, 0.71, 0.50, 0.26.)
■'1 ^2 63 64 65 6„
Vo
"
Vl
■ +
....+.
■ + • -
• ■ + ■ -
• + •
+
2/2
■ • ■ + •
. + . . .
+ ■
■ + ■
■ + •
2/8
...+..
+ • ■
. -u .
Vi
..+...
■ + ■ ■
ya
. + . . .
■ + ■
—
+
+
2/6
+
— .
2/7
.+,...
—
—
• + -
+
...
2/s
. + . . .
2/9
■ + ■
—
• + •
+ .
2/10
..—...
+ •
. .
■ + ■
2/11
■ +
. . . —
- + ■ ■
. .
.+....
_.
2/12
Via
... ._
■•••+■
— .
■ ■ + ■ ■
+
2/14
...
• ■ + ■
— .
. . + . .
...
2/15
2/10
2/17
2/18
+ ■
—
■ • +• ■
■•+■■■
-. ...
—
+ .
__
J/in
...
. . .
. + .
. . + . .
_
— .
2/20
. .
. .
. . + . .
• • + .
2/21
—
-■
— .
■ + • •
+
2/22
. —
. .
— •
. — . .
— .
2/23
. —
. .' .
.
. . _ . .
— ....
_
2/0
2/1
2/2
2/0
2/7
2/8
2/0
2/10
y-Ll
2/12
2/13
2/14
Via
2/10
2/17
2/18
2/19
3/20
yai
2/22
2/28
• + + ■ • •
■■+•■■ • + ■
+ ■■-■■. ■
— +
. . . .+ ..+...
• ■ + — •
+ _. . .
+
-i- ■ -i- • i
• — + + ■ • ■ ■
■ +
■ ■ + . ■ + . ■ ...
■-■
• — ■ +
■ + • ■ -
• ■ + ■ •
■ ■ ■ +
+ —
• •- • • ■+
— ■+■ • ■
••••+• ■-• +
- ■ ■ +
._. . +
■ +■ •+
• •- + • •
- •
• + ■ •
. . + . . + ....
■ • •- .. + ... .
• • • + ■ ■ • ■
• + ■ ■
• — • + ■
+ ■ • • —
• • • +
+ —
■- +
Schedules are given here for 12 and 24 ordinates. These will
probably be sufficient for most purposes.
570 APPENDIX I
Hermann writes the values for y on centimeter paper (with
horizontal and vertical parallel lines dividing the surface into
centimeter squares), and then cuts out squares in a similar sheet
of paper so that when this is placed over the table only the
numbers needed for one constant can be seen. Thus there is a
pattern for a^, another for a^,, etc. ; the pattern for b^ is obtained
by turning a^ over, likewise for b^, b^, . . . by turning over those
for a^, tti . . . Ko patterns are needed for ai, b^, a^Q, and b^a-
The various additions and subtractions can be conveniently
performed on the margins of the paper on which the table is
written. The final sums would have to be divided by the num-
ber of ordinates used to give a-i, bx, a^, b^, . . ., but this can be
omitted as only the relations of amplitude are desired. More-
over, for a like reason the decimal points in the table itself may
be omitted, as this simply multiplies all numbers by 100.
After several analyses have been made it will be found that
nearly every value of y with its products has occurred in some
previous analysis ; thus the results can be copied into the new
table. With these methods a complete analysis of a curve-period
with measurements of 40 ordinates and computations requires
from two to three hours of work. Although the measurements
are recorded with only two places of figures, the error of com-
putation remains much smaller ; the error of measurement ■' is
also much smaller and the results are perfectly trustworthy to
the second figure.
The results do not generally show so decided a prominence of
one partial as in the case worked out above. An example is
found in the following series of values for a certain period of
the vowel a by Hermann :
Ci C^ C3 C4 C5 Cq Gj Cg Cg CiQ
4.2 8.5 3.2 7.3 2.2 13.9 44.7 50.2 13.6 14.6
The 6th to 10th overtones are most strongly represented. If it
were true that the vowel was composed strictly of a fundamental
and its overtones, these tones would be selected as a characteristic
of the vowel. It is, however, established that in general the
higher partials in a vowel sound are not harmonics of the lowest
partial. An analysis into a series of partials in such a case gives
a series of harmonics in which those nearest the inharmonic
^ Hermann, Phonophotographische Untersuchungen, IV., Arch. f. d. ges. Physiol.
(Pfluger), 1893 LIU 44.
FOURIER ANALYSIS 571
partial appear of greater amplitude, while those further away
are less influenced. To find this inharmonic the weighted mean
(or centroid) of the neighboring overtones is obtained by multi-
plying each amplitude by its ordinal number and dividing by
the sum of the amplitudes :
(6 X 13.9) + (7 X 44.7) + (8 x 50.2) + (9 x 13.6)
13.9 + 44.7 + 50.2 + 13.6 = ^•^^'
The number 7.63 gives the relation of frequency between the in-
harmonic and the fundamental. The frequency of the funda-
mental in this case, 98, multiplied by the number 7.53 gives
the frequency of the inharmonic, 737.
In using the centroid of a set of harmonics to indicate the in-
harmonic partial several rules should be observed : ^ 1. when a
harmonic of large amplitude appears with neighboring harmonics
of very small amplitude, it may be considered alone as indicating
approximately the partial ; 2. when the strong harmonic is accom-
panied by two neighboring strong harmonies, all three should be
considered; 3. when one of the neighboring harmonics is more
than twice as great as the other, only the former should be con-
sidered with the strong harmonic.
It may occur that a harmonic partial of very small amplitude
may lie between two of very large amplitude. The centroid
method is applicable here also and the inharmonic partial may
even coincide with an overtone of small amplitude.^ Attempts
at improving Hermann's formula for averaging the harmonics do
not seem to give any advantage.' Lloyd's criticisms * of the
FouEiEE analysis I am unable to understand ; they seem to rest
on mistaken views of the nature of vowels, and of the analysis.
The results of a Foueiee analysis may be graphically expressed
by laying off the distances 1, 2, 3, ... on the X axis to repre-
sent the series of harmonics, and erecting at each point an ordi-
nate proportional to the calculated amplitude of that harmonic.
The curve of Fig. 49 analyzed into a series of harmonics gave the
1 Hermann, as before, 50.
'■* Hermann, as before, 276.
8 Pipping, Zur Lehre v. d. Vokalkldngen, Zt. f. Biol , 1895 XXXI 564 ; Lloyd,
Interpretation of phonograms of vowels. Jour. Auat. and Physiol., 1897 XXXI 240;
Hermann, Weitere Untersuchungen iiber d. Wesen d. Vokale, Arch. f. d. ges.
Physiol. (Pfluger), 1895 LXI 169, 181, 182.
* Lloyd, On the Fourierian analysis of phonographic tracings of vowels, Proc.
Koy. Spc. Edinb., 1897-99 XXII 97.
572
APPENDIX I
resulting amplitudes as indicated by the values for c on p. 667.
The resulting plot is shown in Fig. 348.
KouDET^ has devised a computing machine for determining
the values of the coefficients a and b. It consists of a cardboard
rectangle ABCD (Fig. 349) covered
with millimeter divisions on which the
axis 00' is traced parallel to the sides.
The side AB is graduated in each di-
rection from 0'. Points are marked
on 00' in such a way that each divides
it into two parts whose ratio gives the
cosine of an angle of the first quad-
rant; thus, OE: 00' = sin 30° = cos
60°, OK: 00' =sin 60° = cos 30°,
I etc. Perpendiculars erected at these
I t I I points are divided like the line AB.
./ 2' 3 * ff 6 On the edge AB a small runner of
j-j^_ 34g metal or cardboard is placed ; a thread
is attached to a pivot at 0. The use of
the instrument may be readily explained. Suppose, for example, 12
ordinates to have been measured and the coefficient a^ is to be
o'
Fig. 349.
computed. The number of divisions is ra = 12 ; the whole period
considered as 27r (p. 562) and divided into n parts gives h = 30°
1 RouDET, Abague pour I'analuse des courbes pgriodiques, La Parole, 1900
II17.
FOURIER ANALYSIS 573
as the difference between successive angles. Thus, remembering
that cos 120° = cos 240° = - cos 60°, cos 360° = 1, etc., we have
from the general formula for «< on p. 563, when i — 4,
6«i — yo — yi- cos 60° — y^ . cos 60° + 2/3 — 2/4 • cos 60°
— 2/5 ■ cos 60° + 2/6 — 2/7 • cos. 60° — y, . cos 60°
+ 2/9 — 2/xo • cos 60° — yu .
Let 2/o = 8, 2/i = 23, etc., for a given curve. The rider is moved
upward 8 divisions from 0' toward B. To obtain y^ . cos 60° the
thread is stretched from 0 to the point 23 above 0' ; the distance
above 00' of the point where it cuts the ^i^ gives the value of
2/1 . cos 60°, which is found to be 12. Since this is a negative
value, the rider is moved back 12 divisions, reaching — 4. The
succeeding values are found and added in like manner.
A FouKiER analysis of a curve may be executed automatically
by the harmonic anal3'zer of Thomson ^ or of Henrici.'' By mov-
ing the point of the instrument over a period of the curve the in-
dicators of the machine will show the amplitudes and phases of
the partials contained in it. Stkachet's slide rule aids in ordi-
nary computation.
The FouRiEK analysis furnishes a convenient and trustworthy
method of determining the harmonic sinusoid components of a
curve. When the curve is the product of vibratory movements
of this kind, the results give a proper analysis. The method
gives only approximate results when the component vibratory
movements are inharmonic or non-sinusoid.
Refeeences
For theory of the Fourier analysis : Fourier, Th^orie analytique de
la chaleur, Ch. Ill, Paris, 1822; Thomson and Tait, Treatise on Natural
Philosophy, §§ 75-77, Cambridge, (1879) 1896. For list of modifications :
Pascal, Repertorio di matematiche superiori, Milano, 1898 (trans, by
ScHBPP, Leipzig, 1900). For discussion of applicability of harmonic
analysis to speech curves : Ch. XXVIII above. For a discussion of the
relations between the errors of measurement and computation : Hek-
1 Thomson, Harmonic analyzer, Proc. Roy. Soc. Lond., 1878 XXVII 371 ;
also in Thomson and Tait, Treatise ou Natural Philosophy, I 505, Cambridge,
1 896
2 Henrici, Ueber Instrumente zur harmonischen Analyse, Dyck's Katalog
math. u. math.-phys. Modelle, Apparate u. Instrumente, 125, 213 ; Nachtrag, 34,
Munchen, 1892-93,
574 APPENDIX I
MANN, Die Bedeutung d. Fehlerrechnung bei d. harmon. Analyse von
Kuroen, Arch. f. d. ges. Physiol. .(PAiiger), 1901 LXXXVI 92; Kur-
venanalyse u. Fehlerrechnung, Arch, f. d. ges. Physiol. (PflUger), 1902
LXXXIX 600.
For pantographs and harmonic analyzers : Cokadi, Zurich. For a set
of computing patterns for 40 ordinates and a supply of centimeter paper :
Prof. L. Hermann, Physiologisches Institut, Kbnigsberg. For a set
of schedules for 12, 20, 24, 36, and 40 ordinates: E. W. Scripture, New
Haven, Conn.
APPENDIX II
STUDIES OF SPEECH CURVES
The following is a condensed account of some work on
the speech curves obtained as described in Ch. IV and ana-
lyzed as indicated in Ch. V.
The words first studied ^ were those of "William F. Hooley,
a trained speaker, reciting the nursery-rhyme entitled The
Sad Story of the Death and the Burial of Poor Cock Mobin.
The record is contained on the disc numbered 6015 made by
the National Gramophone Company of New York.
The record on the gramophone disc reads as follows :
Now, children, draw your little chairs nearer so that you can
see the pretty pictures, and Uncle Will will read to you the sad
story of the death and the burial of poor Cock Eobin.
Who killed Cock Robin ? Who '11 make his shroud ?
I, said the sparrow, I, said the beetle,
With my bow and arrow. With my thread and needle
I killed Cock Kobin. I '11 make his shroud.
Who saw him die ? Who '11 be the parson ?
I, said the fly, I, said the rook.
With my little eye With my little book
I saw him die. I '11 be the parson.
Who caught his blood? Who '11 dig his grave?
I, said the fish, I, said the owl,
With my little dish With my spade and trowel
I caught his blood. I '11 dig his grave.
Who '11 carry the link ?
I, said the linnet,
I '11 fetch it in a minute.
I '11 carry the link.
1 Scripture, Researches in experimental phonetics (first series), Stud. Yale
Psych. Lab., 1899 VII 14.
576 APPENDIX II
To extend the treatment to prose some cases of ' I ' were
studied in another record by William F. Hooley, entitled
Gladstone's Advice on Self-Help and Thrift, being disc
number 6014 of the gramophone series. The speech begins
as follows:
Ladies and gentlemen, the purpose of the meeting on the
14th instant may, I can say, be summed up in a very few words :
self-help and thrift.
Two examples of this diphthong were also studied in the
word ' thj', ' as it appears in a disc numbered 668Z (name of
speaker not given), which begins as follows :
Our Father which art in Heaven, hallowed be Thy name,
Thy kingdom come . . .
Pig. 350.
These are the three records referred to on p. 58 as Coch
Mobin, Series I ; Self -Help, Series I; Lord's Prayer, Series
I.
The first occurrence of ai is in the verse ' I, said the
sparrow. ' A reproduction of the curve for this word is given
in Fig. 350. Some of the details were lost by the engraver in
making the figure, and others were not quite correctly repro-
duced; the original curve is much sharper and clearer. The
STUDIES OF SPEECH CURVES 577
beginning of the record appears under the magnifying glass
like the drawing in Fig. 351. The dots indicate intervals
of l^-C^ = 0.0010.
This word ' I ' occupies an interval of 452". It is preceded
by a silent interval of 770", or about | of a second; this
is the-fuU stop in the stanza after the question is asked and
. before the answer is given. It is followed by a silent inter-
val of 210".
The beginning of the a is apparently clear, that is, it is not
preceded by any breathing. The vocal cords are apparently
adjusted for voice production before the expiration begins;
the vowel starts with a light vibration of the cords. There
is no explosive sound, or glottal catch, before the vowel.
The vowel a begins with a movement of the vocal cords by
which an extremely weak puff of air is emitted. This puff
Fig. 351.
of air passing through the cavity of the mouth arouses
three or four oscillations of the air contained in it. There is
first a half-oscillation of weak amplitude, then a compara-
tively strong oscillation, followed by very weak ones. Even
the strongest is, however, very weak; the following oscilla-
tions are so weak as to be hardly perceptible. The cavity
vibrations disappear and there is an interval of silence before
the second puff appears. Then the cords emit another puff of
air a trifle stronger than the first, the time from puff to puff
being 18". The six cavity vibrations are slightly stronger
than before. The period of silence is shorter than before.
The third puff occurs 11" after the second one. The cavity
vibrations are a trifle stronger still ; there are seven of them
with a brief interval of silence! The fourth puff begins at
10" after the beginning of the third one. The fourth puff
contains eight cavity vibrations, all slightly stronger than be-
fore- there is no interval of silence because the fifth puff
' 37
578 APPENDIX II
begins just as the last cavity vibration of the fourth puff
ends. The interval occupied by the fourth puff is 9'^.
It is a characteristic trait of this particular a that each
group of cavity vibrations is strongest at the start; this
indicates a sudden and complete opening of the cords
(p. 260). The quickest opening requires, however, a 'little
time and there must be a measurable change from no passage
of air to full passage ; this is shown by the weak half of the
first cavity vibration preceding the large half. The form of
vibration probably indicates a complete closure of the cords
whereby they actually touch each other after the puff is
emitted.
The cavity vibration in the first part of the word has a
period of 1" or a frequency of 1000.
Fig. 352.
As the period of the cord tone becomes shorter, the num-
ber of cavity vibrations to each period becomes smaller.
Beyond the 30th puff of the cord tone the cavity vibrations
show a lengthening of period. In the 39th cord vibration
the cavity tone reaches a period of 2.2°' or a frequency of about
450 ; it thus falls more than an octave in the time of nine
cord puffs, or, in this case, in 33"- Here the cavity tone is
nearly but not quite of the same period as the octave, 2'^, of
the cord tone, 4". This change is shown in the hand-draw-
ing, Fig. 352, which begins with the 31st vibration. This
relation between cavity tone and cord tone is maintained
to the end of the word; the rest of the curve shows the
peculiar alternation of waves seen in the last two vibrations
in Fig. 352.
The vibrations up to the 31st unquestionably belong to the
a. In the vibrations beyond the 39th both the cord tone and
the cavity tone are constant, except for a slight fall at the
STUDIES OF SPEECH CURVES 579
end. They unquestionably belong to the i. The vibrations
from the 8lst to the 39th show a constant cord tone and a
falling cavity tone. They are presumably to be consid-
ered as belonging to the ' glide. ' During the a the cords have
been stretched more and more, until at the 31st vibration
they reach the tension required for the i; the only further
change necessary is the lowering of the cavity tone.
Beyond the portion shown in Fig. 352 the curve shows
strong vibrations so nearly alike that one is naturally induced
to consider each one a cord vibration. This would not be
proper because close inspection of Fig. 350 shows that suc-
ceeding vibrations differ slightly, while alternate ones are
alike. This likeness of all the cavity vibrations in the i as
contrasted with the a is probably due also to a difference in
the action of the cords; this difference appears more clearly
in the word 'eye ' analyzed below.
With the understanding that no definite limit can properly
be made between one sound and the neighboring one in this
case, we may, in view of the facts just mentioned, consider
the a to have occupied the time 203" ending with the 30th
puff', the glide to have occupied 33'^ ending with the 38th puff,
and the i to have occupied the remaining 216"^.
The cavity tone of the i is one of about 450 frequency.
This cavity tone is much lower than the very high tone
assigned to i by Hermann and others, but is not so low as
those assigned by some other observers. There is, however,
the possibility of different tones in the vowels from different
speakers, and also that of several cavity tones in the same
vowel. By careful examination of the curves I find them
often marked by small additional vibrations. These are
frequently quite prominent in the i of ai. Their fineness
rendered it impossible to settle on any definite period for
them.
Beginning with a period of IS"', the cord tone changes
slowly through 11, 10, 9, 8, 7% reaching Q' at the 11th
vibration, b" at the 15th, 4'' at the 30th ; the period of 4' is
maintained to about the 100th vibration, after which it falls
580 APPENDIX II
slightly to 4.2'^ during the last seven vibrations. In other
words, the pitch glides slowly upward from a tone of 56
complete vibrations per second to one of 200 a second, then
more slowly to one of 250 a second, at which it remains
constant except for a slight drop as the diphthong ends.
The changes of pitch in the cord tone and the cavity tones
are indicated in a general way in Plate XV.
The amplitude of a vibration is the distance from the posi-
tion of equilibrium to the extreme position on either side ; it
is thus one-half the difference in altitude between the crest
and the trough of a wave. The initial cavity vibration of
the first puff of this a has an amplitude of less than 0.1™".
This slowly increases to 0.3™™ at the 20th vibration, after
which it sinks only a little to the 38th. Beyond the 38th,
that is, from the beginning of the i, the amplitude rapidly
increases from 0.3™™ to 0.7™™ at the 50th vibration; there-
after it slowly sinks, becoming 0.3™™ at the 60th vibration
and 0.2™™ at the 80th, 0.1™™ at the 88th and 0 at the 96th.
The maximum for the i is 2^ times that for the a. The
course of change in amplitude is given in Plate XVI. The
horizontal scale is greater than that for the pitch curve in
Plate XV.
The word ai ends by a gradual cessation of the expiratory
impulse with hardly a noticeable change in the tension of the
vocal cords; this is the clear ending usual in 'English. The
slight fall in pitch of the i toward the end indicates a change
that may be apparent in the auditory effect of the word,
although it cannot be distinguished separately. It is probably
due to a relaxation of the cords.
To the ear the sound of this word ' I ' appears from the
record ' colorless, without emotion, without inflectional rise
or fall within the word, a monotone ' ; ' a mild statement.'
The mildness of this word seems related to its length and its
gradual changes in pitch and intensity.
In the original monograph the details of several other
examples of ai are given as for this first one. The curves
of pitch for all of them are given in Plate XV, those of am-
STUDIES OF SPEECH CURVES 581
plitude in Plate XVI ; the horizontal scale for all except the
last two figures on Plate XVI is twice that used in Plate XV.
Thirteen cases of ' I ' from the Cock Rohin record were
studied. In general the fundamental characteristics already
considered were found in all. Some peculiarities, however,
are to be noted.
For convenience the term ' primary ' will be applied to the
strongest one in a group of cavity vibrations, and ' secondary '
to the others. In the first ai the primary is the first of the
group.
In the 4th ' I,' of ' I saw him die,' one of the secondaries
appears almost as strong as the primary. This large secon-
dary keeps at the same distance behind the primary. As the
pitch of the cord tone rises, the primaries come closer to-
gether ; the large secondary, being at a constant interval be-
hind the preceding primary, thus comes steadily closer to the
following primary until it disappears in it.
I do not believe that this larger secondary is due to an
overtone-vibration of the cords. The curves in such a case
would show the first overtone-vibration always half-way be-
tween the two primaries. In the curve for this vowel the
strong secondary keeps at the same distance after the preced-
ing primary, while the distance to the following primary
steadily decreases.
It might be suggested that the primary and the strong secon-
dary may represent two waves of a lower cavity tone, while
the primary and the other secondaries represent the waves of
a higher one ; the lower tone would have a period of S^^" or a
frequency of about 286. There would then be at least three
tones present in the a : the rising cord tone ; the lower cavity
tone of 286, which finally coincides with the cord tone; and
the higher cavj||;y tone of 1000.
This large secondary appears strongly in most cases of ai
in the Ooch Rohin records, but also sometimes weakly (Fig.
350 above ; also ' I ' in Plate II) and sometimes not at all
(see table of tones, p. 587; also 'my' in Plate II).
Sometimes the first period of the cord tone is shorter than the
582
APPENDIX II
following one. This occurs, for example, in ' I '11 make his
shroud,' and ' I '11 be the parson.' In the former case the periods
are 9.8% 11.6% 10.9% 9.8% etc., and in the latter 8.1% 10.-5%
9.8% 8.8% 8.8% 8.1% etc. The cords seem to receive an
excess of tension before the breath begins, and to be then
relaxed to the tension desired.
In one case the fall of the upper cavity tone appears to take
place from the very beginning of the word ; the cavity tone
is thus steadily falling, while the cord tone is steadily rising.
Fig. 353.
This occurs in the a of ' I ' in ' I, said the fish.' The period
of the cavity tone begins with 1.4% reaches l.S"' at about the
10th vibration, 1.8°^ at the 40th vibration and then remains
constant to the end of the word. The cord tone, however,
starts low and rises.
The curve for ai of ' eye ' in the phrase ' with my little
eye ' follows immediately on the last vibration of the final 1
in the word ' little.' The three words ' my little eye ' are
here spoken with no separation. It is interesting, in pass-
ing, to consider the possibility that this fusion of the three
words goes parallel to a fusion of thought. It is evident
STUDIES OF SPEECH CURVES 583
from the very tone of the speaker that he is thinking of
one thing, a certain ' eye, ' and that the facts of ' being mine "
and ' smallness ' are not of any particular account to him
(p. 126).
The curve for dai in ' Who saw him die ? ' is given in Fig.
353. The word begins with 20 vibrations belonging to the d.
These vibrations have a period of 2.0'^ or a frequency of 500.
The a begins promptly and loudly, as might be expected
from the fact that the expiration is already in progress and
the cords are in vibration. The cord tone of the a in the
first vibration is higher than in the subsequent vibrations, as
might be expected on the assumption that the cords are
already stretched to give a period of 2.0"^ for the d, and must
Fig. 354.
be relaxed to produce the lower tone of the a. While this
relaxation is going on, the cords must pass through all inter-
mediate positions between that for a period of 2.0'' and that
for one of 3.2"^. This occurs to a large extent apparently
within the time required for one vibration of the d. At the
same time the mouth is changing from the d position to the a
position. These facts seem sufficient to explain the curve at
the change from d to a, shown in the drawing. Fig. 354; the
three vibrations on the left are the last of the d, the strong
one on the right is the primary cavity vibration of the first
puff of the a, and the connecting line shows the curve dur-
ing the glide.
The curve of dai in ' I saw him die ' is given in Fig. 355.
The word begins with 11 vibrations rapidly increasing in
amplitude from 0 to 0.4" and having a constant period of
2.5" or frequency of 400. These are the vibrations for the
584 APPENDIX II
d; they resemble those of dai in the first example. The
sudden fall in pitch after the d is quite marked. The d
curve is lost at once. The following interval of 1" can
hardly be said to be the first vibration of a, as its secondaries
are very irregular in form; during this interval the mouth
is changing from the d shape to the a shape. The peculiar
form of the vibrations of this glide is well shown in the figure.
The a curve differs from that of most cases of ai, in having
less difference between the primary cavity vibration and the
rest ; the first and second in a group are, in fact, of almost
equal intensity. The curve changes from the a form to the
Fig. 355.
i form so gradually that it is quite impossible to place any
dividing line; each element of the diphthong may be said
roughly to occupy half the total time.
The curve for lai of ' fiy ' in ' I, said the fly ' is shown
in Fig. 356. No specific details concerning the f can be
derived from the curve. The strong vibrations just preced-
ing those of the a are presumably from the 1 sound. They
rise rapidly in intensity and greatly resemble those of the d
in the two cases of dai above; their period is LO'^ and their
frequency 526. Immediately after the last vibration of the
1 there follows a short a puff with the primary cavity vibra-
tion not so strong as in the following puffs. The cord ad-
STUDIES OF SPEECH CURVES
b%l
justment seems not to be perfected for the a till the second
characteristic a puff occurs.
The curve of ' thj^ ' in 'hallowed be thy name' begins
with 7 vibrations belonging to 3 having a period of 2.5' or
a frequency of 400. The beginning of the a is prompt and
loud.
The curve of ' thy ' in ' Thy kingdom come ' shows 6 faint
vibrations at the beginning belonging to the 8 and having a
period of 2.8'^. The curve of the a suggests a more gradual
Fig. 356.
opening of the cords and a less explosive effect. There is
no strong secondary of the kind described on p. 581.
A comparison among the curves of pitch given in Plate XV
shows that general resemblances occur among most of the
curves of pitch for the cord tone in the cases of ' I. ' In
' eye ' the curve differs radically from all the other examples ;
it- starts with a moderately high pitch and falls continuously.
There is the possibility that the fall in pitch in this word
may have something to do with its position at the end of a
phrase. If the word had been followed by a long pause, it
would naturally have fallen on account of its position at the
end of a sentence ; the pause, however, was extremely short
586 APPENDIX II
and we cannot very well assume a short pause as the equiva-
lent of a full stop unless we give up entirely the theory of rela-
tion between punctuation and time. It is, nevertheless,
possible that this theory may have to be modified, as later
researches have shown that comma pauses may be long and
semi-colon and colon pauses may be very short. The upper
cavity tone is to some degree stable in the two end portions
of all cases in Plate XV, but undergoes a more or less rapid
change within the various words. The lower cavity tone
is constant when it is present.
Both the ear and the curves indicate that the first part of
the diphthong in the cases just studied is to be considered a
form of a as in ' parson.' The second part is probably i as in
' kin ' in all cases except the second ' thy.' In this last case
the cavity tone differs greatly from all the others ; the sound
may be e as in 'let.' The difference between the two
cases of 'thy' may arise from the following sound; the
forward contact of n favors an i while the backward one of
k favors an e.
Various words like ' ein, ' ' weisser, ' ' Eis, ' ' Zeiten, '
' Schein,' etc., were closely studied in the tracings from
disc number 1500, Die Lorelei und Der Fichtenbaum,hj W. L.
Elterichc. When examined under the magnifying glass, the
a portion of the record showed in most cases curves analogous
to those in the cases of 'I,' whereas the i portion was ex-
tremely weak. This peculiarity of the weak i in the German
ai and the very strong i in most cases of the American ai
gives the former the effect of containing a longer a. It must
be noted, however, that many sounds usually treated as the
same are really different. Thus the vowel of ' weiss ' in ' Ich
weiss nicht was soil es bedeuten ' gave a curve differing
greatly in character from that of ' weisser ' and the other
words mentioned above. Again, some of the cases of the
American ai described above showed a weakening of the i
that indicates a tendency toward the German form.
The analyses show that ai is not the sum of the two
vowels a and i, but an organic union into a new sound ai.
STUDIES OF SPEECH CURVES 587
There is no necessary pause or very sudden change of inten-
sity or change in pitch or even change in character. This is
what would be expected on psychological grounds. The
speaker does not think of and speak two sounds separately
but only one; the execution of this one impulse by two
distinct processes would be unusual. The various degrees of
perfection of the synthesis of the two elements would corre-
spond to various expressive characters of the resulting sounds.
In so far as they can be considered to be constant, the
cavity tones in these cases of the a and the i were found
to be as in the following table.
Tone of
d, 1, ?S,
/, 1st example
/, 2d
I, 3d
/, 4th "
J, prose "
Ei/e,
Die, 1st example 500
Die, 2d " 400
Fli/, 526
Thy, 1st example 400
Thy, 2d " 417
The following view of the physiological action of the vocal
cavities in producing ai in the cases studied above may be
proposed tentatively. The depressed position of the tongue
for the a (Plate XXV) leaves open a large cavity reaching
from the teeth to the vocal cords ; the uvula offers no great
interruption. The lower cavity tone of the a may be con-
sidered to arise from the vibration in this cavity. The upper
cavity tone of the a may be supposed to arise from the rear
cavity, that is, the throat cavity from the cords to the
slight elevation of the tongue at the uvula. As the a
changes to i (Platq XXVI) this elevation of the tongue
moves forward, enlarging the rear cavity by including more
and more of the mouth ; this continuously lowers the upper
cavity tone until the tongue comes to rest in the typical
i position.
Tones
of the a.
Tones
of the i.
Cord,
Lower
Upper
start.
end.
cavity.
cavity.
Cord.
Cavity.
56
250
286
1000
250
450
83
250
286
1000
250
555
131
250
286
1000
250
500
111
286
286
1000
286
400
102
180
360
1000
180
360
400
160
435
1000
160
476
179
200
1000
200
473
217
133
1000
133
473
160
204
256
625
256
500
143
149
588
149
416
84
143
416
143
288
588
APPENDIX II
Another theory is possible. The lower cavity tone of the a
may be assigned to the trachea (p. 294) ; the rise and fall of
the larynx may explain the differences among the various
cases. This view is favored by the constant character of the
tone within each case (see dash-line in Plate XV). The
higher cavity tone for a is then to be assigned to the mouth
cavity. The resonance tone of i, as found in these records,
is probably from the mouth, the effective capacity of the
rear cavity being increased from a to i.
n -«t
-&
,.
ll
-^
I
tS
-^
1
-4
1 1
1
1
S'^
■d
1
1
1 -^
i tf"^
1 *
1 1
■tt-d
ttW^
*
-S/^
its!
ft" 1
y^"^^^
^V^
'^A-^
1 1
{m\»
1
ttJ tt^
1
4 iJ
•
w
tt*
a
(I, sa
spa
1
i
/
id the
rrow)
-5'-
-J
a
(Ik
E
, 1
/
illed
obin
1
i
Cocl
a
^ (I
th
, I
/
saic
efly
1
)
a
/
(I saw
die)
1 II
1
i
lim
a
/
(may, I
say)
1
i
can
a i
eye
(with my
little eye)
T
1 1
^
1 1
■t'i i
^
<i 1
N
4
•
' Jt
' 4t
** 1 '^
r
V ^
it^i
3i_
ff^
1 1
Mm
4t--l
'Mm'
4*^
1
1
[/
-»•
tj
jj* )
1
Hj^
^
riV
tt. #• ftiJ
* 1 '
#•
W.
! J
^
1 « /^'
b-^
, • cJ
1 • «*
1
1
1
1
«'__^
da idailai Sai 3a i
rfi'e rfie flij thy thy
(Who saw (I saw him (I, said the fly) (hallowed be (Thy kingdom
him die?) die) Thy name) come)
In all cases of ai there is no sudden jump of the cord tone ;
the i continues the cord tone of the a, forming with it the
easiest musical interval, a unison. This tone is, however,
different in different Cases; the cord tone of the a rises to a
certain point selected for that of the i. The selection of the
pitch of the cord tone for the i is influenced by the preceding
cavity tones of the a, as may be seen from the table. A
STUDIES OF SPEECH CURVES 589
Study of these tones reveals some tendency toward musical
relations within each ai. This probably is one of the factors
of the musical character of the voice of this speaker. The
musical relations are roughly indicated in the accompanying
notation ; the lowest notes indicate the cord tone, the others
the cavity tones.
In ai of ' die, ' in a manometric flame record (p. 29) sent
me by Nichols and Mekeitt, I find a cord tone rising
through 7.5, 5.2, 4.8, 4.6, 4.6, 4.6, 4.6, 4.6, 4.6, 4.6, 4.6,
4.6, 4.6, 4.6, 4.6, 4.6, 4.6, 4.6'°°', at which, point the record
is cut off. In another example of ai in ' die ' the cord tone
rises more slowly through 6.1, 6.0, 5.8, 5.7, 5.6, 5.5, 5.5, 5.5,
5.5, 5.5, 5.5, 5.5, 5.5, 5.5, 5.5, 5.5, 5.5, 5.5 (the 18th
vibration), at which point the record is cut off. The begin-
ning is thus like that in the cases studied above.
A curve of ' fly ' has been sent me by Prof. Beviee (p. 49)
with the explanation that the a appears to change gradually
through ae (as in ' hat ') to e (as in ' let ') ; that the two
cavity tones of a are of 600 and 1150 frequency ; that those of
e are of about 600 and 1700 ; that the entire sound lasts 0.32%
including 0.09* for a, 0.11^ for ae, 0.1 2= for e; and that the
maximum amplitude lies in the a. This example of ae re-
sembles the diphthong of the second ' thy ' above.
Some of the characteristics of fusion (Ch. XXX) may be
illustrated by the following study of the phrase ' Who '11 be
the parson?' of Oock Robin, Series I.^ The entire curve is
given in Fig. 357; an inspection of the curve shows the
arbitrary nature of the divisions into vowels, glides, etc.
The breathing h probably does not appear in the record.
The vibrations in line 1 may perhaps be considered as be-
longing to the passage from h to u, or the h-u glide. They
show cavity vibrations beginning 'with a period of 2.8' and
shortening to 2.5''. It is impossible to say definitely whether
or not they are in groups that indicate cord vibrations. The
first two vibrations in line 2 have periods of 2.3°'; they
1 In the following phonetic analysis, of a, complete phrase I have been
assisted by Miss E. M., Comstock.
590
APPENDIX II
Fig. 357.
STUDIES OF SPEECH CURVES 591
are also cavity vibrations. Thereafter the cavity vibrations
remain practically constant at 1.9" and are found in groups
clearly indicating a cord tone; they belong to the u..
The curve for the u (lines 2 and 3) closely resembles
that for ai in its general character. The first part shows a
rising cord tone and a nearly constant but afterwards falling
cavity tone. In the latter portion the cord tone is approxi-
mately constant while the cavity tone falls. The change in
the character of the action of the cords appears clearly also
as in ai (p. 578). It is, in fact, very evident that this sound
is somewhat diphthongal with possibly less difference be-
tween the two elements than in the case of ai. This
diphthongal character of the English u is well known to
phoneticians; the sound is generally indicated as uw.
The curve at the beginning of the u shows a vibration of
6.3'' from the vocal cords acting on a cavity whose period
1.9'^ is not a sub-multiple of the cord period. As the cord
period is gradually shortened, the cavity period (remain-
ing the same) steadily modifies the form of the resultant
vibration, and the curve is seen to change its form gradually.
The relation between cord tone and cavity tone is closely
analogous to that in the a of ai.
The successive vibrations of the u occupy the periods of
6.3, 6.1, 6.1, 5.6, 5.4, 5.4, 4.9, 4.9, 4.9, 4.9, 4.9, 4.6, 4.6,
4.6, 4.2, 4.2, 4.2, 4.2, 4.2, 4.2, 4.2, 4.6, 4.6, 4.6, 4.6, 4.6,
4.6, 4.6, 4.6, 4.6, 4.6, 4.6, 4.6, 4.6, 4.6°-. The total time
occupied by the u is IQl".
The u thus shows a tightening of the cords to the tension
necessary for a tone with a period of 6.3% and thereafter a
gradual increase of tension to a maximum represented by
4.2°' ; after this there is a fall to 4.6"^, at which the tone
remains constant.
The cavity tone begins with period of 1.9", or a frequency
of 526, or at approximately <?. It is, however, not constant
throughout the u in this case. This is especially evident
during the last part where the cord tone is constant. In
this region the curve steadily changes its form from the
592 APPENDIX IT
earlier u form toward the 1 form; during the last eight or
ten cord vibrations it is difficult to say whether the curve
belongs to the u or the 1. The cord vibrations of the u
persist in their own constant period, however, to a point
which can be detected. We are thus justified in attributing
these vibrations to the u although the mouth cavity has been
presumably steadily shaping itself for another sound. The
cavity tone thus rises toward the end.
Repeatedly observed facts of this kind have forced upon
me the belief that the view of a word as composed of a set of
fixed sounds with glides between them is a somewhat inade-
quate one. It is derived from the attempt to get away from
the artificial character of spelling, but it still largely retains
that character. The usual view of the words ' who '11 ' would
represent them as composed of h — glide — u — glide — 1. The
vocal organs are supposed to occupy three distinct positions,
the glides representing the intermediate positions during the
moments of change.
A somewhat different view seems better fitted to the actual
curves. The unit of speech is sometimes a phrase, some-
times a word, and never a vowel or a consonant unless this
is at the same time a word. In speaking a word the vocal
organs pass through a series of positions of a special char-
acter without stopping in any one position. Thus the words
' who '11 ' represent a continuous change in the force of ex-
piration following a definite plan, also a continuous change
in the tension of the vocal cords, likewise continuous move-
ments of the parts of the mouth. The force of expiration
rises to a maximum of 35'' in the h-u glide, continues with
slight fluctuation during 171'^ in the glide and u, and finally
dies away at 277'^ at the end of the 1. Before the breath
begins the mouth has adjusted itself to a tone of a period of
2.8"; this position changes very slightly during the 35°' of h ;
then it makes a rapid change through 2.3, 2.1 to 1.9'^ in the
u, remains constant during 167% and rises suddenly to the
mouth tone of the 1 (not determinable here).
The cord tone has a somewhat similar course. It begins
STUDIES OF SPEECH CURVES 593
with a period of 6.3'' in the u at 39^ after the beginning of
the word; it rises steadily to 4.2"^ and then falls to a con-
stant pitch of 4.6" for the latter part of the u ; suddenly
it rises to 2.1°- for the 1 and remains practically constant
for 71-.
On speaking the words ' who '11' I perceive apparently con-
tinuous movements of the lips and tongue; they do not
assume fixed positions at any moment. This would agree
with the changes just described.
There are thus at least three distinct but cooperating con-
tinuous processes following different courses throughout the
words, namely, the force of expiration, the cavity tone and
the cord tone.
It seems therefore somewhat artificial to divide the words
' who '11 ' into 3 or 5 sounds ; we may preferably say that for
the sake of discussion 5 stages in the changing sound may be
picked out as typical of the whole process. To illustrate by
an analogy, we might take single pictures out of a series of
views of a runner made for the kinetoscope and treat the
whole movement as made up of a series of positions in which
the runner remains at rest. This treatment has its advan-
tages for certain cases, but we should never lose sight of the
fact that the true movement occurs otherwise.
This view is not inconsistent with the fact that some of
the elements of a vocal sound may remain approximately
constant for a short time. Thus, the pitch of the h-u glide
is nearly constant — as far as our methods can discover —
though the intensity is changing, and the pitch of the u is
fairly constant for a while.
The sound 1 apparently does not begin suddenly but arises
from a modification of the u. The u itself has been steadily
changing in character from the very beginning; during its
last five or more cord vibrations it gradually approaches the
form of curve that characterizes the 1. After this point the
curve takes the 1 form which differs completely from that of
the u at the start (Fig. 357, line 4). As stated above, the
explanation is presumably (1) that the cord tone remains
38
594 APPENDIX II
on the u pitch until a certain moment at which it suddenly
rises to the 1 pitch, whereas (2) the mouth cavity begins to
modify itself from the u form to the 1 form before the cord
tone changes.
The 1 occupies a total time of 71". It shows 34 cavity
vibrations with a fairly constant period of 2.1" or 576 fre-
quency. There seems to be a grouping by twos that indicates
a cord tone an octave lower, that is, of 4.2" period, or 238
frequency. The form of the vibration steadily changes as
shown in the figure ; there is a change either in the tone of
the mouth cavity or in that of the cords.
The changes in pitch in these words ' who '11 ' follow
the same general course as in ai, namely, that in a succes-
sion of sonants (speech elements with tones) the cord tone of
a sonant tends to be a multiple or a sub-multiple of the cord
tone or the mouth tone of the preceding sonant. The rela-
tions are not exact but only approximate. The mouth tone
2.5" of the h is followed by a cord tone for the u having
a general average of 5.0" or an octave below the former.
The mouth tone of the u, 1.9", is followed by a cord tone for
the 1 of pretty nearly the same period 2.1". Such relations
are what would be expected in a voice — at any rate in one
that was not unpleasant; for the human ear finds pleasure
in a succession of tones whose periods stand in certain
relations. Possibly some of the explanation of disagreeable
voices may be found in the violation of this law.
In the spoken words on the gramophone disc the sound b
follows immediately upon the 1 without pause. The speech
curve at this point (Fig. 357, line 5) shows no measurable
vibrations, the enlargement not being great enough to reveal
the details of the weak tone of the b. The interval occupied
is 96".
The cavity vibrations of i (lines 6 and 7) have a constant
period of 2.8", or a frequency of 857. They start with ah
amplitude of 0 and rise steadily to an amplitude of 0.2™™;
at the end they fall to 0 suddenly in four vibrations (line 8).
They are grouped in twos, indicating a cord tone an octave
STUDIES OF SPEECH CURVES 595
below with a period of 5.6'^, or a frequency of 179; the rela-
tion is like that often found in i of ai. The glide to S is
seen in the first part of line 8.
The scale of enlargement is not sufficient to give definite
information concerning the waves of the ?5 : it occupies a
time of 56''-
The indefinite vowel 9 of ' the ' rises somewhat rapidly to
its maximum, remains at an even amplitude (line 9), and
drops suddenly to 0 in the last 4 vibrations. It has a pitch
of 6. 7°' on an average and a maximum amplitude of 0.4™".
The entire vowel contains 12 cord vibrations and occupies a
total time of 84'^.
The unstressed vowel 9 of ' the ' is cut short by the closing
of the lips for p. This suddenly reduces the amplitude of
the vibrations till they are very faint (line 9), yet the cords
Fig. 358. Fig. 359.
continue to vibrate after the closure as may be seen from the
faint vibrations (lines 9 and 10). The sound can no longer
be considered to be the vowel 9 and cannot in the usual sense
be called a p. It may be treated as a glide although it occu-
pies fully two-thirds of the interval of 112°' between the 9 in
' the ' and the a in ' parson.' If the period of sonancy after
' the ' is to be considered as a glide, the remaining third of the
112'' may be assigned to the p (line 10).
The word ' parson ' appears to the ear to have an inflec-
tional force of the form indicated in Fig. 357, as often
appears at the end of questions; the circumflexion appears
to lie in the a and the deep fall to be in the n. The word
seems to contain a brief r. The word differs from the same
word three lines later (p. 575) which appears to the ear to
have a deep inflectional tone, at first level and then falling
as in deciding a matter; this is indicated in Fig. 359. The
latter word seems to contain no r. The word ' parson ' is in
696 APPENDIX II
both cases apparently continuous with the word ' the ' and
would be phonetically written 33parsn.
The vowel a in this case occupies a period of 180'- It is
preceded by the interval of 112'' belonging to the p and is
followed by a glide of 12. 3^
The a shows 36 cord vibrations. The pitch rises gradually
as shown by the following measurements of the successive
periods: 6.7, 7.0, 6.7, 6.0, 6.0, 6.3, 5.3, 5.3, 5.3, 5.3, 5.3,
5.3, 4.9, '4.9, 4.6, 4.6, 4.6, 4.6, 4.6, 4.2, 4.2, 4.2, 3.9, 3.9,
3.9, 3.9, 3.9, 3.9, 3.9, 3.9, 3.9, 3.9, 3.9, 3.9, 3.9, 3.9, 4.0,
4.2. It contains a constant lower cavity tone with a period
of 2.8°' or a frequency of 357. The upper cavity tone is one
of about 714 vibrations per second.
The amplitude rises through the first four vibrations from
zero to 0.3™™ and is maintained at this to the end.
The vowel a in ' parson ' has undoubtedly a diphthongal
character. The first portion resembles the a sound discussed
above (p. 577) in the rising cord tone but differs radically
in the falling cavity tone, in which respect it is somewhat
like the a in ' die ' (Figs. 353 and 355). The latter portion
(Fig. 357, line 13) is related to the earlier portion much as
the i is related to the a in ai in respect to amplitude, the
lowering of the cavity tone and the maintenance of the
cord tone. Although this latter portion is not so long as in
most cases of ai, the resemblance is sufficient to justify the
statement with which this paragraph begins. The sound
might be written af, where the sign t indicates a brief
vowel not yet determined. It may be suggested that this
brief vowel may arise from the weakening of the r, whereby
a vowel sound partially or completely replaces the full r.
It seems, however, to be a general rule, that in English the
long vowels have a diphthongal character.
The sudden fall in amplitude and the change in pitch of
the vowel indicated by f is continued through an interval of
8.8"^ in which 3 vibrations with a period of 2.4'^ appear (line
13, middle). During this time the tongue is presumably
passing to the r position. This portion might be called the
a-r glide.
STUDIES OF SPEECH CURVES 597
The very brief r is distinctly heard in the word 'parson; '
it occupies a time of 63'^ (line 13 middle to line 14 begin-
ning). The r shows clearly 3 ' pseudobeats ' (p. 19) with a
period of 19^ or a frequency of 53. The vibrations within
the beats are grouped in pairs indicating a cord tone acting
upon a cavity. The period of the cord tone is at first
constant at 3.5'^ (frequency 286) but falls slightly in the
third beat. The cavity tone has a period apparently con-
stanb at 1.4°- (frequency 714). Still higher cavity tones
are probably present. The explanation of this curve of r
seems clear. The r consists of a cord tone with a frequency
of 286 acting upon a resonating cavity adjusted to a fre-
quency of 714, The tongue is adjusted to vibrate with a
frequency of 58 ; this vibration of the tongue closes and opens
the air passage so that the intensity of the sound escaping
from the mouth is regularly varied from zero to a maximum
and again to zero at the rate of 53 times a second.
The pseudobeats with the cord and resonance vibrations
are shown in the curves of WbndelerI and in those of
Nichols and Mbbritt (p. 28). The German rolled r of Wen-
DELER has a much longer beat period, in general over 250'^
or \ sec. ; the Finnish r of Pipping has a beat of J to |- sec.^
The American rolled r of Nichols and Meeritt has also
apparently a long beat-period as far as can be judged from
the pictures. The brief r in three examples given by these
last observers seems to have a shorter beat-period than that
of ' parson.' The cord period in Wendelbr's examples
varies apparently from 2.3" to 3.3'' (Wendeler's own com-
putation of a frequency of 200, or a period of o'', can hardly
be correct) ; the cavity tone has a period in the neighborhood
of 1.7"^, according to my calculation from his records. The
later observations of Hermann on r have been given on
pages 44 and 337.
1 Wbndelbe, Ein Versuch, die Schallbewegung einiger Konsonanten und anderer
Gerdusche mit dem Hensen'schen Sprachzeichner graphisch dnrzustellen, Zt. f. Biol.,
1887 XXIII 303, Tafel II, B.
2 Pipping, Zur Phonetik d. jinn. Sprache, Unters. mit Hensen's Sprachzeichner,
Mem. de la goc. finno-ougrienne, XIV Helsingfors, 1899.
598 APPENDIX II
The s follows directly upon the r. The vibrations in the
curve are hardly distinguishable and no very definite limit
can be set to them.
The n follows immediately on s (Fig. 357, line 14 to end).
It occupies an interval of lOT". The successive vibrations
occupy periods of 4.2, 3.5, 5.1, 3.7, 5.3, 4.1, 4.1, 5.3, 4.2,
4.9, 4.9, 5.3, 5.3, 5.3, 5.3, 5.3, 5.6, 5.3, 5.3, 5.6, 5.6, 5.3,
6.7, 6.3, 6.7, 6.7, 7.0, 7.0, 7.0, 7.0, 7.0, 8.4, 8.8, 8.8, 9.1,
8.8. The maximum amplitude is 0.1°™
The tracings on Plate I were made with the double record-
ing lever described on p. 59 ; they belong to Cock Rohin,
Series II. They comprise ohi of sohim ' saw him ' (analyzed
and discussed above, pp. 63, 276), o of bow ' bow ' (see also
pp. 66,433), au of Sraud 'shroud,' o of 'sparrow,' oju of
'draw your.' A preliminary study ^ of these curves is here
condensed and continued.^
The word ' bow ' in ' with my bow and arrow ' appeared
to the ear to be melodious and prolonged ; it might even be
called mellifluous. The tracing (Plate I) gives the curve of
ow. It begins with three faint vibrations that presumably
occur as the mouth begins to open. Thereafter the vibrations
follow in groups of four, beginning with a length of 5.5""" and
decreasing slowly to 4.8"" in the middle of the line ; this indi-
cates a cord tone of rising pitch. The cavity tone remains
practically constant at 1. 5"" for each vibration, or a period of
0.0024" and a frequency of 417.
The amplitude rises steadily to a degree that indicates con-
siderable loudness ; it then falls rather suddenly (middle of
second line). The vibratiors beyond this point show so
many peculiarities that their difficulties can best be attacked
by working backwards from a later point where the grouping
is more regular. About one-third of the distance from the
middle in the second line the vibrations fall into groups hav-
ing two main crests with two subordinate ones. The entire
group arises presumably from one cord vibration. This con-
' ScEiPTUKE, Speech Curves, I., Mod. Lang. Notes, 1901 XVI 72.
^ Tu this I have been assisted by Miss E. Jelliffe, of Mt. Holyoke Seminary.
STUDIES OF SPEECH CURVES 699
elusion is drawn because further on to the right the group
gradually changes to two main crests only, a typical form for a
cord tone accompanied by a cavity tone nearly an octave higher.
Starting from the strong vibrations (third quarter of line 2),
we mark off backward the alternate higher vibrations as the
points of maximum for each cord puff. We thus have the
vibrations in pairs ; the period of the cord tone at any moment
will be given by the distance between two such marked
vibrations.
As we go towards the left, we see that each of the vibrations
of the pair shows a tendency to split up into two minor vibra-
tions; this indicates the presence of higher cavity tones.
Measurements of the periods of the cord tone show that it
steadily rises in pitch from the middle to the third quarter of
the line. They also show that the smaller cavity vibration
keeps very closely at the middle of the cord period, though in
the first portion it is generally a little behind the middle point.
This indicates a cavity tone in general an octave higher than
the cord tone, but a little lower in the first portion. The con-
dition of a cord tone with an octave cavity tone is modified in
the first part by higher tones that do not form an exact har-
monic interval with either of the other tones ; these give rise
to the minor fluctuations. The higher tones are of changing
pitch, as can be seen by the steadily changing form.
The puffs of air from the cords are not generally of the
even nature found in sinusoid -nbrations ; they rather re-
semble more or less sharp explosions. In this sound they
are not so sharply explosive as in au of ' shroud ' or ae of
' sparrow,' yet the puff has its greatest intensity in the first
part of the interval of time it occupies.
A third maximum is found in the latter portion of ' bow '
(third line). This vowel sound is to be considered as a triph-
thong; careful listening to the gramophone plate enables the
ear to hear two maxima clearly and the third faintly. The
maxima are due to coincidence of the cavity period with a
sub-multiple of the cord period (p. 13). As a triphthong
the sound might be writen cow or ouw.
600 APPENDIX 11
The word 'shroud' occurs in ' Who'll make his shroud? '
The portion of the record on line 3 and the first quarter of
line 4 gives the curve of the r with one pseudobeat (p. 19)
at the flat place in line 4. The r-a glide after this is followed
by the long record for au reaching to the middle of line 5.
The latter half of line 5 contains the faint vibrations of the u-d
glide, the still fainter ones of the d-occlusion and the strong
ones of the d-explosion. After the occlusion of the pseudobeat
the tongue again allows the cord-and-cavity vibrations to appear.
The form of the vibration is different, indicating a changing ad-
justment of the mouth from the r-position to the a-position ; this
portion is to be considered as the r-a glide. There is no possi-
bility of definitely limiting the r from the a, or' of marking off a
distinct r-a glide ; the change is gradual throughout (p. 451).
The r shows a rise and fall of amphtude. The occlusion
during the one pseudobeat is complete, as indicated by the
entire cessation of vibrations near the beginning of line 4.
During au the cord tone rises from the frequency 120 to
111 and then falls steadily to 92. The diphthong au is of
circumflex pitch. It is of crescendo-diminuendo intensity, the
crescendo being gradual and diminuendo rather sudden. In
the d the cord tone rises to 109.
The word ' sparrow ' occurs in ' I, said the sparrow.' The ae
of ' sparrow ' begins at the first quarter of the sixth line ; it ends
in r just beyond the third quarter of the same line. The o
extends over the remainder of this line and the whole of the
next. This o is quite different from that in 'bow' above.
The vowel is a crescendo-diminuendo sound ; its amplitude
rises slowly to a maximum and then falls to zero. The vowel-
sound in ' bow ' has three maxima ; the fall from the maximum
is in two cases very sudden. In general the curve of the o of
' sparrow ' differs greatly from that of the o of ' bow, ' although
there is some resemblance of the former to the middle portion
of the latter.
The cord tone of ' sparrow ' has at the beginning a frequency
of 125. This frequency increases to 202 and then falls by
degrees as low as 136. The amplitude increases slowly, then
STUDIES OF SPEECH CURVES Qpl
descends suddenly and almost reaches zero at the r ; after
this the amplitude increases again quite rapidly and contin-
uously during o to a maximum, and then gradually decreases.
The words ' draw your ' occur in the introduction ' Now,
children, draw your little chairs nearer.' The last five lines
of Plate I give the curve for the sounds oju, omitting a piece at
the end. The recording-surface was run at about three times
the speed used for the previous curves. Measurements of the
groups of vibrations show that the cord tone rises from about
the frequency 75 at the beginning of o (line 8) to about 189
(line 10, first quarter), after which it remains practically con-
stant until it begins to fall in the oj glide (line 10, last part).
During the j and u the tone seems to fall steadily.
APPENDIX III
FREE RHYTHMIC ACTION
As a measure of the irregularity in a voluntary act we may
use the probable error. ^ When a series of measurable acts
are performed they will differ from one another, if the unit
of measurement is fine enough. Thus, let x-^, x^, . . ., x„ be
successive intervals of time marked off by a subject beating
time, or walking, or running, at the rate he instinctively
takes. The average of the measurements,
x^ + X2+ ■■■ + x„
n
can be considered to give the period of natural rhythm under
the circumstances. The amount of irregularity in the meas-
urements is to be computed according to the well-known
formula
P = i\'^
2 + V + - + <^
where v^ = x^ — a, v^ — x^ — a, . , ., t)„ = a;„ — a The quan-
tity p is known as the ' immediate absolute probable error. '
The quantity
.P
r =-
the ' immediate relative probable error,' expresses the prob-
able error as a fraction of the average. Both p and r are
1 Scripture, Observations on rhythmic action, Science, 1899 X 807 ; also in
Stud. Yale Psych. Lab., 1899 VII 102.
FSEE RHYTHMIC ACTION 603
called ' immediate ' in order to distinguish them from the
' final ' ones,
F = JL
and
Vn
used to indicate the precision of the average.
If all errors in the apparatus and the external surroundings
have been made negligible, the ' probable error ' is a personal
quantity, a characteristic of the irregularity of the subject in
action. If, as may be readily done, the fluctuations in the
action of the limbs of the subject are reduced to a negligible
amount, this probable error becomes a central, or subjective,
or psychological, quantity. Strange as it may appear, psy-
chologists have never understood the nature and the possi-
bilities of the probable error (or of the related quantities,
'average deviation,' 'mean error,' etc.). In psychological
measurements it is — when external sources of fluctuation are
rendered negligible — an expression for the irregularity of
the subject's mental processes. Nervous or excitable people
invariably have large relative probable errors; phlegmatic
people have small ones. Thus a person vpith a probable error
of 25% in simple reaction time will invariably have a large
error in tapping on a telegraph key, in squeezing a dynamo-
meter, and so on. I have repeatedly verified this in groups
of students passing through a series of exercises in psycho-
logical measurements. I do not believe it going too far to
use the probable error as a measure of a person's irregularity.
This is equivalent to asserting that a person with a probable
■error twice as large as another's is twice as irregular, or that,
if a person's probable error in beating time at one interval is
rj and at another interval r^, his irregularity is rjr^ times as
great in the second case as in the first. This concept is
analogous to that of precision in measurements. We might
-use the reciprocal of the probable error as a measure of
604 APPENDIX III
regularity. The positive concept, however, is in most minds
the deviation, variation or irregularity, and not the lack of
deviation, the non-variability, or the regularity. In the case
of the word ' irregularity ' the negative word is applied to
a (^oncept that is naturally positive in the average mind.
The irregularity in an act is a good expression of its difS-
culty. Thus, if a person beating time at the interval T has
an irregularity measured by the relative probable error R and
at the interval t by the relative probable error r, it seems
justifiable to say that the interval * is ^ times as difficult as T.
If Tis the natural interval selected by the subject, then the
artificial interval t would be more difficult than T, and
we should measure the diificulty by comparing probable
errors.
It is now possible to state with some definiteness the law
of difficulty for free rhythmic action. Let T be the natural
period and let its relative probable error — that is, its difii-
culty — be r. It has already been observed ^ that any other
larger or smaller period (slower or faster beating) will be more
difficult than the natural one and will have a larger probable
error. Thus any interval t will have a relative probable error
r' which is greater than r, regardless of whether t is larger or
smaller than T.
Continued observations during several years enable me to
give an idea of the general relation. The results observed
can be fairly well expressed by the law
-<i-^-^>
in which T is the natural period, r the relative probable error
for T, t any arbitrary period, r' the relative probable error for
t, and c a personal constant.
This may be called the law of difficulty in free rhythmic
action. A curve expressing the equation for T= 1.0% r —
1 ScKiPTURE, The law of rhythmic action, Science, 1896 IV 535.
FREE RHYTHMIC ACTION
605
0.02^ and c = 1 is given in Fig. 360. It will be noticed that
periods differing but little from the natural one are not much
more difficult and that the difficulty increases more rapidly
for smaller than for larger periods. In plotting this curve I
have assumed unity as the value for all personal constants.
These personal constants will undoubtedly vary for different
persons, for different occasions and for different forms of
action.
In case it is desired to know what periods are of a diffi-
culty 2, 3, . . ., w times that of T, a table of values for ^ may
.10
.08
.06
M
■02
1.0 1.5
Fig. 360.
ZO
be drawn up in the usual way and tha value for t sought for
(with interpolation) which gives for / a value 2, 3, . . . , n
times as great. Thus, in a table for the above example it is
found that the periods 0.38' and 2.6' are twice as difficult.
This law can be stated in another form which is of special
interest to the psychologist. To the person beating time a
period of 0 is just as far removed from his natural period as
one of 00 ; both are infinitely impossible. The scale of seconds
does not express this fact; objectively a period of 0 is as
different from a period of 1' as a period of 2' would be.
Similar considerations hold good for the lesser periods ; the
scale by which the mind estimates periods is different from
606 APPENDIX III
their objective scale. This difference may be expressed by-
asserting that the following relations exist between the two :
(t-Ty
X — c- — — ^-,
where x is the measure on the mental scale, T the natural
period, t any other period, and o a personal constant. By
this formula the various periods may be laid off according
to their mental differences from the natural period. Every
difference from the natural period is mentally a positive
matter. With the mental scale the law of difficulty becomes
/ — r (1 + car),
where r' and r are the relative probable errors for t and T
respectively, x the measure on the mental scale, and c a per-
sonal constant. This is the equation of a straight line. The
law states that the difficulty of any arbitrary period is directly
proportional to its mental difference from the natural period.
This law of difficulty as depending on the period is, of
course, only one of the laws of free rhythmic action. It is
quite desirable that other la,ws of difficulty and of frequency
should be determined. For example, observations on ergo-
graph experiments tend to show that the irregularity and the
natural period both change with the weight moved ; they also
change with the extent of the movement.
ADDITIONS AND COERECTIONS
Page 24. — Hekmann's earlier records were made by a beam of light
reflected from a mirror attached to a diaphragm ; see first four refer-
ences in note 1 on page 38.
Page 32. — A simpler and better form of recorder has lately been
devised for the phonograph. The sapphire knife is fastened to a mica
diaphragm with no intervening link.
Page 36. — The earliest reference to the use of large cylinders for the
phonograph seems to be Dussaud, De V amplification des sons dans les
phonographes, C. r. Acad. Sci. Paris, 1899 CXXVIII 552.
Page 37. — To note 3 add Boeke, On the derivation of the curves of
vowel sounds by means of microscopical research of their phonograms, Proc.
Koy. Soc. Edin., 1897-98 XXII 88.
Page 42. — ■ The record for x shows pseudobeats resembling those of
uvula r; see also p. 461.
Page 61. — Gramophone (or zonophone) discs are now made of large
diameter; the records are remarkably clear and truthful. Discs contain-
ing typical English sounds are now being traced off.
Page 63. — The statement concerning sohim should bo corrected as
indicated on page 277.
Page 75. — The results of the analysis of the curve in Fig. 49 are not
correctly given; see pp. 567, 572.
Page 248. — The pull on the tongue must modify to some extent the
adjustments of the larynx.
Page 313. — In Figs. 192 and 198, ca, ja should be xa, xa.
PHONETIC SYMBOLS
Some of the less frequently used symbols are not included in this list ; their
meanings are explained in the text.
The following language-names are abbreviated to their initial letters : Ameri-
can English, British English, English, French, German.
a — A. 'yacht, ah.'
ae — E. 'pat,' A. 'pass.'
b — E. b.
p — Dutch w.
c — surd palatal occlusive.
9 — G. ' recAt, buc/ier.'
5 — E. ch.
d — E. d, F. d.
8 — mouille d.
S — E. sonant th.
e — F. e, e, G. eh, a,.
a — E. 'her, fungus,' F. 'pre-
mier,' G. ' gebirge.'
f — E. f, F. f.
g — E. " hard " g.
■^ — mouille g.
Y — sonant fricative in G.
' sa^e.'
— E. h, G. h.
— E. 'pick, pique,' F. 'pique.'
— sonant palatal occlusive.
— North G. 'jageT.'
— E. ' yet; F. ' lieu.'
-E.j.
h
i
J
J
J
J
k — E. k.
K — a k-sound (k, K, c).
K — mouille k.
X
— G. ' acAt, bucA.'
1
— mouille y^.
1
— E. 1, F. 1.
A
— mouille 1.
m
— E. m.
n
— E. n, F. n.
n
— mouille n.
11
— E. 'sinffer, fireger, sink.'
r»
— mouille t|.
o
— F. o, G. 0.
0
— E. ' gnawed,
B.
' nod.'
oe
— F. eu, G. 6.
P
-E.p.
<t>
— Japanese f.
r
— flapped r.
J
— unflapped r.
s
— E. surd s.
cr
— mouille s.
V
s
— E. sh, F. ch.
G.
sch.
0-
— mouille s.
t
— E. t, F. t.
T
— a t-sound (t,
t).
T
— mouille t.
e
— E. surd th.
u
— F. ou, G. u.
V
— E. V, F. v.
w^
— E. 'way; F.
'loi
, loMis.'
A\ — surd w.
y — F. u, G. ii.
q — F. 'lui.'
z — E. z, F. z.
PHONETIC SYMBOLS 609
1^ — mouilM z.
z — E. ' vision,' F. j.
r — mouille I.
' — G. 'be()enden.'
Smaller letters on the line indicate weak sounds.
'' — aspiration, as in G. ' t()at().'
" — nasal modification of preceding sound.
" — labial " " " "
' — palatal " '' " "
^ — sonant " " "
„ — surd " _" " "
The combination ^„ means that the sound is partly sonant,
partly surd.
1, 2, 5, etc. are sometimes used to indicate different varieties of
a sound, as explained in text.
Quantity is sometimes roughly indicated by the macron (long)
and breve (short).
Accent is occasionally marked by an acute over the vowel.
39
^
S
a
a
lO
02 t- CO
t- CO
t—
CO
(M
tHt-h
t-
c
<D
OJ
.s
-+
0 -* CD
0 C-1
CO
CM
CO
0 CO
r^
T
^
d
S
00
00 00 <M
CO
03
-f ^ ^0
CO Ci
CM
c
c
a;
rt
CO
CO CM 05
^
1—1
GO CO -^
CM c:^
Oi
0
&
o
o
^
CO CO M
C>4
CM
T— 1
tH
T— 1
1-1 1-1
i
s
0
^2 Tg ~»
c c
'T! 9 J=-i
4^ «*
« 5
O
I— I
H
<1
O
a
I— I
m
s-^ .= .H ^ g -s-
Hn c«^ n|m i-h|m w|co
i
CDOCOCMOO'*c>OOCOCOiOO(M'*OCOC
g a|^6| g
05^1— It— COCOO'+ICMOCOCOOOlOCMCDLOCi
N
OOO-TtiOt-LOCOOOit— lOCOCMTHOOJOOt-
=-s-|a§ = -
-*COCOCO<MCMC<l<Mi-lTHTHi-(-i-lT-lr-l
Frequency,
half vibra-
tions (even
tempera-
ment)
[French].
LqTjHOt-i-JTHCMTlHCMOC'Oi-lT-JCDCMT-HOiT
coc<^oO'*-*Lra"o6cDc^o'^^t-^'^^05cooc
t-THCDO<M— M<C0O00OC0OCMCD»0-+IU
CMCX)COCMlOCMCOi-HC^'^-r^t~COCOOC^t— u
OOt-t-COiOiO^-+COCOCOC<l(MCMCMi-(rHT-
Frequency,
complete
vibrations
(even tem-
perament) .
Tj; CM 0 CO T-H tj^ CO c<i T-i 0 oi 0 uo CO CO lo 0 T-
OOCDo'tOCM't-^CMoicOOOT-icOr-I^cdOir
C0O00OCDOCMCD".0-*iO00OC0C0t--b-t-
1— lOSTtir- it-COCOOCJit— lOCOCO-rMOCJOOt-
-*C0C0C0C5CMCMCMrHT-lT-lTHT-lT-H,-|
t- to to CO «0 O CD '■-D lO lO 10 lo !.-»
Viergestrichene Octave Dreigestrichene Octave Zweigesi
\: i< "■« "^^"'^ "^ \ i<i ~>S ""^S^^^ "^ \ 5^ %3 ^-
Fol-
lowed
in this
book.
Fourth octave Third octave
Sec
(
>>,'^TtO (MH
03 <S t~
O O '-O
to 1— ( ;o
T^l Ttl CO
o o
o t~
CO <M
■-ti CO en
O lO oi
CO O 00
CM (M tH
O O
iC CO
CI 'O -H
id ci i-i
T-( W' 03
CO CO
I— in
CO CO t-;
b-'i-l LO
lO lO Tfl
lO CO
00 -rt!
CO CO
>»-.«,
« ttj'
^
« 5>i=e,
& g ©
'Ij
OD OS
^il
^.'^
S S 3 S s
•^
L-^1-?^,-S»°'^'^ OCO^jT-jOOOOOCOCNOO^OOOCDCOiOOCMTjHOCOOOCqOCDCq
r'OacOSO'.-OlO ■^■*C0COCOCN(N(MC^T-lt-li-li-lTHTHT-l
coMwq?ioM?q iqOi-;u3oqt-_co(Noioco_05cot--rHicoo'ffl05C^coTHaooicoiotot-
WtSMOOrtSlH-* CO O lO O T-H O b-^ 00 lO b-^ LO lO O OC? rjH t~^ CO CN (>i lO Oi C<i OO' to' CO T-H (>i ^
(N!OU3-t«)COOIOCOt-lr-t-ailOOOT-IOOCOOOTtH<MCiiLO-^t-iait-COTtl(M<Moa3000Dt-CO
MOat-lOeOMHO OSCOt-COCOLOLO-^-^COCOCOOltMCMtMT-lT-l-i-liHTHT-lT-l
J)(MHrlHnrtHr-i
3COlOOH>00)NCO_ CNOlOCOOSCOt— -I— I'OOOCOOitMCOi— lOOOSCOLOCOt— O-^-t^dOOCOCO
■ T(!ffldiddr^;i>-a5LOb-^L6Looo6Tj5i>^co(?4c^'Looi(No6cd<fflTH(N-^T-H-H^
ICOt-Nt-ffiCSi-IOOCOOO-^CMOSLO-^T-iast-CO-^tNC-lOOSOOOOt— COCOLO'*-^-*ICOCO
005a)N-;'-'LOTj<TtlCOCOCOH<M(M(MT— li— Ii—ItHt— (tHt— I
T ■*
CO
CO
^
ffJ
■"?1>
■t-J
e«a
<l
«;>
rS
**w
^
-.a
"'«^-l
•+0
1 1
s
f^
r-.=
^ ^
rsg
GO
s
5.
;s
00
5.
5S
CC
CO
5>
00
^
^-^
CO
^i
^
Contra Octave
Zireigestrichene Odtsve Eingeatrichene Octave Ungestrichene Octave Grosse Octave
Second octate First octave
Zero octave FirEt negative octave Second negative octave
0000000^^'-tr-<'-^'-''-'C«CaClC<lClMC3
~ tji'-*~i « 1^ ^ -<> is bs'^ 'a '>3
INDEX
INDEX
a, 26, 28, 32, 39-43, 49, 114, 115, 122,
221, 222, 223, 224, 227, 266, 287, 305,
307, 312, 318, 321, 330, 332, 333, 336,
339, 342, 343, 344, 352, 403, 595
ai, 114, 122,400, 402, 576
au, 114, 122
a", 315, 339, 345, 359
BE, 117, 330, 600
B, 459
Abdominal breathing, 212
Abstract rhythm, 552
Accent, 506-516
Accessory nasal sinuses, 339
Accommodation in vowels, 410 ; in the
ear, see Tensor tynjpani
Accuracy, of judgment, 104; of move-
ment, 385
Acoustic impressiveness, 114
Acoustic penetration, 115
Action, voluntary, 188 ; rhythmic, see
Ehythmic action
Acusticus nerve, 194
a^wah, 465
afwas, 465
Adam's apple, 339
Adaptation, 454, 467 ; see also Assimila-
tion
addentro, 466
Addition of curves, 67-70
Adjarian, 369
Affricates, 120
Age, influence on highest audible sound,
99
aggettivo, 446
ain, 274
alTT6Kos, 372
Air, surplus, 224; see also Breathing
Air transmission, 195 ; see also Tam-
bour
Alcaic, 537
Alexia, 86
d\K^, 461
Alliteration, 172
American, 302, 305, 358
Amphibrach, 535
Amplitude, 2
Analogy, 170
Analysis, immediate, 62-71 ; harmonic,
72-75, 561-574, see also Fourier ; in
ear, 82; of tone complexes, 112
Anapest, 535, 553, 556
Anterior pillars, 232
Anticipation, 165
Anvil, 78
Aphasia, 84, 128
Aphonic consonants, 443; see also Surd
Appendices, 559
Arabian, 274
Arches, 232
Aristotle, 93
Aristoxenus, 268, 473
Armenian, 369
Arsis, 537
Articulation, 326, 427; basis of, 113
Artificial larynx, 258
Artificial palate, 298
Arvers, 545, 554
Aryepiglottic muscle, 242
Arytenoid cartilage, 235, 244
Arytenoid muscle, 241
Aschaffenburg, 146, 156, 157, 160
Aspiration, 120
Assertion, 218
Assimiliation, 164-172, 372 ; see also
Adaptation
Assimilatory condensation, 167
Association, of ideas, 135-151 ; of move-
ments, 377
Association fibers, 193
Association Phonetique Internationale,
428, 445
Association time, 152, 155, 208
Associative effectiveness, 164
Associative stammering, 158
614
INDEX
Associative suggestion, 115
Atkinson, 330
Attack, 429, 434
Audible, see Lowest, Highest, Sliortest,
Faintest
Audiometer, 109
Auditory aphasia, 85
Auditory basis of speech, 113
Auditory economy, 121
Auditory habits, 113
Auditory ideas, 126 ; see also Internal
word
Auditory impressiveness, 448
Auditory learning, 181-186
Auditory memory. See Auditory
learning
Auditory preference, 123
Auditory rhythm, 517
Auditory words, 83
Auerbach, 106, 288, 413
avKd, 461
Aural, see Auditory
Auricle, 76
Auzoux, 192
Average, 156, 201
b, 47, 114, 131, 203, 224, 225, 226, 285,
307, 317, 333, 339, 342, 344, 355, 357,
358, 376, 594
P, 47
Babbington, 247
Bagley, 131
Balassa, 305
Band, ventricular, 242 ; vocal, 242
Bar, 98
Barlow, 17, 607
Basis of articulation, 113, 377 ; of audi-
tory perception, 113
Bass register, 272
Batteries, 20$
Beating time, 538
Beats, 99, 106
Bell, 133
Bergstrom, 158
Berliner, 52
Bert, 214
Bevier, 49, 75, 412
*bhero, 458
Bigham, 181
Binet, 157, 493
Blake, C, 18, 99
Blake, E. W., 24
Blake transmitter, 26T
Blind, 132
Boeke, 32, 37, 75, 607
Bohemian, 499
bok, 464
Bolton, 520
Bonn dialect, 362
bouche, 465
Bourdon, 160, 500
Brain, connection with ear, 82 ; speech
centers in cortex, 83 ; general struc-
ture, 193
Breath, 268, 276
Breathing, 212-228
Breath recorder, 220
Breathy tone, 273, 274
briller, 461
bring, 460
British English, 103
Broca, 83
Brondgeest reflex, 382
Browning, 168
Briicke, 276, 443, 512, 537
Brugmanu, 168
Bulb, 192
Buccinator muscle, 231
buoh, 464
c, 437
cq, 439
g, 47, 114, 309, 328,344, 376
c, 224, 304, 305, 307, 321, 333, 368, 439,
441 ; see also ts
Cacuminal, 297
Camera, 73
Caual, auditory, 76; semicircular, 79
Canine muscle, 232
capio, 464
Cartilaginous glottis, 240
Catch, see Glottal catch
Cattell, 156
Cavity vibration, 281, 420
Cellefrouin, see Rousselot
Center of density, 126, 163
Centers, in cortex, 83 ; of muscular
control, 86; of speech, 86; of reflex
action, 192; in the bulb, 192; of
breathing aud laryngeal movements
246
Ceutroid, 448
INDEX
615
Centroid interval, 544
Cerebellum, 192
Cerebral, 296
Cerebrum, 192
Certainty, 103
Change, progressive, 462-471 ; just per-
ceptible, see Just perceptible cliange;
phonetic, see Phonetic change
Charriure, 247
Chavanon, 24
Cheney, 53
Chest register, 259, 260, 272
Chest tone, 222, 288, 294
chevals, 461
chevaux, 461
Child speech, 1 1 9, 460, 468
Chin key, 154
Chlintement, 396
Choana, 339
Choice, 208
Chondroglossus muscle, 236
Chord, 95
chortus, 470
Chronometer, 200
Chronoscope, 152
Circumflex amplitude, 504
Circumflex pitch, 476
Cities, poem, breath pressure in recit-
ing, 218; studied as verse, 355
Clockwork drum, 198
cnihtas, 460
Cochlea, 79
Cock Robin, 58, 62, 63, 126, 218, 276,
400, 484, 547, 553, 375, 595
Comerius, 181
Compensation, 461
Complex tone, 95
Compound tone, 95
Compound words, 127
Comstock, 589
Condensation, 167; see also Smoothing
Consciousness, 82, 107, 380
Consonant i, see j
Consonant intervals, see Musical inter-
vals
Consonants, 432-445 ; Hermann's curves
of, 43-49 ; high tones in, 99
Consonant u, see w
Constrictors of the pharynx, 237
Contact wheel, 12, 91, 92
Contamination, 166
Contiguity, 135
Contraction of muscle, 188
Contrast, 135, 148
Control of muscles, 86
Convolutions, 83
Coordination, 457
Coradi, 73
Cord, 192
Cordes, 129, 142, 146
Cord tone, 267, 413
Corniculate cartilage, 240
Cortex, 83, 193
Corti, 81
Costal breathing, 42
Coup de glotte, 428
Cricoarytenoid muscle, 241
Cricoid cartilage, 239
Cricoid muscles, 241
Cricothyroid muscle, 241
Cumberland, 133
Current of thought, 124
Curtis, 112, 133, 206
Curve, sinusoid, 3 ; frictional sinusoid,
6 ; adder, 68 ; specimen for analvsis,
74
Curves of speech, 1-75
Cushion, 257
Cushion pipe, 258
Czermak, 247, 257, 274, 276, 342, 343,
344
d, 47, 131, 224, 226, 285, 305, 316, 317,
321, 333, 336, 339, 342, 344, 376
dz, 307, 321, 442
d2, 224, 304, 307
8, 304, 316, 440
6, 117, 305, 595
d mouille', 304, 305
Dactyl, 535, 553, 556
Danish stod, 279
D'Arsonval, 152
David, 461
dead, 468
Deep bass register, 272
Definiteness in movement, 204
Deprez marker, 92
Demeny, 353
Dennert, 94
Density, 126, 163
Depth of breathing, 213
Desonation, 203
Diagram, sagittal, 296
616
INDEX
Dialectal progressive change, 462
Diaphragm, 212
Difference, see Just perceptible differ-
ence, Imperceptible difference
Difference tones, 99
Differences in auditory perception, 463
Differences in structure of vocal organs,
463
Differential audiometer, 109
Digastricus muscle, 234, 335
Diminished vitality, 467
Diphthong, 430
Diphthong, nature of, 20 ; rise from
long vowel, 103, 122
Discrimination, 208
Discrimination of speech sounds, 113
Dissimilation, 164, 172, 203
Distinction, 456
Disturbance in brain, 87
Dodart, 255
Dodge, 128
Donders, 268, 288
Double occlusives, 466
Double octave, 104
Drum, 7, 198; continuous-paper — , 8
Duodecime, 107
Duration, 89, 91, 488-502; method of
measuring, 500
Duration rhythm, 517
Dussaud, 607
Dutch vowels, 32 ; diphthongal, 459
e, 26, 39, 43, 103, 114, 115, 117, 122,
222, 227, 266, 287, 302, 305, 306, 307,
309, 312, 318, 321, 330, 332, 333, 339,
343, 344, 352, 404, 466
ei, 103, 122, 332
e», 315, 318, 339, 346
a, 117, 305, 328, 329, 465, 595
Ear, 76-88; function in rhythmic ac-
tion, 529
Ear bones, 78
Ear drum, 77
Ease of innervation, 454
Ebbinghaus, 178
Ebhardt, 532
Economy, 121, 172, 466
Edison, 32
ef. 465
Electric fork, see Fork
Electric motor, see Motor
Electrical resistance, see Resistance
Element, see Phonetic
Elements of speech, 113-125
Elevator of velum, 232
Emotional tinge, 89, 111, 174
Emotions, effect on muscular control,
390
Energy, decrease of, 463 ; increase of,
465, 467
Energy of sound, 109
English vowels, diphthongal, 458
enhance, 168
Entrance, 429
Entrance of sounds into consciousness,
107
Epiglottis, 229, 240, 246
Equality judgments, 104
Erdman, 128
Error, of execution, 201 ; of perception,
202 ; of movement, 202
Euler, 418
Eustachian tube, 78
Ewald, 152, 258, 261
Exaggeration, 467
Exaggeration of differences, 122
Excess of energy, 465
Execution, error of, 201
Exit, 429
Bxner, 106, 249
Expenditure of breath, see Breathing
Expiration, 212, 220
Exploratory bulbs, 332
Explosive, 223, 224
Explosive, mouiUe, 440
Expression, 468
External association, 157
External ear, 76
External pterygoidal muscle, 230
Extravagance, 467
f, 47, 89, 114, 117, 224, 226, 302, 307,
315, 317, 329, 333, 336, 342, 376
Faber, 290
facere, 460
Facialis nerve, 1 94
Factors of speech, 399
fahl, 464
Faintest audible tone, 109 ; audible
speech sounds, 114
faire, 460
Faist, 105
INDEX
617
fallow, 464
Falsetto, 118 ; see also Head register
Familiar habits, 468
fate, 103
Fatigue, 205
Faults of perception in speech, 113-125
Favored association, 160, 170
Fechner, 104, 109
Fedor, 117
Feueiitra ovalis, 78
Fenestra rotunda, 80
fero, 458
Ferrari, 1 60
Ferrein, 255
Fihers, 193
Fiftli, 104
filium, 465
fille, 461
Fillmore, 426
Filial vowels, 203
Final surplus air, 227
Finger movements, 200
Finnish vowel harmony, 121, 204
Firnmess of association, 159
Flechsig, 380
Fluctuation of effort, 202
Fluorescent screen, 238
Foot, defined, 553
Foot, inverse, 537, 538, 552
Force of movement, 383
Forced vibration, 420
Foreign language, see Learning
Forgotten associations, 147
Fork, 1 5 ; pure tone, 89 ; for lowest
tone, 93 ; for highest tone, 98
Formant, 39
Formation of speech associations, 175-
187
Formula of sinusoidal vibration, 2, 4 ;
of frictional sinusoid, 6
Fourier, 72, 73
Fourier analysis, 72, 559-574
Fourth, 104
Free rhythmic action, 602
Free rise of ideas, 148
Free vibration, 2, 286
French, 249, 251, 357
French, acoustic penetration of various
sounds, 114
French palatograms, 312
Frencli vowels, 26
Frenum linguaj, 237
Frequency, 4, 64, 65
Frequency of association, 149
Fricative, 223, 224
Fricatives — >• occlusives, 466
Friction, 5
Frog muscle, experiment with, \i
Frontal lobe, 192
Functional association, 149, 167
Fundamental, 72, 96
Fusion, 446
g, 47, 119, 131, 203, 224, 226, 285, 302
315, 317, 321, 333, 340, 342, 344
\, 440
7, 224, 461, 465
(J mouille, 305
Gad, 220
gadge, 168
Galen, 255, 391
Galilei, 94
Galle'e, 336, 355, 370
Galton, 98
Garcia, 247, 260, 266
Gauthiot, 515
gegen, 465
Gelle', 106
General voluntary control, 392
Genioglossus muscle, 235
Geniohyoid muscle, 234, 335 ; tambour,
335
Gentzen, 342
Geographical progressive change, 462
German, 357 ; acoustic impressiveness
in, 114; weakest audible sounds of,
114
German palatograms, 308
*ghend, 450
*ghortus, 470
gladly, 460
Gladstone, see Self-Help
gliedlice, 460
Glass recorder, 9
Glass tubing, 217
Glossopalatine arch, 232
Glossopalatine muscle, 233
Glossopharyngeus nerve, 194
Glottal catch, 119, 223, 268, 278, 429,
515
Glottis, 240, 245
Goldscheider, 128, 224
Gramophone, 52-61, 575, 607
618
INDEX
Gramophone tracing apparatus, 55-61
Grandgent, 327
Graphic chronometer, 200
Graphic method, 188, 211
Graphic recorder, 9
Grassmann, 407
Gre'goire, 447, 494
Grimm's law, 470
Groan, 215
Grutzner, 265, 266, 308
Guicciardi, 160
Guttural vowels, 427
Gutzmann, 343
Gyrus angularis, 86
h, 114, 119, 120, 171, 276, 457, 589;
sonant h, 24, 276
Habits, 468 ; auditory, 113 ; speech, 119 ;
association, 152-174; interference of,
158
hafjan, 464
Hagelin, 316
Hair cells, 81
Hallucination, 115
Hamburg dialect, 362
Hammer, 78
Hansen, 132
Haplology, 167
Hard palate, 229
Harmonic analysis, 72-75 ; analyzer,
see Synthesis; series, 13, 72; vibra-
tion, 2
Harmonic dissimilation, 172
Harmony of vowels, 121, 204, 372, 588
Head register, 260, 272
Hearing, 76-88 ; function in rhythmic
action, 531
heben, 464
Helmholz, 405 ; on cord action, 268 ; con-
sonants, 106 ; difference tones, 99 ;
ear, 97 ; summation tones, 100; vowel
instruments, 291 ; vowel tones, 287
Hemispheres, 192
Henri, 157, 493
Henrici, 573
Hensen, 18, 269, 416
Herbart, 135
Hermann, 307, 405, 411, 422, 442, 561,
570, 571, 579, 607; on analysis, 75;
cord action, 268 ; intermittent tones,
95 ; tracings, 38-49
Hexameter, 537
Hiclcey, 425
High vowels, 427
Highest tone, 98
hijo, 465
Hipp, 152
Hirt, 513
Hochdorfer, 327
Hodge, 393
Hooley, 575
hortus, 470
Hungarian palatograms, 306 ; vowel
harmony, 121, 204
Hurried movements, 387
Ilurrieduess, 204
Hurst, 538
Hyoglossus muscle, 235
Hyoid bone, 233
Hypoglossus nerve, 194
i, 26, 39, 43, 64, 103, 114, 115, 122, 222,
223, 224, 227, 266, 287, 302, 305, 306,
. 307, 314, 318, 321, 330, 332, 333, 336,
339, 342, 343, 344, 376, 432, 594
i ->8, 466
i, consonant, see j
Iambus, 535, 537, 553, 556
Idea, defined, 126, 132, 136, 137
Identification of similar sounds, 120
Ideogram, 128
Ideo-motor associations, 389
Ideophone, 132
Ideophonic texts, 428
Imitation, 218
Immediate probable error, 155
Imperceptible difference, 103
Impressiveness, 448
Incisivus muscle, 231
Incus, 78
Independent sounds, 454
Inductorium, 189, 207
Inferior longitudinal muscle, 235
Inner ear, 78
Inspiration, 212
Instruction, see Learning
Intensity, 89, 91, 109,221
Intensity and interval in rhythmic ac-
tion, 532
Intensity rhythm, 513
Interference experiments, 423
Interference of association, 158
INDEX
619
Intermittent tone, 94
Internal association, 158
Internal ear, see Inner ear
Internal pterygoidal muscle, 230
Internal speech, 132
Internal word, 132
Interrupter, 267
Intervals, 104
Intonation, 251
Inverted, 297
Involuntary whispering, 132
Irish, 305
Irrational rhythm, 552
Irregularities in pitch, 202
Irregularity, measure of, 602
Italian, 347, 365
Italian palatograms, 321
j, 47, 131, 328, 333, 344, 432
j [9] -> 8 [t], 466
], 224, 306, 307, 316, 317, 318, 324, 329,
368, 376, 432, 601
] -^ S, 466
J, 304, 305, 321, 368, 441 ; see also dz
Jaw registration, 355
Jefferson, 479, 489, 551
Jefferson record, 61, 276
Jelliffe, 598
Jespersen, 296, 353, 457
Josselyn, 118, 321, 347, 350, 365, 442,
500'
Jost, 178
Just perceptible difference, 100, 104,
111, 114
k, 45, 114, 119, 120, 131, 203, 224, 226,
285, 302, 306, 307, 315, 317, 321, 333,
342, 344, 376, 434
k, see also k mouille
k — > X or h, 464
K, 437
Ko-, 439
KX, 439
X, 46, 114, 224, 328, 344, 461, 465, 607
J, 436
g->c -^X. ''64
k mouille', 305, 315, 436, 438, 440, 442
Karsten, 453
Kempelen, 290
Kemsies, 184
Key, 154, 207
X^lpofiai, 458
XP^voi TrpwTot, 552
Kinetocamera, 353
Kingsley, 298, 302
Kirkpatrick, 181
Klunder, 270
Knights, 460
Kojnig phonantograph, 17; manomet-
ric flames, 26 ; synthesis of curves,
68 ; intermittent tones, 94 ; highest
tone, 98
Koppel, 168, 171
Koschlakoff, 262
Krapelin, 147, 160
Krdl, 499, 537
Kriiger, 99
Kiilpe, 543
Knrschat, 514
Kymograph, 198
1, 19, 43, 114, 117, 120, 131, 203, 227,
304, 307, 308, 310, 316, 318, 321, 333,
344, 432, 592
1 -> X, 466
S, 119, 224, 307, 316, 318, 324, 440,
466
/ mouille, see A
Labials, mouille, 441
Labialization, 364
Labyrinth, 79
Laclotte, 122, 372
Lahr, 413
Lamp, batteries, 209
Language habits, 156 ; see also Pho-
netic basis
Lantern recorder, 9
laogh, 459
Lapses, 117, 130, 163
Laryngeal ventricle, 243, 265
Laryngoscope, 247
Laryngostroboscope, 249
Larynx, 229, 239
Larynx, registration from, 266-268
Latent time, 92
Laugh, 215
Lautstottern, 158
Law of free rhythmic action, 605
Laws of association of ideas, 135
Lax and tense, 384
Lay, 182
620
INDEX
Learning associations, 150; languages,
175-186 ; new sounds, 205 ; sounds,
124; syllables, 113, 175, words, 133
Least perceptible, change, see Just per-
ceptible change
Least perceptible difference, see Just
perceptible difference
Lebedeff, 25
Lehraann, 133
Lenz, 296, 308, 434, 464, 466
Lettic, 513
Lichtheim, 86
Lieben, 51
Ligamentous glottis, 240
Likeness, 103
Lines, in verse, 554
Linguistic unit, 126
Lip key, 154
Lip position, 352
Liquids, 432-445
Lispers, 395
Lithuanian, 513
Lloyd, 291, 424,426, 571
Lobes, 192
Logograph, 17
Long, 501
Longitudinal fibers, 193
Longitudinal muscles of tongue, 235
Lord's prayer, 58, 485, 576
Lost sounds, 364
Loudness, 502-505
Low vowels, 427
Lowest tone, 93, 98
Lubrifactiou, 264
Ludwig, 257
m, 19, 27, 44, 114, 117, 224, 307, 317,
333, 340, 342, 343, 355, 358, 432
Mach, 106
McKay, 538
Mackenzie, 273
Magnetic marker, see Marker
maison, 465
Major, 95, 104
Malleus, 78
Manometer, 225
Manometric fiame, 26-31, 73
Marbe, 160, 170
Marcet, 220
Mares, 499. 537
Marey pneumograph, 211
Marey tambours, 195
Marichelle, 51, 479
Marker, 91-93, 207
Martens, 475, 489
Martens, phonantograph curves, 19
Masseter muscle, 230
Mayer, 108, 130, 143, 144, 163
Measure, 551
Median septum, 234
Mediate association, 145
Medulla, see Bulb
Meillet, 515
Meinong, 105
Melica, 461
Melody, 472
Membrana "basilaris, 80
Membrana tympani, 78
Membrane, 257
Memorization, 175-186
Memory, 123, 137
Mental intensity of sounds, 109
Mental work, 101
Meutalis muscle, 232
Meringer, 130, 163
Merkel, 327
Merritt, 28, 589
Mersenne, 94
Metathasis, 172
Method, graphic, 188
Method of teaching, see Learning
Meunier, 320, 398
Meyer, 105, 106, 450, 478, 538, 543
Michels, 173
Microphone, 267
Middle ear, 77
Middle register, 272
Mid vowels, 427
Minor, 104
Misprinted words, 128
Misreadiugs, see Mistakes
Mistakes of perception, 128-132
Mitford, 571
Miyake, 279, 509, 529, 539
Model to illustrate vibration, 6
Mora, 501, 551
Morgagni, 243
Mosso, 393
Motor, for drums, 10
Motor aphasia, 84
Motor economy, 123
Motor learning, 183, 186
Motor nerves, 191
INDEX
621
Motor weakening, 463
Motor words, 83
mouche, 465
Mouillc sounds, 304, 305, 307, 315, 316,
318, 440, 441 ; see also iC, n, t, 8
Mouillure, 434, 441, 464; see also
MouilM sounds
Mouth, breath pressure from, 217
Mouth mapping, see Tongue positions
Mouthpiece, 219
Movement, 84 ; error of, 202
Mucous membrane, 243
Mucus, 244, 265
Muller, 113, 128, 175
MuUer, C, 257
Muller, J., 257, 264, 512
Munsterburg, 129, 160, 181
Muscle, aryepiglottic, 242 ; arytenoid,
241 ; buccinator, 231 ; canine, 232 ;
chondroglossus, 236 ; cricoarytenoid,
241 ; constrictor of pharynx, 237 ;
digastricus, 234 ; elevator of the
velum, 232 ; external pterygoidal,
230 ; gastrocnemius, 188; genioglos-
sus, 231 ; geniohyoid, 334 ; glossopala-
tine, 233 ; hyoglossus, 235 ; incisivns,
231 ; inferior longitudinal, 235 ; in-
ternal pterygoidal, 230 ; masseter,
230 ; mentalis, 232 ; mylohyoid, 234 ;
oblique arytenoid, 242; orbicularis
oris, 230 ; phstryngopalatine, 233 ;
petrygoidal, 230 ; quadratus, 231 ;
risorius, 232 ; stapedius, 79 ; sterno-
hyoid, 242 ; styloglossus, 234 ; stylo-
pharyngeal, 242 ; superior longitudi-
nal, 236 ; temporal, 230 ; tensor of
the velum, 233 ; tensor tympani, 78 ;
thyroarytenoid, 240; thyrohyoid, 242;
transverse arytenoid, 241; transverse
lingualis, 236 ; triangularis, 231 ;
uvula, 233 ; vertical lingualis, 236 ;
vocal, 240 ; zygomatic, 232
Muscles, control of, 86 ; curve of
contraction of, 189; nature of, 188
Muscles of the ear, 78
Muscles of the face, 230
Muscles of the larynx, 240
Muscles of the pharynx, 277
Muscles of the tongue, 234
Muscles of the velum, 232
Muscular process, 240
Musehold, 258, 259
Musical intervals, 104-106
Myograph, 189
Mylohyoid muscle, 234, 335
n, 1 9, 27, 44, 1 14, 1 1 7, 203, 224, 302, 306,
307, 316, 317, 324, 333, 336, 339, 342,
343, 432, 597
n-J-ii, 466
n, 224, 306, 307, 316, 318, 324, 339, 440,
466
1), 117, 303, 307, 324, 339, 342, 343, 432
Jl, 440
n-mouille', 440
i)-mouille', 440
Nagel, 29
Narrow vowels, 427
Nasal, 224
Nasal cavity, 229, 339
Nasal olive, 219
Nasal sinuses, see Accessory nasal si-
nuses
Nasal twang, 347
Nasal vowels, 339
Nasal whispering, 132
Nasalization, 346, 359
Nasopharyngeal meatus, 339
National progressive change, 462
Natural interval in rhythmic action, 528
Natural period, 2
Natural rate, 204
Neale, 305
Neglect of a sound, 459
Neglect of difference, 122
Nernst, 51
Nerve, acnsticns, 194; facialis, 194;
hypoglossus, 194 ; glosaopharyngeus,
194; laryngeal, 246; motor, 191 ; of
ear, 81-82 ; sciatic, 190; sensory, 191 ;
trigeminus, 194 ; vagus, 246
Nerve-muscle preparation, 190
Neutral, 427
Nichols, 28, 589
Nichols and Merritt, 412
Noise, 89
Noiseless key, 529
Nose, registration of breath, 217
Nostril, 229, 339
Note, 256 ; see Tone
nothing, 117
Nourishment, effect on vocal action, 392
nuffin, 117
622
INDEX
o, 26, 29, 39, 43, 66, 103, 114, 115, 122,
222, 223, 224, 227, 266, 287, 305, 307,
315, 320, 321, 329, 333, 336, 339, 342,
343, 344, 598, 600
oi, 122
ou, 103
o°, 315, 339, 346
3, 39, 63-67, 329-330, 339, 343, 352, 403,
404, 600, 601
oi, 114
ce, 39-43, 114, 115, 287, 306, 307, 309,
314,320
oe", 315, 339, 345
Objectivity, 89, 111
Observation of organs of articulation,
238
Observation of larynx, 239
Occipital lobe, 192
Occlusiveness, 466
Octave, 104
Oertel, 113, 262
Ohm, 97
Olive, 219
On-glide, 429
Oral cavity, 229
Orbicularis oris muscle, 230
Organ of hearing, 76-88
oriie, 461
Orth, 143
Ossicles, 78
Oval window, 78
Overtone, 72, 96, 256 ; in musical inter-
vals, 106
p, 45, 114, 131, 203, 224, 226, 285, 307,
317, 333, 344, 355, 357, 358, 374, 595
p-^<|>->f, 464
p, 438
Palatal vowels, 427
Palate, 229
Palatogram, 296
Palatine tonsil, 232
pallidus, 464
Pantograph, 73
Parietal lobe, 192
Parisian dialect, 312
Partial, 72, 106, 256
Passy, 459, 461, 464
Paul, 149, 166, 453
Pause, in rhythmic groups, 534
Pause rhythm, 517
Pearson, 68
Pendnlar vibration, see Sinusoid
Pendulum chronoscope, 152
Penetration, acoustic, 115
Perception, 208
Perception, errors of, 202
Perception, dependence on production,
118 ; of sounds, 89 ; of speech, 75 ; of
speech elements, 113-125
Perceptive economy, 123
Periodic change, 95
Periodic time, 2
Persistence of sounds, 107
Personal progressive change, 462
peuple, 460
Pfeil, 91
Pharyngeal cavity, see Pharynx
Pharyngopalatine arch, 232
Pharyngopalatine muscle, 233
Pharynx, 77, 229, 338
Phase, 4
Phonautograph, 17-24
Phonetic basis, 113
Phonetic change, 114, 116, 118, 123,
127, 158, 169,365
Phonetic element, 1 27
Phonetic laws, 468
Phonetic spelling, 452
Phonetic unit, 126, 132
Phonic consonants, 442
Phonogram, 37
Phonograph, 32-51 ; 607
Photography of manometric flames, 27
Photography of the larynx, 249
Physical definition of a vowel, 400 ; of
the consonants, 442
Physical intensity of sound, 109
Pictures, use of, 187
Pillars, 232
Pillsbury, 128, 129
Pilzecker, 113
Pipping, 409, 477, 491, 500 ; analysis,
75 ; Finnish vowels, 21 ; on nature of
vowels, 23 ; on r, 24 ; on sonant h,
24 ; phonautograph, 20 ; Swedish
vowels, 21
Pitch, 64 ; irregularities of, 202 ; nature
of, 89-94 ; of speech sounds as per-
ceived by ear, 473 ; range of, 97
Pitch rhythm, 517
plaisir, 465
Planck, 105
INDEX
623
plenum, 461
Pneumograph, 214
Poltern, 387
Pomeranian dialect, 362
Pons, 86, 192
populum, 460
Posterior pillars, 232
Postponement, 165
Poulsen, 51
Practice, 205
Practice of Fourier analysis, 566
Precision of movement, 386
Preece and Stroh, 18, 69
Pressure of treath, see Breathing
Preyer, 105
Principle of substitution, 549
Pringsheiiu, 476
Probable error, 149, 155, 166, 602
Probableness, 201
Production of speech, 188-398
Progressive change, 462
Prominence of a sound, 460
Propagation of vibration, 4
Prose, 551
Pseudobeats, 19
Psychophysic law ; for pitch, 93
Pterygoidal muscles, 230
Puffs, 90, 94, 96
Pulls, 200
Pure tone, 89
Pythagoras, 93
Quadratus muscle, 231
Quality rhythm, 517
Question, 218
Quickness of movement, 387
Quickness of response, 386
1', 19, 24, 28, 44, 114, 117, 131, 203, 204,
304, 307, 308, 310, 318, 321, 328, 329,
336, 344, 432, 461, 465, 597
J, 432, 459, 465
r mouille, 441
Range of voice, 272
Rate of breath expenditure, 221
Rational rhythm, 552
Rayleigh, 414, 420
Reaction time, 206
Reading, 132
Receiving tambour, 196
Recognition, of speech sounds, 113; of
tones, 1 06 ; of words and letters, 128;
of v/ords and objects, 182
Recording drum, 198
Recording tambour, 196
Rectification of tambour records, 197
Reed, 257 ; for lowest tone, 93
Reflex activity, 191
Reflex centers, 191
Reflex-tonus, 382
Register, 252, 272
Registration of movement, 195
Regulated rhythm in action, 526
Regulation of movement, 202 ; see Reg-
ulative sensation
Regulation of muscular movement, 191
Regulative sensations, 191, 325
Reissner, 80
Repetition, 178
Replacement of sounds, 460
Resistance, 10
Resonance, 13, 73
Resonator, 14, 73
Re'thi, 262
Rhythm, 179; auditory, 517; experi-
ments on, 508 ; laws of, 518 ; of breath-
ing, 213
Rhythmic action, free, 602
Rhythmic grouping, 533
Ribs, 212
Right and wrong cases, 104
RigoUot, 24
Rime, 172
Rip Van Winkle's Toast, 61, 479
Risorius muscle, 232
Rontgen, 237, 337, 341
Rosapelly, 226. 267, 358
Roudet, 221, 226, 572
Rounded, 427
Rousselot, 114, 115, 116, 217, 219, 223,
267, 274, 304, 312, 316, 335, 347, 354,
3.57, 362, 365, 374, 396, 422, 429, 433,
440, 441, 456, 458, 459, 483, 491, 502
Roy manometer, 225
Russian mouiM labials, 441
Russian vowels, 25
s, 46, 76, 89, 114, 117, 119, 120, 223,
224, 304, 305, 310, 316, 321, 328, 329,
394, 597
s — > (T — » (T — )^ I, 465
(T, 436
624
INDEX
s, 46, 114, 119, 223, 224, 304, 305,
307, 308, 310, 316, 317, 321, 328,
333, 344, 395
«r — >-c— > Xj "^^^
s, 44l'
Sacculi, 79
Sagittal diagram, 296
Samojloff, 25, 29
Santorini, 240
Sapphic, 537
Saubevschwarz, 423
Scanuiiif,' speech, 192
Schedule for Fournier's analysis,
569
Schiller, 183
Schischmanow, 105
Schmidt-Wartenberg, 513
Schneebeli, 18, 75
Schuh, 432
Schiilze, 104
Schumann, 175
Schwann, 476
Sciatic nerve, 190
Scott's phonautograph, 17
Second, 104
Seebeck, 94, 97
Seelmann, 279
Self-help, 58, 576
Semicircular canals, 79
Semi-occlusives, 441
Senn, 247
Sensation, 84
Sensiti'i'eness to difference, 113;
also Just perceptible difference
Sensorj"" motor control, 387
Sensory nerves, 191
Septum, median, 234
Seventh, 104
Sharpe, 111
Short, 501
Shortest audible tone, 106
Shortest audible vowel, 107
Sievers, 113, 434, 450, 454, 507, 552
Sigh, 215
Similarity, 135, 148
Simple reactions, 207
Simple tones, 95
Simultaneous action, 543
Simultaneous movements, 357-378
Sinusoid, 2
Sinusoidal vibration, 2
Siren, 89
306,
329,
567,
Sixth, 104
sliofan, 464
sliupan, 464
Slurring, 204, 467
Smith, 179, 528
Smoked drum, see Drum
Smoothing of articulation, 454
Sniff, 215
Sob, 215
Soft palate, see Velum
Sonant, 223, 226, 227, 251
Sonant h, 24, 276
Sonation, 360, 365
Song, 215
Sonnet d'Arvers, 545
Sonority, see Auditory
Sound, 89
Sound compensation, 461
Sound fusion, 446
Southern British English, 103
Spark coil, 12
Speaking, 215
Special association, 163
Specific sounds, 454
Speech, basis of, 113 ; centers in cortex,
83; curves of, 1-75; ideas, 126-134;
perception of, 76-88 ; production of,
188-398
Speech curves, studies of, 575-601
Speech elements, perception of, 113-125
Speech rhythm, 537-557
Speed, 467, 468
Spelling, 182
Spelling pronunciations, 168, 170
Spinal bulb, 192
Spinal cord, 86, 192
Spirant, see Fricative
Spirometer, 221
Spontaneous, see Free rise
Spread, 427
Squire, 508
Stammering, 118
Standard, 7
Stapes, 78
Stapedius, 79
Steffens, 180
Stereoscopic photography of larynx, 249
Stern, 101
Sternothyroid muscle, 242
Stirrup, 78
Stod, 279
Stopwatch, 152, 160
INDEX
625
Stork, 261
Storm, 423
.Stosston, 514
Strachey, 573
Straw-bass register, 272
Stress, 502
Striug, 72, 255
Stroboscope, 249
Stroh, 18, 69
Strong, 502
Structure of larynx, 279
Structure of vocal organs, differences,
and their effect on progressive change,
463
Stumpf, 105, 106
Stuttering, 158
Styloglossus muscle, 234
Stylopharyngeal muscle, 242
Substitution, principle of, 549
Successive movements, 317-378
Suggestion, 115
Superior longitudinal muscle, 238
Surd, 223, 226, 227, 251
Surd and sonant, 304, 317, 367
Surdation, 360, 365
Surplus air, 227
Sustained pitch, 485
Swain, 261
Swedish vowels, 24
Sweet, 427, 467
*sweks, 465
Syllables, 449; learning, 113, 178; nor-
mal, 175
Syntax, 163
Synthesis, 67-70
Synthesis of vowels, 290
t, 45, 114, 119, 120, 123, 127, 131, 224,
226, 285, 302, 305, 306, 307, 308, 310,
316, 317, 321, 326, 333, 336, 344, 437
t ->• 9, 464
t", 120
ri 439
ts, 120, 123, 306, 307, 321, 439
ts, 304, 306, 307, 439
t9, 439
T, 437
T, 304, see / mouille
T, 306, 307, 316, 437, 438, 440
8 117, 304, 329
t mouille', 304, 305
Talking machine, use of, 29 ; see also
Gramophone, Thonautograph, Phono-
graph
Tambour, 195, 219
tad innam, 458
tat, 458
Teaching, see Learning
Techmer, 308
Telephone, 207
Temporal lobe, 192
Temporal muscle, 230
Tense and lax, 384, 427
Tensor of the velum, 233
Tensor tympani, 78
Tests for formation of associations, 176
Tetanic contraction, 189
Tetanus, 189
than, 117
Theodor, 117
Thesis, 537
Third, 104
Thompson, 573
Thorax, 212
Thought, see Current
Thought, defined, 126
three, 464
Thudichum, 352
Thumb, 160, 170
Thyroarytenoid muscle, 240
Thyroepiglottic muscle, 242
Thyrohyoid muscle, 242
Thyroid cartilage, 239
Thyroid prominence, 239
Timbre, 89, 96, 97
Time, see Reaction time
Time estimates, 501
Time marker, see Marker
Time of association, 152
Tone, 268 ; defined, 89 ; difference, 99 ;
faintest audible, 109 ; highest, 98; just
perceptible change, 101 ; just percep-
tible difference, 100; lowest, 93, 98;
shortest audible, 106 ; simultaneous
tones, 105, 106
Tones, of the vocal cavities, 281
Tongue, guidance of, 325; position
and movement of, 325-327 ; sensation
from, 325
Tongue contacts, 296-324
Tongue tambour, 335
Tonsil, 332
Tonus, 382
40
626
INDEX
Trachea, 212, 229
Transmission, see Air transmission
Transmission of sounds, 469
Transverse fibers, 193
Transverse lingualis muscle, 236
Trautmann, 289
Trautscholcl, 156
Tremolo, 108
tres, 464
Triangularis muscle, 231
Trigeminus nerve, 194
Trill, 108
Trochaic, see Trochee
Trochee, 179, 535, 553, 556
trolley, 460
Tuning fork, see Fork
Turbinal bodies, 339
Tympanum, 27, 99
Typical sounds, 454
u, 26, 39, 43, 103, 114, 11.5, 222, 227,
266, 287, 305, 306, 307, 315, 320, 321,
329, 332, 333, 336, 339, 342, 343, 344,
590, 601
u, consonant, see w
ulmum, 461
Uncertainty, 201
Unconscious modification, 115
Unconscious movements, 206
Unconscious whispering, 132, 456
Unfamiliar habits, 468
Unintentional movements, 206
Unison, 104
Unit, see Phonetic unit
Unnoticed association, 147
Unnoticed variations, 123
Unvoiced, see Surd
Urbantschitsch, 108
Utricle, 79
Uvula, 232, 233
Uvular r, 461
V, 47, 119, 223, 224, 225, 302, 307, 317,
333, 342, 360, 376
Vagus nerve, 194
Variation, 102, 201
Velar vowels, 427
Velum, 229, 232, 338-352
Ventricle, 243
Ventricular band, 243, 265
Ventriloquism, 355
Ventriloquistic speech, 227
Verdin, 221 ; pneumograph, 214
Verse, 551
Vertical lingualis muscle, 236
Vibrating springs, 7
Vibration model, 6
Vibratory movement, 1
Vietor, 218, 308, 309, 377, 428, 444, 478,
497
vif, 458
Visual learning, 181-186
Visual memory, 181-186
Visual words, 183
Vitality, 467
vive, 458
Vividness of impression, 185
Vocal band, 243, 251-280
Vocal cavities, tones of, 287
Vocal control, 379-398
Vocal cord, see Vocal band
Vocal harmony, 204, 272
Vocal muscle, 240
Vocal organs, 229-238
Vocal process, 240
Vocal reaction, 208
Vocal tambour, 219
Voice key, 154
Voice tone, see Cord tone
Voiced, see Sonant
Voices, soft and sharp, 265
Voigt, 99
Volition, 205, 208
Volume of air expended, 219
Voluntary action, 188
Voluntary centers, 192
Voluntary. contraction, 190
Von Lieben, 51
Vowels, auditory nature, 422 ; depen-
dence on speed of reproduction, 422 ;
desouation of final, 203 ; diphthongi-
zatiou of, 103 ; expenditure of breatli
during, 223; harmony of, 121, 204,
272; long, 20 ; motor nature of, 425 ,
natui-e of, 19, 23, 39, 94; physical
nature of, 391 ; pitch of, see Melody ;
relations of loudness of, 504 ; relaxa-
tion of, 224
Vowels, Finnish, 24 ; French, 26 ; Ger-
man, 39-43 ; Russian, 25 ; Swedish, 24
w, 131, 318, 329, 355, 369, 433
Wageu, 465
INDEX
627
Wagner, 494
Warning, 218
was, 458
Watch, see Stopwatch
Wavelength, 5
Weak, 502
Weakest audible tone, 109; audible
speech sounds, 1 14
Weeks, 226, 345
Wehnelt, 237
Wendeler, 337, 595 ; phonautograph re-
cords, 18
Wernicke, 83
Wheatstone, 407
Wheeler, 158
Whitney, 169
Whisper, 268 ; mechanism of, 274 ; pres-
sure in, 226 ; weakest audible sounds,
114
Whistle, 98
Wide Towels, 427
Wieu, 111
Willis, 290, 399, 405, 416
Wiltse, 116
Wifasek, 105
Wolf, 114
Wolmar, 514
Word, 126
Word agraphia, 85
Word blindness (alexia), 86
Word deafness, 85
Word dumbness, 84
Word movements, see Motor words
Written words, 83
Wundt, 141, 468
y, 39, 43, 114, 115, 224, 287, 306, 307,
309, 314, 320
^, 320, 433
z, 47, 114, 117, 223, 224, 305, 316, 333,
344, 360, 376
z, 47, 119, 223, 224, 304, 305, 316, 317,
321, 333, 344, 360
z, 441
Zimmermann, 65, 567
Zonophone, 607
Ziind-Burguet, 394, 396
Zygomatic muscle, 232
(First seven liues, 1mm = O.OOlCi ; last Bve lines, Imm = 0.0007».) PLATE I. Cui'
Ul'ves
irresfm"' Cock Rohin, Series II.
(1 Cock Rohin, Series II.
MAA/n/W
ArVV^
WVWIMAAAAA/VWWWV^
what do you
■'v^/V
/vv
AAAA/VVVVV
vv/v
v-^\A.-y^A-V\AA/\A^
^^ /VV\/VV\/V^A /'V'VYAA/V
(lramzzO.OO(n».)
PLATE III.
Rip Van Winkle's
(BlocV
(Imm = 0.0007b.)
PLATE IV. Rip Van Wi
(I
s Toast, by Joseph Jefferson.
II.)
(Imm = 0.0007«.;
PLATE V. Hip Van Winkle:
(Hlocl
Toant, by Joseph Jefferson.
III.)
(Iram = 0.0007«.)
PLATE VI. Rip Van Win.
(Bl
e's Toast, by Joseph Jefferson.
k IV.)
(Imm = 0.0007'.)
PLATE Vir.
Rip Va
^^--y^^/^ ' -^ \f ^ y ^\r^'-^^~~^\/^
fine schnapps.
-T-y r^ -/"--',
-yU^^^- ^^^^^y ^■.._J^- .
That's
I wouldn't keep it as longf as that.
^^^-^-''^_^■
(Iram = 0.0007».)
PLATE VIII. Rip y
e's Toast, by Joseph Jeffekson.
i VI.)
'•'^\/-^\/^-^\/^y\r^,/
r^~- r-^, r^-J-^\
--j-^\r-JAf--J'^V-^^\f-^ y--./— ^^v/
~^, r\ — -^/^--
xAAAAMAM/^^WWVWVVAAAA/
,/ V V V ^ -^
— ^ — ~_^.,.~ '
would I?
Hulit huh*
^^^A^nA-n/
/w^M^V\A
aV\/v^A/Ma-AA''AA'Aa-
^/^--Aa-^IaAa^v/^A
(Imm = O.OOOT'.)
PLATE IX. Rip Van Winkle^
(Block
(lmm = O.0OOTs.)
PLATE X. Rip Van
s toast, by Joseph Jefferson.
. VIII.)
(1mm =: 0.0007b.)
PLATE XI.
Rip Van WM
(Bl(
, Yan Wife's Toast, by Joseph Jefferson.
(Block IX.)
COME RIP
340'
I7S
ISO-
I2S
100
I7S
ISO
125
too
SAY TO A GLASS?
200
toe
t7S
_^-o^ '"
/\ A
ISO
^ —
V"^ "0
"• ■^^-'\^j^-^ — V\ — ^
125
^
12$
^ ~^~\
100
— ^- ^^ ,_^^
100-
N^.^
75
-
'""^^
2S
0
WHAT DO YOU
298'"«
2007
■ 7t'
ISO-
OS
100'
71-
so-
il'
795'
175
lU
tis
100
75
so
25
- 0
3287'
WHAT DO I
ISO'
175
ISO
I2S-
100'
75
50
25
- 0
SAY
108'
175
ISO
I2S
100
75
SO
■ 0
TO A CLASSr
585'
171-
ISO-
fis-
100'
7S-
10-
IS-
HOW WHAT DO I GEHERALLY
128'
t7f-
M4-
ni
aoo-
in-
ISO'
IK-
100-
»!■
SO-
tS'
SAY TO A GLASS!
■ so-
us
100
75
0
140-
ISO'
IIS'
100'
7S'
SO'
25'
- 0'
SAY IT ISA
175-
150
125
100
75
SO-
SO"" IS
0
s_,r^>_A^^-^ '"
y^-^-^~^-y\^.^^ "*■
^V^iso
y ^"^^''"^-\_ "*■
125
^ ^.^^_,^ 125-
100-
^-\ 100'
75-
FIHE THIHG
so
25
25
WHEH THERE*S PLEHTY IH IT
ISO
125
100
PLATE XII. Curres of Pitch. (Ch. XXXII.)
,vA/~A
TEN YEARS AGO EH?
200
175
150
125
100-
200
1>5
150
125-
100
7$
3495-'' "
0
^'^ -^
~^-^
150
•S^ 125
100
FINE SCHNAPPS
2749MM
I WOULDN^T KEEP IT AS LONG AS THAT
150-
I2S'
100-
75-
50-
25-
0
PLATE XIII. Cin-r.s of Pitch. (Ch. XXXII.)
250
200
ISO-
100-
50
250
200
ISO
100-
50
250-
200-
ISO
100-
50-
Z50-
ZOO
ISO
100
50
r
zoo 400 600 800 WOO
b zoo 4bo 600 800 lo'oo
zoo 400 6d0 800 lo'oo
b zoo 4do 600 800
1000
Z50
ZOO-
ISO
100
50
zoo 400 600 800 1000 1200
Z50
ZOO-
ISO
100
50-
200 400 600 800 1000
Z50
zoo-
150-
100
so
250-
200-
150-
100-
50-
Z5lf-
zoo-
150-
100-
50-
0
Z50-
200-
150-
100
50
ZOO 400 600 800 WOO
zoo 400 600 800 WOO
ZOO 400 600 800 WOn
r
250
200
ISO
100-
50-
ZOO 400 600 800 WOO 1200 1400
ZOO 400 600 800 WOO I20O 1400 1600
Z50
ZOO
150-
100
SO
ZOO 400 600 800 1000 IZOO 1400 1600
250-
200-
150-
100-
50
ZOO 400 600 SCO 1000 0 200 400 600 800 WOO 1200 1400 1600
TLATK XTV. Ciims of l^ilrh and Ihiralion it, Uhijllimic Sounds. (Ch. \X\V.)
2JSD-
ZOO
ISO
100-
50-
ZSO-
ZOO-
ISO-
100
50
ZSO
zoo
ISO
100-
50-
250
200-
ISO-
100
SO
r
ZOO 400 600 800 1000 1200 1400 1600 1600
a
a'
•
/
6 zoo
400
600 800
lo'oo
1200
1400
1600
ib'oo
ZOO 4io 6d0 800 woo 1200 1400 1600 IBOO
zoo 400 600 800 WOO 1200 1400 1600 1600
/ (I, said the sparrow)
/ (I saw him die)
a 1
'lie (Who saw him die ?)
/ (I killed CocH
a 1
/ (may, I caii
die (I saw hii
thy (hallowed he Thy name)
PLATE XV. Curves of\
Cordi
? Reso
Robin)
I (I, said the fly)
eye (with my little eye)
die)
a 1
fly (I, said the fly)
a 1
thy (Thy kingdom come)
itch. (Appendix II.)
ne.
ance tones.
zoo Joo too
I (I, said the sparrow)
7 (I, said the fly)
aoo 100 too
I (may, I can say)
die (Who saw him die ■?)
fly (I, said the fly)
'00 ido joo
I (I killed Cock Robin)
'Of lio jio
I (I saw him die)
/io i.!>o JOO
eye ( with my little eye)
— r
roa
iha
100 Joo
die (I saw him die)
roa
thy (hallowed be thy name)
Tiy kingdom come)
,vA/~A
TEN YEARS AGO EH?
200
175
150
125
100-
200
1>5
150
125-
100
7$
3495-'' "
0
^'^ -^
~^-^
150
•S^ 125
100
FINE SCHNAPPS
2749MM
I WOULDN^T KEEP IT AS LONG AS THAT
150-
I2S'
100-
75-
50-
25-
0
PLATE XIII. Cin-r.s of Pitch. (Ch. XXXII.)
250
200
ISO-
100-
50
250
200
ISO
100-
50
250-
200-
ISO
100-
50-
Z50-
ZOO
ISO
100
50
r
zoo 400 600 800 WOO
b zoo 4bo 600 800 lo'oo
zoo 400 6d0 800 lo'oo
b zoo 4do 600 800
1000
Z50
ZOO-
ISO
100
50
zoo 400 600 800 1000 1200
Z50
ZOO-
ISO
100
50-
200 400 600 800 1000
Z50
zoo-
150-
100
so
250-
200-
150-
100-
50-
Z5lf-
zoo-
150-
100-
50-
0
Z50-
200-
150-
100
50
ZOO 400 600 800 WOO
zoo 400 600 800 WOO
ZOO 400 600 800 WOn
r
250
200
ISO
100-
50-
ZOO 400 600 800 WOO 1200 1400
ZOO 400 600 800 WOO I20O 1400 1600
Z50
ZOO
150-
100
SO
ZOO 400 600 800 1000 IZOO 1400 1600
250-
200-
150-
100-
50
ZOO 400 600 SCO 1000 0 200 400 600 800 WOO 1200 1400 1600
TLATK XTV. Ciims of l^ilrh and Ihiralion it, Uhijllimic Sounds. (Ch. \X\V.)
2JSD-
ZOO
ISO
100-
50-
ZSO-
ZOO-
ISO-
100
50
ZSO
zoo
ISO
100-
50-
250
200-
ISO-
100
SO
r
ZOO 400 600 800 1000 1200 1400 1600 1600
a
a'
•
/
6 zoo
400
600 800
lo'oo
1200
1400
1600
ib'oo
ZOO 4io 6d0 800 woo 1200 1400 1600 IBOO
zoo 400 600 800 WOO 1200 1400 1600 1600
/ 1 1, said the sparrow)
a :
/ (I killed C(
/ (I saw him die)
/ (may, I
liic (Who saw him die ?)
die (I saw
thy (hallowed be Thy name)
PLATE XV. Curves
Co
::V
V
[IHslkRobin),
a 1
I (I, said the fly)
a 1
eye (with my little eye)
jllmlm die)
fly (I, said the fly)
thy (Thy kingdom come)
^J Pitch. (Appendix II.)
,((' tone.
' Ijonance tones.
10
t
i-
*
i
I (I, said the sparrow)
I (I, said the fly)
zoo 3 00 ^OO
I (may, I can say)
100 zoo iopj too roo
die (Who saw him die ?)
2-
r
fly (I, said the fly)
/(I killed Cock Robin)
100 iho jio
I (I saw him die)
/io lio JBO
eye (with my little eye)
200 JOO
die (I saw him die)
roo
thy (hallowed be thy name)
; ; ; m ja .
i : ich in schi
s : i in id
PLATE XVII.
u:um mut .,
u ; u in mutter..
o : o in sohn.
o : 0 m Sonne.
PLATE XVin.
PLATE XIX.
PLATE XX.
y : ic in hiite
oe : o ih hohle .
PLATE XXI.
y : M in hiite
oe ; a in hohle .
oe : £> in
PtATE XXII.
» ; i- in rid.
"vf-.tu in wet
] :_y in yet.
^: sk in shut.
PLATE XXIII.
u ; latter part o£ win pool.
o . latter part cf o mpok
PLATE XXIV.
PLATE XXV.
PLATE XXVI.