UC-NRI

B 3 BDD flEl ^m

Tflntversiti? of Chicago

The Influence of the Time Interval

upon the Rate of Learning

in .the White Rat

A DISSERTATION

SU! fITTED TO THE FACULTY
OF THE

GRADUATE SCHOOL OF ARTS AND LITERATURE

IN CANDIDACY FOR THE

DEGREE OF DOCTOR OF PHILOSOPHY

(DEPARTMENT OF PSYCHOLOGY)

BY

JOSEPH USSERY YARBROUGH, PH.D.

A Private Edition Distributed by
The University of Chicago Libraries

A Trade Edition is Published as No. 135 of the Psychological Mono-
graphs by Psychological Review Co., Princeton, N. J.

EXCHANGE

BIOLOGY
LIBRARY

G

Gbe Tflntt>er6it£ of Chicago

The Influence of the Time Interval

Upon the Rate of Learning

in the White Rat

A DISSERTATION

SUBMITTED TO THE FACULTY

OF THE
GRADUATE SCHOOL OF ARTS AND LITERATURE

IN CANDIDACY FOR THE

DEGREE OF DOCTOR OF PHILOSOPHY

(DEPARTMENT OF PSYCHOLOGY)

BY
JOSEPH USSERY YARBROUGH, PH.D.

A Private Edition Distributed by
The University of Chicago Libraries

A Trade Edition is Published as No. 135 of the Psychological Mono-
graphs by Psychological Review Co., Princeton, N. J.

y

BIOLOGY
LIBRARY

G

ACKNOWLEDGMENTS

The experiments herein reported were carried out under the
direction of Dr. Harvey A. Carr of The University of Chicago
who has been unsparing in suggestion, supervision, and criticism.
To him I wish to express my sincerest gratitude for his un-
wearying kindness and interest.

My thanks are due Professor James R. Angell, Dr. W. S.
Hunter, and Dr. Clarence S. Yoakum for their thorough and
stimulating instruction. My initial interest in Animal Psychol-
ogy is due to my contact with the enthusiasm and scientific
ability of Dr. Hunter.

For helpful suggestions I am indebted in particular to the
work of Wylie; in general, to those other investigators whose
work is cited in this report.

Acknowledgment is also due Mr. J. Hall Sheppard, a graduate
student in the University of Texas, for assistance in the early
training of some of the animals, and to Mrs. Yarbrough for
assistance in preparing the manuscript.

Jos. U. YARBROUGH.

University of Chicago,
Feb. 25, 1920.

465004

TABLE OF CONTENTS

PAGE

I. Introduction i

II. Historical 3

III. Apparatus and Method 13

IV. Experimental Results 19

1. The Influence of the Time Interval 19

A. Learning the association in the forward direc-

tion 19

a) Association of sound with pain 19

b) Association of light with sound 24

B. Learning the association in the backward di-

rection 30

2. Simultaneous vs. Successive Presentation 35

3. Learning the Negative Response to Pain 39

V. Interpretation of Results 44

VI. Summary and Conclusion 50

I.

INTRODUCTION

The experiments reported in this monograph1 are concerned
with three closely related problems in the field of animal psychol-
ogy: i) The influence of the time interval upon the rate of learn-
ing. The animals were divided into groups, and the terms X
and Y were presented in a temporal relation to each group. The
comparative number of trials necessary to establish a given
strength of functional connection constitutes the measure of the
relative efficacy of the several modes of presentation. 2) The
readiness with which a given temporal association will function
backward, and the most effective time interval for this function-
ing. The terms X and Y were here presented in the order Y-X,
but in the test experiments they were required to function in the
normal order X-Y. 3) The claims of the rival thories of simul-
taneous and successive presentation in relation to the rate of
learning. The animals were required to associate two factors,
X and Y. To one group the terms were presented simul-
taneously; to the others, successively.

The nature of the present experiments may be further set
forth by additional explanation of the problems under considera-
tion. The various statements made in the modern texts on gen-
eral psychology indicate that simultaneous or successive presen-
tation is regarded as an essential condition for the formation of
an effective association. And by successive is meant immediate
succession; i.e., the two experiences must be in temporal contact
with each other. Our aim is to separate these terms by a certain
interval and observe its influence upon the rate of learning the
association when the terms are presented in the forward and
in the reverse order. Those who hold to the theory of simul-

1 This research was begun in the Psychology Laboratory of The Uni-
versity of Texas in October, 1918, and continued until the following June,
at which time the work was transferred to the Psychology Laboratory of
The University of Chicago.

2 JOSEPH U. YARB ROUGH

taneous association would have us believe that association can-
not be formed unless the facts or objects associated be ex-
perienced simultaneously. They recognize the fact that in every-
day life things are associated which succeed each other in time,
but they account for this on the assumption that "with the ces-
sation of the actual mental experience the nervous excitation is
not abruptly finished but continues, gradually dying away."2
The second factor of the association, then, coincides in time with
this "gradually-dying-away" nervous excitation of the first, and
it is due to this simultaneity that the association of the two fac-
tors is formed. The rival theory of successive association, based
upon the hypothesis that the range of consciousness is not wide
enough to attend to two things at once, denies the possibility
of association between two factors simultaneously presented.
Those who hold this theory say that although the two factors are
apparently presented simultaneously they are in reality successive
experiences, since the attention is constantly vacillating between
the two. If successive association is correctly explained by the
simultaneity of the succeeding experience with the after-phase of
the preceding one, the data here presented show to a certain ex-
tent the duration of this "akoluthic" phase and its rate of dimi-
nution.

Although our prime interest is in the first of the problems
listed above, it was found both practical and profitable to gather
data on the second and third. Very little work on the main
problem has been reported either in the field of animal or of
human psychology, hence it is on this point that these experi-
ments will probably have their greatest significance. The second,
likewise, has received scanty attention as yet. The third problem,
on the other hand, has been the subject of considerable experi-
mentation, although but little of this investigation has been in
the field of animal psychology. It is our purpose to present data
upon these timely problems, with especial emphasis upon the
question of the influence of the time interval in successive asso-
ciation upon the rate of learning in both its backward and its
forward directions.

2Semon, Richard, "Die Mnene," 2te Auflage, Leipzig, 1908.

II.

HISTORICAL SECTION

The purposes of the present historical section are: i) To bring
together for critical consideration all the data reported up to the
present time on the rate of learning in successive association;
2) To present a brief critical review of the experimental litera-
ture upon the subject of simultaneous and successive presentation
in association; 3) To examine all data now published on the
strength of function of association in the backward as com-
pared to the forward direction; 4) To show that the evidence
is inconclusive because of its scantiness and, in some cases, its
general character. Due to the fact that in some of the experi-
ments reviewed more than one of the above-named issues is
discussed, it was deemed expedient to present the full report in
each case rather than to attempt presentation by topics. Crit-
icism and discussion have in some cases been deferred to the
Theoretical Section.

Bigham3 (1894) was among the first to investigate these
questions. For subjects he used five college students of an
average age of 24; for material, series of varied numbers and
colors (separate or in mixed series) mounted on cardboard 3J^
cm. in size. Five simultaneous series were used and five suc-
cessive. In every respect the series were alike save in the manner
of presentation. The length of the series consisted of either 10 or
20 presentations. A 40 second period was allowed for each
series, 20 simultaneous presentations being made in this period,
and, in the successive series, two seconds being allowed for each
of the 20 cards. By this method the exposure time was two
seconds, whether the cards were presented simultaneously or in
succession. The subjects were asked to use no memory aids in
the learning. After a given series had had its due presentation
the subject immediately arranged his colors or numbers in the

3 Bigham, ]., Psychol. Rev., 1894, Vol I, pp. 34-38.

4 JOSEPH U. Y ARE ROUGH

order of the presented series as it was recollected. He saw the
series but once. Obviously, only the visible series gave a basis
for comparison, and in these Bigham found that for the simul-
taneous series the percentage of error for the five subjects ran :
13.6, 9.8, 15.3, 17.9, and 15. In the successive series the cor-
responding percentages of error were 18.5, 23.3, 16.6, 24.1, and
21.6. The averages for the two series were 14.3 percent and
20.8 percent for the simultaneous and the successive, respectively.
Hence, he concludes: "With each observer the memory was
stronger for the simultaneous than for the successive presenta-
tions."

Bergstrom4 (1907) performed a series of experiments pat-
terned after the work of Ebbinghaus. Three subjects were each
required to memorize 12 lists of nonsense syllables, 12 syllables
to the list, for each of four days. Each syllable was exposed to
view for a period of 820-, and a given time interval was allowed
to elapse between the successive exposures of the syllables. The
syllables were to be pronounced but once at the time of exposure,
and after four repetitions of a list a recall test was given. In
this test a written reproduction was made and the amount memo-
rized was gauged by the number of errors made. The intervals
between the exposures were 3O2er, 6860-, and 1454^, and the
average number of errors was, respectively, 10.3, 8.9, and 7.5.
By increasing the interval between exposures the number of er-
rors was decreased. This fact is at variance with the assump-
tion that a nervous impression begins to fade soon after it is
received and gradually dies away, and must be explained by the
greater strength either from the clearer original impressions or
perhaps by other factors which are favored by a longer time
interval.

Again, 30 subjects were used by this same experimenter, and
lists of 10 words and lists of 10 letters, presented orally at a
metronome-controlled rate, were used as material. The time
intervals between presentations here were .5, i, and 2 seconds.
One presentation was made for each list, and the errors made in

4 Bergstrom, J. A., Amer. Journ. PsychoL, Vol. XVIII, pp. 206-238.

RATE OF LEARNING IN THE WHITE RAT 5

immediate written reproduction again tested the amount memo-
rized. For the three intervals the average percentages of error
were, respectively, 51.12, 36.52, and 23.9 for the word lists, and
44.09, 52.65, and 38.44 for the letter lists. In both, the amount
learned increased with the lengthening of the interval. Berg-
strom concludes : "The effect of increasing the interval is greatly
to decrease the number of errors. . . . Indications are that an
interval longer than any employed would have still further re-
duced the errors."

Wohlgemuth5 (1914) reports an experiment the purpose of
which, as he states it, is i) "to decide by experiment, if possible,
between the claims of the rival theories of Simultaneous and
Successive Association," and 2) "to investigate the influence of
the closeness of connection between the members presented simul-
taneously." He used eight subjects, whose age and qualifica-
tions were undefined. To them he gave a series of tests con-
sisting of pairs composed of a colored field (cardboard) fol-
lowed by a black shape mounted on a white ground, or vice-
versa. The exposure was timed by a Miiller memory apparatus,
a certain definite period of exposure thus being given each card.
Obviously, under such conditions the simultaneous pair would
have just the same period of exposure as each member of the
successive pair. He found the successive mode of presentation
superior to the simultaneous, but accounted for the difference
on the basis of the unequal exposure time for each pair of stimuli.
To overcome this defect two groups were formed: in the first,
only half the number of exposures was given in the successive
series, while each member of the pair was exposed for the same
length of time as the pair of simultaneous stimuli — i.e., the rate
of the memory apparatus remained the same; in the second, the
same number of exposures was given in both series and the same
time given each pair, but only half time was given each member
of the successive pairs — i.e., the apparatus was set to work twice
as rapidly in the successive series as in the simultaneous. Wohl-
gemuth finds that in the first group the average scores were 35.5

5 Wohlgemuth, A, "Simultaneous and Successive Association," British
Journ. of Psychol., VII, 1914-15, pp. 434-452.

6 JOSEPH U. Y ARE ROUGH

for the simultaneous and 33.2 for the successive; in the second
group, 36.8 and 34.4, respectively. Hence, he concludes: i)
"The simultaneous experience is more favorable for the learn-
ing of pairs than the successive experience." 2) "All associations
are due to simultaneity or simultaneity of the succeeding ex-
perience with the akoluthic phase of the preceding experience."
On the whole, Wohlgemuth's article is far from convincing.
It seems that his evidence is insufficient to warrant such positive
statements, and one is inclined to agree with Froeberg6 in his
criticism of the work. Froeberg says that Wohlgemuth's method
of presentation must be rejected, but that even if this might be
admitted "no safe conclusion could be drawn from the results
because the differences in favor of simultaneity are in every case
less than the P. E., and in five cases out of sixteen they are
negative." Froeberg calls his own investigation an "attempt
to repeat Wohlgemuth's experiment with the objectionable
features removed." He used psychology students for subjects,
and had them memorize series of five pairs of syllables. Each
term of the pairs was exposed for a period of 1/3 of a second,
and the time between the terms varied from o to 5 seconds. The
interval between the successive pairs was a constant. The recall
came 10 seconds after the last presentation (each series having
been presented twice), and was conducted by presenting the
stimulus syllables in a new order and requiring the subjects to
supply the missing member of the pair. As usual, all memory
devices were barred during the learning. The average percent-
ages of correct response for the group of seven subjects were
61, 49, 45, 48, 51, 49, and 49 for the simultaneous and con-
tinuous presentations and the intervals of i, 2, 3, 4, and 5 sec-
onds, respectively. — a finding which would indicate that the dif-
ference between the simultaneous and successive series is small,
and that in successive presentation the length of the time interval
is not effective. The simultaneous exposure yielded better re-
sults in every case, however, and the average difference of 12
percent in its favor seems significant. Froeberg does not think

6 Froeberg, S., Psychol Rev., 1918, Vol. XXV, pp. 156-163.

RATE OF LEARNING IN THE WHITE RAT 7

that this slight superiority of simultaneous presentation neces-
sarily proves the theory of simultaneity, but would explain it by
a "tendency to articulate and combine the simultaneous syllables
into a single word." From the introspection of his subjects he
found "essential agreement on three points: i) that simul-
taneous exposure taxes the attention more than successive,
2) that there is a persistent tendency toward articulation in the
simultaneous series, and 3) that when in successive association
there was an appreciable interval between the stimuli, the first
would remain in consciousness, though with varying degrees of
intensity, until the second arrived."

In order to eliminate this persisting memory image during
the interval between the two stimuli in successive presentation
the experiment was repeated with a group of four subjects and
with the added condition that the subjects should read numerals
presented to them during the intervals between the terms of the
pairs, a device calculated to prevent memory survival of the
first term during the interval separating it from the second. In
this test the average percentages of right response for the simul-
taneous and continuous presentations and the time intervals from
one to five seconds were 54, 45, 34, 36, 40, 24, and 22 — i. e., the
rate of learning decreased with an increase in the length of the
interval when a new activity was interpolated. This decrease in
the rate of learning is distinct though irregular. From these
data and the statement by each of his subjects that "the reading
of the numerals effectively and completely obliterated the memory
image of the preceding syllable," he concludes that association
may still be formed between two experiences where the first has
already passed out of consciousness when the second one appears.
This is in contrast with the findings of Bergstrom.

Since Wohlgemuth had rejected nonsense syllables as improper
material for the study of association, Froeberg formed another
group of subjects to which he presented colors and letters of the
alphabet. Other details were the same as in the previous ex-
periments. He found under these conditions that the successive
was superior to the simultaneous mode of presentation. This

8 JOSEPH U. Y ARE ROUGH

finding in the data of the first group led him to the conclusion
that when the material used is such that it forms an organic unit
or can be attended as a unit, simultaneous presentation is prefer-
able; when such that this is not possible, successive presentation
is the better.

Chamberlain7 sought to ascertain the effect upon the power of
recall when a number of objects were displayed i) singly and
2) three together. For subjects he used 60 pupils each from the
third, fifth and eighth grades. By means of a circular disc of
wood certain objects were exposed in fixed order for a given
length of time. In the first tests these objects were exposed
three at a time; in the second, one at a time. The two modes of
display were never used with the same subject. One minute was
allowed for the observation of 15 objects. In the first tests
each object was exposed for nine seconds, while in the second
series three seconds were allowed each object. It is obvious,
then, that although one minute was given in both cases for the
observation of the fifteen objects, the exposure time per object
was unequal. The immediate recall was tested in every instance
by having the pupil recall at the close of the experiment all the
objects he could remember. These objects were listed by the
experimenter in the order recalled. The subjects were also pro-
vided with circular sheets of paper and asked to arrange the
objects thereon as remembered from the circular disc. The
author concludes that both recall and arrangement of objects are
stronger when the objects are seen three at a time than when
shown singly, a conclusion scarcely warranted by the insignificant
percentage of difference between the two modes of presentation.
The average number recalled by the three grades when the ob-
jects were presented singly was 9; when presented three at a time,

9-37-

In the human field investigations on the direction of associa-
tion have been made by Ebbinghaus, Muller and Schumann,
Miiller and Pilzecker, and Wohlgemuth. In each investigation

7 Chamberlain, A. H., "A Memory Test with School Children," Psychol.
Rev., 1915, XXII, 71-76.

RATE OF LEARNING IN THE WHITE RAT 9

memory material was used. Ebbinghaus8 employed the relearn-
ing method and found a strong tendency for an association to
function in the backward direction, but its strength was never
equal to that of the forward tendency. Miiller and Schumann9
using practically the same method, confirmed these results.
Later, Miiller and Pilzecker,10 using the method of recall, found
that when pairs of syllables had been learned as trochees and
the first syllable was shown, in 50 percent of the cases the second
syllable was reproduced, while when the second syllable was
shown the first member of the pair was recalled in only 38
percent of the answers. Here a backward tendency is shown,
but it is of less strength than that of the forward direction.

Working in the Psychological Laboratory of the University
of London Wohlgemuth11 experimented rather extensively, vary-
ing both method and material. He had come to doubt the uni-
versal application of the results obtained by the use of nonsense
syllables in the study of memory, and to verify the results he
used colors and diagrams. To make a further check on his find-
ings he used syllables in exactly the same way as he did colors
and diagrams. With five of his subjects the forward tendency
was the stronger, and with two the backward and forward ten-
dencies were about equal in strength. He concludes that the
forward tendency is the stronger when articulation is used in
memorizing, but that the two tendencies are of equal strength
when articulation is prevented. Just in proportion as the motor
element approaches complete elimination do the results indicate
an equal functioning strength in both directions.

In the field of Animal Psychology there is, as we have said
before, very little work reported upon the problem. Carr12 is

8 Ebbinghaus, "Uber das Gedachtniss," Leipzig, 1885.

9Muller and Schumann, Experimentelle Beitrage zur Untersuchung des
Gedachtnisses," Ztsch. f. Psychol. u. Physiol. Sinnesory, 1894, VI.

10 Miiller, C. E., and Pilzecker, A., "Experimentelle Beitrage zur Lehre
von Gedachtniss," Ztsch. f. Psychol., 1900, Erganzungsbd. i.

11 Wohlgemuth, A., "On Memory and the Direction of Association," Brit.
Journ. of Psychol., 1912, Vol. V, p. 447.

12 Carr, H. A., "Length of Time Interval in Successive Association,"
Psychol. Rev., Vol. XXVI, pp. 335-353.

io JOSEPH U. Y ARE ROUGH

the chief contributor. He reports a series of brief but suggestive
experiments. First, he reports three short experiments, all bear-
ing upon the subject of the influence of the time interval. White
rats were used in each of the experiments, and the same problem
box. The problem for the rats was to learn to choose between
two pathways to a food-box in a given temporal sequence, R, L,
R, L, etc. In case of incorrect choice the entrance to the food-
box was barred. Always the day's work began with pathway R.
Between the successive runs of a day's series the rats were
allowed to eat for a definite time interval. In the first experi-
ments three groups of eight rats each were used, and the inter-
vals for the groups were 15, 25, and 35 seconds, respectively.
The results indicate no essential differences between the groups
in the initial stages of the learning, and, indeed, no essential dif-
ferences throughout for the 15 and 35 second intervals. The
group tested on 25 seconds interval manifested a quite rapid rise
from the loooth to the i2OOth trial. This, however, is an in-
conclusive finding, as other factors than the time interval may
have influenced the result. The group was too depleted, unfor-
tunately, to permit of further continuation of the experiment.
In a second experiment three groups, six rats each, were tested
for intervals of 5, io, and 15 seconds, respectively. In this
test the entrances to the pathways R and L were alternately
blocked so that errors were impossible. Test series in which the
pathways were left open were interpolated every fifth day.
Thus the rats were forced to choose between the pathways and
the learning was measured by the number of correct choices
made in these test series. To quote the results : "The group dif-
ference is not very significant, but the results indicate that io
seconds is a more favorable interval than 15 seconds." In brief,
he admits that his findings are inconclusive as regards the in-
fluence of the time interval upon learning, but states that they do
prove that "an association nexus can be established over a con-
siderable interval of time," and that they make possible the ac-
ceptance of the hypothesis of direct connection between the
factors in successive association.

RATE OF LEARNING IN THE WHITE RAT u

Carr and Freeman13 report an experiment upon the problem of
simultaneity vs. succession, and upon backward vs. forward direc-
tion in forming an association. A box with zigzag runways was
used. In preliminary tests rats were taught to run these path-
ways in order to obtain food. Afterwards in 80 per cent of
the trials the door from the chosen pathway into the food-box
was closed, and the rats were forced to retrace and run the op-
posite pathway in order to reach food. At this stage the rats
were divided into three groups for the purpose of testing the
formation of an association between an auditory stimulus, buzzer,
and the act of turning in response to the closed door. With one
group the buzzer was sounded just as the rat was in the act of
turning — simultaneously with the response to the closed door.
With another the sound was given at a definite point, P, ap-
proximately one second before the door was reached. The third
group was given the sound at point P in the return pathway
approximately one second after the rat had turned around from
the closed door. Test series wherein the sound was given at any
point within the pathway were used to test the formation of the
association. It was found that "successive presentation of the
two terms constitutes a much more favorable method for their
association than does simultaneous presentation." The groups
on backward association showed slight evidence of learning when
the experiment closed. These results are extremely suggestive,
but some points of weakness should be pointed out: i) The
time interval between the two terms to be associated was not
constant. Those who are acquainted with the behavior of white
rats would not expect them to traverse the distance from a point,
P, to the closed door at the same speed each time. 2) The
buzzer which was the source of the sound stimulus was in contact
with the maze. The animal may, therefore, have been responding
to a kinaesthetic or cutaneous sensation and not to sound at all.

Watson's experiments also contribute data in the comparative
field bearing upon this problem of backward association. He
tested white rats for their ability to learn a maze backward after

13 Carr, H. A., and Freeman, A. S., "Time Relationships in the Formation
of Association," Psychol. Rev., Vol. XXVI, p. 465.

12 JOSEPH U. Y ARE ROUGH

having learned it in the forward direction. These data are not
considered conclusive, for all who have trained rats know that
considerable retracing is done while mastering a maze.

Summary, i) The majority of the experimental data re-
viewed favors the simultaneous mode of presentation. 2) In
regard to the time interval, Bergstrom is convinced that with the
increase in the length of the unfilled time interval the amount
learned increases. Froeberg is of the opinion that an increase in
the length of the time interval increases the difficulty. 3) All
investigators have found a tendency toward backward association
— a tendency varying in strength, but seldom equal to and never
stronger than the forward tendency.

III.

APPARATUS AND METHOD

The ground plan of the problem box used in this experiment
is shown in Fig. i . The outside dimensions are 89" x 30". The
box is constructed of J4" soft pine, and is 6" deep. All runways

*«,

GROUND PLAN OF APPARATUS
FIG. i

and doors are 4" wide and 5" deep. The doors are made of very
thin metal and are so hung that they may be opened and closed
without appreciable noise. Especially is this true of doors Di
and D2 which are controlled by rods R and Ri. The entire sur-
face of the box is painted black. The food-box F is 16" x 20"
and is the only portion of the box not covered with glass. The
entire floor of the maze in front of the food compartment, with
the exception of the long runway from the center door D, to the
point O, is laid with No. 20 copper wire at J4" intervals. By
pressing a contact key a current from three dry cells is sent
through an induction coil which sends an induced current over
the entire system of wire on the floor of the maze. The strength

I4 JOSEPH U. Y ARE ROUGH

of the current passed over these wires is so controlled that the
animal never becomes frightened by a severe shock.

This box was developed from the one used by Carr and Free-
man1 in their experiments. The following improvements were
made in the construction of the present box : i ) The glass cover
was extended so that the animal is not required to come out in the
open until it enters the food-box. 2) There is but one door to
the food-box. 3) The source of the stimuli is above the glass
cover, making contact with the animal impossible. Further-
more, it was necessary to increase the box in length and number
of compartments to secure sufficient time between the point
of choice and the food-box for the presentation of the two
stimuli with an interval of six seconds between them. The
animals seldom made the distance in so short a time, but in their
most rapid responses the entire distance was traversed. The
additional apparatus used in presenting the pair of stimuli in
certain temporal relations are rather tediously described below.

About 1 8" above the center of the box a light of very low
candle power is suspended. The light from this source is just
sufficient to enable the experimenter to see the animal move
through the maze. Similarly, an electric buzzer is suspended
10" above the glass cover of the middle portion of the box and
connected with the batteries and the timing apparatus as is in-
dicated in Fig. 2. The timing apparatus is, in turn, attached to
a Zimmerman kymograph by screws E and E1 in a way shown
very well in Fig. 3.. The revolving disc upon which the arm A
rests is in direct connection with the shaft of the kymograph,
which turns it around the dial at any speed desired. The dial is
graduated so as to read seconds, half seconds, and quarter sec-
onds, when the kymograph is adjusted at the rate of one revolu-
tion for every 15 seconds. Each of the numerals, therefore,
represents the number of seconds necessary for the arm A to
move from the zero point to the respective numeral as it appears
on the dial. The contact B may be moved along the dial and
firmly set at any point by the thumb screw. The arm A is not

iCarr, H. A., and Freeman, A. S., "Time Relationships of Association,"
Psychol Rev., Vol. XXVI, p. 465.

RATE OF LEARNING IN THE WHITE RAT

TIMING APPARATUS
FIG. 2

fastened to the shaft of the kymograph, but is carried along by
friction on the large revolving disc. If the circuit is closed and
the electro-magnet pulls the lever C into place, the arm A is held
in the position shown in the drawing. The strength of the
magnet is great enough to overcome the friction on the revolving
disc.

By pressing the contact key the circuit of the electro-magnet
is broken, and is closed on the electric buzzer. This presssure
not only gives the sound stimulus, but it also releases lever C,
allowing arm A to be carried along with the revolving disc and
to pass over contact B, which closes the circuit through the in-
duction coil and the system of wires on the floor of the box.
When the contact key is released the current again flows through
the electro-magnet, attracting lever C, which keeps the arm A
from moving when it again reaches that point. Thus the ap-
paratus automatically completes the process, and after 15 seconds
it stands ready for a repetition at the will of the experimenter.
The friction contacts at D and B are so constructed that the
current is broken at the end of one-half second. Each stimulus
is, therefore, given for this period of time. By this apparatus
it is possible very accurately to present two stimuli in a certain
definite temporal relation to each other.

With certain groups of animals light was used as one of the
two terms to be associated. The new apparatus added for pre-

16 JOSEPH U. YARB ROUGH

senting the light may be seen in Fig. 3. Directly above the long
central runway and in contact with the frame of the glass cover
of the box a piece of soft pine i" x 2" - 4' is attached, and to
it is affixed two lamps connected in parallel. The first lamp is 18"
from the end of the box; the second is 22" nearer the food-box.
National Mazda 100 Watt lamps are connected on a 115 Volt
circuit. In order to direct the light upon the zigzag pathways
a reflector is suspended directly over the lights and so adjusted
that the light falls on these runways only. With proper electric
connections the timing apparatus enables us to present sound and
light in a certain known temporal relation, just as the buzzer
and pain were previously presented.

The method of procedure may be described as follows : The
animal to be tested is put through the door D from the food-box
F into the central runway C. He is required to go up to point
O and turn either to the left or to the right and return through
the zigzag pathway X or Y, through the door Di or D2, and
through D3 into the food-box F where he is given food. It is
of little interest in these experiments whether the animal turns
to the one side or the other, or to one side altogether, for the
problem is in no way affected by the alley chosen or by position
habit. The alley chosen in no way determines what is to follow.
Seven times out of ten trials the animal is given an electric
shock in the return zigzag runway; no shock is given in the
three remaining trials. This shock may be given at any point
along the runway, lasts for a period of one-half second, and is
given only once during the trial. The problem here is to learn
to turn around for pain or to go on in case no pain is given.
If pain is given and the animal does not turn around but goes
on toward the food-box he finds the door at the end of the alley
closed. Thus, for example, if the door is D2 he must, to gain
entrance to the food box, run back through the zigzag alley Y,
pass the point O, and approach through the alley X and the door
Di. When no pain is given the door of the alley selected is
always open and the animal enters the food-box direct. The
problem here may, therefore, be thought of as an association

RATE OF LEARNING IN THE WHITE RAT 17

between a closed door and an electric shock, and farther on in
this report it will often be referred to as learning the negative
response to pain.

From the above conditions it will be noted that no stimuli
are given in 30 per cent of each series of trials. This, together
with the fact that chance and position habits are of so little
significance in the solution of the problem, led us to hold a rather
rigid standard as to the number of trials and the percentage of
correct reactions to be required of all animals. No animal was,
therefore, considered to have mastered the problem until be had
made 90 per cent correct responses on the last 100 trials. All
animals used in the experiments reported below were first re-
quired to learn this negative response to pain. When this was
accomplished an electric buzzer was introduced, by means of the
timing apparatus, in a certain definite temporal relation to the
electric shock, pain. The animals were divided into groups and
each group was set to work upon the problem of transferring
this negative response from pain over to buzzer, but each group
worked with a different time interval. To some groups the
stimuli (buzzer and pain) were presented in a forward direction;
to others, in a backward — i.e., some groups received the buzzer
followed by pain, others, pain followed by buzzer. In the test
series the order was always buzzer followed by pain. The prob-
lem in each group is to learn to respond to the sound stimulus
as they have been responding to pain. It is a transfer of re-
sponse from pain to sound over certain given time intervals.
Continuous presentations and one, two, four, and six second
intervals were studied. In the work with the first of the above
groups special test series of 10 trials each were given after each
20 trials to determine to what extent the transfer had been made.
In the remaining groups the time interval was sufficiently large
to determine this without these special tests. The standard for
mastery in each of these groups is the same as before, 90 per
cent on the last 100 trials.

As the animals completed this problem they were regrouped
and set to work on a third problem, but in the regrouping no

i8 JOSEPH U. Y ARE ROUGH

animal was changed from a group where the stimuli were pre-
sented in the backward to a group where they were presented in
the forward direction or vice versa. The situation really re-
mained the same with the exception of the changed stimuli — i.e.,
the substitution of the response is now from the auditory stimulus
to light instead of from pain to sound as before. The same
time intervals were studied as before. As in the previous
experiment, it was necessary to enter the special series of
test trials only in the study of the first interval to determine the
extent to which the transfer had been made at any given time.
The same standard is required here as in other problems.

IV.

EXPERIMENTAL RESULTS

i. The Influence of the Time Interval in Successive Associa-
tion upon the Rate of Learning. — The experiments bearing di-
rectly upon this problem may be divided into two sets : A. Those
in which the association was learned in the forward direction;
B. Those in which the association was learned in the backward
direction. Groups I to XII were tested in the first set and Groups
XIII and XIV in the second. All the animals were of about the
same age and health conditions. Twenty trials were given each
day. To have mastered the problem the animal must have made
not less than 90 per cent of correct responses over a period of
five consecutive days, i.e., 90 per cent on the last 100 trials.

A. Learning the Association in the Forward Direction. — As-
sociations were made between stimuli from different sense fields.
Two pairs of stimuli were used : a) auditory (buzzer) and pain
(electric shock), and b) visual (light) and auditory. The
auditory-pain experiments are presented first.

a) Association of buzzer with electric shock. — In order to se-
cure comparative data upon the influence of the time interval
between the presentation of buzzer and the presentation of pain,
the animals, upon mastering the problem of negative response to
pain,1 were divided into groups, and each group was set to work
to transfer this negative response from pain over to the buzzer.
The sound was always presented before pain but in a different
temporal relation to pain in each of the groups. Other factors
remained constant for all of the groups. In addition to present-
ing the stimuli in immediate succession, i.e., in temporal contact,
presentations at intervals of one, two, four, and six seconds were
made. In all, 38 animals were used.

i ) Continuous presentation. — Group II was trained on buzzer
followed by the pain immediately. The results of the work of
this group are summarized in Table II.

1 See division 3 of this section.

20 JOSEPH U. Y ARE ROUGH

TABLE II

Animal Trials

25 120

26 120

27 140

28 120

29 120

50 TOO

Average 120 91 . 1

A set of test trials was interpolated after each 20 trials of the
regular series in order to measure the gradual growth of the asso-
ciation. The average number of trials necessary to master this
problem was 120, but a study of the curve in Fig. 4 and the
records of the individuals of the group show that after 60 trials
some of the group held their record to and above the 90 per cent
mark, and that after 80 trials no animal fell below this point.

2) One-second Interval. — Group III was used in these tests.
The buzzer was presented one second before pain, to which they
had previously learned to respond negatively. If the old re-
sponse to pain was transferred to the sound stimulus, and the
animals turned around for it, no pain was given. But if they
failed to turn around to the buzzer and instead continued toward
the food-box, the pain was given at the end of one second.
Results are given in Table III. It was unnecessary to give the

TABLE III

Animal Trials

I 118

12 131

22 128

24 126

35 148

37 120

Average 128.5 91.1

special test series in this group in order to determine the gradual
growth of the association, for if, in the regular training series,
the animals failed to turn around before pain was given, the
response was recorded an error: if they turned at the sound no

RATE OF LEARNING IN THE WHITE RAT 21

pain was given. These regular tests, therefore, show the gradual
growth of the association if properly analyzed. The analysis
of Table III is given in the curve presented in Fig. 4. Each
point of this curve was determined by the average per cent of
correct responses for the group in 20 trials of the regular train-
ing series. The growth of the association, as determined by the
percentage of correct responses, is recorded for each 20 trials
during the entire learning period.

The 30 per cent of trials in which no stimuli were given proved
to be an excellent control for determining whether or not the
animals had developed the habit of turning each time in the first
alley selected, or were turning only when the sound was encoun-
tered. Without an exception they continued to the food-box
during these trials. Very often they would proceed slowly and
manifest hesitancy, much as the human does during a period of
expectancy and uncertainty, but in no case did they turn around.
The following conclusion is warranted: The turning response
was transferred from pain to buzzer rather than to some other
factor; 128.5 trials were necessary to perfect this transfer.

3) Two-second interval. — The two stimuli were presented to
Group IV with an interval of two seconds. Again the special
tests were omitted and the rate of of learning measured in the
regular trials. Tabulated results for the group are to be found
in Table IV. With the exception of Nos. 10 and 39 the indi-
vidual differences are small. With No. 10 the association began
early in the period, but it was very difficult for him to hold the
percentage of correct response over a period of five days. Num-

TABLE IV

Animal Trials % on Last 100

10 316 92

11 210 91

21 166 92

23 211 90

38 203 93

39 269 oo

Average 229 91.2

22 JOSEPH U. YARBROUGH

ber 39 was very slow in beginning the transfer, but when once
started the percentage increased normally.

The gradual growth of the association in this group is obvious
from the curve in Fig. 4. Each point of this curve represents
the average for the group in percentage of correct response for
20 trials of the regular learning series. Although we had used
the 30 per cent of undisturbed trials throughout these experi-
ments, at the end of the work each animal was given a special
series of ten test trials in which pain was withheld. In these tests
the response was unaffected. The group required an average of
229 trials to reach the standard imposed.

4) Four-second Interval. — With Group V the sound was pre-
sented four seconds before pain. Seven animals were used in
the group, and Table V gives their data. There are no marked
individual differences, a gradual increase in correct responses
being maintained by each animal, as is graphically shown in the
learning curve in Fig. 5.

TABLE V

Animal Trials % on Last 100

4 236 94

13 265 92

1 6 240 91

19 187 91

40 228 93

41 216 94
52 269 91

Average 236 92.4

After the work was completed the special set of test trials was
given with no effect upon the per cent of correct responses. The
average number of trials necessary for mastery was 236.

5) Six-second Interval. — Group VI made the association over
an interval of six seconds. Table VI shows the number of trials
necessary to make the association and the standard of correct
reaction reached. From the table and the learning curve in Fig.
5 it is possible to appreciate the general disturbance during the
first 60 to 100 trials, as well as the gradual growth of the asso-
ciation. During this first period of learning the animals not only

RATE OF LEARNING IN THE WHITE RAT 23

ignored the buzzer, but also the negative response to pain often
broke down.

TABLE VI

Animal Trials % on Last 100

43 229 92

45 226 91

46 217 9i

47 224 90

48 274 90

49 280 93
51 288 89

Average 248.3 90.9

In other words, if the animal failed to turn for the buzzer
he also often failed to turn for the pain and continued his
course toward the food-box until he encounted the closed door.
This same thing happened in all the groups but not so frequently
as in this one. This common period of confusion prehaps ac-
counts for the larger number of trials necessary for this group
to master the association.

Summary. — The results for these five groups are brought to-
gether in Table VII. From these results the following con-

TABLE VII

Group Time Interval Trials

II o seconds

III i "

IV 2 "
V 4 "

VI 6 "

Average 192.3

elusions are warranted: i. Measured in terms of the number
of trials necessary to learn the association, the difficulty of mak-
ing the transfer of response does not increase uniformly for the
increase in the length of the time interval between the two stimuli
presented. 2. There is a disproportionate increase in dif-
ficulty when the interval is increased from one to two seconds.
3. There is practically no increase in difficulty when the interval

JOSEPH U. YARB ROUGH

100

40

100

*Q|

60

40

TrUls 80

140
Figure 4

220

Trial*

.120

180

240

300

Figure 5

Curves 2, 3 and 4 (Fig. 4) are the learning curves for the Groups II,
III and IV, respectively. Curves 4, 5 and 6 (Fig. 5) are the learning curves
for Groups IV, V, and VI.

is increased from two to four seconds. 4. The six-second inter-
val is more difficult than the four-second, but the increase in
difficulty is not so great as from the one-second to the two-second
interval.

b) Association of Light with Sound.' — When the animal had
made the association of sound with pain, the pain had in reality
dropped out, and the animal responded to the buzzer whenever
presented. To demonstrate the accuracy of this statement a
special set of test trials was given at the end of the learning
period in which pain was not given at all, and the percentage of
correct response was not affected. It was possible, therefore, to

RATE OF LEARNING IN THE WHITE RAT 25

introduce another stimulus affecting still another sense at certain
definite time intervals before the buzzer, just as the buzzer had
been entered before pain in the groups previously reported.
Light was selected for this new stimulus. By properly connect-
ing the timing apparatus in the light circuit as described above,
it was possible to present the two stimuli (sound and light) with
a known time interval between them. The animals were re-
grouped and five new groups were set to work on learning to
transfer their motor response from sound over to light. With
the exception of the time interval, conditions were constant for
all the groups. As before, experiments were made with con-
tinuous presentation and with one, two, four, and six second in-
tervals. The 90 per cent standard was again adopted. In all,
24 animals were used on this problem.

i ) Continuous presentation. — Here the problem was the asso-
ciation of light with sound when presented in immediate suc-
cession. The data for this group are given in Table VIII. One

TABLE VIII

Animal Trials

48 100

49 ioo

50 ioo

51 ioo

Average ioo 91 . 5

hundred trials were given to each animal, but the test trials show
that the association was complete at the end of 60 trials. The
curve in Fig. 6 shows a period of fluctuation in correct response
between the regular series and the test trials. After 60 trials in
the regular series the curve remains almost uniform at 90 and
above. This uniformity was reached in Group II at the end of
40 trials. A comparison of this curve with that of Group II,
Fig. 4, indicates that the light and the buzzer offered a problem
of practically equal difficulty with the sound-pain association in
the first part of the period, but the transfer to light was more
readily made. From a study of the individual records in the
two groups it was found that under these conditions when ani-

26 JOSEPH U. Y ARE ROUGH

mals once learn to respond to light they hold a more uniformly
high record than when responding to sound. On the basis of
the number of trials necessary to make the association it was
easier for the animals to associate light with sound than sound
with pain, when the two stimuli were presented continuously.

2) One-second interval. — The animals in Group IX made the
light-buzzer association over a one-second interval. They had
just transferred the negative response from pain to buzzer, and
the new problem called for this response to be transferred to
light, the latter being presented one second before the former.
When the animal turned to the light, the buzzer was not sounded.
Failure to respond negatively to light was recorded an error.
Table IX gives the data for this group. No special test trials

TABLE IX

Animal Trials % on Last 100

37 1 60 94

45 140 92

46 140 94

47 140 93

Average 145 93.2

were given, the results of the regular training series being suf-
ficient to show the rate of learning the association. The per-
centage is very low at the beginning of the learning period and
increases rapidly, giving a positive accelerating curve. When
compared with the learning curve for Group III, where all con-
ditions are the same except the stimuli, two marked differences
are found : i ) The percentage of correct response in the first 20
trials is considerably lower for Group IX; 2) no learning is
indicated by Group IX until after 40 trials, while learning appears
to have begun immediately with the other group. Under these
conditions, 145 trials were necessary to transfer the negative
response from the sound to the light.

3) Two-second interval. — To Group X the stimuli were pre-
sented with an interval of two seconds between them. The
animals dropped low in efficiency in the first series, but when
learning began the curve rose as rapidly as that for Group IX.

RATE OF LEARNING IN THE WHITE RAT 27

See data in Table X and the learning curve in Fig. 6. The prob-
lem for Group X differed from that for Group IV only in the
matter of stimuli used, but there is a marked difference between

TABLE X

Animal Trials

13 160

16 180

19 160

35 180

Average 170

the curves for the two groups. In the Group IV curve is shown
a period of no learning, followed by a brief plateau and a second
period of gradual growth, which brings the record up to the
standard. In contrast, the Group X curve shows a shorter period
of no learning, followed by rapid progress. There are two pos-
sible explanations for this dissimilarity. First, it may be due
to the previous training the animals in Group X had had in the
problem of pain-sound transfer. Second, the pain-sound trans-
fer may be more difficult than the sound-light transfer.

4) Four-second Interval. — The animals in Group XI were
trained to associate light with the buzzer when presented with
a time interval of four seconds. Table XI records the results.

TABLE XI

Animal Trials

4 160

23 140

38 160

39 160

Average 155 91.5

From these data and the curve in Fig. 7 one may see the rate
of learning the association and the number of trials necessary
to master the problem. There are no marked differences between
the data of this group and those of Group X above. In each
case there is a complete breakdown in the early part of the period
followed by a period of rapid learning. With this group there

28

JOSEPH U. Y ARE ROUGH

40

Trials

80

120

Figure 6

180

100

80

60

40

Trials

80

120

180

Figure 7

Curves 8, 9, and 10 (Fig. 6) are the learning curves for Groups VIII, IX,
and X, respectively. Curves 10, n and 12 (Fig. 7) are the curves for
Groups, X, XI and XII.

appears to be no increase in difficulty with an increase of the
time interval from two to four seconds. But when compared
with the data of Group V in which sound and pain were pre-
sented in the same temporal relation, two points of difference

RATE OF LEARNING IN THE WHITE RAT 29

appear. First, the breakdown in the beginning of the learning
period is more marked in the latter. A percentage of 45 was
made in this group, while in Group V an efficiency of 62 per
cent was maintained. Second, the learning curve rises much
more rapidly for Group XL The percentage of increase rose from
45 to 95 m f°ur series of trials, while in Group V the increase
for the corresponding period was from 67 to 74 per cent. Seven
days were required for Group XI to reach the 90 per cent mark
of correct response. Previous training is an explanatory factor,
but indications are that this difference is largely due to difference
in the stimuli used. The data thus far gathered indicate that,
other things being equal, it is more difficult for white rats of the
type used in these experiments to associate sound with pain than
light with sound.

5) Six-second Interval. — Group XII learned the association
when light was presented six seconds before sound. In Table
XII is given the data for this group. The association over this
time interval required more trials than did that for any other
group, but the difference is not great enough to be of significance.
The curve in Fig. 7 will assist in showing the rate of learning.
In form it follows rather closely the curve for Group XI — low
at first, followed by a period of rapid learning.

TABLE XII

Animal Trials % on Last 100

i 160 91

10 200 94

11 180 92

12 180 95

Average 180 92.5

The effect of stimulus difference is shown by comparing this
curve with that of Group VI. The curve for the former group
begins somewhat higher, and instead of rising higher it first
tends downward and later ascends, gradually reaching the 90
per cent mark after 180 trials; the curve for the latter exhibits a
very brief period of no learning, followed by very rapid progress.

Summary : To enable the reader to see at a glance the results

30 JOSEPH U. Y ARE ROUGH

obtained in each of these five groups the following table was
constructed. From these data and the curves presented the fol-

TABLE XIII

Group Time Interval Trials

VIII o seconds 100

IX I " 145

X 2 " 170

XI 4 " 155

XII 6 " 180

Average 150

lowing conclusions were reached : Judged on the basis of the
number of trials necessary to make the association, the difficulty
of the problem of transfer of response from light to sound
appears not to have increased proportionately with the increase
in time interval between the presentation of the two stimuli.
Practically the same increase in difficulty is manifest in going
from continuous presentation to the one-second interval as in go-
ing from the one to the two-second interval. The four-second
interval, on the other hand, seems easier to master than the two-
second but less easy than the one-second. The learning curves
start low, go through a short period of no learning, and then
rise rapidly in each group.

B. Learning the Association in the Backward Direction. — In
reading the literature on the phenomena of association one is
lead to the belief that psychologists generally hold to the con-
clusion that recall is predominantly in the forward direction.
Some experimental data, however, indicate that an association
will function backward as well as forward. Freeman found
that "the formation of an association between a stimulus and a
motor response by animals is exceedingly difficult and perhaps
impossible when the stimulus is presented after the act has oc-
curred." Since our method of procedure and experimental con-
ditions were so different it was decided to include an examination
of this point.

Two groups of animals (Groups XIII and XIV) were used
in these experiments, the problem being the transfer of the pre-

RATE OF LEARNING IN THE WHITE RAT 31

viously learned negative response from pain over to the auditory
stimulus (electric buzzer). The buzzer was given in each case
after the pain had been given : to Group XIII it was presented
in immediate succession; to Group XIV, with an interval of one
second separating it from pain. Thus, in the latter case the two
factors were presented in the backward order with a time inter-
val of one second between them, i.e., the sound stimulus was pre-
sented after the motor response to pain had been initiated and in
some cases completed. All the animals were given, in addition
to the regular training series, a series of test trials. The object
of the training series was to establish an effective association
between the turning response to pain and the auditory stimulus ;
the object of the test trials was to determine the degree to which
an effective association had been formed. In these tests the pro-
cedure was the same as in the training series except that no pain
was given. If the negative response had been transferred to the
auditory stimulus, they should turn and retrace the path when
the buzzer was sounded. A correct response consisted of an
immediate turn when the auditory stimulus was presented, or
of traversing the whole length of the path into the food-box
in the 30 per cent of trials in which no stimulus was given.
Progress was measured by the increasing percentage of correct
response during the test trials. For in each group the established
negative response to pain prevented the animals from responding
to sound during the training series. They would most often
have turned around or be turning around when the auditory
stimulus was given. It is obvious, therefore, that the new stim-
ulus which was to be associated with the turning response oc-
curred after the act of turning had been initiated. A correct
response in the regular training series is the same as it was before
the auditory stimulus was introduced, i.e., an immediate turn
when the pain stimulus was presented, or traversing the whole
length of the path into the food-box in the 30 per cent of trials
in which no stimulus was given.

The number of animals used, the number of training trials
necessary to make the association, and the percentage of correct

?2 JOSEPH U. Y ARE ROUGH

responses for Group XIII are given in Table XIV. In the test
series only one term of the pair to be associated was presented.
These trials would, therefore, not be added to the regular train-
ing series in order to secure the number of trials necessary to

TABLE XIV

Rat Training Trials Test Trials % on Last 100

53 J40 60 90

54 120 50 92.5

55 140 60 94

56 120 50 91

Average 130 55 91.8

make the association. Their effect would really be expected to
hinder the rate of learning, since each time they took the place
of a regular day's work, thus making a break in the regular
training series. Since each animal completed his work on the test
series, the percentage of right response on the last 100 trials, pre-
sented in the last column on the table, includes four series (40)
of test trials and three (60) of trainnig trials. It was planned
to use only the test trials as a basis for this percentage, but since
each animal had held a percentage of 90 or above for a period
of seven days we were satisfied that the transfer of response
was completely made.

From Graph 13 in Fig. 8 the rate at which the association was
made can easily be observed. The values in this curve represent
the percentages of correct response in the test trials given at the
end of the first 40 trials and following each series of 20 trials
thereafter. The test series was not introduced in these groups
until after the first 40 trials, because it was assumed that not
until after then would the transfer of response be sufficiently
made to make it worth while to interrupt the regular training
series. Since the correct response in the 30 per cent of trials in
which no stimulus was given had become thorougly learned in
the previous training, mistakes in these trials seldom occurred.
Consequently, the percentage records of this group may be ex-
pected to begin at 30 and increase to 90 and above as the motor
response is transferred to the auditory stimulus. Since the test

RATn OF LEARNING IN , THE WHITE RAT 33

trials interpolated after the first 40 training trials show 79 per
cent of correct response, it is safe to infer that considerable
transfer was made during these 40 trials. The association was
readily made; with, in fact, but little more difficulty than was
experienced when the two stimuli were presented continuously in
the forward direction. This small difference may be due to
chance or to the fact that a smaller group was used in these last
experiments.

When the animals of Group XIII had learned the association
of negative motor response to sound, light was presented in the
same temporal relation to sound as sound had been to pain in
the previous training, i.e., in immediate succession. By this
means it was hoped to get some data on the relative merit of the
different stimuli. Table XV depicts the animals used, the num-
ber of trials necessary to master the problem, and the percentage
of right response on the last 100 trials. A comparison of this

TABLE XV

Rat Training Trials Test Trials % on Last 100

53 120 50 92

54 ioo 40 90

55 120 50 92

56 120 50 91

Average 115 47.5 91.5

table with Table XIV and of Graph 13!. in Fig. 8 (the curves
for the present group) with Graph 13 of the same figure shows
no marked difference between the two — so little, in fact, that we
are not warranted in making the statement that it is easier to as-
sociate a motor response with light than with sound when all
other conditions are constant. It is preferable to account for
this small difference on the basis of the effect of transfer of
learning.

In Group XIV where the sound was presented one second
after the pain (in most cases after the motor response to pain)
the conditions were in all other respects the same as those for
Group XIII. Under these conditions we hoped to secure data
showing the relative difficulty in making the association in the

34 JOSEPH U. YARB ROUGH

backward direction when the two members to be associated are
presented continuously and when they are presented with an
interval of one second between them. The data for this group
are given in Table XVI. The last column of this table is mis-

TABLE XVI

Rat Training Trials Test Trials % on Last 100

57 320 70 73

58 340 80 77

59 380 90 68

60 320 70 76

Average 340 77-5 73-5

leading unless it is understood that this last 100 trials include
four series (80) of the regular training trials and only two series
(20) of the test trials. This is accounted for by the fact that
the test trials were interpolated after the first 60 regular trials
and after each 20 subsequent trials. This was done because of
the apparent slow rate of making the association under these con-
ditions. The rate of increase in the strength of the association
can be observed more accurately by reference to Graph 14 in
Fig. 8. This curve is constructed on the basis of percentages of
correct response on the test trials only. Although all the animals
made occasional correct responses to the sound stimulus during
the test trials no real learning is observable at the end of 380
training trials. It is to be regretted that the work of this group
was not continued, but sickness caused the experiments to close
at this point. Yet, several features of the results deserve com-
ment: i) Measured in terms of the number of trials necessary
to make the association function in the backward direction, it is
much more difficult to form the association when the terms pre-
sented are separated by one second than when they are presented
continuously. The animals in the group to which the stimuli
were presented continuously learned the association in an average
of 130 trials, while those in the former group showed no im-
provement at the end of 340 trials. With more training the
animals of the group might have learned the association, but we
have no evidence that they would have done so. 2) Judged upon

RATE OF LEARNING IN THE WHITE RAT 35

the same basis, it is more difficult to learn to associate a motor
response with an auditory stimulus when the latter is presented
one second after the former has occurred than when it is pre-
sented one second prior to the initiation of the motor response.
Under the latter condition Group III learned the association on an
average of 128.5 trials, while under the former conditions Group
XIV failed to make the association, although they had an average
of 340 trials. 3) When the stimuli are presented in immediate
succession it is more difficult for the association to function in
the backward than in the forward direction. To account for
the comparative ease with which the animals made the associa-
tion in the backward direction under these conditions, one may
assume that the new stimulus was not associated with the old
stimulus. This response and the new stimulus were experienced
simultaneously, and association is due to this simultaneity.
Summary :

1. Association in the backward direction is only slightly more
difficult than in the forward direction when the terms to be asso-
ciated are presented in immediate succession. This small differ-
ence is by no means conclusive.

2. It is much more difficult to learn an association in the
backward direction when the terms are separated by one second
than when they are presented in immediate succession.

3. It is more difficult to associate a motor response with an
auditory stimulus when the latter is presented after the former
has occurred than when it is presented before- the initiation of
the motor response.

4. The data found do not warrant the statement that it is
easier to associate a motor response with light than with sound
when the stimuli are presented in immediate succession.

2. Simultaneous and Successive Presentation. — The present
investigation was not undertaken primarily for the purpose of
gathering experimental data relating to the well known contro-
versy between simultaneous and successive association advo-
cates, but the condition of our problem made it profitable to
do so. Experimental data thus far reported are divided on this

36
too

80

40

JOSEPH U. Y ARE ROUGH

Trials

160

JOO

40

Trials 80

FIG. 9

In Fig. 8 curves 2 and 3 are reproduced from Fig. 4; curve 13 is the
learning curve for Group XIII on sound, and curve I3L is for the same
group on light. Curve 14 represents the work of Group XIV. In Fig. 9
curve 15 is the composite curve for simultaneous groups (Groups I and
VII) ; curve 16 is the composite curve for successive groups (Groups II
and VIII).

point, but the better controlled experiments report data indicat-
ing the superiority of the simultaneous mode. The difference in
materials used, experimental conditions, and methods of pro-
cedure, no doubt account for the conflicting results. Our data
support the superiority of simultaneous presentation. Almost
all results thus far reported have come from investigations with
human subjects, a fact which gives our data the more significance,
i ) Simultaneous Mode of Presentation. — Two groups of ani-
mals learned to make the association when the two stimuli were

RATE OF LEARNING IN THE WHITE RAT 37

presented simultaneously. With the one, electric shock and
buzzer were used ; with the other, electric buzzer and light. The
previously discussed test trials were interpolated after each series
of 20 training trials in order to check the rate of learning. These
tests were not made with Nos. 2 and 9. These continued on the
regular learning series until they had had 100 trials before the
test series was entered. These animals do not appear to have had
any advantage over the other animals of this group where the
special tests were entered earlier. These special tests were not
expected to assist in learning the association, but they do not
appear to have hindered it.

When the pain-buzzer group (Group I) had learned to respond
negatively to the first stimulus, pain, buzzer was sounded simul-
taneously with pain, and the problem was to transfer the negative
response from pain over to buzzer. In Table XIV the number
of trials necessary to master the problem and the strength of the
association are given. The number of test trials given is not
shown in this table, but is used in constructing the curve in Fig. 9.

TABLE XIV

Animal Training Trials % on Last 100

2 120 91

9 100 96

14 100 91

31 100 91

32 120 90

33 140 92

Average 113.3 Qi-6

No marked individual differences appear. Although the average
number of trials necessary to learn the association is 113.3, m
each case where the test trials were given the animal had so com-
pletely made the transfer that at the end of 60 trials he made
90 per cent on the test trials. However, because 90 per cent
on the last 100 trials had been adopted as a standard, each of the
animals was given additional trials on the training series.

The four animals of Group VII had not only learned the
negative response to pain, but had transferred this response to

38 JOSEPH U. Y ARE ROUGH

the sound stimulus. After the habitual response to the buzzer
had been built up an electric light was presented simultaneously
with the buzzer. It was hoped by this means to determine the
rate and number of trials necessary to transfer this response
from sound over to the simultaneously experienced light. During
the first series of 20 trials only a slight disturbance was shown,
the average percentage of correct response for the group being 86.
The animals continued to respond to the sound but in the presence
of light, a rather disturbing factor. The special series of 10 test
trials broke down the response. Only 24 per cent above the
minimum was made by the group, with two animals falling to
10 per cent above. Table XV contains the individual records.

TABLE XV
Animal Training Trials % on Last 100

49 ioo 93

41 ioo 96

43 ioo 91

52 ioo 92

Average ioo 93

At the end of ioo trials each animal had made the association.
The transfer again was very rapid, and only a comparatively
small number of trials were necessary to make the association.

The learning curves for these two groups, given in Fig. 9
depict the rapid growth of the association. The construction is
based on the per cent of right responses in both the regular series
and the special test trials. The first and each alternate point
thereafter is determined by the average per cent of right response
for the group in 20 regular training trials. The other points are
the average per cent of efficiency for the group in 10 special
test trials. From the curve it is seen that Group VII formed the
association by the end of 40 trials, for at no time after this did
the percentage of the group fall below 90.

When these two groups are compared there are two points to
be noted, i) In both groups the percentage of right response on
the first 20 learning curves is practically the same. In these trials
the animals were reacting to an old stimulus but in the presence

RATE OF LEARNING IN THE WHITE RAT 39

of a new one; namely, sound for Group I and light for Group
VII. The two stimuli seemed to offer about equal disturbing
effect at this point in the learning. 2) When the first special
test series was given a complete breakdown occurred in Group
VII while Group I only partially broke down. The drop in per-
centage of right response was from 85 to 55 in the former, and
from 84 to 62 in the latter. This breakdown was only temporary,
for in the next special series the percentage rose above that of
Group I. Light seems to be more difficult in the beginning, but
measured on the basis of the required number of trials in the
two groups there is little or no difference.

2) Continuous Mode of Presentation. — The work of these
two groups (Groups II and VIII) was reported with the other
groups on successive association, but in the table below the re-
sults are reproduced in order to get a quick comparison of the
two modes of presentation. The curves presented in Fig. 9

TABLE XVI

Simultaneous Trials

Group 1 113.3

Group VII 100

Continuous Trials

Group II 120

Group VIII 100

show the different rate of learning the association by the two
groups.

With these results before us we are able to observe the follow-
ing facts: i) Associations are possible with either the simul-
taneous or the successive mode of presentation. 2) Our data
indicate the superiority of the simultaneous mode of presentation,
in that under these conditions the association is more readily
made — i.e., fewer trials are necessary to establish it than when
the stimuli are presented in succession. 3) The theory that
associations are possible only when the terms to be associated
are presented simultaneously is not supported by our results.

3. Learning the Negative Response to Pain.1 — All the animals

1 For method of procedure, one may with profit re-examine the last few
pages of Section III above.

40 JOSEPH U. YARB ROUGH

used in the experiments of this monograph learned the negative
response to pain before beginning work on their respective prob-
lems. In training the animals to make this response two mazes
were used. With one group of 23 animals the maze shown in
Fig. i was used; and the maze described by Carr and Freeman
in the report of their studies2 was selected for the training of the
latter group. The results of the former group are presented
first.

The results for the 23 animals trained on the first mentioned
box are given in Table XVII. The first column gives the animal ;
the second, the number of trials necessary to establish the desired
response, and the third the actual number of turns made by each
animal before he attained the adopted standard. Since 90 per
cent of correct responses must be made within the last 100 trials,
it is obvious that few turns were made until near the end of the

TABLE XVII

Animals Trials No. of Turns to Pain

1 820 112

2 770 98
4 1015 157
9 745 107

10 684 99

11 890 135

12 901 151

13 1024 136

14 1009 113
16 1079 n8
19 753 132

21 715 119

22 914 128

23 803 143

24 926 99

25 1045 123

26 III9 102

27 H34 106

28 937 ioi

29 1159 160

31 1376 124

32 979 109

33 1251 n6

Average 960.6 121.2

2 Op. cit., p. 466.

RATE OF LEARNING IN THE WHITE RAT 41

training. The number of trials necessary to master this problem
varied from 684 for No. 10 to 1376 for No. 31, while the
average for the group was 960.6. Since no pain was given in
30 per cent of the trials these animals on the average encountered
pain 672 times before learning to make the desired response.
Because of the large number of trials required and the amount
of time lost in the long zigzag runways of this box, it was thought
expedient to use a simpler box in training the remaining animals.
The apparatus used by Carr and Freeman was selected with the
hope that time could be saved each day by its use. Future ex-
periments required that the animal be able to hold the standard
of correct response in the box presented in this paper and not in
the simpler one. It was necessary, therefore, that both apparatus
be learned, a necessity which made it seem probable that the
group working in the new box would require more trials than the
other group. In order to avoid any delay in making the transfer
of correct response from one box to the other, the animals were
each day given one run in the long maze after having completed
the day's work in the short one. In this way the two mazes
were learned together: the first being partially learned, then the
second partially learned, and finally the first and then the second
completely learned. In the light of the findings of Wiltbank1
this was thought to be the best possible method. Table XVIII
gives the number of animals, the number of trials, and the num-
ber of turns made by each animal before reaching the required
standard of efficiency. The range of trials required is from 394
to 1196; the average for the group of 15 rats is 694,4. Since
in 30 per cent of these trials no pain was given, the animals
received pain 486 times before learning to respond to it nega-
tively. Not only did this method save time in giving the series
from day to day, but on the average it reduced the number of
presentations of pain from 672 (the number required for the
long-maze groups) to 468 — a saving of 204. There are at least
three possible explanations of this difference: i) The increased
complexity of the longer box increased the difficulty of the prob-

1 Wiltbank, R. T., "Transfer of Training in White Rat Upon Various
Series of Mazes," Beb. Mono., Vol. IV, 1919.

42 JOSEPH U. Y ARE ROUGH

TABLE XVIII
Animals Trials No. of Turns to Pain

35 772 103

37 1196 102

38 787 H5

39 817 108

40 429 98

41 840 04
43 578 91

45 651 104

46 679 93

47 394 88

48 652 94

49 585 96

50 654 9i

51 614 98

52 768 107

Average 694.4 98.8

lem. The larger number of compartments greatly increased the
number of turns and the distance to be traversed in ecah trial,
thereby causing confusion. The records show a small number
of instances where an animal, after having run the maze for
several hundred times, became lost and ran back and forth
through three to five adjacent compartments several times before
he proceeded to the food-box. 2) With the short box the trials
were more frequent than with the long one. It took less than
half the time to run the smaller maze; thus, the trials could be
given twice as frequently. 3) The getting of food was delayed
by the long runways of the larger box. The influence of delay-
ing food, the apparent motive for the animal's effort, is not
known, but we are of the opinion that it was an important fac-
tor with our animals.

Although the number of trials was reduced when the shorter
box was used there was really no difference in the rate of learn-
ing between the two groups after learning once began. A study
of the learning curves in Fig. 10 confirms the accuracy of this
statement. In constructing these curves only the last 500 trials
for each animal were used. This was done for two reasons:
i) All observable learning is included in this period. 2) It per-

RATE OF LEARNING IN THE WHITE RAT

43

100

Trial*

200

300

.400

500

FIG. 10

Curve 15 in Fig. 10 represents the rate of learning in the last 500 trials
for groups of animals trained in our maze : Curve 16, the rate for those
trained in the Freeman maze.

mits a comparison of the rate of learning at the exact time learn-
ing was in progress in the individual cases. This is not possible
in either group if the curve is plotted on the basis of trials made
from the beginning of the problem, for some completed the
training before others showed signs of improvement. In each
group there is a period of no apparent learning followed by a
short period of rapid learning.

The large number of animals in each of these groups and the
almost uniform difference between the number of trials required
to master the problem are conclusive evidence that the negative
response to pain can be more easily learned in a simple maze
than in a more complex one, and that the rate of learning is
approximately the same from the time obvious learning first ap-
pears until the problem is mastered.

V.

INTERPRETATION OF RESULTS

Now that the experimental results have been presented there
remains the important task of ascertaining, as far as we can
do so, the relative scope and value of the different factors in-
volved. In making this interpretation and evaluation the dif-
ferent theories that have been advanced by the various investi-
gators to account for the experimental results previously reported
will be discussed. The following questions set forth the chief
points of discussion: i) How can association be accounted for
when the two terms to be associated are presented with a certain
time interval between them? 2) How can association be made
when the two terms are presented in the reverse order? 3) Is
the simultaneous mode of presentation superior to the successive?
These questions are considered below in the order given.

i. The nature of the connection established between the two
members of a pair of external stimuli, the first member of which
is presented for a definite time and is succeeded by the second
after a definite interval, is the problem to the solution of which
the numerous experimental investigations, beginning with Eb-
binghaus and continuing through the past 25 years, have aimed
to contribute. Thus far, three theories have been advanced for
the explanation of this phenomenon, i) One may assume the
survival of a memory image of the first member of a pair of
stimuli during the time interval which separated it from the
second stimuli. The second term of the association, then, coin-
cides in time with the "gradually-dying-away" memory image
of the first, and it is due to this simultaneity that the association
is formed. 2) The connection between the stimuli may be me-
diated by some third factor which is not only common to the two
stimuli, but is also present during the interval which separates
them. For example, a series of syllables may be more success-
fully memorized by associating each syllable with its position in

RATE OF LEARNING IN THE WHITE RAT 45

a temporal sequence. The continued running activity of the
animal and the external environment were possible third factors
in our experiments. If the running activity, which was almost
always present, did serve such a mediating function no increase
in difficulty in forming the association over different time inter-
vals should have been evidenced. Nor can this variation among
the different groups be accounted for on the basis of the external
environment, for it remained constant. 3) The third theory
assumes that the two terms may be associated directly, i.e., when
the first member of a pair of stimuli ceases the actual nervous
excitation is not finished but continues for a time, gradually
diminishing in strength. The second member coincides with the
diminishing phase of the preceding one, and the association is
due to this simultaneity of the two experiences. Although the
external stimuli are presented in succession the internal ex-
periences are in a large measure coexistent.

In the light of our data, this "gradually diminishing nervous
excitation" continues for at least six seconds. This after-phase
may account for the distinct, though disproportionate, increase
in difficulty in making the association over larger and larger
intervals. If this difficulty in making the association is a measure
of the waning strength of the after-phase of the first term, our
data warrant the statement that little or no loss in strength is
experienced during the first second, but by the end of two seconds
a marked weakening is observed. For longer intervals the diminu-
tion is gradual with no marked breaks observable. This is in
agreement with Froeberg's finding that association can be formed
over an interval of at least five seconds.

Although the hypothesis of direct association seems to be the
best explanation of successive association, the following pos-
sible objections to its acceptance should be pointed out: a)
When human subjects are used it may be assumed that all
memory and thought connection are not wholly excluded. These
connections are many, and some are very difficult to overcome,
b) Where activities are interpolated to destroy the memory image
of the preceding term it may be assumed that these activities

46 JOSEPH U. YARB ROUGH

themselves mediate the association. However, in our experi-
ments with animals these objections are not valid, for few be-
lieve that white rats master problems by the use of memory
images and concepts, c) A third and perhaps a more vital ob-
jection to this theory has been made by Carr; namely, that while
one can easily conceive of the persistence of a neural activity
for a short interval of time on the principle of inertia, yet the
assumption of an indefinite continuance rather taxes one's cre-
dulity. For this reason a fourth type of explanation is per-
haps the most acceptable. 4) This, the last theory advanced,
explains the connection formed between the two stimuli in terms
of the decreased resistance of the motor center. The connection
is mediated by so decreasing the resistance of the motor center
that the one term will become the adequate motor outlet for the
nervous impulse aroused by the other. Such a conception is
based upon the well known and generally accepted theory of
synaptic resistance. The two theories have in common two
assumptions: i) Each sensory impulse tends to follow that
motor path which for the time offers the minimum of resistance.
2) The resistance of any center is decreased by an increase in
metabolism in that center. If the resistance of one motor center
has by continued practice been so reduced that its readiness for
response is much greater than that of any other motor outlet,
there are no reasons why an effective association cannot be
formed over a considerable interval of time.

Since there was no observable third factor mediating the asso-
ciation in our work and since survival of the memory image
most likely is not a factor in the behavior of the white rat, there
remain but two possible explanations for our results: namely,
either the theory of the "akoluthic" phase or the reduced resist-
ance of the motor center.

2. Our discussion has thus far been concerned with the
phenomenon of successive association when the two members to
be associated were presented in a certain definite temporal rela-
tion. We shall now examine successive association when formed
under the same conditions except that the order of the presenta-

RATE OF LEARNING IN THE WHITE RAT 47

tion of the terms is reversed. Under these conditions the con-
nection between the terms may be explained upon the same hypo-
thesis as in the experiments in the forward direction. For there
is no reason for the assumption that, since the animal has made
a certain motor response to the first term, the neural excitation
produced by this term suddenly ceases. On the contrary, if the
stimulus gets sufficient attention to call out this motor response
its neural effect might be expected to diminish more gradually
than when no reaction is made to it. The assumption is made
here that the "akoluthic" phase diminishes at a rate proportionate
to the intensity of the stimulus, which, of course, we have not
proved. Although our data are not conclusive on this point
they indicate greatly increased difficulty in learning the associa-
tion when the terms are presented in the backward order. This
increase in difficulty may be explained by factors of the situation
other than the "akoluthic" phase — factors incident to the reversal
of the stimuli, i) The animal had previously been thoroughly
trained to turn when pain was presented. As this response was
learned it was easy to observe a change in his behavior from a
careless, random run to a slower movement which resembled
human behavior when in a state of expectancy. This method of
approach made him more responsive to any stimulus presented at
the time. He, therefore, readily transferred the response from
pain to sound when the buzzer was presented before the motor
response. If, however, the buzzer was presented after the motor
response had occurred this state of "expectancy" was absent,
and the transfer of response was more difficult. 2) The previous
training had taught the animals that when once turned they must
retrace the maze and select the opposite side in order to secure
food. Without exception this was true. Hence, after they
turned around for pain and started on a never-failing road to the
food-box, it was scarcely probable that they would proceed in
the same state of "expectancy" as before. Daily observations
showed that they did not. 3) One may assume that the two
terms were associated, but that it was more difficult for them to
function in the backward direction. The conditions for learning

4S JOSEPH U. YARB ROUGH

here are probably more difficult than in learning the negative
response to pain.

3. In most of the experiments reported the problem of suc-
cessive and simultaneous mode of presentation has been im-
portant. In these, successive presentation has been thought of
as immediate succession of the members of the stimuli. The re-
sults are divided as to the better mode of presentation, but the
majority of the data favor the simultaneous method. The writer
is of the opinion that the difference in results is best accounted
for by the difference in the material used in the several experi-
ments and difference in the method of procedure.

The most important point of difference in method of pro-
cedure is the attempt to secure an exposure time in successive
presentation equal in length to that in simultaneous presentation.
If each member of the pair in the successive series is exposed
for the same length of time as the pair of simultaneous stimuli
the learning time is twice as long in the successive as in the
simultaneous presentation. Wohlgemuth felt a necessity for
making this exposure time equal in length if comparative data
were to be secured. To accomplish this he devised two methods :
i ) The exposure time for the simultaneous pair was made twice
as long as for each member of the successive pair, the number of
presentations remaining the same; 2) The exposure time re-
mained the same while the number of presentations were twice
as great in the simultaneous as in the successive pairs. Froeberg
objects to both of these methods. To the first his objection is
that it "implies that the strength of an association depends upon
the exposure time of the stimuli only, regardless of whether or
not the pair is exposed at the same time, and thus contradicts
the theory which was to be proved, namely that association takes
place only while the two experiences are present in conscious-
ness." Of the second method his statement is that "it disregards
the effect of the distribution of presentations."

The criticism of the second method of equalizing the ex-
posure time is pertinent, but so much cannot be said for the
first. This first method is good or bad, depending upon which

RATE OF LEARNING IN THE WHITE RAT 49

phase of the problem is under investigation. If the problem
attempts to determine the effect of the exposure time this is a
good method, but if the aim is to determine whether or not two
terms when presented simultaneously offer greater interference
to each other than when presented successively, thereby hin-
dering the process of learning, the method is inept. The method
used in our experiments and also by Froeberg is preferable in
such investigations. By means of it each term is exposed for an
equal time, regardless of the mode of presentation. In simul-
taneous presentation the exposure time of each of the two terms
is not only equal, but the presentations actually coincide in time.
In the successive mode the exposure time of each of the terms
is also equal, but the presentations succeed each other in time.
It is obvious that such a method reveals the relative value of the
two modes of presentation on the basis of term interference.

Our data do not indicate term interference when the two terms
are presented simultaneously. On the other hand, they indicate
a more rapid learning with the simultaneous than with the suc-
cessive mode. These findings are in agreement with those of
Chamberlain, Bigham, Wohlgemuth, and Froeberg, when the
latter used nonsense syllables as material.

As indicated above, the results have varied when different
materials were used. Disconnected words, associable words,
nonsense syllables, syllables, letters, colors, and diagrams have
served as materials for the experiments thus far reported in the
field of human psychology. From a study of all the results ob-
tained the following conclusion, reached by Froeberg, appears
to be true : If the material is such that the pair of stimuli form
an organic whole or can be attended as a unit, simultaneous
presentation is preferable; if it is such that the materials stand
as separate entities, or must be attended to separately in order to
be apprehended, successive presentation is preferable.

In the field of Animal Psychology a certain motor response
and a definite external stimulus have served as the two terms to
be associated. After a motor response to one term is built up
the second term is presented in a certain temporal relation to

50 JOSEPH U. Y ARE ROUGH

the first — i.e., either before or during or after the motor response
has occurred. In the present experiments light and buzzer were
introduced in this way in relation to the motor response, and the
results indicate in both instances the superiority of simultaneous
presentation. The learning was more rapid when light was used
than when sound, but this fact is due rather to transfer of train-
ing than to a difference in stimuli.

VI.

SUMMARY AND CONCLUSION

1. The number of trials necessary to learn the association
does not increase proportionately with the increase of the time
interval between the presentation of sound and pain.

2. Upon the basis of the number of trials necessary to make
the association, the difficulty of the problem steadily increases
with the increase of the time interval between the presentation of
sound and light, but this increase is not proportionate.

3. Our data indicate no marked difference in difficulty for
the continuous and one-second interval presentations; likewise,
no marked difference is manifest for the intervals above two
seconds. The point of disproportionate increase is between one
and two seconds.

4. The excess in the number of trials required for the animals
to transfer a motor response to sound over the number required
to transfer this response to light is perhaps due to the sense
material used, yet the transfer of training is an important factor
here.

5. Association in the backward direction is very little, if any,
more difficult than in the forward direction when the continuous
mode of presentation is used. If, however, the two terms to be
associated are presented with a time interval of one second be-
tween them, the association in the reverse direction is perhaps
impossible.

6. Under our experimental conditions and with the sensory
material used, our data indicate a slight superiority of the simul-
taneous mode of presentation. In order to show the relation of
our results to those of earlier investigators of simultaneous and
successive association, we may classify the findings of the latter
upon the basis of exposure time per term. Wohlgemuth (two
experiments), Bigham, and Chamberlain used an unequal ex-
posure time per term, the inequality favoring the simultaneous
mode, and in every instance they found the simultaneous mode

52 iOKMG IN THE WHITE RAT

to be the better. In the experiments of Wohlgemuth (one ex-
periment), Froeberg, and Freeman the exposure time per term
was equal, and, except for Froeberg's work with nonsense syl-
lables, all results favor succession as the better mode. Froeberg
thinks his exception is not a proof of the superiority of the simul-
taneous mode, but is rather to be accounted for by the sense
material used. Although our results indicate a slight superiority
of the simultaneous mode, this is not sufficient to be of great
significance. These data seem to warrant the conclusion that
when the exposure time per term is unequal the simultaneous
mode is the better ; when it is equal the better mode of presenta-
tion depends upon the materials used.

a/-

4C5004

LJBRAKY

8

UNiVERSITY OF CALIFORNIA LIBRARY