*M,:^^^;,

-..l-xS . /^

^^■"^ ' . '^

'Xa.-«^C'^

V-' <(^\^'

i^^:*-. '

' Vh-

iy>^;

'^<'^•^^.

iJ-T' ■- ■< .i--:jv>fjna/!

HARVARD UNIVERSITY.

LIBRARY

OF THE

MUSEUM OF COMPARATIVE ZOOLOGY

GIFT OF

Ui^^.Q I^V'j- SjijUl^uloL a jQ3,Zj.

5Lr ^ vs^c.

BEHAVIOR MONOGRAPHS

ofif.inraii

Edited by
JOHN B. WATSON and WALTER S. HUNTER

Volume 4
1919-1922

COMPOSED AND PRINTED AT THE

WAVERLY PRESS

Bv THE Williams & Wilkins Company

Baltimore, Md., U. S. A.

CONTENTS

1. Transfer of training in white rats upon various series of mazes.

RUTLEDGE T. WiLTBANK. 1919. Pp. ill + 65.

2. Redintegration in the albino rat: A study in retention.

Thomas William Brockbank. 1919. Pp. iii + 66.

3. Discrimination of light of different wave-lengths by fish.

Cora D. Reeves. 1919. Pp. iv + 106.

4. Visual perception of the chick.

Harold C. Bingham. 1922. Pp. vi + 106.

5. Heredity of wildness and savageness in mice.

Charles A. Coburn. 1922. Pp. iv + 71.

I

mi

Behavior Monographs

Volume 4. Number 1. 1919 Serial Number 17

• Edited by JOHN B. WATSON

The Johns Hopkins University

Transfer of Training in White Rats
Upon Various Series of Mazes

RUTLEDGE T. WILTBANK

Instructor in Psychology, The University of Washington

From the Psychological Laboratory of the University of Chicago

Published

at Cambridge, Boston, Mass.

HENRY HOLT & COMPANY

34 West 33d Street, New York

G E. STECHERT & CO.. London, Paris and Leipzig, Foreign Agents

[Publication temporarily suspended]

The Journal of Animal Behavior

, An organ for the publication of studies concerning the instincts,
habits and intelligence of organisms

The Journal contains a Department of Notes in which appear brief accounts of
especially interesting and valuable observations of behavior.

Published bi-monthly at Cambridge, Boston, Mass., by Henry
Holt and Company, New York. Each volume contains at least 450
pages with plates and text-figures.

The subscription price is $5.00 per volume (foreign subscription
$5.50) postpaid.

Manuscripts, subscriptions, and all business correspondence should
be addressed to

THE JOURNAL OF ANIIMAL BEHAVIOE

Emerson Hall, Cambridge, Massachusetts

Behavior Monographs

For the publication of studies in behavior and intelligence which are

too lengthy or too expensive for acceptance

by The Journal of Animal Behavior

Published at irregular intervals at Cambridge, Boston, Mass., in
connection with the Journal of Animal Behavior, by Henry Holt and
Company, New York. Each volume contains approximately 450
pages with plates and text-figures.

The subscription price is $5.00 per volume (foreign subscription
$5.50) postpaid. ^

Monographs may be purchased separately at prices varying with
the cost of manufacture.

Manuscripts and inquiries concerning terms of publication should
be addressed to the Editor of the Behavior Monographs,
JOHN B. WATSON,

The Johns Hopkins University,

Baltimore, Md.

Subscription to The Journal of Animal Behavior and the Behavior
Monographs should be sent to Emerson Hall, Cambridge, Massa-
chusetts.

JUL £9 1919

Behavior Monographs

Volume 4. Number I. 1919 Serial Number 17

Edited by JOHN B. WATSON

The Johns Hopkins University

Transfer of Training in White Rats
Upon Various Series of Mazes

RUTLEDGE T. WILTBANK

Instructor in Psychology, The University of Washington

From the Psychological Laboratory of the University of Chicago

Published

at Cambridge, Boston, Mass.

HENRY HOLT & COMPANY

34 West 33d Street, New York

G E. STECHERT & CO.. London. Paris and Leipzig. Foreign Agents

Acknowledgments

It is a pleasure for me to acknowledge a deeply felt
indebtedness to my teachers, Professors Angell and Carr,
for their instruction and inspiration.

Professor Carr kindly proposed these experiments, and
patiently and helpfully criticized this presentation of their
methods and results.

CONTENTS

I. Introduction 1

II. Transfer of Training between Pairs of Mazes 5

III. Transfer of Training through Series of Mazes 13

IV. Transfer of Training between a Maze Partially Learned and One Com-

pletely Learned 40

V. Transfer of Training between a Maze Completely Learned and One Already

Partially Learned 45

VI. The Learning of Two Mazes When They are Learned One after the Other

Compared with the Learning of the Same Mazes When One is Partially
Learned, Then the Other Completely Learned, and Finally the Learn-
ing of the Former Completed 56

VII. The Releaming of a Maze after the Learning of Four Intervening Mazes 59

VIII. Summary of Results 63

I. INTRODUCTION

The object of these experiments, stated in general terms,
was to study the kind and the degree of transfer of training
in white rats which were allowed to learn a maze after
they had learned, either completely or partially, one or
more other mazes. Specifically, the object was to find
out:

1. Whether there was transfer of training between a
pair of mazes which were learned completely one after
the other, and if so whether it was positive or negative.
These expressions, " positive transfer " and " negative
transfer," are familiar to those interested in this phase of
psychological research; but it may not be useless to re-
mark, for the sake of some chance reader, that by the
former is meant that the mastery of one or more mazes
renders easier the mastery of a subsequent one, while by
the latter is meant that the mastery of one or more mazes
renders more difficult the mastery of a subsequent one.

2. Whether, if either positive or negative transfer is
present, it will persist through a series of mazes; and
whether, if both present and persistent, it will also be
cumulative.

3. The effect upon the learning of a maze of having
partially learned a preceding maze.

4. The effect of the complete learning of a maze upon
the mastery of another maze already partially learned.

5. Whether it is more advantageous to learn two mazes
in succession; or to partially learn one, then completely
learn the other, and finally perfect the learning of the
former.

6. The effect upon the relearning of a maze of having
learned four intervening mazes between the original learn-
ing of the maze and its relearning.

Diagrams of the mazes are presented in Plate I, figs. 1-5.
In all the mazes except C the blind alleys were twenty

1

RUTLEDGE T. WILTBANK
PLATE I

\ \ 3 <

rio. I
MOZE A

no. a

I

_; I , I I

s
■^ 1

3
I Z

riG 3
M/JZC C

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 3

inches long, four inches wide and five inches deep. In
the C maze they were thirteen inches long, four inches
wide and four inches deep. The mazes were built of half-
inch boards, the partitions between the runways in each
section being of metal, although the partitions between
the sections themselves were of wood. By a section is
here meant one of the four main divisions of the maze,
in which in all the mazes except C the blind-alleys run
in the same direction relative to the observer. By a run-
way is meant one length of the maze, such as a — b in
maze A, whether it forms part of the true pathway or
constitutes a blind alley. The terms, " true pathway "
and " blind alley," are open to no ambiguity. The mazes
were painted black without and within, and were covered
with glass tops.

The rats were between two and three months of age at
the beginning of the particular experiment in which they
were used. About one third of them were born in the
laboratory where the experiments were carried on, and the
remainder were obtained elsewhere. The rats were allotted
to the various groups in a thoroughly promiscuous way, so
that the groups approximated unselected collections. They
were fed in the food-box of the maze every day for a week
before they were introduced into the maze proper. They
were given one trial a day for the first five days, and after-
wards two trials a day, one in the forenoon and one in
the afternoon. They were fed for two minutes after the
trial in the forenoon and for five minutes after the trial
in the afternoon. The food was bread and milk with an
occasional allowance of sunflower seed. The rats were
transferred from one maze to another by groups. Those
individuals in a group which learned a maze earliest were
still given one trial a day upon the same maze until all
had learned it.

Records were kept of the number of trials necessary
to learn to run a maze four times out of five without an
error, of the errors made during the course of learning
and of the time required for each trial. The only errors
that were counted were those made by the rat running
into blind alleys while moving in the forward direction,

RUTLEDGE T. WILTBANK

i.e., along the true pathway toward the food-box. These
are the only errors which become fixed, and the actual
task of the rat, stated in negative terms, is to overcome
the tendencies to enter blind alleys while moving in the
forward direction. The time was taken with a stop-watch,
and was reckoned from the moment when the rat was
introduced into the maze until it reached the food-box.

The groups contained the following numbers of rats
and proportions of males and females:

Group 1
Group 2
Group 3
Group 4
Group 5

8 rats — 4 males, 4 females
10 rats — 5 males, 5 females
10 rats— 6 males, 4 females

9 rats- — 5 males, 4 females
10 rats — 6 males, 4 females

II. TRANSFER OF TRAINING BETWEEN PAIRS OF MAZES

The Positive Nature of the Transfer

An experiment to discover whether there was transfer
of training when white rats which had learned one maze
were carried over to another, and whether this transfer
if present was positive or negative, was performed by
Webb.i He effected transfers between ten pairs of mazes,
in five of which the first maze remained the same, while
the second maze in each pair was different from all the
rest; and in five of which the first maze in each pair was
different from all the rest and the second maze remained
the same; so that the order of the transfers was from
A to B, from A to C, from A to D, from A to E, from A
to F; and from B to A, from C to A, from D to A, from
E to A, from F to A. He found in every case that the
number of trials required, the total number of errors made
per rat and the total time consumed per rat were less for
a group learning a given maze as its second to be learned
than for the group that learned the given maze as its first.
He is of the opinion that the evidence furnished by his
experiments both with rats and with humans is strongly
in favor of the presence of positive transfer. There was
no accounting for these savings on the ground of chance
or of group-differences; and, moreover, Webb constructed
his mazes with the express purpose of rendering the condi-
tions as favorable as possible for negative transfer.

The present experiment was designed to further test
this matter of transfer as between pairs of mazes, using
mazes of another pattern and in a different arrangement
from those of Webb's. The results of this experiment
corroborate the results of Webb's with regard to the presence
and the nature of the transfer. The percentages of transfer
were computed on the same basis as Webb's — that of the

*Webb: Transfer of Training and Retroaction. Psych. Rev. Mon. Sup. Vol. 24
No. 3.

6 RUTLEDGE T. WILTBANK

average number of trials required to learn the maze, of
the total number of errors and the total time per rat.
The percentages are given in table 1.

TABLE 1

Percentages in Savings in Trials, Errors and Time in Transfers Between

Two Mazes

Trials

A to B 59

B to C 33

C to D 27

D to E 69

E to A 20

E to D 15

Tors

Time

88 ...

95

49 ...

84

54 ...

57

93 ...

89

65 ...

89

54 ...

49

Although the present experimenter did not try, as did
Webb, to devise mazes which would differ so much from
one another that the learning of one would exert a detri-
mental influence upon the learning of the other, neverthe-
less an examination of plate I, figs. 1 to 5, will disclose
that there were marked differences between certain of
them. Consider for example mazes B and C, and note
the differences in the number, length and arrangement of
the blind-alleys. In the first section of maze B, the animal
at the end of the first runway turns to the right, but in
the corresponding section of C it turns to the left. In
B immediately after this there is a blind alley to the right,
while in C there is none either to the right or left. After
the second turn in C, the animal encounters a blind alley
to the left; after the second turn in B, it follows, the true
pathway. At the fifth turn in C, the animal faces a situa-
tion different from any to be met in B, for not only is it
at this point in C far more likely to meet the true pathway
at a right angle, but there is a blind alley at its left im-
mediately as it turns into the true pathway, a similarly
placed one being lacking in B. An examination of the
other three sections of the mazes will multiply discrepan-
cies of this kind.

Again, consider the differences between A and E. In
the first section of the former, the true pathway consists
of three runways and the connecting runways, Q and R;

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 7

in the first section of E, there are five runways in the true
pathway, and there is only one connecting runway, for
at T, U and V the animal passes directly from one runway
into another, although instead of turning at about a right
angle to do so it must turn through 180 degrees. In the first
section of A, there are three blind alleys, and in the cor-
responding of E there is but one. In the second section
of E, the true pathway turns abruptly to the left at the
entrance to the section, while in A a blind alley opens at
this place. In the third section of E, two blind alleys are
placed side by side; in the third section of A the blind
alleys alternate. In the fourth section of E, the entering
animal turns into the true pathway to the left; in the cor-
responding section of A, there is a blind alley to the left.
But there was positive transfer as between these two mazes,
judged by all three criteria.

The Effect of Identical Parts

In order to discover whether the presence of identical
parts would affect the kind or the degree of transfer, the
first sections of mazes A and B were made alike, and also
the third sections of mazes D and E. The savings in
trials and errors in the transfers between these mazes with
identical parts were notably greater than in the transfers
between the mazes without identical parts, as may be
seen by reference to table 1. The saving in time in the
transfer between A and B was the largest of all the time-
savings, but that between D and E was equalled by that
between E and A.

If the greater positive transfet in the case of the mazes
with identical parts was due to the presence of the identi-
cal parts,— and it would seem too large to dismiss as merely
fortuitous,- — still it is not plain that this greater transfer
was brought about within the identical parts themselves.
It would seem reasonable to maintain that, if the presence
of the identical part in the second maze exerted a favor-
able influence upon the learning of the maze, the benefit
would be most likely to appear in this identical part,
although it is conceivable that the beneficial effect might
appear in some other part of the maze. For the purpose

Sec. 2

Sec. 3

Sec. 4

9.4
.9
90

13.9
.8
94

5.5
3.8
31

8 RUTLEDGE T. WILTBANK

of finding out whether there was a greater saving in this
identical part of maze B than in the other sections, the
distribution of errors within the four sections of maze B
made by the control-group and by the trial-group may
be compared. The average number of errors per rat made
by the control-group, group 2, in the four sections of B,
and the average number made by the trial-group, group 1,
which had already learned maze A, are given in table 2,
together with the percentages of savings made by group 1.

TABLE 2

The Average Number of Errors per Rat Made by Groups 2 and 1 in the
Sections of Maze B, and the Percentages of Savings

Sec. 1

Group 2 16 . 5

Group 1 1.7

Savings by Group 1 89

While group 1 effected a saving in the first section of
89 per cent, the savings in the second and third sections
were still larger. It cannot be argued on the prima-facie
evidence, therefore, that the saving in the identical part
was due to the ta... "^^ this identity, when the savings in
sections dissimilar from the corresponding sections of maze
A were even larger. It might be expected that much con-
fusion would arise at the entrance of the second section,
inasmuch as the animal had gained undoubtedly as much
momentum in the first section of the second maze as in
the corresponding section of the first maze, and at the
entrance to the second section of the maze it encounters
quite a different pattern. Contrary to any preconception,
however, the group effected as great a saving in the second
section as in the first. In connection with this fact, it
may be noted — as has been shown by Webb^ and is shown
also in one of the experiments now being reported^ — that
an animal brings over from the learning of one maze to
the learning of another some factors making for positive
transfer and some making for negative. There is, accord-

2 Webb: op. cit. p. 52.

3 Infra, p. 18, ff.

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 9

ing to the view adopted here as a result of the observations
made in the course of these experiments, a practise in error-
elimination which is useful and which an animal passing
from one maze-learning to another enjoys, so that this
animal may be expected to display a greater facility in the
the dropping of errors in the second maze than an animal
learning this maze as its first. But among the specific
habits, including entrance into some runways and the
avoidance of others, there will be some that aid the learning
and some that deter it; and an instance of the latter may
be found in the transfer now under consideration. The
group trained in the first maze had learned to pass a blind
alley on its left upon entering the fourth section and then
to turn into the next runway, which formed a part of the
true pathway. When transferred to the second maze, there
was no blind alley at the left immediately on entering this
section, so that no difficulty was experienced here; but
instead of the adjoining runway being a part of the true
pathway it was a blind alley, and as a consequence the
trial-group made 53 per cent of the total number of errors
in this section, where the control-group made but 14 per cent.

It is impossible to predict, with regard to the factors
making for positive and for negative transfer, how they
will interact in a given situation. It would seem that,
in the second of the two- mazes involved in the present
comparison, the tendency to turn into the blind alley in
the fourth section is so deeply implanted, as a result of
the learning of the former maze, that any general habit
of error-elimination is nearly if not completely suppressed;
while on the contrary there are no specific habits in the
second and third sections so strongly formed as to counter-
act the working of the general elimination activity, and
the results of this activity were so great that it effected
in these sections savings larger than in the identical part.

Turning to the other mazes which contained identical
parts, these mazes being the D and E and the parts the
third section, we find that positive transfer was evident
in the learning of the E maze by the group which had
first learned the D maze, and that the greatest saving
was effected in the identical part, although the saving in

10 RUTLEDGE T. WILTS ANK

the first section was nearly as great. The distribution of
errors in the various sections of the E maze, and the per-
centages of savings are given in table 3.

TABLE 3

The Average Number of Errors per Rat Made by Groups 5 and 4 in the
Sections of Maze E, and the Percentages of Savings

Sec. 1 Sec. 2 Sec. 3 Sec. 4

Group 5 (Control) 11.1 11.7 8.7 .4

Group 4 (Trial) 33 1.7 .11 .11

Savings by Grout) 4 97 85 99 72

The situation in maze E differs from the situation in B,
in that the identical part is not encountered in the former
until the animal has run through two sections which differ
from the first two of the maze which he has already learned;
so that in the situation presented in the B maze the ques-
tion was not only as to whether the animal would retain
the habit acquired in the first section of the maze in which
it was first run, but also as to whether there would be
any disconcerting effect from meeting the strange second
section after travelling the familiar first one; while in the
situation presented in the E maze the transition was from
a pathway unfamiliar at the beginning and through half
of its whole length to a section which had been learned
before. Although the saving in the third section was 99
per cent, this was but little more than the saving in the
first section. There would seem to be no ground for the
assertion that, if the presence of the identical part in the
two mazes results in the more expeditious learning of the
second maze, this is due to the saving of errors in the
identical part.

It might be contended that the comparison ought to
be made between the average number of errors and per-
centage of saving in the identical part and the average
errors and savings of the other three sections taken to-
gether, instead of being considered severally as above.
When this method is followed, it is found that the aver-
age number of errors per rat made in the dissimilar sections
of maze B was 1.8 as compared with the average of 1.7
in the identical part, and that the' average saving of the

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 11

dissimilar sections was 71.7 per cent as compared with
89 in the identical part. The same method applied to
maze E discloses that the average number of errors in the
dissimilar sections was .71 and the average saving 85 per
cent as compared with .11 and 99 in the identical part.
It may be remarked that the smaller average savings for
the three dissimilar sections were due in the case of both
mazes principally to the low records of the fourth section;
to which a defender of this method might reply that the
low gains in the fourth sections may have resulted from
the same cause as the high gains in the identical parts:
namely, the presence of the identical part, which exerted
more of a beneficial effect in the learning of other sections
than it did in the learning of the fourth. But we have
seen reason to believe that the smaller gain in the fourth
section of the B maze was due to a peculiarity of its pattern
in connection with the maze-habit acquired in A; and the
smaller gain in the fourth section of E as compared with
D may be explicable according to the same principle.

The Dependence of Transfer upon the Relative Difficulty

of the Two Mazes

The query is a logical one, whether as between mazes
of different degrees of difficulty it matters from the point
of view of savings if the more difficult or the less difficult
is learned first.

Two of the transfers in this experiment were from the
harder to the easier maze, judging by the records of the
control-groups; and four were from the easier to the harder.
In the case of one of the transfers from the harder to the
easier, the saving in errors was less than in any of the
transfers from an easier to a harder maze, the saving in
trials was less than in two of these transfers, and the saving
in time was also less than in two of these transfers, the
computations being on both the basis of the average total
number of errors and the average total time per rat and
the basis of the average errors and the average time per
trial. In the case of the other transfer from a harder
to an easier maze, the savings in trials and in errors were
greater than in any of the transfers from an easier to a

12 RUTLEDGE T. WILTBANK

harder maze, judging both by the average total errors and
the average total time per rat and by the average number
of errors and the average time per trial; and the saving
in time was less than the saving in one of the transfers
from an easier to a harder maze, according to the average
total errors and average total time per rat, and less than
the saving in two of these transfers, according to the average
number of errors and the average time per trial.

TABLE 4

Percentages of Savings in Transfers from Less Difficult to More Difficult
AND FROM More Difficult to Less Difficult Mazes, on the Basis of
THE Average Total Errors and Average Total Time per Rat

From the Less to the More Difificult

From the More to the Less Difficult

Trials Errors
From A to B . . . 59 88
From C to D... 27 54
FromEtoA... 20 65
From E to D... 15 54

Time
95
57
89
49

Trials Errors Time
From B to C... 33 49 84
From D to E... 69 93 89

TABLE 5

Percentages of Savings in Transfers from Less Difficult to More Difficult

AND FROM More Difficult to Less Difficult Mazes, on the Basis of the

Average Number of Errors and the Average Time per Trial

From the Less to the More Difficult From the More to the Less Difficult

Trials Errors Time Trials Errors Time

From A to B... 59 70 88 From B to C... 33 33 78

From C to D... 27 37 41 From D to E... 65 77 69
From E to A . . . 20 53 86
From E to D... 15 46 40

It will be seen from the figures in tables 4 and 5 that,
although the average savings are higher when the transfer
is from the harder to the easier maze, there is no ground
for the statement that these savings are in every instance
higher than when the transfer is from the easier to the
harder. The transfer from the easier to the harder may
be greater, as we may . note by comparing the transfer
between A and B with that between B and C.

III. TRANSFER OF TRAINING THROUGH SERIES OF MAZES

The Persistence of the Transfer

If we accept the statement that there is positive transfer
of training when rats are carried over from one maze com-
pletely learned to another which is then also completely-
learned, the question arises as to whether the transfer of
training will persist and retain its positive character if the
learning is continued through a series of mazes. It might
be presupposed with some plausibility that habits acquired
in situations which present marked peculiarities in the
midst of general resemblances, as do the situations in
the various mazes, would conflict with one another in such
a way that the transfer would cease to be positive. On the
other hand, it might be presupposed with perhaps equal
plausibility that in these situations practise tends toward
perfection, and the ability which an animal has to thread
the maze may consequently be expected to improve in
proportion to the number of mazes it learns to run.

It was to throw light upon this question that the present
experiment was undertaken. The five groups of rats were
allowed to continue their learning through series of mazes,
each group learning one series. The series were composed
of the same five mazes, but the mazes were learned in a
different order by each group, according to the scheme
presented in table 6, in which the letters represent mazes
and the numbers groups of rats.

TABLE 6

The Order in Which Various Series of Mazes Were Learned by Various

Groups

Mazes ABCDEABCD

Group 1

13

1

1

1

1

2

2

2

2

2

3

3

3

3

3

4

4

4

4

'4

5

5

5

5

14

RUTLEDGE T. WILTBANK

In pursuance of this method, there were twenty-five
maze-learnings, in twenty of which evidence might be
sought as to whether the positive transfer persists. Atten-
tion is drawn to the fact that the first group to learn a
given maze constituted a control-group for that maze. The
records of the groups upon these series of mazes are given
in tables 7 to 16. The average number of trials necessary
to learn a maze is given, and in connection with it the
mean variation of the individual rats from this average;
then the percentages of transfer for each maze are given,
these being computed by comparing the record of the
trial-group with that of the control-group in this particular
maze. The same method is followed with respect to errors
and time, and these are reckoned both on the basis of
the average total per rat and on that of the average per
trial. The percentages of transfer are shown graphically
in figures 6, 7, and 8.

TABLE 7

The Average Number of Trials per Rat in the Learning of a Series of Mazes

Maze A. .
Maze B. .
Maze C. .
Maze D. .
Maze E. .
Maze A. .
Maze B. .
Maze C. .
Maze D. .
Maze E. .

Group 1
26.62 ± 6.87
n.62± 5.28

9.75± 3.31
25.56 ±12.67

4.00

6.62 ± 3.56

Group 2

28.'70±' 6.66

12.20± 4.14

27.60 ±10.93

4.40± .56

6.40± 1.92

7.50± 3.91

Group 3

16.20±7.24

18.00±8.03

7.30± .49

6.50±2.32

13.20± .67

6.20±1.84

Group 4

24.67±7.24
7.11±3.88
6.89±3.23

11.44±3.06
9.00±2.64
8.89±3.96

Group 5

23.10±7.10
21.40±8.95
19.00±4.91

7.00±1.51
22.10±4.87

5.00± .83

TABLE 8

The Percentages of Transfer of Training in Trials through a Series of

Mazes

Group 1

Maze B. .

59

Maze C. .

39

Maze D. .

—3.58

Maze E. .

83

Maze A. .

75

Maze B. .

Maze C. .

Maze D. .

Maze E. .

Group 2

33

—12

81

76

74

Group 3

27
68
75
54
62

Group 4

69

74
60
44
64

Group 5

20
34
57
4.33

78

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 15

TABLE 9

The Average Total Number of Errors per Rat in the Learning of a Series

OF Mazes

Group 1
Maze A. . 48.45±13.52
Maze B. . 8.15± 2.52
Maze C. . 6.82± 2.42
Maze D. . 26.58±13.27
Maze E. . 0
Maze A.. 1.31 ± .91

Maze B

Maze C

Maze D

Maze E

Group 2 Group 3 Group 4

68.88±3a52 '.'.'.'.'.'.'.'.'.'.'. '.'.'.'.'.'.'.'.'.'.'.

14.50 ± 6.20 28.35±8.72

22.36±10.61 21.42±9.38 46.63±14.96

.31± .12 1.31± .95 2.20± 1.26

1.22± .70 1.75±1.20 2.07± .92

3.97 ± 1.30 6. 74 ±3. 29 9.27 db 3.55

1.92± .82 7.29± 4.45

4.53± 2.21

Group 5

31.88±12.27
16.69 ± 6.56
18.05 ± 5.99
3.64=h 1.29
23.87± 8.45
.90 db .72

TABLE 10

The Percentages of Transfer of Training, on the Basis of the Average
Total Number of Errors per Rat, in the Learning of a Series of Mazes

Group 1

Group 2

Group 3

Group 4

Group 5

Maze A. .

Maze B. .

88

Maze C. .

76

49

Maze D. .

43

52

54

Maze E. .

100

99

96

93

Maze A. .

97

97

96

96

65

Maze B. .

94

90

87

74

Maze C. .

93

74

87

Maze D. .

90

49

Maze E. .

97

TABLE 11
The Average Total Time per Rat in The Learning of a Series of Mazes

Group 1
Maze A. . 50.05 ±29.74
Maze B. . 4.65 ± 2.27
Maze C. . 5.75 ± 3.30
MazeD.. 9.20± 4.42
Maze E. . 1.56 ± .38
Maze A. . 1.39± .81

Maze B

Maze C

Maze D

Maze E

Group 2 Group 3 Group 4

93.85±4a36 '.'.'.'.'.'.['.'.'.'. '.'.'.'.'.'.'.'.'.'.[

12.71± 9.17 77.92±39.49

15.46 ± 8.41 16.20 ± 7.17 37.50 ±19.45

1.53± .34 3.36± 1.57 3.27± 1.79

1.60± 1.02 2.01 ± .80 1.93± .90

2.02± 1.51 4.22± 1.23 3.09± 1.04

7.56± 4.81 17.73±10.37

6.49± 5.98

Group 5

30.72 ±16.20
5.56 ± 1.48
4.75± 1.44
4.55 ± 1.38
5.52± 2.13
2.15± 1.05

16

RUTLEDGE T. WILTBANK

TABLE 12

The Percentages of Transfer of Training, on the Basis of the Average
Total Time per Rat, in the Learning of a Series of Mazes

Group 1

Maze A. .

Maze B. .

95

Maze C. .

93

Maze D. .

75

Maze E. .

95

Maze A. .

97

Maze B. .

Maze C. .

Maze D. .

Maze E. .

Group 2

84
59
94
97
96

Group 3 Group 4

57
89
97
96
90

89
96
97
77
83

Group 5

89
96
94
84
93

TABLE 13

The Average Number of Errors per Trial in the Learning of a Series of

Mazes

Maze A

Group 1
1.82±.33

.71 it. 30

.70±.35
1.04±.34
0

.17±.16

Group 2

Group 3

Group 4

Group 5

Maze B. .

2. 40 it. 40
1.18it.54
.81±.23
.07±.04
.19±.01
.53±.39

Maze C.

1.75it.40
1.19±.26
.18=t.21
.27±.17
.51=b:28
.31±.21

Maze D

1.89±.37
.31±.17
.30±.24
.81±.13
.81±.34
.51it.26

Maze E. .
Maze A. .
Maze B. .

1.38±.32
.78±.24
.95±.21

Maze C. .

.52db.25

Maze D. .

1.08±.30

Maze E. .

.18±.08

TABLE 141!

The Percentages of Transfer of Training, on the Basis of the Average
Number of Errors per Trial, in the Learning of a Series of Mazes

Group 1

Group 2 Group 3

Group 4 Group 5

Maze A. .

Maze B. .

70

Maze C. .

59

33

Maze D. .

45

58

37

Maze E. .

100

95

87

77

Maze A. .

91

90

80

84

53

Maze B. .

78

79

65

60

Maze C. .

82

54

71

Maze D. .

76

43

Maze E. .

87

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 17

TABLE 15
The Average Time per Trial in the Learning of a Series of Mazes

Maze A. .
Maze B. .
Maze C. .
MazeD..
Maze E. .
Maze A. .
Maze B. .
Maze C. .
Maze D. .
Maze E. .

Group 1
1.88±1.23
.40± .12
.59± .13
.36± .16
.39±. 10
.21± .04

Group 2

3.'27±2;i2
1.05± .68
.56± .40
.37± .09
.25± .12
.27± .05

Group 3

4.81±1.77

.90± .41

.46±

.31±

.32±
1.38±

,14
.06
.06
.97

Group 4

1.52± .77
.46± .19
.28± .08
.27± .05

1.97±1.62
.73± .67

Group 5

.33 ±.68
.26±.03
.25±.02
.65±.12
.25±.06
.43±.13

TABLE 16

The Percentages of Transfer of Training, on the Basis of the Average
Time per Trial, in the Learning of a Series of Mazes

Group 1

Group 2

Group 3

Group 4

Group 5

Maze A. .

Maze B. .

88

Maze C. .

88

78

Maze D. .

76

45

41

Maze E. .

71

73

65

65

Maze A. .

89

86

30

85

86

Maze B. .

92

90

92

90

Maze C. .

70

59

54

Maze D. .

52

83

Maze E. .

68

6

O

E

C

80

70
60
50

do

30
10

c
/o

y

^-.

■\^^.

J^^e Pet'ceata^e3 of Transfer
of Tra/ntnff in TrtaJs-fhrGi^
! a Series of Mirzes.

Oroup / —^ • — ' ■ ' '

Orm/p Z. —' ^

"Ordup 3 ■

18 RUTLEDGE T. WILTBANK

/OC7^

TO

V /

77?^ Percentages, ^f Ttiao^fer 12 f
7r<^in/ng on the 3^fs^ of the^
Average Totzj/ /yumber of Errors
per Raf:

Group I •

Groups. ■ —

O roup 3

Group 4

Group 5 —

D E

The Fercenta0e-3 of Transfer
60 ^\// of Tra/n/r^ on the Ba5/3 of

the A/en^^e Total T/me per ffat.

so

Group I

Group Z

O^oup 3

Group 4

Group 3

An examination of tables 7 to 16 will reveal the fact
that there was positive transfer manifest throughout the
various series of maze-learnings with the exception of the
learning of maze D by groups 1 and 2, as judged by the
criterion of trials, the former group requiring an average
of 3.58 per cent more trials than did the control-group
upon this maze, and the latter requiring 12 per cent more
(Tables 7 and 8).

In examining the records made by groups 1 and 2 in
maze D for the purpose of explaining these occurrences of
negative transfer, the outstanding fact brought to light is
that it is the number and persistence of errors made in
the first blind alley of maze D which account for the nega-

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 19

tive transfer. This fact comes out upon a comparison of
the records of the control-groups and trial-groups concerned.
Out of 420 errors made by the control-group in learning
the maze, 102 were made in the first blind alley, and the
rest were distributed among the other blind alleys. Out
of a total of 223 errors made by group 2 in this maze,
166 were made in the first blind alley; and out of a total
of 212 made by group 1, 126 were made in the first blind
alley. It was the large number of trials necessary to elim-
inate these errors made in the first blind alley which resulted
in the negative transfer, as may be seen from an examination
of tables 17 to 20, and figure 9, in which respectively the
number of errors and the percentage of errors made in
each blind alley by each group are shown.

TAI

5LE

17

The Errors Made by Group 4

(Control

IN Maze D

, Arranged According

TO

Blind Alleys

AND '

Trials

Blind Alleys

1

2

3

4

5

6

7

8

9

10

11

Trial 1

5

2

2

4

1

1

2

1

2

2

2

2

5

2

1

3

1

1

2

2

1

3

4

1

2

5

1

1

2

3

4

2

4

6

1

5

4

3

3

4

5

5

2

5

3

2

1

3

6

4

2

4

1

2

2

1

4

2

7

4

2

2

1

3

1

1

2

1

8

5
3

3
2

3
3

3

4

1
4

1

2
2

1
4

3
3

9

10

6

4

2

4

2

1

3

3

11

5

2

3

3

2

1

2

4

12

7

4

1

2

3

3

2

3

13

5

3

2

2

3

4

2

14

3

2

3

1

3

15

3

3

3

2

1

1

16

4

1

1

3

17

2

2

3

1

2

2

18

2

3

2

1

1

19

5

2

1

20

3

1

1

21

4

2

3

1

1

22

1

2

4

1

1

23

3

1

24

1

3

25

2

1

1

26

1

27

1

28

1

2

29

1

2

30

3

31

1

1

32

1

1

33

20

RUTLEDGE T. WILTBANK

TABLE 18

The Errors Made by Group 3 in Maze D, Arranged According to Blind

Alleys and Trials

Blind Alleys

1

2

3

4

5

6 7

8

9

10

11

Trial 1

.. 14

2

4

2

3 2

1

i

i

2

.. 14

1

1

1

1

3

7

1

4

4

.. 6

1

1

2

1

1

2

2

5

.. 4

2

2

2

1

6

4

1

1

1

1

1

1

7

4

1

4

8

7

1

2

3

9

.. 5

1

3

1

1

10

.. 4

3

2

11

.. 3

12

5

4

2

2

13

4

2

14

4

2

15

4

1

1

16

.. 3

1

17

.. 3

1

18

2

19

2

1

20

^

i

21

1

22

1

23

1

24

25

26

27

28

29

30

31

32

33

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 21

TABLE 19

The Errors Made by Group 2 in Maze D, Arranged According to Blind

Alleys and Trials

Blind Alleys

1

2

3

4 5

6 7 8

9

10

11

Trial 1

4

1

2 1

4

b

1

2

6

1

3

2

3

1

3

8

1

2

4

8

1

2

1

5

.. 8

1

1

6

8

1

2

1

7

.. 7

8

8

1

9

6

1

1

10

.. 9

1

1

1

1

11

4

1

12

1

1

13

7

1

14,. .

6

1

15

6

1

16

.. 5

17

4

18

.. 4

19

5

20

3

21

3

22

5

23

4

24

2

25

3

26

1

27

2

28

.. 3

29

2

30

.. 3

31

1

32

.. 2

33

2

34

.. 2

35

1

36

2

37

.. 2

38

.. 2

39

.. 2

40

2

41

.. 2

42

1

43

.. 0

44

1

22

RUTLEDGE T. WILTBANK

TABLE 20

The Errors Made by Group 1 in Maze D, According to Blind Alleys and

Trials

Blind Alleys

1

2 3

4

5

Trial 1

.. 9

3

1

2

.. 9

2

1

3

.. 5

1

4

.. 5

1

5

.. 2

1

6

.. 6

1

1

7

.. 5

1

1

1

8

5

2

1

9

4

10

5

1

11

4

1

12

>/

1

13

4

14

4

15

^.

16

. 2

17

.. 2

18

2

19

.. 3

20

.. 2

21

.. 2

22

1

23

.. 3

24

1

25

2

26

.. 3

27

.. 2

28

.. 2

29

.. 2

30

.. 2

31

2

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

7

8

9

10

11

2

1

7

2

1

2

2

5

2

3

4

4

2

4

1

3
3

2

2

1

1

2

1

2
3

1

1
1

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 23

TABLE 20— (Continued)

Blind Alleys 1234 56789 10 11

Trial 54

55 1

56 1

57 1

58

59 1

^345 6789/0//

\ The Percentages of fj-ror^ M<a<^e /r? the

7D,^ B/Ind A//ey3 of the D M<^ze p^

f^v Groi^ 4 (Control)

' 1 Group Z ■■

50-li Or^up I —- — - — .

30 \\

k %

It is apparent that these instances of negative transfer
cannot be treated as if the transfer in question were simply
that between the C maze and the D maze. The groups
exhibiting this negative transfer had learned — one the A
and B mazes before the C, and the other the B maze before
the C. The fact that, as between the C and the D mazes
when no other maze had been learned before C, there was
27 per cent of positive transfer, while the transfer was
negative when maze B or mazes A and B were learned
before C, would suggest that the final element of the ex-
planation is to be sought in the learning prior to that of
the C maze.

The increase of more than 100 per cent in the errors
made by group 3 in blind alley 1 of the D maze, as com-
pared with the errors made by the control-group, would
seem to have been due to the facts that group 3 came to
the maze with the habit already formed of traversing a
runway of similar construction to those confronting it on

24 RUTLEDGE T. WILTBANK

entering D, and that the behavior of the animals on being
placed within the entrance of D accounts for their frequent
entrances into blind alley 1. The former of these facts
needs no special comment, for the connection between
running forward through a series of runways and obtaining
food at the end of the journey must be implanted in the
animal's neural mechanism by the close of the first maze-
learning. Why the animals entered the blind alley so
frequently, however, is not to be explained solely on the
ground of the dispositions which they brought with them
to the maze, but also on the ground of their behavior
within the maze itself. It was the almost invariable be-
havior of the rats, — invariable both as respects the number
of rats concerned and the number of trials through which
it continued, — to turn immediately on being introduced
into the maze and sniff at the door. This brought their
bodies into a position in which the likelihood of their
entering blind alley 1 was much greater than that of
their entering the true pathway, as may be seen by re-
ferring to the diagram of the maze, Plate I, figure 4. When
they were in this position, it required movement through
only a few degrees to the right in order to bring their
heads into line with the blind alle^^ or even into the blind
alley. In any case, it was a situation similar to that which
in the C maze called forth the first of the motor habits
eventuating in the animals' obtaining food, and the animals
plunged into the blind alley with great speed and great
force. After sniffing at the door, it was necessary for them,
if they were to enter the true pathway, either to pass close
to the blind alley by turning to the right, without being
overcome by its solicitation; or else by turning to the
left pass through an arc of more than 180 degrees. It is
plain that, once the animals turned to the door, their
chance random movements would favor decidedly the
selection of the blind alley.

The question remains to be dealt with, why this tendency
to enter the blind alley should have been so greatly exag-
gerated in the cases of groups 1 and 2 — so greatly, in fact,
as to be the cause of the negative transfer. If 52.8 per
cent of the errors made by group 3 were made in blind

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 25

alley 1, 60 per cent of those made by group 1 and 74
per cent of those made by group 2 were made in the same
blind alley. All the conditions under which the three
groups learned the maze were the same, and the only
difference among the groups was in the number of mazes
they had learned before being placed in D. Group 3
had learned one maze, group 2 had learned two mazes,
and group 1 had learned three. It would seem therefore
that the difference in the records of group 3 on the one
hand and of groups 1 and 2 on the other must have been
due to some factor or factors acquired prior to the learning
of the C maze. Reference to the patterns of the mazes
concerned will furnish a clue as to the nature of this factor.
In maze C the true pathway proceeds directly from the
entrance without necessitating any change in the orientation
of an animal that was introduced straight through the en-
trance; but in mazes A and B the true pathway turns
to the left, demanding a turn to the left in an animal so
introduced. The marked tendency of the rats in groups
1 and 2 to turn into blind alley 1 in maze D would seem
accordingly to come from a habit brought to maze C,
leading them to seek the true pathway in the same position
which the blind alley occupies in maze D.

It would be reasonable to affirm that, on the basis of
the principles adduced to explain the cases of negative
transfer, we should look for negative transfer in the group
which learned mazes E, A, B and C before learning the
E maze, for in the E maze as well as in mazes A and B
the true pathway lies to the left of the entrance. It would
seem equally reasonable to expect that the tendency to
take the first runway to the left of the entrance in maze
D would be more pronounced in group 1 than in group 2,
and more pronounced in group 5 than in either 1 or 2,
since group 5 learned three mazes in which the true path-
way takes this course, group 1 learned two, and group 2
learned one. Contrary to these assumptions, we find that
the transfer of group 5 on maze D is positive. But it is the
least of all the positive transfers, being only 4.33 per cent;
and when we examine the distribution of errors according
to the blind alleys and the trials, we see the same tendencies

26 RUTLEDGE T. WILTBANK

at work as in the case of groups 1 and 2. Of the 239 errors
made by the group, 137 were made in the first blind alley,
or 56.9 per cent of the total number; and the errors per-
sisted there as with the other groups. In the case of this
group, however, the tendencies were not quite sufficient
to turn the tide of positive transfer. This reversal of such
justifiable expectations as those above might be due to
an increase in the number of deviations from the rule of
turning to sniff at the door as the rats learned an increasing
number of mazes. That there was an increase in these
exceptions cannot be asserted, for no record of the number
of sniffings was kept; but it would seem probable that the
futility of such behavior would lead to its becoming less
frequent. It would not take many deviations of this
sort from the usual behavior to account for the difference
between the 3.58 per cent of negative transfer shown
by group 1 and the 4.33 per cent of positive transfer shown
by group 5.

The negative instances dealt with above indicate cer-
tainly that, when three mazes are learned in succession,
the transfer is not a function of the last two mazes alone,
but of all three; and they probably indicate that, when
four or five are learned in succession, the transfer is not a
function of the last two alone, but of all the mazes involved.
They also prove — it needs hardly to be remarked — that,
although transfer is generally positive, there may be a
particular kinaesthetic habit occurring in a particular maze-
situation so as to produce negative transfer.

The Question as to Whether the Transfer is Cumulative

The experiment demonstrates that the transfer as between
the first and second mazes of a series is positive, and that
the transfer as between any other two contiguous mazes
of the series is also positive, judging by all the three criteria
employed, except in the two negative results in trials.
The query which next presents itself is whether this gen-
erally persistent positive transfer increases with each addi-
tional maze-learning, or in other words whether it is
cumulative.

There was no series in which the transfer as estimated by

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 27

any one of the criteria was cumulative throughout the series,
but there were some instances where there was cumulation
for three maze-learnings and one where there was cumula-
tion for four. These are given in table 21.

TABLE 21

CuMtJLATivE Savings in Average Trials, in Average Errors per Trial and
IN Average Time per Trial

Savings in Trials

Group Mazes Percentages of Transfer

in Mazes Named

3 C to D, D to E, E to A 27, 68, 75
5 E to A, A to B, B to C 20, 34, 57

Savings in Errors
2 B to C, C to D, D to E 33, 58, 95

5 E to A, A to B, B to C 53, 60, 71

Savings in Time
2 CtoD.DtoE, EtoA,AtoB 45, 73, 86, 92

4 D to E, E to A, A to B 65, 85, 92

Over against these five instances of cumulation for three
mazes and one of cumulation for four, there were eight
instances in which the transfer decreased in amount through
three maze-learnings. Three of these were in trials, four
in errors and one in time. They are presented in table 22.

TABLE 22
Successive Decreases in Transfer of Training

According to Average Trials

Group Mazes Percentages of Transfer

in Mazes Named

1 A to B, B to C, C to D 59, 39, —3.58

2 D to E, E to A, A to B 81, 76, 74
4 E to A, A to B, B to C 74, 60, 44

According to Average Errors per Trial

1 A to B, B to C, C to D 70, 59, 45

2 D to E, E to A, A to B 95, 90, 78

3 D to E, E to A, A to B 87, 80, 79

4 E to A, A to B, B to C 84, 65, 54

According to Average Time per Trial
1 B to C, C to D. D to E 88, 76, 71

28 RUTLEDGE T. WILTBANK

When the errors and the time were computed on the
basis of the average total per rat, there were six instances
in errors and time taken together where the transfer was
cumulative for three maze-learnings, and six where it de-
creased for three maze-learnings in succession. In the
case of the errors, the mazes upon which these cumulations
and decreases occurred were the same, whether the errors
were computed according to one or the other method,
except that according to the average per trial there was
one more series than according to the other method. But
in the case of time, cumulation appeared upon the same
two series of three mazes as it did according to the average
time per trial and also upon two additional series of three
mazes. There was no coincidence in the decreases in time
as reckoned by the two methods.

There is no evidence therefore that this generally positive
and persistent transfer is also cumulative invariably and
continuously; nor, it may be added, that there is a regular
and continuous decrease.

The Effect of Preceding Maze- Learnings upon the Learning
of a Particular Maze

In the foregoing discussion, the data are treated from
the point of view of the different groups as they proceed
to learn the series of mazes, and we have followed the
various groups and examined their records to see whether
these records show a persistence and a cumulation of
positive transfer. But the same data can be considered
from another point of view — that of a particular maze,
where we can take our position and inquire if this maze
can be learned more readily by a group that has had two
previous learnings than by a group that has had but one,
and if by a group that has had three more readily than by a
group that has had two, and so on. The number of trials
required by the various groups to learn the various mazes
is given in table 23, and figure 10. Disregarding the desig-
nations of the groups, which can be obtained by reference
to table 6, p. 13, they are treated merely as the groups
which were the first, second, third, fourth or fifth to learn
a certain maze.

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 29

TABLE 23

The Average Number of Trials per Rat Required in Learning a Maze by
Groups Having Previously Learned One, Two, Three, or Four Mazes

Maze A Maze B Maze C Maze D Maze E

Group having learned one pre-
ceding maze 21.40 11.62 12.20 18.00 7.11

Group having learned two pre-
ceding mazes 6.89 19.00 9.75 27.60 7 30

Group having learned three pre-
ceding mazes 6.50 11.44 7.00 25.56 4.40

Group having learned four pre-
ceding mazes 6.40 13.20 9.00 22.10 4.00

■^ 3 6 O £

30

7?;e /7i^e'rz?g& Ai<rr7^/^ cf 7?/a/5 /:?er ^z- /rhoi/y're'c:^ /r?
^/7€> or A/orv Afi^:z-<^3.

♦ " " 7>W " "

" * " -T TTynee " "

* " " f^tyr- " " V

The figures in table 23 show that the groups which have
learned more than one maze before attacking another do
not necessarily and invariably enjoy any advantage in
the learning of the new maze, as compared with the trial-
record of the group that came to the maze with but one
prior maze-learning. The groups learning maze A showed
a nominal increase of savings with the increase in the num-
ber of previous maze-learnings, but the group that had
had four maze-learnings saved but .1 of a trial as compared
with the one having had three, and the one that had had
three saved but .4 of a trial as compared with the one
having had two. If the group with two previous maze-

30 RUTLEDGE T. WILTBANK

learnings showed a large saving relative to the group with
one, this is deprived of any significance by the fact that,
in the case of the other four mazes, two of them proved
easier for the group with one prior learning than for the
group with two. A comparison of the groups with three
prior learnings and the groups with four discloses the fact
that in two instances — namely, in mazes B and C, the
groups with three prior learnings learned the maze more
readily than the groups with four; and, if there were
three instances where the groups with four prior learnings
did better than those with three, one of these was but
.1 of a trial better. This table might accordingly be brought
forward in support of the statement that three maze-
learnings are as good as four, measured by the beneficial
effect upon the learning of a subsequent maze according
to the standard of trials. But the actual differences in
these figures representing the number of trials necessary
to learn the third and fourth mazes are not large enough
to give much weight to such a statement.

The D maze presents the least evidence of any beneficial
effect accruing from the learning of preceding mazes, and
the principle which is to be invoked in explanation is the
one used with reference to the negative figure representing
the transfer in trials, as may be seen on pp. 18 ff. The num-
ber of trials needed by the groups coming to this maze with
two and with three prior learnings was greater than the
number made by the control-group, and the number needed
by the group coming with four learnings was also high.
These figures were all greater than any in the A and B
mazes, the control-figures of which show them to be some-

TABLE 24

The Rankings of the Groups upon the Mazes, According to the Average
Number of Trials per Rat

Maze A Maze B Maze C Maze D Maze E Av. Rank
Group having learned

four preceding mazes 1 3 2 2 1 1.8

Group having learned

^Aree preceding mazes 2 1 1 3 2 1.8

Group having learned

two preceding mazes 3 4 3 4 4 3.6

Group having learned

one preceding maze .4 2 4 1 3 2.8

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 31

what more difficult mazes. It is the position of blind
alley 1 which accounts for this. When the animals came
from the C maze, they found themselves in a situation
which in the maze or mazes before C had called forth
the habit of turning into a runway to the left on entering
the maze; and this position in the D maze being occupied
by a blind alley which called forth this habit, the learning
of the maze was consequently retarded.

The evidence that the average total number of errors
per rat tends to decrease as the number of prior maze-
learnings increases is more convincing than any figures
furnished by the trial criterion in favor of a similar benefit
from the latter point of view, as may be seen from an
examination of table 25 and figure 11.

TABLE 25

The Average Total Number of Errors per Rat by Groups Having Previously
Learned One or More Mazes

Maze A Maze B Maze C Maze D Maze E
Group having learned one pre-
ceding maze 16.69 8.25 15.40 21.42 2.20

Group having learned two pre-
ceding mazes 2.07 18.05 6.82 22.36 1.31

Group having learned three pre-
ceding mazes 1.75 9.27 3.64 26.58 .31

Group having learned four pre-
ceding mazes 1.22 6.73 7.29 23.87 00

so
^ 3 C D jf

v\

o

The /9i^3r<ape '^/<u/ A/cmher- cf £?-r(Pr^ /?<?/- /Phf J?y

(proilp /^i^/'/?p /.etofrnea/ C^ne /-^£>ce(pfjng Maze

"/^ " " Two " "

* " ' Three " "

^ /' > /^^^ " "

32

RUTLEDGE T. WILTBANK

In two mazes, A and E, the average total number of
errors per rat decreased with the increase in the number
of mazes previously learned, and these decreases were not
slight ones, proportionately speaking. If the various
groups are ranked upon the mazes according to the average
total errors made by the rats of the group, it will be found
that the average ranking of the groups having had four
prior learnings is highest, the average total errors of the
rats in these groups taken together being least; and the
groups having had three, two and one prior learnings
follow in order. This may be seen in table 26.

TABLE 26

The Rankings of the Groups upon the Mazes, According to the Average
Total Number of Errors per Rat

Maze A Maze B Maze C Maze D Maze E Av. Rank
Group having learned

jour preceding mazes
Group having learned

three preceding mazes
Group having learned

two preceding mazes
Group having learned

one preceding maze .4 3 4 1 4 3.2

1

1

3

3

1

1.8

2

2

1

4

2

2.2

3

4

2

2

3

2 8

According to the average total time per rat, there was
only one maze in which this time decreased with the in-
crease in the number of previous maze learnings by groups

B

D

jS

2a

T/?e /^i^era^e 7^/a/ ^me y:yer /4S?/ J?y <$r-i:Jcr£?^

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 33

coming to the maze — maze A. When the groups were

ranked upon the various mazes, the groups with four and

those with three previous learnings had the same average

ranking, these having required the least total time per rat;

and the groups with two and with one previous learning

followed. In tables 27 and 28 are given respectively the

average total time per rat and the rankings on this basis.

The records of the different groups are shown graphically

in figure 12.

TABLE 27

The Average Total Time per Rat by Groups Having Previously Learned

One or More Mazes

Maze A Maze B Maze C Maze D Maze E

Group having learned one pre-
ceding maze 5.56 4.65 12.81 16.20 3.27

Group having learned two pre-
ceding mazes 1.93 4.75 3.51 15.46 3.36

Group having learned three pre-
ceding mazes 2.01 3.09 4.55 8.20 1.63

Group having learned four pre-
ceding mazes 1.60 4.22 17.73 5.52 1.56

TABLE 28

The Rankings of the Groups upon the Mazes, According to the Average

Total Time per Rat

Maze A Maze B Maze G Maze D Maze E Av. Rank

Group having learned

/owr preceding mazes 1 2 3 3 1 2.0

Group having learned

//zree preceding mazes 2 1 1 4 2 2.0

Group having learned

two preceding mazes 3 4 2 2 3 2.8

Group having learned

one preceding maze .4 3 4 1 4 3.2

The rankings according to the average errors per trial
and the average time per trial are given in tables 29 and 31.

TABLE 29

The Rankings of the Groups upon the Mazes, According to the Average

Errors per Trial

Maze A Maze B Maze C Maze D Maze E Av. Rank
Group having learned

/owr preceding mazes 1 1 3 3 1 1.8

Group having learned

//zreg preceding mazes 2 3 1 2 2 2.0

Group having learned

two preceding mazes 3 4 2 1 3 2.6

Group having learned

one preceding maze .4 2 4 4 4 3.6

34 RUTLEDGE T. WILTBANK

TABLE 30

The Ranking of the Groups upon the Mazes, According to the Average

Time per Trial

Maze A Maze B Maze C Maze D Maze E Av. Rank
Group having learned

/o?<r preceding mazes 1 3 4 1 2 2.2

Group having learned

/Aree preceding mazes 4 2 2 2 1 2.2

Group having learned

two preceding mazes 3 1 1 3 3 2.2

Group having learned

one preceding maze .2 4 3 4 3 3.2

It is apparent from tables 26 to 30 that, as in the case
of trials so in that of errors and time, an increase in the
number of preceding maze-learnings does not necessarily
and invariably result in a decrease in errors or in time.
But that there is a tendency for the errors to decrease
with the increase in the number of maze-learnings may
be seen from the average rankings.

The Theory oj Transfer

The foregoing paragraphs show that the degree of trans-
fer is not a function of the serial position of the maze.
It appears to be a function of the maze to which the transfer
is made, irrespective of the position of this maze. Thus
maze E, whatever its position, tends to be the one in which
according to all the criteria the greatest savings are effected.
The degree of transfer may be a function also of the maze
or mazes from which the transfer is made. This became
apparent in our search for the causes underlying the in-
stances of negative transfer in trials on maze D. The
transfer is not a function of the difficulty of the maze, for
the E maze — to use it again as an example — was not the
least difficult maze, as measured by the records of original
mastery; and the B maze, which was the most difficult
with respect to original mastery does not appear as the
one in which the least savings were made.

In attempting an explanation of the presence and per-
sistence of positive transfer, we may assume that the
animal having learned one or more mazes and being in-
troduced into a new one is to be looked upon as an organism

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 35

which has acquired some tendencies making for positive
transfer and some making for negative. The kind and
the degree of transfer may be viewed as the result of the
interaction of these tendencies with the new situation.

When the hungry animal is introduced into the maze,
its behavior indicates that it is controlled by the instincts
of curiosity and timidity, the former being undoubtedly
quickened by hunger. The instinct of curiosity accounts
for the explorations of the animal, in the course of which
it sooner or later reaches the food-box; and the instinct
of timidity accounts in the main for the retreating move-
ments, which are so frequent in the early stages of the
learning. It would be baseless to say that there is never
a retreating movement due to curiosity, or that curiosity
never enters as a factor into the induction of such move-
ments; nevertheless, the behavior of the animal when
advancing through the maze in the earlier stages of the
learning would seem explicable only on the ground that
it is under the control of a combination of curosity and
timidity, with curiosity in the ascendency. Its change
of behavior into that of retreating is so precipitate and
its movements back toward the food-box are so direct
and rapid that it would seem due to timidity's gaining
the upper hand over curiosity and coming to dominate
the animal. The early cessation of these retreating move-
ments in the course of the maze-learning would seem
to indicate that the instinct of timidity soon recedes far
into the background and becomes almost if not quite
inactive. That this recession of the instinct of timidity
carries over from one maze to another is apparent from
the fewer retreating movements and their prompter ces-
sation in a second maze than in the first. This recession
of the instinct of timidity is a factor making for positive
transfer.

Included among the kinaesthetic habits which the animal
brings to the new maze, there are some tending to lead
it into the true pathway and some tending to lead it into
blind alleys. Since it is generally the errors made during
the first few trials which become fixed and constitute
the hindrance to be overcome in order to learn the maze,

36 RUTLEDGE T. WILTBANK

it is the tendencies driving the animals into these bHnd
alleys which are to be looked upon as the chief factor
accounting for negative transfer. Conversely, the kinaes-
thetic habits acquired while traversing the true pathway
under the influence of the exploring instinct of curiosity
constitute one of the factors making for positive transfer.

The association which is formed between the animal's
running through the maze to the end of the true pathway
and its gaining access to the food-box is also a factor making
for positive transfer. The only purpose for which the
animals are taken from the cage before they begin their
trials in the maze is to feed them, and before their first
trial they have been taken out daily for a week and fed
in the food-box of the maze. When they are placed in
the maze for the first time, their behavior denotes the
frustration of their food-securing movements and help-
lessness to deal with the new situation; but they need
to pass through the maze only a few time and to gain
ingress to the food-box every time before they will eagerly
attack the problem presented to them by the maze, which
is that of reaching its end and the food there as rapidly
and easily as possible. What at first was a mere m-
pediment to the'r obtaining food becomes consequently
a part of the process of obtaining food, which begins
with their being taken from the cage and ends with their
entering the food-box. This association is no doubt formed
by the end of the first maze-learning, and must affect
favorably the promptness and persistence with which
they attack the next maze, irrespective of the elements
which affect unfavorably the learning of the second maze.

The practise in error-elimination which an animal has
had while learning one maze would seem to exert a beneficial
influence upon the learning of another. The blind-alley
situation is one that is repeated in maze after maze, and
calls for the same reaction in every instance, which is an
about-face and retreat. This situation is the one unam-
biguous situation between the entrance and the exit of
the maze, and the uniformity and repetition of response
would seem to constitute what may properly be called
practise in error-elimination; and we have encountered

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 37

evidence that uniformity of response would tend to appear
in a uniform situation, in whatever maze this situation
might occur.

The tendency to turn into blind-alleys which have been
eliminated remains after the animal's last entrance into
them, for not only do the animals occasionally reenter
these alleys once or twice after they have been apparently
avoided for good, but the swaying inward of their bodies
as they pass certain blind-alleys into which they once
regularly entered, sometimes bringing their bodies into
contact with the far corner of the alley, would point to
the same conclusion. The successful running of the maze
must therefore include the overcoming of these tendencies
to enter blind alleys, and this practise in error-elimination
with its attendant resistance to the blind alleys once fre-
quently entered cannot but leave its impress upon the
reactions of the organism in any similar situation. In
other words, the animal comes from its first maze-learning
with a maze-habit formed, an integral part of which is
this error-elimination activity including the habitual re-
sistance to certain runways once frequently entered.

In seeking an explanation of the positive transfer of
training, it would seem necessary to take into account
this practise in error-elimination, while at the same time
being careful not to read into the expression, " error-
elimination," more than is implied in the discussion above.
The rat which is learning its second maze forms habits
under the combined influence of its instinctive reactions
and the environment in which they are made, just as was
the case in its first maze-learning; and under the spell
of its exploring instinct it forms in the same manner habits
of running into blind alleys. Its problem, from our point
of view and little as the rat itself may realize it, is to
eliminate these blind alleys, so as to reach the food-box
as speedily as possible. The advantage which the rat
that has learned at least one maze enjoys as compared
with a rat that is attacking the same problem without
having learned another maze beforehand is several-fold.
The retarding influence of the instinct of timidity is not
felt as strongly nor does it continue as long, the best evi-

38 RUTLEDGE T. WILTBANK

dence for this fact being furnished by the large saving
in the group's time of learning the second maze, as com-
pared with the time of the control-group. In addition
to this, the association between running the maze and the
procurement of food is firmly fixed at the beginning of
the maze-learning in the case of the rat that has learned
a preceding maze, while this connection has to be formed
by the inexperienced group in the course of the learning.
Still again, and perhaps most important of all, is the prac-
tise in error-elimination. This does not secure the animal
against entering blind alleys nor forming habits of doing
so, neither does it assist the animal in finding the true
pathway; but it enables the animal that has had this
practise to eliminate more promptly its entrances into
blind alleys.

The Correlations among the Records of Transfer through a
Series of Mazes

It was found that a positive correlation existed among
the savings in trials and errors, in trials and time, and in
errors and time, effected by the various groups as they
learned the various series of mazes; the savings being
estimated on the basis of the average number of trials,
the average total errors and the average total time per rat.
A positive correlation existed among the savings when
they were estimated on the basis of the average errors
and the average time per trial, except in the case of that
between trials and errors in the third group and that
between trials and time, and errors and time, in the fifth
group, in which cases it was negative. On this basis of
estimation there were also two .cases of zero correlation,
those of errors and time in the second and in the third
groups. The method of rank-differences was used in de-
termining these correlations, and the two sets of figures are
given in tables 31 and 32.

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 39

TABLE 31

The Correlations among the Records of Transfer through a Series of
Mazes, on the Basis of the Average Total Errors and Average Total

Time per Rat

Group Series Learned

1 A to B, B to C, C to D, D to E

2 B to C, C to D, D to E, E to A

3 C to D, D to E, E to A, A to B

4 D to E, E to A, A to B, B to C

5 E to A, A to B, B to C, C to D

TABLE 32

The Correlations among the Records of Transfer through a Series of
Mazes, on the Basis of the Average Errors and the Average Time per Trial

Trials

Trials

Errors

and Errors

and

Time

and Time

1.00

.70

.70

.80

.80

.60

.70

.80

.50

1.00

.40

.40

1.00

.80

.80

Group

Trials and Errors

Trials and Time

Errors and Time

1

1.00

.40

.40

2

.80

.40

00

3

.80

— .40

00

4

1.00

.40

.40

5

1.00

— .20

— .20

IV. TRANSFER OF TRAINING BETWEEN A MAZE PARTIALLY
LEARNED AND ONE COMPLETELY LEARNED

We have seen that there is always transfer of training
when both mazes of a pair are completely learned, and
that the transfer is positive. The inquiry may be made
as to whether there is transfer of training when the first
maze of the pair is only partially learned and the second
completely learned.

In order to answer this inquiry, groups of rats were
transferred from one maze to another after two, four,
eight, or sixteen trials in the first of the two mazes, and
also after the complete mastery of the first maze. The
groups consisted of from seven to twelve rats, with an
average of nine. The mazes used were the ones in which
the greatest degree and the least of positive transfer of
training had occurred in the last experiment considering
the average of all the groups — mazes E and D respectively.
The results are presented in tables 33 to 3S, and graphically
in figures 13 and 14.

TABLE 33

The Records of Groups Learning the D Maze after the Partial Learning

OF the E Maze

After After After After After complete

2 trials in E 4 trials 8 trials 16 trials mastery

Av. Trials. . . . 25.29±2.90 28.43±6.20 26.10±7.46 16.20±7.46 20.86±4.00
Av. Errors per

Trial 2.09 ± .31 1.94± .77 2.64 ± .63 1.07± .40 1.03 ± .35

Av. Time per

Trial 38± .08 .32± .06 .31± .03 .41± .17 .91di .25

TABLE 34

The Percentages of Transfer of Training in Groups Learning the D Maze

AFTER the Partial Learning of the E Maze, on the Basis of the

Average Errors and the Average Time per Trial

After After After After After complete

2 trials in E 4 trials 8 trials 16 trials mastery

Trials —3.73 —15.24 —5.81 34.63 15,44

Errors —10.60 —2.65 —39.84 43.50 44.97

Time 75 11 78.76 79.62 72.82 40.13

40

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 41

TABLE 35

The Percentages of Transfer of Training in Groups Learning the D Maze
After the Partial Learning of the E Maze, on the Basis of the Aver-
age Total Number of Errors and the Average Total Time per Rat

After After After After After complete

2 trials in E 4 trials 8 trials 16 trials learning

Errors 13.36 —18.27 —48,18 63.01 53,83

Time 74.37 75.73 78.40 82.37 93.07

TABLE 36

The Records of Groups Learning the E Maze after the Partial Learning

OF THE D Maze

After After After After After complete

2 trials in D 4 trials 8 trials 16 trials learning

Av. Trials. ... 28.33 ±6.80 23.86±4.98 23.80±6.06 12.20±6.60 7.11±3.81
Av. Errors per

Trial 2.05 ± .90 1.35± .44 1.99± .81 .68± .27 .31±.17

Av. Time per

Trial 33± .05 .28± .03 .33± .06 .38± .14 .46± .1

TABLE 37
The Percentages of Transfer of Training in Groups Learning the E Maze

AFTER THE PARTIAL LEARNING OF THE D MaZE, ON THE BASIS OF THE

Average Errors and the Average Time per Trial

After After After After After complete

2 trials in D 4 trials 8 trials 16 trials learning

Trials —22.51 —11.72 —3.03 47.19 69.22

Errors —48.55 —2.17 —44.19 51.01 77.69

Time 75.11 78.95 75.11 71.36 65.41

TABLE 38

The Percentages of Transfer of Training in Groups Learning the E Maze

AFTER the Partial Learning of the D Maze, on the Basis of the Average

Total Number of Errors and the Average Total Time per Rat

After After After After After complete

2 trials in D 4 trials 8 trials 16 trials learning

Errors 79.02 —1.07 —44.36 73.96 93.09

Time 69.56 77.93 76.56 84,89 89.36

The most notable fact disclosed in tables 33 to 3S is
that the transfer of training is negative, according to the
trial and error-criteria in all the changes from one maze
to another previous to the change after 16 trials. This
is in accord with our study of the factors affecting transfer

42

RUTLEDGE T. WILTBANK

(see pp. 34 ff.)- As was there pointed out, it is the errors
made in the blind alleys during the finst few trials which
become fixed and constitute the chief obstacle which the
animal has to overcome in order to learn the maze. The
transfer is negative prior to the change of mazes after the
sixteenth trial for the reason that the factors making for
positive transfer, and particularly the practise in error-

ZTria/s in S'

lifter
4 Tria/s

/Jfter
f Tria/s

t6 rrtefs

^

/ifter Com/t/efB
Learnirrg o^ £

Percentages of Transfer of Thf/r?/n^ trf Orotps Leafnin^
the D A^^ze after the Partj^a/ Leofrning of tiie € Aiaze
on the Bas/s of the /lyera^e Total /dumber of Zrrors
ancf the /^yenff^e TotisJ T/me per Pat

TrjaJs

Errors __ . ^

Time . — — ■ "

/V<S,/5?

elimination, do not gain the ascendency over the negative
tendencies until the latter stage of the learning process.
Concerning the factors making for positive transfer, we
may assert with confidence that the kinaesthetic habits
tending to impel the animal into the blind alleys are formed
at the beginning of the learning-process, and that the
practise in error-elimination does not exert its beneficent
effect until the latter part of the process; but we cannot

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 43

affirm with equal confidence at what part of the process
the overcoming of fear and the association between running
to the end of the maze and obtaining food begin to tell.
With regard to the transfer-effect according to the
time-criterion, it will be observed that the degree of transfer
did not increase regularly with the increase in the number
of trials in the first maze of the pair, except in the average

2 Tr/als in E

4 Tna/s

/fffer
STnals

/Ifter
/6 Tria/3

/9fter Comp/efe
Leafn/na of £

PercenftPT^e^ of Transfer
of Tra/nin^ in Groups
Learning the f Maze
after the fhrtial Learning
of the O Maze • on the
Bas/s of the A^/era^e yiDta/
Number of Errors ang^ the
Average Total Time per Rat.

Tr/a's

Errors

Firne

total time per rat of the group carried over from E to D.
But a regular increase of this kind was brought to light
when the transfer was reckoned on the basis of the average
surplus time required by the rats to run the mazes. This
surplus time was obtained by subtracting from the average
time per trial within a group the average time of the last
trial made by the rats of the group that completely learned
the same maze.

44 RUTLEDGE T. WILTBANK

TABLE 39

The Percentages of Transfer of Training as Measured by the Savings
IN Surplus Time by Groups Learning the D Maze after the
Partial Learning of the E Maze

After 2 trials on E 81 After 8 trials on E 86

After 4 trials on E 85 After 16 trials on E 88

After learning E 93

TABLE 40

The Percentages of Transfer of Training as Measured by the Savings

IN Surplus Time by Groups Learning the E Maze after

THE Partial Learning of the D Maze

After 2 trials on D 76 After 8 trials on D 80 . 7

After 4 trials on D 80 . 6 After 16 trials on D 92

After learning D 96

It is not surprising that the transfer in time should be
positive throughout the entire series of transfers, for the
first two trials are almost w^ithout exception the greatest
time-consumers. The animals during these trials, and
especially during the first, proceed very slowly and with
every sign of extreme timidity. It rarely takes more than
one or two successful passages of the maze for the animals'
behavior in this respect to radically change, so that they
proceed to their task with much greater speed. It is prob-
ably this increase in speed which, continuing through the
four-trial partial learning and the eight-trial, secures slight
increases in transfer of training.

V. TRANSFER OF TRAINING BETWEEN A MAZE COMPLETELY
LEARNED AND ONE ALREADY PARTIALLY LEARNED

We have just considered the transfer of training as
between a maze partially learned and a maze in which
the rat has never run and which is completely mastered.
Now we shall consider transfer of training as between a
maze completely mastered and one which has been partially
learned before the transfer is effected. The plan was to
give two, four, eight or sixteen trials on D, then master E,
and then return to D and complete its learning. A similar
process was carried through beginning with E, then master-
ing D, and then completing the learning of E.

In the absence of any such data as are presented in the
experiments herein described and by other experiments
in the same general field, it might have been contended
with some reason that the learning of the new maze would
blot out the effects of the partial learning of the old, so
that the animals would have to rebegin it. But, con-
sidering the evidence set forth above which is strongly
in support of the positive transfer of training, the con-
trary presupposition might be formed, that the effect of
such an interpolated maze-learning would prove beneficial.
It might seem accordingly, on the one hand, a serious
hindrance to the acquirement of a habit to break off
the formation of that habit at a certain point in order
to acquire completely a different habit before resuming
practise upon the former. On the other hand, we have
found ample evidence to believe that the beneficial effects
of learning one maze can be carried over to the learning
of another, and consequently we should expect that the
animal will return to the learning of the original maze
with some results of his training in the new maze which
will prove available in the old. It becomes therefore a
question of relative advantage and disadvantage, so that
it may be asked with regard to the completion of the learn-

45

46 RUTLEDGE T. WILTBANK

ing of the maze already partially learned: Does the ad-
vantage of the experience in the new maze, consisting
in the formation of the association between the running
of the maze and the satisfaction of hunger and in practise
in error-elimination, counterbalance the disadvantage of
bringing in another habit to confuse a habit in process
of formation, especially when as in this case the habit
which is being formed is the first maze-habit of the organism?

The groups which were used in the foregoing experiment
in transfer after partial learning were carried back, after
they had learned the second maze, to the mazes in which
they had had two, four, eight and sixteen trials respectively,
and were allowed to complete the learning. This furnished
eight groups of rats, four groups of which while learning
D were interrupted at these various points and taught E,
after which they were returned to D and continued their
learning of it; and four groups of which were trained
according to the same plan, except that they began in E,
were taught D, and then returned to E.

The effect from the point of view of trials was beneficial
in all the transfers, except that in which sixteen trials
were given in the first maze. This may be explained
on the ground that in the last-named instance the habit
was nearly formed, and much of the benefit of the sixteen
trials were stored up — so to speak — in it; and when the
new maze-learning was effected the disadvantage which
it brought by inculcating particular kinaesthetic habits
which conflicted with the nearly perfected maze-habit out-
weighed the advantage gained through the additional
practise afforded by the new maze-learning in those activ-
ities which are beneficial in a general way irrespective of
the design of the particular maze. The work which the
group had done — to express it somewhat differently — by
means of its sixteen trials was to a large extent undone
by the new habit, and it took a number of trials to bring
the group up to the efficiency it had attained before the
interruption. While the new habit afforded exercise for
the factors making for positive transfer, still not enough
trials were saved as a consequence of this additional prac-
tise to make up for the trials lost through the conflict

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 47

of the habit completely formed in the second maze with
the habit nearly formed in the first maze. Where the
transfer was made after two, four or eight trials, the forma-
tion of the maze-habit had not progressed nearly so far,
and a larger number of errors were being made than were
being made by the sixteen-trial group when it was trans-
ferred. The consequence was that with these groups
any interference with the imperfect maze-habit would not
render futile, to all intents and purposes, many of the
trials which had already been made, and that furthermore
any interference with the imperfect habit, including a
relatively large number of errors, would be not unlikely
to correct an error instead of diverting the animal from
the true pathway.

Table 41 presents the number of trials necessary to
learn the maze — the trials before the interruption plus
the trials after the interruption, and also the percentages
of savings and losses. The savings and losses were com-
puted by comparing the records thus obtained with the
records of the control-group for the particular maze, and
besides being shown in the table just named, they are
presented graphically in figs. 15 and 16. Accordingly,
the total number of trials in D before and after the learning
of E were compared with the number of trials in D before

TABLE 41
The Average Number of Trials Required in a Maze Partially Learned

BEFORE THE COMPLETE LEARNING OF A SECOND MaZE AND COMPLETELY

Learned after the Learning of the Second Maze; and the Per-
centages OF Transfer of Training

Maze D Maze E

Trials Percentages of Transfer Trials Percentages of Transfer

(Two Trials in Partial Learning)
17. 9± 7.8 27.36 8.43±2.20 63.51

(Four Trials in Partial Learning)
16.43± 4.61 33.39 15.70± .42 32.06

(Eight Trials in Partial Learning)
23.00± 5.23 6.77 14.40±2.24 37.66

(Sixteen Trials in Partial Learning)
29 .1± 4.70 —17.91 24.00±4.00 —3.90

(Control-groups: Group 4 for D, Group 5 for E)
24.67±12.67 23.10±7.10

48 RUTLEDGE T. WILTBANK

TABLE 42

The Average Total Number of Errors per Rat Made in a Maze Partially

Learned before the Complete Learning of a Second Maze and Completely

Learned after the Learning of the Second Maze; and the Percentages

OF Transfer of Training

Maze D Maze E

Errors Percentages of Transfer Errors Percentages of Transfer

(Two Trials in Partial Learning)
33.11 29.24 5.81 81.71

(Four Trials in Partial Learning)

37.62 12.89 24.81 —22.18
(Eight Trials in Partial Learning)

80.27 —72.14 29.81 6.18

(Sixteen Trials in Partial Learning)
80.61 —72.88 36.00 —12.92

(Control-groups: Group 4 for D, Group 5 for E)

46.63 31.88

TABLE 43

The Average Total Time per Rat Made in a Maze Partially Learned before
the Complete Learning of a Second Maze and Completely Learned
after the Learning of the Second Maze; and the Per-
centages OF Transfer of Training

Maze D Maze E

Time Percentages of Transfer Time Percentages of Transfer

(Two Trials in Partial Learning)
22.91 44.24 32.52 — 5.86

(Four Trials in Partial Learning
19.55 50.55 51.32 —67.06

(Eight Trials in Partial Learning)
30.36 19.04 56.16 —82.81

(Sixteen Trials in Partial Learning)
20.08 46.46 30.49 .75

(Control-groups: Group 4 for D, Group 5 for E)
37.50 30.72

and after the learning of E w^ere compared v^ith the number
made by the control-group for D, which v^as group 4;
and the same method v^as followed with respect to errors
and time.

We cannot make, from the point of view of time, as
broad a statement concerning the beneficial effect upon a
maze-learning of interrupting it in order to learn another
maze as was made concerning the same matter from the
point of view of trials. The transfer in the E maze, with

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 49

^ 2Tr/a/s

■4 Trials

8 Tnah /6 Tr/a/s

50

.— -

"X

,

x

X

40

X

y'

^^"^^C

—

—

\

x

s,

X
y

x

zo ^^^

\

N/

"^^-^

\

"^

\

lO

\

\

0

\

\

10

\

\

zo

30

\

40
SO
60
70

The Percentages of Tnarnsfer of Ir^^/n/n^ in
fhe D Ma^e Fart/<^f/y Learned before the
Complete Learn /n^ of the E M^i^fze (!^nL<i
Completely Learne'dl /:;ffter the Learn /n^
of fhe 3econc:^ Ma^e.

"Tri^^ls

Lrror3 J^/G /S

respect to time, was negative in three of the four changes
from D to E. But there is nothing in this negative transfer
to invalidate the findings of the experiments reported here
and those reported elsewhere as to the generally positive
nature of transfer, for this negative transfer was due to
the fact that the animals made large time-records in the
trials before the interruption, and then on being returned
to the maze completed the learning in comparatively few

50

RUTLEDGE T. WILTBANK

fO

Z7'r-/c3/s

4 Tna/s

/ Tr/'a/s

/6 Tr/ah

^^. Percentages of Transfer of 7r^jn/n^ /n the

E Maze Fart/al/i/ Learne^:^ before the
^■^ ^ Learning of the D Maze an^^ com/:>/ete/y
Learne/s^ after tt?e Learn tn^ of the
^ecopi^ Maze

50
<-0
30

zo

IC
0
I0\

ZO

30
40
50
60
70

eo

<J0

60 \

T~r/a/5

Errors —

\

\

/=7<^, /6

trials, relative to the records of the control-group. Tb«
result was that, although the time-records of the trials
following the return to the partially learned maze wa^
low, still they were too few to counteract the high records
made in the trials before the interruption; a large time-
total therefore had to be divided by a small number of
trials. For example, the average time per trial made by
the group that was taken from the E maze after two trials

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 51

and brought back to E after having learned D was for
the two trials 13.32 minutes, and for the remaining trials
necessary to learn E but .28 of a minute; but the average
number of trials necessary to complete the learning of
E was but 6.43.

The negative transfer in errors after sixteen trials would
seem to be due to the fact that a nearly perfect habit is
interrupted in order to learn another maze; and the dis-
concerting nature of such a proceeding is denoted by the
increased number of trials necessary to learn the maze,
as compared with the control-group and by the increased
number of errors per trial. It would seem that this prin-
ciple is to be invoked to explain the negative transfer
also when the partial learning consisted of eight trials.

The only criterion logically applicable here, when the
two entire group-performances are compared with each
other, is the trial-criterion, for as has been just indicated
the favorable effect of practise as it is seen in the error and
time-records of the group in the trials following its return
to the maze whose learning has been interrupted is swamped
by the high average made during the trials preceding the
interruption. Accordingly, when we compare the records
of the control-group with regard to errors and time with
the records of the group whose learning is interrupted, we
are not bringing together two learning processes which are
analogous in these respects. The transfer in errors and
in time, however, may be estimated by bringing into com-
parison the performances of the rats in the trial-group
upon their return to the first maze after having learned
the second with the performances of the rats in the control-
group for the first maze, beginning with that trial of the
control-group which corresponded with the first trial made
by the trial-group upon its return. Thus in the case of
the interruption after four trials, the first trial of the re-
turning group after having learned the intervening maze
would be its fifth trial upon the former maze; and the
transfer which it has effected from the maze learned during
the interruption may be estimated with reference to the
record of the control-group beginning with its fifth trial
and continuing to the end of the learning. In table 46

52

RUTLEDGE T. WILTBANK

these records of the trial-group and of the control-group
are placed side by side.

TABLE 44

The Records Made upon the Partially Learned Maze after the Resump-
tion OF Learning, Compared with Records of Control-Group for
Same Part of Learning
Maze D Maze E

Trial-Group Control-Group Trial-Group Control-Group

(Two Trials in Partial Learning)

Trials. ... 15. 92 ±6. 06 22.67 .09 6.43zt2.20 21.10±7.14

Errors.... 1.35± .48 1.66 .40 36± .31 1.32± .32

Time 22± .03 .71 .48 28± .04 .43± .17

(Four Trials in Partial Learning)

20.67 7.09 11.71±6,10 19.10±7.06

1.43 .42 61± .22 l.Udb .30

.38 .18 26± .04 .32± .06

(Eight Trials in Partial Learning)

16.67±7.26 6.40±3.16 15.10±6.98

l.lOi .64 20± .32 .89± .22

.44± .26 25± .03 .30± .04

(Sixteen Trials in Partial Learning)

13.10±4.70 8.67±6.14 8.00±4.00 7.10±6.70

1.06± .51 .65± .23 30± .02 .40± .23

.21± .02 .38± .18 20± .01 .23± .03

Trials.
Errors.
Time. . ,

Trials.
Errors.
Time. .

Trials.
Errors.
Time. . -

12.43± .56

1.31d= ,29

.24± .02

15.00±6.75

1.55± .64

.24± .05

All the records of the groups whose learning was in-
terrupted showed a saving in that part of the learning
which followed the return to the maze already partially
learned, except the error-record of the group whose learning
was broken into after the eighth trial in the D maze, the
trial and error-records of the group whose learning was
interrupted after the sixteenth in the D maze, and the
trial-record of the group whose learning was interrupted
after the sixteenth trial in the E maze.

The figures in table 44 help to bring out the fact, which
has been already stated, that the negative result in the
case of time, when the learning of the maze before and
after the interruption is compared with the record of the
control-group, is due to the small time-records made in
the part of the maze-learning succeeding the interruption
not being numerous enough to counterbalance the large
time-records for the trials before the interruption.

These figures make plain the basis for the similar observa-

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 53

tion concerning the error-records, although in this case
there are two exceptions. In the record of the group whose
learning of the D maze was interrupted after the sixteenth
trial, the negative transfer in errors in the part of the
learning succeeding the interruption is to be ascribed to
the same cause which produced the negative transfer in
trials. In the part of the learning of the E maze that
followed the interruption after the sixteenth trial, there
was a small negative transfer in trials, but the causes
producing it did not suffice to produce negative transfer
in errors, although they did bring about the closest ap-
proach between the average errors of the trial and control-
groups for any of the interrupted learnings in the E maze.
As a part of the present inquiry, the average error and
time-records of the trial-group for the first three trials
after its return to the maze already partially learned
were compared with the records of the control-group for
the corresponding trials. For example, the records for
these first three trials after the resumption of learning
by the group whose learning of the maze had been inter-
rupted after the second trial were compared with the
records of the control-group for the third, fourth and
fifth trials upon the same maze. The figures setting forth
this comparison are presented in table 45.

TABLE 45

The Average Errors and the Average Time per Trial for the First Three

Trials after the Return to the Maze Already Partially Learned,

Compared with the Corresponding Records of the Control-Group

Maze D Maze E

Trial -Group Control-Group Trial -Group Control-Group

(Two Trials in Partial Learning)

3.29 67 2.30

1.30 34 1.01

(Four Trials in Partial Learning)

2.52 81 2.00

.89 30 .39

(Eight Trials in Partial Learning)

2.00 41 1.57

.50 28 .37

(Sixteen Trials in Partial Learning)

.86 50 1.10

.52 23 .29

Errors. . . .
Time

2.89
.26

Errors....
Time

2.52

.27

Errors

Time

2.90
.28

Errors

Time

2.10
.25

54 RUTLEDGE T. WILTBANK

The trial-groups all effected a large saving in the first
three trials upon the resumption of the E maze. It may
be said, in other words, that while they were learning
the D maze following their two, four, eight or sixteen
trials on E, they were continuing the learning of E, for
on their return they showed both an improvement up-
on their earlier performances in E and a marked saving
as compared with the control-group. Even more, some
of the rats on their return to E ran that maze four times
in succession without an error. Of those whose learning
had been interrupted after two trials, three out of the
seven rats composing the group made a perfect score upon
their return to E, and none of these three had made an
errorless run on either of their two former trials in E.
Of those whose learning had been interrupted after four
trials, one rat out of seven made a perfect score, so far as
errors were concerned, and another made but one error
in five trials; and neither of these had made an errorless
run in its four previous trials. Of those whose learning
had been interrupted after eight trials, six out of ten made
a perfect score, and one made a single error in five trials;
and among these was one rat which had made one errorless
run, which was its first trial, and another which had made
two errorless runs, on its fifth and sixth trials respectively.
Of those interrupted after the sixteenth trial, two out
of eight made a perfect score, one of these having made
one errorless run, on its second trial, and the other no
errorless runs.

If the general statement is justified that, while the rats
of all four groups were learning D after having begun to
learn E, they were continuing their learning of E at the
same time they were learning D, there is equal justifi-
cation for the statement that certain of these rats while
being taught D completed their learning of E. It would
be tempting to believe that these rats making errorless
runs on their return to E had learned to recognize a blind
alley at sight, if both their behavior in the maze and also
the results of the experiments which have been made
upon the rats' sensory control in the maze did not preclude
such a belief. The present experiment does not furnish

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 55

the data for the solution of the problem, and any attempt
at an explanation here would be but the advancement of
conjecture.

The first three trials of the groups returning to the D
maze after they have learned the E maze do not all show-
positive transfer, as is the case in the E maze after having
learned the D. The group that had had two preliminary
trials in D showed a saving both in errors and in time;
but the group with four preliminary trials showed the
same average number of errors as did the control-group,
while those with eight and sixteen preliminary trials re-
spectively made more errors than the control-group. With
all four groups there was a constant saving in time.

VI. THE LEARNING OF TWO MAZES WHEN THEY ARE LEARNED

ONE AFTER ANOTHER COMPARED WITH THE LEARNING

OF THE SAME MAZES WHEN ONE IS PARTIALLY

LEARNED, THEN THE OTHER COMPLETELY

LEARNED, AND FINALLY THE LEARNING OF

THE FORMER COMPLETED

After the learning of the first maze had been completed,
as described in the last section, records were at hand giving
the total number of trials, of errors and the time required
for the mastery of the two mazes. These records enable
us to compare the records of learning two mazes when
these mazes are learned one after the other with the records

TABLE 46

The Learning of Mazes D and E Together: The Average Trials, and the
Average Errors and the Average Time per Trial

1. Learning E and then D (Control)
Total for E and D

Trials 43.72±3.17

Errors 1.78± .20

Time 1.19± .26

3. Learning E after 2 trials on D, then
completing the learning of D

Trials 46.23±9.46

Errors 2.03± .58

Time 1.17± .64

5. Learning E after 4 trials on D, then
completing the learning of D

Trials 40.29±4.64

Errors 2.25±1.37

Time 1.14± .61

7. Learning E after 8 trials on D, then
completing the learning of D

Trials 46.80±8.33

Errors 2.96± .40

Time 79± .35

9. Learning E after 16 trials on D,
then completing the learning of D

Trials 41.30±8.30

Errors 2.12± .27

Time 66± .31

2. Learning D and then E (Control)
Total for D and E

Trials 31.78±5.57

Errors 1.52 ±.32

Time 1.22± .48

4. Learning D after 2 trials on E, then
completing the learning of E

Trials 33.72±3.59

Errors 1.39± .73

Time 1.17± .32

6. Learning D after 4 trials on E, then
completing the learning of E

Trials 44.13±4.71

Errors 1.19± .32

Time 1.17±1.32

8. Learning D after 8 trials on E, then
completing the learning of E

Trials 40.50±7.33

Errors 2.53± .44

Time 1.73± .66

10. Learning D after 16 trials on E.
then completing the learning of E

Trials 40.12±3.75

Errors 1.55± .25

Time 96± .49

56

TRMNING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 57

when one is partially learned, then the other completely
learned and finally the learning of the former completed.
The group which after learning the D maze as its first
was carried over to the E maze in the course of the first
experiment described above, having to do with the per-
sistence and cumulation of transfer of training, served as
the control-group for these mazes in this order. A special
group of seven rats was trained in E and then transferred
to D, giving a control for the learning of these mazes in
this order.

The figures of the control-group are given in table 46,
and after them the figures of the other groups learning the
mazes according to the method indicated in connection
with each group.

From the control-figures — those numbered 1 and 2
above — it will be apparent that the two mazes were mas-
tered with less effort when the more difficult maze, judging
according to the degree of transfer, which was D, was
learned first. But the advantage of learning the more
difficult maze first was counteracted if before learning it
four trials were made upon the easier maze, to which the
transfer was to be made. Moreover, when the order:
sixteen trials in a maze, the learning of another maze, the
completion of the learning of the maze in which the sixteen
trials were made, — when this order was followed it mattered
little whether the trials began in the more difficult or less
difficult maze. The sixteen trials when made upon the
more difficult maze resulted in a large saving in the easier
maze, but this saving was counterbalanced by the com-
paratively great number of trials needed upon the return
to the harder maze in order to carry its learning to com-
pletion. When the sixteen trials were made in the easier
maze, the beneficial effect upon the learning of the second
maze was far less marked, but this disadvantage was fully
compensated for by the less effort needed to complete
the easier maze.

If the series numbered 2, 4, 6, 8, 10 is viewed as one in
which the order of the learning is from the D maze to the
E maze, except that after the first maze-learning of the
series the order is modified by the giving of a certain num-

58 RUTLEDGE T. WILTBANK

ber of trials in the second maze, E, before the first is learned,
then it will be observed that no maze-learning in this series
subsequent to the first equalled the first, as far as the
trial-criterion was concerned, while learnings 4 and 6 bet-
tered it judging by the error-criterion, and 6 and 10 bettered
it by the time-criterion. If the series 1, 3, 5, 7, 9 is viewed
in the same way, as if the order E to D were the fundamen-
tal one and the learnings subsequent to it but modifications
of it, we find that learnings 5 and 9 excel the first as
judged by the trial-criterion; and that, although none
of the after learnings equal the first from the point of
view of errors, all surpass it from the point of view of
time. It may be said therefore that there is no one method
among all the arrangements, taking the D-to-E series and
the E-to-D together, which enjoys signal preeminence.
If it is said that the learning of the harder maze before
the easier without any previous trials in the easier results
in the best record according to trials, the learning numbered
4 is a very close second to it, and there are better arrange-
ments according to the error and time-criteria. Never-
theless, there is an average advantage in learning the
harder maze before learning the easier, from the point of
view of trials and errors, as may be seen through the series
in which the learning of D and then of E. is viewed as
fundamental. When the time-criterion is applied, the
series in which the passage from the easier to the harder
maze is viewed as fundamental enjoys a striking advantage.
These facts may be seen in table 47.

TABLE 47

The Average Records of the D-to-E and of the E-to-D Series

D-to-E Series E-to-D Series

Trials 37.77±4.01 43.60±2.25

Errors 1.64=b .21 2.03± .41

Time 1.44±.20 81±.21

VII. RELEARNING A MAZE AFTER HAVING LEARNED FOUR
INTERVENING MAZES

It was possible, by carrying the groups of rats which
had learned a series of mazes back to the first maze learned,
to ascertain whether any beneficial effect of the first maze-
learning appeared in the relearning of that maze after
four intervening maze-learnings. Thus group 1, which
learned the series from A to E, was brought back and
retaught A; and group 2, which learned the series from
B to A, was brought back and retaught B. It was con-
sidered that, if any beneficial effect of the original learning
remained, the maze relearned would be mastered with
less effort than any maze of equal or greater difficulty
among the intervening four. If, conversely, one or more
mazes of at least equal difficulty were learned more readily
than the first maze was relearned, this would be accepted
as evidence that the first maze, now being relearned, was
practically speaking a new maze to the rats. The records
of the relearnings may be found in tables 7, 9, 11, pp. 14 ff.

Looking at the relearnings from the trial point of view,
we see that the A maze was relearned by its group more
easily than any one of the intervening mazes was learned,
except the E maze. This E maze, however, was an easier
maze, according to the records of the control-group; al-
though the ease of its mastery by group 1, as compared
with the relearning of A was proportionately much greater
than the difference in the difficulty attending the original
mastery of the two mazes by the control-groups. The
E maze was but 13 per cent easier as judged by the trial-
criterion of the groups last named, but in the series of
learnings which we are now considering the E maze was
learned with 40 per cent less effort, by the same criterion.
The group relearning the A maze saved 75 per cent as
compared with the record of original mastery, but this
same group learned E with a saving of 83 per cent as com-

59

60 RUTLEDGE T. WILTBANK

pared with the control-group. This fact would seem to
render unsafe the statement that the first group enjoyed
any advantage in the relearning of A over and above the
other mazes.

The B maze was relearned by the second group with an
average of 7.5 trials, but the E and A mazes were learned
with an average of 4.4 and 6.4 trials respectively. These
mazes were less difficult than the B maze, which was the
most difficult of all from the point of view of original
mastery; but in these instances, as in the one just con-
sidered, the saving in the learning of the mazes as compared
with the relearning of B was much greater than the records
of the control-groups in these mazes as compared with
the control-group in B. The A maze was 7.0 per cent
easier according to the records of the control-group, but
15 per cent as shown by the group learning it in the series
now being considered when this group's records in A and
B are compared. The E maze was 20 per cent easier
according to the records of original mastery, but it was
41 per cent easier in this connection. If the mazes A and
E were as difficult as B, it would be defensible to say that
no benefit from having learned B remained through the
four intervening learnings and reappeared upon the re-
learning of B, since the group failed to evince any superiority
in the relearning of B relative to the learning of an equally
difficult maze. As it was, the group relearning B enjoyed
a saving of 75 per cent as against the record of the control-
group for A; and in learning E it saved 81 per cent, as
compared with the control-grou-p for E. Its saving in the
B maze, as compared with its own record of original mas-
tery, was 74 per cent. It is unwarrantable to see in the trial-
record any survival of the benefit of the former learning of B.

The third group relearned the C maze in 6.2 trials.
Although there was no learning of any intervening maze
in fewer trials, the A maze was learned in nearly the same
number — 6.5 trials. The C maze was 39 per cent easier
according to the control-groups, but its relearning was
but 4.6 easier than the learning of A. The group relearned
C with 62 per cent less effort than was needed in the original
mastery, while it learned A with 76 per cent fewer trials
than the control-group.

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 61

The fourth group learned the A maze, which was more
difficult than the D maze, in fewer trials than were needed
in the D maze, its saving in the A maze being 74 per cent
and in the D maze 64. The fifth group failed to learn
any of the intervening mazes in fewer trials than the re-
learning of the first maze called for, the nearest approach
to the five trials in the relearning being the seven trials
in C; but the C maze was much easier than the E. The
fifth group was the only one that furnished any evidence
in favor of the retention of the habit through a series
of four mazes. With the other four groups, the weight
of the evidence inclines to the other side.

When we consider error-records, we find that here also
the weight of the evidence is against the retention of a
habit through a series of this number of mazes. The third
and fourth groups both learned mazes of greater difficulty
than the original maze with fewer average total errors
per rat than they made in the relearning of the original
maze. The first group made the low average total of
1.31 errors per rat in relearning its first maze; but it made
no errors in learning E. The second group made an average
3.97 per rat while relearning B, but it learned A with an
average of 1.22 and E with an average of .31. It saved
94 per cent in errors as compared with its former learning
of B, but it made 69 per cent fewer errors in A than in B,
although A was but 29 per cent easier. It learned E with
92 per cent fewer errors than it made in A, although E
was but 39 per cent easier. Coming to the fifth group,
it will be found that its error-record in E was far better
than its error-record in any of the other mazes of the
series; and with regard to errors, as well as to trials and
time, this group is the only one of the five that furnishes
evidence of the retention of the beneficial effect of the
original learning through four intervening learnings.

Looking at the matter from the point of view of the time-
criterion, we find that the third and fourth groups learned
one or more mazes of greater difficulty than the first of
the series with a smaller time-average than the relearning
of the first necessitated; and that the second group learned
A, which was 7.8 per cent easier than B, in 37 per cent
less time than was needed in B. With the first and fifth

62 RUTLEDGE T. WILTBANK

groups, there was no maze-learning that rivalled the re-
learning of the first maze, as far as the time-criterion was
concerned. The evidence furnished by the time-criterion,
accordingly, is more favorable than that furnished by the
other criteria to the continuance of beneficial effect of the
first learning throughout the series.

When the records are surveyed together, it is seen that
only one group showed, according to all three criteria,
signs of the retention of the beneficial effects of the habit
formed in the first learning of the first maze. If the first
and second groups showed evidences of the retention of
the beneficial effect in time, they showed no unmistakable
evidence of such benefit in either trials or errors. The
most that can be said therefore is that the favorable results
of having learned a maze do not always appear upon the
relearning of it after the learning of four intervening mazes,
irrespective of the patterns and order of the intervening
mazes; but that there may be such patterns and such an
order as taken together will result in a saving, according
to one, two or three criteria, although within the limits
of the present experiment there was but one series upon
which the beneficial effect was seen as judged by all three
criteria. It cannot be determined on the basis of these
results whether, in those cases where the maze on being
relearned was, to all intents and purposes, a new one,
the loss of the benefit was due to the disintegration of the
benefit was due to the disintegration of the habit through
the lapse of time since its original learning, or to the in-
terference of the habits acquired since the first; or if
due to both, in what proportion to each.

VIII. SUMMARY OF RESULTS

The evidence furnished by these experiments supports
the position that the learning of one or more mazes is
usually beneficial with respect to the learning of another
following them immediately, as determined by the fewer
trials required and lesser time and smaller number of errors,
compared with those of the control-group. There were no
instances of negative transfer in either errors or time,
but there were two instances of negative transfer in trials.

When two adjacent mazes contained identical parts,
the more expeditious learning of the second was not due
to the saving of errors in the identical part. Whether the
identical part exerted a favorable influence which was
effective in the saving of errors elsewhere than in the
identical part is a problem that these experiments did not
attempt to solve.

Although the average savings are higher when the trans-
fer is from the more difficult to the less difficult maze,
there is no ground for the statement that the transfer
from the harder to the easier maze is in every instance
higher than from the easier to the harder. The transfer
from the easier to the harder may be greater.

The transfer of training was found to be not only positive
but persistent, except in two instances of negative transfer
in trials, throughout the learning of five series of mazes by
a different group in each series. Although it was thus
persistent, it was not cumulative; that is to say, it did
not increase regularly with the increase in the number
of previous maze learnings. There was no case in which
the transfer, as determined by any one of the three criteria,
increased throughout an entire series; and if there were
a number of instances in which there was an increase for
three maze-learnings, there were more successive decreases
for three mazes than there were successive increases for
three.

63

64 RUTLEDGE T. WILTBANK

Looking at the matter from the point of view of the
various mazes, rather than from that of the various groups,
it was found that as far as the trial-criterion is concerned
increase in the number of previous maze-learnings did not
always result in an increase of positive transfer. According
to the error-criterion, in the case of but two mazes did the
favorable effect of prior learnings increase regularly.
There was no maze in which the favorable effect according
to the time-criterion increased regularly with the increase
in the number of prior learnings.

It was found that a positive correlation existed between
the savings in trials and errors, in trials and time, and in
errors and time, though a series of mazes, estimated on
the basis of the average number of trials, the average
total errors and the average total time per rat; and that
a positive correlation existed among the savings when
they were estimated on the basis of the average errors
and average time per trial, except in three instances.

The transfer-effect as between two mazes when the first
is only partially learned did not become positive according
to all three criteria until the transfer following the sixteenth
trial upon the former maze, transfers having been effected
after the second, fourth, eighth and sixteenth trials. The
transfer according to the time criterion was positive after
the change of mazes following two trials, and remained
positive throughout the series, and nearly uniform when
computed on the basis of the average time per trial. But
when it was computed on the basis of the average surplus
time per rat, it was both positive and cumulative.

When a transfer was made from a maze completely
learned to one already partially learned, and the learning
of the latter then completed, the latter maze was learned
with a saving in trials when the partial learning consisted
of two, four or eight trials, but the transfer was negative
when the partial learning consisted of sixteen trials. The
transfer of training, as shown in the average total number
of errors per rat, was positive when the partial learning
consisted of two or four trials in D, and negative when
it consisted of eight or sixteen. It was positive when
it consisted of two or eight trials in E, and negative when

TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 65

it consisted of four or sixteen. As shown in the average
total time per rat, it was positive throughout the entire
series in which the preliminary training took place in the
harder maze, and negative throughout the series in which
it took place in the easier maze, except that when these
sixteen preliminary trials took place in the easier maze
there was a positive transfer of .75 per cent.

When two mazes were learned together, the first being
partially learned, then the second completely learned, and
finally the first completely learned, it was found that
no one of these methods enjoyed a superiority according
to all three criteria. The mazes chosen for this purpose
were the ones in which the greatest degree and the least
degree respectively of positive transfer had appeared in
the preceding experiments. When the one in which the
least had appeared and the one in which the greatest had
appeared were learned in this order for purposes of control,
it resulted in the two being learned with fewer trials and
errors than when the order was reversed; the time was
slightly greater, being an average of 1.22 as against 1.19.
The advantage, from the point of view of trials and from
that of errors, in learning the maze first in which the least
transfer-effect had appeared was manifest throughout the
entire series of learnings where this may be viewed as the
fundamental order. The advantage from the point of view
of time resides with the other order, in which the maze
manifesting the greatest transfer was learned first.

When a maze was relearned after four intervening maze-
learnings, only one group showed, according to all three
criteria, signs of the retention of beneficial effects from
the original learning of the maze. Two groups showed
signs of the retention of a beneficial effect in time, but
with no unmistakable evidence of such benefit in either
trials or errors.

The Behavior Monographs

Edited by JOHN B. WATSON

The Johns Hopkins University, Baltimore, Md.

VOLUME 1
No. 1 The development of certain instincts. and habits in chicks. By

Frederick S. Breed. Pp. iv + 78, $1.00, postpaid.
No. 2 Methods of studying vision in animals. By Robert M. Yerkes

and John B. Watson. ""Pp. iv + 90, $1.25, postpaid.
No. S An experimental study on tiie death-feigning of Belostoma

(-Zaitha Aucct.) flumineum Say and Nepa apiculata Uhler.

By Henry H. P. Sevei-in and Harry C. Sevei'in. Pp. iii -1

47, $.G5, postpaid.
No. 4 The biology of Pliysa. By Jean Dawson. Pp. iii + 120, $1.50,

postpaid.
No. 5 The function of the vibrissae in the behavior of the white rat.

By Stella Burnham Vincent. Pp. iv + 81, $1.15, postpaid.

VOLUME 2
No. 6 The delayed reaction in animals and children. By Walter S.

Hunter. Pp. v + 86, $1.15, postpaid.
No. 7 The Canada porcupine : a study of the learning process. By

Leroy Walter Sackett. Pp. iii + 84, $1.15, postpaid.
No. 8 Audition and habit formation in the dog. By Harry Miles John-
son. Pp. iv + 78, $1.00, postpaid.
No. 9 Habit formation in a strain of albino rats of less than normal

brain weight. By Gardner Cheney Basset. Pp. iv + 46,

$.65, postpaid.
No. 10 Distribution of effort in learning in the white rat. By John

Linck Ulrich. Pp. iii + 51, %.65, postpaid.
No. 1 1 The effect of age on habit formation in the albino rat. By

Helen B. Hubbert. Pp. v + 55, $.65, postpaid.

VOLUME 3

No. 1 2 The mental life of monkeys and apes : a study of ideational be-
havior. By Robert M. Yerkes. Pp. iv + 145, $1,50, postpaid.

No. 13 A study of perseverance reactions in primates and rodents. By
G. V. Hamilton. Pp. iv + 65. $.75, postpaid.

No. 14 An analysis of the learning process in the snail, physa gyrina Say.
By Elizabeth Lockv/ood Thompson Pp. iii + 97, $1.25,
postpaid.

No. 15 The effect of length of blind alleys on maze learning: An ex-
periment on twenty-four white rats. By Joseph Peterson.
Pp. iii + 53, $.75, postpaid.

No. 16 An experimental study of transfer of response in the white rat.
By Harry H. Wylie. Pp. iii + QQ, $1.00, postpaid-

VOLUME 4.

No. 17 Transfer of training in the white rats upon various series of
mazes. By Rutledge T. Wiltbank. Pp. iii 4-65, $1.00,
postpaid.

No. 18 Redintegration in Albino rat. By William Brockbank, A. M.
Pp. iii + 66, $1.00, Postpaid.

Regular subscription $5 per volume, postpaid.

THE BEHAVIOR MONOGRAPHS

Emerson Hall, Cambridge, Mass.

Directory of American Psyciiological Periodicals

American Journal of Psychology — Worcester, Mass.: Florence Chandler.
Subscription $5. 600 pages annually. Edited by G. Stanley Hall.
Quarterly. General and experimental psychology. Founded 1887.

Pedagogical Seminary — Worcester, Mass.: Florence Chandler.
Subscription $5. 575 pages annually. Edited by G. Stanley Hall.
Quarterly. Pedagogy and educational psychology. Founded 1891.

Psychological Review — Princeton, N. J. : Psychological Review Company.

Subscnption (with Psychological Bulletin) $6. 480 pages annually.
Bi-monthly. General. Founded 1894. Edited by Howard C. Warren.

Psychological Bulletin — Princeton, N. J.: Psychological Review Company.
Subscription $3.25. 480 pages annually. Psychological literatvu-e.
Monthly. Founded 1904. Edited by Shepherd I. Franz.

Psychological Monographs — Princeton, N. J.: Psychological Review Company
Subscription $4 per vol. 500 pages. Founded 1895. Ed. by James R. Angell.
Published without fixed dates, each issue one or more researches.

Psychological Index — Princeton, N. J.: Psychological Review Company.
Subscription $1.25 200 pages. Founded 1895. Edited by Madison Bentley.
An annual bibliography of psychological literature.

Journal of Philosophy, Psychology and Scientific Methods — New York:

Sub-Station 84. Bi-weekly. 728 pages per volume. Founded 1904.
Subscription $3. Edited by F. J. E. Woodbridge and Wendell T. Bush.

Archives of Psychology — Substation 84, New York: Archives of Psychology.
Subscription $5. 600 pp. ann. Founded 1906. Edited by R. S. Woodwortb.
Published without fixed dates, each number a single experimental study.

Journal of Abnormal Psychology — Boston: Richard G. Badger.
Subscription $5. 480 pages annually. Edited by Morton Prince.
Bi-monthly. Founded 1906. Entire field of abnormal psychology.

Psychological Clinic — Philadelphia: Psychological Clinic Press.

Subscription $1.50. 280 pages annually. Edited by Lightner Witmer.
Monthly (9 numbers). Orthogenics, psychology, hygiene. Founded 1907.

Training School Bulletin— Vineland, N. J.: The Training School.

Subscnption $1. 160 pages annually. Ed. by E. R. Johnstone. Founded 1904.
Monthly (10 numbers). Psychology and training of defectives.

lournal of Race Development — Worcester, Mass. : Louis N. Wilson.
Subscription $3. 460 pages annually. Founded 1910.
Quarterly. Edited by George H. Blakeslee and G. Stanley Hall.

Journal of Educational Psychology — Baltimore: Warwick & York.
Subscription $3, 600 pages annually. Founded 1910.
Monthly (10 numbers). Managing Editor, J. Carleton Bell.
(Educational Psychology Monographs. Edited by Guy M. Whipple.
Published separately at varying prices. Same publishers.)

Journal of Animal Behavior — Cambridge, Mass.: Emerson Hall.
Subscription $5. 450 pages annually. Founded 1911.
Bi-monthly. Robert M. Yerkes, Managing Editor.

The Behavior Monographs — Cambridge, Mass.: Emerson Hall.
Subscription $5. 450 pages per volume. Edited by John B. Watson.
Published without fixed dates, each number a single research.

Psychoanalytic Review— Washington, D. C: 3617 Tenth St., N. W.
Subscription $5. 500 pages annually. Psychoanalysis.
Quarteriy. Founded 1913. Ed. by W. A. White and S. E. JeUiffe.

Journal of Experimental Psychology — Princeton, N. J. : Psychological Review
Company. 480 pages annually. Experimental. Founded 1916.
Subscription $3.25 Bi-monthly. Edited by John B. Watson.

Journal of Applied Psychology — Worcester, Mass.: Florence Chandler.
Subscription $4. 400 pages annually. Quarterly. Founded 1917.
Edited by G. Stanley Hall and L. R. Geissler.

The Brandow Printing Company
Albany N. Y.

Behavior Monographs

Volume 4, Number 2, 1919 Serial Number 18

Edited by JOHN B. WATSON

The Johns Hopkins University

Redintegration in the Albino Rat

A Study in Retention

BY
THOMAS WILLIAM BROCKBANK, A. M.

Published

at Cambridge, Boston, Mass.

HENRY HOLT & COMPANY

34 West 33d Street, New York

G. E. STECHERT & Co., London, Paris and Leipzig, Foreign AgenU

[Publication temporarily suspended]

The Journal of Animal Behavior

An organ for the publication of studies concerning the instincts,
habits and intelligence of organisms

The Journal contains a Department of Notes in which appear brief accounts of
especially interesting and valuable observations of behavior.

Published bi-monthly at Cambridge, Boston, Mass., by Henry
Holt and Company, New York. Each volume contains at least 450
pages with plates and text-figures.

The subscription price is $5.00 per volume (foreign subscription
$5.50) postpaid.

Manuscripts, subscriptions, and all business correspondence should
be addressed to

THE JOURNAL OF ANIMAL BEHAVIOR

Emerson Hall, Cambridge, Massachusetts

Behavior Monographs

For the publication of studies in behavior and intelligence which are

too lengthy or too expensive for acceptance

by The Journal of Animal Behavior

Published at irregular intervals at Cambridge, Boston, Mass., in
connection with the Journal of Animal Behavior, by Henr}'^ Holt and
Company, New York. Each volume contains approximately 450
pages with plates and text-figures.

The subscription price is $5.00 per volume (foreign subscription
$5.50) postpaid.

Monographs may be purchased separately at prices varying with
the cost of manufacture.

Manuscripts and inquiries concerning terms of publication should
be addressed to the Editor of the Behavior Monographs,
JOHN B. WATSON,

The Johns Hopkins University,

Baltimore, Md.

Subscription to The Journal of Animal Behavior and the Behavior
Monographs should be sent to Emerson Hall, Cambridge, Massa-
chusetts.

Behavior Monographs

Volume 4, Number 2. 1919 Serial Number 18

Edited by JOHN B. WATSON

The Johns Hopkins University

Redintegration in the Albino Rat

A Study in Retention

BY

THOMAS WILLIAM BROCKBANK, A. M.

Published

at Cambridge, Boston, Mass.

HENRY HOLT & COMPANY

34 West 33d Street, New York
G. E. STECHERT & Co., London, Paris and Leipzig, Foreign Agents

Acknowledgments

The Author desires to take this opportunity to express
his thanks to Dr. J. L. Ulrich for his continued interest
in th'^ -^--.--^^ work, and for his valued aid, encourage-
ment, and criticism, without which this dissertation would
not have been completed.

CONTENTS Page

Introduction 1

Apparatus and Method for Maze 4

Experimental Results:

I. Learning 11

II. Redintegration:

(1) Dominant " Error " 16

(2) Effect of Distribution of Trials 25

(3) Redintegration After a 30- Day Retention Period:

(a) Retention Period for \the Maze 32

(b) Influence of Learning a New Problem During the Reten-

tion Period 34

(c) Influence of Incomplete Learning 38

(4) Influence of Previous Training at 45-Day Retention Period 43

(5) Inclined Plane Problem 48

III. Individual Differences 61

Conclusions 64

Bibliography. 66

INTRODUCTION

In reviewing the field of animal psychology the fact
becomes evident that no special phase of retention has
ever been studied to any great extent, nor has there been
to date any single research devoted entirely to this subject.
The need of some definite knowledge, other than what is
speculative, concerning the nature or characteristics of
retention has been more and more emphasized year after
year as the facts of experiment point more inevitably to the
conclusion that the basis and foundation of all learning
is inseparably bound up with the capability of the organism
to retain what has been learned., If what is learned can
not be retained, there would be no advance on any line.
And if what is learned can be retained, we are vitally con-
cerned in possessing all attainable facts which will shed a
clearer light on the obscurity of present knowledge on the
subject. The present dearth of facts on retention is thus
not to be underestimated.

Experimental literature on the present subject is so
limited that the available references covering the history
of experiment in retention comprise for the most part only
incidental observations, supplementary to experiment which
had another specific object. For the sake of comparison
the following data has been summarized from these refer-
ences in work on animals*

Kinnaman (1) in his work on the monkey has shown by
some brief experiments that, after an absence of 50 days
from the problem, perfection in point of speed, and in-
cidentally errors, is not as marked as in the last trials of
learning. The problem was one of manipulation. Allen's
experiment (4) shows that the Guinea Pig retains a simple
labyrinth without great loss for 63 days. Porter in his
work (5) on the Vesper Sparrow, the Cowbird, the English
Sparrow, and the Pigeon shows that retention was good
for the first three subjects at 30 days, but there was great

2 THOMAS WILLIAM BROCKBANK

loss at 120 days and 140 days, respectively, for the Cowbird
and the Pigeon. Among other conclusions Rouse (6) men-
tions that the Pigeon retains fairly well for 6 weeks. The
period of retention on a simple labyrinth is given by
Hunter (11) as 28 days for the Pigeon.. Cole states (7)
that the memory for a combination of 7 fastenings was far
from perfect in the Raccoon at the end of approximately
5 months. With the same animal Davis (8) found that
retention for simple fastenings was very good at the end
of one year and even longer. Colvin and Burford tested
memory in a dog and a squirrel on color discrimination.
The dog exhibited no loss of discrimination for the standard
color at the end of 3 weeks; after 2 months the squirrel
showed no evidence of discrimination for the color, " red-
red-orange." In reference to the dog, Johnson (16) notes
that, after 60 days, loss in accuracy was only about 10 per
cent in the retention of habits of manipulation. Thorndike
observed (10) that chicks are practically perfect on a simple
labyrinth after 20 days. Breed (14) made a further ad-
vance on the conclusions of Thorndike and found that 30
days was the approximate periods for retention in the
chick. He states further that the chick which learns most
rapidly retains the best. Casteel (13) shows that the
Painted Turtle retains a simple sensory habit without loss
for 2 weeks. In his work on the Canadian Porcupine,
Sackett (15) shows that it retained the Hampton Court
maze for 10 days with some loss. Basset (17) and Ulrich
(18) have found that after a 60-day period most rats did
not retain the maze perfectly. Further experimental con-
clusions by Ulrich seem to indicate that the effect of the
distribution of effort in learning is maintained in reten-
tion within certain undefined limits; but his results on
this point are inconclusive. Hubbert (21) tested rats
90 days on the Round Maze, but established nothing
definite.

The foregoing summary of experiments on retention in
animals clearly shows that the little already done on the
subject lacks systematic study and a definite statement
of results however brief. Those methods which employ
only an increase or decrease of time or speed, and also

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 3

percentage of loss or gain, in explaining facts of retention
have been barren of results. From this it appears that a
new manner of approaching the problem along with some
determining factor for retention is necessary. When the
question is one in the learning of movement habit of some
sort the important consideration is movement, while all
else is subsidiary or of a secondary nature. This point has
certainly been overlooked in the past, and as a consequence
a confused mass of useless conclusions has accumulated.
It is the object of the present investigation to obtain
some definite results on retention. It may have been
readily gleaned years ago from experiments of long stand-
ing that the secret of retention may be found in the learn-
ing. Retention can be studied only by the effects learning
has produced in the organism. As a consequence the most
important thing is to compare the results obtained from
the learning with those of the retention tests. This con-
stitutes the primary factor to be considered.

APPARATUS AND METHOD FOR MAZE

The apparatus used in the present experiment comprised
for the most part the Watson Circular Maze (Figs. I. & II.)
somewhat modified. The paths were 12.7 cm. in width
and were partitioned off by vertical sheets of aluminum
13 cm. in height, set in grooves of a circular wooden base
189 cm. in diameter. The goal was 35.5 cm. in diameter.
Two semicircular sections of wire mesh covered all but the
center of the maze, which was covered by two separate
semicircular sections. The distance from the entrance to the
center of the maze was approximately 560 cm. This enlarged
pattern of the original maze was used because it allowed the
animal greater freedom of movement and at the same
time presented a greater difficulty in obtaining cues from
the walls of the alleys.

The entrances to the alleys were located alternately in
adjacent quadrants of the same arc. The radial stops in
the alleys were placed diametrically opposite the entrances
to the same alley, thus rendering it possible for a rat to
travel only one half the circumference of the maze in any
one direction from the entrance. No stop was placed in
the alley around the center.

The camera lucida attachment was employed in the
present experiment. This attachment consists fundament-
ally of two mirrors and an achromatic lens by means
of which the maze and its contents are reflected in miniature
on a convenient holder where the observer may plot the
course of the rat's pathway through the maze. The camera
lucida is described fully by Watson (20). In the present
maze a further alteration from the original consisted in
placing the lights higher up from the pathways, thus pro-
viding uniform illumination and avoiding shadows to a
greater extent than possible in the original design.

The object in using the above attachment was to obtain
a graphic record of each day's trial. It was the desire in

4

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 5

the present work to compare not so much the time and
distance of the learning trials with the trials after the
period of disuse of the habit, but to attempt to compare
the movements in the two series of trials, namely, learning
and redintegration. It was hoped that in this manner a
more definite disclosure of essentials in retention could be
thereby attained. The camera lucida attachment made
possible the permanent graphic records of all gross move-
ments and their ultimate comparison, and thus definite
consideration of the degree of precision of movement was
eliminated in the present work.

The rats used in this experiment were bred in our own
laboratory, inbreeding being avoided by the occasional in-
troduction of a strain from other laboratories. Young rats
were used in all experimentation, except in such cases
where previous training was introduced as a preliminary
factor. When 30 days old, as a rule, the litter was sep-
arated from its mother; at this time it was found the young
would thrive well by themselves. Before the time of ex-
perimentation had arrived, the young rats were tamed by
handling them every day. Actual work commenced on
the fiftieth day when the sexes were separated in cages
of six or eight. The incentive to all activity during the
experiment was food. The kind of food and the amount
used was constant, as far as lay within the power of the
experimenter. The usual food was milk-soaked bread, and
after the day's experiment some additional sunflower seed
and cracked corn. A hard grain biscuit was occasionally
soaked with the bread in order to afford a slight change of
diet. The living cages were thoroughly washed and disin-
fected once per week in a strong solution of " Kreso Dip"
before fresh shavings were placed in them. The animals
were carefully watched and treated on the appearance of
any symptom of disease.

The method employed in research on the maze, with
all but two groups later mentioned, was as follows: The
groups were divided into those that should receive three
trials per day and those to receive one trial per day. Be-
ginning on the fourth day preceding the first actual trial
of learning, the group of rats which was to begin the prob-

6 THOMAS WILLIAM BROCKBANK

lem was fed every day for 30 minutes in the center of the
maze, the entrance to which had been temporarily closed.
The object of this procedure was to allow the rat, first, to
adjust itself to the new environment, and, second, to ac-
quire a certain association between this experimental en-
vironment and its daily food. On the first day of learning
a bowl of bread and milk was placed in the center, and
a rat of the group was placed by hand in the entrance
box. The rat was permitted to emerge into the first alley
of the maze, immediately after which the door was closed
behind it and a stop watch started. The graphic record
was also begun at this time and the course of the rat traced
until the center was reached or a failure was recorded at
the end of 30 minutes. Each rat of the group was ex-
perimented with in turn. The first day's experiment was
completed for the one trial per day group at the end of
the first trial.

With the three trials per day rats, the rat on reaching the
center was immediately replaced in the entrance box for a
new trial, and the previous procedure was repeated as also
in the third trial. When the experiment for the day had
been completed the group was fed in its respective cage
on bread and milk, and a small ration of grain; no more
food was allowed until the next day. Miss Hubbert (21)
has stated that ill effects may arise from this method of
feeding. In the present work this method was employed
throughout all experiments and nothing detrimental was
detected.

In this research the method of learning and its norm were
important, inasmuch as both were to be applied in the
series of trials after the retention period. No time norm
was adhered to rigidly. Six seconds is usually considered
the time norm for this maze, but rats were frequently
found that ran the maze perfectly, although in somewhat
slower time. The norm for learning, therefore, covered a
period of the last fifteen trials, the first six trials of which
were required to show perfect integration of all movements,
and at the same time such speed as would indicate the
establishment of the integrations. Less than fifteen trials,
the first six of which are perfection, can not be called a norm.

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 7

This has been estabHshed by experimentation in this work.
If fewer than six perfect trials are regarded as the norm,
additional trials are likely to reveal the imperfection of
response. The rat thus was required to run the last nine
trials, whether perfect or not. The majority of these last
trials were generally perfect, according to the norm of the
first six. In the foreging the word, " integrations," is to be
understood as meaning an absence of so-called " errors,"
or imperfect movements. Thus, if a rat exhibited perfect
movements in making a turn from one alley into the next
these movements were considered perfectly integrated, ir-
respective of speed and hesitations. The employment of
the camera lucida denied any further determination, as
regards details of the method by which the animals learned
the problem.

All criteria of learning in the maze have been based on
these three — time, " error," and distance. In deciding on
a norm for the present work, the time and distance criteria
have both been eliminated to a certain extent because
they were considered as only minor factors in aiding the
pointing out of essentials in learning and retention. The
learning of the maze consists in the acquirement by the
rat of certain integrated movements; and the trials after
the period of disuse of these movements shows the loss in
integrations of movements, if any. Time and distance as
primary factors can not be employed except under almost
impossible conditions to show what actually takes place
in the learning of the maze, but they are valuable as sec-
ondary factors. The primary criterion for us has thus
resolved itself to one of movement, and therefore precision
of integration is primarily concerned. , " Error," therefore
will not be considered except in the sense of imperfect
integration.

There has never been any definite agreement as to what
may be adjudged an " error." Not considering hesitation,
an " error " may be said to have been made by the rat
when, in the process of solving the problem, it enters a
cul-de-sac or retraces its steps. Here is where distance as
well as time may enter as factors, and they do so every
time an " error " is made. But they are eliminated as

8

THOMAS WILLIAM BROCKBANK

primary factors because they do not show where, how, and
why the " error " was made, and this is all important.
Time — that is, speed of integration — is not retained nor is
distance; but precision of movement is always retained,
more or less. The place of " error " in the problem must
be known. To facilitate this specific location of " error "
in the maze, the method was adopted of dividing the
maze into areas corresponding to the turns the rat was
required to make in a perfect trial to solve the problem.
Thus, namely, an " error," or absence of precision of in-
tegration at the first turn (Fig. 1) was designated at turn

FIG. I - MAZE

1 and was qualified as minus, being to the right of the turn;
at turn 2, plus, being to the left of the turn. " Error "
is not to be regarded here as a positive qualifying factor
in the learning.

The question then arose whether the trials after the

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 9

period of disuse, or the retention period, of the habit could
be called " Retention " trials if the above norm of learning,
namely, first six trials perfect, etc., would be applied to
them. According to Watson (20), behavior has not set
any fixed methods to make retention tests; nor could any
other author be quoted to answer the question adequately.
It was therefore decided to apply the norm, of learning
to the series of trials after retention, and call this series

10 THOMAS WILLIAM BROCKBANK

the " redintegration " trials. These trials were so desig-
nated because, following the norm of learning in the redin-
tegration trials after the period of disuse of the habit, the
rat did actually reestablish the integrated movements of
the original learned habit. It was seen further that, by
completing the redintegration trials, one could arrive at a
more definite conclusion as to what movements of the
habit had been lost and what retained. Retention here is
to be understood as the potential perservation in the
organism of those certain precise integrations of movement
necessary in the habit. The word " retention " will be used
in reference also to the period of disuse of the habit — the
time between learning and redintegration.

After the completion of learning, all rats were held over a
retention period for redintegration. During the retention
period, which according to the group was respectively 70
days, 45 days, or 30 days, the rats were required to run
through a runway about 900 cm. in length, which was
considerably more than the exercise they received in the
last few trials of learning. The amount of exercise thus
given was found to be sufficient to avoid the evils that
arise from allowing the rats to remain inactive during the
retention period. The rats were thus kept in normal con-
dition throughout the whole experiment, comprising the
successive periods of learning, retention, and redintegration.

The first trial in the redintegration series was not pre-
ceded by any preliminary feeding in the center of the maze.
On the day the redintegration trials were to begin, instead
of being exercised in the runway the rat was taken directly
to the entrance box of the problem and the procedure as
in learning was followed. It was in this stage of the work
that our opinion concerning the time norm as a secondary
criterion was confirmed. More than one rat was observed
to travel through the maze in relatively slow time, but
without loss in integration of movements, while others,
moving more quickly, a certain and definite loss. The
redintegration trials followed to the end the same norm
as the learning.

This is an adopted and modified system, taken from the theory which Dr. Ul-
rich recently advanced, that nearly all " errors " in the maze, after those in the first
few trials, are consequent upon the rat's making an imperfect turn (integration), or
not making a turn where one should be made.

EXPERIMENTAL RESULTS
I. Learning

As originally outlined in the general plan of research in
this study of retention every group of rats, following a
certain method in learning, retention, and redintegration,
should total at least ten. As the work progressed, however,
and unforeseen difficulties and varying circumstances pre-
sented themselves, it was found that the original plan could
not be adhered to; and, furthermore, it was seen that neces-
sity did not demand a rigid following of the set rule. Each
group therefore is not uniform in number.

Throughout the learning period of all rats, a general
similarity in the behavior of each group has been noticed.
The " cautious treading of the pathway in the early trials,"
the " hesitancy on leaving an explored alley to enter a
new one," the " excitement on encountering a stop," the
" usual increase of activity on drawing closer to the center,"
and other characteristics of early learning were noted care-
fully. Whether this general behavior is due to " hunger,"
" curiosity," "fear," " some kinaesthetic sensibility," or
a combination of any or all of these can not be definitely
stated; but, in part, it may be said to be due to response
to stimuli from environment.

That this behavior does not persist throughout the
whole learning process is certain. It is to be noted that
this general behavior as above stated is only the beginning
of subsequent integration. After the first few trials it may
be noted that integration begins to develop more definitely,
finally resulting in increase of speed until the norm is reached.

This treatise is not one on learning. Be that as it may,
before treating of retention and redintegration a careful
perusal of learning experiments on the maze, to date, has
forced us to a reconsideration of maze learning in order
to analyze, briefly, for our purpose the process of the elim-
ination of " error," and thus the acquirement of integration
of movement necessary in the perfect maze habit.

11

12 THOMAS WILLIAM BROCKBANK

Every experimenter has observed that in the early trials
there is little, if any, definite integration of movement
directed by the external senses toward the center; and
further, that " errors " may be made in almost every spot
in the problem, especially in the first few alleys. That this
is true of some rats more so than of others is a matter of
individual differences. But that these " errors " have an
influence on subsequent behavior is not to be doubted.
As the learning progresses it may be observed more clearly
than in the first few trials that, as a rule, there seems to
be some certain place or places in the maze where the
acquirement of perfect integration is more difficult than
elsewhere; and as. a consequence a delay arises here in the
integration of certain movements which are important links
in the chain of responses constituting the habit. In every
learning process, such as this in the maze, the series of suc-
cessive movements in the habit may be compared to a
chain which in the process of making has naturally some
relatively weak and strong links. The locus of an imperfect
response would indicate a weak link; and as the same
imperfect response recurs continuously, or at irregular in-
tervals, down through the whole process of learning, it is
quite evident that of all the imperfect responses possible
to be made in the treading of the maze, this particular
response or " error " is dominant. And, therefore, there is
to be observed in the process of learning what may be
called the " dominant ' error '." By the dominant " error "
is to be understood that imperfect integration which is
made most frequently as the result of some difficulty in
the problem which delays the establishment of perfect
integration. The " error " which persists is nearly always
the dominant "error "; in any case it is the dominant
" error " in the last trials of learning which, as will be
shown, has an important bearing on the subsequent trials
in redintegration. In such a problem as the round maze,
where each turn is a separate problem in itself, there may
be expected numerous " errors " more or less dominant
and equal in number to the number of separate problems.

Records 1 and 2 show the daily dominant imperfect
response, or " error " records for two individual rats in
the learning of the maze. These records present typical

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 13

examples of the points treated above and show quite clearly
the case of the dominant " error," with the occasional
recurrence of other " errors." Thus record 1 shows the
plus " error " dominant at the start, the minus " error "
at turn 2, the plus " error " at turn 3, etc. Turn 3 shows
the minus " error " persisting, and likewise dominant in
later trials. This point of dominant " error " in later
trials will be treated of latef. Numerous other daily
records could be readily cited in full if space permitted.
Suffice it to state that all other records, with very few
exceptions, substantiate the two given herewith.

RECORD 1
Dominant Imperfect Responses in Learning
70-Day Period. 1 Trial per Day
Rat 7. " Errors " At Turns In Daily Trials

Turn

Start

1

2

3

4

5

6

*P M

P

M

P

M

P

M

P

M

P

M

P M

Trial 1

2

4

5

2

2

2

4

10

18

4

2

9

1

3

6

6

4

4

7

4

2

4

1

3

1

16

4

1

8

3

2

5

1

3

6

24

10

5

3

6

3

1

6

1

4

4

4

2

9

8

2

2

7

1

1

1

1

5

1

2

8

1

1

2

1

9

1

1

1

1 !

10

2

1

1

11

1

2

1

12

'

2

13

1

1

14

1

1

1

15

1

1

16

1

1

17

1

18

1

1

19

1

1

20

1

1

21

1

22

1

1

23

24

1

25

1

1

26

1

27

1

♦*28

1

Totals

6 1

24

35

75

24

23

58

29

7

1

4

2

* p. — Plus " errors," to the left of turn.

M. — Minus " errors," to the right of turn.
** Finished in 43 trials; no " errors " after the 28th trial.

14

THOMAS WILLIAM BROCKBANK

Rate.

RECORD 2

Dominant Imperfect Reponses in Learning

70 Day Period. 3 Trials per Day.

" Errors " At Turns In Daily Trials.

Turn

Start

1

2

3

4

5

6

Trial

P M

P

M

P

M

P

M

P M

P M

P

M

1

1

3

4

1

1

2

2

2

2

1

3

3 1

3

3

1

5

1

4

1

3

2

1

1

1

1

1

5

1

1

1

1

1 1

1

2

5

6

2

1

4 3

1 1

7

5

7

2

1

1

2

2

8

8 1

3

1

2

1

2

9

1

4

2

,

1

10

1 1

1

1

11

1

12

1

1

13

1

14

2

1

1

15

1 1

1

2

1

16

1 3

1

2

1

1

17

2 1

1

1

18

1 1

1

19

1 1

20

1 1

1

21

2

1

22

1

1

23

2

2

2

24

1 1

1

25

1

26

1 2

1

1

27

28

1

1

29

1

30

1

1

31

1

32

1

1

2

33

1

2

2

1

1

1

34

1

1

2

35

1

2

2

1

36

1

1

37

1

*39

1

42

1 2

1

44

1 1

45

2

46

1

'

47

1

49

1 1

50

1

51

1

1

53

1

* Omitted trials are free of " error."

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 15
Record 2 (Continued)

Turn

Start

1

2

3

4

5

6

P M

P

M

P

M

P M

P

M

P M

P M

Trial 54

1

1

1

55

1

56

1

1

58

1

59

1

60

1

62

1

1

63

1

1

64

1

66

1

67

1

1

68

1

1

1

69

1

1

70

1

71

1

74

1

75

1

77

1

1

78

1

85

1

90

2

1

1

1

1

92

1

1

93

1

Totals

40 40

29

52

30

10

6 36

20

4

1 3

13 16

II. Redintegration
(1) Dominant " Error "

Records 1 and 2, showing " errors " in daily trials of
two individual rats, have been cited in full to illustrate
not only the dominant " error " but also the method of
obtaining the total " errors " in the tables to be given.

Table I-A presents the total " errors " in both learning
and redintegration for the 3 trials per day rats of the
group at 70^day retention. Rats 1, 2, and 3 comprise one
complete litter; rats 4, 5, and 6 are from a litter of 7, the
four remaining being in the 45-day group. Rat 6 is the
one cited previously in record 2. The figures running
horizontally across the page and preceded by " L." and
" R " are the total " errors " in learning and redintegration,
respectively. The figures in perpendicular columns under
" P." and " M." are totals of plus and minus " errors,"
respectively. All the " error " tables to follow are con-
structed in this manner.

It has been observed in the consideration of records
1 and 2 that of the two " errors," plus and minus, one is
an established dominant " error " in almost every case.
Thus, beginning at the very start of the maze and con-
sidering the learning totals, individually or collectively,
the dominant " error " runs in the order plus, minus, plus,
minus, etc., omitting alleys 5 and 6 for the present. E. g.,
in the " errors " of rat 2, plus 85 is dominant over minus 18,
minus 83 over plus 12, plus S3 over minus 20, etc. The
qualitative totals present a similar situation.

Coming to a consideration of the redintegration " errors,"
it may be seen that the dominant " errors " at the turns,
with a few unimportant exceptions, follow those dominant
in learning. This is notable. There are not as many
" errors " in redintegration as in learning, and thus the
plus, minus, plus, minus order is so decided; but that it is
there is beyond question. Therefore, at the locus or place

16

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 17

where integration was difficult to establisli in learning a
similar difficulty appeared in redintegration trials after
retention. In other words, the weak link in the chain of
the habit, or the integration difficult to acquire, has appeared
after a retention period, although the weakness apparently
had disappeared after the long period of learning.

Table I-B presents " errors," tabulated as in Table I-A,

TABLE I-A

70-Day Period. 3 Trials per'Day

" Errors " At Turns

Turn

Start

1

2

3

4

5

6

Total

No. of
Trails

*P M

P M

P M

PM

PM

PM

PM

P M

Ratl.**L.

72 58

36 88

54 15

11 53

39 7

2

10

5 11

219 242

150

R.

5 1

5 5

3

4 12

1 3

1

1 1

19 23

45

2. L.

85 18

12 43

83 20

5 30

18

5

5

2 12

210 128

138

R.

2 3

1

6

3 3

1

5 14

22

3. L.

61 38

55 111

52 13

4 60

31 5

6

13

14 11

223 251

153

R.

3

3 4

1 1

2 15

1 1

1

1

2 16

13 28

29

4. L.

43 20

12 49

46 23

6 54

8

4

2

6 10

125 158

101

R.

4 2

5 3

1 2

1 2

1 3

12 12

33

5. L.

62 41

24 76

73 28

10 57

21 9

11

14

20 21

221 246

107

R.

3 2

6 8

3 3

4 14

4

2

2

6 3

28 32

28

6. L.

40 40

29 52

30 10

6 36

20 4

1

3

13 16

139 161

93

R.

4 4

2 7

4 2

1 3

1

1

1

13 17

42

Totals

Qualit. L.

363 215

168 419

338 109

42 290

137 25

29

47

60 81

1137 1186

R.

19 9

23 30

12 9

12 52

10 7

4

4

10 15

90 126

Quant. L.

578

587

447

332

162

76

141

2323

R.

28

53

21

64

17

8

25

216

Aver's. L.

96.3

97.8

74.5

55.3

27.

12

.7

23.5

387.1

123.6

R.

4.7

8.8

3.5

10.7

2.8

1

.3

4.1

36.

33.1

* p. — Plus " errors," to the left of turn.
M. — Minus " errors," to the right of turn.
** L. — Learning; R. — redintegration.

but compares the total " errors " of redintegration with
the " errors " of the last 35 trials of learning. It was
hoped by summing these trials of learning that further
light on " dominance " would be obtained. The table
shows that dominance in later trials of learning, as a
whole, both in reference to complete learning and also
redintegration, follows closely what has already been noted.
It may be seen further that in some cases " errors " not

1!

THOMAS WILLIAM BROCKBANK

recording the learning totals of this table appear in redin-
tegration, and also that the number of " errors " in redin-
tegration frequently exceed that in the learning. From
this it seems that the " errors " which appeared earlier in
the learning are tending to reappear in redintegration;
or, in other words, a recapitulation may have taken place.
The results with the individuals are not as definite here
as in Table I-A, but totals follow the rule.

Table II-A presents the total " errors " in both learn-

TABLE II-A

70-Day Period. 1 Trial per Day

" Errors " At Turns

No. of

Turn

Start

1

2

3

4

5

6

Total

Trials

P M

P M

P M

PM

PM

PM

PM

P M

Rati.

L.

10 3

14 28

10 4

2 19

4

1 3

2

43 57

52

R.

2 1

1 3

4

3 8

24

2.

L.

7 7

16 20

9 4

11

14 2

2 5

1

49 49

48

R.

4 1

2 3

1

1

1 1

8 6

17

3.

L.

25 1

33 52

29 6

4 19

16 4

1

1

107 84

43

R.

3 1

2 7

4

4 1

9 13

17

4.

L.

15 5

7 18

30 13

2 20

22 2

1

77 58

51

R.

26 2

8

4

3 8

7

40 18

65

5.

L.

16 6

16 12

13 4

8 18

4

2

2 1

61 41

48

R.

3 1

4 12

3 3

2 10

2

14 26

43

6.

L.
R.

12 1
3 1

3 26
2

17 6

1 33

10

1

3 1

46 67
3 3

61
17

7.

L.

6 1

24 35

75 24

23 58

29 7

1 4

2

160 129

43

R.

4 3

1 7

1

13

3

1 1

10 24

29

8.

L.

8 3

3 10

5 2

3 14

4

1

1

24 30

54

R.

4 1

5 11

4

3 9

1

17 21

28

9

L.

11 3

1 11

9 1

6 10

3 1

1

1

31 27

29

R.

1

4 7

1 2

1 15

1

1

9 24

29

10.

L.

9 4

7 20

10 5

11 13

4 1

3 5

2

46 48

43

R.

Is

1 2

1 6

2

1

5 8

22

Tota

Qualit.

L.

119 33

124 232

207 69

60 215

110 17

11 20

13 4

644 590

R.

L.

51 13

20 66

16 5

9 64

20 2

2 1

118 151

Quant.

152

356

276

275

127

31

17

1234

R.
L.

64

86

21

73

22

3

269

Avar's.

15.2

35.6

27.6

27.5

12.7

3.1

1.7

123.4

47.2

R.

6.4

8.6

2.1

7.3

2.2

.3

26.9

29.1

ing and redintegration for the 1 trial per day rats of the
70-day group. Rats 1, 2, 3, 4, 5, 6, and 7 are from a litter

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 19

of eight, the remaining rat being in the 45-day group.
Rats 8, 9, and 10 are from a litter of five, the remaining
two being in the 45-day group. What was noted in Table
I-A applies in like manner to the present table as far as
totals and general results are concerned. But there are
some few individual discrepancies, such as in rat 2, turn 9,
rat 5, turn 1, etc.

Table II-B was compiled with the similar intention as
was Table I-B; that is, to ascertain dominance in later
trials. This table confirms the findings of table I-B. The
later trials of learning beyond doubt have an important
influence on the subsequent behavior in redintegration.

TABLE II-B

70-Day Period. 1 Trial per Day

" Errors " At Turns

Last 35 Trials of Learning; Complete Redintegration

Turn

Start

1

2

3

4

5

6

Total

No. of
Trials

Rati. L.
R.

2. L.
R.

3. L.
R.

4. L.
R.

5. L.
R.

6. L.
R.

7. L.
R.

8. L.
R.

9. L.
R.

10. L.
R.

P M
5 2
2 1

2 4
4 1

12

3 1

4 3
26 2

4 5
3 1
3

3 1
2

4 3
1

4 1
11 3

1

5 3
1 2

P M
3

1 a

2

2 3

1 4

2 7
5
8

3 1

4 12
1

2
1 3
1 7

4

5 11
1 11
4 7

7
1 6

P M

1

2 2
1

2

3 3

4

3 3

2

1
1
4
9 1

1 2

2 4
2

PM
3
4
1
1
6
4

1 3
3 8

5

2 10

7

1 20

13

1 7

3 9
6 10
1 15
2

1

PM

1
1 1

3 1

4 1
9

7
1
2
4

7

3

3

1

3 1

1

1 1

PM

2

1 2
1 1

1
1

1

PM

1
1
2

P M
6 8
3 8

5 11

8 6

16 13

9 13

17 14
40 18

8 11
14 26

8 7
3 3

14 25
10 24

6 12
17 21
31 27

9 24
12 16

5 8

52
24
48
17
43
17
51
65
48
43
61
17
43
29
54
28
29
29
43
22

Totals
Qualit. L.
R.

49 20
51 13

8 40
20 66

20 12
16 5

11 62
9 64

32 3
20 2

1 6

2 1

3 1

123 144
118 151

Quant. L.
R.

69
64

48
86

32

21

73
73

35
22

7
3

4

267
269

Aver's. L.
R.

6.9
6.4

4.8
8.6

3.2
2.1

7.3
7.3

3.5
2.2

.7
.3

.4

26.7
26.9

47.2
29.1

20

THOMAS WILLIAM BROCKBANK

k

TABLE I-B -
70-Day Period. 3 Trials per Day
" Errors "At Turns
Last 35 Trials of Learning; Complete Redintegration

No. of

Turn

Start

1

2

3

4

5

6

Total

Trials

P M

P

M

P

M

PM

PM

PM

PM

P M

Rati.

L.

9 3

4

5

1

1

2

1

16 10

150

R.

5 1

5

5

3

4 12

1 3

1

1 1

19 23

45

2.

L.

11 4

2

4

3

1

1

2

17 11

138

R.

2

3

1

6

3 3

1

5 14

22

' 3.

l;

7 2

3

6

1

1

2

1

1

11 13

153

R.

3

3

4

1

1

2 15

1 1

1 1

2 6

13 28

29

4.

L.

5 3

1

6

1

1

3

7 13

101

R.

4 2

5

3

1

2

1 2

1 3

12 12

33

5.

L.

3 3

6

8

4

1

1 3

1

15 15

107

R.

3 2

6

8

3

3

4 14

4

2 2

6 3

28 32

28

6.

L.

7

10

6

2

1

4

2

14 18

93

R.

4 4

2

7

4

2

1 3

1

1

1

13 17

42

*7.

L.

1

1

1

1

1

2 3

61

R.

3

4

4

3

2

5

5 1

2 1

11 4

28 17

21

8.

L.

3

1

3 1

43

R.

Is

4 2

5

6

3

3

1 4

3 4

6 2

6 1

28 22

21

Tota

Qualit.

L.

39 22

25

35

13

8

2 14

6 1

4

85 84

R.
L.

26 11

32

40

18

14

13 61

18 12

12 7

27 20

146 165

Quant.

61

60

21

16

7

4

169

R.
L.

37

72

32

74

30

19

47

311

Aver's

7.62

7

.5

r

!.62

2

.87

.5

21.1

105.7

R.

5.25

g

1 A

{

9.25 ! 3.74

2.37

5.87

38.8

30.1

* 7 and 8 omitted in Table I-A due to incomplete records of " errors " in learning.

Table III-A presents the total " errors " in both learning
and redintegration for the 3 trials per day rats of the 45-
day group. Rats 1, 2, 3, and 4 are from a litter of seven,
the remaining three of the litter being in the 70-day group.
In Table III-B rats 5, 6, and 7 are from a litter of six;
of the remaining three in the litter one failed to complete
the experiment and the other two are recorded in the
70-day group as 7 and 8. The dominance in " errors " of
both learning and redintegration in these tables confirms
the findings of the former tables. Table III-B is the
comparison of " error " totals of redintegration and the
last 30 trials of learning, which number was thought suffi-
cient for the 45-day retention period. These totals con-
firm those of Tables I-B and II-B.

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 21

TABLE III-A

45-Day Period. 3 Trials per Day

" Errors " At Turns

No. of

Turn

Start

1

2

3

4

5

6

Total

Trials

P M

P M

P M

PM

PM

PM

PM

P M

Rati. L.

86 43

49 113

93 49

11 70

32 6

9 11

18 19

298 311

146

R.

4 1

6 10

5 3

2 9

3

1 3

2 3

23 29

34

2. L.

88 26

18 73

86 39

8 65

32 3

1 8

7 8

240 222

170

R.

3 4

2 5

3

1 3

9 12

46

3. L.

63 33

25 60

58 15

9 37

26 4

1 8

9 9

191 166

143

R.

11 10

16 21

12 5

3 7

4 1

4

2 5

48 53

61

4. *L.

123 47

61 162

127 51

18 128

81 7

7 14

7 7

424 416

252

R.

5 2

2 4

1

4

7 11

27

Totals

Qualit. L.

360 149

153 408

364 154

46 300

171 20

18 41

41 43

1153 1115

R.

23 17

26 40

20 9

6 23

7 1

1 7

4 8

87 105

Quant. L.

509

661

518

346

191

59

84

2268

R.

40

66

29

29

8

6

12

192

Aver's. L.

127.25

140.25

129.5

86.5

47.75

14.75

21

567

177.7

R.

10

16.5

7.25

7.25

2

2

3

48

42.

* Did not complete learning; completed redintegration.

TABLE III-B

45-Day Period. 3 Trials per Day

"Errors" At Turns

Last 30 Trials of Learning; Complete Redintegration

No. of

Turn

Start

1

2

3

4

D

6

Total

Trials

P M

P M

P M

PM

PM

PM

PM

P M

Rati.

L.

2 4

1 3

1

4 7

146

R.

4 1

6 10

5 3

2 9

3

1 3

2 3

23 29

34

2.

L.

3 2

3 4

1

6 7

170

R.

3 4

2 5

3

1 3

9 12

46

3.

L.

6 2 7 8

5

1 2

1

20 12

143

R.

11 10

16 21

12 5

3 7

4 1

4

2 5

48 53

61

4.

L.

9 7

10 19

7 2

3 6

1

1

30 35

252

R.

5 2

2 4

1

4

7 11

27

*5.

L.

4 1

1 3

5 4

56

R.

2

2

1

3 2

27

6.

L.

1

2

1

2 2

63

R.

2 • 1

5

3

9

1

6 8

21

7.

L.

3

3

1

3 4

60

R.

Is

8

4

1 1

4

1 1

10 10

26

Tota

Qualit.

L.

28 16

22 42

14 2

4 10

2

1

70 71

R.
L.

35 18

26 51

25 10

6 29

7 1

1 7

6 9

106 125

Quant.

44

64

16

14

2

1

141

R.
L.

53 j 77

35

35

8

8

15

231

Aver's.

6.28

9.14

2.28

2.

.28

.14

20.14

R.

7.57

11,

5.

5.

1.14

1.14

2.14

33.

* 5, 6, and 7 omitted in Table III-A due to incomplete records of "errors" in
learning.

22

THOMAS WILLIAM BROCKBANK

TABLE IV-A

45-Day Period. 1 Trial per Day

" Errors " At Turns

No. of

Turn

Start

1

2

3

4

5

6

Total

Trials]

P M

P M

P M

PM

PM

PM

PM

P M

RatL

L.

27 16

16 30

14 3

2 35

14 1

1

74 85

118

R.

1

3

2 3

2 2

5 8

20

2.

L.

40 19

19 23

32 10

20 10

33 3

1 3

2 2

147 70

105

R.

4

1 5

1

2

1

1

7 8

32

3.

L.

40 23

14 36

19 4

2 40

4

1 2

1

80 105

84

R.

10 1

1 3

1 4

1 2

2

13 12

34

*5.

L.

30 19

17 33

27 13

6 26

10 7

3 9

11 11

104 118

59

R.

1

4 2

1 1

1

1

1

6 6

27

6.

L.

21 9

11 24

24 7

7 14

15 1

3 1

81 56

45

R.

5 5

2 2

2 1

1 4

10 7

25

7.

L.

12 24

5 22

15 5

3 14

13 2

3

48 70

57

R.

1 5

6 3

2 1

9 9

29

8.

L.

21 15

8 19

11 3

3 16

5

1 1

1 1

50 55

62

R.

Is

5 5

2

3

5 10

17

Tota

Qualit.

L.

191 125

90 187

142 45

43 155

94 14

7 15

17 19

582 562

R.
L.

26 12

14 20

8 6

4 16

3 2

3

1

55 60

Quant.

316

277

187

198

108

22

36

1144

R.
. L.

38

34

14

20

5

3

1

115

Aver's.

45.14

39.65

26.7

28.28

15.4

3.1

5.1

163.4

75.71

R

5.4

4.85

2.

2.8

.71

.43

.14

16.33

26.21

* -4 did not complete redintegration.

Table IV-A presents the total " errors " in both learning
and redintegration for the 1 trial per day rats of the 45-
day group. Rats 1 and 2 are from a litter of 5, and rat
3 is from a litter of 7, the remaining rats of both litters
being in the 70-day group. Rats 5 and 6 are from a litter
of 3, the remaining rats of which did not complete the
experiment. Rats 7 and 8 comprise a complete litter.
This table adds nothing further to the previous tables
other than a confirmation of their general result — of domi-
nance. Table IV-B is modeled after Table III-B. The
results are not as definite as in the former tables but general
results are confirmed.

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 23

TABLE IV-B

45-Day Period. 1 Trial per Day

" Errors " At Turns

Last 30 Trials of Learning; Complete Redintegration

No. of

Turn

Start

1

2

3

4

5

6

Total

Trials

P M

P M

P

M

PM

PM

PM

PM

P M

Rati.

L.

4 1

2 3

1

1

7 5

118

R.

1

3

2

3

2 2

5 8

20

2.

L.

2

2 1

1

1

1

1 1

6 4

105

R.

4

1 5

1

2

1

1

7 8

32

3.

L.

2 4

5

7

2 16

84

R.

10 1

1 3

1 4

1 2

2

13 12

34

5.

L.

5 8

4 2

3

3

1

2 3

15 16

59

R.

1

4 2

1

1

1

1

1

6 6

27

6.

L.

12 2

3

2

3 1

18 5

45

R.

5

2 2

2

1

1 4

10 7

25

7.

L.

13

3

2

2

3 17

57

R.

1 5

6 3

2

1

9 9

29

8.

L.

3 11

1

4

1

1

4 17

62

R.

Is

5 5

2

3

5 10

17

Tota

Qualit.

L.

28 39

11 11

7

8

2 15

4 2

1 2

2 3

55 80

R.
L.

26 12

14 20

8

6

4 16

3 2

3

1

55 60

Quant.

67

22

15

17

6

3

5

135

R.

L.

38

34

14

20

5

3

1

115

Aver's.

.9.57

3.14

2

.14

2.42

.85

.42

.71

19.28

75.7

R.

5.42

4.85

2

2.85

.71

.42

.14

16.42

26.2

TABLE IV-D

45-Day Period. 1 Trial per Day

Distance in Centimeters

Alleys

1

2

3

4

5

, 6

Total

Rat 1.

L.

30341

18729

13497

17206

7898

8792

96463

R.

3445

2622

2634

1978

1120

1500

13299

2.

L.

27935

20169

15562

16285

10562

8360

98873

R.

5338

3961

3564

2548

2057

2400

19868

3.

L.

30382

14291

12392

11206

5298

6270

79839

R.

5427

4211

3332

3408

2032

2475

20885

5.

L.

27681

14355

12513

7790

6098

7785

76222

R.

4560

3704

2646

2145

1514

2271

16840

6.

L.

17545

13739

9117

7132

4658

3675

55866

R.

4510

3766

2673

2428

1400

1875

16652

7.

L.

21074

16093

10303

8383

5278

4814

65945

R.

5819

5150

2842

2262

1624

2175

19872

8.

L.

22681

13249

8878

6918

4121

4976

60823

R.

4977

2579

1764

1475

1008

1350

13153

24 THOMAS WILLIAM BROCKBANK

Table IV-D presents the distance totals of the 45-day
group at 1 trial per day. The respective distances of each
day's trial were computed by ration from the measure-
ments of tracings obtained through the camera lucida,
and the totals as here presented were summed on the
adding machine. The object of computing the totals
which this table presents was to ascertain whether distance
in learning had influence in redintegration, either in par-
ticular alleys or in total result. No positive result was
obtained on this point, as the table clearly shows. However,
a subsidiary fact is evident, namely, that the distances
generally diminishes toward the center of the maze with
the exception of the last alley, where " fear " or timidity
may intervene; but at the same time there are striking
instances in which the rule does not hold, e. g., in the first
record of the present table. The reduction of distance
toward the center may have been expected, inasmuch as
the construction of the maze is such that this result would
be almost inevitable. But that distance in learning seems
to bear no specific influence on distance in redintegration
is an important fact and the necessary finding to eliminate
distance as a prime factor in the consideration of final
conclusions. The distance traversed in successive alleys,
therefore, is not of such importance as the establishing of
integration of turning where the dominant " error " is
produced. There would not be excessive distance if no
" error " was produced.

As has already been observed in previous pages, the
maze is a compound of lesser problems which are located
specifically at the successive turns. Turning to Figure 1,
it may be seen that the serial order of integrations in the
treading of the maze proceeds, first, w4th the acquiring of
perfect integration of movement away from the entrance,
or start, to the right; second, the solution of turn 1, which
consists in the acquiring of perfect integration of move-
ment in making the 180-degree turn to the left in passing
from alley 1 to alley 2 ; third, the acquirement of similar
integration, as in turn 1 at turn 2, except that here the
turn is to the right and so on through the series of successive
turns to the center.

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 25

It is quite true that to turn either way from the start
would be equally easy. But when the rat has finally begun
to run the correct pathway from the start the more easily
integrated movement by which to enter alley 2, through
turn 1, is in the direction which leads to a minus " error,"
i.e., straight away into the cul-de-sac. It is easier for the
rat to continue straight ahead, past any turn, in the direc-
tion in which it is running, or straight ahead through the
turn into the cul-de-sac. These " errors " are indicated by
a minus sign in the turns 1, 3, and 5; and plus at the other
turns. All figures show that many more " errors " are
made here, both in learning and redintegration, than in
making the more acute turns, which require established
integration to be made continuously in a perfect manner.
Thus it is clearly seen that the integrated movement at
the respective turns must be learned before the habit can
be established.

Some authors have attempted to show by logic rather
than by research in the laboratory that this question is a
mere matter of chance — which way a rat will go through a
turn. But experimental observation clearly proves that it
is a hard and fast question in experiment on the establish-
ment of perfect integration of movement. It is not a
matter of chance but of difficult integration versus easy
integration. The minus " error " at turn 1 is thus dominant
over the plus " error "; at turn 2, the plus over the minus,
etc., as far as already limited. In other words, as already
stated, the weak link in the chain of the habit — the domi-
nant error or the integration difficult to acquire — appears
in the process of learning and reappears after retention in
redintegration.

(2) Effect of Distribution of Trials on Redintegration

It has been shown quite conclusively in animal experi-
mentation that the wider the distribution of effort, within
certain limits as yet undefined, the greater will be the
economy [Yerkes (3), Hunter (12), Ulrich (18)]. Any
attempt to ascertain the effect upon retention of different
distributions of effort in learning should be begun with
full cognizance of these conclusions.

26 THOMAS WILLIAM BROCKBANK

As has been shown in the foregoing pages, in the process
of learning, in the acquirement of a habit, the establish-
ment of complex integrations of movement has a very
marked effect on the retention of the habit. All methods
of learning produce finally the same result, viz, the estab-
lishment of the perfectly integrated habit. Yet since each
method of learning, according to the distribution of effort,
has such a marked effect on the characteristics and general
progress of the learning process, a legitimate working
hypothesis may be laid down to the effect that retention
and subsequent redintegration may be affected in a some-
what similar manner by the method of distribution of
trials previously employed in learning. Within a definite
limit it is the object of the present chapter to show some
effects of the distribution of effort in learning on the re-
tention and redintegration.

It was with the twofold object of attempting to throw
some light on retention that the different groups of rats
considered in the foregoing pages were divided into three
trials per day groups and one trial per day groups. The
" error " Tables I-A, II-A, III-A, and IV-A have already
been explained in reference to the dominant " error." The
economy of these same tables is summed in the " error "
totals. It may be noted in these totals that the number
of " errors " in learning were far more numerous for the
three trials per day group than for the one trial per day
group. This fact is significant in the explanation of the
economical superiority of the one trial per day method
over the three trials per day method. The group following
the three trials per day method display a more pronounced
difficulty in establishing the perfect integrations of move-
ment in the solution of the successive turns and the con-
sequent acquirement of the habit, if one is to judge from
the " errors." The three trials per day method is pro-
ductive of the greater number of " errors," and evidently
a greater expense of energy. Therefore, the one trial per
day method is the method of wider distribution of trials
and is more economical.

Turning to the influence of retention, in redintegration,
it may be seen that a somewhat similar situation exists.

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 27

But the fact is most evident that although the one trial
per day method evokes a more rapid reestablishment of
perfect integration and hence a lesser number of " errors,"
yet the three trials per day method has relatively gained
in economy. This fact is illustrated in Plates I and II,
where it may be observed that the average " error " curves
tend toward a mean in redintegration.

It has been noted that every method of learning, in
regard to the problem under consideration, arrives approx-
imately at the same end, namely, the establishment of
the perfectly integrated habit. That is, at the beginning
of the retention period — and with the exception of the
process or history of the establishment of integrations — the
three trials per day group may be considered equal to the
one trial per day group as far as habit integrations are
concerned. In integration as is here considered this seems
to be the fact. Otherwise there would be no tendency of
the " errors " of the three trials per day group and the
" errors " of the one trial per day group to tend to a mean,
as the curves in Plates I and II show graphically; and
likewise with the trials. In redintegration the absolute
number of " errors " of the one trial per day method is
less than the number in the three trials per day method;
but the relative decrease of " errors " in the three trials
per day redintegration is greater than in the one trial per
day redintegration. And thus in economy the one trial
per day method after retention is absolutely superior to
the three trials per day method, but relatively inferior.

Tables I-C, II-C, III-C, and IV-C contain total trials
in learning and redintegration; and also time and perfect
trials in the last 15 trials of learning and the first 15 trials
of redintegration of those groups in the " error " tables
already considered. Curves (b), (c), and (d) in Plates
I and II are constructed from the averages in the above
tables. Curve (b) in both plates, showing average trials,
displays the tendency to a mean which appears in redin-
tegration totals, and further, also the superiority of the
one trial per day method. Curve (c) in both plates of
average perfect trials in the last 15 trials of learning and
the first 15 trials of redintegration shows no marked

28

THOMAS WILLIAM BROCKBANK

TABLE I-C

70-Day Period. 3 Trials per Day

Time in Seconds

No. ot Trials

First 15 Trials of R.; Last 15 Trials of L.

Av.

Rati. L.

150

*6

*7

*e

*7

*6

*7

*6

*6

*6

12

*7

*6

8

*6

*6

6.8

R.

45

241

55

75

35

154

28

*12

*14

27

16

20

n2

no

n2

8

47.9

2. L.

138

*6

*6

*6

*6

*6

*6

*6

*6

*7

7

*6

*8

8

*6

*6

6.4

R.

22

105

15

15

13

no

*9

9

*8

*9

*8

*8

*9

*7

*6

*7

15.5

3. L.

153

*6

*6

*6

*6

*6

*6

*6

*6

12

8

*7

20

*6

7

*7

7.5

R.

29

87

145

127

27

170

68

35

*13

14

12

*8

8

12

17

*9

50.1

4. L.

101

*6

*6

*7

*7

*6

*6

*7

*7

*6

*7

*6

13

12

7

13

9.9

R.

33

100

23

13

15

10

17

17

*7

*7

*9

24

27

15

*8

no

20.1

5. L.

107

*6

*6

*6

*6

*6

*6

11

8

47

61

14

20

*7

*7

17

15.2

R.

28

197

255

356

15

217

73

19

125

35

13

51

20

16

*8

*7

93.8

6. L.

93

*6

*6

*6

*6

*6

*6

8

*6

*7

*6

*6

72

*8

15

8

11.4

R.

42

59

12

10

*9

*7

*7

*7

15

*8

*7

14

*8

*7

29

11

14

7. L.

61

*6

*6

*7

*6

*6

*6

*6

*6

*7

*6

*6

*6

25

*7

*6

7.4

R.

21

581

147

55

85

187

23

*7

*7

*6

*9

*7

*6

*8

*7

*6

76

8. L.

43

*6

*7

*6

*7

*6

*6

*6

8

no

*7

*6

*8

*7

*7

*8

7

R.

21

297

587

432

1330

27

8

*7

*6

*7

*6

*6

*6

*7

*6

*6 182.5

Averages-

—

Trials.. L

105.7

* (Perfect) Trials.. L. 11

.8

Time..L. 8.9

R

. 30.1

R. 6.2

R. 62.4

* starred trials were perfectly integrated.

TABLE I I-C

70- Day Period. 1 Trial per Day

Time in oecgnds

No.

of Trials

First 15 Trials of R.; Last 15 Trials of L.

Av.

Rat 1.

L.

52

*6 *6

-6

*6

*7

*6

*6

*7

*6

*6

7

*6

*6

20

*6

7.1

R.

24

16

*6

*6

*6

27

10

*6

*6

25

*7

*7

*6

*6

*6

*7

9.8

2.

T,.

48

*6

*6

*6

*6

*6

*6

20

*6

*7

*6

20

11

*6

*6

*6

8.9

R.

17

53

11

*7

*6

*6

*6

*7

*6

30

*6

*6

*6

59

*6

*7

14.8

3.

T,.

43

*7

*6

*6

*6

*6

*6

*6

*6

15

*6

8

*7

22

*6

*6

7.9

R.

17

20

13

*7

*6

*6

*e

*7

*6

Q

*6

9

153

*6

43

10

20.4

4.

L.

51

*6

*6

*6

*6

*6

*6

*8

25

10

*7

8

11

*6

7

30

9.8

R.

65

36

14

*6

13

9

44

10

*7

*7

21

10

27

17

8

*7

15.7

5.

T,.

48

*6

*6

*6

*7

*6

*6

*6

*6

*6

12

11

*6

*6

30

20

9.3

R,

43

72

n2

10

9

8

*7

14

*7

26

*7

34

15

63

*7

9

20.

6.

T,.

61

*6

*6

*6

*6

*6

*6

*6

16

*6

*6

*6

*7

*7

10

*6

7.

R.

17

?A

13

*9

*7

*7

*7

*7

*9

*7

*7

*7

*7

*7

15

*8

9.4

7.

L.

43

*6

*6

*6

*6

*6

*6

*7

*6

*6

*6

*7

*6

*7

*6

*6

6.2

R.

29

135

W

29

12

19

15

9

15

*7

13

11

10

*7

20

*7

21.6

8.

L.

54

*7

*7

*6

*6

*7

*7

*6

*6

12

8

*7

*6

*6

*6

*6

6.8

R.

28

31

15

16

9

9

11

*6

*6

30

9

10

*7

12

*7

*7

12.3

9.

L.

29

*6

*6

*6

*6

*6

*6

*6

*6

*6

*9

*6

*6

*6

*6

*6

6.2

R.

29

57

10

40

9

8

7

7

14

15

9

13

8

7

7

*6

14.4

10.

T<.

43

*6

*6

*6

*6

*6

*6

*6

15

*6

*9

10

*8

*8

*7

*6

7.4

R.

22

36

*7

8

*6

*6

*6

7

*6

*6

*6

*6

*6

*6

11

*6

8.6

Averages —
Trials L. 47.2
R. 29.1

* (Perfect) Trials L. 12.6
R. 7.1

Time L. 7.6
R. 14.7

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 29

^ Si S

^ ^. *»

H
LJ
<

Q.

CO

h
<

>■

o

Z
o

u

55S5,S»(b^*^l»lr,(Vj^

|§§gS&SSi$Si

T^~T

superiority of the one trial per day method. Curve (d)
in both plates, showing average time in the last 15 trials
of learning and the first 15 trials of redintegration, points
clearly to a superiority of the one trial per day method

30

THOMAS WILLIAM BROCKBANK

over the three trials per day method. These curves confirm
the results in the " error " curves in the fact that the one
trial per day method is absolutely superior in both learning
and redintegration; but, owing to the tendency to the mean
in redintegration, the one trial per day method becomes
inferior in a relative sense only.

TABLE III-C

45-Day Period. 3 Trials per Day

Time in Seconds

No.

of Trials

First 15 Trials of R.; Last 15 Trials of L.

Av.

Rati.

L.

146

*6

*7

*7

*6

*7 *6

*6

*6

*6

*6

*8

*7

*7

10

*6

6.7

R.

34

155

195

205

17

no

*7

no

13

12

*8

*8

*7

*9

*7

26

45.9

2.

L.

170

*6

*6

*7

*7

*7

*6

*7

*6

*6

*6

*6

*6

*6

*6

*6

6.2

R.

46

46

9

*7

*8

*6

*6

*7

13

14

*8

*6

10

no

11

10

11.4

3.

L.

143

*7

*6

*6

*6

*6

*7

20

*6

*7

*7

*6

9

*7

*6

*6

7.4

R.

61

54

26

37

19

110

71

25

75

95

20

70

27

8

10

20

44.4

4.

L.

252

45

85

51

no

35

17

45

20

17

16

12

10

9

10

8

26

R.

27

21

21

ni

*11

14

13

*9

*8

14

*8

20

19

*8

*7

*8

12.8

5.

L.

56

*6

*6

*6

*6

*6

*6

8

*6

*6

*6

9

8

*7

*7

*6

6.6

R.

27

41

*8

*7

*7

*6

14

*6

*6

*6

*6

*6

12

*6

*6

*7

9.6

6.

L.

63

*7

*6

*6

*7

*6

*7

*7

n5

*7

*8

7

*7

7

*6

*7

7.3

R.

21

192

*8

*8

*9

*7

*7

*8

*7

*6

17

*7

16

*8

*8

8

21.2

7.

L.

60

*6

*6

*6

*7

*6

*6

*6

*6

*7

*6

*6

*6

*9

*6

*6

6.3

R.

26

63

18

21

*6

*6

13

*6

32

*6

*6

10

*6

*7

*6

no

14.4

Averages —
Trials. L. 127.1
R. 34.5

(Perfect) Trials.. L. 11.7
R. 8

Time. .L. 9.5
R. 22.8

TABLE IV-C

45-Day Period. 1 Trial per Day

Time in Seconds

No. of Trials

First 15 Trials of R

.; Last 15TnaIsof L.

Av.

*6

*6

*6

*7

*7

*7

*7

♦6

7

*6

37

*6

*6

*6

8

8.5

115

no

*8

7

10

*/

*7

10

*6

*7

*7

*9

51

*7

*6

17.8

*7

*6

*6

*7

*7

♦6

*6

*6

*6

14

*7

9

*6

*6

*6

7

175

26

26

12

*9

n5

*14

25

n2

*8

*7

*7

25

*7

*7

25

*6

*7

*6

*7

*6

*7

*7

*6

*6

*6

*6

11

13

9

14

7.8

15

51

11

15

*6

*7

8

*7

*7

15

*8

*7

10

♦8

*8

12.2

*7

*7

*7

*7

*8

*8

*7

10

13

*7

*8

*7

19

*7

*8

8.6

9

*7

*9

*6

*7

115

8

*8

ni

*9

27

11

*7

*7

*6

16.4

*7

*7

*6

*6

*7

*9

*6

*6

*7

125

10

*7

10

*9

*6

15.6

39

*9

8

*7

7

11

*7

*7

8

*6

*6

*7

*6

*7

*7

9.4

*7

*6

*6

*6

*6

*6

8

♦7

*6

8

*6

14

*8

14

*9

7.8

73

14

11

*7

*6

13

*6

*6

12

8

*7

*6

*6

8

*6

12.6

*6

*6

*6

*6

*6

*6

*9

9 8

*8

*7

12

*6

13

9

7.8

17

8

*8

*6

*6

*6

*7

*6

*7

*6

*6

*7

10

*6

*6

7.f

Averages —
Trials.. L. 75.7
R. 26.2

♦(Perfect) Trials.. L. 11.5
R. 9.5

Time..L. 9.
R. 14.4

REDINTEGRATION IN ALBINO RAT — AJSTUDY IN RETENTION 31

c2^m

- to

f^

l(!l

/5 53

y^§Sl«jt--SOif") N

u

h
<

Q

« i

S 5$

o -2

o o o a s o 9

2 S 8 ? I S 8

GU

These results do not present an exhaustive study of
economy in retention. In learning, the question of economy
has not yet been carried to its limits; and until that time

82 THOMAS WILLIAM BROCKBANK

comes the question of economy in retention must remain
more or less suspended. The results of the present experi-
ment show conclusively that of the two distributions here
considered, namely, three trials per day and one trial per
day, the greater distribution in trials produces the greater
economy both in learning and after retention. The totals
given in the tables and the curves in the plates only in-
dicate that economy is present; but, as may be expected
from any statistical method, they do not give any physio-
logical explanation for the economy itself.

(3) Redintegration After a 30-Day Retention Period
(a) Retention Period for the Maze

The effects of the respective retention periods of 45 and
70 days, which have been considered in the preceding
pages, aid only negatively in the determination of the
approximate period during which the maze habit may be
retained with a precision approaching that of the norm
in learning. No rat, after either of the two retention periods
already considered, completed redintegration tests in the
learning norm of 15 trials. This would have required a
practically perfect retention of the habit and also its perfect
exercise in the first six trials of redintegration.

In order to ascertain the approximate period after which
integration appears perfect in the redintegration tests, it
was determined to experiment with a group of rats by
the one trial per day method and try redintegration after
a 30-day retention period. With this immediate end in
view, two litters, the first of three rats, namely, 1, 2, and 3
in Tables V-A and V-B, and the second of four rats, namely,
4, 5, 6, and 7 in the same tables, were set to learn the
maze. All the rats of both litters completed learning and
redintegration. The same norm and method was used with
this litter as with the groups already considered.

The " error " results of these seven rats are presented
in Table V-A, and the trial and time records in Table V-B.
In two rats of the seven, redintegration was as good or
even better than the learning; while a third rat required
but one more trial, or 16. This may be understood more

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 33

clearly by reference to Table V-B, records 1 and 6 being
perfect in redintegration, while in record 5 the first trial
of redintegration was was imperfect, and thus 16 trials
were required to complete the series. As individual dif-
ferences show, it seems quite probable that there are some
rats which would never exhibit perfect integration from
the very first of the redintegration series of trials even if
the learning period were excessively prolonged or the re-
tention period shortened. And, therefore, although redin-
tegration in the other rats of the group did not evidence
such perfect retention, the fact that two retained the
maze perfectly for 30 days seemed to indicate that this
period may be laid down as the approximate maximal
period of perfect retention for the maze. This was con-
firmed by a later experiment.

The records of this group of rats in Tables V-A and
V-B will not be considered separately to confirm what

TABLE V-A

30-Day Period. 1 Trial per Day

"Errors" At Turns

Last 25 Trials of Learning; Complete Redintegration

No. of

Turn

Start

1

2

3

4

5

6

Total

Trials

P M

P

M

P

M

PM

PM

PM

PM

P M

Rati.

L.
R.

2

2

3

1

6 2

1 4

3 4

1

4

13 20

74
15

2.

L.

5 1

1

1

5 3

39

R.

4

3

5

2

6 8

28

3.

L.

1

1

2

1 1

2 4

47

R.

4 1

6

1 7

1 1

1 1

7 16

30

4.

L.

4 1

5

4

1

1

10 6

74

R.

1

5

5

4

1

1 2

4

14 9

29

5.

L.

5

4

5 4

130

R.

1

2

3

1

1

4 4

16

6.

L.
R.

1

5

2
1

1

2

6 4
2

100
15

7.

L.

1 3

2

2

2

1

1

2 1

8 7

72

R.

Is

6

5

2

3

4 2

2

1 2

2

5

15 19

22

Tota

Qualit.

L.

11 7

19

16

3

4

8 7

3 5

4 5

1

4

49 48

R.
L.

9 3

13

23

6

4

6 17

8 3

2 3

2

5

46 58

Quant.

18

35

7

15

8

9

5

97

R.
L.

12

36

10

23

11

5

7

104

Aver's.

2.57

5

1

2.14

1.14

1.28

71

13.84

76.5

R.

1.71

5.

14

1.

42

3.45

1.57

0.71

1.

14.85

22.1

34

THOMAS WILLIAM BROCKBANK

TABLE V-B

30-Day Period. 1 Trial per Day

Time in Seconds

No

of Trials

First 15 Trials of R.; Last 15 Trials of L.

Av.

Rat 1.

L.

74

*15

*13

*11

*n

no

*14

n6

n9

*15

*10

*13

*16

*13

*16

42

15.6

R.

15

*13

*11

*11

*9

*ii

*13

*14

*22

*8

*8

*8

*11

*9

*7

*8

10.8

2.

L.

39

*6

*6

*8

*7

*6

*7

*6

*6

7

*7

*6

18

*6

31

*7

9.3

R.

28

35

*7

*7

8

*8

*7

33

21

*7

12

*8

*6

12

12

*7

12.6

3.

L.

47

*7

*6

*7

*9

*6

*7

*7

31

*7

*7

*6

*6

*7

*7

*7

8.4

R.

30

76

28

11

11

*7

8

*7

*6

*7

16

13

*7

*6

*7

7

14.4

4.

L.

74

*6

*6

*6

*6

*7

*8

*8

190

*7

*13

*8

*9

*13

*12

ne

21

R.

29

28

16

*20

*19

*14

38

23

*15

27

38

23

*16

*11

30

ni

21.9

5.

L.

130

*11

*19

*9

*12

*11

*13

*10

*12

*10

*11

*8

*9

*8

*9

no

10.8

R.

16

12

*7

*11

*8

*7

*6

*7

53

26

*9

24

*11

*8

*6

*6

13.4

6.

L.

100

*9

*11

*8

*8

*8

*9

*8

*8

*9

*10

* 7

17

98

46

17

18.2

R.

15

*9

*8

*10

*7

*7

*7

*7

*7

*6

*6

*7

66

*7

*15

*17

12.4

7.

L.

72

*7

*7

*7

*6

*6

*7

14

*7

*8

66

255

36

57

35

*14

35.4

R.

22

197

27

63

n3

255

n5

49

*23

*16

*8

*10

*12

*12

no

no

48

Averages —
Trials... L 76.5
R. 22.1

* (Perfect) Trials.. L. 12.7
R. 10.2

Time..L. 16.9
R. 19.

has already been proven concerning dominant " error,"
but will be used as a norm for comparison with a second
group, at 30-day retention, which will be taken up now.

(6) Influence of Learning a New Problem during the
Retention Period

This group was undertaken with the view of obtaining
some light in regard to the effect of the acquirement of a
new habit on the retention and redintegration of a habit
already acquired. The group comprised two litters, the
first of three rats, 1, 2, and 3, respectively, and the second
of seven rats, 4, 5, 6, 7, 8, 9, and 10, respectively, in Tables
VI-A and VI-B. Rat 3 of litter 1 failed to complete
redintegration, while rat 10 of litter 2 died before com-
pleting retention. Each litter was set to learn the maze
by the one trial per day method. The object of the present
problem required that at the beginning of the retention
period, instead of following the method of exercise in the
runway during retention, each rat was to begin the learning
of a new problem. This new problem was the " rope-
ladder " problem. As the name suggests, it was constructed
of first, a rope about 1 cm. in thickness and 93 cm. in

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 35

length, fixed at one- end to a broad topped upright, 46 cm.
in height, from which the rat started; and at the other
end to a second upright of the same height as the first.
From the second upright, a ladder, 87 cm. in length and
8 cm. in width, with rope sides and wooden crosspieces,
extended upwards at an angle of approximately 45 degrees
to third upright 100 cm. in height, where food was placed
as the incentive. The pathway was approximately 190 cm.
in entire length. The learning of the rope ladder is of
little concern to us here. Suffice it to say that the integra-
tion of movements required in running the rope and ladder
in succession and maintaining equilibrium on them was
difficult to acquire. No single rat of the group under
consideration learned this problem perfectly in the 30-day
retention period allowed.

At the end of the retention period the rats were not
given preliminary feeding in the center of the maze but
were put directly into the starting box of the maze and the
first trial of redintegration was made. Thus it is to be
noted that with this group the former norm and method
was followed with the single exception that the learning
of a new problem during the retention period took the
place of daily exercise.

Tables VI-A and VI-B contain the complete records of
this group. In Table VI-A, the last 25 trials of learning
are compared with redintegration totals as is the case in
Table V-A. The compilation of Tables VI-B and V-B was
made in the same manner as the earlier tables of this
nature. Coming to a consideration of the results of the
present group. Table VI-A, compared with the records
of the norm group at 30 days' retention, contained in
Table V-A, it is evident that there is no increase in number
of " errors " in redintegration with the group learning a
new problem during retention, over that of the norm.
In fact the number of " errors " in redintegration in Table
VI-A is the lesser, relative to the learning total, while the
number of " errors " in redintegration in Table V-A is
the greater, relative to the learning total. In Tables V-B
and VI-B, it may be seen that the learning of the new
problem in the retention period has not interfered in point

36 THOMAS WILLIAM BROCKBANK

TABLE VI-A

1 Trial per Day

30-Day Period, During Which A New Problem Was Learned

" Errors " At Turns

Last 25 Trials of Learning; Complete Redintegration

No. of

Turn

Start

1

2

3

4

5

6

Total

Trials

P M

P M

P M

P M

P M

PM

PM

P M

Rat 1. L.

4

1

1

1 5

114

R.

1 1

1

2

2 3

21

2. L.

1

4

1

1

2 6-

106

R.

1

2

2

3

5

1 1

8 7

15

*4. L.

3

1 2

1

4 3

74

R.

.15

5. L.

1

2 3

1

3 4

65

R.

1

1

10

6. L.

1

1 1

2

4 1

61

R.

1

2 4

3 4

16

7. L.

1

4 6

1 1

1 5

1

6 13

69

R.

1

1

1 1

2 2

16

8. L.

1 2

4

1 1

3

2 1

7 8

68

R.

1

1

1 1

16

9. L.

2

1

1

2 2

56

R.

3

3 3

1 5

4 4

3 2

5 3

19 17

29

Totals

Qualit. L.

9 3

8 25

4 3

2 8

4

1

2 2

29 42

R.

5 2

7 11

3

2 12

9 4

3 2

6 4

35 35

Quant. L.

12

33

7

10

4

1

4

71

R.

7

18

3

14

13

5

10

70

Aver's. L.

1.5

4.12

0.67

1.25

0.50

0.12

0.50

8.87

R.

.87

2.25

.37

1.75

1.62

.62

1.25

8.75

*3 Omitted, due to incomplete records of " errors " in learning.

of trials, nor in average number of perfect trials, nor in
speed of movement. On the contrary, it seems from the
results that the rats have been benefited in retention by
the learning of the new problem during the retention
period. A summary of these conclusions is presented in
Plate III, the curves of which are constructed from the
data in the four tables of the two groups under considera-
tion. The norm group, according to these curves, is inferior
to the group which learned a new problem during the
retention period.

In considering the results of the present experiment
where a new problem was learned during the retention
period, there is on the one hand the fact that at the be-

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 37

5
u

_i

CD
O

a:

^S
u o

^ CL

i|

< I-

UJ Z

-• lij

li. »-

o uj

I Z tD

!<-?
a. z o

u

\

flsWK***^ 5^5 5»><oe~^» 'o*'nf<-»

5 $ 5 Si i

'=CC

<o ki

^ <3 C

o g o

^ S IT)

ginning of such an experiment the ordinary working hypoth-
esis would tend to anticipate the appearance of an ad-
verse effect in redintegration. While on the other hand
the fact is presented by the results, as stated above, that
a greater number of the individuals of the group which

38

THOMAS WILLIAM BROCKBANK

TABLE VI-B
1 Trial per Day
30-Day Period, During Which A New Problem Was Learned
Time in Seconds

No.

of Trials

First 15 Trials of R.: Last 15 Trials of L.

Av.

Rat 1.

L.

114

*9

*15

*11

*14

*9

*9

n2

*11

ni

n2

no

n2

54

no

*21

14.6

R.

21

35

*9

*8

*8

35

17

*12

*11

*9

n2

ni

*14

14

*7

*8

14.

2.

L.

106

*7

*7

*7

*7

*7

*8

8

*7

36

*9

*7

*9

*7

*7

16

9.9

R.

15

*20

*8

*8

*8

*8

n2

53

21

no

16

36

15

17

*9

no

16.6

3.

L.

48

*6

*6

*7

*6

*6

*6

*6

*6

*6

13

*7

*7

*6

*6

*6

6.6

R.

16

14

*6

*7

*7

*9

*6

*6

*6

*7

*7

*6

*6

*6

*6

*6

7.

4.

L.

74

*7

*7

*8

*6

*7

*7

17

*7

*6

*6

*6

*7

*6

24

n6

8.4

R.

15

*10

*7

*6

*7

no

*8

*6

*7

*8

*8

*7

*9

*7

*7

*9

7.7

5.

L.

65

*8

*7

*8

*8

*ii

*7

*8

*6

n2

ni

no

13

*7

no

*7

8.8

R.

19

*9

*9

*8

20

*8

*7

*7

*6

*8

*8

*7

*6

*8

*7

*7

8.3

6.

L.

61

*11

*7

*7

*8

*8

*7

*8

*7

*7

*7

*7

*7

*6

*7

*9

7.5

R.

16

27

*7

*7

*7

*8

*7

*7

*7

*7

*6

*7

*8

*6

*6

*7

8.2

7.

L.

62

*11

*8

*12

*13

*15

*9

*10

*20

n8

*9

*8

ni

*9

*9

*20

12.3

R.

16

46

*11

*7

*9

*8

no

*7

*7

*7

*7

*9

*6

14

*6

*7

10.7

8.

L.

68

*47

*22

*22

*11

*21

*14

*10

*39

*14

*38

n9

n5

61

*29

n7

25.6

R.

16

44

*10

*13

*11

*11

*9

*19

*9

n2

ni

n3

*14

no

no

*8

13.6

9.

L.

56

*8

*8

*8

*7

*7

*8

*7

*7

*8

*8

*8

*8

*9

13

*14

8.6

R.

29

*13

12 11

15

*7

*9

*14

347

249

n9

n8

*9

no

11

*8

50.1

Averages —

Trials .

L.

72.6

* Trials.. L. 13.8 Time.X. 11.3

R.

18.1

R. 12.4 R. 15.1

learned a new problem during retention were more nearly
perfect in redintegration than of the group which repre-
sented the norm. The learning of the rope-ladder problem,
in this instance, has not interfered with the retention of
the acquired habit, and it would seem that no interference
would result from the learning of any similar problem
during the retention period. There is a possibility, however,
that some problem different in nature from that of the
rope ladder, and more difficult in point of the establishment
of integrations may be detrimental to the retention of
the maze habit. But this can be proven only by further
experiment.

(c) Influence of Incomplete Learning

The individuals of every group already considered have
completed learning and redintegration according to the
standard norm of fifteen trials, as already explained. Al-
though, according to this norm, any of the last nine trials

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 39

may or may not have exhibited perfect integration, yet
it was necessary for each individual to have exhibited
perfect integration of movement in at least the first six
trials of the norm. From these facts it is evident that
although at the end of the fifteen trials of the norm it is
possible that no two individuals began retention with the
necessary integrations established in the same degree of
precision, yet all acquired the habit, as was shown in the
first six perfect trials of the norm, and all underwent
experiment according to the same standard.

Experiments with the present group of rats were under-
taken in the attempt to ascertain some influence of in-
complete learning on redintegration after retention. Two
litters composed of five rats each were chosen for the
present work and set to learn the maze by the one trial
per day method. The standard norm and methods were
applied to both litters, except that when the norm of fifteen
trials had been completed by the first rat of each litter
the whole litter was put on retention regardless of how
far the learning had progressed with the remaining four
rats. The same method was employed in redintegration.

Tables VII-A and VII-B contain the records of these
two litters in " errors," trials, and time. In Table VII-B
it may be seen that, in litter 1, rat 5 was the first to com-
plete learning, in S3 trials; but in redintegration, rat 2
was the first to finish, in 15 trials. In litter 2, rat 9 was
the first to finish learning, in 25 trials; but in redintegra-
tion, rat 10 was the first to finish, in 15 trials. At first
sight it may appear somewhat questionable that in both
cases cited the individual which finished first in learning
according to the norm did not finish first in redintegration.
But this fact simply points out that the rat which learned
the problem first according to the norm seems not to have
been the one which had most firmly established integra-
tion during learning, and hence was not the best in redin-
tegration. The apparent discrepancy in the redintegration
records must be solved, at least in part, from the records
of the learning.

Considering first the learning of litter 1, in Table VII-B
(supplement), it may be seen, first, that three rats of the

40

THOMAS WILLIAM BROCKBANK

TABLE VII-A

30-Day Period. 1 Trial per Day

Norm: When The First Rat Finished

" Errors " At Turns

Last 15 Trials of Learning; Complete Redintegration

Start

1

2

3

4

5

6

Total

Litter 1. .

Rati. L.

R.

2. L.
*R.

3. L.
R.

4. L.
R.

5. *L.

R.
Litter 2.

6. L.
R.

7. L.
R.

8. L.
R.

9.*L.

R.

10. L.

*R.

P M

2

1

2
1
8

1
1
2

7

n

1

1 1

6

7

P M
3

1

2

1

2 2

2

1

1 3

2

2

4 1

1

3

P M

1
1

1 1

1

1 1

3
1
1 1

P M

4
1 1

1

9

1 1

1

1

2

• 4

P M

1
1

1
1

4 3

P M

2

1

P M

P M
3 8
2 2
1 1

1 1
4

2 2

3 10
2 6
1 2

1 3

2 10

7

10 8

1 2

5 8

1
1

14

Totals
Quant. L.
R.

4 32
1 10

7 17
2 5

5 2
3 3

2 11
1 5

6 3
,2

*2 1

26 66
9 23

Quant. L.
R.

36
11

24

7

7
6

13
6

9
2

3

92
32

Avar's. L.
R.

3.6
1.1

2.4
.7.

.7
.6

1.3
.6

.9
.2

.3

9.2
2.3

* First to complete norm.

five had 13 perfect trials in the last 15 trials of learning;
second, that rat 2 is the best in redintegration in point
of perfect trials; and finally that rat 5 is the only rat which
did not make as many or more perfect trials in redintegra-
tion as in learning. It may be inferred from these facts
that, first, either rat 5 did not retain as well as the
other rats due to innate constitution, to some metabolic
distrubance during retention, or to some extraneous dis-
turbance; or second, at that particular time, when rat 5
completed the norm of learning, perfect integration was

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 41

TABLE VII-B
30-Day Period. 1 Trial per Day
Norm: When the First Rat Finished
Time in Seconds

No.

of Tnals

First 15 Trials of R; Last 15 Trials of L.

Av.

Litter

1

Rati.

L.

10

*8

*7

*7

12

10

*7

7

12

13

*7

*9

8

*7

*7

8.73

R.

*n

*12

*7

*7

*8

9

*8

*7

*6

*6

9

*6

10

*8

*8

8.13

2.

L.

*6

*6

*6

*6

10

*7

*7

*6

6

*6

*6

*6

*6

*6

*6

6.4

**R.

15

*20

*7

*7

*10

*9

*9

15

*7

*6

*7

*6

*6

*6

*7

*6

8.53

3.

L.

15

*6

*6

*6

*6

*6

*6

*6

*6

*6

*6

*6

*6

*6

*6

6.66

R.

20

*40

*11

*13

*7

*7

18

*7

*6

*7

*7

*6

*6

*6

*6

9.8

4.

L.

*6

9

10

*8

*6

10

15

25

15

*9

*7

*7

15

92

no

16.26

R.

51

11

*8

*11

*8

*7

27

*8

*7

*7

*7

*6

8

13

*7

12.4

**5.

L.

33

*7

*7

*7

*7

*6

*8

*7

*6

*7

12

*7

*7

8

*7

*6

7.93

R.

17

*7

9

no

*8

10

*8

*8

*7

*7

*6

*6

*6

*6

*7

8.13

Litter

2.

6.

L.

17

26

*6

*6

9

8

*7

8

*6

*7

10

*6

*6

9

9

9.33

R.

18

*9

*8

*7

*7

*9

*7

8

16

10

9

*7

11

9

*7

9.46

7.

L.

15

*6

*8

46

*24

14

n2

36

55

47

74

*27

72

65

29

35.33

R.

16

*11

14

no

no

*9

*8

14

*9

*9

*7

*7

*6

*6

*7

9.53

8.

L.

13

9

9

11

9

8

13

21

17

*8

*7

*7

*6

10

*9

10.46

R.

8

*7

*7

*6

*6

*7

*6

*7

*6

*6

*6

*6

*6

*6

*6

6.4

**9.

L.

25

*11

no

*6

*8

*7

*6

*9

9

n2

*8

ni

*7

n2

n2

*20

9.83

R.

*9

*7

15

*6

*7

*6

*6

ni

*6

*6

*7

*6

*6

*8

*7

7.53

10.

L.

27

15

*7

no

29

21

42

47

*4

17

22

*9

*9

n7

*14

20

A

^*R.

15

*14

*7

i*12

*9

ni

*9

*9

no

*6

*7

*7

*6

*6

*8

*7

8.53

**First to complete norm.

Av. * Trials . . L. 9.3 Av. Time per Trial . . L. 13 . 09
R. 12.4 R. 8.84

Supplement.— No. of * Trials

Litter 1. Rat— 1. 2.
L. 8 13
R. 12 14

3. 4. 5. Litter 2. Rat— 6. 7. 8. 9. 10.
13 7 13 L. 7 5 5 14 7
13 10 12 R. 8 12 14 14 15

not as firmly established in this rat as it may have been
in rat 2 and, as a consequence, rat 2 would be more likely
to exhibit perfect integration than rat 5. Although it is
possible that the first inference may be a fact, there is no
data at hand to prove it, while there is some proof in
Table VII-B in favor of the second view. Thus it may
be noted that the last trials of learning in the records of
rat 2 show perfect integration and maximum speed. There
is no other record at this particular time in the learning
which shows so much evidence in favor of the inference
that at the end of the norm perfect integration at the
locus of dominant " error " was more firmly established in
any other rat than in rat 2. Metabolism, extraneous

42 THOMAS WILLIAM BROCKBANK

disturbances, and innate constitution may have had some
influence in the retention, but as already stated there are
no means at hand to prove their influence in the present
case.

The redintegration records of rats 1 and 4 could almost
be taken as a direct continuation of the learning, were it
not known that retention had intervened; redintegration
was practically perfect for both these rats but the norm
in both cases had not been reached. The redintegration
of rat 3 is rather difficult to explain when compared with
the learning. It appears that the habit had been fairly
well established, excepting the " error " in the next to
the last trial; and thus it seems that the evidence in this
case, more so than in any other yet treated, points to poor
retention on the part of the organism.

In litter 2 (Table VII-B), the final records of learning in
rats 9 and 10 show that the integrations had been more or
less firmly established by both rats at the completion of
the norm. In the first two trials of redintegration the
records of both rats, in integration of movement as well
as speed, are equal if not superior to their records in the
last few trials of learning. Rat 10 continued the perfect
exercise of the habit throughout redintegration, but rat
9 in trial three of redintegration exhibited in alley 2 im-
perfect integration, due to an extraneous disturbance.
Had it not been for this disturbance, it seems most likely
that rat 9 would have had a perfect redintegration, as
is credited to rat 10. Rats 9 and 10 both had the habit
firmly established at the beginning of retention. Rats
6, 7, and 8, as those of a similar status in litter 1, had not
firmly established integration, but retained the habit so
that it was exercised as well after 30 days' retention as
when learning ceased. This was likewise found true of
rats 1 and 4 in litter 1, and to a somewhat less degree in
rat 3.

It may be concluded from this experiment that in most
cases, and under ordinary circumstances, the better estab-
lished is integration in the organism when the learning
ceases and the retention period begins the better will be
the redintegration.

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 43

(4) Influence of Previous Training at 45- Day Retention Period

As a final experiment in retention on the maze, it was
decided to give a group of rats some definite preliminary
training before setting them to learn the maze. Following
this plan two litters of six and four rats, respectively, were
set first to learn the inclined-plane problem. (This problem
is treated more fully on page 48, and for this reason it will
not be considered here in detail. Only the important
points of all preliminary training will be mentioned.)
Rats 1, 2, 3, 5, 7, and 9 comprise the litter of six; 4, 6, 8,
and 10 the litter of four. Rats 5 and 7 of the litter of six
and rat 10 of the litter of four failed to complete the whole
experiment. Concerning the remaining rats, 1, 2, 3, 4,
6, 8, and 9, it is a fact of importance here to note that
three — rats 2, 3, and 4^ — failed to solve the inclined plane
once after fifteen successive days. This point of informa-
tion is mentioned here specifically to show that the rats
used in this experiment were far from being above the
average; in fact, if we may judge from the above fact,
they were below the average, because the percentage of
failures on the inclined plane is usually low.

At the completion of the inclined-plane problem, a
definite period of 70 days was allowed to elapse in the
case of each rat before the maze was begun. The last
3 days of the 70-day period was used to test the redintegra-
tion of each rat on the inclined plane. During the 70-day
period each rat was required to continue the learning of
one or more problems, the rope-ladder problem being first,
and the sawdust box second. The sawdust box was a
problem which necessitated the digging of a tunnel through
the sawdust before the rat could reach its food.

Following the usual method, at the completion of a
problem each rat was fed in the next problem for four
days to accustom it to the environment. Rats 1 and 4
failed to establish perfect integration on the rope ladder,
but rats 2, 3, 4, 8, and 9 had learned the rope ladder and
had begun the learning of the sawdust box when the re-
spective dates for their preliminary training had ended.
Rats 1, 6, 8, and 9 which had learned the inclined plane
were then tested in redintegration on the plane, while

44

THOMAS WILLIAM BROCKBANK

the other rats continued the learning of their respective
problem until the end of the 70-day period. At the com-
pletion of all this preliminary training, each rat was fed
for four days in the center of the maze, and the methods
of learning, retention, and redintegration, as originally
described in the chapter on methods, were carried out.
The same norm was also applied in learning and redin-
tegration.

Tables VIII-A and VIII-B contain the records for this
group of rats. For the sake of comparison a summary
of the records of the rats in Tables IV-A and IV-C are
given, and also a summary of the records of Table V-C,
but not of Table V-A, owing to the fact that the records
of " errors " in this table are not sufficiently complete to
allow comparison with the present group. These tables

TABLE VIII-A

45-Day Period. 1 Trial per Day

" Errors " At Turns

With Previous Problems.*

From 15th Trial of Learning; Complete Redintegration.

No. of

Turn

Start

1

2

3

4

5

6

Total

Trials

PM

PM

P M

P M

P M

P M

P M

P M

Rati. L.

2

2

2 2

38

R.

1

3

4

17

2. L.

5

5

29

R.

1

1

4

6

15

3. L.

2

1

2 3

34

R.

11

1

2

1 1

15

4. L.

7

3

4 1

9

1

12 13

64

R.

1

1

3

3 2

20

6. L.

1

2

3

40

R.

1

2

1 2

16

8. L.

1

5

3

5

2

3 3

9 13

58

R.

1

1

2

18

9. L.

1

1

1

5

3

5 6

39

R.

3

1

1 3

24

Totals

Qualit. L.

13 1

10

5 1

23

14

2

3 3

35 40

R.

2

4

1 1

1 8

8 1

12 14

Quant. L.

14

10

6

23

14

2

6

75

R.

2

4

2

9

9

.

26

Aver's. L.

2

1.4

.85

3.28

2.

.28

.85

10.7

43.1

R.

.28

.55

.28

1.28

1.28

3.7

17.8

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 45

(IV-A, IV-C, and V-C) contain the records of rats at
the 45-day retention period, and also at the 30-day reten-
tion period, hence afford a norm both of learning and

" Error " Totals From Table IV-A

45-Day Period. 1 Trial per Day

From 15th Trial of Learning; Complete Redintegration

Totals

Qualit. L.
R.

PM

114 92
20 6

PM

32 50
8 15

P M

16 13
6 6

P M

7 53
4 13

P M

26 3
3 2

P M

2
3

P M

5 4

1

P M

200 217
41 46

Quant. L.
R.

206
26

82
23

29
12

60

17

29
5

2
3

9
1

417
87

Aver's. L.
R.

29.4
3.7

11.7
3.2

4.1
1.7

8.5

2.4

4.1
.7

.28

.42

1.2

.14

59.5
12. 4

75.7
26.2

*A11 individuals of this group learned the inclined plane first. After this problem
a 70-day period was allowed to elapse, when the maze was begun. During the
70-day period each rat was required to learn one or more new problems, the rope
ladder being first, the sawdust box second. During the 45-day period after the
learning of the maze, the rats were exercised in the runway, as outlined in the chapter
on methods.

TABLE Vlll-B

45-Day Period. 1 Trial per Day

With Previous Problems

Time in Seconds

Noi of Trials

First 15 Trials of B

..; Last 15 Trials of L.

Av.

Rat 1.

L.

38

*10

*8

*7

*8

*7

*7

*8

9

*7

*7

*9

*8

*8

*7

*8

7 86

R.

17

11

*9

*7

*6

*7

*7

*7

*7

*8

*8

*8

*7

*8

no

*7

8

2.

L.

29

no

no

no

ni

*9

*17

*8

no

*9

*8

*9

n6

10

*8

*9

10 27

R.

15

*20

no

*8

*8

*9

*9

*8

*7

*9

*8

*8

8

12

*8

*8

9,33

3.

L.

34

*7

*6

*7

*7

*7

*7

*8

*7

*7

*7

9

7

*8

*7

*7

7.2

R.

15

*8

*7

*7

*6

*9

'*7

8

*6

*7

*7

*7

*7

*7

*7

*7

7.13

4.

L.

64

*9

*9

*9

*9

*9

*8

*9

*8

*8

*9

*7

*7

10

*8

*9

8.53

R.

20

46

*8

*8

*8

14

*9

*8

*8

*8

*8

*8

*9

*8

*8

*8

11.06

6.

L.

40

*7

*7

*7

*7

*7

*7

*7

*6

*7

*8

*7

*7

*7

*7

*7

7.

R.

16

15

*7

*6

*6

*7

*7

*6

*7

*8

*6

*7

13

*7

*7

*8

7.8

8.

L.

58

*9

*9

*8

*9

*8

*7

*9

*9

*9

*9no

*8

*8

*8

*9

8.6

R.

18

*9

*8

25

*9

*8

n2

no

n5

*8

*8

*7

*8

*7

*6

*7

9.7

9.

L.

39

*8

*7

*9

*8

*8

*6

*7

*7

*7

*7

*7

*8

*8

*8

*8

7.53

R.

24

*11

14

*8

*7

9

*8

*7

*7

10

*7

n2

*8

*7

*7

*8

8.66

Averages —
Trials.. L... 43.1
R. 17.8

Averages —
Trials.. L. 75.7
R.26.2

Av. * Trials.. L. 14.2
R. 13.4

Av. Time per Trial . .L. 8. 14
R.8.81

From Table IV-C
45-Day Period. 1 Trial per Day

Av. * Trials.. L. 10.5
R.(9.55

Av. Time per Trial ,

.L. 9.02
R. 14.42

46

THOMAS WILLIAM BROCKBANK

Averages —
Trials.. L. 76.5

R.22.1

From Table V-C
30-Day Period. 1 Trial per Day

Av. * Trials.. L. 12.7
R. 10.2

Av. Time per Trial . .L. 16
R. 19.

^ -<

o ^ t;

c

I

o

Z

Z

I

h

CO

Z)

o

>
U
Ct
Q.

kj

Q- 2

u

-J

o tt. t

o

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 47

redintegration by which results of the group under con-
sideration may be judged. A summary of the results of
the present group and the 45-day norm group is contained
in Plate IV.

In considering results, it may be noted particularly that
in the learning records of the rats of the present group
which had previous training, the average " errors " show
that this group had far fewer " errors " than the 45-day
norm group; and this fact follows with consistency in
redintegration. It is evident in the comparison of Tables
VIII-B with Tables IV-C and V-C that the learning
of the rats with previous problems was complete in far
fewer trials than the norm, one trial per day group. And
in redintegration the average number of trials of the group
under consideration is not only less than the 45-day norm
group but less than that of the group with 30-day retention.
In average perfect trials and in average time per trial the
superiority of the group with previous training is shown
not only in learning but also in redintegration; and further,
what is most significant, the superiority holds in this in-
instance, not only in comparison with the 45-day norm but
also with the group at 30 days.

In considering the individual records the fact is brought
out that the approximate retention period for the maze
is increased in the group with previous training. This
may be seen when one notes that of the group with previous
training there were two perfect records in redintegration,
as recorded in Table VIII-B, while not one is recorded
perfect in Table IV-C of the group without previous train-
ing. A further fact is evident, namely, that after the
retention period of the individuals with previous training,
the best individual record is far superior to any record
of the norm group and equal to the best of the group with
30-day retention period, while the poorest record of the
group with previous training is superior to the majority
of the records of the 45-day norm group and some of the
records of the 30-day group.

From the results of these tables the facts are thus
established: First, for learning, that previous training so
affects the rat that the subsequent acquirement of a new

48 THOMAS WILLIAM BROCKBANK

habit is less difficult; second, there are fewer " errors "
in the learning, and integrations are more readily estab-
lished. As a consequence, redintegration must neces-
sarily be improved. In other words, when a rat has passed
through a relatively long period, acquiring integrations,
and then is set to learn a standard problem, the process
of learnng the integrations necessary for this problem will
proceed with much more rapidity than with the rat that
has had no previous training, and the redintegration as a
consequence will be better than with the untrained rat,
not because of better retention — for retention is probably,
not bettered by learning — but because of the fact that
the learning is better than ordinary.

(5) Inclined Plane

The standard norm and methods which were described
in the foregoing experiments on the maze were adopted
for the present experiment. These comprise the feeding
of the respective groups in the problem for a half hour
on each of four days preceding the first day of learning,
the exercise in the runway each day during the retention
period, and the requirement of the fifteen-trial norm, the
first six trials of which must be perfect. Three seconds
is the usual time set for the exercise of the habit in the
present problem, but this time norm was not rigidly
adhered to, the same flexibility being allowed as in the
maze, providing that perfect integration of movement was
exhibited.

The inclined plane (Fig. Ill) was a problem made up
essentially of a box to which access was gained through a
small door that opened on the pressing down of an inclined
plane located at the rear of the box. The inclined-plane
box was 25 cm. square. In the lower center of one side of
the box was a door, 11.2 cm. square. The box and also
the door was covered with wire netting of 1 cm. mesh.
A solenoid coil was so placed above the door that its arma-
ture projected approximately 0.3 cm. over the edge of the
door, holding the door shut against the force of a small
spiral spring. The plane itself was 7.5 cm. long, 4.5 cm.
wide, and was hinged to the plate beneath, which was in

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 49

turn imbedded in the base of the problem. The plane
was raised at an angle 15 degrees, approximately, and was
held in this position by means of a spring beneath. The
weight of a rat stepping on the plane was sufficient to
force the plane down far enough that a contact would
be made under the plane, thereby raising the armature
in the solenoid coil and the door would open inward. The
plane and plate were constructed of aluminum but were

Figure III

insulated by wood fiber. The plane was located 12 cm.
behind the problem box.

The plane and box were placed on a base 92 cm. square,
inside measurement, and were covered by a glass top,
68 cm. from the base, and glass sides in a wooden frame-
work. The object of the glass sides was to prevent excess
waste of movement and effort which always occurs in
problem boxes with a mesh covering. Because the glass
sides limited the rat's activities to the floor of the problem
box, there resulted a greater conservation of movement

50 THOMAS WILLIAM BROCKBANK

and effort, and concentration of these on the problem itself.
The glass sides were firmly fixed in grooved wooden up-
rights located at each of the four corners of the base of
the problem box. They could easily be slid upw^ard, and
access readily gained to the problem box within. The
entrance to the problem box was located at a distance
of 28 cm. directly in front of the door of the inclined-plane
box. This entrance box was constructed of wood with a
wire-mesh top.

In the learning of the inclined plane-problem the be-
havior of the rat appeared more complex than in the
learning of the maze. This point is well worthy of notice.
The present problem is one requiring a habit of manipula-
tion to solve it, while the maze evokes a general motor
habit. Motor habits are in general the simplest habits
in the entire repertoire of the rat's movements, while habits
of manipulation require more complex integrations of
movement in their process of establishment. The maze
evokes a habit requiring the integrating of movements of
running in a certain direction and turning the whole body
at certain intervals in response to the environment. The
inclined plane likewise calls forth the motor movements
of running and turning the body at certain stages; but
it calls forth the additional integration of certain definitely
adjusted movements of the body and particularly the
forelimbs in the pressing down of the inclined plane. Sim-
pler methods of pressing down the plane may be used,
such as stepping on the plane in running over it; but
the method of pressing down the plane with the forelimbs
is the most satisfactory. Although these movements may
be included under motor movements in the widest sense,
yet they are movements of manipulation in the strict
sense and are much more complex and difficult of adjust-
ment than the movements of running, etc. Thus the
establishment of integration in the learning of the inclined
plane is much more difficult than the establishment of
integrations in the learning of the maze.

As in the first trial on most new problems, the rat usually
entered slowly and investigated every accessible place in
the problem. If it so occurred that in the first trial the rat

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 51

stepped on the plane and thus opened the door to the
food, this action may be said to have been a response
to the stimuH which the plane presented to the rat and
no more a matter of chance than the walking of the rat
in any other part of the inclosed area. Similarly in the
first few trials the solution of the problem can neither
be attributed to chance on the one hand nor to any estab-
lished integration on the other. Gradually, at approxi-
mately the 8th or 9th trial, response to the plane became
established as the stimulus to the successive integration
of going direct to the food box. Whether the establishment
of the integration — going to the plane preceded the estab-
lishment of the integration — coming from the plane to
the door of the food box can not be definitely stated. But
the fact is certain that these integrations as a whole com-
prised the first part of the learning, while the linking of
the two by the establishment of the integration of pushing

The most difficult integration to acquire in the learning,
and likewise the redintegration of the inclined plane, is
that of pushing down the plane. Invariably this move-
ment appears to be what has been already designated
as the " weak link " in the chain which constitutes the
successive integrations of the habit. When the integrations
of this movement are beginning to be established, and
when precision of movement first appears in the pushing
down of the plane, it may then be observed that the learning
of the problem has been delayed by this difficulty of push-
ing down the plane. The mass of qualitative data on
the learning of the plane problem confirm these facts.
In redintegration a similar situation presents itself. All
responses which the problem calls forth may be exhibited
on the first trial of redintegration, but frequently the
stepping on the plane is imperfectly integrated and the
habit can not be perfectly exercised. And, therefore, a
confirmation is here found ""for the results which were
noted in the study of the maze, namely, dominant " error "
appears in the process of learning and also in redintegration.

Dominant " error " and imperfect integration can not be
shown in tabulated form for the inclined plane as they
were shown for the maze. But Table IX, of the three

52 THOMAS WILLIAM BROCKBANK

trials per day group, and Table X, of the one trial per day
group, presenting records of time, will serve to indicate
to some extent the effect which was produced by the diffi-
culty of establishing integration at the locus of dominant
" error." Each table presents the complete time records
of three rats, the first three columns (a, b, and c) containing
records of the poorest individual of the group, the second
three columns containing records of the best, and the last
three columns containing records of the second best. In
the respective columns, records 2, 4, and 8 in Table IX
correspond to records 2, 4, and 8 in Table XI; records
5, 9, and 8 in Table X correspond to records 5, 9, and 8 in
Table XII. Column "a" presents time in seconds from the
moment the rat entered the problem until it stepped on
the plane; column " b " presents time in seconds from the
moment the rat stepped on the plane until it entered the
food box; and column "c" is the total time. The effect of
the dominant " error " may be seen in column "a" where
the time is considerably greater than in column " b." The
distance from the entrance of the problem to the plane
is approximately equal to the distance from the plane to
the entrance of the food box; or, in other words, the distance
the rat must cover to produce the time of column "a" is
approximately equal to the distance the rat must cover
to produce the time in column " b." But the rat must ex-
ercise the integration of stepping on the plane before the
time of column "a" is recorded; and the difficulty of exer-
cising this integration of stepping on the plane frequently
called forth such responses as merely touching the plane,
going round the plane, etc., which greatly increased the
time of column " a." These records of time thus give some
intimation that the locus of dominant " error " is at the
plane.

Table XI contains perfect trials and time in the first
fifteen trials of redintegration and the last fifteen trials of
learning for the three trials per day group at the 70-day
retention period. Table XII contains a like summary for
the one trial per day group at the 70-day retention period.
One rat of each group failed to complete learning; and
three of each group failed to begin the norm in redintegra-
tion within 30 trials. According to these tables the averages

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 53

TABLE IX
70-Day Period. 3 Trials per Day
Time in Seconds.

Rat-

8

2

4

Trial

a

b

c

Trial

a

b

c

Trial

a

b

c

1

632

40

672

1

380

38

418

1

364

108

472

2

1800

2

181

14

195

2

16

58

74

3

3

54

9

63

3

158

59

217

4

848

39

887

4

136

6

142

4

193

14

207

5

1800

5

149

6

155

5

150

25

175

6

6

30

3

33

6

309

36

345

7

1800

7

68

4

72

7

100

19

119

8

8

12

2

16

8

122

10

132

9

9

4

3

7

9

291

15

306

10

1098

86

1184

10

15

4

19

10

81

10

91

11

123

16

139

11

42

2

44

11

50

8

58

12

123

13

136

12

16

2

18

12

33

7

40

13

1546

25

1571

13

5

2

7

13

74

5

79

14

616

5

621

14

2

2

4

14

19

5

24

15

576

14

590

15

2

1

3

15

41

6

47

16

146

10

156

16

8

2

10

16

31

5

36

17

55

4

59

17

2

3

17

34

4

38

18

24

7

31

18

11

12

18

9

3

12

19

7

9

16

19

8

10

19

20

4

24

20

24

3

27

20

2

3

20

16

2

18

21

218

5

223

21

6

7

21

27

9

36

22

570

6

576

22

14

15

22

14

3

17

23

38

4

42

23

2

3

23

35

4

39

24

18

3

21

24

2

3

24

31

7

38

25

69

3

72

25

2

3

25

16

5

21

26

25

11

36

26

1

2

26

2

2

4

27

14

5

19

27

1

o

27

12

4

16

28

4

3

7

28

2

3

28

2

6

8

29

4

2

6

29

3

4

29

6

5

11

30

5

2

7

30

2

3

30

7

2

9

31

32

4

36

31

2

3

31

15

4

19

32

3

2

5

32

1

1

2

32

10

3

13

33

2

2

4

33

1

2

33

2

6

8

34

57

2

59

34

9

10

34

22

2

4

35

2

1

3

35

9

3

35

2

1

3

36

3

2

5

36

2

3

36

2

1

3

37

4

4

8

37

8

9

37

19

10

29

38

12

9

21

38

8

2

10

39

8

2

10

re

DINTEG

ratic

)N

39

11

3

14

40

5

1

6

40
41

12

1

2
1

14

41

1

12

13

2

42

2

1

3

2

2

3

42

2

4

6

43

100

2

102

3

5

6

43

13

2

15

44

35

2

37

4

1

2

44

3

4

7

45

14

1

15

5

2

3

45

4

2

6

46

20

2

22

6

2

3

46

10

1

11

47

103

3

106

7

2

3

47

3

1

4

48

10

1

11

8

2

3

48

9

2

11

49

46

2

48

9

1

2

49

7

1

8

50

31

1

32

10

8

9

50

3

2

5

51

12

2

14

11

3

4

51

2

1

3

52

40

3

43

12

4

5

52

5

1

6

53

10

2

12

13

4

5

53

7

1

8

54

THOMAS WILLIAM BROCKBANK

TABLE IX— {Continued).
Time in Seconds

Rat-

8

2

4

Trial

a

b

c

Trials

a b

c

Tnal

a

b

c

54

9

3

12

14

2

3

54

10

2

12

55

17

2

19

15

2

3

55

12

2

14

56

8

2

10

16

1

2

56

2

2

4

57

82

2

84

17

1

2

57

4

4

8

58

64
9

2
2

66
11

18

2

3

58
59

5

4

1
3

6

59

7

60

7

2

9

8

(concluded

60

5

2

7

61

31
5

3

1

34
6

61
62

7
8

2
2

9

62

111

100

3

103

10

63

11

1

12

112

202

2

204

63

2

1

3

64

52

2

54

113

130

2

132

64

2

1

3

65

32

2

34

114

89

2

91

65

2

1

3

66

19

2

21

115

66

30

96

66

11

1

2

67

55

2

57

116

29

4

33

67

2

1

3

68

4

3

7

117

39

2

41

68

2

1

3

69

19

1

20

118

65

10

75

69

1

3

4

70

233

3

236

119

50

4

54

70

9

4

13

71

61

2

63

120

26

2

28

71

2

7

9

72

107

2

109

121

138

2

140

72

2

1

3

73

70

3

73

122

91

2

93

73

3

1

4

74

33

2

35

123

39

2

41

74

3

2

5

75

5

2

7

124

107

2

109

75

5

1

6

76

145

2

147

125

31

2

33

76

4

1

5

77

10

41

1
2

11
43

126

49

2

51

77

2

1

3

78

79

48

5

53

RE

DINTEGRATIO

N

REDINTEGRATION

80

18
15

1
2

19
17

81

1

23

2

25

1

6

3

9

82

16

2

18

2

4

3

7

2

15

2

17

83

9

2

11

3

2

3

3

2

1

3

84

35

4

39

4

5

6

4

2

1

3

85

21

4

25

5

2

3

5

2

1

3

86

40

2

42

6

1

2

6

3

1

4

87

10

2

12

7

2

4

7

2

1

3

88

10

2

12

8

2

3

8

2

2

4

89

3

1

4

9

2

3

9

15

3

18

90

6

1

7

10

2

3

10

2

1

3

91

55

2

57

11

2

3

11

2

1

3

92

10

3

13

12

2

3

12

12

2

14

93

25

1

26

13

3

4

13

1

1

2

94

59

3

62

14

3

4

14

2

1

3

95

55

2

57

15

4

5

15

7

1

8

96

51

1

52

16

6

4

10

16

8

2

10

97

81

2

83

17

27

2

29

17

2

2

4

98

5
37

2

1

7
38

18
19

10
28

3
2

12
30

99

100

12

1

13

20

15

2

17

101

3

1

4

21

4

1

5

102

45

2

47

22

7

2

9

103

8

1

9

23

3

1

4

104

17

1

18

24

4

1

5

105

3

2

5

25

15

2

17

106

4

2

6

26

12

2

14

107

38

16

54

27

5

2

7

108

35

2

37

28

6

1

7

109

195

2

197

29

11

2

13

110

62

3

65

30

5

1

6

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 55

TABLE X

70-Day Period. 1 Trial per Day

Time in Seconds

Rat-

8

9

5

Trial

a

b

c

Trial

a

b

c

Trial

a

b

c

1

28

4

32

1

395

8

403

1

145

53

198

2

58

3

61

2

155

8

163

2

28

9

37

3

118

3

121

3

1800

3

168

27

195

4

19

5

24

4

1800

4

30

7

37

5

66

25

91

5

582

22

604

5

491

32

523

6

50

5

55

6

145

8

153

6

68

17

85

7

35

2

37

7

94

8

102

7

55

3

58

8

58

3

61

8

82

5

87

8

13

2

15

9

346

4

350

9

51

30

81

9

48

2

50

10

19

2

21

10

699

11

710

10

17

2

16

11

10

3

13

11

70

3

73

11

25

2

27

12

5

2

7

12

70

3

73

12

745

10

755

13

16

4

20

13

32

2

34

13

320

3

323

14

8

4

12

14

22

2

24

14

166

3

169

15

50

3

53

15

6

2

8

15

13

2

15

16

24

1

25

16

5

2

7

16

4

2

6

17

6

2

8

17

6

2

8

17

12

2

14

18

5

2

7

18

3

2

5

18

6

2

8

19

10

2

12

19

3

1

4

19

9

2

11

20

8

2

10

20

2

2

4

20

3

1

4

21

22

23

21

6

7

21

5

2

8

22

14

15

22

6

7

22

8

9

23

7

2

9

23

7

12

23

2

3

24

7

8

24

2

3

24

6

7

25

5

2

7

25

2

3

25

7

2

9

26

25

3

28

26

1

2

26

14

16

27

3

4

27

5

11

27

6

7

28

6

7

28

6

8

28

2

3

29

7

8

29

2

3

29

2

3

30

6

7

30

2

3

30

5

7

31

6

7

31

1

2

31

1

2

32

3

4

32

1

2

32

1

2

33

7

1

8

33

1

2

33

5

6

34

4

2

6

34

1

2

34

1

2

35

5

6

35

1

2

35

14

16

36

4

5

36

6

8

36

1

2

37

10

11

37

4

6

37

8

9

38

7

8

38

7

8

38

7

8

39

4

5

39

6

7

39

14

15

40

2

3

40

5

6

40

1

2

41

5

6

41

1

2

41

10

12

42

5

6

42

1

2

42

1

2

43

5

6

43

2

3

43

8

9

44

3

2

5

44

11

12

45

6

7

re

DINTEGl

^TIO

N

45

9

10

46

9

4

2

10
6

46
47

6

1

7

47

1

10

11

2

48

4

5

2

2

3

5

48

1

2

49

5

6

3

4

5

49

9

10

50

5

6

4

5

7

50

8

9

51

2

3

5

1

2

51

5

2

7

52

3

2

5

6

2

3

52

2

3

53

5

1

6

7

2

3

53

6

7

54

5

2

7

8

2

3

54

2

3

56

THOMAS WILLIAM BROCKBANK

TABLE X — (Coni!nn?d)

Rat-

Trial

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

5
6
3
2
2
3
5
2

10
2
4
2
2
3
5
4
3
2
5
3
2
3
5
2
2
4

6
4
5
4
5
5
8
4
5
3
7
3
6
6
5
6
9
6
7
5
3
3
4
6
3
11
3
5
3
3
4
6
5
4
3
6
4
3
4
6
3
3
5

Trial

9
10
11
12
13
14
15
16
17
18
19

8. (concluded)

112
113
114
115
116
117
118
119
120
121
122
123
124
125

2

1

3

1

3

3

3

2

2

3

2

2

. 2

2

3

4

REDINTEGRATION

1
2

3

4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

4
6
8
10
8
5
5
4
3
8
7
3
3
3
3
4
4
6
5
4
3
3
4
3
3
4
4
3

Trial

55
56
57
58
59
60
61
62
63
64
65
66
67

redintegrat:

ON

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

REDINTEGRATION IN ALBINO RAT— A STUDY IN DETENTION 57

TABLE XI

70-Day Period. 3 Trials per Day

Time in Seconds

No. of Trials

First 15 Trials of R

.; Last 15 Trials of L.

Av.

Rat 1. L.

73

*3

*3

*3

*3

*3

*3

*4

*3

*4

6

*3

*3

*3

*2

*5

3.46

R.

30

17

*5

*5

*4

7

9

21

18

6

6

*3

7

10

9

8

9

2. L.

37

*3

*3

*3

*2

*2

*3

:t^4

*3

*3

*2

*2

10

*3

*3

9

3.66

R.

18

13

*3

6

*2

*3

*3

*3

*3

*2

9

4

6

5

*3

*3

4.53

3. L.

83

*3

*3

*2

*2

*3

*3

*2

*4

*3

*3

6

116

10

*3

*3

4.4

**R.

30

45

10

8

10

16

*6

*2

*3

5

7

5

*3

11

7

*3

9.33

4. L.

77

*3

*3

*3

*2

*3

*3

*4

13

9

*3

*3

*3

5

4

*3

4.26

R.

17

9

17

*3

*3

*3

*2

*3

*3

8

*3

*3

14

*2

*3

8

5.6

5. L.

70

*2

*2

*2

*2

*2

*3

*2

*5

6

*2

*3

6

*2

*2

*3

2.93

R.

20

29

*3

*3

17

7

*3

*3

*2

*3

*2

*3

*3

7

*3

6

4.53

6. L.

72

*3

*3

*3

*3

*3

*3

8

*3

*2

*3

*3

24

*3

*3

21

5.86

R.

19

12

*4

*3

13

*3

*3

*3

*3

*3

*3

8

6

*3

*4

19

6

7. L.

75

*3

*3

*3

*3

*3

*3

*3

*3

*4

*4

*3

*3

11

*3

*4

3.73

**R.

30

8

7

10

11

7

7

*3

9

*5

9

*3

*5

*3

15

12

7.6

***8. L.

126

204

132

91

96

33

41

85

54

28

140

93

41

109

33

51

82.66

R.

23

25

7

*3

6

*3

4

4

4

*3

*3

*3

*3

*4

*4

*5

5.4

Averages —
Trials.. L. 76.6
R.23.3

Av. * Trials .

.L. 11.1
R. 7.7

Av. Time per Trial

.L. 13.8
R. 6.4

** Had not begun norm of perfect trials at end of 30th trial.
*** Had not begun norm of perfect trials at end of 126th trial.

TABLE XII

70-Day Period. 1 Trial per Day
Time in Seconds

No. of Trials

First 15 Trials of R.

Last 15 Trials of L.

Av.

Rati. L.

88

*3

*3

*3

*3

*3

*3

*2

11

12

*3

*3

*3

5

*3

*4

4.26

R.

22

43

*5

*3

25

*3

11

11

*4

*3

*6

*3

*3

*3

*3

*4

7.93

2. L.

91

*3

*3

*3

*3

*3

*2

*3

5

*3

*3

*3

*4

*3

*3

11

3.66

**R.

30

136

23

*6

8

*3

5

*4

9

7

11

*3

*3

7

8

6

15.93

4. L.

45

*2

*3

*2

*2

*2

*2

8

7

*3

*3

*2

*2

10

*3

*3

3.6

R.

21

13

*3

*3

5

*3

6

*3

*3

*3

*3

*2

*3

8

7

6

4.73

5. L.

68

*3

*3

*3

*3

*2

*2

*4

6

*2

*3

*2

*3

*2

*2

*2

2.8

R.

16

7

*3

*3

*3

*3

*2

*3

5

*2

*2

*2

9

*3

*3

6

3.73

6. L.

45

*3

*3

*3

*3

*3

*2

11

10

*4

*5

*3

9

*3

*3

*3

4.53

R.

29

8

5

5

*4

5

7

6

*4

*3

20

*4

*3

*3

5

*4

5.73

7. L.

83

*3

*3

*3

*3

*3

*2

*3

*3

*2

*3

9

6

*2

*3

*2

3.33

R.

27

5

*3

6

*3

6

*3

6

5

*3

*4

*3

5

*3

*3

*3

4.66

***8. L.

125

5

3

4

4

4

4

3

3

4

3

3

3

3

4

4

3.66

**R.

30

8

8

4

6

8

10

8

5

5

4

*3

8

7

3

3

6

9. L.

43

*3

*3

*2

*2

*2

*2

*2

8

6

8

*7

*6

*2

*2

*3

3.86

R.

19

11

5

5

7

*2

*3

*3

*3

*3

*3

*2

*3

*2

*3

4

3.93

10. L.

61

*2

*2

*2

*2

*3

*3

*4

*3

*2

*2

*2

*2

*2

4

*2

2.46

♦*R.

30

8

22

*3

*3

*3

5

5

4

5

5

4

9

*3

4

10

6.33

Averages —
Trials.. L. 72.1
R.24.8

Av. * Trials .

.L. 12.1
R. 7.5

Av. Time per Trial . . L. 3 . 5
R.6.5

** Had not begun norm of perfect trials at the end of the 30th trial.
*** Had not begun norm of perfect trials at the end of the 125th trial.

58 THOMAS WILLIAM BROCKBANK

of learning fall very close together, but it seems that the
method of wider distribution is the most economical. A
similar situation developed in redintegration but the
averages, if any conclusion may be drawn from them,
show that the three trials per day group is superior. From
these tables it must be admitted that the results of these
two groups are inconclusive in regard to the question of
economy. Because the averages of time and trials are so
nearly equal it may be said that on the whole neither the
three trials per day method nor the one trial per day method
has any decided advantage in point of time and trials.

The data from observation indicate a greater number
of movements, such as in the attempt to push down the
plane, due to the three trials per day methods, and this
fact may be interpreted to indicate a greater economy of
the wider distribution of trials. Because of the persistance
of the dominant " error," the inclined plane presents one
of the most difficult problems with which to carry on any
satisfactory and consistent experiments. In any case, a
larger group of rats should be experimented with before
one can be certain of anything definite; and, certainly, as
far as economy is concerned in redintegration, the problem
must be approached on a broader scale than the one here
presented before there may be any hope of a conclusive
result.

In order to ascertain the effect of retention on incomplete
learning, two litters of rats were set to learn the inclined
plane by the one trial per day method. The results of
this experiment are presented in Table XIII. In this ex-
periment the usual norm was adhered to, except that each
litter was taken from the problem when the first rat of
the respective litter finished both in learning and redin-
tegration. Under similar conditions on the maze it was
seen that the rat which completed learning first did not
complete redintegration first. The present table shows that
the result on the inclined plane follows that on the maze
in this respect.

Probable explanations of this fact, as already stated in
the consideration of incomplete learning on the maze, are
as follows: That the rat which finished first in learning

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 59

TABLE XIII

70-Day Period. 1 Trial per Day

Time in Seconds

No. of Trials

First 15 Trials of R.; Last 15 Trials of L.

Av.

Litter 1.

Rat 1. L.

**34

*3

*3

*3

*3

*3

*2

6

*3

*3

*3

*3

5

8

*3

*2

3.53

R.

77

9

*4

*3

*3

4

*3

5

5

5

*3

6

5

17

17

11.66

4. L.

*5

16

24

46

8

*4

*5

*3

*3

*3

*2

*3

*3

4

16

9.66

R.

11

*5

*4

*3

*3

30

*4

*3

*3

35

*5

*4

*4

*3

6

8.2

5. L.

86

82

59

52

51

23

74

46

39

25

33

87

89

45

48

55.93

R.

30

15

*5

9

7

6

*5

5

10

38

9

25

112

69

149

33.6

6. L.

29

*7

*6

18

10

14

12

39

17

22

9

30

26

30

*5

18.26

***R.

17

27

10

*3

*3

*3

*3

*3

*3

7

11

6

6

*5

*7

*8

5.86

Litter 2.

7. L.

19

130

79

21

42

146

*3

*3

*3

*3

*3

*3

*3

*4

*3

31

R.

*3

*2

5

6

*3

5

5

7

5

8

7

*3

*3

5

*3

4.66

8. L.

*7

15

4

10

*3

*3

7

8

*4

*4

*3

7

8

*3

8

6.26

R.

10

14

6

*3

6

10

*4

*4

*5

4

9

22

*3

9

*4

7.53

9. L.

**57

*3

*3

*3

*3

*3

*3

*4

*4

7

15

67

25

*7

*3

*3

10 2

R.

*2

13

*4

7

*3

*3

*3

7

*4

8

*3

*3

8

6

*3

5 13

10. L.

78

34

*4

*5

7

7

7

*3

6

6

*4

*3

*3

*5

*3

11.6

R.

*4

14

6

10

6

6

*3

12

5

7

*3

*3

3

4

,3

5.93

12. L.

252

93

30

28

22

*6

11

14

10

*4

10

*7

*6

6

*5

33.6

R.

24

18

24

19

*4

6

12

7

15

10

123

15

*6

13

6

20 13

13. L.

*

*

* . .

*

*

138

*6

*5

*5

*4

10

*3

*3

*5

*3

18.2

R.

*6

7

22

9

4

6

*3

11

*4

*3

7

*3

7

*3

8

7.26

14. L.

*5

c;

*5

6

*6

6

*5

*4

*5

6

*5

*5

*4

*4

*4

5

***R.

29

26

*5

10

*4

*3

*3

*4

6

*3

*3

9

5

4

6

*3

6.26

** L. — First in Learning.
*** R. — First in Redintegration.

Sick.

Supplement — No. of * Trials.

Litter 1.

Rat—

1.

4.

5.

6.

L.

12

9

0

3

R.

5

11

2

9

Litter 2.

Rat—

7.

8.

9.

10.

12.

13.

14.

L.

9

7

11

8

5

8

11

R.

6

6

9

6

2

5

8

and did not complete redintegration first may be due to
(a) poor retention owing to innate constitution, some
metabolic disturbance during the retention period, or some
extraneous disturbance; (b) at that particular stage when
this first individual completed learning, perfect integration
was not as firmly established in it as in the rat which
completed redintegration first; and, as a consequence, in

60 THOMAS WILLIAM BROKBANK

the redintegration of the rat which had the more poorly
established integrations* the dominant " error " would be
likely to occur first, and persist the longest. As with
the results on the maze it would be well to repeat that,
although innate constitution may bear some influence,
yet the data at hand on the inclined plane point to the
second explanation as more satisfactory, and more open
to proof than the intangible hypothesis of the first.

In experiment on the inclined plane there are numer-
ous cases, not found frequently on the maze, where the
first trial or trials of redintegration were perfect and a
number of subsequent trials were imperfect; e.g., in the
records of rats 7, 9, 10, and 13. Had the present experi-
ment been conducted according to the norm which holds
that the first trial is " pure " retention and evidently
adequate to ascertain how well a habit is retained, the
individuals mentioned would certainly have been credited
with perfect retention. And further, according to this
norm, if the first trial after the retention period was perfect,
it would logically and theoretically follow that, with every-
thing else equal, the subsequent trials would also be per-
fect, in accordance with the theories of recency, frequency,
and repetition or intensity. The subsequent trials in the
cases cited were not perfect; and thus it may be inferred
that the capacity of the individual organism to retain
integrations may not be completely known in the first
trial after retention, but only when a number of successive
perfect trials show conclusively that the habit has been
acquired. It is thus important to remember that all
" errors " and imperfect integrations of the redintegration
series of trials throw light on how well the habit has been
learned in the first place and how well it was retained in
the second. Or it may be said further that the problem
is so difficult that it is impossible to get six consecutive
trials in redintegration since it is so difficult to get these
in learning; but all these rats made six consecutive trials
perfect in order to attain the norm.

From the data of both litters the dominant " error "
was found, as stated in the previous records on the plane,
at the point in the problem where the rat must establish

REDINIEGRATION IN ALBINO RAT — A STUDY IN RETENTION 61

the integration of stepping on the plane, and the same
" error " followed in redintegration. It seems possible
then that a dominant " error " is to be found in every
problem, and that the recurrence of this " error " is to
be expected in trials after retention. In regard to the
effect of incomplete learning on retention and redintegra-
tion, therefore, the conclusion from the maze has been
confirmed by these results on the plane. That is, in general,
the better established is integration in the whole process
of the learning, the fewer will be the imperfect integrations
during redintegration, allowing always that the capacity
of the organism to retain the integrations may tend to
affect the exercise of the habit after the retention period.
Evidently, then, the dominant " error " that appeared in
learning reappeared in redintegration — a recapitulation
being probable; that is, in the inclined plan problem
" errors " in learning and redintegration are chiefly those
pertaining to the plunging of the plane. The maze and
the inclined plane present similar difficulties in establishing
integrations, which with some individuals are more difficult
than with others. These integrations are recorded in the
imperfect integrations usually called " errors."

III. Individual Differences
Before proceeding to the general discussion of results,
the subject of individual differences, appearing in learning
and redintegration, seems worthy of consideration. The
point has been observed in preceding pages that the learn-
ing and redintegration of all individuals have many char-
acteristics in common, such as the reappearance in redin-
tegration of " errors " which appeared in learning. The
object of this chapter is to show some differences which
characterize individuals in learning and redintegration.
It is not necessary for the present purpose to reconsider
the records of every individual. However, in order to
explicitly point out concrete examples of differences, some
certain group should be considered. Tables V-A and
V-B will be most advantageous for the present, owing to
the fact that some individuals in this group were perfect
in redintegration.

62 THOMAS WILLIAM BROCKBANK

The individual " error " totals of Table V-A of the maze
show the relative extremes in redintegration, namely,
where redintegration was completed (a) with numerous
" errors " and (b) without " error." These extremes may
be seen (a) in records of 3, 4, and 7 and (b) in record 1.
In considering the " errors " in records 3, 4, and 7, the fact
significant of individual differences, is that the number
and kind or, in other words, the quantity and quality of
" errors " in each case are not the same. From this fact,
it is clear that the process of the establishment of integra-
tions in each case proceeded with more difficulty with
some than with others. The integrations difficult for one
individual to establish — as in record 4, turn 2, are readily
established by another — as in record 3, -turn 2; while on
the contrary the source of " error " for the second in-
dividual— records 3, turn 5 — is not the source of " error "
for the first — record 4, turn 5. There are individual
differences also in the number of trials necessary to com-
plete learning and redintegration. In Table V-B it may be
noted in record 5 that 130 trials were required to complete
learning and 16 for redintegration; while in record 2,
39 trials were required for learning and 29 for redintegration.

It is true that the differences which characterize the
learning process, and that also of redintegration, are influ-
enced to some extent by extraneous disturbances' and
stimuli. But the important basis of difference is fO be
found in the individual. Every individual possesses an
heridetary neuromuscular endowment, or organization,
capable of performing a limited repertoire of actions. The
functioning required in the performance of certain actions
is called forth by stimuli of environments; but beyond the
limit of the repertoire of actions and their integration in
the organism, stimuli are ineffective to produce action
which may result in a habit. One individual may possess
an endowment whose functions may be called forth by
certain stimuli from environments and rapidly integrated
in a learning process until the integrations are established
in a habit. Another may be so endowed that the process
of learning under the conditions of environment as above
mentioned proceeds much more slowly. While a third

REDINTEGRATION IN ALBINO RAT— A STUDY IN RETENTION 63

individual may be placed in the same environment, but
repertoire of actions is so limited that the individual can
not respond. Examples of each of these cases may be
seen in records of the inclined plane already cited, where
some learned rapidly, some more slowly, and some failed
entirely.

In regard to responses of individuals, as above con-
sidered, it is an open question whether the muscular or
the sensory aspect of the endowment is the most important.
The object of the sensory experiments by Vincent (22)
and others would indicate that the response of the individual
in the acquirement of the maze habit, for example, is for
the most part sensory. It is undeniable that the sensory
endowment of the individual conditions reception of stimuli
to a great extent; but the fact should not be over-
looked that the muscular endowment of movement is of
equal if not of greater importance, because upon this
the habit depends first and always, while the sensory
response is confined more particularly to the beginning
of the learning. Individual differences in endowment
would thus result in individual differences in behavior,
in the establishment of integrations, as may be seen in
the number of " errors," in rapidity of the process of
establishing integrations, as may be seen in the number
of trials and in speed of movement, and as may be seen in
the time totals. This is true because of the persistence
of the dominant " error," and the difficulty of the rat
to establish the integrations of movement at the locus of
dominant " error." For example, the rat has all the sen-
sory cues of going to the plane and coming from the plane
after it is plunged, and these are usually established at
approximately the 9th trial, yet the integrated movement
of plunging the plane is not yet established. Evidently
the sensory cues are present, but integrated movement
required in plunging the plane remains imperfect for
many subsequent trials.

Numerous other cases could be cited to illustrate con-
cretely the fact that scarcely, if ever, are two individuals
found to exhibit the same behavior in any given environ-
ment. One individual may learn as rapidly as another,

64 THOMAS WILLIAM BROCKBANK

but the responses of each differ, and the methods of learning
are never the same, in ioto. The redintegration of two
individuals may likewise seem identical by comparison,
but there are degrees of difference which characterize each
and differentiate one from the other. Individual differ-
ences are thus to be considered as facts which can not be
called into question.

Conclusion

1. The foregoing experiments were undertaken with
a view to attempting to discover some characteristic
in learning which likewise appeared as a characteristic in
redintegration after a period of retention.

2. In the learning of the maze and the inclined-plane
problems, there is a process through which the integrations
of the habit are established. In the chain of integrations
which constitute a habit, the learning process discloses
certain definite integrations which are more difficult to
establish than others. These may be said to be " weak
links " in the chain of the habit integrations. Because
of the fact that the " errors," due to these weak links,
are the more numerous, these " errors " are called dominant.

3. The locus of dominant " error " in the maze specifi-
cally indicates those movements which are difficult to
integrate. They are movements which are nearer the
higher functional limits of the rat's organization than
are other integrations required in the habit.

4. Though at the end of the learning period the domi-
nant " error " may appear to have disappeared, it is the
first one to reappear in the redintegration series. This
discloses the fact that the integration or integrations which
are most difficult to establish are the first to be lost, whereas
the integrations which are most readily established persist
the longest.

6. In order that the problem of establishment of in-
tegrations be approached from as many points of view as
laboratory conditions will allow, additional experiments
were conducted.

7. The method of wider distribution of trials is a better
method of learning; and second, that although both methods

REDINTEGRATION IN ALBINO RAT — A STUDY IN RETENTION 65

establishment of a habit, yet by the three trials per day
method perfect integration is more difficult at the locus
of dominant " error." Such a method is absolutely more
economical in learning, retention, and redintegration, but
is relatively less economical during retention and integra-
tion than it is in the process of learning.

8. The learning of another problem during the period of
retention, before giving redintegration series, does not result
in a loss of redintegration; that is, the acquisition of a
second habit does not produce distrubances in the organism
which might interfere with redintegration tests for the first
habit established.

9. Previous training improves subsequent learning and
redintegration. The condition of the organism is in many
ways superior after a period of training.

BIBLIOGRAPHY

1) KiNNAMAN, A. J. Mental Life of Two Macacus Rhesus Monkeys in Captivity.

Am. Joum. Psych., XIII, 1902.

2) Yerkes, R. M. Instincts, Habits, and Reactions of the Frog. Harv. Psych.

Studies. 1903.

3) . The Dancing Mouse. 1907.

4) Allen, J. The Associative Processes of the Guinea Pig. Journ. Comp. Neu.
and Psych., XIV, 293. 1904.

5) Porter, J. P. Further Study of the English Sparrow and Other Birds. Am.
Joum. Psych., XVII, 248. 1906.

6) Rouse, J. E. The Mental Life of the Domestic Pigeon. Harv. Psych. Studies.

II., 580. 1906.

7) Cole, L. W. Concerning the Intelligence of Raccoons. Journ. Comp. Neu.

and Psych., XVII, 211. 1907.

8) Davis, H. B. The Raccoon: A Study in Animal Intelligence. Am. Journ.

Psych., XVIII, 447. 1907

9) CoLViN, S. S. and Burford, C. C. The Color Perception of Three Dogs, a

Cat, and a Squirrel. Psych. Mon., Ser., No. 44.

(10) Thorndike, E. L. Animal Intelligence. 1911.

(11) Hunter, W. S. Some Labyrinth Habits of the Domestic Pigeon. Joum. An.

Beh., I, 278. 1911.

(12) . The Delayed Reaction in Animals and Children. Beh. Mon.,

Ser., No. 6.

(13) Casteel, D. B. The Discriminative Ability of the Painted Turtle. Journ.

An. Beh. I. 1. 1911.

(14) Breed, F. S. Reaction of Chicks to Optimal Stimuli. Journ. An. Beh. II.,

280. 1912.

(15) Sackett, L. W. The Canada Porcupine: A Study of the Learning Process.

Beh. Mon., No. 2, Vol. 2.

(16) Johnson, H. M. Audition and Habit Formation in the Dog. Beh. Mon.,

No. 3, Vol. 2.

(17) Basset, G. C. Habit Formation in a Strain of White Rats with less than

normal Brain Weight. Beh. Mon., No. 4, Vol. 2.

(18) Ulrich, J. L. Distribution of Effort in Learning in the White Rat. Beh.

Mon. No. 5, Vol. 2.

(19) Meumann, E. The Psychology of Learning. 1913.

(20) Watson, J. B. Behavior An Introduction to Comparative Psychology. 1914.

(21) HuBBERT, H. B. The Effect of Age on the Habit Formation in the Albino Rat.

Beh. Mon. No. 6, Vol. 2.

(22) Vincent, S. B. The White Rat and The Maze Problem. Jour. An. Beh.

5. 1915.

66

The Behavior Monographs

Edited by JOHN B. WATSON

The Johns Hopkins University, Baltimore, Md.

VOLUME 1
No. 1 The development of certain instincts and habits in chicks. By

Frederick S. Breed. Fp. iv + 78, ^1.00, postpaid.
No. 2 Methods of studying vision in animals. By Robert M. Yerkes

and John B. Watson. Fp. iv + 90, $1.25, postpaid.
No. S An experimental study on the death-feigning of Belostoma

(— Zaitha Aucct.) flumineum Say and Nepa apiculata Uhler.

By Henry H. F. Severin and Harri^ C. Severin. Pp. iii -f

47, $.65, postpaid.
No. 4 The biology of Fhysa. By Jean Dawson. Pp. iii + 120, $1.50,

postpaid.
No. 5 The function of the vibrissae in the behavior of the white rut.

By Stella Burnham Vincent. Fp. iv + 81, $1.15, postpaid.

VOLUME 2
No. 6 The delayed reaction in animals and children. By Walter S.

Hunter. Fp. v + 86, $1.15, postpaid.
No. 7 The Canada porcupine: a study of the learning process. By

Leroy Walter Sackett. Pp. iii + 84, $1.15, postpaid.
No. 8 Audition and habit formation in the dog. By Harry Miles John-
son. Pp. iv + 78, $1.00, postpaid.
No. 9 Habit formation in a strain of albino rats of less than normal

brain weight. By Gardner Cheney Basset. Pp. iv + 46,

$.65, postpaid.
No. 10 Distribution of effort in learning in the white rat. By John

Linck Ulrich. Pp. iii + 51, $.65, postpaid.
No. 1 1 The effect of age on habit formation in the albino rat. By

Helen B. Hubbert. Pp. v + 55, $.65, postpaid.

VOLUME 3

No. 12 The mental life of monkeys and apes: a study of ideational be-
havior. By Robert M. Yerkes. Pp. iv + 145, $1.50, postpaid.

No. 1 3 A study of perseverance reactions in primates and rodents. By
G. V. Hamilton. Pp. iv + 65. $.75, postpaid.

No. 14 An analysis of the learning process in the snail, physa gyrina Say.
By Elizabeth Lockwood Thompson Pp. iii + 97, $1.25,
postpaid.

No. 15 The effect of length of blind alleys on maze learning: An ex-
periment on twenty-four white rats. By Joseph Peterson.
Pp. iii + 53, $.75, postpaid.

No. 1.6 An experimental study of transfer of response in the white rat.
By Harry H. Wylie. Pp. iii + 66, $1.00, postpaid-

VOLUME 4.
No. 1 7 Transfer of training in the white rats upon various series of
mazes. By Rutledge T. Wiltbank. Pp. iii +Q5, $1.00,
postpaid.
No. 18 Redintegration in Albino rat. By William Brockbank, A. M.
Pp. iii + 66, $1.00, Postpaid.

Regular subscription $5 per volume, postpaid.

THE BEHAVIOR MONOGRAPHS '

Emerson Hall, Cambridge, Mass.

Directory of American Psychological Periodicals

American Journal of Psychology — Worcester, Mass.: Florence Chandler.
Subscription $5. 600 pages annually. Edited by G. Stanley Hall.
Quarterly. General and experimental psychology. Founded 1887.

Pedagogical Seminary — Worcester, Mass. : Florence Chandler.
Subscription $5. 575 pages annually. Edited by G. Stanley Hall.
Quarterly. Pedagogy and educational psychology. Founded 1891.

Psychological Review — Princeton, N. J. : Psychological Review Company.

Subscnption (with Psychological Bulletin) $6. 480 pages annually.
Bi-monthly. General. Founded 1894. Edited by Howard C. Warren.

Psychological Bulletin— Princeton, N. J. : Psychological Review Company.
Subscription $3.25. 480 pages annually. Psychological literature.
Monthly. Founded 1904. Edited by Shepherd I. Franz.

Psychological Monographs— Princeton, N. J.: Psychological Review Company
Subscription $4 per vol. 500 pages. Founded 1895. Ed. by James R. Angell.
Published without fixed dates, each issue one or more researches.

Psychological Index — Princeton, N. J. : Psychological Review Company.
Subscription $1.25 200 pages. Founded 1895. Edited by Madison Bentley.
An annual bibliography of psychological literature.

Journal of Philosophy, Psychology and Scientific Methods — New York:

Sub-Station 84. Bi-weekly. 728 pages per volume. Founded 1904.
Subscription $3. Edited by F. J. E. Woodbridge and Wendell T. Bush.

Archives of Psychology — Substation 84, New York: Archives of Psychology.
Subscription $5. 600 pp. ann. Founded 1906. Edited by R. S. Woodworth.
Published without fixed dates, each number a single experimental study.

Journal of Abnormal Psychology — Boston: Richard G. Badger.
Subscription $5. 480 pages annually. Edited by Morton Prince.
Bi-monthly. Founded 1906. Entire field of abnormal psychology.

Psychological Clinic — Philadelphia: Psychological Clinic Press.
Subscription $1.50. 280 pages annually. Edited by Lightner Witmer.
Monthly (9 numbers). Orthogenics, psychology, hygiene. Founded 1907.

Training School Bulletin— Vineland, N. J.: The Training School.

Subscription $1. 160 pages annually. Ed. by E. R. Johnstone. Founded 1904.
Monthly (10 ntunbers). Psychology and training of defectives.

Journal of Race Development — Worcester, Mass. : Louis N. Wilson.

Subscription $3. 460 pages annually. Founded 1910.
Quarterly. Edited by George H. Blakeslee and G. Stanley Hall.

Journal of Educational Psychology — Baltimore: Warwick & York.
Subscription $3. 600 pages annually. Founded 1910.
Monthly (10 numbers). Managing Editor, J. Carleton Bell.
(Educational Psychology Monographs. Edited by Guy M. Whipple.
Published separately at varying prices. Same publishers.)

Journal of Animal Behavior — Cambridge, Mass.: Emerson Hall.
Subscription $5. 450 pages annually. Founded 1911.
Bi-monthly. Robert M. Yerkes, Managing Editor.

The Behavior Monographs — Cambridge, Mass.: Emerson Hall.
Subscription $5. 4^ pages per volume. Edited by John B. Watson.
Published without fixed dates, each number a single research.

Psychoanalytic Review— Washington, D. C: 3617 Tenth St., N. W.
Subscription $5. 500 pages annually. Psychoanalysis.
Quarterly. Founded 1913. Ed. by W. A. White and S. E. Jelliffe.

Journal of Experimental Psychology — Princeton, N. J.: Psychological Review
Company. 480 pages annually. Experimental. Founded 1916.
Subscription $3.25 Bi-monthly. Edited by John B. Watson.

Journal of Applied Psychology — Worcester, Mass.: Florence Chandler.
Subscription $4. 4(X) pages annually. Quarterly. Founded 1917.
Edited by G. Stanley Hall and L. R. Geissler.

The Brandow Printing Company
Albany N. Y.

^i,3ri

X

Behavior Mono»'APhs

Volume 4. Number 3, 1919 Serial Number 19

' Edited by JOHN B. WATSON

The Johns Hopkins University

Discrimination of Light of Different
Wave-Lengths by Fish

BY
CORA D. REEVES

Published

at Cambridge, Boston, Mass.

HENRY HOLT & COMPANY

34 West 33d Street, New York

G. E. STECHERT & CO., London, Paris and Leipzig, Foreign Agents

The Journal of Animal Behavior

An organ for the publication of studies concerning the instincts,
habits and intelligence of organisms

The Journal contains a Department of Notes in which appear brief accounts of
especially interesting and valuable observations of behavior.

Published bi-monthly at Cambridge, Boston, Mass., by Henry
Holt and Company, New York. Each volume contains at least 450
pages with plates and text-figures.

The subscription price is $5.00 per volume (foreign subscription
$5.50) postpaid.

Manuscripts, subscriptions, and all business correspondence should
be addressed to

THE JOUKNAL OF ANIMAL BEHAVIOR

Emerson Hall, Cambridge, Massachusetts

Behavior Monographs

For the publication of studies in behavior and intelligence which are

too lengthy or too expensive for acceptance

by The Journal of Animal Behavior

Published at irregular intervals at Cambridge, Boston, Mass., in
connection with the Journal of Animal Behavior, by Henry Holt and
Company, New York. Each volume contains approximately 450
pages with plates and text-figures.

The subscription price is $5.00 per volume (foreign subscription
$5.50) postpaid.

Monographs may be purchased separately at prices varying with
the cost of manufacture.

Manuscripts and inquiries concerning terms of publication should
be addressed to the Editor of the Behavior Monographs,
JOHN B. WATSON,

The Johns Hopkins University,

Baltimore, Md.

Subscriptions to The Journal of Animal Behavior and the Behavior
Monographs should be sent to Emerson Hall, Cambridge, Massa-
chusetts.

iMUV a4 1919

Behavior Monographs

Volume 4. Number 3, 1919 Serial Number 19

Edited by JOHN B. WATSON

The Johns Hopkins University

Discrimination of Light of Different
Wave-Lengths by Fish

BY

CORA D. REEVES

Published

at Cambridge, Boston, Mass.

HENRY HOLT & COMPANY

34 West 33d Street, New York

G. E. STECHERT & CO., London, Paris and Leipzig, Foreign Aeents

CONTENTS

PAGE

I. STATEMENT OF THE PROBLEM 1

II. EXPERIMENTAL WORK 1

A. Introduction 1

B. Apparatus and Methods 4

1. General description of the apparatus 4

2. General method of procedure 8

3. Detailed description of the apparatus 9

a. Experiment aquarium 9

b. The lamphouse 10

c. Food-bar 10

d. Apparatus for the regulation of the intensity and of the

distribution of light 12

e. Color f-lters 13

4. Detailed description of methods 14

5. Determinations of the intensity and the distribution of light

for stimulus patches 15

C. Experiments to Test Discrimination by the Formation of a Food

Association 20

1. Intensity discrimination by dace 20

2. Wave-length discrimination: White-red by dace 23

3. Wave-length discnmination: Blue-red by dace 25

a. Red chiefly at constant maximum intensity 25

b. Tests for dace Md, Blue with Red decreasing in intensity 30

c. Tests for dace H, Red decreasing in intensity 32

d. Tests for Yl and YP, Red decreasing in intensity 32

e. Discussions of Blue-red graphs of the dace M, Md, Yl,

YP and conclusions 32

4. Intensity discrimination by sunfish as indicated l^y response 37

5. Wave-length discrimination: Blue-red by sunfish, with Red

decreasing in intensity 38

a. General account 38

b. Large sunfish 41

c. Small sunfish 42

d. Sunfish 5S2 43

D. Tests of the Apparatus and its Manipulation 44

1. Check tests by changes in manipulation 45

a. Shifts of si it and filters as a clue 45

b. Position of the experimenter as a clue 45

2. Check tests by changes in the apparatus 45

a. Food-bar 45

b. Lenses 46

c. Filters 46

d. Slits 46

iv CONTENTS

II. Experimental Work — (Continued) PAGE

E. Tests with Changed Quality of Light 49

a. Delayed response 50

b. Peculiar behavior 52

F. Tests with Energy Equated -. . . . 53

G. Results of Experimenta IWork with use of Food Association 55

H. Unlearned Responses of Fish 57

a. Modified breathing rate of fish subjected to light of long

wave-length 57

b. Other characteristic behavior of fish in the presence of

stimulus patches of unfamiliar light of long wave-length 58

c. Dark adaptation and sensitivity to light 62

III. REVIEW OF THE LITERATURE 63

IV. DISCUSSION 78

A. Analysis of Methods Employed 78

a. Method of [response 78

b. Method [of training 79

c. Change of color in response 82

B. Sou rces of Error 83

a. From unsuitable environment or care 83

b. From uncontrolled environmental factors 84

c. From uncontrolled response 85

d. From lack of training at matched brightness 85

e. From exclusive use of the method of response 86

C. General Considerations 86

D. Final Statement of Results 89

V. SUMMARY 92

VI. CONCLUSIONS 99

VII. POSTSCRIPT 100

VIII. BIBLIOGRAPHY 104

DISCRIMINATION OF LIGHT BY FISH^

I. STATEMENT OF THE PROBLEM

In the early part of the second half of the nineteenth
century many comparative studies of the anatomy of the
eye were made. Studies of the physiology of the human
eye followed, but experimental work upon animal vision
was delayed. Not until 1885 were Graber's results pub-
lished. He was among the first to test experimentally
the effects of light of different wave-lengths upon fish, and
found that they showed a preference for the shorter wave-
lengths. Then a number of other experimenters decided
that fish discriminate color. But in 1909 and 1910 Hess
published an account of experiments from which he con-
cluded that fish do not show the responses to wave-length
apparently established by his predecessors in this field.
He attributed the responses obtained by him as well as
those already obtained by others to differences in the
brightness rather than to differences in the quality of the
light used and held that fish are color-blind. His evidence
has been given favorable consideration by a number of
investigators of color vision in higher vertebrates (Watson
1914, Parsons 1915). From this discussion arose the
definite problem, — to determine whether fish can respond
to differences in wave-length of light or can respond only
to differences in intensity^ of light.

II. EXPERIMENTAL WORK

A. Introduction. — The first requirement for the solution
of the question was to determine whether our native fish
show responses' available for experimentation. For some

1 Contribution from the Zoological Laboratory, University of Michigan
- In this paper intensity is used as a physical term, and brightness as a sensation
term; quality is used to describe that phase of sensation dependent upon wave-
lengths.

^ The study of first responses will be referred to as the Method of Response. Sec-
tion H of this paper is devoted to the presentation of data gained by this method.

2 CORA D. REEVES

months three small fish, of undetermined species, had been
living in a crystallizing dish on my desk. One morning
the dish was placed so that one-half of it was in the direct
sunlight and one-half in the shade. The fish came to rest
most frequently in the shade. Then a prismatic solar
spectrum was thrown on the bottom of the shaded half
of the dish and the fish were observed to start or leap as
they entered the red-orange area. They often swam
around this area, to rest with the head in the blue or green,
facing in the direction of the incident rays. Some shiners
{Notropis cornutiis) were then tested by throwing the
intense light from a projection lantern, one meter distant
into the glass-walled aquarium in which they were kept.
At first some of the fish gave a start as they entered the
lighted area, while others swam along the side of the lighted
space but did not enter it.

It was evident in both of these experiments that light
called out differential responses, presumably unlearned or
instinctive; but whether intensity or wave-length differ-
ences induced the response in the test with the spectral
band was in no way indicated.

These reactions to the intensity of white light, as well
as to different parts of the spectral band, were evanescent.
They were obtained in the first trials, but were not evident
in later trials, when the fish, having become used to the
stimulus, no longer responded to it. These first responses
showed the behavior of an untrained animal when stimu-
lated but cannot be assumed to have shown the animal's
inherent power of discrimination. To secure sustained
responses and to show by them the full capacity of the
animal it seemed essential to resort to habit-forming ex-
periments. For these experiments two large intense
patches of light of restricted and definitely known wave-
length were used as stimuli and were presented simul-
taneously. In order to form a food association and thereby
to secure continued response the fish were always fed before
the same patch and the position of this positive stimulus
was, of course, irregularly shifted, while it was kept at
constant intensity. When an association had been formed
with the positive patch of light it became necessary to

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 3

determine to which of the two variables involved, intensity
or wave-length, the response was to be attributed. In
order so to control the experiment that response could be
made to but one of these variables the brightness of the
negative plate was varied until it was equal to that of the
positive for the human eye. Continued response to the
positive plate might then be due to the quality of the
stimulus light. But since it could not be assumed, without
proof, that the relative brightness values of the spectrum
were similar for fish and man the response might be due
to a brightness difference between the two plates. To
follow the usual practice of equating the brightness of the
plates by the human eye only could not lead to dependable
results, and so a means was sought by which it might be
possible to determine by the behavior of the fish when the
two plates were of equal brightness for them. It is believed
that such means has been found.

As an aid in interpreting the responses obtained when
two stimuli of different wave-lengths were used, an effort
was made to determine the minimum intensity differences
between two white light stimuli capable of producing
differential response. For this purpose an attempt was
made to form a food association with the duller one of
two white plates, which was kept at constant intensity
while its companion plate was varied in intensity. If
there should occur any difference in the relative ease of
discrimination of the stimulus plates when the quality
difference was absent then the value of difference in wave-
lengths for discrimination would be shown. The minimum
intensity difference necessary for the discrimination of
two lights should be the same for white and for colored
lights if wave-length differences are not able to differen-
tially stimulate fish.

' For normal activity it is essential that the fish be kept
in a normal environment; otherwise response is often in-
hibited by unnatural conditions or by manipulations that
induce fright. The water should be of constant tempera-
ture from one container to another, shifting of containers
from place to place, variations in intensity of the general
illumination, and manipulations likely to frighten the

4 CORA D. REEVES

iish should be avoided. In the experiments to be described
the fish have been kept under normal conditions and have
remained in the same containers during the tests and in
the intervals between them, often for months together.
The experiments have been so conducted as to reduce
fright behavior to a minimum. During three years the
fish have taken food regularly, have remained healthy,
have grown and have been at all times very tame.

The apparatus described in the following section was
devised to meet the experimental conditions already out-
lined. These call for;

(1) An experiment aquarium in which the fish could
live continuously without being taken from the water.

(2) An experiment procedure which would not arouse
fear behavior.

(3) Two large stimulus patches of either mixed light or
light of restricted and known wave-length, interchange-
able in position, one of which could be varied through a
wide range of measurable intensities.

(4) Provision for offering food before one of the patches
of light, which thus becomes the positive stimulus in the
formation of an association.

(5) Constant conditions in the general luminous and in
the aqueous environment of the fish.

(6) The experimental procedure involves further a
method of equating the brightness of the patches of light
of different wave-length by means of the behavior of the
fish.

B. Apparatus and Methods.

1. General Description of the Apparatus

For the experiments a black-lined, galvanized-iron
aquarium was used. Its ground plan is shown in figure .1.
Movable partitions separate it into three parts, referred
to as the stimulus, the discrimination, and the retention-
compartments, respectively (1, 2, and 3, in fig. 1).

The removable partitions A and B separate these com-
partments as indicated by their positions in figure 1. In
partition A is a vertically sliding door, D, shown partly

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH

1

If

■ r

■c

r

/"

-

'

h-

i

J

£

3

yf

. A

2

- as 4 en,—

1^

Fig. 1
Ground plan of the Experiment Aquarium.
\. Stimulus compartment. /' and /" recess for stimulus plate.

2. Discrimination Compartment.

3. Retention Compartment.

A and B movable partitions; C fixed partition; D sliding door in partition A;
efgh indicate the positions of the corners of a frame above this end of the aquarium;
xy plane of section shown in figure 4; 2iv plane of section shown in figure 5.

lifted in figure 2. A vertical fixed partition, C, divides
one end of the stimulus compartment, 1, into two sub-
compartments, 1' and 1". Lengthwise of each of the sub-
compartments, 1' and 1", extends a plate of opal ground
glass set across it so as to form an angle of 45° with the
floor and with the end wall. These plates are illuminated
from above by patches of white or colored light, and from
them this light is diffused into compartment 1. When
the sliding door {D in fig. 2) is lifted, a fish in the discrim-
ination compartment, 3, has before it a patch of light upon
each of the white stimulus plates (5 in fig. 2). If the patches,
are unlike in either intensity or wave-length, its instinc-
tive behavior toward them may for a time be differential.

CORA D. REEVES

Fig. 2

Perspective View of Aquarium and lamphouse with part of each cut away, to
show interior.

AquariiDu and fond rclcnsp.

1. Stimulus compartment: /' /" recess for stimulus plate.

2. Discrimination Compartment.

3. Retention Compartment.

A, movable partition between 1 and 2 (the partition between 1 and 3 not in place).

C, fixed partition.

D, sliding door in A.

EG, parts of the frame which surround the base of the lamphouse.

IJ, slender food-bar in position for feeding the fish.

A", position of electro magnet by which the food-bar may be held in horizontal
position.

Light Apparatus.

L, II. lamphouse. Below L, H, hooded air-inlet.

M' . C', partition through lamphouse, an upward extension of C.

A'^, Nernst lamp in one half of lamphouse, one in the other half partially shown.

The glowers of these lamps are across the compartments and thus parallel with
the long axes of the cylindrical lenses shown at SL. They are at the principal foci
of these lenses:

O, slit opening of variable width.

P. sliding metal plates which form the slit.

Q, frame which supports metal plates PP and in which they slide.

RF. ray filter in its kit.

SL. cylindrical lens.

S, S. stimulus plates.

W, indicates water level.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 7

(See section H.) Such differential instinctive responses
may show the ability of the fish to discriminate, but they
persist only a short time. To secure responses persisting
for a longer period the fish were fed before one of the
stimulus patches, and thus a food association with the one
patch of light was built up in order to test the ability of
fish to discriminate between the two lighted areas.

To secure patches of light of definite form on the stimu-
lus plates two cylindrical lenses {SL in fig. 2) are mounted
in a frame (indicated by EG in fig. 2) which rests above the
sub-compartments /' and 1". The major axis of each
lens is across its sub-compartment. These lenses par-
allelize the rays in one plane and give two stimulus patches
of equal size, having sharp parallel upper and lower edges
across the opal glass plates. The lower margin of each
patch is always adjusted so as to coincide with the lower
margin of its plate. Control of wave-length of light is
obtained by colored filters ( RF in fig. 2) laid above the
cylindrical lenses. A blue filter of gelatine and a red one
of ruby glass are provided and may be shifted from side
to side, so that patches of light of either color may be used
on either plate. The light for the illumination of the stimu-
lus plates comes from two single-glower Nernst lamps {N
in fig. 2) contained in a lamphouse. The house rests upon
the aquarium, above compartments 1' and 1", and its cross
section is nearly that of these sub-compartments taken
together. It is divided by a vertical partition {M'C in
fig. 2), which is an upward extension of partition C of the
aquarium. Access to its interior is through a door above
the end of the aquarium, which is not visible in figure 2.
The long lamphouse extends nearly to the ceiling, from
which it is so suspended that it may be swung out free
from the aquarium. The lamps remain always at the
same height in the lamphouse, and the intensity of the
light reaching the plates is controlled by a slit (O in fig. 2)
of variable width placed beneath each lamp at right angles
to its glower. Thus by varying the width of the slits and
by shifting the filters a patch of either white or colored
light of any desired intensity may be secured on either
plate.

8 CORA D. REEVES

Food may be supplied to the fish from the food-bar (//
in fig. 2) weighted at one end and pivoted at its center.
After food has been placed at both ends, the bar is brought
into a horizontal position and held there by an electro-
magnet {K in fig. 2). When a fish approaches the patch
of light before which he is to be fed, the bar is released by
opening a switch. The bar is then free and turns on its
pivot until its weighted end is beneath the water in front
of the stimulus plate where the fish can secure the food.

The aquarium is supported on a low table and supplied
with running, filtered water. To exclude extraneous light
and secure uniform illumination of the apparatus during
experimentation the space about the table is enclosed by
black building-paper to form a room 6 ft. by 10 ft. This
light-tight room includes a west window with light-tight
shades. Suspended from the ceiling directly above the
aquarium is a 40 watt tungsten lamp at a distance of 157
cm. from the water. This lamp is enclosed in a cylindrical
metal shade 145 cm. high. As the diameter of the shade
is 14 cm. the light is reduced, being only that passing out
from its open lower end. To give the fish an environment
of more nearly normal color and to keep them from seeing
the experimenter, the aquarium table is surrounded by
a curtain of yellow cheese-cloth hung from the ceiling
within the light-tight room. A cord attached to the slid-
ing door {D fig. 2), in partition A, runs straight up to an
eye in the ceiling and ends outside the curtain. A counter
weight on this door-cord enables the operator to leave the
door open at any height. The switch for control of the
food-bar magnets is also located outside the curtain.

2. General Method o." Procedure

Before an experiment is begun, the window shades are
lowered and the ceiling light is turned on, so that the fish
may become accustomed to the light and their eyes adapted
to the conditions of illumination maintained during the
experiment. After half an hour or more all of the fish
in the aquarium are shut into the retention compartment
(3 in figs. 1, 2) by means of the movable partition B. The
experimenter then adjusts the illuminating apparatus and

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 9

the food-bar with its food. Next, by manipulation of the
movable partitions, the individual fish needed for an experi-
ment is allowed to swim from compartment 3 into compart-
ment 2. The further operations of lifting the sliding door
and releasing the food-bar when the fish has reacted suit-
ably are carried on from without the curtain, while obser-
vations are made through a peek hole.

When experiments are not in progress, the ceiling light
is turned off, the yellow curtain drawn aside, the shades
at , the west window raised, and the movable partitions
taken from the aquarium. The fish then live in a black
pool into which the sunlight penetrates for a short time
each day. On the sides of the pool the fish may see the
yellow curtain and above it the gray of the ceiling. At
one end of the pool towers the black lamphouse. The
visual surroundings are then somewhat like the black
pool and the sky of the natural fish habitat.

The tap water supplied to the aquarium is passed through
a Berkefeld filter. All fecal material and excess food are
daily removed from the aquarium. Frequently the aquar-
ium is almost emptied through a drain at the bottom;
then it is refilled. By these means the water is kept uni-
formly clear and its light transmission constant.

That the fish may have all necessary food substances,
pieces of angle worm, fish, chicken, or bits of boiled egg
or fish food are occasionally substituted for the usual
scraped beef. The fish are healthy and active, and are
growing rapidly after over two years of experimentation.

3. Detailed Description o? the Apparatus^

a. The Experiment Aquarium.

The aquarium is 22 cm. deep, 66 cm. long, and 25.4 cm. wide, inside measure-
ments. Each of the three movable partitions (A, B, T in figs. 1, 2, 6) measures 23
cm. high by 25 cm. wide, and has attached to its u pper edge a small rod of metal
extending 3 cm. beyond the ends. By hanging the projecting ends of this rod over
the edges of the aquarium, a partition can be set at different places across the aquar-
ium. One of these partitions (A) has an opening 10 cm. wide, extending 13 cm.
from the bottom. This is provided with a sliding door {D in fig. 2) 13 by 23 cm.
Opposite to the door when it is in place are the stimulus plates (S,S in fig. 2). Each
plate extends entirely across its sub-compartment and reaches from the floor at the
projecting end of the partition C to the surface of the water at the back of the sub-
compartment. The tap water, after passing through a Berkefeld filter, flows over

* The further descriptions in this section and in section 4 are included especially
for those who may wish to duplicate the apparatus and methods.

10

CORA D. REEVES

a glass plate 20 cm. long, not shown in the figures, and enters at the top of the closed
end of compartment 2. The water flows in a thin sheet over the glass plate and
drops from it into the aquarium. This permits the escape of excess gases from the
water. When the plate is not used the fish develop symptoms of gas-bubble dis-
ease. From the plate the water drips into the aquarium at a rate somewhat above
one drop per second. Outflows are provided in the center of the ends of sub-com-
partments 1' and 1" at 17 cm. from the bottom.

b. The Lamphonse.

The larnphouse is a black rectangular wooden box 172 cm. long, with an internal
cross-section 22 cm. by 23 cm. A partition of thin tin divides it lengthwise into
two equal compartments. The inside of the lamphouse is blackened with several
coats of lamp black and shellac. The single glower Nernst lamp inside each half
of the house (A'' in fig. 2) is of the 0.6 ampere, 220 volt D. C. type. Openings, shown
opposite arrow-heads in fig. 4, at the bottom and at the top of the lamphouse, and
below L. H. m fig. 2, afford ventilation and prevent excessive heating. They are
hooded to prevent the escape of light. The lower end of the lamphouse rests on the
experiment aquarium, which is so placed that the partition C (figs. 1 and 2) is
beneath the partition MO (fig. 2) in the lamphouse. Thus two continuous, dark-
lined tunnels are formed with a white diffusing surface set at an angle of45° across the
bottom of each. A wooden frame {EG in fig. 2) rests on the top of the sub-compart-
ments V and 1" of the aquarium in the position of ejgh of fig. 1, and surrounds
the lower end of the lamphouse.

FIG. 3

Diagram of compartments /' and /" as seen from the center of the doorway of
partition A, figure 2, showing the lower end of the lamphouse with food-bar attached
to the frame upon which it rests.

//, rj', food-bar.

K, electro-magnet which holds // horizontal until released.

S, S, stimulus plate.

w, water level.

c. Food-bar.

At the center of the side gh of this frame is fastened a narrow metal strip or food-
bar. It is 0.8 cm. wide and 24 cm. long after being looped at its center to surround

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH

11

a pivot. This strip is folded to form a short trough at each end and has a drop of
solder in one end to give weight enough to cause that end to fall when not held in
place. At the corners g and h of the frame (figs. 1 and 3) are placed electro-magnets
connected with a dry cell. When the switch outside the curtain is closed and the
free end of the food-bar is placed against the magnet, the bar is held in position
along the side of the frame, gh, above the aquarium. When the switch is opened,
the weighted end of the food-bar falls so that it is beneath the water surface imme-
diately above the front edge of the positive stimulus plate, at which point it is stopped
by a peg in the frame efg/i. Food placed in the shallow trough is thus allowed to
fall immediately in front of the stimulus patch. When the stimulus patch is shifted
from side to side, the food-bar is removed from the pivot and the ends reversed.

r

5!!

L.

^XJOfer/rrren/ A^i/ar/u/r?.

L« ^6 cm.

Fig. 4

Section of aquarium and lamphouse along line xy in figure 1.

A'^, Nemst glower in cross section.

0, level of slit.

RF, ray filter.

SL, cylindrical lens in kit ki.

S, stimulus plate of ground opal glass.

12

CORA D. REEVES

d. Apparatus for Regulation of the Intensity and Distribution of Light.

A plano-convex cylindrical lens (Yerkes 1903) 20 cm. long and 10 cm. wide, with
a radius of curvature of 12.5 cm., was divided crosswise into halves. One of these
was fitted into the frame at the bottom of the lamphouse in each side (SL in figs.
2, 4). The Nemst lamps are placed at the principal foci of the lenses with the
glowers parallel to and above the long axis of the lenses. In this position the cylin-
drical lenses serve to parallelize the light rays in the direction indicated in figure 4.
As the walls of the aquarium are black, the diverging rays which strike them at the
sides of the stimulus plates (see fig. 5) are so little reflected as to be scarcely visible.
But as the margins of these areas remain parallel they vary only in brightness not
in shape. Their variation in brightness will be the same as the variation of the
stimulus patches depending upon the slit-width. By this method no direct light
from the glowers reaches the floor of the aquarium, and though the whole of each
stimulus plate (S in fig. 2) is not illuminated, the patches of light on the two are
exactly the same in shape and size. They are rectangular 12.3 cm. by 11.6 cm.,
and their edges are always sharp and straight across the plates.

To regulate the light intensity a slit is used (figs. 2, 4, 5). The slit consists of
two steel plates and a narrow steel frame which is the size of the inside cross-sec-

FiG. 5

Cross-section of one-half of lamphouse through Nemst glower along line uv in
figure 1.

A'^, Nemst glower.
0, slit opening.
RF, ray filter in kit ki.
SL, cylindrical lens.
S, stimulus plate of opal ground glass.
W, groove in which the frame for the slit slides in and out

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH

13

tion of one-half of the lamphouse and can be slid into a groove in either half of the
house. Shding in this frame is a pair of wide steel plates, each 48 mm. by 170 mm.
with beveled knife-edges. The wide plates can be brought close together when a
narrow slit is desired, or separated for slit apertures up to 4 mm. For greater slit
widths two other pairs of plates 35 mm. and 25 mm. by 170 mm. are provided to
replace the wider ones. The widths of the slit openings can be read on a millimeter
scale on the frame. For regulating the slit width of 0.1 mm. intervals, a series of
strips of accurately measured thickness are used. One of these strips is placed in
the slit, the plates are brought against it, and the strip is then carefully withdrawn.
The slit in position in the house is 3 cm. beneath the glower of either Nernst lamp.
The direction of the slit is at right angles to the length of the glower and to the
long axis of the lens.

To prevent any variation in the appearance of the lower surfaces of the lenses
being seen by the fish, a curtain is hung across the top of the front of the sub-com-
partments 1' and 1", reaching to the level of the upper edge of the stimulus patches.
This is not shown in figure 2.

e. Color-fillers.

The color filters, each 9 cm. square, are placed as shown in figures 2, 4, 5, in a
photographic kit or frame 9 cm. by 9 cm., resting upon the lenses. The wave-
lengths of light transmitted by the blue filter are A4600 to A5245 with the maximum
in the blue; those transmitted by the red ruby glass are from A5890 to the end of
the visible spectrum. None of the yellow is let through by either filter, so that the
light used is from two separated parts of the spectrum. The blue filter is a Wratten

Fig. 6

Ground plan to show how the partitions are manipulated in testing the fish.
/'/", compartments for stimulus patches.

A, movable partition with sliding door.

B, T, movable partitions; T is used for cutting out single fish to be tested.

A' B' , movable partitions swung out to allow fish separated from the others to
swim from 3 to 2.

14 CORA D. REEVES

ray filter (Red 3.65 C 3-48 A) secured fr6|m the Eastman Kodak Company. The
red is a piece of copper-flashed ruby glass.

4. Detailed Description of Methods

To make clearer some methods employed during experimentation the following
details of procedure are included. That the fish might behave normally care was
taken to avoid frightening them. They were allowed to live for months in the
experiment aquarium without being lifted from the water or being moved in the
aquarium. The partitions were moved very slowly. Often when they were shifted
from one compartment to another, the fish showed absence of fear by swimming
up to the partition as if to nibble its lower margin. It may be that the partition
when thus moved resembles a clod which gradually settles to the bottom or slides
along the stream margin and offers the fish a good place to feed, because worms
previously covered are exposed. Because the fish exhibit little fear of these moving
partitions, they were used exclusively when changing the fish about in the aquarium.
A third partition, T in figure 6, was often put into the position shown, in order to
separate from the other fish the one that it was desired to test.

Then after swinging the partitions A and B, figure 6, out to positions A' and B' ,
the fish to be tested would swim quickly into the discrimination compartment,
and with A and B brought back into place, was ready for the raising of the sliding
door, D, when a test was to be made. If when the door was raised, the fish swam
toward the plate determined upon for the food association, that is, toward the
positive stimulus plate, then the bar with the food was allowed to fall. After time
had teen allowed for feeding, the fish was brought back into the discrimination com-
partment by lifting the partition A with the door closed and slowly lowermg it so
that the fish was on the side away from the stimulus plates. The oartition was
then slowly returned to its former place. If the fish made a wrong choice it was
not fed but brought back into the discrimination compartment without releasing
the food-bar. The latter was not set free until the fish had come within 15 cm.
of the positive stimulus plate. It was thus possible for the fish to change
direction of response within the discrimination compartment or at any point
between it and the distance of 15 cm. from the positive plate.

It was determined by the use of a mirror (see p. 46) that when the fish was swim-
ming low in the aquarium, a reflection of the slit over the red plate on the surface
of the cylindrical lens was sometimes visible to the fish when within 15 cm. of the
stimulus patches and might guide it in its choice. Choice must therefore be made
at a point farther from the plates. If the fish entered the negative half of compart-
ment 2 but while still more than 15 cm. from the stimulus plates turned in a straight
course toward the positive plate, then the food-bar was released and the choice
recorded as correct.

It was found that caution was required, for when a fish swam up toward the
positive plate, there was a tendency to drop the bar too quickly. The fish soon
learned to go up to one side of the aquarium and wait, and if the bar was not dropped,
to swim promptly to the opposite side. Only by very careful attention as to whether
the fish was more than 15 cm. from the positive plate was this error avoided. If
the fish was more than 15 cm. from either plate, the food-bar was not dropped and
either side might be correct. Waiting for the food-bar could not indicate need for
change of direction when the fish was at a greater distance from the plate.

That there might be no regular order for the side on which the positive stimulus
was presented, six slips of paper were prepared, each with one of these inscriptions
upon it:

1 lej;

1 right

1 lejl

1 right

2 left

2 right

LIGHT OF DIFFER EKT WAVE-LENGTHS BY FISH 15

Before the day's series was begun these slips were shaken together and drawn by
chance, and the order in which they were drawn determined the positions of the
positive plate. It will be noted that there is here possibility for only four consecu-
tive tests to be made on one side. This procedure was followed to prevent the
formation of " position habits." In order to prevent any possible " alternating
swing," i.e. fish going first to one side and then the other, it appeared best occa-
sionally to modify this possible order of choices by substituting for two of the single
choice slips two with 3 left and 3 right. These gave the possibility of six conescu-
tive choices to occur on one side.

When a fish had been given from five to twelve chances to discriminate between
the stimulus patches, it was closed back into the retention compartment by means
of partition B and was not used again until the following day's series. Care was not
taken to keep the number of trials per day exact. The number of food masses
seized and eaten was thought to be a better index of the physiological condition of
the hunger of the fish than the number of tests given. Trials with an individual
were, therefore, kept up until its behavior indicated that it was no longer hungry
(until it had eaten about 6 to 8, or 8 to 10 food masses). This largely depended
upon whether the fish were being tested every day or every second day. Care
was taken to place the food-bar in exactly the same position with reference t) the
magnets. The two masses of food on the ends were made as nearly as possible
of the same size and shape. Frequently, as the day's series progressed, new masses
were placed on the lighter end of the bar. Both ends of the bar were kept wet.
They were alike in appearance.

During the tests the operator observed the fish from behind the yellow curtain
already mentioned. The stop watch and the string running to the sliding door
were manipulated together in one hand while the other hand was placed on the
switch so that no jar or body movement or motion of the curtain might indicate
the correct side. The time was taken from the raising of the door until the food
was snapped at or the fish was before the negative stimulus patch and less than 15
cm. from it. The path taken by the fish was then traced on a plan of the aquarium
floor, the time recorded, and other notes made. The fish was then closed back for
a second trial or closed out into the retention compartment while others were tested .

Matching Brighness.

Whenever it was necessary to match the brightness of two stimulus patches,
either direct judgment of the appearance of the patches was used or the flicker
photometer was employed. In the first case wdiere the Equality of Brightness Method
(Ives 1912) was employed one of the following graduate students in psychology, —
Miss Marion Bills, or Miss Z. Pauline Buck, — directed the changes to be made in
the variable plate until the two plates appeared to be equally bright. For assist-
ance in this matching my gratitude is acknowledged.

The method for the use of the flicker photometer was as follows. The lamp-
hcuse with the frame and lenses was swung out past the end of the aquarium. The
flicker photometer was placed immediately beneath the partition separating the
lalvesof the lamphouse. Plates of ground opal glass, like the stimulus plates,
were set beneath the lamphouse, each at an angle of 45° facing the flicker photo-
rneter. The light that fell upon these plates of ground glass was that which would
give the stimulus patches when the lamphouse was in place, so that, as nearly as pos-
sible, the patches were matched against each other by this means.

5. Determinations of the Intensity and the Distribution of Light

In order to determine the relative brightness of the
colored lights employed, so that the conditions may be
duplicated in the future, the illumination used for the
blue stimulus plate was measured. The method was to
illuminate with a IIU volt tungsten lamp, at a suitable

16

CORA D. REEVES

distance and voltage, an area of the white stimulus plate
equal to that illuminated by the Nernst lamp with the
blue filter interposed and without a slit. These equal blue
and white areas were then similarly placed with refer-
ence to the flicker photometer, and were equated by vary-
ing the distance of the lamp. The tungsten lamp, at its
given voltage, was then calibrated by Dr. E. F. Barker
of the Physics Department of the University of Michigan.

1

-

A

^te

'■n.

'sf%

ifi

y

1

y

cooa

■fif

^

■^^

-^

p-^

f '

9.

/■

_^

-^

■^

fP

P

£i'

y

y-

^

"^

^

■fc

'■'!

'"

0 00(.

.IP

y

y

^

-

^

^

oooc

P't

^

^

\^

y\

\^

ooo<.

.'(T

<

/

^

,^

^

L^

'^

>-*

.,

_

r/'

A

r'/

■ ^

A/

1 H

_

aoa

IV

^\

1

i

P 1 k

■i

c

,

<

? 1 /

o

/

s

/

^

/

S

/

9

2

0

2

2

2\

■f

2

6

2

S

3

v>

3

2

3

^

vj

S

__

_

_

_

_

__

_

t

Fig 7

Graph of light energy in terms of width of slit. The relative intensities of light
reaching the center 0 and the margin P of the lens aperture is represented. Abscissae
give widths of slit, ordinates give relative light intensities.

Measurements against Bureau of Standards Lamp No.
2435 gave 2.35 candle power for the lamp. As it was 1
meter from the plate when equated, the blue stimulus
patch had an intensity of 2.35 meter candles. It is to be
understood that the blue plate was kept as nearly as pos-
sible at this intensity throughout the experiments.

In order to determine the relation between the light
energy reaching the lens aperture and the width of the
slit, and also to determine the relative light at the center

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH

17

and at the margins of the lens aperture, the curves shown
in figure 7 were prepared under the direction of Professor
John F. Shepard. He also prepared the following descrip-
tion and determinations. The conditions of illumination

Fig. 8

Diagram to show the relations of light energy at the center and margin of the
lens aperture.

were as shown in the diagram, figure 8. The effective
portion of the Nernst filament used was 28 mm. long.
The slit was placed 30 mm. below the filament, and the
lens was 188 mm. below the slit. The distance from the
lens to the highest portion of the stimulus plate which
received illumination was 166 mm. The aperture of the
lens was 80 mm. square, and the stimulus plate was 125

18 CORA D. REEVES

mm. wide, measured on a line parallel to the Nernst fila-
ment and to the width of the slit.

Assume the conditions shown in figure 8 and let W equal
the width of the slit. To find the intensity of illumination
at 0, let z be the range of points on the light-source from
which light reaches 0, and let d be the distance from such
points to 0. Also let @ be the angle which d makes with
the line perpendicular to the plane of the lens at 0. Then
the intensity of illumination received at 0 from each point
on the source is:

K Cos2 6)

Intensity received at O from each point on the source =

d^
d2 = 218- f z2

2182
Cos^© =

%* the total intensity at 0= K

218Mz

(2182 + z2)2
109w

Similarly, let d' be the variable distance from points on
the light source to P; let /3 be the angle which d' makes
with the perpendicular from the light source to the line
PQ; and let 7 be the angle which d' makes with the line
perpendicular to the plane of the lens at P; then the in-
tensity at P from a point in filament equals:

1
K X X cos/i X cosx

(di)2

Perpendicular from filament to PQ

= \/ (218)2 + 1600

di = \/(40 +z)2 + (218)2 + 1600
\/(218)2 + 1600

cos 5 = ■

\/(40 +z)2 + (218)2 + 1600
218

cos;' = — . —

\/(40+z)2T(2r8y2 + 1600

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH

19

Intensity at P=K

/I09

w + 6.38

188

Reducing- Intensity at P=K

[218\/(218)2 + 1600]dz

[(40+z)2 + (218)- + 1600p
109

w -h 6.38

188

109

— w + 6.38

188

48374 dz

[z2+80z+50724P

109
— w + 6.38

188

For the limiting case when the slit is removed,

14

The intensity at 0=K

The intensity at P==K

2182dz

(218= +z2)2
— 14

14

[218\/218- + 1600]dz

[2182 + 1600 + (40 tz)2]2
— 14

From these determinations the curves of figure 7 are plotted.

The curve " Point 0 " (fig. 7) shows the intensities of
light at the center of the lens aperture for different widths
of the slit, while the curve " Point P " shows the intensi-
ties of the light for point P, in figure 8, at the border of
the lens aperture, for different slit-widths. The intensities
are expressed on the ordinate by means of an arbitrary
scale. It is to be noted that there is less difference in the
light distribution between the center and the margin with
a wider or with a narrower slit width than with a slit of
24 cm. Whenever a stimulus plate is illuminated without
the use of the slit, or with any width in excess of 35 mm.
the light passing through the lens aperture at its center
will be as 100 per cent to 90 at the border. From the
graph (fig. 7) it is plain that when the slit width is 24 mm.
this difference in intensity will be maximum or the ratio
will be as 100 to 71, and the difference between the center
and the margin will be less with narrower slit widths. In

20 CORA D. REEVES

the blue-red series of experiments, pp. 25 and 48 the blue
plate was illuminated through a 35mm. slit, the red through a slit
constantly changed in width during the series. Relations
were about as follows for the red :

Ratio of illumination

Slit width at centre and border

35 mm 100 90

24 mm 100 71

14 m 100 82

6 m 100 89

3 m 100 90

Im 100 95

For the blue as 100 at the center to 90 at the border.

C. Experiments to Test Discrimination by the Formation
of a Food Association. — In this section are presented all
experiments in which the attempt was made to establish
a food association, together with a few experiments on
innate response to white-light differences. In the case
of each individual fish used the chronological order of the
tests has been preserved, so that the past experience of
each fish may be known in its relation to any particular
series of tests.

1. Intensity Discrimination of Dace

To test the effectiveness of differences in intensity of
white light in control of behavior, a horned dace or creek
chub {Semotilus atromaculatus Mitchill) which had been
in the laboratory for two years but which had not been
worked with, was placed in the experiment aquarium.
The ceiling lamp was lighted. After some hours the fish
was closed into the retention compartment, the two Nernst
lamps were lighted, and a slit with a 5.0 mm. aperture
was placed in one side of the lamphouse, while the oppo-
site side remained without a slit. The slit cut off about
three-fourths of the total light reaching the stimulus plate
upon that side. (See graph, fig. 7.) Both stimulus patches
looked bright, but the one without the slit was of dazzling
brightness.

By moving partitions A and B (fig. 6) this fish, called
Bu, was brought into the discrimination compartment.
On opening the slide door Bu swam up to the stimulus
patches. Of the first seven responses, five were to the

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH

21

duller plate, and each time the fish fed from the bar on
that side. It seemed that, since the fish had fed from the
bar at once and on the side of the duller plate, only a few-
trials would be required to fix the response. But more
than 150 trials were not enough to show any increase in
accuracy of response, and at no time was a more persistent
" position habit " formed than developed here. The lack
of learning is expressed in the curve, figure 9, which was
made by plotting the average percentages of correct choice
of each twenty consecutive trials.

Because there are individual differences and abnormal-
ities among fish, Bu was taken from the experiment aquar-
ium, and two other dace of the same reserve stock were
placed in the aquarium and tested one at a time, under

RRIGHTNtSS

N umber Of trlnlJ io 40

DISCRIMINATION

60 60 100 120 140

i.

Y-

An

70

(n

/

1

/

\

\n

N

~>

/

\

\

' \

4n

S

S

/

i

\

/

\

jr>

\

>

>

\

y

■^^

?o

1

_

Fig. 9

Graph showing the percentage of correct choices in each 20 consecutive trials in
a series of 140 trials of horned dace Bu. The choice was between two white plates
of relative brightness 1 :4. The fish was fed at the duller plate.

the same conditions as described for Bu. Of the first ten
trials given, seven for each fish were to the duller, or posi-
tive plate. Again it seemed that it would be easy to train
the fish to go to that plate for food, but sixty-five trials

22

CORA D. REEVES

for each fish showed no improvement. Fifty per cent of
the choices were to each plate. The fish failed to learn
to discriminate between white plates illuminated as 1 to
4 (see fig. 7). The early preference for the duller plate
in these trials and the slow discrimination of the dace upon
other problems (see fig. 13) made it seem probable that
with a very long series of tests the fish might learn to dis-
criminate. But neither the 150 tests of the fish Bu, nor
the aimless behavior of the three fish in their swimming
about before the stimulus patches, supports the opinion.
Whatever the capacity of horned dace to discriminate
intensity differences, they probably make little use of such

WhiU

5rnm shr

R<.d

no all t

FIG. 10

Path of the dace along the limiting plane between the area illuminated by light
from the red plate and from the white. (See p. 23.)

differences in discriminating familiar objects. By way of
comparison it may be noted that under the same condi-
tions of illumination these two large intense plates differed
for the human eye when illuminated by slits of 8.0 and
8.5 mm., respectively.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 23

2. Wave-Length Discrimination: White- Red by Dace

While the last two dace were failing to discriminate the
two white plates, a red filter without the slit was placed
on the side of the brighter plate. This cut off the rays
shorter than ^ 5890.5 The illumination of the positive
plate continued to be with white light through a 5 mm.
s^it. The results were most surprising. When the red was
first introduced, each of the fish went to it for the first
three responses. But the manner of approach was notice-
ably different from that to the white plates. The fish
swam more swiftly toward the red plate, or after approach-
ing darted away from it, or sidled along the edge of the
red illuminated area, as shown in figure 10. This sort of
behavior is described later in the case of " Large Sunfish,"
in which it was somewhat more striking. But it is to be
noted that it took place here when the red filter had re-
duced the brightness of the total light to which the fish
were accustomed, by cutting off a considerable part of
its energy. It was not then a response to increased inten-
sity. Each fish after from twenty to thirty trials, equally
divided between white and red or showing a strong position
habit, got the " cue," and for the following thirty trials
gave 80 to 90 per cent white, or correct choices. When
ten successive correct choices had been given, it seemed
possible to conclude that discrimination was by means of
brightness.

Hess (1909) has stated that the red rays have little effect
on fish. If the introduction of the red plate had greatly
reduced for the fish the brightness on that side, then the
brightness differences between the red and white plates
might now be greater for them than it had previously
been between the two white plates; it might be greater
for the fish than 1 to 4. This increased brightness dif-
ference might enable the fish to discriminate between the
red and white, although they had previously failed to
discriminate the two white plates illuminated as 1 to 4.
It has been noted that the fish tended to swim along a
plane separating two white illuminated spaces when there

* It is to be noted that the orange as well as the red wave-lengths, are included
whenever the red stimulus is mentioned in this paper.

24 CORA D. REEVES

was a very considerable difference in their intensity. Their
similar behavior toward the same spaces now illuminated
with red and white might then be regarded as an intensity
response. But the occasional peculiar method of ap-
proach to the red area was in no way like the behavior
toward a dull white plate. It indicated that the fish were
reacting to a difference between the quality of the dull
white and that of the red. To learn whether the fish were
reacting merely to a brightness difference between the
white and red patches or, as indicated by their behavior,
to a quality difference, a method was sought of equating
the brightness of the red and white for the fish.

A tendency of the fish to rise to the surface of the water
in pursuit of floating bubbles and dust had been noted
when the red plate was first used in the red-white tests
just described. This rising was like that often seen in
natural waters on dull days or at sunrise or sunset when
the sky is red. It was possible that under the conditions
the rising was a response to light of any quality but of a
definite brightness for the fish. In that case, if the rising
response were shown toward two lights of different quality
when presented separately, these might be regarded as
of such brightness for the fish as to induce similar reaction.
If the fish then showed differential behavior toward the
two lights, the response must be based on wave-length,
or quality differences. This supposition was tested by
using a very dull white plate in place of the red.

The 5 mm. slit was left above the one plate, which thus
remained illuminated with mixed light, while the red filter
was removed from above the other plate and a second
slit was introduced. This slit was gradually narrowed
in successive tests until it was but little more than 0.1 mm.
wide, but the rising response was not noted. When the
slit reached 0.1 mm., the fish rose to the surface and
snapped at small floating objects. Thus while the white
light illumination of one of the plates remained constant,
i. e., that through a 5 mm. slit, that on the other plate was
changed from red without a slit to white with a 0.1 mm.
slit. The rising response took place with these two con-
ditions of illumination. Apparently the light from the

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 25

red plate without the slit was for the fish in some way
equal to white light through a 0.1 mm. slit, or the behavior
was a chance response.

I decided to see whether the fish could discriminate
between the two light patches that had apparently in-
duced the rising action and were presumably equated in
brightness, i. e., the red without the slit and the white
through the 0.1 mm. slit. I tested them with the red on
one side and this very dull white on the other. The fish
which had given the best response to the white continued
to go to it and to feed before it as I shifted the plates from
side to side, though the brightness of the white presented
was now presumably reduced to that of the red. The
other fish gave slight evidence of a " position habit," but
on the second day responded accurately to the reduced
white. The width of the slit used with the white was again
changed to 5 mm., then to 2 mm,, then to 4 mm. Eighty
per cent of the choices continued to the white. There
seemed little possibility that brightness was controlling
the behavior when a continually shifting brightness for
positive stimulus was presented. The avoidance of a red
plate at first and the ability to distinguish it from the
white plate, even when the latter presumably matches it
in brightness, is a more consistent explanation. The shift-
ing brightness of the white plate did not prevent discrim-
ination. That this red plate had a very definite value
and a very different one from any dull white, was shown
also by a tendency on the part of the dace to avoid the
red illuminated area as compared to their indifferent ap-
proach toward the dimly illuminated white. But this
initial tendency to avoid red was very brief on the part
of the dace. It is doubtful whether such a tendency would
be shown by all individuals.

3. Wave-Length Discrimination: Blue-Red by Dace

a. Red chiefly at constant maximum intensity.

To test the power of dace to discriminate longer and
shorter wave-lengths, the apparatus already described
was used with both the red and blue filters in place. By
this means two patches of light were presented to the fish,

26 CORA D. REEVES

the one of wave-length ^. 4600 to ^ 5245, with maximum
in the blue; the other of wave-length from ^ 5890 to beyond
the end of the visible spectrum. The two filters were quite
different in the degree to which they interrupted the light.
When the stimulus patches were observed with the filters
in place but without the slit, the red was brilliant, while
the blue was somewhat dull but a saturated color. In
order to tell whether the illuminations used were effective,
dace that were tame but without experience in the experi-
ment aquarium were observed when put into compart-
ment 2 and presented with the stimulus patches. They
were not fed. At first one plate was illuminated with the
light passed through the blue filter while the other was
unilluminated. The straight approach to the blue as
compared with the zigzag swimming on the non-illuminated
side showed that the particular illumination was effec-
tive in modifying the behavior of the fish. Then with my
eyes dark-adapted I matched this blue and a red in bright-
ness by varying the width of a slit used with the red. The
red thus reduced was presented to the fish in contrast
with no-illumination. It was certain that the fish again
responded to the lighted area by definite straight approach
but moved in a zigzag course in the non-illuminated side.
When the slit was removed and the brilliant red presented,
while the other plate remained unilluminated, the approach
to the red area was in a straight line; but the fish came
only half the distance to the stimulus plate from the slide-
door, then turned suddenly across the aquarium into the
dark half, and retreated. Other tests showed that the
brilliant red gave strong stimulation.

It was evident that both red and blue patches of light
were effective as stimuli for the fish, and that without the
slit the red stimulated more strongly than the blue. It
now remained to discover whether the fish could learn to
go to the blue for food, and whether at any intensity the
red would give the same stimulation as the blue. If the
fish could discriminate the red and blue patches at all
intensities of red, it must be concluded that discrimination
was on the basis of quality and not brightness. If, at any
intensity of red, discrimination of the two patches became

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH

27

impossible, it must be concluded that the patches were
then of equal brightness for the fish and that they were
unable to discriminate them by their quality difference.
The first tests were made without the slit, that is, with
the red much more intense than the blue. They are pre-
sented in this section, while those with shifting values of
red appear in the following section. The plate illuminated
by the spectral band ^ 4600 to ^ 5245 (blue) was chosen
for the positive plate, and the fish were fed before it and

Fig. 11
The First Record of Fish H (see p. 28).

not before the red. The experiments were made with two
horned dace, Md'^ and //J" . These fish had been in the
laboratory a year, and the previous spring had been tested
with red and blue in the experiment aquarium for some
weeks. When they were again introduced into the experi-
ment aquarium on the morning of October 13, 1914, they
showed absence of fear by coming within two minutes to
feed from my fingers. A detailed account of a single ex-
periment will serve to illustrate their general character.

28 CORA D. REEVES

At 1:30 P. M. the curtains were drawn to exclude all
extraneous light from the experiment aquarium, and the
ceiling light was turned on. At 2 P. M. the Nernst glowers
were lighted. By the method already described the fish
were brought into the discrimination compartment. The
slit was not used, so that the brilliant red patch and the
saturated blue were presented. The slide door was raised
a short distance. The two dace were swimming about
the top of compartment 2, figure 2, and did not appear
to note the opening. At 3:21 P. M. the slide door was
again opened. H went into the blue lighted area, and
although the food-bar was released and he remained near
the end of the bar with its piece of worm, the food was
not touched. His path is shown in figure 11. The ap-
proach to the red and then the straight path across in front
of that plate appear to show an inhibition of further ap-
proach to the bright red area. At 4:20 P. M. the door
was again raised and both dace swam out in the wavering
paths characteristic of exploring fish. One went to the
blue lighted area and fed. H was then closed into the
retention compartment, while the fish Md was tested.

A typical record (fig. 12) follows:

P. M.

4:53 Slide door opened.

4:54 Swimming back and forth in the discrimination compartment; pauses fre-
quently with head in midline.

4:58 Pauses on either side of the midline.

5:00 Went out to blue area and up into the direct blue light; stayed there a time;
fed.

In this preliminary work H was next tested. The first
thirteen out of fifteen responses were to the blue. When
Md was again tested thirty-eight out of forty responses
were to the blue.

For the next ten trials of H the slit was used, so that
the light intensity of the red was reduced to 66, 37, and
20 per cent of its maximum value. The fish went to the
blue area nine out of ten times with this shifting intensity
of red. It appeared that there was a decided tendency
to approach the blue rather than the red. It was now
evident that this was true even when the bright red was
reduced in intensity.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH

29

During the following weeks a separate series ot tests
with the red at maximum value was made for each fish.
The percentages of blue choices for each ten trials were:

For H: 80, 90, 80, 60, 40, 90, 70, 30, 60, 70 per cent.

For Md: 85, 50, 70, 70, 70, 80, 50, 80, 70, 90, 100 per cent.

The variations in the records from day to day, as shown
in the above percentages and in the graphs, indicate that

ffeJ

^-)o

Fig. 12
Path of Md going from the discrimination tank to the blue illuminated area, p. 28

uncontrolled factors modified behavior. In figure 13, at.
the left over the heading " Red Maximum " the records
of these fish are shown in graphs plotted from the average
percentages of correct choices in each twenty successive
trials. In the graphs of Md and H the first 20 preliminary
trials are also included. During the period covered by
this series there were variations in the quality of the water
supplied to the aquarium, due to changes in the filter
cores, which become porous, and variations in the light
due to changes in the electric current, which affected the
intensity of the ceiling light. Besides these external vari-
ables some manifestations of breeding behavior indicated

30 CORA D. REEVES

an internal factor. A further indication of an internal
factor was the behavior of fish H. He jumped from the
aquarium once or twice and this jumping was on days
when the fish had approached the bright light. In its
native brook the male fish in the spawning season remains
in the brightly lighted shallows rather than hiding in the
dark pools (Reighard 1910). His jumping from the aquar-
ium may have been a positive reaction toward stronger
light. In spite of these modifications of behavior the
record indicates that the fish discriminated between the
red and blue when the red was very bright. This is espe-
cially true of the fish Md.

b. Tests of dace Md. Blue with red decreasing in intensity.

Since discrimination appeared to be possible with the
red at the maximum intensity as well as at 2/3, 1/3, and
1/5 maximum, Md was again tested while the slit on the
red side was gradually reduced in width so as to give a
wide range of reduced intensities. The results are shown
in figure 13, graph Md, at the right over the heading " Red
decreasing." Her percentage of red choices remained be-
tween 85 and 95 as long as the slit opening was more than
1 mm. At a slit width of about 0.9 mm. the percentage
of correct choices was only 60 for the first twenty trials
(trials 201-220). But when the slit width was kept at
about this value for twenty additional trials, the fish made
^3) per cent of correct choices (trials 221-240). With greater
reduction of the light intensity of the red area the fish
continued to show a high percentage of positive or blue
responses. This happened when the red was reduced in
brightness so as to be duller than the blue for the human
dark-adapted eye, and finally so dull that it had little color
value to the human eye. The slit width was reduced suc-
cessively to 0.9 mm., 0.6 mm., and 0.4 mm. Following
this a photographic negative" was interposed to further
reduce the light and was used with slit widths of 0.6 mm.,
0.4 mm., and 0.2 mm., successively.

^ A photographic negative was made by passing the Hght through the blue filter
so that any local differences in the amount of light transmitted by the filter might
be duplicated on the other side.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH

31

JiSmm l^/c///? o/j////n mrr?

^ Afc7/cA,9e/ ^ J>u//er- ^

20M F-2 2/ as ac ae-a^' m^Arr

Md

' 1

QO

1

-A

V

A{/

1

'eTSd

-~^

\ 1

ffO

\

1

\

?~

L

\

a

u'°

u

\

%,

<r)

£f^

^

^

1^

^

\

/

'\

aSr

^ilO

4>

^«7

\

/

\

tvn

i>>

ea

eio

so

,

/OO

•^/Tjm

JiS.-£S

SSS

J

S-/.5

G/.5'

neoa/zt^f*

1

QO

/4//SO

41

f^

\

f!n

S

\

\

y

'-.

^ ^

,<

\

^o

\

a-

io

\

/6I

/SO

/S/iOO

\

\

<^

\

\

\

1

\

^c

1

£/-

*o

V

'°'-

/!0

420

H

yoo

J^/^/77

JS-i5

/sz

/■S

a.9

as-of

6-/fS'/'rQ''//yfi .

2!/-,t«0_

»l/-260

90

V

/-PO

2/

«o

/

so

r

-1

/IV

/an

/

^

y

\

'W

f^

/

FO

\

„,

4

\

N

,^

P\

2t

Vil

•o

*4

4f-N

k^

90

/

/<4

'/so

f^O

\

/^

\

s

7

■w

\

/

*/■

^

:fSZ

'

YP

,3iS^

"

JO-20

30/0

3-2

ae/s-

•^s

oA

'<«

ry?

\

\

-

!^n

\

/«

y/20

L

s/

/oo

f~~^

~1

■--.

/z/

/'M

•^0

>

P

\

^

r

/W

/60

fiO

\^

\

1

^

^

^

<JrJ

\

1

M

1

/?«/.

rr7c?Ji fm cyr?^

/r*e<^ ^p<rr'&€^j ^^^

Fig. 13

Graphs made bv plotting the percentages of correct or blue choices for each suc-
cessive 20 trials of the four dace, //, Md, YL, YP.

Abscissae indicate successive slit widths in millimeters.

Ordinate? percentao;e of correct choices

Short Dash hnes show rapid changes in slit widths.

Long Dash lines indicate that no tests were made at intermediate widths.

Numbers on the graphs indicate successive groups of 20 trials from which
percentages were calculated.

32 CORA D. REEVES

c. Tests for the dace H. Blue with red decreasing in
intensity.

The fish H was tested day after day in the same manner
as Md (fig. 13, graph H at right), except that the reduc-
tions in the width of the sHt, and therefore in the intensity
of the red plate, were more rapid. Changes in slit width,
cutting off 15 per cent and 50 per cent of the light, did not
cause failure to discriminate. When the intensity of the
red plate approached the point where it matched the blue
in brightness for the human dark-adapted eye (slit width
1.5 mm. to 0.9 mm.), the changes in slit were made more
slowly. No failure to discriminate resulted, as the plotting
of correct choices shows (fig. 13, graph H, red decreasing).

d. Tests of dace Yl and Y P. Blue with red decreasing
in intensity.

Other dace, Yl and Y P, were introduced into the aquar-
ium two months later than H and Md. Their records, also
shown in figure 13, are less regular, but they show some
points where the discrimination was very accurate at slit
widths at which the records for Md and H were not taken
for as long a series.

e. Discussion of blue-red graphs of the dace H, Md, Yl>
YP, and conclusions.

The graphs of these fish, made by plotting the percent-
age of correct choices for each successive twenty trials,
and distributing these according to the width of slit used
with the red variable, are shown in figure 13. The slit
widths appear at the top of the graphs, and the corre-
sponding relative intensity of illumination of the red filter
may be read from the curve, figure 7.

Each graph falls into two parts, as indicated at the
bottom of figure 13. During the first part the red remains
at maximum intensity. During the second part the red
is slowly decreased in intensity by means of the slit. Each
graph (except YP) may also be divided into three parts,
determined by the human dark-adapted eye as indicated
at the top of the figure. In the first of these the red is
brighter than the blue; in the second the region of matched
brightness for the human dark-adapted eye is reached;

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 33

while in the third the red is duller than the blue. The
introduction of a negative to further reduce the intensity
of the red is also indicated at the top of the graphs.

The general features of the entire graphs may first be
noted. At first there was avoidance of the bright red plate.
This shows in all the graphs in the high percentage of blue
choices at the beginning. After the initial avoidance,
there follows in the record of each fish a time when the
bright red and the blue were approached with nearly equal
frequency, as shown by the fall in the curves toward 50
per cent. This drop in the curves probably indicates that
the fish were becoming accustomed to the bright red so
that they no longer avoided it, while the association blue-
food had not yet been formed. After the curves have
descended they again rise while the red is still at maximum
value. The curves thus reach a maximum indicating 80-
100 per cent of correct or blue choices. Their rise is evi-
dence of the formation of the association blue-food. With
reduction of intensity of red the curves again descend to
a level indicating but 50 to 70 per cent of correct choices.
Thus as the brightness of the red patch is reduced to that
of the blue for the human dark-adapted eye (slit width
about 1.0 m), the ability of the fish to discriminate the two
diminishes, but in differing degrees in different individuals.
Failure to discriminate is evident in Md at a slit width of
about 1 mm. (trials 201-220). In Yl the failure comes
at slit width 5 mm. (trials 101-140). If no other trials
than these had been made at these slit widths the curves
would show for these two fish failure to discriminate at
a certain value of the red variable. It would be fair to
conclude that at this red value the red and blue were
matched in brightness for the fish and were indistinguish-
able by means of any quality difference. The entire
curves might then be interpreted in terms of brightness
differences in the two stimulus patches. But when the
fish failed to discriminate at a brightness of the red corre-
sponding to slit width 1.0 mm. to 0.9 mm., in the case of
Md and width 5 mm. in the case of Yl, the trials were
continued at these widths (trials 221-240 for Md. 121-160
for Yl). The curves of Md and Yl then rise, and a return

34 CORA D. REEVES

of discrimination is thus shown. This return of discrim-
ination is probably due to the failure to secure food, which
has made the fish hungry and alert. With further reduc-
tion in the intensity of the red, discrimination is main-
tained in varying degree by all the fish. Discrimination
continues when slit widths of red are reduced to 0.6 mm.,
0-.4 mm., and 0.2 mm., and finally when the interposition
of the photographic negative has made the red so dull
that it has scarcely any color value for the human eye.
That the difference in distribution of light over the plates
does not control the behavior of the fish appears from the
results. For it is at slit width of 35 mm. for both plates
that discrimination is easiest and at that width the light
distribution is identical for the two plates. On the other
hand discrimination is more difficult at slit width 5. to
0.9 mm. for the red plate and in this region there is a dif-
ference varying from 0.1 per cent to 10 per cent in the
relative intensities of center and border in the variable
plate. The ease or difficulty of discrimination does not
vary with variation in illumination of the variable plate
(Note the graph of fish H.) It is shown elsewhere that
the dace did not learn in my own experiments to dis-
criminate intensity differences less than 1 :4. The most
accurate brightness discrimination found by any worker is
that reported by Hess (1909) in which differential response
was obtained to two white plates illuminated as 1.0 to 1.23.
When the graphs of the four dace are compared, it is at
once evident that they are not alike. Identity is not to
be expected since fish differ in the degree to which they
are influenced by external stimuli, in readiness of response,
and perhaps in ability to discriminate wave-lengths. If
we compare first the ' red maximum ' part of the curves,
it is seen that the initial fall is more gradual in Md and H
than in the other two fish. In both of them it involves
100 trials as compared with 20 in YP and 17, and in the
case of H there is no descent of the curve for the- first 40
trials. Both H and Md had to be taken from the aquar-
ium during the first part of this series of trials in order
to perfect the adjustment of the apparatus. This prob-
ably accounts for the greater number of trials required

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH

35

for them to " get used " to the red. H was of distinctly
different temperament from Md; he gave more evidence
of fright reactions and was more " cautious." Doubtless
his failure to approach the red until after 40 trials was
due to these peculiarities. The curves of Yl and YP are
then more typical than those of // and Md in so far as they
show a rapid adaptation to the red stimulus patch. Dur-
ing the ' red decreasing ' part of the curves, temporary
failure to discriminate is shown for Md at slit width 0.9
mm. In Yl it occurs at width 5.0 mm. In the case of
YP, as the graph shows, no trials were made between
slit widths 2.0 mm. and 0.6 mm. Had trials been made
at this width, i. e., between trials 120 and 121 of this fish,
a drop in the curve would probably have taken place.
Although trials were made with H at slit width 1.0 mm. and
0.9 mm., no considerable drop occurs at this point or else-
where in his curve. The curve shows merely a lower region
between slit widths 2.0 mm. and 0.9 mm. Somewhere in
this region the stimulus patches probably matched in
brightness for this fish. The fact that this fish discrimin-
ated more accurately than any of the others when the red
was' greatly reduced in intensity (trials 221-260) is con-
sistent with his better discrimination at matched bright-
ness. The greater caution of this fish has been referred to.

JO ^O /O S-2 /OSI OS'n^o/'^e

SO

eo

i

X

/

X

/

\

i

V

\

/

1

Y

'

—

-D^r-f

.

FIG. 14

Graph made by plotting the percentage of correct choices found by averaging
all the records from the four dace for each successive width of slit. The individual
graphs of these dace are shown in figure 13. Abscissae indicate successive slit widths
in millimeters, ordinates percentage of correct choices at the given widths.

36 CORA D. REEVES

When we compare the curves of the fish Md and 17, in
which a distinct drop marks failure to discriminate, we
find that this failure does not occur at identical slit widths
In the two. It occurs about 5.0 mm. for 17 and at 0.9 mm.
for Md. It is probable that in this region of decreasing
red-brightness the exact point at which discrimination is
lost and recovered depends on the chance rushing of the
fish toward a red plate similar to the blue in brightness.
When this occurred, it was shown by a fall in the percent-
age of correct choices — a fall in the curve. The slit- width
was then held constant. As hunger increased through
failure to get food, it may be that the fish became more
alert and looked at the plates before it swam toward them.
The percentage of correct choices then increased. Ability
to discriminate between the plates was present at all these
slit widths, but became manifest only when hunger helped
the association blue-food to function in behavior.

If we take the records of these four fish, more than eight
hundred in number, and plot the averages of correct choices
at the successive widths of the slit, the curve (fig. 14) shows
a region from 5.0 mm. to 0.9 mm. in which discrimination
is less accurate. This region includes the point wher'e the
blue and red are matched in brightness for the human
dark-adapted eye. But even here there is over 70% of
correct choice; while the average percentage of correct
. choices for the whole series is a little above eighty. The
individual curves show that at every intensity of the de-
creasing red the fish sooner or later discriminated. The
composite curve shows no region of very low discrimina-
tion. It appears to follow from these experiments that:

(1) Dace do not easily learn to differentiate white
stimulus patches of considerable difference in intensity
(1 to 4).

(2) They discriminate promptly between red and white
with variation in the intensity of the white, which in-
cludes a white intensity inducing a similar response to that
of the red used.

(3) They discriminate between blue and red at all in-
tensities of red. While there is at first lack of accuracy m
response when these two colors were similar in brightness

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 37

for the fish, response becomes more accurate as trials with
similar brightness are continued.

4. Intensity Discrimination by Sun?ish as Indicated by Response

To study the relative sensitivity of sunfish to white
light of different intensities, tests were made with some
small (4 cm.) untrained sunfish. These fish had shown
themselves sensitive to lower intensities of light, than the
dace, for they stayed motionless in the farthest corner of
the aquarium if brilliantly lighted patches were presented.
The stimulus patches were therefore reduced in intensity
by the use of slits on both sides. The fish were not fed,
but their reactions were observed when the stimulus patches
were exposed by lifting the slide door. When one plate
was illuminated through a slit of 0.4 mm. and the other
through one of 0.6 mm. there were slight differences in
behavior toward the two. One record shows a straight
return along the midline; another shows that in approach-
ing the patches the fish swung back and forth to the mid-
line. These turns, doubtless, were responses to a change
of stimulus intensity at the mid-line. When one stimulus
plate was illuminated through a 0.2 mm. slit and the other
plate through slits successively 0.2 mm., 0.25 mm., and
0.3 mm., I could determine no difference in the behavior
of the fish toward the two plates. The approach to either
plate was in a smooth, straight line at these narrow widths
of the slit. But when the two slit- widths were 0.2
mm. and 0.4 mm., then the approach to the brighter side
was no longer straight but slow and wavering. When in
this series the slit- width for the brighter plate was increased
to 0.6 mm., while the other remained at 0.2 mm., the fish,
instead of approaching the plates, lurked in the corner
farthest from them. Reference to the graph (fig. 7), which
indicates the light intensities reaching the aperture of the
lens as varied by slit-widths, shows that with slit-widths
0.2 and 0.4 mm. the intensities were about 1 to 2 (as they
would be for 2 mm. and 4 mm.). Hence a capacity to
respond to intensity differences of white light somewhat
greater than a 1 to 2 ratio is present in sunfish. This result
was obtained by the method of response.

38 CORA D. REEVES

Two years after the tests just mentioned, one of the
sunfish' which had meantime been trained in the discrim-
ination of the colored plates, was tested by training meth-
ods for brightness discrimination of white light. A 1 to
2 intensity difference was not found effective even after
long training. At first, slits of 4 mm. and 0.6 mm. were
used, and food was given in front of the duller plate. The
first fourteen responses were positive, i. e., to the duller
side. The width of the wider slit was then reduced. When
the slits were 1.8 mm. and 0.6 mm. the fish showed 70 per
cent of correct choice. When the slit- widths were 1.2 mm.
and 0.6 mm., a long training failed to secure more than 50
or 60 per cent of responses to the duller plate. It is there-
fore certain that under training methods this sunfish was
able to make use of a difference of intensity of 1 to 3 corre-
sponding to slit-widths 0.6-1.8 (see fig. 7, graph for Point
0) but not of a smaller difference under the conditions of
illumination already described. It is possible, as indicated
in the preceding experiment, that its unlearned response
is the more accurate.

5. Wave-Length Discrimination. Blue-Red by Sunfish, with Red Decreas-
ing IN Intensity

a. General Account. — Three sunfish were used for this
test, and each was separately tested.

A 10 cm. female sunfish (Eupomotis gibbosus) named
Large Sunfish was put into the experiment aquarium at
the same time and subjected to the same training test
under the same conditions as the dace Md and H (p.
25). The red and blue stimulus patches were presented,
but for a time the fish lurked in the dark corners at the
farther end of the discrimination compartment. Then the
fish when tested approached to the blue illuminated area
in a series of rapid mouse-like movements, but it refused
to eat before the lighted plates. It required more than a
month to so build up the food association with the blue
area that shifts could be made in the intensity of the red
plate without inhibiting the response; and even then sev-

^ This sunfish was " Small Sunfish " and was about 8 cm. long when used for
these tests.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH

39

Sf/'a/?/)?/'

y /^o/cAec/ ^

jDa//e.

Mc/A 0/^/,/,,

1 ^rtrrj

36

1

20

1

/0-P

1

J

/L6

1

/.3

1

/.3-/J5

1

OP

1

o.e-

1

ntya/tre ~

.5

/4J /so

1 1

9<

,

/

'\

i6/2S0

/a

/"'

/

1

\

\

1

^a/'xo 1

S<

3

•#

•-so

e>^

so

^

-

■--

^^

/

\

' 1

32/-340

1

,/

p

a

/OO

/It

■/40

\

/

\

/

?o

,,'i

,

/

^

/

N

/

\

s

/

1 1

fiC

\

/

\

^

^0/^30

\

/

/J/

^100

so

2/- 40

1

Z or^f xyiy/i/zbA

^90

/oo

jiS^^ js^^ 2 /^ 09 0.9 09 as aeo^i

' 'I'll' „^^/,-,^

1 1 1

.QO

/

|S

'

■fO

/O/

-■

\.

vac

m

\

61

■so

/O/

/PO

\

-"

^

■^

/-

/'

1

FO

4/-

SO

SI

-/Ol

j

\

eo

\

\

30

/2/-

140

^c

'-PO\

. 1

<-5^rpa//JL^uy7/y3A

/OO

S5^n^ JSmm P 2 /.5 /.S 09 a9 OS S-/S

i

\

00

4/-S0

h

\

so

/

\

/O/^'SO

/

^

/

\

/4'/-/sa\

\

^

A

N

\

A

\

y^

\

/ 1

,/

a/

so

/2/-/40

!

1

/

/

1

1

.10

^/-ao

1

/?ec/

€Cf rTTaAfrrfur?-?

^ea c/<pcr'e<73fn^

Fig. 15

The per-cent of correct choices of 20 successive trials for each of three sunfish
Dotted lines indicate rapid changes in width of slit.
Abscissae indicate successive slit widths in millimeters.
Ordinates percentages of correct choices at the given widths.
The numbers on the graphs indicate the successive groups of 20 trials from which
the percentages were calculated.

40 CORA D. REEVES

eral minutes might elapse after the sliding door had been
opened before the fish would leave the discrimination com-
partment. The inconvenience of this slow approach was
in part compensated for by the rapid rush back to the
discrimination compartment when the test was over. A
series of tests was finally made with this fish, with gradual
reductions of the width of slit.

Two small sunfish, which had had the same treatment
as Large Sunfish, were brought into the experiment aquar-
ium about the time that dace Y P and Yl were introduced.
Records were made of their first approach and behavior
toward the red and blue stimulus patches in contrast with
no-illumination. The eye movements of the fish, their
orientation, and the straighter paths followed in approach-
ing the illuminated plates, showed that both colored patches
were effective in modifying behavior. Tests for the two
small fish, Small Sunfish, of undetermined sex, and 552,, a
male, began in November, 1914.

The graphs of these three fish, made by plotting the
percentage of correct choices for each twenty successive
trials and distributing these according to the width of slit
used with the red variable, are shown in figure 15. The
slit widths appear at the top of the graphs and the corre-
sponding intensity of illumination of the red filter may be
read from the curve, figure 7. Like the graphs of the dace
these may be divided into two parts, one in which the
red was held at maximum intensity (35 mm. slit) and one
in which the red was of decreasing intensity. They are
also divisible into the three parts shown at the top of the
figure, in the first of which the red is brighter than the
blue for the human dark-adapted eye, while in the second
the colors are matched in brightness, and in the third the
red is the duller.

In the ' red maximum ' part of the graphs it is to be
noted that that of Large Sunfish is essentially like the
graphs for dace (fig. 13). It shows (1) an initial avoid-
ance of red. indicated by a high per-cent of blue choices
(trials 1-20), (2) a subsequent getting used to the red
(adaptation) (trials 21-40), shown by a drop in the curve
to the 55 per cent level, and (3) a final establishment of

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 41

the blue-food association, /. c. the learning to discriminate,
indicated by the rise in the curve to the (i^S per cent level,
The curve of 552 shows the same divisions (trials 1-60), but
since this fish was given no trials with the red at maximum
value the whole of its curve falls within the red decreasing
part of its series of trials. In the case of Small Sunfish
there is no initial avoidance of red. Its curve rises imme-
diately from the 45 per cent (trials 1-20) level and ascends
to 85 per cent (trials 21-40) as the food association is
established.

In the red decreasing part of the series of trials the
graphs again show a general likeness to those of the dace.
In each of these curves there is a drop, shown most typ-
ically in the curve of Small Sunfish. Here it corresponds
to slit width 0.9 mm., at which point the curve falls to the
55 per cent level (trials 121-140). This is in the region of
matched brightness for the human dark-adapted eye and
shows an initial failure to discriminate at this value of the
red. As the slit-width is held at this value, discrimination
is re-established in vSmall Sunfish (trials 140-180), and its
curve rises above the 90 per cent level. A similar but less
pronounced drop in the curve occurs in 552 (trials 61-80)
and is followed by recovery. In Large Sunfish the drop
occurs at 1.80 mm. slit (trials 181-200) and is followed by
recovery. In general the sunfish graphs show the same
characteristics as those of dace. In the composite sunfish
graph (fig. 16) there is, however, a shorter region of inac-
curate response with decreasing red brightness than in
that of the dace. This might be expected from the greater
ability of the sunfish to discriminate intensities of white
lights. Their brightness discrimination has been shown
to be more accurate. The peculiarities of the individual
sunfish graphs may now be considered.

b. Large Sunfish.— The long graph of this fish differs
from those of the other sunfish in showing a lack of dis-
crimination at slit- widths 1.8-1.5 mm. rather than at
0.9 mm. None of the other experiments with red and blue-
indicates that in this region the brightness of the red
matches the blue for sunfish. Small Sunfish and 55o gave
from the first high percentages of blue choice at these

42 CORA D. REEVES

widths, as did three untrained sunfish which were tested.
Two years after the experiments represented by the graph
of this fish, Large Sunfish was again tested with blue against
red with a 1.8 mm. slit. The first responses, made after
16 months absence from the experiment aquarium, showed
great differences in behavior when compared with the
earlier tests of this fish. It at once went toward the stimulus
patches and either waited at a little distance or jumped
toward the end of the food-bar. There was apparently
retention and recall of kinesthetic behavior. I could dis-
cover no memory of the blue-food association. The avoid-
ance of red was less marked than in the first responses of
two years previous. A " position habit " soon developed.
More than ten days passed before I discovered any differ-
ential response toward either plate but then noted that
there was a difference in the length of time for the response
to blue as compared with that to red. Then some days
later the discrimination between the blue and red with 1.8
mm. slit became definite and exact. The ability of this
fish to discriminate at slit widths 1.8-1.5 mm. does not
then differ from that of the other sunfish. Its earlier failure
at these widths may have been due to some internal factor
or some undiscovered effect of manipulation. No further
trials were made with it at slit width 0.9 mm. The curve
of this fish indicates that it formed the blue-food associa-
tion more slowly than the smaller sunfish. Eighty trials
are required to bring the curve to the 85 per cent line (trials
40-120), while in the other fish this level is reached in twenty
trials. This slowness is perhaps due to the greater age
of Large Sunfish and the fear behavior at first so dominant.

c. Small Sunfish. — Small Sunfish learned the food asso-
ciation very quickly and gave high percentages of correct
choice with the red at maximum intensity, but did not
show the initial avoidance of the bright red plate shown by
all the other fish. For this exceptional behavior I have no
explanation other than that the fish lacked fear of the
bright red plate as shown by the fact that it fed immedi-
ately from the bar. As the red light was reduced in intensity,
there was no lack of discrimination. The percentage of
correct choice remained about 80 per cent until a 0.9 mm.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH

43

slit was reached (tests 121-140). There was then, at first,
a very great drop in correct discrimination, but with the
slit held at this width the fish learned to go regularly to
the blue for food, and when tested further with the intensity
of the red lessened, showed no confusion.

d. Sunfish SS2. — The small male fish, 552, was first
tested with the slit at 7 mm. (graph 55.2 in fig. 15). For a
time there was avoidance of this relatively weak red as
shown by the 70 per cent average of correct choices for
the first twenty trials. As the red was weakened, the
tendency to approach it is shown by a marked fall in the
percentage of correct choices at 2 mm. slit (trials 21-40),
but the learning of the food association followed (trials
41-60). The record then gives a curve of learning more
typical than that of the other fish (fig. 15, trials 21-160) ,

^^ ^^. / _-

X-/0 2

/s as as-rri^i

7//>^e

/fyfi

-

-

—

1

'^0

80

1,

t

p— J

^

~

^sj

\

-

P"

T

PO

f!0

■5(9

Fig. 16

Graph made by averaging all the percentages of correct choice of the three sunfish
at the successive widths of slits indicated.

with fluctuations as the red became duller, but no lack
of discrimination after the food association had been estab-
lished. When the negative already described was used
to further reduce the light, there was at first a few red
choices (trials 121-140); but, as with the other fish, there
was recovery, and discrimination became more accurate.
The percentage of correct choices was then 75 or over.
When the graphs (fig. 15), made by plotting the per-
centage of correct choices for each twenty tests for each

44 CORA D. REEVES

of the three sunfish, are compared, it appears that with
all three fish discrimination was easier and more exact
when great brightness differences were present; but there
was no intensity of red light tested where some fish did
not show ten consecutive tests, with 90 to 100 per cent of
correct choices. The average of all the sunfish records
plotted, as in figure 16, fails to give a learning curve for
two reasons: (1) there is at first an instinctive avoidance
of the bright red plate; this gives a high initial percentage
of correct or blue choices; (2) although the curve of 552 is
similar in shape to that of the other fish, its successive high
and low points occur at different values of red light intensity.
The composite curve shown in figure 16 is, therefore, not
a curve of learning; it shows merely that under the condi-
tions of the experiment there is, at each intensity of red,
some individual in the group that discriminates, so that
the composite curve never drops to the 70 per cent level.
When the composite sunfish graph is compared with
that of the dace it is seen (1) that the percentage of correct
choices in the region of matched brightness for the fish is,
on the average, higher for sunfish than for dace (compare
figs. 14 and 16); (2) that the range of slit- widths where dis-
crimination is less accurate is shorter for the sunfish than
for the dace. Discrimination fails for sunfish at slit-widths
2-0.9 mm. and for dace at widths 5-0.9 mm. This is in
accord with the greater capability of sunfish to discriminate
white light intensities. The more exact aim of the sunfish
in capturing their food is perhaps evidence of their greater
visual acuity.

D. Tests of the Apparatus and its Manipulation.

In the preceding account of this experimental work it
is assumed that the fish are guided in their choice of stimu-
lus patches solely by the patches themselves. But the
operator must frequently shift filters and slit from side to
side and must at each test pull the door-string and operate
the electric switch while looking through the peep-hole.
It is, therefore, possible that the fish were guided to a
correct choice by some one of these manipulations. It is
possible that some difference in food-bars or filters of the

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 45

two sides may have aided the fish. Since the slit was used
chiefly on the red side, on which the fish were never fed,
they may have learned by seeing the reflection of the slit
on the lens to go to the opposite side for food. Check
tests were, therefore, made in order to learn whether any-
thing else than the stimulus patch served to guide the fish.

1. Check Tests by Changes in Manipulation

a. Shifts of slit and filters as a clue. — The fish were in-
variably shut into the retention compartment before the
apparatus was adjusted for the day's trials. Since they
were then unable to see the stimulus plates or other parts
of the apparatus, these could afford them no clue in making
their first choice for the day. To see whether their choice
in subsequent trials was possibly influenced by the shift-
ing of filter and slits from^ side to side, I often introduced

make believe " shifts. These were interpolated in the
course of the regular series of the day at times when the
fish were discriminating accurately, and consisted in going
through the manipulation of shifting slits or filters but
without actually making the change. The responses of
the fish were in no way affected.

b. Position of the experimenter as a clue.— To learn
whether a possible clue was afforded by the movement
of the string or curtain or of my eyes, when I was in the
usual position at the peek-hole outside the curtain, I fre-
quently made tests while inside the curtain or in some
other unusual position. None of these altered the responses
of the fish.

2. Check Tests by Change in the Apparatus

a. Food-bar. — Although, as already stated (p. 11), both
ends of the food-bar were kept as nearly alike in appear-
ance and position as possible, nevertheless, new food-bars
were introduced from time to time, or the bar was occa-
sionally removed and the fish were fed by hand if they
remained before the blue plate. These changes made no
difference in the percentage of correct responses of the
fish.

46 CORA D. REEVES

b. Lenses. — At the close of the red-blue series I slipped
•out the lenses and tested the fish with only the red and
blue filters in place. This reduced the intensity of the red
and blue plates and changed the distribution of the light.
Although the fish acted uneasy, they continued to dis-
criminate.

c. Filters. — Because the blue filter showed some irregu-
larities in transmission, which might be the sign by which
the fish knew where to come for food, a new filter was
obtained. The red filter was changed several times, but
these changes did not affect behavior.

d. Slits. — In the experiments described a slit was used
to regulate the light on the red side only. A mirror was
used to determine the distance at which the image of the
slit on the cylindrical lens was visible. The curtain already
mentioned was used to hide the lenses from the fish (p. 14.)

After the experiments had been completed a further
test was made on untrained fish to learn whether by any
possibility the difference in the appearance of the lenses
could have influenced the behavior of the fish. This con-
sisted in extending the vertical partition C, fig. 1, to the
doorway D. The fish thus chose while still within the
discrimination compartment at a point from which there
was no possibility of their seeing the lenses. It was found
that the fish could be trained to go to the blue plate after
the partition was introduced. The fish had not been
guided by the reflection of the slit on the lens above the
red stimulus patch.

As a further check a suspension of a very fine bone-black
in a glycerine-water mixture was substituted for the slit.
This suspension in different concentrations was placed in
a glass-bottomed container above the red plate. By this
arrangement there was no slit on either side. The blue
patch remained rectangular and sharp bordered, but,
owing to reflections from the carbon particles, the borders
of the red patch were now indefinite. The small sunfish
swam up to the blue at all intensities of red, including
those which matched the blue for the human dark-adapted
eye, as did also the dace H . The dace 17 circled about and
gave indefinite responses, while YP went to the red plate.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 47

The difference in the appearance of the red plate might
easily account for the behavior of 17 and YP, while the
discrimination of H and the sunfish showed that they were
not depending upon the presence of the slit in order to
reach their food.

Further evidence was furnished when a " new " blue
filter was put into use. This let through the same range
of wave-lengths as the old, or standard blue filter, but
transmitted more light. In order to reduce the intensity
of its color patch to match the standard blue, it was neces-
sary to use a slit. The seven trained fish were tested with
this " new " blue with the slit, and the red at maximum
value, i. c, without the slit. The result of this arrange-
ment was to transfer the slit to the blue side and leave
the red side without a slit. Of 34 trials 30 were to the blue.
It was clear that when the colors were not matched in
brightness the fish could discriminate with the slit on the
blue side quite as well as they had hitherto discriminated
when it was on the red side. The slit did not afford them
a clue.

A short time after this the same fish were tested again,
by the use of the slit with the new blue filter as in the last
tests. To reduce the intensity of the red without the aid
of a slit three separate means were used; the carbon suspen-
sion already mentioned, a solution of saffranin, or a solu-
tion of potassium sulfocyanide with ferric chloride. By
use of the flicker photometer the new blue with the slit,
and red reduced by one of the solutions were matched by
varying the amount of the solution above the red plate.
By this arrangement the slit was on the blue side while
the red was without a slit. The experiment differs from
the other experiment only in that the colored plates were
matched in brightness. The distribution of the light over
the two stimulus plates was again different because of
reflections from the surface of the solutions employed on
the red side. The results obtained are presented in
Table I.

The dace YP and 17 failed to discriminate. The record
of H may also show lack of discrimination. In the case
of the sunfish the definite straight lines of their paths to

48

CORA D. REEVES

<:

CO ^
CD ,£5

o
a; a;

rv

•i3 k:i

w

o ^

0) •-M

O CD

O

Ph

0)
(h

CD

0)

Z

in

u

U

<

Q

c/5

X

M

•xxxx

30

g

N

3

M

xx«xxxx-x..><

a;

■■•■XXX

CO

3

;;;X;;;;-X—XXXXX---

■~o

xxx--x-x^^xxx-^-x

ffi

■ xxxx -x x xx XX x ■

^

;;;;;;;;;;;;--XX-X--

3

XXX X

00

73

•o^

;;XX----X-X^XX

t>

"=-

3

X ■ ■ -xxx -x ■ ■x -xx ■ ■xxx -xx

2

•X--X--X

?

■■■XXXXXX-^XX--X-

o

>-)

3

■X-XXX X-XX--

IM

•xxxxxx-^xx

o

; ; X X

3

X X

1

^

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 49'

the blue, as well as the number of correct choices of the
three (35 blue to 7 red) made it plain that they discrimin-
ated without the aid of the slit on the red side when bright-
nesses match for human dark-adapted eyes. For the
dace the record shows 41 red and 41 blue choices. These
fish had long since ceased to avoid red. If their high per-
centage of blue choices in the preceding tests had been
merely choices of the no-slit side, they should, in the series
now described, have made a high percentage of no-slit
choices, and in that case would have chosen red. Taken
alone this set of data certainly indicates lack of ability
on the part of dace to discriminate patches of light which
for man match in brightness. The results, however, show
that the slit was not the means for discrimination by these
trained fish.

In some of these check tests the slit was not present on
either side, in others it was transferred to the blue side.
In the later procedure the fish did not show by high red
choice that they were avoiding the slit and so using it as
a guide. Further evidence that they were not guided by
the slit appears in the section of this paper dealing with
equated energy values (p. 53).

The changes in the sharpness of outline of the red stimu-
lus patches may account for the tendency of the dace to
go toward it to investigate the unusual, Table I. In other
words, changes in the appearance of the stimulation patches
were sufficient to inhibit temporarily the association in
those individuals in which it was weakest. The series was
not long enough to re-establish the habitual response.

E. Tests with changed quality of light.

A further attempt was made to determine whether the
quality of the light of the stimulus patches was the means for
the discrimination shown by the fish in the red-blue series.
The red variable had been presented at all intensities
from very bright to very dull as described in section C.
In the last tests (described in section C) the changes in
intensity were rapid. It was thought that substituting
a patch of light of moderate intensity but of different quality
for this red variable might show the significance of wave-
length for fish.

• 50 CORA D. REEVES

Therefore the red filter was removed and in its place
was put the photographic negative already referred to.
By means of this negative there was obtained a dull patch
of mixed (gray) light which was then matched to the blue
with the aid of a flicker photometer by varying the width
of the slit. The two patches were thus made alike in bright-
ness for the human eye dark-adapted. The dace H and
Md and three sunfish were presented, each separately,
with the matched patches, and were fed before the blue
plate. The behavior of the fish was so significant that
this discussion might be placed in the division of this paper
on Unlearned Responses but is placed here to preserve the
chronological order so that the previous experience of the
fish may be clearly in mind.«

The results are shown in Table II. Large Sunfish, Small
Sunfish, and 552 chose the blue plate seventeen times and
the gray plate but three times. The record for dace H
shows 76 per cent of blue choices, not a high percentage.
Of the 43 choices made by the five fish used 72 per cent
were to be blue. These records are similar to those in which
matched blue and red were used. As in that series the
percentage of blue choices increased when the red was
held at matched intensity, so here a longer series might
have shown increase of percentage of blue choices. How-
ever, the number of tests made was relatively few. Fur-
ther work will doubtless show what significance the per-
centage of choices may have as evidence that blue and
gray are different for fish. That red was different from
gray was evident from the behavior of the fish, for both
delayed response and peculiar behavior occurred.

a. Delayed Response. — On the first trial each sunfish
swam rather quickly up to the blue, but for the succeeding
trials the time of response was either greatly increased or
there was no response, and the fish remained lurking in

* Chronological Summary for //, Md, and Large Sunfish. 1914. Oct. 13 Intro-
duced into Experiment Aquarium. 1. Blue-red Experiment. Oct. 13-Jan. 1, Red
Maximum (A few early tests with decreased red). 1915. Jan. 1-March 20, Red De-
creasing. 2. Check tests. Mar. 20-Apr. 2, With solutions. Ap.-il 3-5. Blue-Gray
Experiment. April 8-10, Equated Energy.

The data of introduction of Yi, YP, also Sma". Sunfish and S.S- was the last of
November 1914. Their dates differ in that the Blue-red series is shorter but in the
following experiments the dates are the same.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH

51

CQ

<

ro

(1)

Q)

;^

n

O

Uj

0)

<-U

a

UJ

Ph

T-!

^

t/2

a;

Uh

:3

'-I— 1

03

:3

hfl

?1

0)

rn

^ ^^ -
^ "-■ t-i

+-> CAl rG

oj a; ^

r^ );i V)

o (S rn

O '^
CO

CD

Oh^

o

s

^ 9

<D tn ^H

^^ O O

8 XI 3 ^

CD O -^

a

CD
3

CD

1 X

-

3

1 X X X X X

OO

X XX

he

foCLot>Loai^c Soo

■-HMtO ^

re

0

c

"a
E

CM M

1 ,

^0 0

r

3 3
O 0

CO

X o X 0 ><
C C

Time
in sec.

60
480
660

30
360

n

(M X

-

^ O

3

O

U3

X X X X 0 X

c

Cm re —1 ■-< 'H

c

K

X X

^

X X

X X X X X X

X XXXXX X XX

a; (J

He

LOOOt-~OOtDLO-S-CD!r!<£lO t^;Dt^0505C

1

X XCM X X X

in

o

3

3

0

-

O X

c

1) <J
ES
H.E

OOOWOlDt-

C^CO W i-H 1-1

1

!

52 CORA D. REEVES

the discrimination compartment (2 fig. 1). They remained
for perhaps six to ten minutes quite motionless. Table II
in the column marked " time in seconds " shows these
delayed responses. " Not out of 2," in the table indicates
no response.

In the case of Small Sunfish the usual time for response
in the blue-red series was eight to twenty seconds. In the
series described on page 38ff, I have records of 30 to 50
seconds, but none of 480 seconds and none of failure to
leave the discrimination compartment.

Before testing Md with this blue-gray I had tested her
three times with red-gray, twice with this same gray against
very bright red, and once with the gray against a duller
red. She had gone each time to the gray and been fed.
The time for these successive responses was 20, 10, and 6
seconds. Her delay (70 and 230 seconds and refusal to
respond) when gray was substituted for the red, convinced
me that for this fish red difi^ered from a matched gray.
The fact that this fish had just been fed before the gray
may explain the continued gray response shown in Table
II. A somewhat lengthened time for response often oc-
curred in the blue-red series. This was more frequent on
the third or fourth test for the day than on the first, appar-
ently because hunger prevented delay in the first response.
By delay in response' these fish showed that the change
from light of longer wave-length to mixed light had been
effective in stimulating them. The light was evidently
the means of discrimination, and quality rather than in-
tensity was the effective stimulus.

b. Peculiar Behaiior. — When the blue-gray patches were
exposed to the three sunfish, each advanced toward them,
halted, retreated, and acted now as Large Sunfish had
acted when the blue and red were first presented. The
behavior of Md when the blue-gray patches were first
presented was like her behavior when the bright red and
blue were first presented, some six months earlier. She
now, as then, swam back and forth across the discrimina-
tion compartment before the door and finally up toward
the gray patch where she had just fed.

To summarize, the sunfish by peculiar behavior and

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 53

delayed response clearly showed that the change from red
to gray light of presumably equal brightness was effective
in modifying behavior. The dace Md showed the same
modifications of behavior, but the dace H did not. Just
before these tests were made, this fish had been presented
once with a dull red and gray. After going up to the gray,
he shied off from it as if frightened. It was plain that the
gray stimulated. The lack of uniformity in the response
made by the different individuals at any one time and
shown by these fish is characteristic of fish.

F. Tests with Energy Equated. — -It has been held (Laur-
ens 1911) that animals are able to differentiate between
various light stimuli, not by difference in luminosity or
subjective value as measured by the eye, but by difference
in radiant energy, i. e., by mechanical value as measured
by physical instruments. The possibility, therefore, that
the radiant energy differences of the stimuli might account
for the discrimination already described, must be tested.
If the radiant energy reaching the two stimulus plates
were made equal, and the fish, especially the sunfish, could
still discriminate, then the possibility that they discriminate
by radiant energy would be eliminated. In order to equate
the energies on the stimulus plates a bismuth-silver thermopile
was used (Coblentz 191vS). It was placed in the center of
a heavy glass container, fitted into a cardboard support.
This made it possible to place the thermopile beneath
the lenses above the edges of the stimulus plates.
The thermopile was connected with a sensitive galvano-
meter. On one side were placed a water-cell, a slit, and
a new blue filter; on the other was the red filter and a solu-
tion of potassium sulfo-cyanide and ferric chloride. The
energy passing through the red filter was altered by ad-
justing the depth of the solution above it until the gal-
vanometer deflections were the same when the hi-ag ther-
mopile was shifted from one side to the other. The energy
of the two plates was also equated by means of a slit on
the red side, while a water cell and slit were used with the
blue. The relative brightness of the energy-equated plates
varied for the human eye according to the method used.
When the red solution was used, the red plate was some-

54

CORA D. REEVES

what the duller at equated energy, but with the slit on the
red side, and water cell and slit on the blue side it was

TABLE III

Table showing percentages of red and blue choices for four
dace and three sunfish when the radiant energy of the two
stimulus patches was equated.

Dace

Si'N'FISH

YP

YL

Md

H

Large

Small

SS2

Totals

6

3

33

2

8

80

5
13
72

0

10

100

7
12
63

0

15

100

2
17

84

22

Blue choices ^

78
78

possible to so adjust them that the red was brighter or
duller than the blue or matched it at equated energy values.

When the energies of the two plates had been equated,
the fish were tested by feeding before the blue plate as in
the red-blue series. The results are shown in Table III.

The record of Large Sunfish is difficult to interpret.
For the first six tests this fish went to the blue. Then
while keeping the energies equated, I increased the bright-
ness of the red; the six following trials were to the red, then
one to the blue, one to the red, and again five to the blue.
The conditions were duplicated some days later, and when
this fish was retested, discrimination was accurate. In
the record of Y P of nine trials, six choices were red and
three were blue. The curve for this fish (fig. 13) shows
that there were abrupt changes in the slit widths during
its series of red-blue trials, and that in the region of matched
brightness no trials were made. The fish apparently had
not retained the blue-food association. The remaining
fish show percentages of correct choices ranging from 72
to 100.

It appears that when the energy is equated, the fish can
still discriminate, and that consequently relative energies
do not condition the discrimination. (See postscript.)

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 55

During these tests the intensity of the blue Hght was
changed so that there was added evidence that the choice
of blue by the small sunfish (p. 42) was not merely a re-
sponse to light of a given constant intensity. But the
disturbances in the behavior of Large Sunfish and YP
did not occur when the blue intensity was changed, hence
cannot be attributed to that cause. Since the slit was
used in a part of these trials on the blue side only and in
a part of them on both sides, the series affords additional
evidence that the position of the slit did not guide the fish
in their choice (see p. 46).

G. Results of Experimental Work with the use of Food-
Association.

In the foregoing experiments with stimulus patches of
different wave-length it is believed that the responses of
the fish were made to the light on the stimulus plates and
were not dependent on any other part of the apparatus
nor on its manipulation. The radiant energy of the plates,
as measured by physical instruments does not appear to
have been a controlling factor. The fish were therefore
responding either to the wave-length difference between
the plates or to a brightness difference. At one point in
the series the colored plates were matched in brightness
for the human dark-adapted eye and the behavior of the
fish indicated that they were matched in brightness for
them. Since it is possible that fish have a capacity to dis-
criminate very small intensity differences between stimulus
patches of like quality, a capacity even greater than that
of man, it became necessary to test them with respect to
this. When thus tested the dace failed, under the condi-
tions of the experiments, to learn to discriminate intensi-
ties of white light which differed as 1 :4 while sunfish did not
discriminate intensity differences of 1:2. As it is certain
that the intensity difference between the colored patches
discriminated by the fish are far less than these, it follows
that this discrimination is based on a difference in the
wave-lengths of light of the two plates.

At the critical value of the red variable (slit width of
about 0.9 mm. in most of the graphs) theVe is temporary

56 CORA D. REEVES

failure to discriminate with return of discrimination as
the tests are continued at this value. This temporary
lapse of discrimination and its subsequent recovery are
held to be evidence of matched brightness of the stimulus
patches.

The evidence for discrimination of the stimulus patches
by a quality difference between them is quite uniform and
is most readily appreciated in the graphs figs. 13 and 15.
A fuller discussion of the whole question is reserved for
the final section of this paper — but attention is here called
to the case of failure to discriminate at matched brightness
that appeared in the training experiments. In this case
the slit was removed from above the red filter and was
replaced by solutions which reduced the intensity of the
red to that of the blue. The result was a change in the
distribution of the light on the red plate. The dace failed
to discriminate. (Table I.) Two explanations of this
failure are suggested. Either with matched brightness of
red and blue discrimination must be maintained by con-
stant use when the red has become familiar, and this had
not been done, or the fish in this case were tending to go to
the modified red plate, changed enough in appearance by
the solutions to awaken their " curiosity." As Bauer
(1910) has shown, they often go toward an unusual object.
Of course, there are those who will be apt to call attention
to the fact that here a psychic factor is resorted to in order
to account for what will appear to them to be proof of
inability to distinguish differences in color as shown in
response of a particular fish. It appears, however, that a
psychic factor is continuously resorted to in this problem;
namely, brightness discrimination. The change in the
appearance of the red plate by the use of solutions modified
response, in that after the first responses another psychic
factor dominated behavior. Bi;t Table I shows that all
first responses, even with a modified red plate, except that
of 552, were to the blue. After a bite or two, hunger did
not inhibit the tendency of dace YP and Yl to go fre-
quently to the slightly modified red plate. In view of the
uniformly positive results obtained in other experiments,
little importance is attached to this single failure and it

LIGHT CF EIFFERENT WAVE-LENGTHS BY FISH 57

is not doubted that a longer series of tests would have re-
established the response to blue

Besides the percentage of correct choices in the two
series of training experiments, there is the incidental evi-
dence in delayed response and in peculiar behavior shown
by the fish when gray was substituted for the red to which
they were accustomed. But in the next section further
evidence of innate differential response to light of different
wave-length will be presented, which gains weight from the
results obtained by the training method.

H. Unlearned Responses of Fish.

a. Modified breathing rate of fish subjected to light of long
wave-length. — That the longer wave-lengths of light have
a peculiar stimulating effect on fish has been mentioned
on p. 2. Parker (1903) found that he could use the respira-
tion rate as a means of detecting effective sound stimuli
for fish and I have twice noted a change in breathing rate
following exposure to light of long wave-length.

(1) A mud minnow, Umbra limi, had grown quite tame
during a year's careful feeding. The fish was somewhat
used to a tungsten lamp above the aquarium. One morn-
ing the breathing rate was counted with daylight illumina-
tion, before the lamp was turned on, and after the
lamp was turned on the rate was again counted. There
was no change. A piece of ruby glass was placed under
this lamp and it was then left for some days with the light
turned off so that the fish might become used to the ruby
glass. On the morning of the test the breathing rate was
normal, about 30 per minute. I turned on the red light
while experimenting with another fish. When I looked to
observe the Umbra after the red light was turned oft', the
fish had changed position. It had settled closer to the
bottom of the aquarium and was trembling and panting
in such a peculiar fashion that I forgot to count the breath-
ing. It seemed that it was more than twice as fast as
when counted just before the red had been thrown on it.

(2) A black lined box with a 16 candle power incan-
descent carbon lamp inside it was placed with an open end
against an aquarium containing a number of shiners (No-

58

CORA D. REEVES

tropis cornutMs), 7-10 cm. in length. The apparatus was
in a room illuminated by daylight.

Counts of respiration rate in diffused daylight gave
about 60 per minute. The respiration rate of a quiescent
fish was then taken while the artificial light in the box
was shining upon it, and gave 85 per minute. A ruby
glass was then placed in front of the carbon lamp. The
effect of this was to change the quality of the light to which
the fish was exposed but to lessen its intensity. The respira-
tion rate in this less intense illumination by red light was
found to be increased to 150.

b. Other characteristic behavior of fish in the presence of
stimulus patches of unfamiliar light of long wave-length.

(1) Sunfish. In describing the red-blue training ex-
periments I have already referred to the peculiar avoidance
of red shown by sunfish unaccustomed to it. During the
first weeks of the experiments in which the blue and red

Fig. 17

Path of Large Sunfish in the experiment aquarium to show the effect of red light
on its movements. The red illuminated region is bounded above by the straight
broken line. For explanation see text.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 59

stimulus patches were presented to Large Sunfish there
were many records of peculiar behavior. One such record
from my notes follows (see fig. 17):

"As the slide-door went up, the sunfish made straight for the opening, went about
one-third the distance to the stimulus patches; halted six seconds (position I),
then went up toward the blue (position 2) . Food was released and fell close in front
of its mouth. The fish did not snap at it. It was a piece of angle worm and nearly
white against the black floor. Then followed a movement of retreat to the further
end of the stimulus compartment in front of the door^. The same fish then returned
nearly to its former position (2) in front of the blue. I let a piece of worm fall
through the water, but it was not touched. The fish circled around to the red side
(position 3). There followed a swift dart toward position 4 indicated by the dotted
line in figure 17. This was not like the usual slow swimming up toward the stimulus
patches, nor like the retreat around the outer edge of the retention compartment
which followed it and is indicated in the figure (positions 5, 6, and 7). During all
these movements except the swift dart from 3 to 4 and in all positions except 4 the
head was toward the red plates."

Avoiding reactions toward red were observed with all
widths of slit, even those so narrow that the light trans-
mitted was very weak. They were not so vigorous with
weak red light as the reaction just described, but they
showed, at least, that intensity was not the only factor
in their conditioning. Besides this peculiar response to
red there was noted an equally interesting mode of be-
havior of the sunfish toward blue. Large Sunfish, even
in the same tests in which it was giving this vigorous avoid-
ance reaction toward red, repeatedly came close to the
margin of the blue patch of light and remained there, some-
times with the head just above the margin of the stimulus
plate. Often at such times the body of the sunfish was
turned slightly on one side so that the sagittal plane formed
an angle of about 30° with the vertical. My notes fre-
quently record, " Cocked its eye at the blue." This was
frequent behavior for the other sunfish also.

(2) Minnows. — Peculiar behavior toward red was less
marked in minnows than in sunfish but was occasionally
quite striking.

Pimephalcs. — Some nearly full-grown minnows which
had been one week in a stock aquarium, and were not yet
tame, were tested as to first responses toward blue-red
and also toward blue-gray patches which were matched
in brightness. Seven Pimephales notatus were tested. Each
fish was presented with the matched blue-gray patches

60 CORA D. REEVES

without being fed. There were 28 approaches to the
patches, 14 of them to the blue. The two patches were
approached by all the fish in the same way and with equal
frequency; the time of approach varied from three seconds
to three and one-half minutes. When identical matched
patches were presented immediately afterward except
that one was now red, the other blue, the behavior changed.
Two of the fish, after patching for a time at the doorway,
slipped out of the discrimination compartment, went an
inch or two into the stimulus compartment, and then
returned by backing out of the stimulus compartments.
They repeated this manoeuver. One did not leave the
discrimination compartment for seven minutes; another
still lurked in the discrimination compartment after fifteen
minutes.

Semotilus. — A horned dace, an untrained fish from the
same stock as the Pimephales, was placed in the experiment
aquarium one noon. The ceiling light was turned on. At
3:15 that afternoon the sliding door was raised and matched
blue and red plates were exposed. For a time the fish was
quiet, then it went slowly a short distance into the stimulus
compartment, then lay in the doorway across the shadow
of partition .4. It then swam back into the discrimination
compartment and circled around the back of the compart-
ment facing the lights. After 295 seconds the fish went
up to the blue plate. It was closed back into the discrim-
ination compartment by use of partition ^4, and after fifteen
minutes the sliding door was again raised to expose the
matched blue-red patches. Five times the fish upon com-
ing to the door-way turned with its head away from the
stimulus patches so that the posterior half of the body
projected out of the door into the stimulus compartment,
then swung its head across the doorway and moved back
into the discrimination compartment. It then circled across
the end of the aquarium and after 340 seconds swam to
the blue plate.

This dace was tested with matched blue and gray patches,
two hours later. Twenty seconds after the slide door was
raised the fish was at the gray plate, and 30 seconds was
the time for its second trial. Three hours later the blue

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 61

and red were again used. When the door was raised the
fish stood in the doorway three minutes, then went for-
ward an inch into the stimulus compartment and backed
away. It next turned straight toward the red plate and
darted forward to within 20 cm. of it, then drew back
and darted up to the blue stimulus patch. It is to be noted
that the waiting in the discrimination compartment of
the untrained minnows and their circling across the door-
way in the presence of the red-blue patches are exactly
the sort of behavior that the dace, habituated to blue-red
stimulus patches, gave when the blue-gray was presented
to them. When the matched blue and gray were presented
to the untrained minnows, all the fish tested swam promptly
to these plates. They went an equal number of times to
each plate. They often swam from one plate to the other.
I could detect no differential behavior toward either, but
these same untrained minnows, when presented with the
red and blue, often remained at a distance from the red
or turned from it or swam vigorously toward it. In other
words, the introduction of red in place of gray brought
about a differential response in those untrained fish. This
took place when no part of the apparatus was modified in
appearance except the stimulus patch. Without the evi-
dence of the long blue-red training series the tendency of
fish to change their behavior when the red plate is first
exposed might be attributed to brightness differences,
although the plates matched in brightness for human eyes.
But in the red-blue food association series with the variable
red intensity, the least discrimination was shown by ex-
perienced fish when the red and blue first approached
matched brightness for the human eye. Therefore, if
untrained fish showed a differential response when the
red plate was first exposed at this matched brightness,
this could scarcely be due to its brightness.

When untrained fish were presented with blue-gray
patches that matched for the human dark-adapted eye,
they showed no preference for either plate and gave no
indication that the two patches were in any way different
for them. Therefore, it seems clear that a difference in
brightness was not the reason for the peculiar behavior

62 CORA D. REEVES

noted when the fish used to blue-red patches were shown
patches matched in brightness but with the gray substi-
tuted for the red. The evidence that red differs from gray
is emphasized by the differential behavior of untrained
fish. It is equally emphasized by similar behavior of the
fish which were modified by months of experience, as
recorded in the blue-red series. They showed toward
gray the same differential response which wild fish showed
toward red.

C. Dark-adaptation and sensitivity to light. — To test this
question I made tests from time to time to determine
whether there was a change in the relative brightness of
colors when the fish were in strong and weak light. These
tests were made during the red-blue training experiments.
They were never long continued but merely interpolated
in the red-blue series, when a point had been reached in
the adjustment of the red at which its brightness for the
fish was judged to be equal to that 'of the blue. The fish
were then tested at night after two hours dark-adapta-
tion. This dark-adaptation did not greatly change the
accuracy of the response. There were no marked changes
in behavior even when the two stimulus patches were of
the intensities to which the fish had responded by similar
behavior before the food association was established. This
seemed to show either that there was no shift in relative
brightness, or that the response was to color rather than
to light intensity. It was too dark for the fish to secure
the food as it fell. Only when it fell through the water
between the fish and the colored plate was it ever seized.
As there was danger under these conditions of breaking
up the reaction of the fish to the positive stimulus plate,
the experiments were not carried further with the trained
fish. The work was continued with red and white stimulus
plates and with untrained dace, sunfish, and blunt nosed
minnows (Pimephales). While these were light-adapted,
they were presented with the stimulus patches until red
avoidance disappeared, and the intensity of the plates
was then adjusted until the fish went to the two with equal
frequency and in the same manner. The fish were then
allowed to become dark-adapted. No marked change

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 63

could be seen in the method or in the frequency of approach
to either of the two plates.

Although I obtained no evidence of a shift in relative
brightness of the red and blue used, or in that of red and
white, there was evidence with both dace and sunfish of
an increased sensitiveness to light after dark-adaptation.
If the intensities of the red and blue plates were so ad-
justed that fish adapted to the illumination from a tungsten
lamp went without hesitation to either plate and visited
the two plates with equal frequency, and if the fish were
then allowed to become dark-adapted and the plates again
presented, they no longer went close to the plates but
halted at some distance from them. Their sensitivity to
light had increased, but my evidence does not show that
the relative brightness of the plates had changed for them.

III. REVIEW OF THE LITERATURE

It is not uninteresting to note that the early workers
in this field used a variety of methods which involved both
(1) training of the fish and (2) their unlearned responses.
But the discordant results obtained are not correlated with
the type of method employed.

Graher (1885) was one of the first to test the color vision
of fish by the method of response. He illuminated dif-
ferent parts of the aquarium simultaneously with lights
of different sorts, and he noted in which of them the fish
gathered. He found a difference in the behavior of Gaster-
osteus spinachia toward darkness as compared with their
behavior toward light, and a difference in their behavior
toward red as compared with their behavior toward blue.

Zolotnitsky (1901) fed fish with red Chironomus larvae
until they were accustomed to this food. He then put
against the glass aquarium wall a cardboard, upon which
were pasted bits of wool of various colors and of the size and
shape of the larvae. The fish jumped most frequently
for the red pieces. He decided that fish could distinguish
red from other colors. He did not consider the brightness
values of the colors used.

Washburn and Bentley (1906) attacked the problem by
a method which involved more exact discrimination by

64 COKA D. REEVES

the fish. They fed a single horned dace, Semotilus atro-
maculatus, from forceps, two pairs of which were fastened
to a wooden bar so that they could be lowered into the
water simultaneously. One pair of the forceps had small
red sticks fastened to the legs with rubber bands, while
the other was equipped with green sticks. Food was placed
in the red forceps to form an association red-food. When
the association was formed, both pairs of forceps were
presented empty; the fish selected the red, and continued
to do so even when a much lighter red was used.

The conclusion reached by Washburn and Bentley is
that the fish can associate food with pigment color and
discriminate between red and green even when the shade
of the red is changed.

Reighard (1908) studied color vision in the gray
snapper. The primary purpose of his work was the in-
vestigation of warning coloration, not of color vision. If
the fish upon which he worked should be shown to be color
blind, his conclusions that the coloration of coral reef
fishes does not serve as warning, would be strengthened.
He describes a school of gray snappers (Lutianus griseus)
which were fed in the open sea with blue and red bait.
The fish showed an instinctive avoidance of red. When
they had become accustomed to the feeding, the red bait
was presented together with the blue. The fish avoided
the red. They also gave preference to white when offered
with blue. The brightness of the blue, red, and white
were measured by the color wheel and found to be in the
order white, red, blue. Professor Reighard concludes that
since the snappers preferred the white to the blue, and the
blue to the red, they chose in the first case the brighter
bait and in the second the duller one of a pair. They were,
therefore, guided not by the brightness of the bait but by
the color. The quickness and accuracy of the fish in form-
ing associations show, in this experiment and in others, the
advantage of working with fish in a natural environment.

Since Zolotnitsky did not determine the relative bright-
nesses used by him for either the fish eye or the human
eye, it is clear that these may have been such that the fish
were able to select the red yarn by its brightness alone.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 65

Washburn and Bentley used a brighter red in non-food
tests to detect a possible brightness discrimination by the
fish.

Reighard varied the brightness of one of the two colors
so that one was used both lighter and darker than the other
color. Following the practice current at the time and
finding that the fish discriminate between two similar
objects when one is used both brighter and less bright than
the other for the human eye, he infers the presence of color
vision. The brightness values of the color used should
have been determined with dark-adapted eye but since
in no case was the attempt made to obtain matched
brightness, but rather to insure a considerable difference
in brightness, this criticism has little force. The conclusions
of the writers so far considered are based on the assumption
that the brightness values of the colors used are approxi-
mately the same for the fish eye, as for the human eye,
light-adapted. For this assumption there was no evidence.
Had they conducted a long series of experiments in which
one of the two colors was changed in intensity by small
degrees, they might have found a region in which the two
colors matched for the fish eye. This region might have
made itself known by temporary failure of the fish to dis-
criminate. A recovery of ability to discriminate in this
region of matched brightness would have been evidence of
color vision. A persistent failure to discriminate at these
matched values would have been evidence of color-blindness.

Hess (1909) held the proofs of color vision so far offered
to be inconclusive. When he illuminated an aquarium
placed in a dark room, half with blue light and half with
red by the use of colored glasses, young (15 cm.) Atherina
hepsetus gathered in the blue, even when it appeared to
his light-adapted eye somewhat darker than the red. When
the red was made sufficiently bright, the fish gathered in
the red instead of in the blue. By properly adjusting the
relative brightness of the blue and the red he succeeded
in getting the fish distributed evenly in both halves of the
aquarium. These results indicated to him that the fish
were responding to intensity differences rather than to
wave-lengths. He observed that these young fish swam

66 CORA D. REEVES

toward the light and in a dark room gave this response
to light of very low intensity. They went to the brighter
of two white plates which differed in illumination as 1 to
1.23. Taking advantage of this phototactic behavior, he
found that fish gathered in the yellow-green of a spectral
band thrown into the aquarium. By using a piece of
cardboard he slowly cut off the successive portions of the
spectral band from the violet end until only red was left.
The fish gathered in whatever part of the spectrum re-
mained except in the extreme red. If the spectrum was
shortened from the red end, they continued to gather in
it until they were finally in the violet. Since the fish did
not gather in the red of longest wave-length, Hess decided
that it stimulated them very little. For them the red end
of the spectrum appeared to be shortened as it is for the
totally color blind.

By means of their brightness response Hess determined
the relative brightness of different parts of the spectrum
for fish. To do this he used monochromatic light on one
side of the aquarium and a white light of variable intensity
on the other, and found intensities of the variable at which
the fish distributed themselves equally on both sides. He
decided that for fish the yellow-green was the brightest
part of the spectrum, that blue, yellow, and orange were
less bright; that red was the darkest; and that for fish
orange and red have little brightness value. These deter-
minations of brightness values of the spectrum were so
like those already found by others for the human color-
blind that Hess was led to conclude that fish are color-
blind.

The first work of Hess seems to be open to criticism on
several grounds:

(1) He used very young fish. The condition of the
retinal cones of young fish has been shown by Shafer (1900)
to be quite difTerent from that of adults. It is not impos-
sible that their color vision is different. Whether this is
the case must always be difficult to determine, for such
young fish cannot well be used in training experiments
because their behavior is too mechanical. They feed on
minute plankton animals and have perhaps little need to

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH G7

exercise discrimination in the choice of their food. Gold-
smith (1914) has shown that when the fish are very small
they do not chase their prey. The need for discrimination
grows as the fish develop and change their food habits.

(2) The experiment in which the fish are found to show
no differential response toward red and blue when the
relative brightness of the two is suitably adjusted, is taken
as evidence for the lack of wave-length discrimination.
The fish are here not responding to a difference in wave-
lengths of light, but it does not follow that they are unable
to respond to such differences. The experiment shows only
that under the conditions they do not respond differentially
to the wave-length differences. Training experiments
might show a capacity not revealed by this method. It
is further possible that during the time required for ad-
justment of the blue and red lights the fish became accus-
tomed to the red so that they no longer avoided it. When
the adjustment was completed they might then distribute
themselves equally in lights of the two colors. My own
experiments contain much evidence of such rapid " getting
used " to red. In such work on response to light of different
wave-length it is the initial responses that have greatest
value.

(3) Hess' experiments show that the fish do not gather
in the red of a spectral band, and this is taken to mean
that the red has little stimulating power. The experiment
could, however, be explained by supposing that the red
affords a stronger stimulus than other rays and is avoided.
My own observations as well as those of Reighard, Bauer,
Mast, and Goldsmith show an avoidance of red.

(4) The experiments in which Hess determines the rela-
tive brightness of different parts of the spectrum for these
young fish in the dark is of interest but seems to me to
have no bearing on the question of wave-length discrim-
ination. That the brightness values of the different colors
are the same for fish as for the color-blind may be thus
established. Hess' work with the pupilloscope is confirma-
tory (Hess 1917). My own work adds to this evidence.
It does not necessarily follow that fish are color blind. Von
Kries (1896) shows that for the periphery of the human

68 CORA D. REEVES

retina the brightness values of different regions of the gas-
spectrum are the same as for the central area of the retina
■of the normal eye in daylight. It should then follow, if
Hess' mode of reasoning is sound, that the periphery has
color vision, but it is well known that it is color-blind. But
if the periphery of the human retina has the same bright-
ness values of the spectrum as the normal eye in daylight
but is color blind, may not the fish eye have the same
brightness values as the human color-blind and yet the fish
distinguish colors?' With more knowledge of the color
discrimination possible in lower vertebrates, theories of
human color vision may be enriched and improved.

Bauer (1910), noting that the results of Hess were at
variance .with those of other writers decided that the fact
that the work of Hess had been done with fish in the dark
rather than in the light might account for some of these
differences. Accordingly he placed above the aquarium
in which he kept his fish a light, which furnished a constant
moderate light intensity for those experiments in which the
fish were light-adapted. The stimulus light reached the
narrow, black-lined glass aquarium through a clear rec-
tangular area in one of its ends. When two colored glasses,
used as filters, were placed so as to divide this opening
through the center, the halves of the aquarium were illumin-
ated by light of two colors. When different thicknesses
of white paper or ground glass were before the halves of
this opening, the two parts of the aquarium were illumin-
ated by white light of different intensities. When white
light was used, Bauer was not able to cause a gathering
of the fish in either half by differences in the. intensity of
the two sides. But with red and blue filters in place certain
light-adapted fish tended to gather in the blue half. With
the Charax and Atherina tested, he found, in contradiction
to Hess, no brightnesses of these two colored lights at which
the fish were equally distributed in both halves. When
spectral bands were substituted for the filters, the light-
adapted fish went into the blue green or yellow area. When
the orange or red (between 620 /^/^ and 630 f^f^) were pre-
sented to the fish, they rushed to the unilluminated end
of the aquarium as though frightened. When tested with

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 69

the white Hght, no such behavior was observed, and with
different intensities of white light no avoidance of a definite
moderate brightness appeared. Bauer concluded that
when light-adapted, the fish have color vision, a conclusion
reached both from his observations that the fish avoided
red, and from the fact that he was unable so to equate the
brightnesses of red and blue that the light-adapted fish
responded to them similarly. When the same fish were
dark-adapted, they entered the red and orange lighted
portions which they had avoided when light-adapted. He
regarded this as evidence that when dark-adapted, they
might be color blind. He presented evidence for a " Pur-
kinje phenomenon."

Hess (1911), stimulated by the work of Bauer, tested
the relative sensitivity of fish to light when dark-adapted
and when light-adapted. He used carp between 8 and 9
mm. long. He put the fish in their containers in the sun
to get them light-adapted and then rapidly moved the
container into the dark room, in which the tests of both
light- and dark-adapted fish were made. He noted the
strength of the stimulus necessary to call forth a photo-
tactic response in light-adapted fish and compared that
with the strength of stimulus for fish which had been for
some time undisturbed in the dark. He found their sen-
sitivity increased more than 1,000 times by 15 minutes
dark-adaptation. By varying the intensities of the light
used he, like Bauer, was able to make dark-adapted fish
distribute themselves uniformly in the red and blue halves
of an aquarium illuminated by filtered light. When these
fish were light-adapted, it was necessary to increase the
intensity of the blue illumination six to eight fold, in order
again to secure uniformity of distribution. The relative
brightness of red and blue for the fish had been changed
by light adaptation, but whether light- or dark-adapted,
they showed no preference for red or blue.

One factor seems to have affected unfavorably the de-
terminations of Hess concerning the relative sensitivity
of fish after dark-adaptation. It appears that for certain,
experiments with light-adapted fish the aquarium contain-
ing the fish was removed from a lighted window to a dark

70 CORA D. REEVES

room. For experiments with dark-adaptation the aquarium
was not shifted. I have found fish more sensitive to Hght
after dark-adaptation although I have not worked with
such young fish as Hess used. It is certainly to be expected
that fish, whether dark-adapted or light-adapted, would
more readily respond to light after being quiet for 15 min-
utes as in the case with Hess' dark-adapted fish than after
the earthquake-like experience of being moved about a
room, as were his light-adapted fish. It therefore seems
to me that the relative lack of sensitiveness of Hess' light-
adapted fish may be due quite as much to inhibition of
response through mechanical disturbance as to light-
adaptation. Certainly dependable comparative results are
to be expected only when the light- and dark-adapted fish
are handled in the same way.

Hess repeated some of the experiments of Zolotnitsky.
He fed light-adapted fish upon red Chironomus larvae. He
then offered them at the same time bits of red and gray
yarn that for his dark-adapted eye matched the larvae
in brightness, and he found that the fish snapped with equal
frequency at both the red and the gray yarn. When the
imitation larvae matched the background in brightness
but not in color, the same fish passed them without trying
to eat them, quite regardless of their color. Again areas
of red light of the size and shape of the red larvae were
projected against a green-lighted area and were snapped
at by the fish when they were darker or lighter than the
background, but not when they matched it in brightness
for his dark-adapted eye. These tests seemed to show that
fish are color blind.

Bauer shows that after dark-adaptation fish go into a
red which they avoided when light-adapted when this
red is presented along with a blue. Hess shows that a
blue which matched a given red for dark-adapted fish must
be increased in brightness to match the red for the fish
after light-adaptation. It is probable that both men are
dealing with the same fact, namely that the stimulus
afforded by the blue is increased by dark adaptation. It
is not impossible that the real brightness of the red has
remained little changed with changed adaptation. But

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 71

the blue, according to Hess' explanation, is greatly reduced
in its effect by being absorbed by the pigment which in
the light-adapted eye has drawn forward and therefore
must absorb the light before it reaches the ends of the
rods and cones. In Bauer's work the dark-adapted fish
in entering the red may have been giving a negative re-
sponse to a too brilliant blue, a blue past the optimum
because of the retractions of the retinal pigment. The
red, however, had not so increased in brightness and was
entered. The dark-adaptation had increased sensitivity
to brilliant blue light. The fact that fish tend to go to an
optimum light will equally well explain the behavior of
the fish used by Hess when they are recorded as going to
a red after light adaptation and entering the blue half of
the container when the blue was increased in brightness
six to eight times. This would be expected for fish where
red avoidance is absent as it is in Mugil according to Bauer.
Frisch (1912) placed fish in glass-bottomed dishes with
colored papers beneath them and found that they became
of a brightness and color like the background upon which
they rested. The stimulus for this matching is received
through the eye.-' He found that the fish changed their
brightness tone before they changed color. When he used
at the same time papers of different colors he found that
with properly selected colors the fish could be shifted from
one to another without change in their brightness tone.
Such colors were thought to be of matched brightness for
the fish. After a time these fish changed color to match
the background, thus indicating that the color change is
a response to wave-length and not to brightness. In other
experiments Frisch fed his fish from the upper end of a
glass tube lined with a certain color. This tube was placed
in an aquarium as a member of a series of tubes with linings
varying from white, through a series of grays, to black.
To make impossible a choice by position the order of the
colored tube in the series was changed from time to time.
When the fish had learned to seek food only in the colored
tube, one similar in color but empty was substituted. The
fish continued to seek food in the colored tube. As they

^ See Parker 1909, Mast 1916.

72 CORA D. REEVES

did not confuse with the color any intensity of a long series
of grays varying from white to black, he decided that
their response had been to wave-length and not to bright-
ness. He further tested their color vision by seeing whether
the fish trained to go to a given color would respond to it
positively when other colors of matched brightness were
presented. He found that blue and green were discrimin-
ated from each other and from other colors as well as from
the grays of the series, but that the fish confused yellow
and red with each other though not with grays or with
black.

Uhlenhuth (1911) after a critical and illuminating review
of the literature, accepts as conclusive Frisch's evidence
that fish react differently by color change to lights of dif-
ferent quality but of equal brightness. It does not follow
that they are aware of the color differences in their back-
grounds. Where a method of choice is used, as in my
own work, it becomes probable that the fish are stimulated
by quality and not intensity alone and that the quality
determines their choice, (cf. Mast 1916.)

Frisch (1913) fed fish upon yellow-colored meat until
the habit of approaching it was fixed; then he tested them
by substituting colored papers for the meat. For these
tests small pieces of yellow paper were pasted upon sheets
of gray of the same texture. These gray backgrounds
included one the brightness of which matched that of the
yellow paper for his dark-adapted eye. Upon each gray
background, in addition to the yellow bit of paper, was
pasted one of similar shape of a darker gray and another
of a lighter gray. When these sheets of gray were placed
against the walls of the aquarium, the fish that were used
to the yellow food swam more frequently to the yellow bits,
while fish that had not been fed with yellow meat swam
with equal frequency to all three pieces of paper. From
this Frisch maintained his previous conclusions, that for
fish yellow and red light are different from white and from
other colors.

Hess (1912) resumed his attack on the problem and
followed the method used by Frisch. He placed his fish
in a rectangular aquarium and fed them on yellow meat.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 73

which he allowed to fall through the water while a gray
back ground was placed against the outside of the aquar-
ium. He also fastened pieces of yellow meat and pieces
of colored wool of similar shape upon one side of a glass
plate which could be lowered into the water, so that the
colored bits were visible to the fish against the same gray
background but the glass surface which separated them
from these pieces was invisible. The fish went with equal
frequency to all the bits of color as the glass was sunk
into the water. Next a glass was used to cut off a small
space along one side of the aquarium. For one week yellow
food w^as allowed to fall on the side of this glass plate where
the fish might reach it. Then for two weeks, while yellow
food was given on the same side, where it could be seized,
blue food was allowed to fall at the same time on the re-
verse side of the plate, where it was visible but inaccessible
to the fish. After the three weeks of feeding of yellow
meat, with or without the inaccessible blue, bits of yellow,,
blue, gray, red, and green were dropped into the aquarium
behind the glass plate, the fish swam at them freely and
showed no preference for yellow. Again Hess insists that
fish are color blind.

In this work of Hess, in which he used training methods,
he allowed the food masses to fall through the water and
found that the fish rushed to a mass of any color, though
they had previously been fed upon yellow-colored food
only.

Hess has here adapted the well known experiment of
Mobius (1873) with pike, afterward confirmed by Triplet
(1901) with perch. Mobius found that when a pike in
an aquarium was separated from minnows by a glass plate,
the pike at first dashed at the small fish and received a
bump on the nose whenever it did so. After a time the
pike stopped rushing at the minnows, and when the glass
plate had been removed, the minnows were not at first
seized by the pike. A severe bump had become associated
with trying-to-seize-minnows. Had Hess, while feeding
his fish on food of quite other form and appearance, first
succeeded by punishment in establishing an immunity
for his blue food masses so that they were not seized when

74 CORA D. REEVES

made accessible, and had he then offered the fish, along
with the blue, masses of equal brightness but of other colors,
and found all of them refused, he might then have concluded
that the fish did not discriminate them by color. What
he did was to establish an association with yellow by feed-
ing the fish on it. He then offered this yellow along with
blue, but without showing that the blue had been made
immune. The fish showed no differential choice, and there
was no reason to expect it.

This experiment, used to prove the lack of ability on the
part of fish to discriminate on the basis of wave-length,
illustrates the difficulty of trying to prove a negative.
Fish are keen in detecting movement. There was appar-
ently no severe bumping against the glass when blue bits
were falling through the water with only a plate of glass
between them and the fish. The accessible yellow meat
dropped near the plate satisfied the hunger of the fish.
The yellow food served only to form the association " fall-
ing bits-food. ' But the inhibition against reacting to blue
was not established.

Hess (1913) argues against the view that the red color
of the belly of the breeding " Saibling " may be observed
by other " Saibling " and is therefore an indication of color
vision. As evidence that red rays do not reach the spawn-
ing grounds of the brilliantly colored Konigseesaibling he
tested the penetration of red rays into water by a simple
experiment. He used a brass tube 4 m. long. On the side
of this at one end was a window 4 cm. in diameter, and
within the tube facing the window was placed a mirror
set at an angle of 45° with the axis of the tube. Attached
to the lower end of the tube opposite the window and at a
distance of 20 cm. was a red plate 12 by 20 cm. which could
be set at various angles to the window. The end of the
tube to which the red plate was attached was immersed
in the water to a depth of 4 meters, and the reflection of
the plate was observed in the mirror through the upper
end of the tube. When thus observed, the plate appeared
yellow or brown but not red. I have repeated the experi-
ment in air in diffused daylight. When the light is suf-
ficiently reduced, the results are as Hess states. This

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 75

does not mean that the surface of the plate does not appear
red if observed directly. In Hess' experiment only a small
fraction of the light waves reaching the mirror are those
reflected from the plate. Inside of the tube, which appar-
ently was not blackened, the light waves were probably
reflected back and forth. The illumination of the mirror
is weak, and the visual angle for its surface at 4 m. distance
is small. Conditions are unfavorable for observing the
image of the red surface in the mirror, but not because red
rays are lacking. Hess recalls that many fish spawn at
20 m., that this especially brilliant variety may spawn at
60 m. and, because of the absorption of red at this great
depth, the red of their lower surfaces can have no biolog-
ical significance. There are few data as to the absorption
of light of different wave-lengths in fresh water lakes. It
has been determined, however, by Murry and Hjort (1912)
by photographic methods that in the ocean at 100 meters
there are many rays of all kinds, though fewest of the red.
From data of v. Henri et Larguier des Bancels it is deter-
mined that the eye is about 3,000 times as sensitive, as
the most rapid photographic plate. Although the red rays
are absorbed with increasing depth in shallow water, on
the other hand, spectral colors abound. This is shown by
v. Aufsess (1913) who placed a mirror beneath the water
of an Alpine lake and in it viewed the terrestrial landscape.
The branches of trees and all other external objects then
appeared fringed with spectral color, ranging from red
near the horizon to blue higher up. Most young fish live
in such water but may retire to deeper water as they grow
larger. Hess' evidence does not satisfactorily show that
no red rays reach the spawning grounds. But if none do,
that does not prove the absence of color sensitivity in fish.
If the absence of a particular stimulus in the normal en-
vironment of an animal were to be accepted as evidence
of the lack of response to it, we should be forced to hold
that Paramoecium is not affected by the electric current.
Experiment shows it to be sensitive to the electric current,
and experiment alone can decide the response of any animal
to a particular stimulus, whether that stimulus is or is
not normal to the environment. I have shown that the

76 CORA D. REEVES

color of the male probably plays no role in selection by
the female Elheostoma coeruleum (Reeves 1907), but that
displays of the male may nevertheless play a part in the
breeding activities. The color of the breeding fish need
not indicate the presence or absence of color vision in a
given species, but the absence of ability to distinguish
wave-length can not be demonstrated by assuming or by
proving that the red wave-lengths do not penetrate to the
breeding grounds of any species. (See also postscript.)

Hess (1913) showed that if the integument of young
eels is subjected to the action of violet light, a peculiar
fluorescence appears. He did not find this when he used
older fish. I found that a certain very large sunfish (not
the one whose graph is shown in fig. 15) failed to learn to
discriminate a blue and red which the smaller sunfish, SS2,
discriminated. This may have been due to a difference
in effect of light on the retinas of young and old sunfish
comparable to the difference in effect of light on the inte-
gument of young and old eels. It may also have been due
to greater readiness of the younger fish in forming a sen-
sory habit.

Before leaving the work of Hess, it may be well to char-
acterize it in more general terms. The conditions of his
first work, which was carried out in the dark, were unfavor-
able for observing small differences in behavior. His sub-
sequent failure to find color vision is in part due to the
earlier failure, in part to the fact that he worked often in
the dark, and in part to his method of handling the fish.
Small differences in behavior are difficult if not impossible
to observe under the conditions of illumination he used.
His method of handling his fish introduces too many dis-
turbing factors. Fish picked up and put into the light and
then put into the dark will show excitation or inhibition
effects because they are especially sensitive to jarring and
shifting of the aquaria. But differences between response
to red and other colors exist; and Hess found these and
misinterpreted them. His work in establishing the simi-
larity in brightness values of different spectral regions for
fish and for the totally color-blind will probably be con-
firmed.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 77

Goldsmith (1914) in a most suggestive paper gives the
results of a study of the perception and memory of colored
objects by fish. Her work shows differences in this respect
in the various species used. While Gobius avoided red,
the stickle-back came to rest over red and approached the
red objects presented.

Freytag (1914) reports work to determine whether fish
become colored like their surroundings. The fish were
usually placed for a few hours in a glass dish above a given
background. Twenty-four hours seems to be the maxi-
mum time. He finds no proof for color vision in the Phox-
inus laevis used. The fact that there is in his specimens a
tendency to match the background in brightness leaves
one in doubt as to whether with as much time for color
adaptation as was allowed by Mast for flounders these
fish would not also have shown consistent color changes.

Mast (1916) has published evidence of color vision and
of the avoidance of red by fish. In experiments on change
of color to conform to background, he used red, yellow,
green and blue paints. He had boxes painted half with
one color and half with another one of the four colors used,
so that all possible pairs of these colors were provided.
Fish (flounders, Paralichthys and Amylopsetia) were al-
lowed to become adapted for six weeks or more, to each
of these colors in boxes painted of one color. Colored
photographs show that they became colored like the back-
ground. They were then released above the dividing line
of a two-colored box in order to see toward which half
they would swim. It appears from an examination of
Mast's data that fish adapted to blue showed 88 per cent
of blue choices, those adapted to green 70 per cent of green
choices, but those kept in the red box and then released in
the two-colored boxes went to the red in only 26 per cent
of the choices. Of the 120 times when red and blue were the
pair of colors presented to the red-adapted fish, it went
but five times to the red. This record, given by Mast,
of natural differential response in avoiding red has added
value from the fact that his experiments were not per-
formed for the purpose of testing the existence of a differ-
ential response to that color.

78 CORA D. REEVES

IV. DISCUSSION
A. Analysis of Methods Employed

In the study of color vision in fish, investigators have
availed themselves of three methods — that of training
(formation of a food association), that of instinctive re-
sponse, and that of change of color of the fish with change
of background. The second and third of these methods
might, of course, be considered together since both involve
instinctive response, but it is convenient for purposes of
discussion to separate them.

a. The method of response has been used chiefly by Reig-
hard, Bauer, and Hess, and results obtained by it are in-
cluded in this paper. No attempt is made by this method
to form a food association, but the fish are presented with
light stimuli of different intensities and qualities, and
their behavior is noted. Two patches of light of restricted
wave-length, or one of restricted wave-length and one of
white light, may be presented together or a considerable
part of the spectrum or the whole of it may be shown at
one time. Where two qualities of light are used, the fish
may show by difference in frequency or method of approach
or by tendency to avoid one of them, that the lights differ
for it. The intensity of one light is then varied (Bauer
and Hess). If at any intensity of the variable, behavior
of the fish toward the two lights is found to be the same,
it is concluded that their past differential responses toward
the lights have been due to a difference in brightness; the
cessation of these responses is taken as evidence of matched
brightness. If the fish do not, under the conditions of
the experiment, show differential response at this matched
brightness, they are regarded as color-blind. The evidence,
failure to respond at matched brightness, thus leads to a
negative conclusion. It has been pointed out that responses
of this sort are evanescent. The response may be given
when the stimulus is first presented, but may not appear
on repetition of the experiment: because the fish may have
become " adapted " to the unwonted stimulus and no
longer respond (see adaptation to red in figs. 13, 15 and
discussion p. 33). Failure to respond may also result from

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 79

inhibition by other stimuli (delayed response in sunfish
p. 50). For these reasons a negative conclusion based wholly
on evidence from the method of response seems unwar-
ranted. The method may answer for lower organisms,
but animals as highly organized as fish may be quite capable
of responses that do not appear under it. Those who use
it need to take particular pains that they are not relying
upon innate responses which have disappeared through
adaptation or which are inhibited by uncontrolled stimuli.

If on the other hand, when positive results are obtained
by the response method (Reighard, Bauer), they appear to
have the greater value. Here, at least, it is certain that
response is not lacking through adaptation to the stimulus
or through inhibition. Positive results are obtained when
the fish show differential response toward light-stimuli
of different quality at all intensities of the variable used.
Owing to the evanescence of the response (adaptation)
it is not practicable to repeat the experiments at many
intensities of the variable. It is, therefore, possible that
matched brightness for the fish is neier secured by this
method. But if, at various intensities of the variable,
the fish continue to show a differential behavior toward
the two stimuli of different quality and do not show this
toward stimuli of the same quality at presumably like in-
tensities, then we have evidence of color vision. Such
evidence we get, for instance, from the peculiar behavior
of fish toward red obtained by Reighard, Bauer and myself.
Hess' failure to obtain similar differential behavior is merely
la,ck of evidence, and as already pointed out, permits of
other explanation than inability of fish to respond differ-
entially to red.

b. The method of training gets rid of the difficulty in-
volved in evanescence under the response method, for it
insures response over long periods of time and makes it
possible to use the variable at many intensities in the
effort to secure matched brightness At the same time
the fact that the fish take food during experimentation
shows that their responses are not inhibited. In work that
has been done hitherto on color vision of fish by training
methods, three procedures have been used in the attempt

80 CORA D. REEVES

to determine whether the fish were guided by brightness
differences rather than by quality of light.

In the first of these procedures (Washburn and Bentley)
no attempt is made to secure matched brightness of the
two colored objects presented to the fish as stimuli. One
of them is kept constant in brightness value and is pre-
sented to the fish in contrast to the other colored stimulus,
which is used at two values, one brighter and one duller
than the first. The relative brightness of the two stimuli is
judged by the human eye either light- or dark-adapted.
The method is probably adequate to insure the end sought,
namely, that the variable stimulus be used both distinctly
brighter and distinctly duller than the constant. When,
under these conditions, the fish learns to avoid the con-
stant stimulus in contrast to both a brighter and a duller
variable, he chooses in the one case the duller object and
in the other case the brighter, and it has been assumed
that the discrimination is, therefore, made by means of
a quality difference between the two stimuli. Since con-
siderable brightness differences for the human eye always
exist under this procedure, it is possible that they exist for
the fish eye also, and it follows that the fish may discrim-
inate the constant stimulus from the variable by brightness
difference only. A totally color blind person might so dis-
criminate them by avoiding the stimulus of definite con-
stant brightness, with preference for one either brighter
or duller.

The second training procedure referred to above attempts
to make the two stimuli of equal brightness for the fish.
To this end a colored stimulus object is matched with
another colored object or with a gray by means of the
human dark-adapted eye. The two objects may be iden-
tical except for color and may be offered to fish as food,
or one may be offered as food and the other form a back-
ground against which the food is seen by the fish. If the
fish learns to distinguish the two objects assumed to be
matched in brightness for it, the conclusion is drawn that
quality differences in the light are the basis of discrimina-
tion. However probable it may be that colors matched
in brightness for the human dark-adapted eye are also

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 81

matched for the fish, the method under discussion affords
no proof that this is so. If it is not so, the fish may dis-
tinguish the two stimulus objects because of brightness
difference, although to the human eye they appear equally
bright.

The third training procedure hitherto employed attempts
to meet the difficulty of the first two by varying the in-
tensity of one of the two stimuli while the other is held at
constant intensity. It assumes that some intensity of the
variable will be used at which its brightness equals, for
the fish, that of the constant. If at this intensity the fish
continue to discriminate, they must do so by means of
quality differences, and they, therefore, have color vision;
if they fail to discriminate, they are color-blind. Frisch
used this method when he offered his fish food concealed
in a colored tube, which was one of a series of gray tubes.
On the assumption that the colored tube is of the same
brightness as one or more of the gray tubes, a fish that
learns to go to it and continues to ^o to it when it is empty,
must distinguish it by the quality of light coming from it,
and has, therefore, color vision. A fish that is capable
of learning, but cannot distinguish the colored tube amongst
the grays is color-blind. The danger in the employment
of this method lies in the assumption that the colored tube
is matched in brightness for the fish by one or more of the
gray tubes. If a long series of closely approximated gray
tubes is used, it is probable that one of them matches the
colored tube, but it always remains possible that none of
them matches for the fish. There may remain a bright-
ness difference between the colored tube and any of the
grays great enough to enable the fish to discriminate them.
To obviate this, the brightness limen must first be found.
It cannot be too often pointed out that brightness is not
measurable by physical means, but is a psychical or physio-
logical quality, so that colors matched in brightness for
one animal may or may not be matched for another. What
is brighter or duller each of us can know directly only for
himself. It is, therefore, important to use a training method
in which the behavior of the fish affords, a means of judging
when the two color stimuli are matched in brightness.

82 CORA D. REEVES

The training procedure used in my own work belongs
in the third category outlined. It resembles the procedure
of Frisch with the colored and gray tubes, in that a posi-
tive stimulus of constant intensity is presented to the fish
along with a variable. In Frisch's work the constant was
a colored tube, while the variable was a series of gray or
colored tubes, all of which were shown at the same time
with the constant. In my own experiments the constant
is a patch of blue light, the variable is a series of like patches
of red light of different intensities shown one at a time
in contrast to the constant. The intensity of the illumina-
tion used for the patches is known, and in the case of the
variable the intensities form a series the members of which
are close together. At one point in this series, usually at
the intensity of the variable corresponding to slit width
0.9 miji., the fish temporarily fail to discriminate (cf.
graphs figs. 13 and 15). This failure is believed to indi-
cate the region of matched brightness of the two stimuli
for the fish. If this is true, the fish themselves show by
their behavior when the colored patches are matched in
brightness for them. As the variable is held at this matched
brightness, the fish again come to discriminate the two
patches. The conclusion seems inescapable that the quality
of the light is the basis of the discrimination. The method
thus appears to meet the difficulties encountered in the
other training methods, in that it affords a behavior-indi-
cator for matched brightness. It has not hitherto been
used in training experiments on fish.

c. Change of color in fish in response to background was
long ago shown by Pouchet (1876) to be response to stimu-
lation received through the eye. It does not occur in
blinded fish (see also Mast, 1916). The use of this method
to afford evidence of color vision in fish has been discussed.
Compared with the time required for discrimination of
colored stimuli by trained fish, a relatively very long time
is required for these color changes in the integument to
occur. While Mast's experiments with blinded flounders
show that the color changes are mediated through the eye,
Frisch has shown that in the minnow, Phoxinus, they are
influenced by illumination of the integument of the pineal

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 83

region. In eyeless Amblystoma Laurens (1917) has shown
that the melanophores contract in light and expand in
darkness. It may then be suspected that the mechanism
by which the changes in the melanophores are controlled
is more complicated than we know. The slowness of these
changes and the fact that they may not be wholly medi-
ated through the eye make it difficult to draw conclusions
from them with regard to color vision, and especially diffi-
cult to compare such conclusions with those derived from
training experiments. In the one case we have a slow
reaction to background, while in the other we have, if
my results be accepted, instantaneous choice of one of
two stimuli which differ only in quality of light. In the
latter case we may speak of color vision, or wave-length
discrimination. In the former it is doubtful whether
either discrimination or vision is involved. The choice by
Mast's flounders of a background of the same color as that
to which they had been previously adapted affords evidence
more nearly comparable to that obtained by training.

B. Sources of Error

From the foregoing review of the literature it appears
that workers have used many methods of experimental
attack on the problem of color vision in fish. All workers
except Hess conclude that fish have the capacity to dis-
criminate wave-lengths of light. My own results support
this conclusion., but before discussing them further I shall
take up, on the basis of my own experience and from a
study of the literature, the sources of error involved in
such work.

a. Errors may arise from lack of proper environment and
care. — The need of keeping the fish in normal physiological
conditions has already been referred to. The chemical
contents of the water, the size of the containers (Franz,
1910, 1911), the tameness or wildness of the fish, are all
factors to be considered in judging of the effectiveness of
environment in modifying appearance and behavior.
Through neglect of these factors Hess seems greatly to
have lessened the value of his experiments. He tells us
(1913) that while he was experimenting two of his twelve

84 CORA D. REEVES

fish died. This indicates that his other fish were not in normal
condition. They may have failed to give responses that
could have been obtained from normal fish.

b. Errors may arise during experiment from uncontrolled
environmental factors. — The procedure of Hess (1909) in
comparing the sensitivity of dark- and light-adapted fish
has been already described. The aquarium containing the
light-adapted fish was moved just before testing them,
while that containing the dark-adapted fish was not moved.
To compare the responses of the two sets of fish under
these conditions cannot be regarded as a sound procedure.
Again, Hess (1913) in studying the influence of the back-
ground in modifying the colors of fish puts several into the
same containers or into separate containers so near together
that the fish are visible to each other. He thus ignores
the fact that the colors of one fish may be influenced by
the presence of another fish. I have several times observed
the colors of fish when two are in one container. If one
chases the other, the pursuer, as observed, may become
lighter in color and the pursued darker or at times a re-
versed change may take place. In the breeding season,
I have found the male rainbow darter to change strikingly
in color from second to second and without change of back-
ground (Reeves 1907). Colors flash out with the incidence
of various stimuli, such as the presence of other fish or the
movement of the observer. Similar changes may be seen
at other seasons. Yet it appears that both Hess (1913)
and Frisch (1911), in some of their experiments, went to
the aquarium and stood over the fish to observe whether
they changed to match the background. This procedure
in itself would cause color change in some of the species
with which I have worked. (See also Freytag, 1914).
Through neglect of these factors Hess seems greatly to
have lessened the value of his earlier experiments. Frisch
appears to have avoided putting a number of fish into one
container without running water and into such poorly
adjusted conditions as to cause the death of his fish, but
for the control of all the complex physiological and psycho-
logical factors that cause response, his methods are not
entirely satisfactory. The recent work of Mast (1916)

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 85

gives evidence that some fish can and do change color to
match the background upon which they are resting. Mast
presents pictures made by color photography of fish which
have become like the background. These were taken after
three months of adaptation.

c. Errors may arise from inhibition of response. — Hess
(1911) shows that fish fail to eat when red light is thrown
upon them. Of his experiments in which a worm was
upon white sand on the bottom of the aquarium illuminated
by a Nernst lamp with a red glass filter in front of it, he
says, in effect, that more convincing than all else to the
lay mind is the fact that the fish left the worm undisturbed,
but on illuminating with blue light, the fish snatched the
worm at once though it was then less conspicuous to the
human eye. Here Hess observed the same response that
Bauer got and called " Rotscheu," but Hess misinterprets
and says that the red was of low illuminating value. He
adds that when brightness of the red was slowly increased,
the worm was seized and eaten. Many times my fish have
failed to feed upon a bit of white worm when it lay wrig-
gling on the black aquarium bottom in front of the colored
plates, but when I turned off the lights for the stimulus
plates and left only the general illumination from the
ceiling light the fish have fed at once, though the total
illumination was reduced. The fact was, the colored patches
so stimulated the fish that the feeding behavior was in-
hibited. When Hess slowly increased his red, it is not
stated how long the fish was allowed to become used to
the red illumination. It may be that the time required
for the slow increase in the red sufficed for the fish to be-
come used to it, if fear of red was originally present.

d. Errors may arise through lack of training at matched
brightness. — When stimuli of different wave-lengths are
employed in training experiments, it is always essential
to vary the brightness of one of the two, for the purpose
of finding a point at which the two stimuli match in bright-
ness. If at this point discrimination fails, it may be argued
that the discrimination shown in other parts of the series
has been due to brightness difference and that such a series,
therefore, affords no evidence of color vision. If discrim-

86 CORA D. REEVES

ination occurs at matched brightness for the fish, it may
be taken as evidence of color vision. This general form
of experiment was used by Frisch (1911) in the experi-
ments in which the fish discriminated a colored food tube
in a series of grays or colored tubes. The results were
rightly judged to be an indication of color vision. Had
discrimination not taken place in this case, the result
would ordinarily have been taken to indicate a lack of
color vision. This conclusion could be accepted only in
case the trials were long enough continued at matched
brightness of the two colors to make it certain that failure
to discriminate persisted. In other words, in such a series
temporary failure to discriminate at matched brightness
is not conclusive. This source of error has not "appeared
in previous work with fish. A reference to the graphs
(fig. 13) shows how it might appear. If in the case of Small
Sunfish, upon failure to discriminate at slit- width 0.9 mm.,
(matched brightness) the slit width had been at once con-
siderably reduced, discrimination would have been re-
established. ■ The curve might then have been interpreted
as evidence of brightness discrimination with failure at
matched brightness. Only by continuing the trials at matched
brightness and securing return of discrimination at this slit
width does it give evidence of discrimination of wave-
lengths.

e. Error may arise from exclusive use of the method of
response. — This method does not accurately test the pDtenti-
ality for discrimination for fish. It shows not inability but
merely failure to respond under the conditions of the ex-
periment. Only a method which necessitates the formation
of a definite habit of response to a definite sensory stimulus
can be expected to show the ability of the fish to respond.

C. General Considerations

a. Difference in species of fish investigated. — All fish
that have been used in investigations on color vision belong
to the group Pisces, one of the three classes of vertebrates
popularly referred to as fish. Few species have been in-
vestigated. It has been pointed out that the capacity to
discriminate by means of wave-lengths is not the same in

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 87

the two species chiefly investigated by me and that the
capacity of individuals of the same species is not necessarily
the same. Even a single individual may be expected to
show responses varying with age or condition. That the
conflicting results obtained by different investigators may
in some cases have arisen from difference in species used,
in individuals, or in ages, is possible.. It is also to be ex-
pected that color blindness occurs among fish as among
men.

b. The responses to wave-length mediated through the
eyes. — The question may be raised whether the responses
to wave-length are conditioned through the eye. Parker
(1909) has shown that in the species of fish (dogfish, AIus-
telus canis\ eel, Anguilla chrysops; killifish, Fundulus
heteroelitus] scup, Stenotomus chrysops; cunner, Tauiogo-
labrus adspersus; tautog, Tautoga onitis; puffer, CJiilo-
mycteriis schoepfi; toadfish, Opsanus tan; tomcod, Micro-
gadus tomcod) investigated by him, the skin is not sensi-
tive to white light even when of great intensity. The
eyes remain as the only possible mediating sense organ.
Mast (1916) found that flounders " with both eyes removed
do not simulate the background either in shade, color, or
pattern." The evidence that with the eyes functioning
these fish become colored like the background upon which
they rest has been mentioned. AVith the sunfish there
never was reason to doubt that the behavior toward light
of different wave-lengths is governed by stimuli received
through the eye. The eye movements toward the source
of light are striking, and occur at all intensities of red light.

c. Peculiarities of the teleostean retina. — Hess suggests
that fish do not necessarily have color vision because men
can see how it would be of use to them if present. With
equal reason it may be said that fish do not necessarily
lack color vision because men cannot see how it would
be of use to them if present. Yet Hess argues the lack of
red vision on the ground that red rays do not reach the
environment of certain fish. The truth is that only ex-
perimental evidence can determine whether color vision
exists in fish. This evidence has been considered and the
conclusion drawn that the fish used in my experiments

88 CORA D. REEVES

and probably those used by others have color vision, but
certain peculiarities of the eyes of fish make it probable
that their color vision differs from that of terrestrial forms.
The fish eye shows peculiarities of the visual purple and
of the cones which suggest that the longer wave-lengths
of light may have more marked effect upon them than
upon man. These peculiarities are as follows:

(1) Kottgen and Abelsdorff (1896) have published curves
showing the relative absorption of light of different wave-
lengths by the visual purple of different animals. These
show the maximum absorption for fish at wave-length
540 /A/A, while for mammals, birds, and Amphibia it is at
500 MAi.

(2) According to Ewald u. Kuhne (1878) this visual pur-
ple in fish has not the true purple color found in terrestrial
vertebrates but is somewhat reddish-violet. This is sig-
nificant, for researches^" on organic compounds show that
the farther their absorption band lies toward the red, the
more sensitive is the chemical to light. From the differ-
ences in color and absorptive power between the visual
purple in fish and higher vertebrates, we should expect to
find the relative sensitivity for light of longer and shorter
wave-lengths different from that in terrestrial vertebrates.
We might even predict from the work of Kottgen and
Abelsdorf that fish would show the greater sensitivity to
red subsequently found by Bauer and others.

(3) In another respect the retina of fish differs from
that of higher vertebrates. It possesses (Eigenmann and
Shafer 1900) large cones and peculiar twin cones, structures

■ which may have advantages in the reception of the reduced
number of red rays of light. Whether wave-length dis-
crimination is confined to the cones (Schultze 1866 and v.
Kries, 1896) or is in part the function of the rods and the
visual purple (Kiihne 1879, Siven and Wendt 1903), the
fish eye has modified anatomical structures which may be
the basis for differences in the functioning of the eye as
compared with the human eye, and for increased sensi-
tivity to red.

" Reported by Bayless from Luther and Nikolopoulos.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 89

D. Final Statement of Results

This paper embodies an attempt to learn whether fish
are capable of responding differentially to lights of longer
and shorter wave-lengths. The experimental evidence
presented is of two sorts — that in which the fish went for
food to a patch of light of shorter wave-length when pre-
sented at the same time with a patch of longer wave-length,
and that in which the unlearned responses of the fish to
light of different wave-length were observed.

By the first method all the seven fish used learned to
go to the patch of blue light of constant intensity before
which they were fed, when this was presented at the same
time with a red patch. There was then established a food
association or sensory-motor habit. When the intensity
of the red patch was gradually reduced the -fish continued
to go to the blue until the brightness of the red was ap-
proximately that of the blue for the human dark-adapted
eye. At this value of the red most of the fish at first failed
to discriminate the two patches and went about as often
to the one as to the other. But as the trials were continued
at this matched brightness the fish again discriminated and
continued to do so with further reduction in the intensity
of the red. The value of the red variable at which tempor-
ary failure to discriminate and subsequent recovery take
place is believed to be that at which the red and blue are
of matched brightness for the fish. The behavior of the
fish at this value thus becomes an indicator for matched
brightness. The fish then learned to discriminate the red
from the blue at all values of the red variable used includ-
ing one of matched brightness for them. (See graphs figs.
13 to 16 and discussion of them elsewhere). Check tests of
the apparatus used and of the method of manipulating
it seem to have excluded the possibility that the fish were
guided in their discrimination by anything else than the
stimulus patches, and at matched brightness these patches
differed for the fish and for man only in the wave-length
of light coming from them.

This evidence from training experiments is reinforced
by that obtained from a study of the unlearned response
of the fish. From this evidence it appears that in the

90 ■ CORA D. REEVES

training experiments the fish show an initial avoidance of
red and that they rapidly become adapted to red and then
no longer avoid it. (Graphs 13 and 15 and discussion of
these in text.) It appears further that when first presented
with red and blue patches of matched brightness for the
human eye fish show without training, by manner and fre-
quency of approach to the patches, that the stimulus value
of the two is not the same for them. When in experiments
they have become adapted to red and blue patches, their
behavior is greatly altered when a gray patch of matched
brightness is substituted for the red. Many innate responses
indicate the different stimulus values for the fish that
lights of longer and shorter wave-lengths have.

In spite of the fact that the fish discriminated red and
blue stimulus patches when they were matched in brightness
for the human eye and when the behavior of the fish indi-
cated that they were matched in brightness for them; it
is possible to postulate extraordinary powers of brightness
discrimination for the fish, powers far beyond those of
man, and to explain the results obtained with the use of
food association on that basis. Thus, postulating high
capacity to differentiate slight brightness differences, we
might explain the graph in figure 13 on the basis of such
discrimination. Failure to discriminate at slit width of 0.9
mm. would then be due to the fact that the stimulus patches
were nearly but not quite matched in brightness for the
fish. Recovery of discrimination at this slit-width could be
attributed to continued training at these slightly differing
brightnesses. To determine whether the fish possesses
extreme powers of brightness discrimination tests were
made to find by how much two white stimulus patches
must differ in intensity in order that the fish may learn
to go to the duller patch for food. It was found that in-
tensity differences of white light of 1 to 4 were necessary
for discrimination in the case of dace and of 1 to 3 in the
case of sunfish. The fish do not therefore under the con-
ditions of the experiments show powers of accurate bright-
ness discrimination sufficient to explain the graphs shown in
figures 13 and 15, nor sufficient to explain their innate dif-

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 91

ferential responses toward patches of light of different wave-
length but of presumably matched brightness.

It is a priori not impossible that the responses of the
fish to the stimulus plates are determined by the total
energy reaching the fish from them, as measured by physi-
cal instruments, rather than by the stimulus value of the
light rays alone. Tests were therefore made to determine
this point but no relation could be detected between total
energies and the responses of the fish. (p. 53.)

Throughout the work the attempt has been made to
escape the sources of error enumerated in the last section.
The fish have been kept in normal physiological state and
it is believed that environmental factors likely to modify
normal response or to inhibit it have been detected and
eliminated.

The conclusion is reached that the fish used are capable
of responding differentially to light of longer and shorter
wave-lengths. This capacity is made evident by the use
of a food association. In the absence of modifying or
inhibiting environmental factors differential response may
be apparent in the unlearned responses of the fish of some
species. The two sources of evidence thus supplement
each other. Evidence is collated to show that the response
to light of longer and shorter wave-lengths is mediated
through the eye. We may therefore speak of color vision
in fish, but attention is called to the differences found by
others between the retinal structures and visual purple of
fish and higher vertebrates, differences which indicate that
the stimulus value of the longer and shorter wave-lengths
of light may not be the same in the two cases.

The evidence presented in this paper that the longer
wave-lengths of light have a physiological or psycholog-
ical effect upon some fish different from that produced by
the shorter rays is, I think, sufficient, even in the sense
demanded by Watson (1914, p. vS5-i) when he writes,

" Ordinarily we mean when we say that an animal is sensitive to difference in
wav^e-length? that such stimuli play a role in the adjustment of the animal to food,
sexual objects, shelter, escape from enemies, etc., t. e. that such stimuli initiate
activity in arcs which end in striped muscle."

I prefer to designate what I have found as difference
in stimulating value of the longer and shorter wave-lengths

92 CORA D. REEVES

of light not entirely dependent upon the intensity of the
light.

V. SUMMARY OF EXPERIMENTS

A. Apparatus

1. A T-shaped experiment aquarium of
galvanized iron supplied with running
filtered water was used, and in this the
fish remained not only during experi-
ments but in the intervals between
them. p. 9, figs. 1, 2.

2. The aquarium was divided into com-
partments. The first of these con-
tained two stimulus plates of opal
ground glass, which could be exposed
to the fish to be immediately tested
by lifting a sliding door separating it
from a second compartment. The third
compartment contained fish not being
tested, p. 5, fig. 1.

3. The partitions between the compart-
ments were movable, and by manipu-
lating them the fish could be shifted
from compartment to compartment
without causing fear reactions, p. 9,
figs. 1, 2, 6.

4. The stimulus plates were illuminated
by lamps in fixed positions, contained
in a light-tight house, pp. 7, 10, fig. 2.

5. The intensities of the illumination on
the plates were varied by means of
adjustable slits beneath the lamps, pp.
7, 12, fig. 2.

6. Well defined areas of light of practi-
cally uniform distribution were secured
upon the stimulus plates by the use
of cylindrical lenses, pp. 7, 12, fig. 2.

7. Red and blue filters were used to
secure light of restricted wave-lengths,

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 93

and by shifting the filters a red or a
blue light could be secured on either
plate.

8. By use of slits the intensities could be
varied. The aquarium was illuminated
by a ceiling light of constant intensity,
and other light was excluded from it
by a light-tight enclosure.

9. A curtain was hung about the aquarium
within the light-tight enclosure in such
a way as to conceal the experimenter
from the fish.

10. Food was released before one of the
stimulus plates (blue or white) by
means of a device operated by an elec-
tro-magnet controlled by a switch,
which could be manipulated from out-
side the curtain. The sliding door by
means of which the stimulus plates
were exposed to the fish could also be
controlled from outside the curtain.

B. Method of experimentation

11. Training methods were used in which
a food association was established by
feeding the fish before a stimulus-
patch of constant intensity in contrast
to one of variable intensity. By shift-
ing filters and adjusting slits the posi-
tions of constant and variable stimuli
were irregularly interchanged. In ex-
periments with light of restricted wave-
length, blue was used as the constant
and was the positive stimulus.

12. The unlearned differential responses of
the fish toward lights of different wave-
lengths and different intensities were
studied by the use of the same appa-
ratus.

94 CORA D. REEVES

C. Care of the fish

13. The fish used in experiments with food
association were horned dace (Semo-
tilus atromaculatus, Mitchell) and Com-
mon Sunfish ( Eupomotis gibbosus Lin-
naeus). Individuals of these species as
well as of the blunt-nosed minnow,
Pimephales notatus, Rafinesque, the
Common Shiner, Notropis cornutus,
Mitchill, and the mud minnow, Umbra
limi, Kirtland, were used in experi-
ments on differential response.

14. When not in the experiment aquarium,
the stock fish were kept in separate
aquaria. When needed for experiment,
they were transferred to the experi-
ment aquarium, where they lived often
for months together. During more
than two years these fish have shown
by feeding regularly, by lack of fear,
by continued growth, and by breeding
behavior that the conditions were nor-
mal for them.

D. Experinients with food association

I. Intensity jdiscrimination of white light

15. Three full-grown individuals of the
horned dace were given an aggregate of
307 trials in an attempt to form a
food-association with the duller of
two white plates, the intensities of
which were as 1 :4. They failed to
discriminate these plates although they
learned to discriminate plates of greater
intensity difference, (p. 20.)

16. Three sunfish, 4 cm. long, were tested
for intensity discrimination without
training by noting their behavior to-
ward two white plates. When the

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 95

intensities of the two plates were in a
ratio somewhat greater than 1 :2, the
fish showed differential response to-
ward them by a difference in frequency
or method of approach to them. With
less intensity difference no differential
response was observed, (p. 37.)

17. A single sunfish, 8 cm. long, was given
160 trials before two white plates of
different intensities, and was fed be-
fore the duller plate. The fish dis-
criminated at once plates differing as
1 to 4. In an attempt to use the food
association thus established the fish
discriminated plates differing in in-
tensity as 1 :3 poorly but not those
that differed less. (p. 38.)

II. Discrimination of light of different wave-lengths
a. White-red

18. Fifty trials were given to each of two
horned dace by feeding before a white
plate presented at the same time with
a red of different intensity. These
fish were accustomed to red and were
not red-shy. After 30 trials each fish
made 80% to 90% of correct or white
choices, and each finally gave 10 suc-
cessive correct choices. (p. 23.)

19. By means of a peculiar unlearned re-
sponse of the dace of section 18 to red
and white plates these plates were
matched in brightness for them. (p.
24.)

20. Twenty trials were then given to each
of the horned-dace of section 18 before
these red and white matched plates,
while the fish were fed before the white.
Seventy per cent of correct or white
choices were obtained. (p. 25.)

96 CORA D. REEVES

b. Blue-red

21. Four untrained but tame horned dace
were tested before a brilliant red and
duller blue plate in order to learn
whether these plates were effective
stimuli. There was an initial avoid-
ance of the bright red plate but this
soon ceased and the behavior of the
fish showed that both plates were
effective stimuli. (p. 26.')

22. Four horned dace were given 420 trials
before blue and red plates with the
red varied in intensity so as to include
an intensity equal to the blue. They
were fed before the blue. The results
are plotted in the graphs, figure 13,
p. 31, in average percentage of correct
choices at different intensities of red.
At an intensity of the red obtained
by a slit-width of about 1 mm. the
percentage of blue choices diminishes
and the curve falls to near the 50%
level. This is believed to be the
region of matched brightness of red
and blue for the fish. As the trials
are continued with the red held at this
brightness, the percentage of blue
choices again increases as is shown by
the rise in the curve. With subsequent
lessening of the intensity of the red
the percentage of correct choices con-
tinues to be high. The fish then dis-
criminates the red and blue patches
at all intensities of the red. (p. 30
also figs. 13 and 14.)

23. Three sunfish were given 680 trials
before a blue in contrast to a red plate
as in the preceding experiment; they
were fed before the blue. The graphs
showing the percentage of correct

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 97

choices at various red intensities are
shown in figure 15, p. 39. The graphs
have the same general form as those
of the dace so that the statements of
sec. 22 apply to them.

c. Blue-gray

24. A brief series of trials was made before
blue and gray plates of matched
brightness for the human eye, with
blue as the positive plate, with two
horned dace. One went to the gray
and blue with about equal frequency:
the other chose blue in 13 out of 17
trials, (p. 49 and Table II.)

25. A brief series of trials was made before
blue and gray plates of matched bright-
ness, with the blue as positive plate,
with three sunfish previously trained
to go to blue for food. Seventeen out
of tw^enty choices were to the blue
plate, (p. 49 and Table II.)

E. Unlearned responses
1 Breathing rate

a. Umbra limi

26. After stimulation with red light a mud-
minnow. Umbra limi, was found pant-
ing rapidly. The red was the only
unusual factor in the environment,
(p. 57.)

b. Notropis cornutus

27. The breathing rate of a shiner, Notropis
cornutus, increased by one-fourth when
the illumination of white light to which
it was accustomed was increased by
the addition of a carbon lamp. But
when this increased illumination was
lessened by a red filter the breathing
rate more than doubled, (p. 58.)

98 CORA D. REEVES

2. Red avoidance and behavior toward blue

a. Eupomotis gihhosus

28. One sunfish circled about the red
lighted area as far from it as possible
and with her head toward the light.
This fish often seemed to bask in the
blue light, remaining with the direct
rays of light shining upon her and with
one side turned so as to receive more
light than when in a vertical position,
(p. 58.)

b. Pimephales notatus

29. When matched blue and gray plates
were presented to several untrained
Pimephales notatus, they promptly ap-
proached them; but when a red plate
was substituted for the blue, they
remained at a distance, (p. 59.)

c. Semotilus atromaculatus

30. When an untrained fish was tested
with matched blue and gray patches,
one-third to one-half of a minute was
sufficient for the fish to swim up to
the lighted areas, but when the same
blue patch was matched with red,
approach was completely inhibited or
was inhibited for five or six minutes

3. Delayed Response

31. Two dace and three sunfish, fed before
the blue plate from October to March,
in the blue-red series, usually reached
the stimulus patches three to twelve
seconds after the opening of the door
D (fig. 2). They were shown the blue
with a matched gray. They failed to
leave the discrimination compartment
or left after a long wait. The average
time of response was over 100 seconds
for the four fish showing the best dis-

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 99

crimination. Instead of the usual
direct approach there was a weaving
back and forth in front of the door or
circHng around the end of the discrim-
ination compartment or remaining still
in a corner.

F. Experiments with equated energies.

32. When in the blue-red discrimination
experiments the energies reaching the
two stimulus plates were equated by
the use of a silver-bismuth thermopile
the percentage of correct choices con-
tinued to be significantly high. (p. 53
and postscript.)

G. Tests of apparatus and manipulations

33. Check tests to determine whether shift-
ing of slits and filters, changes in the
position of the experimenter, or dif-
ference in appearance of any part of
the apparatus on the positive and
negative sides might afford the fish a
clue gave negative results.

VI. CONCLUSIONS

Sunfish (Eupomotis gibbosus L.), horned-dace (Semo-
tilus atromaculatus, Mitchill) discriminate light of longer
wave-length from light of shorter wave-length and from
white light. This is shown not only by innate differential
reaction but by the results of experiments in which a feed-
ing response was associated with a stimulus of restricted
wavelength.

Acknowledgments

I am obligated to Professor Jacob Reighard, under whose
direction the work has been carried on, for unstinted aid
in planning the work and for careful oversight of the details
of the problem. To Professor John F. Shepard I am much
indebted for constant suggestions and stimulating criti-
cism. For the use of physical apparatus and for advice
thanks are due to Professor H. M. Randall.

100 CORA D. REEVES

PoSTvSCRIPT

While this paper was being made ready for the press in
the fall of 1917, Doctor Reeves took up her residence
at Ginling College, Nanking, China. War has delayed
the publication. Owing to Br. Reeves' absence and the
lapse of time I am venturing to add the following notes
for which I alone am responsible.

1. The absorption of rays of the visible spectrum by the
water of Wisconsin Lakes has been recently studied by
Pietenpol (Trans. Wis. Acad. Sci., XIXI (1918), 562-593).
In connection with the discussion on pp. 74-75, it seems
best to record here a recent simple experiment of my own
on the visibility of red through lake water. In February,
1919, when Winan's Lake in Livingstone Co., Mich., was
covered by about eight inches of clear ice, without snow,
I suspended through a hole in the ice and at a distance of
three feet below the surface a cylindrical tin can covered
with red paper (Klinksieck's color code No. 56). At a
distance of thirteen feet from this hole was a small spear-
house which shut off a good part of the light from a second
hole cut through the ice which formed its floor. By means
of a mirror immersed to a depth of a foot through this
hole I was able to view the can. Its color, as thus viewed,
was unmistakably bright red, merely a little duller than
when seen through air. The red rays which reached my
eye had travelled a distance of 16 feet through water.
With the great sensitivity of fish to red I have no doubt
that it stimulates them through this lake water at much
greater distances.

2. Miss Gertrude Marean White (Jour. Exp. Zool. Vol.27,
pp. 443-498, 1919) has recently published an account of
ingenious experiments on color vision in sticklebacks and
mud-minnows. The m.ethods used were those of the third
training procedure (this paper p. 81) and the results reached
are those of other investigators except Hess and are in
accord with those of Doctor Reeves. They do not involve
the equation of brightness through the behavior of the fish
and are therefore open to the same criticism as other ex-
periments of their type. Miss White also reports ex-
periments on the inability of these fish to discriminate
grays of very unequal brightness and in this respect con-
firms Doctor Reeves (this paper pp. 20 and 37).

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 101

3. In all her measurements of the energy reaching the
two colored stimulus plates (pp. 53-54) Dr. Reeves used
a new blue filter. Since the filter transmitted more light
than the old filter, a slit was used with it to reduce the
intensity of the blue to that obtained with the old filter
to which the fish were accustomed. The width of the
slit, once adjusted, was not varied. A water-cell on the
blue side served as an infra-red filter, if indeed such a
filter were needed, since no such rays should have been let
through by the blue filter. The energy which reached the
blue plate during the measurements with the thermopile is
therefore the same as that which reached the fish from
the blue plate during their tests. The two might differ some-
what on account of the unequal thickness of the strata
of water involved but in both long rays should be wholly
excluded.

But in measuring the energy reaching the red plate two
methods were used to vary the intensity of the red light — a
red solution alone and a slit alone. Since red solution and
aquarium water both act as infra-red filters in greater or
less degree, whatever energy reached the red plate through
the one should reach the fish through the other though in
less intensity. Infra-red was presumably reduced under
the two conditions. When a slit alone was used to vary
the red intensity there was no infra-red or long-ray filter
present on the red side in the energy-equation measure-
ments. By this method infra-red rays would reach the
red plate unhindered. When the energies had been thus
equated and the fish were then tested the aquarium water
served as an infra-red filter on both sides. The energies
presumably equated with water cell on one side only
would then no longer be equated for the fish.

It appears that both methods of varying the intensity
of the red permit the energies of the two plates to be equated
but that only one of them, that in which a red solution
was used, retains approximately those equated energies
in the stimuli which reach the fish through the aquarium
water.

Table IV, p. 54, presumably includes trials in which
both methods of equating energy were used. The table

102 CORA D. REEVES

shows that two out of three sunfish discriminated while
from the text it appears that the third, large sunfish,
discriminated accurately in later trials not included in
the table. Three dace discriminated, but one did not.
Its failure is perhaps adequately explained on p. 54. The
table might be interpreted as showing response to unequal
energies if only the slit method of varying the red had
been used in equating. Since both methods were used,
energies were equated in a considerable but unknown
number of trials and in these the fish should have failed
if they were reacting to energy differences. The high
percentage of correct choices for five (or six) of the fish
indicates that Doctor Reeves is right in holding that they
were not responding to total energy differences.

Undoubtedly the tests should be repeated and in equating
energies of the red and blue plates the light for both should
be passed through filters of the identical water used in the
experimental aquaria and as thick as the layer of water
through which the fish ordinarily see the stimulus patches.

The recent work of Pietenpol gives reliable data on the
absorption of rays of different wave-lengths by the waters
of Wi-sconsin lakes. It is clear that absorption is unequal
and that either the red or the blue or both may be absorbed
more rapidly than other rays. The total-energy ratio of
two stimuli of unequal wave length probably could not
be constant for a fish unless its choice is made always
at the same distance from the stimulus plates. This was
by no means the case in Doctor Reeves' experiments. As
judged by the behavior of the fish (change of direction)
choices were made at all possible distances and with all
possible variation in the energy ratios. This seems to me
to show that choices could not have been made on the basis
of energy differences nor of brightness differences, but
only as the result of quality difference. But the truth
is that no value of the red variable was found at which
the fish did not learn to discriminate the longer and shorter
wave lengths. Amongst the values tried was one which
showed temporary failure of the fish (slit width 0.9-1.0
mm. for red). Whether we assume that at this point we
have to do with matched brightness or equated energies

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 103

is immaterial — for even here the fish learned to discriminate
and could have been guided only by quality differences.

Jacob Reighard.
Zoological Laboratory,
Ann Arbor, Michigon.
May 23, 1919

104 CORA D. REEVES

VII. BIBLIOGRAPHY

Abney, Sir William.

1912. Researches in color vision. London.

AuFSESS, Otto Freiherr Von u. Zu.

1905. Die Physikalischen Eigenschaften der Seen. In Sammlimg Nalnrivis-
senchajilicher imd Mathematischer Monographien, Braunschweig.

1913. Seeing under water: how things look from a fish's point of view; ab-
stracted by the Berlin correspondent of the Scientific American from the
Deutsche Alpenzeitung, Jan., 1913. Scienl. Amer., Vol. 108, pp. 469-470

Bauer, Victor.

1910. Ueber das Farbenunterscheidungsvermogen der Fische. Arch, ges
Physiol., Vol. 133, pp. 7-26.

1911. Zu meinen Versuchen iiber das Farbenuntersheidungsvermogen der
Fische, idem.. Vol. 137, pp. 622-626.

BoviE, W. T.

1915. The action of light on protoplasm. Amer. Jour, of Trop. Diseases and
Preventive Med., Vol. 2, No. 8, Feb., pp. 506-517.

Coblentz, W. W.

1913. Instruments and methods used in Radiometry, Bull. Bureau of Stand-
ards. Vol. 9, pp. 7-63.

ElGENMANN AND ShAFER.

1900. The mosaic of single and twin cones in the retina of fishes. Amer.
Nat., Vol. 34, pp. 109-118.

EWALD AND KUEHNE!

1878. Untersuchungen fiber den Sehpurpur. III. Veranderungen des Seh-
purpurs in der Netzhaut in Leben. Heidelberger Unters I. P. 370.

Franz, V.

1910. Phototaxis und Wanderung. Intern. Revue d. ges. Hydrobiol.u. Hydrogr.,
Vol. 3, pp. 306-334.

1911. Weitere Phototaxisstudien, ibid, Biologische Supplemente, Series 3
(1911-1912), 23 pp.

Freytag, G. V.

1914. Lichtsinnuntersuchungen bei Tieren. Archiv. f. vergleich. Ophthalm.,
Vol. 4, pp. 151-161.

Frisch, Karl V.

1911. Beitriige zur Physiologic der Pigmentzellen in der Fischhaut. Arch. f.
ges. Physiol., Vol. 138, pp. 319-387. 2 pis. 8 figs.

1911a. L'eber den Farbensinn der Fische. Verhandl. d. deutsch. Zoolog. Ges,.
20-21 Vers. pp. 220-225.

1912. L'eber farbige Anpassung bei Fischen, Zool. Jahrb. Abt. f. allg. Zool.
und Physiol., Vol. 32, pp. 171-230, 2 pis. 4 figs.

1912a. Sind die Fische farbenblind? Zool. Jahrb., Abt. f. allg. Zool. und Physiol.,

Vol. 33, pp. 107-126, 2 figs;
1912b. Ueber die Farbenanpassung des Crenilabrus, ibid, pp. 151-164.

1913. Weitere Untersuchungen iiber den Farbensinn der Fische. Zool. Jahrb.,
Vol. 34, pp. 43-68. 5 figs.

Goldsmith, M.

1914. Reactions physiologiques et psychiques des poissons. Bull. I'lnstitut
general psychologique, Paris, Vol. 14, pp. 97-225.

Graber. V.

1885. Ueber die Helligkeits und Farbenempfindlichkeit einiger Meertiere.
Sitzungsber. d. Akad. d. Wissensch. in Wien. Math. Klasse, Vol. 91,
Abt. 1. pp. 129-150.
Henri, Victor, et Larguier des Bancels, J.

1911. Photochimie de la retine. /. de physiol. et de pathol. gen.. Vol. 13, pp.
841-856.

LIGHT OF DIFFERENT WAVE-LENGTHS BY FISH 105

Hess, C. v. . , . ^. , . , , , ■„

1909. Untersuchungen iiber den Lichtsinn bei Fischen. Arch. t. Augenheilk.

Vol. 64, Ergiinzungsheft, pp. 1-38, 12 figs.
1910 Ueber den angeblichen Nachweis von Farbensinn bei Fischen. Arch.
/. d. ges. PlivsioL, Vol. 134, pp. 1-14.
■ 1911. Experimentelle Untersuchungen zur vergleichenden Physiologie des
Gesichtssinnes. Ibid., Vol. 142, pp. 405-446, 5 figs.

1912 Untersuchungen zur Frage nach dem Vorkommen von Farbensinn bei
Fischen. Zool. Jahrb., Abt. f. allg. Zool. u. Physiol., Vol. 31., pp. 629-643.

1913 Neue Untersuchungen zur vergleichenden Physiologie des Gesichtssinnes.
Zool. Jahrb., Abt. f. allg. Zool. u. Physiol., Vol. 33, pp. 387-440, 2 pis.,
9 figs.

1914. Ueber die Entwicklung von Lichtsinn und Farbensinn in der Tierreihe.

Vortrag gehalten bei der Versammlung deutscher Naturforscher and

Aerzte in Wein am 25. September, 1913. Weisbaden.
1916 Messende Untersuchung des Lichtsinnes der Biene. Arch. f. d. ges.

Phvsiol. Vol. 163, pp. 289-320.
1917. New Experiments on the light reactions of plants and animals. Jour.

of An. Beh., Vol. 7, pp. 1-10.'

Ives, H.

1912. Photometry of hghts of different colors. Phil. Mag., Vol 24, Series 6,
pp. 149-189.

KOETTGEN AND ABELSDORFF. .

1896. Absorption und Zersetzung des Sehpurpurs bei Wirbeltieren. Zeitsch.
f. Psychol, u. Physiol, d. Sinnesorg, Vol. 12, pp. 161.

KRIES, J. V. „ , r,, • 7 J

Ueber die Funktion der Netzhautstabchen. Zeit. j. Psych, u. Physiol, d.
Sinnesorg, Vol. P, pp. 81-123.
1905. Die Gesichtsempfindung. NageVs Handbuch d. Physiol, des Menchen
Vol. 3, pp. 109-279.

KUEHNE, W. ^^ , rr J, , J

1879. Chemische Vorgange in der Netzhaut. Hermann s Handbuch der
Physiologie, Vol. 3, pp. 235-337. Leipzig.

Laurens, Henry.

1911. Reactions of amphibians to m.onochromatic lights of equal intensity.

Bull. Mus. Conip. Zool., Vol. 53, pp. 253-302.
1917. The reactions of the melanophores of Amblystoma tignnum larvae to
light and darkness. Jour. Exp. Zool., Vol. 23, pp. 195-205.

Luther, R. and Andreas Nikolopoulos.

1913. Ueber die Beiziehun? zwischen den Absorptionsspektrum und der
Konstitution der Komplexen Kobaltamminsalze. Zeitsch. f. physik. Chern.,
Vol. 82, pp. 361-378.

Mast, S. O.

1911. Light and Behavior of organisms. New \ork.

1916. Changes in shade, color, and pattern in fishes, and their bearing on the
problems of adaptation and behavior with e3pe:ial reference to the flounders.
Paralichthys and Amylopsetta. Bull, of Bureau of Fisheries, Vol. 34,
(1914), pp. 177-238, 19 pis.

Moebius. ^^ . ^ , . ,

1875. Die Beweiungen der Thiere und ihr psych.isher Honzont. bchrijt, d.
Naturwiss. Ver. f. Schleswig- Holstein, Vol. I, pp. 111-130.

Murray, Sir John and Hjort, Johan.

1912. The depths of the ocean. London.

Parker. G. H. ^ ^ „ ,r i no

1903. Hearing and allied senses in fishes. U. S. Fish. Com. Bull., Vol. 22,
(1902), pp. 45-64, one plate.

■106 CORA D. REEVES

1909. The integumentary nerves of fishes as photoreceptors and their signifi-
cance for the origin ot vertebrate eyes. Amer. Jour. Phys., Vol. 25, pp.
77-80.

Parsons, J. Herbert.

1915. An introduction to the study of color vision. Cambridge.

POUCHET, G.

1876. Les changements de coloration sous I'influence des nerfs. Joiirn. de
I'Anat. Phys., Vol. 12, pp. 1-90; 113-165. Paris.

Reeves, Cora D.

1907. The Breeding habits of the rainbow darter (Etheostoma coeruleum,
Storer). A study in sexual selection. Biol. Bull., Vol. 14, pp. 35-59.

Reighard, Jacob.

1908. An experimental field-study of warning coloration in cola reef fishes.
Pap. Tortugas Lab. Carnegie Inst., Vol. 2, pp. 257-325, 5 plates, 15 tables.
Washington.

1910. Methods of studying the habits of fishes with an account of the breeding
habits of homed dace. Bull. U. S. Bureau of Fisheries, Vol. 28, 1908.,
pt. 2, pp. 1111-1136, 7 plates, 5 figs.

Schultze, Max.

1866. Zur Anatomie und Physiologie der Retina. Arch. f. mikr. Anat., Vol. 2,
pp. 185-266, 6 plates.
Secerov, Slavko.

1909. Farbenwechselversuche an der Bartgrundel. Arch f. Entw-Mech. der
Organismen, Vol. 28, pp. 629-660. 2 plates.

Shafer, George Daniel.

1900. The mosaic of the single and twin cones in the retina of Micropterus
salmoides. Archiv. f. Entw-Mech. der Organismen., Vol. 10, pp. 685-691,
one plate.

SiVEN AND WENDT.

1903. Ueber die physiologische Bedeutung des Sehpurpurs. Skand Arch. f.
Physiol., Vol. 14, pp. 196-223.
Sumner, F. B.

1911. The adjustment of flat-fishes to various backgrounds. Jour. Exper.
ZooL, Vol. 10, pp. 409-505, 13 plates.

Triplett, N. J.

1901. The educability of the perch. Amer. Jour. Psych., Vol. 12, pp. 354-360.

Uhlenhuth, E.

1911. Zur untersuchung des Farbensinnes. Biol. Centralblalt, Vol. 31, pp.
76F4771.
Washburn, M. F. and Bentley, Madison.

1906. The establishment of an association involving color discrimination in
"the creek chub, Semotilus atromoculatus. Jour. Comp. Neur. and Psych.,
Vol. 16, pp. 113-125.
Watson, J.

1914. Behavior. An introduction to comparative psychology. 8vo. pp.
XII + 439, 68 figs. New York.
Yerkes, R. M.

1903. Reactions of Daphnia pulex to light and heat. Mark Anniversary
Volume, pp. 361-377. New York.
Yerkes and Watson.

1911. Methods of studying vision in animals. Behavior Monographs, Vol. 1,
No. 2, pp. IV + 90.

ZOLOTNITSKY, N.

1901. Les poissons, distinguent-ils les couleurs? Arch. Zool. Exp. Gen. 3 Ser.
Vol. 9, pp. 1-5.

The Behavior Monographs

Edited by JOHN B. WATSON

The Johns Hopkins University, Baltimore, Md.

VOLUME 1
No. 1 The development of certain instincts and habits ixi chicks. By

Frederick S. Breed. Pp. iv + 78, $1.00, postpaid.
No. 2 Methods of studying vision in animals. By Robert M. Yerkes

and John B. Watson. Pp. iv + 90, $1.25, postpaid.
No. S An experimental study on the death-feigning of Belostoma

(-Zaitha Aucct.) flumineum Say and Nepa apiculata Uhler.

By Henry H. P. Severin and Harry C. Severin. Pp. iii -f-

47, $.65, postpaid.
No. 4 The biology of Physa. By Jean Dawson. Pp. iii + 120, $1.50,

postpaid.
No. 5 The function of the vibrissae in the behavior of the white rat.

By Stella Burnham Vincent. Pp. iv-f81, $1.15, postpaid.

VOLUME 2
No. 6 The delayed reaction in animals and children. By Walter S.

Hunter. Pp. v + 86, $1.15, postpaid.
No. 7 The Canada porcupine : a study of the learning process. By

Leroy Walter Sackett. Pp. iii + 84, $1.15, postpaid.
No. 8 Audition and habit formation in the dog. By Harry Miles John-
son. Pp. iv + 78, $1.00, postpaid.
No. 9 Habit formation in a strain of albino rats of less than normal

brain weight. By Gardner Cheney Basset. Pp. iv + 46,

$.65, postpaid.
No. 10 Distribution of effort in learning in the white rat. By John

Linck Ulrich. Pp. iii + 51, $.65, postpaid.
No. 1 1 The effect of age on habit' formation in the albino rat. By

Helen B. Hubbert. Pp. v + 55, $.65, postpaid.

VOLUME 3

No. 12 The mental life of monkeys and apes: a study of ideational be-
havior. By Robert M. Yerkes. Pp. iv 4" 145, $1.50, postpaid.

No. 13 A study of perseverance reactions in primates and rodents. By
G. V. Hamilton. Pp. iv + 65. $.75, postpaid.

No. 14 An analysis of the learning process in the snail, physa gyrina Say.
By Elizabeth Lockwood Thompson Pp. iii + 97, $1.25,
postpaid.

No. 15 The effect of length of blind alleys on maze learning: An ex-
periment on twenty-four white rats. By Joseph Peterson.
Pp. iii + 53, $.75, postpaid.

No. 16 An experimental study of transfer of response in the white rat.
By Harry H. Wylie. Pp. iii + 66, $1.00, postpaid-

VOLUME 4.

No. 17 Transfer of training in white rats upon various series of

mazes. By Rutledge T. Wiltbank. Pp. iii -1-65, $1.00,

postpaid.
No. 18 Redintegration in the Albino rat. By William Brockbank, A.

M. Pp. iii 4- 66, $1.00, Postpaid.
No. 19 Discrimination of light of different wave-lengths by fish. By

Cora D. Reeves. Pp*. iv 4r 106, $1.25, postpaid.

Regular subscription $5 per volume, postpaid.

THE BEHAVIOR MONOGRAPHS

Emerson Hall, Cambridge, Mass.

Directory of American Psychological Periodicals

American Journal of Psychology — Worcester, Mass.: Florence Chandler.
Subscription $5. 600 pages annually. Edited by G. Stanley Hall.
Quarterly. General and experimental psychology. Founded 1887.

Pedagogical Seminary — Worcester, Mass. : Florence Chandler.
Subscription $5. 575 pages annually. Edited by G. Stanley Hall.
Quarterly. Pedagogy and educational psychology. Founded 1891.

Psychological Review — Princeton, N. J.: Psychological Review Company.
Subscription (with Psychological Bulletin) $6, 480 pages annually.
Bi-monthly. General. Founded 1894. Edited by Howard C. Warren.

Psychological Bulletin — Princeton, N. J. : Psychological Review Company.
Subscription $3.25. 480 pages annually. Psychological literature.
Monthly. Foimded 1904. Edited by Shepherd I. Franz.

Psychological Monographs — Princeton, N. J.: Psychological Review Company
Subscription $4 per vol. 500 pages. Foimded 1895. Ed. by James R. Angell.
Published without fixed dates, each issue one or more researches.

Psychological Index — Princeton, N. J. : Psychological Review Company.
Subscription $1.25 200 pages. Founded 1895. Edited by Madison Bentley.
An annual bibliography of psychological literature.

Journal of Philosophy, Psychology and Scientific Methods— New York:

Sub-Station 84. Bi-weekly. 728 pages per volume. Founded 1904.
Subscription $3. Edited by F. J. E. Woodbridge and Wendell T. Bush.

Archives of Psychology — Substation 84, New York: Archives of Psychology.
Subscription $5. 600 pp. ann. Founded 1906. Edited by R. S. Woodworth.
Published without fixed dates, each number a single experimental study.

Journal of Abnormal Psychology — Boston: Richard G. Badger.
Subscription $5. 480 pages annually. Edited by Morton Prince.
Bi-monthly. Founded 1906. Entire field of abnormal psychology.

Psychological Clinic — Philadelphia:' Psychological Clinic Press.
Subscription $1.50. 280 pages annually. Edited by Lightner Witmer.
Monthly (9 numbers). Orthogenics, psychology, hygiene. Founded 1907.

Training School Bulletin— Vineland, N. J.: The Training School.
Subscription $1. 160 pages annually. Ed. by E. R. Johnstone. Founded 1904.
Monthly (10 numbers). Psychology and training of defectives.

Journal of Race Development — Worcester, Mass. : Louis N. Wilson.

Subscription $3. 460 pages annually. Founded 1910.
Quarterly. Edited by George H. Blakeslee and G. Stanley Hall.

Journal of Educational Psychology — Baltimore: Warwick & York.
Subscription $3. 600 pages annually. Founded 1910.
Monthly (10 numbers). Managing Editor, J. Carleton Bell.
(Educational Psychology Monographs. Edited by Guy M. Whipple.
Published separately at varying pnces. Same publishers.)

Journal of Animal Behavior — Cambridge, Mass.: Emerson Hall.
Subscription $5. 450 pages annually. Founded 1911.
Bi-monthly. Robert M. Yerkes, Managing Editor.

The Behavior Monographs — Cambridge, Mass.: Emerson Hall.
Subscription $5. 450 pages per volume. Edited by John B. Watson.
Published without fixed dates, each number a single research.

Psychoanalytic Review— Washington, D. C. : 3617 Tenth St., N. W.
Subscription $5. 500 pages annually. Psychoanalysis.
Quarteriy. Founded 1913. Ed. by W. A. White and S. E. JeUiffe.

Journal of Experimental Psychology — Princeton, N. J. : Psychological Review
Company. 480 pages annually. Experimental. Founded 1916.
Subscription $3.25 Bi-monthly. Edited by John B. Watson.

Journal of Applied Psychology — Worcester, Mass.: Florence Chandler.
Subscription $4. 400 pages annually. Quaiterly. Founded 1917.
Edited by G. Stanley Hall and L. R. Geissler.

The Brandow Printing Company
Al'bany N. Y.

JUL 7 1922

BEHAvioR Monographs

/ I '

r.lUlKM

Volume 4, Number 4, 1922 Serial Number 20

Edited by
WALTER S. HUNTER

Univertity of Kania*

Visual Perception of the Chick

BY
HAROLD C. BINGHAM

A dissertation submitted to the Board
of University Studies of the Johns Hop-
kins University in conformity with the
requirements for the degree of Doctor

of Philosophy.

Publifhcd by

WILLIAMS & WILKINS COMPANY

Mount Royal and Guilford Avenues

Baltimore, U. S. A.

The JotTRNAt bWOoMPiuATiVE PsTCHOLOQT wiU accept for publication, studies which
contribute to out kpipwleflge of mental function and behavior in any organism.

Manuscripts may be sent to Knight Dunlap, The Johns Hopkins University, Baltimore,
Md.; to Robeyt M. I'ieTl^a, 1701 Massachusetts Ave., N. W., Washington, D. C; or to any
other member of the editorial board.

Twenty-five reprints of articles will be furnished gratis to contributors when ordered in
advance. An order slip, with table of cost of additional reprints, is sent with proof.

The Journal of Comparative Psychology is published every other month, appearing
February, April, June, August, October and December. At present, one volume of approxi-
mately 500 pages is issued a year. Subscription is by the volume only, and not by the
year. The Behavior Monographs are issued irregularly.

Correspondence concerning business matters should be addressed to Williams &
WiLKiNS Company, Publishers of Scientific Journals and Books, Mount Royal and Guil-
ford Avenues, Baltimore, U. S. A.

Subscription Price: Current Volume; Volume II, United States and dependencies,
Mexico, Cuba, $5.00; Canada, $5.25; other countries, $5.50. These rates apply also to the
Behavior Monographs. Prices are net, postpaid. Foreign subscriptions to the current vol-
ume accepted at par of exchange.

Back Volume: Vol. I: Price, $6.00, United States, Mexico, Cuba; $6.25, Canada; $6.50
other countries. Prices are net, postpaid.

Back Volumes: Complete sets of Psychobiology (Volumes I and II), of the Journal of
Animal Behavior (Volumes I to VII, inclusive) and of the Behavior Monographs (volumes
I to IV), may be obtained from the publishers of the Journal of Comparative Psychology
at the price of $5.00, United States and dependencies, Mexico and Cuba; Canada, $5.25; other
countries, $5.50 per volume. Prices are net, postpaid. Foreign subscriptions to the back vol-
umes accepted at the current rate of exchange.

Single Volumes, when available, may be obtained at $5.00 a volume. The complete
sets are not broken to supply single volumes.

Subscriptions are received at the following addresses:

For Argentina and Uruguay: Beutelspacher & Cia,, Sarmiento 815, Buenos Aires,
Argentina.

For Australia: Angus & Robertson, Limited, Sidney; Stirling & Co,, 317 Collins St.,
Melbourne,

For Belgium: Henri Lamertin, 58 Rue Coudenberg, Bruxelles.

For the British Empire, except Australia and Canada : Order from your own agent,
or from any British bookseller, or order direct from the publishers.

For Canada: Wm. Dawson and Sons, Ltd., 87 Queen Street, East, Toronto, Canada.

For Denmark: H, Hagerup's Boghandel, Gothersgade 30, Kovenhavn.

For France: Emile Bougault, 48, Rue des Ecoles, Paris,

For Germany: R, Friedlander & Sohn, Buchhandiung, Carlstrasse 11, Berlin, NW. 6.

For Holland : Scheltema & Holkema, Rokin 74-76, Amsterdam, Holland.

For Japan and Korea: Maruzen Company, Ltd. (Maruzen-Kabushiki-Kaisha), 11 to 16
Nihonbashi Tori-Sanchome, Tokyo; Fukuoka, Osaka, Kyoto, and Sendai, Japan.

For Spain: Ruiz Hermanos, Plaza de Santa Ana, 13, Madrid.

For the United States and other countries except as above : Williams & Wil-
KiNs Company, Mount Royal and Guilford Avenues, Baltimore, U. S, A,

No claims for copies lost in the mails will be allowed unless such claims are received
within 30 days (90 days, foreign) of the date of issue. Claimants must state that the publi-
cation was not delivered at their recorded address. The publishers will not be responsible
for loss due to change of address unless notification is received at least one week in advance
of issue.

JUL 7 1922

Behavior Monographs

Volume 4. Number 4. 1922 Serial Number 20

Edited by
WALTER S. HUNTER

University of Kansas

Visual Perception of the Chick

BY

HAROLD C. BINGHAM

A dissertation submitted to the Board
of University Studies of the Johns Hop-
kins University in conformity with the
requirements for the degree of Doctor
of Philosophy.

Published by

WILLIAMS & WILKINS COMPANY

Mount Royal and Guilford Avenues

Baltimore, U. S. A.

CONTENTS

Part I. Historical

PAGE

Chapter I. Origin and history of the domestic fowl 1

Selected bibliography 5

Chapter II. Experimental literature on vision in the chick 7

Bibliography 20

Part II. Experimental

Chapter III. Apparatus 23

Chapter IV. Problem, method, and technique 31

Chapter V. Size perception 43

Chapter VI. Form perception 59

Chapter VII. Flicker perception 69

Chapter VIII. Problem learning 76

Chapter IX. Ideational behavior 96

Chapter X. General conclusions and summary 101

PREFACE

The observations on which the experimental portion of the
following pages is based were made upon four different groups
of Plymouth Rock chicks. Work with the first three groups
was conducted in the Harvard Psychological Laboratory during
1911-1912 when the visual factors in space perception described
in Chapters V and VI were studied. The data on flicker per-
ception reported in Chapter VII and the greater portion of the
material on problem learning, presented in Chapter VIII, were
gathered in the same laboratory during 1915-1916. The results
on relativity and specificity of reactions reported in Chapter
IX were obtained quite incidentally during the two years of
research.

In this Monograph there is presented an account of the re-
sults as written in 1916. A preliminary report had been pub-
lished in 1913 in the Journal of Animal Behavior, following
which there were several published comments relative to certain
questions raised in the paper. Cognizance of these discussions
and the related literature was taken at the time the manuscript
was prepared. Rather than isolate the unpublished portion,
it now seems advisable to present the entire study in the form
it had taken before the United States became a participant in
The World War. A revision of the manuscript to make it
strictly current is not deemed worth while.

The problem in spatial perception was proposed by Professor
Robert M. Yerkes, under whose guidance the entire study was
made. During the vear preceding: my initial work on this
problem, he completed in the Harvard Psychological Laboratory
the construction of the apparatus which has been described in
a Monograph^ on "Methods of studying vision in animals."
For the use of this apparatus, for his numerous suggestions in
method, for his precise formulation of problems, and above all
for his continued and sympathetic encouragement, I wish to

^ Behavior Monographs, Vol. 1, No. 2.

vi PREFACE

express sincere appreciation. My only contribution to the
apparatus Is the simple mechanism described In Chapter VII,
yet in the development of this device his helpful suggestions
always were at my command. Likewise, my only contribution
to the main problem of visual perception In the chick is that
represented by the preliminary data on the reactions to flicker
stimuli. Gladly I acknowledge my Indebtedness to Professor
Yerkes for his patience in starting me on this experimental task
and for his kindly Interest throughout the study.
' This opportunity is also taken to express my obligations to
Miss Doris M. Wood, who generously read the entire manu-
script, and to my wife, who assisted in various ways with the
preparation of the manuscript.

Harold C. Bingham.
Washington, D. C, January 25, 1922.

PART I. HISTORICAL

CHAPTER I

Origin and History of the Domestic Fowl

In the jungles of southern Asia lives a species of birds known,
from the nature of its habitat, as the Jungle Fowl. As the
name suggests, it is a wild type of Gallus, the comb fowl, to
which our domestic chick belongs. Several varieties of these
little \\\\d fowls are to be found in India and the Malay regions,
but all have general characteristics in common. Four groups
are generally recognized, namely, the Banklva Fowl, the Son-
nerat Fowl, the Ceylon Jungle Fowl, and the Fork-tailed Jungle
Fowl or Gangegar. These jungle fowls are commonly regarded
as the ancestors of all the present breeds of domestic fowl.

Gallus bankiva, the wild type which seems to be most closely
related to our domestic fowl, is found throughout India, in the
Malay Islands, and in the Philippines. It closely resembles the
domestic Black-Red Game Bantam. The feathers of the head
and neck, including the long hanging feathers at the back of the
head, possess a glistening golden hue. The golden hue of the
collar is matched with the long supra-tail feathers, but the
shorter straight tail feathers beneath are all black. The middle
feathers of the wings are a lively brown, the remainder, in the
male, being dark gray with pale borders. The iris is orange red,
the head dress red, the bill brown, and the foot slate black.
The body length reaches 56 cm., the wing length 22 cm., and
the tail length 27 cm.

The tail of the smaller female stands more erect, her comb
and wattles are barely noticeable, and her long neck feathers —
black with yellowish borders — add brownish black spots to the
mantle plumage. The dorsal side is cream colored. The tail
and supra-tail feathers are a brownish black.

Gallus sonnerati was the first of the wild breeds to become
known. It occurs in west, south, and central India. The long
slender feathers on the head, neck, and upper back of the male'

2 HAROLD C. BINGHAM

are black with gray edges and white tips; near the end they bear
an ornamental yellowish spot. The other feathers of the dorsal
side are narrowly striped with white. The ventral side of the
body is black, dotted with little white lines, and becomes almost
white at the median line.

Domestication of the Sonnerat Fowl is more difficult than
that of Bankiva. We are told that these two wild species will
cross but the hybrids are said to be sterile. The sterility of the
hybrids and the characteristic differences between tamed Ban-
kiva and Sonnerat fowls have frequently been urged as evidence
that the latter is not the ancestor of our domestic chick.

The Ceylon Jungle Fowl, Gallus lafayetti, closely resembles
Bankiva. The male is distinguished from Bankiva by the
orange-red breast, the long black-striped yellow neck feathers,
and the glistening blue-black throat and supra-tail feathers.
The slightly notched red comb, bearing on the lower part a
large oval yellow spot, is also distinctive. The Ceylon and
Bankiva females differ with respect to voice. According to a
report of the National Poultry Conference, Reading, England,
(cited by Houwink), Lafayetti has been crossed with common
fowls and fertile hybrids were obtained.

The extreme shyness of the Fork-tailed Gangegar, Gallus
varius, is a characteristic phase of its behavior indicating that
it would not be readily domesticated. At the least noise, it
hides in the alang-alang thickets, which makes close observation
of this fowl in its natural habitat very difficult. Were it not
for the male's noisy voice it would scarcely be noticed. Early
morning is the best time to observe it, for it then leaves the
thickets for the open places in search of food.

Gangegar seems to be very remotely related to the three
preceding classes. It differs from them in the number of supra-
tail feathers, the untoothed comb, and single neck lobe. It
also has certain characteristic differences in coloring.

Of these four groups of wild fowl, Bankiva is most generally
regarded as the ancestor of the common chick. Darwin, basing
his conclusions chiefly upon structural evidence, is inclined to
regard Bankiva as the original stock. "Having kept," he says,
"nearly all the English breeds of the fowl alive, having bred and
crossed them and examined their skeletons, it appears to me
almost certain that all are the descendants of the wild Indian

VISUAL PERCEPTION OF THE CHICK 3

fowl, Gallus bankiva." Others think that the jungle and do-
mestic fowls sprang from a common ancestor of intermediate
size which has since become extinct, but Houwink thinks it is
improbable that the domestic fowl has an extinct ancestor, for
it, like the known wild poultry races which have maintained
themselves in the midst of the oldest and most populous Indian
towns, surely have been preserved somewhere in Central Asia.

The behavior of Bankiva, as reported by certain naturalistic
observers, would lead one to regard it as the most probable
ancestor of the domestic fowl. Guillemard is cited in Brehm's
Tierleben to the effect that captive Bankiva fowls on the island
of Sulu are very tame and are easily crossed with domestic
fowls. In the same source, but in a different citation, it is
asserted that in the mountainous regions of Pegu, where the
Bankiva abound, they come into the villages and yards of the
natives where they mix with the tame fowls. Another citation
has it that crossing with the domestic chick is a common occur-
rence in Ceylon and that the male hybrids resulting are far more
spirited and have sharper spurs than the parent males.

At the present time there are domestic fowls which are very
similar in structure to the wild Bankiva. It has been found
that modifications in food and environment may cause changes
in the general form and color of poultry, if we may believe
Houwink on this point. It is rather resonable that the breeds
resembling Bankiva have descended from that origin. Natural
and scientific selection might readily account for the variations
and further development that have resulted.

From discoveries in mounds and other ancient settlements of
man, it appears that the chick has long been domesticated.
Along with the duck and the goose, the chick could have been
taken and held in captivity with comparative ease. On this
fact is based the assumption that the origin of the fowl's history
dates back to ancient times. Robinson deems it quite probable
that it was domesticated before any of the mammals. Houwink
on the contrary, reasons that it was not kept by man until
population took on a permanent aspect. Unlike the dog, its
domestication would have been neglected by the hunter. Such
animals as the cow and the horse would naturally be first tamed
by the nomads.

Turning to evidences which ancient literature furnishes, the

4 HAROLD C. BINGHAM

first reference to the domestic fowl is probably made in the great
Chinese Encyclopedia which was prepared about the fourteenth
or fifteenth century before Christ. This book suggests that
the chick was brought from the West and from India into China.
This would indicate that it is an old domestic animal with the
Chinese and Japanese. It is possible, of course, that this
account is based upon nothing more reliable than tradition.

The Egyptians are also said to have had domestic chickens
at an early date. It is thought that they were received from the
Orient. It is probable that the fowls were transported from
Egypt or Persia to southern Europe, Greece, and Rome. The
chick is not mentioned by the Old Testament, Homer, and
Hesiod, but it was to be found in Babylonia as early as the
seventh and sixth centuries before Christ. It was surely known
to the Greeks as early as the fifth century before Christ, for
Socrates, after taking the poison which caused his death, re-
minds Crito that he owes a cock to Asclepius and asks him to
pay the debt. Darwin is of the opinion that the domestic fowl
appeared in Europe about 600 B.C.

The early history of the chick is thus quite uncertain, despite
the fact that a number of students have given it their attention.
Numerous theories have been advanced, evidence more or less
convincing has been submitted in favor of them, yet the avail-
able facts are not conclusive. The evidence thus far furnished
does little more than provide a basis for theories, and in con-
sequence the statements which occur in the literature must be
accepted with caution. For the most part it seems probable
that Galliis bankiva is the common ancestor of the numerous
breeds of domestic chicks which we now possess. It is quite
reasonable to conclude that man caught the wild fowl, domes-
ticated and developed it for cock fighting, fortune telling, or
domestic use according to his peculiar interests. Chickens
used for fighting tended to remain fighters ; those accustomed to
domestic life tended to retain their utilitarian characteristics;
and from some of the deviations which have occasionally oc-
curred the numerous breeds w^hich we possess today have been
developed.

Thus the wild fowl, after its domestication, has been per-
manently adopted by man. In a general way, its distribution
over the earth seems to have followed the migrations of the

VISUAL PERCEPTION OF THE CHICK 5

Aryan peoples. The Chinese record, however, if its reliabiHty
be not questioned, suggests an important exception. At any
rate, the Hfe of the chick has been so long connected with human
activities that it has become quite dependent upon man. If
left to itself, it is no longer able, to maintain an existence.
Since becoming domesticated, it has changed in form and has
been bred toward gigantic or dwarfish structure according to
climatic conditions and the influence of natural and artificial
selection.

As a result of these variations, there were in the year 1911
over 160 breeds of domestic fowl. Five of these are recognized
as distinctly American breeds, of which the most typical is the
Barred Plymouth Rock. An accurate knowledge of the Plym-
outh Rock's blood heritage is impossible, for in the duplica-
tion of the original stocks and in the perfection of the breed
numerous elements were intermingled. It is generally believed
that the original stock was secured by crossing the American
Dominique with black Javas or Cochins. Light Brahma, Dark
Brahma, and Pit Game are said to have been used in its making.
This breed was briefly known as Plymouth Rock until the white
variety appeared, after which it became known as Barred Plym-
outh Rock.

This particular type of fowl appeared about the middle of the
19th century. Some of the more intelligent poultry breeders
conceived the idea of a breed of chicks particularly adapted to
American conditions. Accordingly, they set about developing
birds with the desired characteristics. As a result of this
activity there were exhibited in 1869 the first Plymouth Rocks,
These birds were so favorably received that they soon became
more numerous in America than all the other standard bred
birds combined. The observations reported in Part II were
made upon this strain of chick.

Selected References on the Origin and History of the Domestic Fowl

Brehm, A. Tierleben: DieVogcl. Bd. 7. 4th ed.

1911.
Darwin, C. The origin of the species.

1889.
Houwink, R. The history of domestic poultry. Assen (Holland). Also.

1911. De Hoenderrassen.

6 HAROLD C. BINGHAM

Howard, G. E. Standard varieties of chickens. U. S. Dept. Ag., Farmers'

1907. BuUetin, No. 51.
Plato. Phaedo, 117. Tr. Jouett, B. The works of Plato translated into

English. Vol. 2, p. 266.
Robinson, J. H. Principles and practice of poultry culture.

1911.
Thomas, J. L. Hybridization experiments with the Ceylon jungle fowl. Nat.

1907. Poultry Conf., Reading Off. Rpt., 2, pp. 98-115.

CHAPTER II

Experimental Literature on Vision in the Chick

During the course of the investigation which is reported in
Part II, the question was asked by a Harvard undergraduate:
Why is it that a chicken always wants to get on the other side
of the road when it sees an automobile? The question has
nothing in common with my investigation but it is suggestive
of a prevalent notion regarding the stupidity of the domestic
fowl. The chick seems to rank low in intelligence according to
unscientific observers. Eulogistic literature, in which anec-
dotes and remarkable stories abound, often contains accounts of
exceptional and almost incredible behavior of dogs, cats, horses,
parrots and the like, but in such collections of stories and inci-
dents, it is noteworthy that the chick has been generally
neglected.

One may incidentally hear of such stories as that of a hen
finding her w^ay through an open window to a bedroom in a farm
house and selecting the top of the old-fashioned bureau on which
to make a nest. In her efforts to collect all of the articles that
might be used in a hen's nest, she knocked from the bureau
a kerosene lamp and made use of a doily on which the lamp had
been standing. In this extraordinary nest, composed of pin-
cushions, ribbons, laces, doilies, and handerchiefs, she laid an egg.

The above anecdote was related to me by people who per-
sonally observed "biddy's stolen nest" and the results. Though
trivial to the extent of absurdity in science, it is nevertheless
typical of much that one may read about cats and dogs in the
eulogies of the anecdotists. Chicken anecodotes of this sort
seldom appear in print, but let one cat claw at the knob of a door
supposedly as a singnal .to be let out, and she becomes (see
Thorndike (30)) "the representative of the cat-mind in all the
books." It seems that no chicken has performed any act which
might furnish the aspiring eulogist with a representation of the
chick-mind.

Considering the apparent delight of the naturalists and semi-
experimentalists in repeating animal stories of an exceptional

8 HAROLD C. BINGHAM

type, it is rather surprising that so Httle chick behavior has
found its way into print. Apparently the animal anecdotists
saw in the behavior of the chicken only acts of stupidity, rather
than intelligence, and as a result of their eulogistic inclinations
they neglected to report such facts. The chick does not fill
the place of a household pet; hence the stock of chick stories like
that related above has generally been unknown to the reporters
of animal anecdotes.

Despite the tendency of Aristotle to report unusual facts
about animals, he has said very little about the chick. The
brief description of it is typical of his usual method of describing
animal behavior. He naively says :

Young hens are the first to lay, and they do so at the beginning of spring and
lay more eggs than the older hens, but the eggs of the younger hens are compara-
tively small. As a general rule, if hens get no brooding they pine and sicken.
After copulation hens shiver and shake themselves, and often kick rubbish about
all around them — and this, by the way, they do sometimes after laying. . .
(1).

With such generalization, Aristotle's description of chick be-
havior ends. His metaphysical interests did not encourage
intensive observations that would lead to a description of the
fowl's perceptions.

Offering little or nothing for the pure naturalists to eulogize,
therefore, chick behavior is not reported until the scientific
spirit spreads to the field of animal behavior. Naturally, then,
the first publications involving the chick are those of the semi-
experimentalists. Stimulated, as they must have been, by the
theory of natural selection, one might well expect these first
observations to be turned upon the instincts. From a consid-
eration of Inherited types of behavior, it Is a logical step to
individual modifiability. And the Investigation of habit forma-
tion, involving the sensory equipment of the organism, finally
leads to studies of perceptual discrimination.

Early reference to the chick's vision is incidentally made
by Spaulding (28), representing the semi-experimental group,
whose observations were primarily centered upon the nature of
the pecking and drinking reactions. In a general way, Preyer
(25) observes that the perfection of sight In newly hatched
fowls is astonishing when compared with the Imperfection of
this sense In newly born humans. The work of Romanes (26),

VISUAL PERCEPTION OF THE CHICK 9

Eimer (8), Mills (22), and Morgan (23) also primarily Involves
the instincts. There Is a tendency for this transitional group to
become controversial over the problem of Instinct and acquisi-
tion. No definite experimental data are produced to settle the
disputed questions. Mills loosely reports some of his observa-
tions In diary form. The diary method also appears a few years
later in reports by Hunt (12) and Kline (19) in each of which
certain visual discriminative ability is described.

Irritated by the controversies of this transition group and the
failure to answer on empirical grounds the disputed questions,
Thorndike (30) sets out on a procedure distinctively experi-
mental. However, he barely touches the problem of visual
acuity in the chick and, for the most part, his observations are
confined to crude studies of the chick's space and color per-
ceptions.

Nearly ten years later, Hess (9-11) takes up the problem of
the chick's color vision as well as light and dark adaptation.
The method of Hess involves for the first time the use of the
dark room for studying visual acuity in the chick. Although
he greatly Improves on the method of Thorndike and deserves
praise for introducing the controllable method of the dark room
which has later been highly refined, Hess nevertheless often
leaves the reader uncertain of the experimental conditions.
He should be credited with ingenuity and alertness, but his
method in certain details is too crude for the problems with
which he deals.

A study, certainly deserving more severe criticism than the
work of Hess, has been jointly reported by Katz and Revesz
(18). These authors have hastily considered a variety of
topics relative to the chick's visual discriminative ability.
Their crude "Klebmethode" has been applied to several phases
of the problem, including the topics of size and form perception
as well as color discrimination in "unsuppressed daylight."

In its simpler form, the "Klebmethode" of Katz and Revesz
consists In pasting on a cardboard kernels of grain at which the
chick's pecking response Is to become Inhibited. Among these
"glued" kernels is scattered a different kind of grain which the
bird Is allowed to pick up. To Illustrate, It was found by the
experimenters that the chicks preferred rice to wheat. The
bird was classed as "Fehlerfrei" when It bad picked up the loose

10 HAROLD C. BINGHAM

wheat without pecking at the glued rice. The observers had
two quantitative measurements for the rapidity of learning:
(1) The number of series necessary for a perfect habit; (2) the
number of reactions to rice. To make sure that the errorless
chicks did not avoid the rice by means of the glue which might
be visible on the pasted kernels, the rice was scattered loosely
upon the cardboard in the same manner as the wheat. The
wheat w^as again eaten by the errorless chicks and the rice was
left. The discrimination, the authors concluded, was between
wheat and rice.

Having found that the chick could pick up wheat and avoid
rice, Katz and Revesz sought an answer to the question: By
what means does the chick discriminate the grains? Accord-
ingly they turned to a study of size and form discrimination.
Because the chicks could be trained to eat only half grains of
rice when scattered among whole grains, the authors were con-
vinced that there was discrimination of size and form. Further
observation leads them to conclude that the chick discriminates
squares and triangles. They say:

Knowledge of this fact we secured through a variation of the experimental pro-
cedure. Out of green peas (some of which had been readily eaten), we cut three
and four cornered pieces. On account of their moisture, they could not be glued
down, so we laid the four sided pieces upon a glass plate and the three sided pieces
under it. The chick found that the three sided pieces could not be reached and
soon ceased to peck at them. Then if we laid both forms upon the glass plate,
only the squares were picked up. By means of the same method we found that
the chick discriminates between triangles and circles as well -as squares and
triangles.

The report of Katz and Revesz leaves the reader too often
uncertain of the experimental conditions. They have tried to
do too much and have not treated any one task intensively.
Their report indicates carelessness and indifference to details.
No doubt the assertion that the chick discriminates squares and
triangles as well as triangles and circles is true for the conditions
under which the experiment was made. But from the written
account, one cannot tell what the conditions were. One does
not know, for example, the relative sizes of the different forms.
Were they equal in area? Was the square an inscription or a
circumscription of the circle? Was the diameter of the circle
equal to the side of the square or the altitude of the triangle?

VISUAL PERCEPTION OF THE CHICK 11

Was the third dimension constant? These are some of the
factors which must be known before one can safely say that the
chick perceives form. One could truthfully state that an ani-
mal discriminates between a circle and a triangle if it regularly
chooses the former even though the triangle be twice as large
as the circle, but it is improbable that the basis of discrimina-
tion in such a case would be anything other than size. Since
such points are ignored in the report by Katz and Revesz, one
suspects of the experimenters carelessness in technique.

Furthermore, no information is offered as to the method of
cutting these forms. It is a difficult matter to cut green peas
with regularity. Can we be certain that the chicks discrimi-
nated on the basis of form rather than irregularity of surface?
No mention is made of controls to eliminate this possibility.
Because the authors leave vital points like these unmentioned
and apparently unnoticed, one is inclined to regard the experi-
ment as quite superficial.

A definite answer to the controversy of the semi-experimental
school over the nature of the chick's instinctive reactions is
offered for the first time by Breed (4).^ His study of the in-
stincts is supplemented by his study of certain habits, out of
which grew the problem of accurately determining the nature
of the color, form, and size stimuli in response to which a chick
is able to acquire a habit of reaction (5). The chief significance
of Breed's later study appears in his contribution to the de-
velopment of the dark-room method. For testing the chick's
perception of colors. Breed combines the Yerkes discrimination
method (34) with the Hess (9-11) dark-room method. The
next step is the adaptation of the combined methods to investi-
gations of size and form perception. The stimuli were presented
to the chicks by means of the illumination of two plates of opal
flashed glass over which were set mats of tin or cardboard con-
taining the desired openings. Thus the chick was given an
opportunity of selecting a circle when appearing with a square,
or of discriminating a larger from a smaller stimulus.

^Certain instincts have later been carefully observed by Pearl (24). The
instincts considered relate primarily to behavior having practical significance
for the poultry keeper. In the single reference cited at the close of this chapter,
there appears a selected bibliography referring to a number of other papers on
related subjects.

12 HAROLD C. BINGHAM

For testing size discrimination, two circular areas — one 5 cm.
in diameter, the other 8 cm. — were presented to the chick.
Evidence of a perfect habit appeared at approximately two
hundred trials. Controls were introduced throughout the
training to eliminate "difference in brightness of the lighted
areas" and "difference in brightness of the right and left sides
of the experiment box." In spite of ambiguity in the use of
"brightness" Breed's conception is fairly obvious.

In the study of form perception. Breed used three chicks, one
of which, he declares, "learned to discriminate two optical
stimuli on the basis of difference of form." The forms were a
circle and a square equal in area. Because his studies of the
reactions of other chicks to similar stimuli yielded negative re-
sults, he attributes the positive reactions of this particular
chick to fortune in the choice of animal. In this connection, it
is significant that the chick was credited with a perfect record
in the fifth training series or after only forty trials. A perfect
habit is clearly evident after one hundred trials. For space
stimuli differing in form alone this rate of habit formation ap-
pears, in the light of later results, unusually rapid and suggests
that the successful chick may have discovered some easy cue
other than the form difference. This possibility suggests it-
self rather forcefully when one considers that Breed was using
a mechanism only partially standardized; it rises to the level
of probability w^hen one notes that he does not report controls
for varying the relative positions of the stimuli within the slid-
ing frame.^

It is unfortunate that Breed also failed to make control tests
to determine whether the distribution of light on the chick's
retina was the basis of discrimination. Such a control is easily
made in the case of a triangle by inverting the stimulus. The
inversion of a square would cause no change in the distribution
of light, but such a change might be produced by turning the
square through 45 degrees. Even had there been no possibility
of an unsuspected cue furnishing the chick with a discriminable
factor, there remained this second possibility which Breed did
not consider.

2 A discussion of controls in which this possibiUty was found to be a reality
appears in the Journal of Animal Behavior, 1913, vol. 3, pp. 86-90.

VISUAL PERCEPTION OF THE CHICK 13

A study of the relation of strength of stimulus to rate of
learning in the chick, though subsequent to Breed's work, is
reported by Cole (7) prior to Breed's account (5) of his dark
room work. Cole presented brightness stimuli to his chicks,
using the same apparatus as Breed.

A further stage in the development of the combined "dark-
room-discrimination" method is indicated in a joint report by
Yerkesand Watson (35), on methods of studying vision. They
have greatly refined the apparatus used by Breed and Cole and
limited its use to studying light vision. In addition, they have
developed a special apparatus for studying color vision.

The first use of the Yerkes light vision apparatus with chicks
is my study of size and form perception (2). The aim of this
study was to locate thresholds for standardized forms and sizes.
With a standard circle 6 cm. in diameter, the threshold was
found to be one-fourth to one-sixth. But instead of determining
a limit in form perception, the question became chiefly whether
the chick perceives forms. It was found that the chick could
discriminate between a circle and an upright triangle of equal
area, but the evidence of discrimination disappeared when the
triangle was inverted. Since Breed failed to make a similar
control, his statement that one chick "learned to discriminate
two optical stimuli on the basis of form" may be questioned.
A safer conclusion seems to be that the chick reacts to unequal
stimulation of different parts of the retina.

With reference to this interpretation of the reactions to form
stimuli, Watson {?>2>) says:

Bingham has raised the whole question as to what is meant by form.'
. . . • . While the chick can apparently respond to the difference in form be-
tween the circle and the square, and the circle and the triangle, when they are
equal in area, yet such responses, after all, are really nothing more than keen
perception of size differences. He draws this conclusion from the fact that
after the chick has learned the circle-triangle habit with the base of the triangle
down, the habit will disintegrate if the apex is placed down. (p. 366).

Quite naturally, this interpretation of form perception has
called forth considerable comment. Hunter's contribution (13)

3 For this suggestion I hasten to acknowledge my indebtedness to Professor
Yerkes who proposed this question as a part of my experimental task. Because
he conceived the possibility, however, it does not follow that he agrees with my
view. All responsibility for subsequent use of the conception as an explanation of
the facts, I alone assume.

14 HAROLD C. BINGHAM

to the discussion which has ensued consists in calling attention
to the need of sharper distinction between the study of form
discrimination and pattern discrimination. He argues that we
should not expect the child to discriminate forms in the sense
of my definition. He theorizes that infrahuman animals "have
only a more or less crude pattern vision." As a means of test-
ing the validity of his theory he proposes that the surroundings
of the discriminable forms be changed, since the form, seen
with its surroundings, must be considered as part of a pattern.
Even if no other objects are in the visual field the stimulus "is
seen surrounded by the more or less irregular outline of the
field of vision, and so is again part of a pattern."

To demonstrate experimentally whether the animal is react-
ing to forms or patterns. Hunter proposes that the electric
chambers of the experiment box be fitted, after a perfect habit
has been established, with hollow cylinders or hollow triangular
prisms through which the stimuli may be seen by the reacting
subject. He presents diagrams which, he declares, represent
the discrimination situation in my experiment and in Lashley's
study (20) of form perception in the rat.

Both series of, experiments are concerned with patterns not forms

In problem boxes such as those described by Lashley and Bingham ....
the animal tested is confronted not by two "forms'^ corresponding to the configura-
tions of the opal glass, but by such designs as are suggested in figure 1. The
squares drawn in the figure represent the rectangular tunnels down which the
animal goes in making his responses. What the animal sees is a triangle or a
circle each in more or less of a square setting.

Designs 4 and 5 represent Laskley's "forms."

Johnson (14) points out that Hunter's proposed method of
control would introduce new olfactory stimuli and probably
new tactile stimuli. Is there any means of deciding, he asks,
whether failure to discriminate, after making such a change,
resulted from the change of pattern or from the simultaneous
introduction of other novelties? Furthermore, Johnson be-
lieves the change of a form stimulus from the right to the left
compartment of the experiment box actually changes the back-
ground and foreground. This would therefore make the pat-
tern a variable and the form a constant.

In addition to Johnson's reply to Hunter's contention, I have
pointed out (3) the impossibility of pattern, in Hunter's sense,

VISUAL PERCEPTION OF THE CHICK

15

serving as the discriminable factor. The conditions of control
leave the rectangular tunnel constant but wholly change, if not
destroy, the perceptibility of the environment.

That the animals could not see the environment is attested by the fact that they
were frequently observed to walk bHndly into the confining walls. Not all of the
time was the environment "darkened," but the control tests were always made to
determine whether, among other factors, setting was a factor in discrimination.

Figure 1, as presented by Hunter, does not accurately illus-
trate the condition of the stimulus areas. With the intro-
duction of a screen as a means of control between the general
illumination and the electric boxes and with the reduction of
the intensity of the source lights, a similar condition to that
illustrated in figure 2 appears. In the compartment where the
triangle appears, the source light fails to illuminate the corners

A

O

V

Fig. 1.
Reprinted from Jour. Animal BeJmvior, vol. 3, no. 5, p. 331.

of the tunnel, and so the perceptible portion of the setting
changes to a sort of circular form as in diagram 1. About the
circular stimulus, the visible setting is more nearly a perfect
circle as is shown in diagram 2. Even if my apparatus offered
a possibility of pattern discrimination, my plan of control
(2 : p. 98) would have made so variable the patterns confronting
the animal that they never could have served as a basis of
discrimination.

Hunter has apparently overlooked one of the essential fea-
tures of the apparatus used in my study. The dark-room ap-
paratus allows the experimenter to control the visibility of sur-
roundings by means of artificial illumination. His criticism

16 HAROLD C. BINGHAM

would be valid for similar experiments conducted in natural and
uncontrolled light. With the discrimination of stimuli con-
tinued under a varied visibility of surroundings the perception
could not have been of patterns.

In a brief review, Washburn (32) dissents from my interpre-
tation of form perception, declaring that she would conclude
that the chick is not possessed of an abstract idea of
triangularity. She says:

A triangle with apex up is a different form from a triangle with apex down:
the two have in common only the abstract quaHty of three-sideness. The per-
ception of form, as distinct from an abstract idea of form, is based precisely on
the unequal stimulation of different parts of the retina.

A portion of my later paper involves this lack of agreement
in defining form (3). In that paper it is maintained that my

Fig. 2
Reprinted from Journal Animal Behavior, vol. 4, p. 137.

conception of form is in keeping with the ordinary usage of the
term, and that we should not depart from the usual meaning
in an effort to have it included in an animal's stock of percep-
tual experiences. If we find that our animals have a power of
discrimination which approaches form perception, but which
Is not form perception in the strict sense of the term, we
should adopt a terminology to fit the special case; we should
not attempt to enlarge the scope of the old term to cover the
special case.

Perhaps a more or less crude pattern vision is the nearest
approach to form perception that animals possess. At any
rate, Hunter has done well in calling attention to the distinction
between patterns and forms. But our definition should not
stop here. Two forms may be identical, but even in the ab-

VISUAL PERCEPTION OF THE CHICK 17

sence of an orienting background or environment (pattern), as
in the dark room studies, there may yet be a fundamental dif-
ference. This difference has been arbitrarily called "shape."
An illustration of form similarities having shape differences is
furnished in the experimental setting of Lashley's study. He
used two identical forms in that both were rectangles 2 mm. by
60 mm. In his use, however, they differ in this respect: one is
extended laterally thirty times as far as its vertical extension;
the other is extended vertically thirty times longer than later-
ally. This has been defined as a difference in shape of two
identical forms.

It is not to be denied that a triangle with vertex up differs
from a triangle with vertex down, but one can scarcely say that
they are different forms. They are both triangles, and more;
they are equilateral triangles. Obviously it is desirable to ad-
here as far as possible to an explanation of the difference between
these patternless stimuli in perceptual terms instead of resorting
to ideational content. It seems useful to employ a term like
"shape," therefore, which may differ in forms that are other-
wise identical. When the extended base of the triangle is so
placed as to stimulate the region of the retina which was for-
merly stimulated by the vertex of the triangle, a condition oc-
curs similar to that pointed out regarding Lashley's forms: the
forms remain identical, but the lines of maximum and minimum
extension become interchanged. This fact led me to conclude
in my paper (2: p. 110) that the apparent reactions to form are
the result of keen perception of size differences. It seems prob-
able that they are due to the perception of these so-called shape
differences. The inversion of the triangle causes certain partic-
ular size changes — vertex or point interchanged with base or
line — which introduces a change in shape, but no general change
of size since the area remains constant. Similarly, the factor of
triangularity remains constant and the form is unchanged. Not
"the perception of form," therefore, but the perception of shape
"is based precisely on the unequal stimulation of different parts
of the retina."

Our definition, then, as separate from the distinction between
forms and patterns, should distinguish between forms and
shapes. Referring to the retinal area stimulated, there is form
which is general, for example a triangle. But there is a partic-

18 HAROLD C. BINGHAM

ular feature about this general distribution of light — it is equi-
lateral, or isosceles, or right angled — which is called shape.
Forms are identical when their areas are equal and their general
retinal distribution is similar. Shapes of forms are identical
when all extension of the identical forms are equal and in cor-
responding directions. Thus, the area remaining constant,
either or both form and shape may change. The form remain-
ing constant, the shape may change. Change in form must
always be accompanied by change in shape.

Unquestionably the test of form perception by inversion of
the triangle is a severe one. But if this test of inversion is
found to interrupt the discrimination, form perception in the
strict sense of the term can scarcely be said to prevail. More-
over, there is evidence that form perception regarded not as an
abstract idea but as perceptual phenomena does not exist.

More in line with Washburn's inclination to regard my inter-
pretation of form as "an abstract idea of triangularity" (32) is
my incidental observation (2) of reactions to relative stimulus
difference. Because a few chicks showed ability to carry the
correct large-small reactions from a 6-4 training to a 4-3 or
9-6 test situation, it would seem that the chick has some sort
of ideational content. Watson, having been informed by John-
son that the latter failed to secure positive evidence in this
relative stimulus problem with a single adult game bantam,
hastily remarks (33): "From experiments now in progress at
Nela Physical Laboratory (Johnson) it would seem that this
observation cannot be confirmed." Johnson, however, is more
conservative and disclaims Watson's unguarded remark by sug-
gesting that the variation is due to individual differences partic-
ularly because it was found that not all of my subjects reacted
positively to this problem. Moreover, I found only one chick
that reacted positively and perfectly to both of the unfamiliar
combinations. Doubtless, the amount of training has much to
do with the nature of the reaction.

One of the best controlled studies employing the dark room-
discrimination method is reported by Johnson (15-16). His
report, appearing as two papers, is on the detail vision of the
dog, the monkey, and the chick. The first paper deals with
standardization of method in which he presents certain improve-
ments upon the Yerkes-Watson apparatus. He commendably

VISUAL PERCEPTION OF THE CHICK 19

points out that my discussion of controls (2) is ambiguous
through my synonymous use of the words "intensity" and
"brightness," The luminous intensity of a source, he declares,
involves the total number of candles presented, but the bright-
ness is the luminous intensity divided by the area of the source.
The brightness, then, is measured in terms of candles per unit
of area. In the light of this definition, Breed's use of the term
"brightness" (5) is also open to criticism.

Regarding Johnson's assertion (15) that I reported "no con-
trol tests to show that discrimination was on the basis of size,
rather than of luminous intensity of the stimuli," it is sufficient
to note that variation of source distances would vary both
brightness and luminosity. Independent variation of these two
factors would be unnecessary so long as neither was allowed to
remain constant. This control is adequately described in my
report (2) on page 98.

Johnson's method consists in presenting striate fields for dis-
crimination, adopting the visual angle subtended by one of the
striae as a convenient measure of the animal's visual acuity.
In terms of the visual angle subtended by individual striae, the
stimulus threshold for the chick was found to be slightly above
4'. The standard individual striae were about 0.11 mm. in
width, the assumption being that, at the given distance of 60
cm., "they were too small to be resolved by the eye." The
minimum size of the variable individual striae discriminated
from the standard striae was accepted for chick 1 as 0.710 mm.
and for chick 2 as 0.743 mm. (16).

The refined dark room-color apparatus is used by Lashley (21)
in studying the spectrum of the chick. Preliminary to his study
of wave length and without using punishment as a motive, he
finds that the brightness threshold is approximately three to one.
In contrast with earlier work, this difference limen is unusually
large. Aside from certain trivial errors in computations, Lash-
ley's work seems to have been carefully done. He thinks that
field experimenters may feel confident that differential reactions
of birds to colored objects are made on the basis of wave length
if the objects do not differ enormously in brightness for the
experimenter. He concludes that "the chick can distinguish
between monochromatic lights of any intensity between thresh-
old and the Pfund standard, irrespective of the brightness or
saturation. The effective stimulus is the wave length."

20 HAROLD C. BINGHAM

Finally, the latest paper dealing with chick reactions is by
Fletcher, Cowan, and Arlitt (8a) / They have made compara-
tive observations of chicks hatched, from normal eggs, from
alcholized eggs, from eggs with distilled water inserted, and
from eggs merely pierced and sealed. Inherited and acquired
reactions of these four groups were observed after the manner of
Breed (4-5). They have contributed nothing to the topic of
visual acuity.

In addition to this chick literature, studies in bird vision have
been made by Porter (24a-24b) on the sparrow, cowbird, and
pigeon; by Tugman (31) on the sparrow; and by Coburn (6) on
the crow. In technique, the work of Porter is similar to that of
Katz and Revesz (17-18) but it reflects less ingenuity than the
German papers. However, Porter's work was done a little
earlier than that of Katz and Revesz and he seems to have been
better acquainted with the historical setting of his task. Co-
burn's and Tugman's studies have made use of recent available
developments in the field of infra-human vision. These papers
will receive further consideration in subsequent chapters where
the results will be compared with my own results from the
chick.

Bibliography of Publications to 1916

1. Aristotle: De Anima.

2. Bingham, H. C. Size and form perception in Gallus domesticus. Journal

1913 Animal Behavior , vol. 3, pp. 65-113.

3. . A definition of form. Ibid., vol. 4, pp. 136-141.

1914.

4. Breed, F. S. The development of certain instincts and habits in chicks.

1911. Behavior Monographs, vol. 1, no. 1, Pp. Ill + 78.

5. . Reactions of chicks to optical stimuli. Journal Animal Behavior,

1912. vol. 2, pp. 280-295.

6. Coburn, C. A. The behavior of the crow, Corvus americanus, Aud. Journal

1914. Animal Behavior, vol. 4, pp. 185-201.

7. Cole, L. W. The relation of strength of stimulus to rate of learning in the

1911. chick. Journal Animal Behavior, vol. 1, pp. 111-124.

8. Eimer, G. H. T. Organic evolution. Tr. London.

1890.
8a. Fletcher, J. M., Cowan, E. A., and Arlett, A. H. Experiments on the
1916. behavior of chicks hatched from alcoholized eggs. Journal Animal
Behavior, vol. 6, pp. 103-137.

9. Hess, C. Ueber Dunkeladaptation und Sehpurpur bei Hiihnern und Tauben.

1907. Arch. f. Augenheilk., Bd. 57, S. 298-316.

* This manuscript was prepared in 1916.

VISUAL PERCEPTION OF THE CHICK 21

10. Untersuchungen iiber den Lichtsinn und Farbensinn bei Tagvoglen.

1907. Ihid., Bd. 57, S. 317-327.

11. . Untersuchungen uber das Sehen und iiber die pupillenreaktion von

1908. Tag und von Nachtvogeln. Ibid., Bd. 59, S. 143-167.

12. Hunt, H. E. Observations on newly hatched chicks. American Journal

1897. Psychology, vol. 9, pp. 125-137.

13. Hunter, W. S. The question of form perception. Journal Animal Be-

1913. havior, vol. 3, pp. 329-333.

14. Johnson, H. M. Hunter on the question of form perception in animals.

1914. Journal Animal Behavior, vol. 4, pp. 134-135.

15. . Visual pattern-discrimination in the vertebrates — I. Problems and

1914. methods. Journal Animal Behavior, vol. 4, pp. 319-339.

16. . Visual pattern-discrimination in the vertebrates — II. Comparative

1914. visual acuitv in the dog, the monkev, and the chick. Ibid., pp. 340-

361.

17. Katz, D. and Revesz, G. Ein Beitrag zur Kentnis des Lichtsinns der

1907. Hiihner. Nachr. d. k. Ges. d. Wiss. zu Gottingen, Math.-physik.

Kl. S. 406-409. (Cited by Lashley).

18. . Experimentelle-psychologische Untersuchungen mit Hiihnern. Zeit.

1909. /. Psych, u. Physiol, d. Sinnesorgane, Bd. 50, S. 93-116.

19. Kline, L. W. Methods in animal psychology. American Journal Psychol-

1899. ogy, vol. 10, pp. 265-277.

20. Lashley, K. S. Visual discrimination of size and form in the rat. Journal

1912. Animal Behavior, vol. 2. pp. 210 ff.

21. . The color vision of birds. I. The spectrum of the domestic fowl.

1916. /Z)zJ.,voL6,pp. 1-26.

22. Mills, W. The nature and development of animal intelligence. New York

1898. and London. Pp. XII -j- 307.

23. Morgan, C. L. Habit and instinct. New York and London. Pp.351.

1896.

24. Pearl, R. Studies on the physiology of reproduction in the domestic fowl.

1914. VII. Data regarding the brooding instinct in relation to egg produc-
tion. Journal Animal Behavior, vol. 4, pp 266-288.
24a. Porter, J. P. A preliminary study of the psychology of the English

1904. sparrow. American Journal Psychology, vol. 15, pp. 313-346.
24b. . Further studv of the English sparrow and other birds. Ibid., vol.

1906. 17, pp. 248-271'.

25. Preyer, W. The senses and the will. Tr. New York. Pp. XXV -1- 346.

1890.

26. Romanes, G. J. Mental evolution in animals. New York. Pp. 411.

1884.

27. Shepard, J. F., AND Breed, F. S. Maturation and use in the development of

1913. an instinct. Journal Animal Behavior, vol. 3, pp. 274-285.

28. Spaulding, D. A. Instinct. With original observations on young animals.

1873. Macmillan's Magazine, vol. 27, pp. 282-293.

29. — . Ibid. Reprinted, Popular Science Monthly, vol. 61, pp. 126 ff.

1902.

22 HAROLD C. BINGHAM

30. Thorndike, E. L. Animal intelligence. New York. Pp. VII + 297.

1911.

31. TuGMAN, E. F. Light discrimination in the English sparrow. Journal

1914. Aninml Bclmvior, vol. 4, pp. 77-109.

32. Washburn, M. F. Recent literature on the behavior of vertebrates. Psy-

1913. chologkal Bulletin, vol. 10, p. 320.

33. Watson, J. B. Behavior. An introduction to comparative psychology.

1914. New York. Pp. XII + 439.

34. Yerkes, R. M. The dancing mouse. New York. Pp. XXI + 290.

1907.

35. Yerkes, R. M., and Watson, J. B. Methods of studying vision in animals.

1911. Belmvior Monographs, vol. 1, no. 2, Pp. IV + 90.

PART II. EXPERIMENTAL

CHAPTER III

Apparatus

The work on visual acuity, as reviewed in the preceding chap-
ter, has been criticised primarily on the basis of method. One
general fact stands out when these experiments on visual per-
ception in the chick are viewed in their historical setting: the
methodological aspect has rapidly come to be the central feature
of the work. Beginning with loose observations in which ac-
curate description of method was either impossible or ignored,
the method has developed into a thoroughgoing scientific pro-
cedure. The early naturalistic observers raised questions for
the later scientific school to solve. The naturalists have con-
tributed a general orientation ; the experimentalists are solving,
one by one, the specific tasks involved. The apparatus used in
the present experimental study represents an important stage
in the development of technique for stud^dng detail vision.
Controllability w^as the primary aim in its construction. It
represents a definite standardization which has been previously
described in detail, hence my description will be a rapid sum-
mary. Supplementary to apparatus, but no less important,
is experimental method and technique to which a special chap-
ter will be devoted.

The dark room-discrimination apparatus has been carefully
described in the report by Yerkes and Watson. In Chapter II,
the origin and development of the method has been set forth.
In figure 3 is presented an isometric view of the apparatus
showing the skeleton of the various parts when assembled for
work. That part of the mechanism labelled I is the experiment
box. The opposite end. III, is the light or source box. Between
the experiment box dindthesourc^hoKis the stimulus shifter, II.

1. Experiment box

An illustration of the experiment box appears as figure 4.
This section of the apparatus is made of one-half inch lumber,

24

HAROLD C. BINGHAM

except where stated otherwise, and is painted within and with-
out a dead black. It consists of four maih parts: (1) A is an
entrance chamber/ 20 by 15| by 22. The floor of A is provided
with a metal tray containing a layer of wet felt. The entrance

Fig. 3
Isometric view of light vision apparatus. Reprinted from Journal Animal
Behavior, vol. 3, p. 67.

box may be removed by raising out of the iron straps, C and C,
the board to which it is attached. (2) 5 is a discrimination

^ Unless otherwise stated, dimensions are given in centimeters and the order of
presentation is length, width, and depth — inside measurements.

VISUAL PERCEPTION OF THE CHICK

25

chamber, 26 by 51 by 22. Leading from A to ^ Is an entrance,
/, 8 by 10. The floor of B, Hke that oi A, is carpeted with wet
felt. The tray of either compartment can be easily removed
and cleaned. (3) W and W are electric boxes separated from
each other by the partition D. On the floors of W and W is a
slab of slate carrying electric wires for punishment. By means
of the interposed sides, 0 0 0, the electric compartments are

Fig. 4
Experiment box as seen from above.
Behavior, vol. 3, p. 69.

Reprinted from Journal Animal

set back 14 cm. from the stim.ulus shifter. The floor of this
extended portion drops down about 10 cm. below that of W-W .
The middle 0 presses closely against the shifter so that the
exposed stimuli are sharply separated from each other. On the
end of D-0 is glued a piece of piano felt, P, which rubs snugly
against the shifter. (4) N and N' are nest boxes, 40^ by 22
by 22. Each nest box is covered by a tightly fitting lid, which is

26 HAROLD C. BINGHAM

hinged at the outside, and is equipped with a 2 c.p. frosted
electric lamp, a watch glass for water, and sand or litter in which
food may be scattered. It is also provided on the outer side
with hidden holes for ventilation.

Between W and iV is a vertically sliding door, E closed and
E' open, 3 mm. thick which fills an opening 8 by 10. Suspended
from an upright frame, F, is a coiled spring, S, which passes
through the walls of the experiment box to the top of E. When
E is closed, 5 is extended and in a state of torsion, hence when
E is released 6* tends to return to a state of rest and the doorway
is opened.

The method of closing the exits is best illustrated in figure 3.
A silk line, e, passing up through a wire loop directly under E,
is attached to the lower part of the door. By pulling on e,
the experimenter can close E which automatically locks when
it is closed. By stepping on or striking T, the chick can release
the lock. The experimenter, however, by catching e on the
hook, h, can prevent the exit from opening when the lock has
been released. This automatic release was devised by Professor
F. S. Breed.

The construction of this mechanical device for locking and
opening E appears in figure 5 where T is the trip that was seen
in the other figures and t is a short piece of No. 30 black thread
attached to T a.t a. Passing down through the floor, F, of the
electric box, t runs over a pulley, P, and is fastened to the spring
lock, L, at b. B is a. block through which L passes and against
which one end of the spring, C, presses. Z) is a stop attached
to L supporting the opposite end of C. This spring tends to
force L in the direction of X and to keep T in the position as
illustrated. But when sufficient pressure is applied to T at any
point above R, L is forced back in spite of the pressure of C,
E is released at X, and the recoil of 5 raises £. When E is
drawn down, L, by reason of the slanting end surface at X, is
forced toward P until the notch of E has passed below the lower
side of L when the constant pressure of C toward X forces L
into the notch of E and the door is again locked.

In several respects, figure 5 is a poor representation of this
tripping device. L is shown as constructed solidly in B, when,
in reality, it slides freely through B. Moreover, L is shown in
the figure as wide as the floor of the experiment box, but it is

VISUAL PERCEPTION OF THE CHICK

27

really a delicate latch narrower than the width of T or E and is
released by a slight touch upon T.

2. Source box

To provide illumination for the two stimuli simultaneously
presented to the chick, the source box represented in figure 6,

w

Fig. 5
Automatic tripping device of the experiment box. Reprinted from Journal
Animal Behavior, vol. 3, p. 70.

— 1 1 1 of figure 3 , — was used . A complete description of this part
of the mechanism appears in Behavior Monographs from which
figure 6 is reprinted.

28 HAROLD C. BINGHAM

3. Stimulus shifter

Figure 6 also presents a general view of the stimulus shifter,
the primary function of which is control of the various visual
details. The stimulus shifter or adapter is also described in
Behavior Monographs along with the description of the source
box.

A considerable number of stimulus plates, used for varying the
size and form of the stimuli, is required for determining the
chick's threshold for these spatial details. The majority of the
plates used in this study is included in table 1. Owing to varia-
tions in sensitiveness among different subjects, a set which meets
the requirements for one animal does not suffice for another,
hence, in those cases where the size difference among the plates
varies by one millimeter, a safe margin has been allowed by
enumerating a few more plates than would ordinarily be
required.

4. Accessories

In my report from which figures 3-5 are reprinted, there
appears, on page 73 and following, a description of the details
designed to aid in the control and manipulation of the mechan-
ism just described. The punishing device and the upper illu-
mination are also described. It is there shown how the experi-
menter from his position for observation, can control by means
of ropes, pulleys, and weights the entire apparatus.

VISUAL PERCEPTION OF THE CHICK

29

&h

® c/

i' nl

. r

^-h>i

e

0 C/'

/

Fig. 6
Reprinted irom Behavior Monographs, 1911. vol. 1, no. 2, p. 18. "Perspective of
light or 'brightness' apparatus. A, light box; C, D, compartments oi A; B,
partition between C and D;E,F, lids of A ; G, H, metal carriages carrying tungsten
lamps; // and KL, tracks for G and H; M, N, Starrett steel millimeter tapes;
0, P, apertures covered by Aubert diaphragms; R, Bausch and Lomb cooling
cell in light box; d, d', metal straps; y, aluminum plate sliding between d and
d'; T, tracks for y; V, stop for y; z, steel plate bolted to wooden end of light box;
h, screws attaching d to s; s, s, standard brass stimulus plates; p, brass frame
about aperture in y;r, hard rubber ring screwed to p."

TABLE 1

Stwndus Plates

NUMBER

USE

FORM

DIAMETER OR SIDE

AREA

PLATES
NEEDED

To test Size

cm.

sq. cm.

Perception

Circle

1.8

2.5447

<(

it

1.9

2.8353

It

it

2.0

3.1416

It

"

2.1

3.4636

it

it

2.2

3.8013

It

it

2.3

4.1548

<(

it

2.4

4.5239

It

it

2.5

4.9088

<(

a

3.0 (standard)

7.0686

tt

a

3,5

9.6211

it

it

4.0

12.5664

tt

it

4.1

13.2026

ti

tt

4.2

13.8545

tt

a

4.3

14.5220

tt

it

4.4

15,2053

it

a

4.5

15.9043

it

a

4.6

16,6191

it

it

4.7

17.3495

it

it

4.8

18,0956

It

a

4.9

18.8575

it

tt

5.0

19.6350

tt

it

5.5

23.7583

tt

a

6.0 (standard)

28.2744

tt

it

6.1

29.2247

it

a

6,2

30.1908

It

it

6.3

31.1725

tt

it

6:4

32.1700

it

it

6.5

33.1832

it

"

6.6

34.2120

it

it

6.7

35.2566

it

it

6.8

36.3169

a
tt

it

6.9
7.0

37.3929
38.4846

'<

it

7.1

39.5920

a

it

7,2

40.7151

it

it

7,3

41.8540

it

it

7,4

43,0085

tt

a

7,5

44 1787

it

it

8,0

50,2656

it

it

8,5

56,7455

tt

tt

9,0 (standard)

63,6174

2

44

To test Form

cm.

sq. cm.

Perception

Circle

6,0 (as above)

28,2744

—

"

"

5.9

27,3397

1

It

it

5.8

26,4208

1

it

<■'

5.7

25,5176

1

it

it

5 . 6 and smaller circles used
in size perception

24.6301

1

tt

Square

5.317 (side)

28.2743

2

it

Equilateral A
Openings t

8.081 (side)
nscribed in circle — 28.2744

28.2743

2

it

Square

4.243 (side)

18.003

2

tc

Equilateral A

5 . 196 (side)

11.691

2
12

Total

. .. 56

30

CHAPTER IV

Problem, Method, and Technique

The experiments' reported in the following pages were made
with about thirty chicks belonging to four different groups.
The first and fourth groups, each consisting of ten chicks, were
secured from poultry breeders when the chicks were about two
days old. The second and third groups were artificially in-
cubated in the laboratory. The chicks used were Plymouth
Rocks and all except two were of the Barred variety. The two
exceptions were White Plymouth Rocks but neither contributes
importantly to the final results. Properly caring for the chicks
and keeping them healthy was one of the most serious difificulties
which had to be overcome. On the whole, the laboratory
hatched chicks proved more satisfactory.

The most common ills seemed to be due chiefly to improper
feeding, irregular temperature, and inadequate ventilation. A
few individuals of group 3 survived until the weather became
warm enough for them to be out of doors during a few hours on
favorable days. When this plan was first tried the birds were
in poor condition. Two-thirds of the group had died. But
as soon as the survivors were placed out of doors their physical
condition began to improve, and it was possible to do nearly
three times the amount of experimental work that could be
formerly done. Evidently, the requirements for healthy labora-
tory chicks include an abundance of sunlight and fresh air with
the necessity of working for a living.

The disease giving trouble with the first two groups was a
type of "leg weakness," so called because the leg joints become
enlarged, the toes curl out of shape, the chicks cannot stand,
and they move about only with great difhculty. This trouble
ultimately carried off nearly all of the birds which did not suc-
cumb to bowel disturbances. With one brood it was probably
the result of excessive heat in the hover. In other cases it was
probably due to overfeeding. More often, perhaps, it was due
to a combination of both conditions. The birds which first

32 HAROLD C. BINGHAM

showed signs of this weakness were the largest and apparently
the strongest of the flock. There was no evidence of it among
the chicks of the third and fourth groups, which were fed
sparingly and for which the temperature of the brooder was
carefully regulated.

The matter of health among the chicks turned out to be a
problem which had not been anticipated. However, it did not
prevent work toward the solution of the primary problem. The
task originally planned was a study of the chick's discriminative
ability between sizes, forms, and brightnesses, but, owing chiefly
to these unfavorable conditions, little consideration has been
given to the third factor. Later tasks added to the original
plan involve the perception of flicker and a study of the learning
process and ideational content.

The aim has been to make the study intensive and quantitative.
The question whether the chick can discriminate between stim-
uli diff^ering with respect to a single visual detail is only a
preliminary aspect of the problem. The primary task has been
that of determining the least perceivable difference in detail
vision. Breed's work indicates that chicks can discriminate
on the bases of size and form. The work of Katz and Revesz
suggests the same possibility. My original plan, then, was to
determine the threshold of difference for each of these factors.

With respect to size, my original plan was carried out with no
essential changes, but with form, it was found necessary to
abandon the original plan. In a short time it appeared that
proper responses to stimuli differing only in form were not so
readily acquired as reactions to size differences. After trying
in vain for several weeks to train different subjects to discrimi-
nate between a circle and a triangle equal in area, the nature of
the problem was considerably modified. It was clearly neces-
sary to determine whether the chick, under the conditions of
this experiment, could perceive form differences.

The task involving flicker perception, like that with forms,
demanded first a preliminary answer. Other than the generally
observed fact that birds are highly sensitive to moving stimuli,
there were no experimental data at hand to indicate that a three
or two to one flicker difference could be perceived. Beyond the
study of a two to one discrimination, time and conditions have
not permitted me to go.

VISUAL PERCEPTION OF THE CHICK 33

The studies involving problem learning and ideation arose
merely as incidents in connection with the more fundamental
tasks. As the time of observing the chick's behavior in the
various problems lengthened, there was a proportional increase
of interest in the behavior itself. At last I gave in to a per-
sistent desire to record quantitatively the chick's characteristic
behavior in learning a particular task. From such an interest
it is a natural step to questions pertaining to evidences of
ideation in this learning. The problem of relative stimulus
difference, incidentally discussed in Chapter IX, represents an
initial approach to the study of the imaginal problem.

Finally, another aspect of the problem, also arising quite
incidentally, was that of the relative value of the different visual
details. My earlier results with forms, largely negative, em-
phasized the desirability of considering the factors in combina-
tion as well as in isolation. In its normal life the chick is not
compelled to rely upon a single visual factor, but, on the con-
trary, it relies upon a natural combination of the visual details,
It thus seemed desirable to start with complex stimuli and from
this complex gradually to eliminate inequalities until a single
visual factor remained. This method might furnish data on
the relative value for the chick of the various visual factors.

The plan adopted for solving these various problems was the
familiar Yerkes discrimination method. The description of the
apparatus sets forth the general features of the method. Condi-
tions both desirable and undesirable are presented to the chick.
Either nest box contains those things which the chick wants.
It provides food, light, warmth, and companionship. The ex-
periment box is arranged to make the chick seek escape. In the
entrance box the chick is closely confined; in both the entrance
and discrimination chambers the floors are wet; the entire re-
action box provides faint illumination, little warmth, no food,
and no companionship. The chick's problem is to learn how
to get from the undesirable to the desirable part of the appara-
tus. The two different stimuli are the signs by means of which
the chick may learn which way to escape.

Each chick was taught the way of escape to the nest box by
means of 20 preliminary trials. The entrance to the nest box
at the side where the right stimulus appeared was open; the
sliding door closed the entrance from the electric box where the

34 HAROLD C. BINGHAM

wrong stimulus was presented. The chick was allowed to go
alternately to each nest box 10 times. It was thus made fa-
miliar with the nest boxes and the reaction box.

By displaying, always on the side of escape, the stimulus
which the chick was later to be trained to choose, the subject
was occasionally aided, prior to training, in acquiring a perfect
habit. ^ This fact was especially noticeable in the experiments
on size discrimination in which a circle 6 cm. in diameter always
appeared in that electric compartment by way of which the
subject escaped. At the end of the preliminary series a few

chicks were responding perfectly to this condition of o 28 H

o 7 + discrimination. 2 This perfect o 28 H o 7 + habit,

however, may not mean that size was the basis of choice. It is
quite probable that the chicks chose the lighter compartment.
Precaution was taken to eliminate this possibility before size
tests were completed, and control tests were introduced to make
certain that it had been eliminated.

This preliminary work was followed by the training series.
Both entrances to the nest boxes were now closed, and the only
cues that remained to aid the chick in reaching the nest box
were the two visual stimuli. Continuing with the example of
size discrimination, o 28 + was the positive sign, ©7+ was the
negative sign of escape from the experiment box. If a chick
chose o 7+ by stepping into that compartment, it was shocked
by momentarily closing the key, Z of figure 3. A single shock
was usually administered but if it were not effective it was re-
peated. The wet floors of A and B regulated the intensity of
the shock. Care in the manipulation of the shock was impor-
tant for it was essential that it be an effective punishment yet
mild enough to avoid frightening the animal. The results of
weeks of work could be destroyed in a moment through a
blunder in the method of punishment.

In the early stages of training the chicks often went beyond
the exit towards the stimulus at the end of the compartment.
If the animal were allowed to do this it spent considerable time
and energy around this illuminated stimulus. By placing a

^ A habit is termed perfect when a chick successively fnakes 20 correct choices.

2 The stimulus demanding a positive response is named first followed by the
stunulus demanding a negative reaction. The dimensions represent the number
of square centimeters in the area.

VISUAL PERCEPTION OF THE CHICK

35

Subject

RECORD SHEET
Date /jz/y y^//— //z/^^Experimeiit^?:^^7':7Wfc'

Tests— •

1

2

3

4

5

6

7

8

9

10

R

w

REMARKS

=

S.-ES i

A

1

r

1

r

I

r

1

r

1

r

B

r

1

r

1

r

1

r

1

r

1

1

r

1

0- b

r

1

o - 'f

r

1

0 - 120

r

1

r

1

0 - 'i-'^

y=

<i

<;/. i' ■

2

1

1

o - J^

r

1

£-,-iz

X

0- 3f

1

0- >f

1

£-X-/tO

X

a- 'J-

X

0 - >33

f

3^

-f^^.^a

3

r

O - Cy

r

1

r

1

1

X
E-r-lto

1

X

0 - J-

1

£-1- CI

^

uS-

U7'

4

1

1

1
E-I-JS

X

o - ^j-

X

X

c - 6^

1
£-X-/<7ff

r
0-30

r

i> - 3Z

1
£-l-t,S

L,

^

•A:?- Z.

5

I

B-Z-S4

1

a - r

r

" - "

1

o - /a.

X

c - /a

1

0 i>

r

<5 - /f

1

0 - zip

X
£-/-.V-J

1

■7

J

ZzSliA

6

1
O - "

1

e-z-7f

1

X

X

1

X
£Z-'zo

1

0- *^

X
£ -Jt-fl/

i>

J^

s/4i

7

X

1

O- J/

1

1

•=- 7

X

X
e-z-XJ

X

0 - :zs-

1

0 - i.

r
0 - y

1
£ - 1 -si

(,

^

.Z.g.J

8

E-Z-fi

r

0 - t.z

1

1

0 - U'T

X

I

X
£-/-i,S

I

r

0 - a.'^

1

£ -i-JW

V

i

J>i,o

9

r

r

r

1

I
£■-/- 7

1

0-3.1

£-^-¥1

1 .

X

0- 'S-

1

£-/■/<.

«<-

^

^cf.^

10

I

1

<7 - -iT

1

1

r

X

X

r

1

X

0 ■ -i^

■ £>

0

A^.J-

11

r

■ O - /O

1

r

X

r

I

1

0 - J^

1

r

a - ar

1

e-iSf

b

-A

hO. 2j

12

r

0 - c

1

r

1

X

r
o-/a.

1
0-g

1

0- g

X

1

g

A

/«. -r

13

r
o- e

1

r

1

1

1

X

X

e-/-xo

X

0 ■ a.1

1

7

J

,3S..T

14

1

0 - iJt.

1

1

1

o - lo

X
El- /O

r

0 - 7*^

X

0 -/*.

X

£ 1-a.y

1
0- s

X

^

v3-

>3S.e

15

r

1

O - 1/

X

X

X
£-/-3i

1

I

£-X{^

1

0-/6

X

1

u

■^

.za.^S-

16

1
O- fg

r

1

1

1

0 -vJ-

r

X

r

1
0 - //

r

i'

z.

X,'^ St

17

r

r

r

<9 .^c

r

o- //

1

1

1

e- ix

1

O-JO

X

0 -a.t

1

O'

I?

^2,-g

18

1

r

I

X

E-i SS

r

1

1

X
0 - IS-

1
£-i->e

X

0 - 1 A

¥-

i,

Sz.g

19

r

1

o - *4o

I

1

X

0 - J.£

1

X

£ ^-«y

1
0-10

X

1

£- i-JJ

v-

(,

^^7

20

1

1

1

o - g

r

^ .2.-/7

1

r

0 - J.S

1
0 - <,

X

e '/-Co

X

0 - ^Z.

X

£ -IIOJ

i.

¥■

33.3

21

r

1

O- / 3,

1

o - .J

r

r
0-/7

I

1

X

X

0- 'f

1

.s

c5-

'S.S-

22

1

1
0-,f

r

X

1
" 7

1

0 - Jt/

r

X

£ a-ja

1

X

■s

-7

g.<f.,9

23

I

O- tz.

1

1

I

1
0 - a-

r

r

X

0 - s.

X

1

i,

^

/3. 3

24

1

o - ^

r

1

1

1

X
0 "^

r

<3 - V-

X

0 - 'A

1
0 - a.

X
E-/-iAi

T

c?

/A.'T

25

T

r

r

r

O- 3.(,

1

0 - Jl^jr

1

0- S.3

1

1

0 - /.s-

X

0 -^s

1

f

/

Xz-f-r

36

HAROLD C. BINGHAM

thin plate of glass across each compartment at a point just
beyond the exit, the chicks were stopped near the exit and yet
were not prevented from seeing the regular stimulus. As soon
as they had learned the way of escape and had ceased trying to
reach the stimulus, the glass plates were removed.

After a perfect habit had been acquired, the amount of dif-
ference between the two stimuli was decreased. When the

o 28 H o 7 + discrimination became perfect, the situation

was changed to o 28 M o 9 +. The large circle was thus the

standard which remained constant. The smaller circle was the
variable which was successively increased in size until the ani-
mal could no longer choose correctly. This limit of correct
choices was accepted as the chick's threshold of difference.
The changes in diameter of the variable, it was found, could be

TEST SHEET

Title of investigation, Form: o28H A 28+

Experimented on, 9

Harvard Psychological Laboratory, December 25, 1912

Record Sheet, 1; Series, 15

TEST

BEHAVIOR

RECORD

1 r

U- III + /// .

0-38

21

/// + /// .

0-11

3r

77+///.

0-22

4r

// + // .

0-8

5r

/// - // Ai + /// .

E-1-38

61

/// - /// A^ + // «

E-2-54

71

// - /// A^' + /// .

E-2-64

81

// + // „

0-16

9r

II + 1 A' + II „

E-1-39

101

1 ^ II „

0-5

6-4-29.5
Remarks: Tendency to choose by position, i.e., to go where it last escaped.
Usually goes immediately to other compartment when it has been shocked for wrong
choice.

as great as 5 mm. between o 7 + and o 12+ when presented
with a standard © 28 + . Above ©12+ the diameter of the
variable was successivel}^ increased only by 1 mm.

For tabulating results, a record sheet which had formerly
been used in the laboratory was well suited to my needs. It

VISUAL PERCEPTION OF THE CHICK 37

provides for a concise summary of results. For the details of
behavior, a test sheet was adopted. The 10 tests, horizontally
arranged on the record page, appear vertically on the test page.
After each one is a space for recording the details of the animal's
behavior. At the right of the page is a narrow column for
recording the results of the test — whether an error was made
and the time for choosing. At the bottom is a space for further
notes.

The manner of recording the details of the chick's behavior
is illustrated in the accompanying test sheet which is a record of
an actual series of tests. A set of symbols was used in order
to follow fairly accurately the movements of an animal while
in the discrimination chamber. Herewith is presented the key
to the symbols used in these experiments:^

1. + = Approach to right compartment.

2. — = Approach to wrong compartment.

3. /, //, /// = Degree of attention based on behavior and time. A horizontal

bar above any one of these symbols indicates unusually long time in
this locality. For example,
+ / = Brief consideration of right stimulus area.

— / = Brief consideration of wrong stimulus area.

, f + // = Longer consideration of right stimulus area.
' \ — // = Longer consideration of wrong stimulus area.

) + /// = Close consideration of right stimulus area.
\ — /// = Close consideration of wrong stimulus area.

f + /// = Long time before and close consideration of right stimulus

, I area.

I — /// = Long time before and close consideration of wrong stimu-
[ lus area.

+ 7 = Long time before but slight consideration of right stimu-
lus area.

— 7 = Long time before but slight consideration of wrong stimu-
lus area.

4. '"""""""" = Attention at entrance end; indifference to stimulus areas.

5. A = Approach on wires before right stimulus followed by retreat.

^ The symbols presented in print only approximate, as closely as is convenient,
the written symbols actually employed during experimental observations. Printed
substitutes, of courpe, are consistently presented throughout the text. The written
symbols used for recording observations were selected as far as possible to repre-
sent concisely and accurately the behavior recorded.

38 HAROLD C. BINGHAM

6. V = Approach on wires before right stimulus followed by wrong turn.

7. A = Approach on wires before wrong stimulus followed by retreat.

a. A^ = One shock

b. A" = Multiple shock.

c. a" = No shock.

8. 9 = Turn around, clockwise; o, counter-clockwise.

9. <\} = Partial turn around, right to left; c/3, left to right.

10. \\ = Position directly before division between each compartment; by turn-

ing head either stimulus area is visible.

11. ,/ = Escape to nest box.

12. E = Error in choosing; this is followed by time in seconds. '

13. 0 = Correct choice; this is followed by time in seconds.

For specific behavior in the electric compartments and remote
corners of the experiment box, which it was desirable to record
in gathering the data presented in Chapter VIII, the following
additional symbols were used:

n. r, U = Right corner, left corner, or stimulus faced and apparently

inspected.
H- n. U = Right corner, left corner, or stimulus approached and inspected.
~], [~, [J = Right corner, left corner, or stimulus more closely approached
and inspected.

n, n. U = Right corner, left corner, or stimulus appropriately approached
and inspected.

The preliminary series were begun during the second week,
usually when the chicks were 10 days old. The experiments
on space perception were conducted during the morning between
the hours of eight and twelve. The experiment on flicker per-
ception was conducted with less regularity during the morning
and afternoon. On the whole, the early morning is the best
time for experimental work with chicks. They are then most
active for that is the time they naturally start out for food.

In the matter of rewarding with food, however, caution is
necessary. A three-weeks-old chick will "fill up" in a short
time if food is abundant. It then becomes sluggish. The plan
was adopted 6i scattering only a few grains in the litter of the
nest boxes so that the chick was kept active in working for its
meal. This plan, when perfected, made it possible to increase
the number of tests considerably above the maximum when the
food was scattered less sparingly.

Up to this point considerable emphasis has been laid upon
method and technique. Certainly an accurate solution of an

VISUAL PERCEPTION OF THE CHICK 39

animal problem depends upon the adoption of a favorable
method. The experimenter must use an apparatus by means
of which he can accurately control the conditions of his study.
He must also show alertness by controlling these conditions in
such a way that his animals behave normally. Much animal
stupidity, so-called, is really a reflection of human ignorance.
The experimenter may lack animal intelligence by setting human
problems for a lower animal to solve.

An animal frequently shows ingenuity in an unexpected direc-
tion, and if the experimenter be not alert he misses the most
important part of the behavior. How many times this is the
case, we cannot tell, for we only know of the cases where we did
not miss the significant fact. A perfect method would make it
impossible for the animal to pick up cues that were not inten-
tionally offered by the experimenter. But it is through experi-
ence that we approach a perfect method, because there is always
the possibility of improvement even in the best standardized
methods.

The importance of method was impressed upon me through
an incident in my experiments with the first group of chicks.
I had inexcusably blundered by making the original conditions
of discrimination too difficult for the chicks. They were being

tested on o 28 H © 7 -f discrimination. The subjects were

allowed to go through the discrimination chamber and experi-
ment box during the preliminary tests without regard to the
position of the right or wrong stimulus; that is, both exits were
open. When the training series were begun the brightnesses
of the two stimuli were made markedly unequal and were irreg-
ularly varied from the first.

The experiment was begun with six subjects. As a partial
result of my initial bungling, I had at the end of a few days only
subjects 2 and 3 in suitable condition for further work. At the
end of two and one-half weeks No. 2 gave up, but No. 3 persisted.
At the end of the 24th series it had successively made three
perfect series (see table 2). At the end of four more series it

had reacted perfectly to the o 28 -\ o 9 -f- discrimination.

The diameter of the variable was successively increased by 5
mm. until the variable o 19 -F was reached after which the diam-
eter of the variable was changed by increments of 1 mm.
The perfect responses continued with very few exceptions.

40

HAROLD C. BINGHAM

TABLE 2
Subject: 3. Hatched: 10/3?/'!!. Sex: Undetermined

SERIES*

DATE

RIGHT

WRONG

TIMEt

O 28+-

O 7+ Discrimination

1-21

(1-21)

Oct. 12-27

22

(22)

" 28

io

0

55

23

(23)

" 30

10

0

41

24

(24)

" 30

10

0

29

O 28+ —

O 9+ Discrimination

1

(1)

Oct. 30

10

0

62

2

(2)

" 31

9

1

79

3

(3)

" 31

10

0

103

4

(4)

Nov. 1

10

0

55

O 28+ —

O 12+ Discrimination

1

(6)

Nov. 2

7

3

71

2

(7)

" 3

9

1

93

3

(8)

" 3

10

0

50

4

(9)

" 4

10

0

52

G 28+ —

O 15+ Discrimination

1

(11)

Nov. 4

6

4

112

2

(12)

" 6

9

1

68

3

(13)

" 6

8

2

44

4

(14)

" 6

8

2

51

5

(15)

" 7

9

1

52

6

(16)

" 7

9

1

40

7

(17)

" 7

8

2

41

8

(18)

" 8

10

0

52

9

(19)

" 8

O 28+ -

10
O 19+ DiscrimitiL

ition

0

62

1

(21)

Nov. 8

8

2

49

2

(22)

" 9

8

2

49

3

(23)

" 9

8

2

90

4

(24)

" 9

10

0

50

5

(25)

" 10

8

2

78

6

(26)

" 10

10

0

53

7

(27)

" 11

10

0

47

O 28+ —

O 23+ Discrimination

1

(6)

Nov. 11

8

2

58

2

(7)

" 13

8

2

116

3

(8)

" 14

10

0

41

4

(9)

" 15

8

2

45

5

(10)

" 15

10

0

16

O 28+ —

O 24+ Discrimination

1

(21)

Nov. 17

10

0

22

2

(22)

" 17

9

1

30

3

(23)

" 18

10

0

16

4

(24)

" 18

10

0

11

* Numbers in parentheses refer to series on record sheet,
t Average time for series given in seconds.

VISUAL PERCEPTION OF THE CHICK

TABLE 2— Continued

41

SERIES*

DATE

RIGHT

WRONG

TIMEf

1 (11)

2 (12)

3 (13)

O 28+ —
Nov. 21
" 21
" 21

O 25+ Discrimination
9
10
9

1
0
1

22
16
17

1 (16)

2 (17)

3 (18)

4 (19)

O 28+-

Nov. 22

" 22

" 23

" 23

O 26+ Discrimination
10

9

7
10

0
1
3
0

15
24
45
23

1 (21)

2 (22)

3 (23)

4 (24)

5 (25)

O 28+ —
Nov. 23
" 23
" 24
" 24
" 24

O 27+ Djscrimina
8
8
6
10
9

lion

2
2
4
0
1

20
25
2,2,
11
• 30

1 (1)

2 (2)

3 (3)

O 28+ —
Nov. 25
" 25
" 25

O 28+ Discrimina
10

6 (crack closed)
0

tion

0
4

5

?
?
?

1 (16)

2 (17)

O 28+ —
Nov. 27
" 27

O 23+ Discrimina
5
5

lion

5
5

62
59

1 (18)

O 28+ —
Nov. 28

O 19+ Discrimination
3

7

71

1 (19)

O 28+ —
Nov. 28

O 15+ Discrimination
9

1

60

1 (20)

2 (21)

3 (22)

4 (23)

5 (24)

6 (25)

O 28+-
Nov. 29

" 29

" 30

" 30

" 30
Dec. 1

O 19+ Discrimina
6

5
7
7
5
5

tion

4
5
3
3
5
5

73
83
31
22
23
27

* Numbers in parentheses refer to series on record sheet,
t i\.verage time for series given in seconds.

Finally the behavior of the chick indicated that it was discrim-
inating between ° 28+ and o 27 + (see record for November
23 and 24).

The discrimination of such a minute difference was incredible,
for the human eye could scarcely detect a difference after the
variable reached o 23+. The next step was to try the chick

42 HAROLD C. BINGHAM

when the variable was made equal to the standard. The result
was a perfect series (see record 1 for November 25).

With positive evidence that the discrimination was not of
size difference, an inventory of possibilities was taken. Among
the features noted was a small crack where the outside extension
of the electric source box jointed to the experiment box. But
a similar crack appeared on both sides in corresponding posi-
tions, so it seemed that in this factor there could be no cue by
which the chick had been guided in its choice of the proper
compartment.

Closer inspection, however, proved that these two minute
cracks lay at the bottom of the matter. Where the rabbeted
edges of the shifter and tracks rubbed, there was a bright edge,
and wherever the shifter rested the bright surface was covered.
Thus when the shifter was at the left, the right end of the track
was uncovered, and from this uncovered part were reflected some
of the light rays from the upper illumination. A small portion
of this reflected light could enter the crack on the uncovered
side. Now, when the standard circle was presented at the left,
the shifter was always moved to the right; when the standard
stood at right, the shifter stood at the left. As a result, the
crack that was illuminated was always the one w^here the stand-
ard circle was displayed; the other crack, being reached by no
reflected light, was comparatively dark.

The effect of closing these small cracks through which the
light was reflected is indicated in the records following the first
on November 25. The perfect reactions abruptly cease; out of
15 tests there are only 6 correct responses. As the following por-
tion of the table indicates, the experiment with the cracks
closed was continued, but the variable had to be reduced to
o 15+ before the correct reactions returned.

The chick, then, had detected a faint streak of light not more
than 6 mm. long and less than 1 mm. wide or else the increase in
general illumination which resulted. By considering the
amount of light that came from the upper illumination, the poor
reflecting surface of a steel strap worn only moderately bright,
and the narrow surface from which the light was reflected, (not
more than 1 cm. in width), the insignificance of this strip of
light is emphasized. It was an example of the chick "outguess-
ing" the experimenter.

CHAPTER V

Size Perception

The foregoing discussion of method suggests the difficulties
that arise when one's experimental problem becomes limited
to a single detail of vision. It emphasizes the uncertainty of
earlier work where exacting controls were not employed. In
the light of my experience, it seems to be a safe guess that prior
studies, presumably of detail vision, were, in fact, studies of
reactions to visual complexes the constituents of which were
not known to the experimenters. The control, in my study,
leading to the discovery of the disturbing crack proves that the
results with chick 3, prior to the control, were untrustworthy.
In other studies where the possibilities for similar disturbances
have been many times greater, results should be accepted with
considerable caution.

After the discovery of the light-admitting crack, my results
become more reliable. Later experiments with chick 3 indicate
that the early positive choices had been made on the basis of a
discrimination of size difference, but as the variable became
larger, and the discrimination proportionately harder, the chick
resorted to a different criterion for choosing. In table 2 one finds
evidence of the place where chick 3 ceased to discriminate be-
tween the two size stimuli. After the controls, the chick was
first tested with a variable o 23 + . The results are clearly nega-
tive since chance alone is sufficient to explain the number of
correct responses. With the variable reduced too 19 + , there
are only three correct choices in ten tests, but when the variable
was changed to o 15+ the results are clearly positive. The
standard was chosen nine times and the variable only once.
While the average results of the tests with the o 19+ are
slightly in favor of the larger stimulus, it seems that one should
not emphasize the fact for the larger circle was chosen, in a total
of 60 tests, only five more times than chance allows.

Table 3 summarizes the results of my study of the chick's
perception of size stimuli. In contrast with earlier reports

44

HAROLD C. BINGHAM

+

^

o

1

'—

^

+

00

«

o

1

+

>

<M

'"'

d

+

C4

+

^

"

+

00

+

^

.o

rO

..

"

+

^^

^

^

OO

Pi

+

^

O

^

■^

6

+

z

00

Pi

+

^

-

+

00

pi,

+

?:

^

r^

"

^

+

^

'b

00

p;

+

^

o

^

^

-

-

o.

o

+

CO

C<

o

O

-

«

o.

00

+

^

^

^

<^

-

-

-

o

-

o

O

o

o

o

+

Pi

^

o

Ov

o

00

o

O

o

o

o

OO

^

^

"^

^

""

"^

+

^

cs

'-

„

•b

+

00

Pi

00

+

^

CN

cs

cs

^

rs

f

rs

—

o

+

00

Pi

00

00

00

VO

00

VO

00

+

^

ro

-

-

cs

o

-

-

■*

-

o

^

o

o

+

1^

00

-o

On

o

00

o

o

s

w

'^

^

^

+

^

-

^

+

00

M

1

+

LO

fN

-*•

fS

„

^

o

ro

C2

''

O

2

+

Pi

l«

00

Ov

vO

o

00

^

+

r^

"*

+

p;

H

0!

u

1 •

^

r~l

VO

t^ 1 00 1 ON

o

_^

cs

'^

1

1

~

""

""

"■

=* o -8

OS 1^ Cl!

-5 +

2+e

rt On 3

■S I c

il 00>'

<u_o j2

a.H ofaH

s I

•f=+

VISUAL PERCEPTION OF THE CHICK

45

1

S

f^

05

w

ij

;;

ra

^•^

<

a

H

+

^

+
00

«

+

^

+

00

Pi

©1 "§•

I 'S "o

+.5 *^
«S .§

•a 2.2 "S

., u rt i2

j; !«■ o in
-u ^ ^ ij

H tnMH

46 HAROLD C. BINGHAM

which emphasize habit formation my results do not always ap-
pear clear cut and definite. Frequently the positive choices
that have been allowed to pass as though satisfactory do not
exceed 75 per cent of the total, whereas earlier work seems to
have been more regularly favored with ultimate perfection.
Does this not indicate that my results are untrustworthy?

In general, there are two answers to this possible criticism.
I shall refer first to the least important, namely, that my task
was not one in habit formation. Perfection of a habit was only
incidental to the primary task of sensory acuity. Experience
with the first group of chicks indicated that time was wasted
by holding rigidly to the perfection requirement. The uncer-
tainty of the chick's health, further, made it advisable to hasten
towards the threshold of discrimination without waiting in all
cases for perfect habits.

Chicks 6 and 7, representing the first group, were held to the
preliminary training, o 28 H g 7 +, until two successive per-
fect series were obtained. Then the variable was changed to
o 12 +. From the first, there was evidence of discrimination
of this new variable but only a single perfect series appeared.
After six or seven series, there was no evidence of improvement
over the beginning of the ©12+ discriminations. This sug-
gested that time could have been saved by increasing the variable
as soon as there was satisfactory evidence of discrimination,
even though positive reactions had not reached 100 per cent.

With the second group, including chicks 15-18, therefore,
this suggestion was tried out. The series numbered 1 in table
3, which was the 14th for this group of chicks, represents the
final series of the preliminary training. In that series chick 17
is the only one to react perfectly. But the reactions of 15, 16,
and 18 were so closely approaching perfection that there was
unmistakable evidence of discrimination. Moreover, it was
more convenient to present to the other chicks the same stimuli
which were presented to chick 17. Without further training,
then, all chicks of the group were confronted with the new
variable, o 12 -f.

That such haste was justified is indicated in the single o 28 +
— o 12 + series for each individual of this second group.
Chicks 15, 17, and 18 responded perfectly while the positive
responses of No. 16 are 8 out of 10. Again, when the variable

VISUAL PERCEPTION OF THE CHICK 47

was increased to o 15 + , chick 16, the only individual of the
group which had failed to make a perfect record in one of the
two preceding tests, is the only one in this series to make a per-
fect record.

Comparing, now, these records of chicks 15-18 with the rec-
ords of chicks 6 and 7 of the first group, the haste is further
justified. Chicks 6 and 7, after continuing in the easy training
until the reactions were well perfected, subsequently fail to
show up superior to chicks 15-18. Furthermore, the records of
chicks 20 and 21, belonging to group 3, are no less clear cut for
the variables © 12 + and o 15 + than the records of chicks 6
and 7.

The second answer to this possible criticism is more theoretical
than the first, but the two are closely allied. A review of the
actual results indicates that it was not essential to insist upon
100 per cent correctness. In the second place, mine is a study
of detail vision in the strict sense; earlier studies, though pre-
suming to present data on visual details, have probably dealt
with visual complexes. For this reason, my results cannot be
justly estimated by means of the same criteria that have been
adopted in connection with earlier work. In studies of detail
vision, animals are compelled to react appropriately to a single
visual factor. Now, when the remaining factors happen to be
so grouped that the complex becomes especially attractive, — ■
and well controlled tests will contain such combinations, — is it
surprising that the attractive complex gains the attention of the
animal for the time being and causes the animal to ignore the
more insignificant single visual detail? In control tests, one is
always demanding, in the face of other attractions, responses
to this more or less insignificant factor. Due allowance should
be made for momentary lapses of the attention of the animal.

Furthermore, as the threshold of the particular visual factor
is approached, the single detail becomes less and less significant.
It is reasonable to suspect, then, that even before the percepti-
bility of the discriminate detail wholly vanishes, the animal
will be tempted to try out novel reactions. On the analogy of
human behavior, it seems reasonable that the animal actually
ignores at times the discrimination problem that has been set.
This supposition is especially plausible when we consider that
the vividness of punishment may decline as the discrimination
habit approaches perfection. Making allowance, then, for

48

HAROLD C. BINGHAM

these interruptions, which are certainly not indicative of inabil-
ity to perceive the visual detail in question, wholly perfect
responses when the threshold is being approached cannot rea-
sonably be expected.

By considering, in addition to the correct records, the pecu-
liarities of behavior which cannot be presented in tabular form,^
I am convinced that the largest variable which the chick can
distinguish, under the conditions of this study, when presented
with a standard o 28 + , lies somewhat above o 15 +. The
quantitative measurements on the basis of correct choices sup-
ports this conclusion. Every subject, excepting 7 and 20, suc-
ceeded in making at least 80 per cent of the choices correct in a
series when the ©15+ variable was used. Even in the re-

TABLE 4
Average time for choosing

SUBJECT

0 28+-0 7+

0 28+ — O 12 +

0 28+-0 15 +

6

15"

13"

7"

7

22

59

55

15

8

5

31

16

30

28

11

17

3

8

13

18

4

3

20

20

3

13

11

21

6

2

9

Average for all .

subjects

11"

16"

20"

spouses of 7 and 20, an efficiency of 70 per cent was indicated.
But more convincing, perhaps, is the fact that two chicks, 3
and 17, reached 90 per cent and two, 16 and 21, made perfect
series.

The time used by the chicks in choosing is a factor which
might throw supplementary light on this threshold question.
Having kept the time for every choice, and having casually
observed changes in time for choosing in the beginning, inter-
mediate, and final stages of discrimination, it is at least inter-
esting to note whether the actual time records confirm this ob-
servation. In table 4 appears the average time in seconds
consumed by each of eight subjects for choosing under the
three different discrimination settings. An average of the

^ See sample test sheet, page 36, for method of recording these peculiarities of
behavior.

VISUAL PERCEPTION OF THE CHICK 49

averages for the eight subjects gives respectively for these dis-
crimination settings 11, 16, and 20 seconds. These averages
indicate that the chicks in the early stage of learning were
faster in choosing than in the intermediate stage. However,
the reliability of these averages as an index of discriminability
is open to question. The frequency of errors indicate that less
discrimination underlies the early responses. While the chicks
were less acquainted with the conditions of the experiment in
the initial training, variable o 7 + , they made many direct
choices without showing evidence of comparison such as fre-
quently occurred when the discrimination habit had become
well established. Less time naturally was recorded for these
direct choices, although errors were more frequent. Apparently
the averages of No. 21 as recorded in table 4 are more reliable
as an indication of the relative time required for each setting
where discrimination underlies the choices. The time of this
chick for each of the three variables is respectively 6, 2, and
9 seconds. If it were possible to obtain an average for the time
of those responses only in which choice was based on discrimi-
nation, it is probable that we would find the time greater in the
beginning and final stages than in the intermediate stage. Dur-
ing the intermediate stages the habit would be well established
and the discrimination settings would not be extremely difficult.

It might be urged that this time average became relatively
longer towards the close of the experiments on account of the
physical condition of the chicks. The facts, however, seem to
refute such a view. Referring again to the individual record
of chick 21 in table 4, one finds evidence that the changes in
time averages are not conditioned by the health of the chicks.
During the course of these tests No. 21 was given open air
range under which conditions it throve and grew normally.
This chick continued in good health even at the end of 1500
subsequent tests in form perception. The records of table 4
indicate that chick 21 was slower with the variable o 7 + when
it was learning, the average time was not increased when the
variablew as changed to o 12-|- because the discrimination re-
mained easy, but the time was increased when the more difficult
discrimination, variable o 15 + , was required.

The records of chick 21 alone cannot be accepted as reliable
on account of the limited number of tests, particularly in the

50 HAROLD C. BINGHAM

intermediate stage. On the other hand, one might assume
with some justice that, in more than 2000 tests distributed
among eight individuals, all irregular fluctuations would tend
towards an equal distribution among the different size relations.
In this event, the time records represented by the final averages
of table 4 might be said to substantiate the preceding conclu-
sion, based upon the number of correct choices, that the maxi-
mum which the chick can discriminate under the conditions of
this experiment lies somewhat above o 15-f-.

Statistically considered, however, the averages of this table
have very little significance. The mean variation, for example,
from the 11, 16, and 20 seconds averages are respectively 8, 13,
and 12 seconds. These mean variations are nearly as great as
the averages themselves, and they are, in every case, as great
as or greater than the differences between any two of the aver-
ages.

These averages of table 4 are even more questionable when
one considers the individual records on which the table is
based. A notion of these individual records may be derived
from Test Sheet 1 and Test Sheet 2 which are presented to
illustrate how the behavior respectively differed under easy
and difficult discrimination. Taking these two series as ex-
amples, we find that the mean variation represented by Test
Sheet 1 is 1.88 seconds which is nearly 59 per cent of the average
for the ten tests. In the records presented by Test Sheet 2,
the mean variation is a little lower being slightly above 51 per
cent. This wide variability suggests that the averages for the
different individuals are themselves questionable. Combining
with this the variability noted within the table itself, these
averages appear as doubtful criteria.

A number of things might be mentioned in explanation of the
variability between the tests of a single series. Wholly secon-
dary influences, such as a severe shock, the slamming of a door,
or a snow slide on the roof, are often responsible for a sudden
fluctuation in the reaction time. A more natural difficulty
with the chick which is responsible for even greater fluctuations
in reaction time is a tendency to go to "roost" in the dark room.
It can usually be aroused by blowing lightly against the feathers,
but such behavior plays havoc with the reliability of the indi-
vidual time records.

VISUAL PERCEPTION OF THE CHICK

51

TEST SHEET 1

Title of investigation, Size: O 28H O 7+

Experimented on, 21

Harvard Psychological Laboratory, February 28, 1912

Record Sheet; 1; Series, 6

TEST

BEHAVIOR

RECORD

11

// + /// .

0-4

21

+ 777.

0-5

3r

+ 777«

0-2

41

+ 777«

0-1

5r

+ 777«

0-2

6r

+ 777«

0-2

71

+ 777«

0-2

8r

+ 777«

0-2

91

+ 777.

0-2

10 r

+ // '""" .

0-10

TEST SHEET 2

Title of investigation. Size: o 28-| o 7+

Experimented on, 16

Harvard Psychological Laboratory, March 6, 1912

Record Sheet 1; Series 16

10-0-3.2

TEST

BEHAVIOR

RECORD

11

/// -// + /// V - /// + /// „

0-47

2r

III - II 0 + III ,

0-38

31

771+- + ii-Fa' + -ii ^'^ II ,

E-2-58

41

II + 177 r\ o - /// + /// ,

0-35

51

/// + /// .

0-11

6r

/// -71 II ^' + III «

E-1-21

7r

II + /// - III ^^ III //

0-22

8r

+ ///„

0-5

91

+ /// o - /// o + /// „

0-35

10 r

II + /// .

0-10

8-2-28.2

52 HAROLD C. BINGHAM

The characteristics of the chicks' behavior offer a less precise
basis for determining the location of the threshold. This type
of evidence, however, is practically limited to the observer.
It can scarcely be conveyed to the reader. Yet, despite the
limitations of such evidence, the fact remains that the observer
has a qualitative criterion which he may employ in addition to
his quantitative measurements. If both agree he naturally
feels more certain of his conclusions.

The two test sheets presented illustrate two different types of
behavior. Test Sheet 1 shows that chick 21 lost no time on
entering the discrimination chamber, but went directly in every
test to the correct compartment. Test Sheet 2 presents a dif-
ferent type of behavior, yet it indicates that chick 16 was dis-
criminating between the two stimuli. In the first test the
chick was quiet for some time after entering the discrimination
box; then it went to the wrong side where it stopped momen-
tarily. It then turned to the left, went close to the left electric
compartment, hesitated, and entered but turned to the right
and came out. Again it looked closely at the wrong (right side)
stimulus, turned back to the correct stimulus (left), and escaped.
The behavior was somewhat similar in the second test except
that in turning away from the wrong stimulus, which was this
time at the left, the chick turned its self completely around,
counter clockwise, and came up before the correct stimulus.
The third test shows that chick 16 made two errors for both of
which it w^as punished. In test 9 the chick made two complete
turns — one counter clockwise from the correct compartment
(left side), and one clockwise turn from the other compartment.

These two test sheets have been selected with an idea of
showing as marked contrast as possible in order to emphasize
the importance of the chick's behavior during each test and
series. The first shows that chick 21 had no difficulty in choos-
ing the proper" stimulus. The second indicates that chick 16
was comparing the stimuli and using considerable care in choos-
ing, although it made two errors. The choices alone do not
indicate as much advancement as the details of the behavior
during the choices. These two series cannot be called typical
yet they represent fairly well some of the essential principles of
behavior in the discrimination problem.

VISUAL PERCEPTION OF THE CHICK 53

On the basis of choices and behavior — -it seems safest to ex-
clude the time factor — it appears that the chick's hmit in dis-
crimination from a standard o 28-f Hes somewhere between
o 15+ and o 19 + . To express the relation in terms of centi-
meters instead of square centimeters, the size in diameter of the
variable is above 4.5 but below 5 ; that of the standard is 6.
The threshold of difference with a standard 6, therefore, is one-
fourth to one-sixth.

To be certain that the reactions of the chicks were not made
on the basis of some detail or complex other than size difference,
it was necessary to employ various controls. There appear, in.
the stimuli themselves, three factors other than size that might
provide a basis for correct response: (1) luminous intensity of
the respective stimuli; (2) brightness of the respective stimuH;
and (3) illumination of the respective electric compartments.
Without adequate controls, it would be impossible to say definitely
whether size or these other characteristics were the bases of the
responses which point to the above conclusion.

Supposing the two stimulus areas to be 6 cm. and 3 cm. in
diameter, the display surfaces equal in brightness, and a com-
plete image of the larger formed on the retina, it follows that
the chick would be stimulated by four times as much light from
the larger as from the smaller stimulus. It is conceivable that
the chick, under these conditions, might respond on the basis
of pain or its physiological concomitant, or even of pupillary
reaction. In addition to this possibility, the general illumina-
tion of the electric compartment containing the larger stimulus
would differ from that of the other compartment. The chick
might choose the lighter or darker compartment. Both of
these possibilities might occur independently of the matter of
brightness.

As a portion of the controls used in this connection, an addi-
tional lamp was placed directly above the experiment box.
This upper illumination (L of figure 3, p. 24), consisting of a
2 c.p. electric lamp enclosed in a vertically suspended cylinder
of galvanized iron, was designed to hide the inequalities in the
distribution of light in each compartment. It hangs about 120
cm. above the floor of the experiment box so that an equal
amount of light is added to each compartment. The diameter
of the cylindrical enclosure is 10 cm. and its length is 20 cm.

54 HAROLD C. BINGHAM

The upper end is tightly closed. The lower end is fitted with a
removable cover in which is centered an Aubert diaphragm for
regulating the amount of light falling upon the experiment box.
To eliminate shadows in the experiment box and to increase the
amount of upper illumination, the diaphragm cover was removed
throughout the greater part of the experiment. With the lower
end of the cylinder open, the amount of light added to the ex-
periment box was that coming from a 2 c.p. lamp, passing
through a circular opening 10 cm. in diameter, and falling a
distance of 120 cm. Further variations in the upper illumina-
tion were made by substituting for the 2 c.p. lamp, after the
chicks had acquired a positive reaction tendency, larger lamps
with a maximum of 10 c.p.

The upper illumination thus serves as a control of the in-
equality of general illumination in the two electric compart-
ments. It can be so arranged as to furnish enough illumination,
equally divided between the two compartments, to reduce the
proportional difference in illumination to a point below the
chick's threshold. The amount of upper illumination should
therefore depend upon the degree of difference between the sizes
and brightnesses of the two stimulus areas. In a quantitative
study, where the difference in size is being continually reduced,
and where variations and reversals of brightness are essential
features of the vital tests, the possibility of consistent reactions
on the basis of difference in general illumination is removable
by means of the upper illumination.

Variations and reversals of brightness, as essential parts of
the system of control, are easily effected by changing the re-
spective distances of the source lamps from the display surfaces.
If the brightness of the respective stimuli be varied by reducing
the illumination of the larger, increasing that of the smaller,
or vice versa, the luminous intensity of the stimuli is at the same
time varied. Thus, a proper combination of controls, including
the upper illumination and relative changes in distance of the
sources, provides a means of determining whether the basis of
the chicks' choices lies in size difference or any of the three
possibilities (1) luminous intensity, (2) brightness, and (3)
general illumination.

The method of adapting these controls to my experiment is
concisely stated in table 5. The location of the source lamps

VISUAL PERCEPTION OF THE CHICK

55

was varied according to the plan of the second and third col-
umns. The illumination of the electric compartments was con-
trolled as indicated by the last two columns. In tests 1 and 2,
the + or correct stimulus area was considerably brighter and
the corresponding compartment lighter than the other. In

TABLE 5

Control series for Q 2S-\ ©7+ Discrimination

Nos. 15, 16, 17, 18, 20, and 21

March 4, 1912

Record Sheet 1. Series 14

TEST

SOURCE DISTANCE FROM

GENERAL ILLUMINATION OF

+ Stimulus

— Stimulus

+ Compartment

— Compartment

11

cm.

240

cm.

125

Uncontrolled except fc

r upper illumination

21

240

125

<<

<<

3r

240

125

Much darker: Cov-
ered with 2 thick-
nesses of ground
glass

Much lighter: Cov-
ered with thin sheet
of plain glass

41

240

125

Darker: Covered
with 1 thickness of
ground glass

Lighter: Covered
with 1 thickness of
plain glass

6r

240

240

a

71

180

180

li

8r

180

180

it

91

180

180

a

10 r

180

180

(<

tests 3 and 4 the general illumination was reversed without
changing the brightness or luminous intensity of the stimulus
areas. After test 5,^ the source controls remain unchanged
although the relative position of the + or correct stimulus
varies from right to left.

This represents the first specific control for o 28+ — Q 7 +
discrimination after a varying amount of preliminary training
had been completed. If a chick could not react correctly under
these conditions, it was assumed that the preliminary training

^In test 5, the source distances were equal and no controls were introduced.

56 HAROLD C. BINGHAM

had not been completed. If the previous correct reactions were
not noticeably interrupted by this control system, the chick was
considered ready for a larger variable. With the exception of
the records for chicks 3,6, and 7 the essentials of this procedure
were carried out for every record given in table 3. The results
presented in tabular form, therefore, are those of controls; they
are not training or learning records.

The significance of having the + compartment darker than
the —compartment in most of the tests of this control series
appears when the plan of the preliminary training is set forth.
In the beginning, the chick was required to choose on the basis
of a visual complex consisting of (1) the lighter compartment in
which the stimulus area was (2) larger, (3) triangular, and (4)
brighter. As soon as consistent choices appeared, the inequali-
ties between the two stimuli, except size difference, were
gradually removed by replacing the triangle with a circle, in-
creasing the upper illumination, and gradually bringing the
source lamps to an equal distance from the diffusing or display
surfaces. Next, the source distances were gradually and irreg-
ularly made unequal. Possibility of discriminating on the
basis of brightness and luminous intensity, or either, was thus
eliminated before reaching the final control stage. In the more
thoroughgoing control series, therefore, the most important
fact to establish was certainty about the possibility of general
illumination. The + compartment was thus made darker after
the first two tests while the other variations were continued.
In the matter of general illumination, then, the conditions of
the final controls were exactly reversed from those of the pre-
ceding training series.

This method of preliminary training was adopted after the
perplexing experience with chick 3, described in Chapter IV,
and the discovery of the crack which admitted reflected light.
The incident suggested that the chick which was being trained,
preliminary to the primary threshold study, should be aided in
this early stage by having the differences between the two visual
stimuli emphasized by a combination of light factors.^ Two
stimulus areas, differing from each other with respect to one
quality only, present to the chick a problem which is quite

^Yerkes, R. M. The discrimination method. Journal Animal Behavior,
vol. 2, p. 142.

VISUAL PERCEPTION OF THE CHICK 57

unnatural. Under natural conditions, it probably relies upon
combinations of many visual factors. The perception of an
isolated visual quality is a technical human problem, not a
natural animal problem.

Johnson^ criticises the use of the visual complex in prelimi-
nary training on the ground that one is risking a large waste of
time. He says:

It has happened that an experimenter discovers after several weeks of training
that the animal had learned to react to difference of e.g. brightness, and had not
been affected by the other stimulus-differences. It would seem that a difference
in luminous intensity is especially objectionable since in dark surroundings the
animal can choose the brighter or darker alley without attending specifically to the
stimulus-forms.

Now the value of the visual complex lies precisely in the fact
that the natural inclination of an animal is to attend to some
other factor than the one under investigation. A discrimina-
tion habit is hastened because the animal may resort to the
most conspicuous cue. Whatever this cue may be, it is used
as a valuable ally in directing the animal's attention to the
proper detail. A patient and alert experimenter can eliminate
brightness, luminous intensity, et cetera so gradually that the
animal's basis of discrimination is quite simply switched to the
detail in question, provided that the detail is perceivable for the
animal. The change should not be made abruptly, nor is it
necessary to wait for a perfect habit. Slight changes should
be commenced as soon as there is evidence of an incipient habit.
My results with chick 3, in contrast with later results from other
chicks, compel me to favor the use of the visual complex in pre-
liminary training. The success of one method or another
evidently depends to a great extent upon the temperament of
the experimenter.

Coburn's study (6, Chapter 2) of the vision of the crow fur-
nishes the only available results on size perception in birds that
are comparable with my results. His study of the crow's per-
ception of size was not primarily a threshold study, but his pre-
liminary results roughly indicate in centimeters a 3-2, 5-3, 6-4,
and 9-6 discrimination of circles. A hasty series of tests to-
wards the close of his observations indicate a discrimination
between a 5 cm. circle and a 4.5 cm. circle. He thinks a finished

^Journal Animal Behavior, vol. 4, p. 324.

58 HAROLD C. BINGHAM

threshold study would reveal a discrimination considerably
smaller than a 0.5 cm. difference.

Evidently the crow's perception of size is much superior to
the ability of the chick in the same task. Although the experi-
mental setting was somewhat different in our studies, it seems
that they are sufficiently alike to warrant comparisons. The
maximum variable that my chicks were able to discriminate
from a standard 6 cm. circle was the same, 4.5 cm. circle, as his
crows discriminated from a 5 cm. circle. From the standpoint
of size perception, it would seem, therefore, that the ability of
the crow is considerably greater than that of the chick.

CHAPTER VI

Form Perception

The development of a definite procedure in the study of form
perception in the chick followed, in general, that which has been
described in the preceding chapter on size perception. A study
of the ability of the chick to discriminate between circles and
triangles which are equal in area was made with two of the
second group, chicks 9 and 11. The latter became afflicted
with "leg weakness" after the 24th series, up to which time it
had given no positive results. The results from chick 9 which
are presented in table 6 might be regarded as slightly positive.

On the whole, the number of choices of the circle, as shown in
table 6, indicate some sort of discrimination. Series 10 was
perfect. Series 25 contains 9 correct choices and several series
reached 70 per cent or 80 per cent correctness. It is not sur-
prising that the chick became discouraged before a perfect habit
was established, for the plan of starting with complex stimuli
and working towards the single detail had not yet been adopted.
A system of control similar to that described in the preceding
chapter was followed throughout the work. The surprising
feature is the high percentage of "right" choices on January 4
and 5 when the chick was becoming discouraged and frightened.
This discouragement resulted in a rush for one stimulus or the
other as though "to get the choice over." It appears that there
was some sort of a difference between the two illuminations
which tended to attract the rushing chick, but this was not
clearly enough perceived for a dependable basis of response.
But even though a discrimination between the stimuli be ad-
mitted, it does not follow that the chick perceived form. As
has been pointed out in Chapter II, the distribution of light on,
or the unequal stimulation of, the different parts of the retina
may have been the discriminable basis.

The result of this experience with chick 9 convinced me that
a better method was necessary to determine with certainty
whether the chick actually perceives form. When the study

60

HAROLD C. BINGHAM

of form perception was begun with group 3, therefore, the
method of the prehminary visual complex was adopted. That
is, form, in the beginning, was made only one of other visual
factors which were gradually eliminated as the experiment pro-
gressed. Only one chick, No. 21, gave anything like positive
results. The work with this chick, including considerable train-

No.

TABLE 6

Form Perception: O 28-1 A 28-f-

9. Hatched: December 1, 1911. Sex:

SERIES

DATE

RIGHT

WRONG

TIME

1

Dec. 9

4

6

2

11

8

2

44 '.2

3

12

5

5

47.1

4

13

6

4

49.2

5

14

7

3

24.8

6

15

6

4

54.6

7

16

6

4

28.3

8

18

4

6

56.8

9

19

5

5

15.9

10

19

10

0

29.5

11

20

6

4

60.2

12

21

8

2

16.7

13

22

7

3

42.7

14

23

5

5

32.8

15

25

6

4

29.5

16

26

8

2

24.2

17

27

7

3

42.8

18

27

4

6

32.8

19

28

4

6

44.7

20

28

6

4

33.2

21

28

5

5

22.0

22

29

3

7

29.3

23

29

6

4

13.3

24

30

7

3

12.7

25

Jan. 2

9

1

27.8

1

2

7

3

30.9

2

3

5

5

25.4

3

4

7

3

12.5

4

5

8

2

4.6 (Rushing)

5

6

(Discouraged; rushed blindly
towards either stimulus)

and wildly

ing in size and brightness vision, represents more than 1500
tests.

The experiment on form discrimination with chick 21 may be
divided into three parts: (1) A preliminary investigation; (2)
reactions to the triangle-circle; and (3) reactions to the circle-
triangle. The method of the visual complex and the general
improvement of my technique enabled me to get much more

VISUAL PERCEPTION OF THE CHICK 61

work out of this subject, and at the same time, called forth less
energy than in the earlier tests. No. 21 knew me quite well by
this time. It was perfectly contented to leave its companions
in the chick-room and 'go alone with me to the dark-room for
work. The chief reason for this was the fact that it always had
its morning feeding in the dark-room. As soon as we reached
the experiment room it was given a little chick food and allowed
to run freely about until the apparatus was made ready for the
tests. During the preparation of the apparatus, the chick
would follow me about the room (then lighted) twittering and
contented, but if it were left alone for a few minutes its dis-
satisfaction was made known by loud and ceaseless peeping.

When everything was ready the chick was placed in the ap-
paratus. At this point the note in its voice changed to a slightly
modified "hovering twitter." Commonly this peculiar sort of
"singing" changed, as the chick entered the discrimination
chamber, to the "food twitter" which was continued all the
time it was inspecting and comparing the two stimulus areas
and, of course, after it had entered the nest box, for there it
was rewarded by finding a few grains scattered in the litter.
In this manner a series could be completed in about fifteen
minutes, after which the chick was taken from the experiment
box and given more food and its freedom in the room. This
procedure could be carried out until the chick's hunger was
satisfied after which the tests went so slowly and with such
uncertainity that it was found best to postpone the work until
the following morning. As the bird became older this mode of
experimentation could be carried on as long as three hours.

While I was preparing test sheets or making notes at my
desk between experiments, No. 21 would crowd about my feet
and "beg" to be taken up. If it were allowed to perch itself
on my arm it would sit there as long as I was quiet, contentedly
preening its feathers and occasionally giving a short "hovering
twitter." All of this indicates that the chick was quite con-
tented and normal throughout the experiment.

Table 7 shows the results of my first study of form with chick
21 in which it was trained to go to the triangle and to reject the
circle or square. While none of the results here recorded are
clear cut, there is strong evidence that the chick was discrimi-
nating between the two stimuli. This experiment was hurried

62

HAROLD C. BINGHAM

lest the chick should not remain in good physical condition, but
at its conclusion the bird was in excellent health, having out of
doors range, so the work was repeated with more thoroughness.
The last two series (24 and 25) of table 7 were introduced as
a different sort of control tests. The circle was larger than the
triangle, hence it afforded an opportunity to see if the chick
would react to form difference, after this training, rather than
size difference. Apparently it was demanding too much of the
chick for it became discouraged at this point and the training
had to be repeated.

No. 21.

TABLE 7

Form Perception

Hatched: February 9, 1912.

Sex: 9

DISCRIMINATION

SERIES 1

3ATE

RIGHT

WRONG

TIME

A 28+— o 7+

6 Ma

rch 13

7

3

10"

" — O 12 +

7

' 13

9

1

11

" — O 15+

8

' 14

8

2

18

" — o 19+

9

' 14

4

6

12

(< «

10

' 14

6

4

8

i( ii

11

' 14

8

2

33

i( «

12

' 14

8

2

12

tt (C

13

' 14

8

2

12

tl il

14

' 14

8

2

12

" - O 23+

15

' 14

8

2

13

(( a

16

' 15

8

2

6

" — O 24+

17

' 15

8

2

13

" — O 25+

18

' 15

9

1

6

" — O 26+

19

' 15

6

4

17

" — O 27+

20

' 15

8

2

18

" — O 28+

21

' 15

7

3

17

" — "

22

' 16

9

1

8

" - D 28+

23

' 16

9

1

40

" — Q 63 +

24

' 18

6

4

56

25

' 18

6

4

85

In table 8 are presented the results of the re-training of No.
21. While the record, as shown in the table, indicates that
the chick discriminated between the triangle and the circle,
it is, nevertheless, equally convincing that the chick did not
perceive the form difference. It had reacted properly to the

A 28 H (D 19+ and — 23+ when the apex of the triangle

was at the top, but when the triangle was inverted (series 20,
March 25), the animal did not choose the triangle more than
half of the time. With the triangle again upright, the correct
reactions returned. Proper responses were made when the

VISUAL PERCEPTION OF THE CHICK

63

areas were equal (series 8 and 9, March 28), but when the
inscribed triangle was substituted for the standard, the chick
was again confused. Series 13, (March 29), another control
where the triangle was inverted, gave more decided negative
results which were immediately preceded and followed by al-
most perfect reactions. As a further test of inverting the tri-
angle, the inversion was made during a regular series, 15,
(March 30). The result of the inversion during tests 5-7 was
one right choice out of three chances with an average time of

TEST SHEET

Title of investigation, Form: A 28H O 63-f-

Experimented on, 21

Harvard Psychological Laboratory, March 18, 1912

Record Sheet, 2; Series, 24

TEST

BEHAVIOR

RECORD

11

/// - 777 + /77 o - 777 + /// .

0-65

2r

/// + IIL,

0-7

31

III - ITT Ai + /// „

E-1-14

41

III -/// + / W + /// n

0-14

51

III - III Ai '"""" 4- /// .

E-1-5S

6r

111--^ III""""" - III o + //-

-/ + //"""- A2+.

E-2-171

7r

/// ^ II - II -V II ^ """' +11 -

"""" + /// .

0-94

8r

II + // W - // W"""' + // «

0-69

91

III + /// .

0-4

10 r

/// - / o + \\ - / + / """ - //

A' + // .

E-1-70

6-4-56.3

39 seconds. These three special tests were preceded and fol-
lowed, during the same series, with perfect reactions (triangle
upright) the average time of which was somewhat less than one-
sixth of the average time for the special tests.

In connection with these results, the details of the behavior
of No. 21 are presented when it was confronted with A 28 +
— o 63+ stimuli. These tests are reported in table 7 as series
24 and 25. The test sheets show how utterly confusing to the
chick was this condition. The results of this test were verified

64

HAROLD C. BINGHAM

in series 10 recorded in table 8 under date of March 28 where
the record is the same and the behavior was very similar.

Finally this chick's response to the circle-triangle was tested.
The reversion of the condition for discrimination was at first
confusing to the chick, and much patience on the part of the

TABLE 8

Form perception: A 28H o 28+

No. 21, Hatched: February 9, 1912. Sex: 9

DISCRIMINATION

SERIES

DATE

RIGHT

WRONG

TIME

A 28+ — O 9+

9

ISIarch 20

10

0

6

" — O 12+

10

ii

20

10

0

14

" — ©15+ (inscription)

of triangle)

11

«

21

9

1

9

<( i<

12

ii

21

10

0

9

(C «

13

a

21

10

0

8

« {t

14

<(

22

10

0

6

" — O 19+

15

It

22

5

5

15

« it

16

tt

23

9

1

6

U 11

17

ti

23

10

0

6

" — O 23+

18

tt

23

9

1

7

<( a

19

tt

25

9

1

20

V 28+ (inverted) — o 23+

20

It

25

5

5

92

A 28+ (upright) — o 23+

21

tt

25

8

2

20

" — O 23+

22

tt

26

10

0

17

" — O 25+

23

tt

26

10

0

12

" — O 27+

24

tt

26

8

2

12

il u

25

tt

26

9

1

6

t( l(

5

tt

27

8

2

35

U l(

6

tt

27

6

4

8

" — O 28+

7

tt

27

8

2

9

(( <(

8

tt

28

10

0

5

tc (t

9

tt

28

10

0

9

A 11+ (inscribed) — o 28+

10

It

28

6

4

40

A 28H o 28+

11

tt

28

10

0

15

« i(

12

tt

29

9

1

5

V 28+ (inverted) — o 28+

13

tt

29

2

5

83

A 28+ (upright) — Q 28+

14

tt

29

9

1

7

" - O 28+

151

tt

30

7

0

6

" — O "

16

tt

30

9

1

7

" — O "

17

it

30

10

0

24

" - o "

18

April

1

10

0

15

iJn series 15 the triangle was inverted during tests 5-7; the result was 2 wrong choices
with an average time of 39 seconds.

experimenter was necessary. The plan now was to begin with

o63H A 28+ and gradually to eliminate the size difference

by substituting successively smaller circles. In this case the
correct stimulus area was the variable and the standard was the
sign of "shock" and "no escape." From the summary of these

VISUAL PERCEPTION OF THE CHICK

65

results, table 9, the early records are omitted since these tests
were made only as a means of getting a discrimination between
the later forms. The size of the circle was gradually decreased
from o 63 + to © 38 + at which point the tabulation begins.
About 400 tests between April 1st and 20th were necessary to
bring the chick to the point in its training at which the records
of the table commence.

Table 9 shows that the chick was able to acquire the circle-
triangle habit. The plan of controlling the qualities other than

TEST SHEET

Title of investigation, Form: A 28H O 63+

Experimented on, 21

Harvard Psychological Laboratory, March 18, 1912

Record Sheet, 2; Series, 25

TEST

BEHAVIOR

RECORD

1 r

/// + //-/// o u - 77 + 77 - 77 p + /" o

-lA'^+ ll„

E-0-145

2r

III + // o """" -111 + lo - llo + lll ,

0-79

3r

III +- 0+//0 """" + ///,

0-45

4r

/// o + /// o """" - / +/// """" + /// „

0-60

+ /// .

51

111(1 - // + /// - /// ' " + /// - // A^

E-1-61

61

II + III - 1 o '""""" + /// „

0-33

II

71

11 + II - III """""""" - /I/O- II + III

0-46

81

III + III - o """"" + 1 - III ^^ + III „

E-1-35

9r

III - /// o"' " - // '""""" "" + /// ,

0-99

_

- A' + .

10 1

1 _ 1 tnnnn _ ^^nnnnnnnnnnn _ ^^nnnnnu,

E-1-243

6-4-84.6

form was carried out as heretofore explained. After No. 21 had

acquired the o 28-| A 28 -f reaction, the order shown in

in the table was followed to determine whether or not the bird
actually depended upon form difference. If it had a perception
of form such as has been found to exist for size, then no marked
change in the results should have occurred when o 28 + was
replaced by a circle that circumscribed or inscribed the standard

66

HAROLD C. BINGHAM

triangle; and if the chick perceived three-sidedness or triangu-
larity on the one hand and circularity on the other hand, there
should have been no important modification of its reaction
when the triangle was inverted.

TABLE 9

Form perception: O 28H A 28+ and O 28H D 28+

No. 21. Hatched: February 9, 1912. Sex: 9

DISCRIMINATION

SERIES

DATE

RIGHT

WRONG

TIME

o 38H A 28+

O 33+ — "
it it

12
13
14

April 20
" 20
" 20

10
6

5

0

4
5

4

7
7

it "

15

" 20

7

3

5

It "

16

" 20

8

2

8

iC tc

17

" 20

6

4

6

it il

18

" 20

10

0

8

Q 28+ - "
O 44+ — "

" — V 28+ ( A inverted)
O 28H A 28+ ( A upright)

19
20
21
22
23

" 20
" 20
" 20
" 21
" 21

10

10

9

5

6

0

0
1
5
4

5
5
5
6
4

tc it .

24

" 21

6

4

7

It it

25

" 21

3

7

6

O 33+ - "
« it

1
2

" 21
" 21

10
10

0
0

4
4

O 28+ - "

" — V 28+ ( A inverted)

O 15H A 28+ ( o inscribed)

O 28+ - "
« It

3
4
5
6
7

" 21
" 21
" 21
" 22
" 22

10
6

4
8
8

0

4
6

2
2

5

5

11

10

7

tt it

8

" 22

10

0

6

tt if

9

" 23

8

2

5

tt it

10

" 23

8

2

5

tt it

11

" 23

9

1

4

it "

12

" 23

9

1

7

it ti

13

" 24

9

1

5

It "

14

" 24

9

1

4

" — D 28+

tt a

16

17

" 24
" 24

5
5

5
5

5
3

it tt

18

" 25

3

7

4

tc it

19

" 25

6

4

4

tt "

20

" 25

5

5

6

it tt

21

" 25

8

2

6

tc tc

22

" 25

5

5

4

The work on form thus indicated that the chick can discrimi-
nate between circles and triangles of equal area, and there are
indications of a discrimination between circles and squares of
equal area (witness series 21, April 25, table 9, 'and series 23,
March 16, table 7), but with the application of control tests
we have indications that this discrimination is on some basis

VISUAL PERCEPTION OF THE CHICK 67

other than form. The results from the Inversion of the tri-
angle indicate that the basis of choice depends upon the unequal
stimulation of different parts of the retina. When the extended
base of the triangle is so placed as to stimulate the region of
the retina which was formerly stimulated by the apex of the
triangle, the chick becomes confused. Under the conditions of
the present experiment, therefore, I am forced to conclude that
the apparent reactions to forms are the result of keen percep-
tion of relative size differences.

The image formed on the retina of the chick is truly changed
when the triangle is inverted by causing certain particular size
reversals. The vertex, which was formerly uppermost, becomes
interchanged with the base. The portion of the retina which
was formerly stimulated by an extended line of light relative
to the base of the triangle becomes stimulated by a mere point
of light which is related to the vertex of the triangle. Inversion
thus causes particular changes in the extension of the stimulated
portions of the retina. It would seem, therefore, that the in-
version of the triangle, though causing no change in form as
such, produces these relative size differences which were evidently
a source of confusion to the chick.

This control of inverting the triangle is a severe test, but it
affords an excellent basis of comparison between the chick and
other birds. Coburn made the same test with his crows with-
out interrupting their correct choices. It therefore seems to be
a legitimate and highly desirable procedure in form studies. In
addition to circle-triangle discrimination, Coburn's crows dis-
criminated a circle from squares and hexagons.

Coburn's crow study again furnishes from the birds the only
results directly comparable with my chick results. Porter tried
to study the perception of forms and designs in sparrows by
using six form boxes made of poplar boards. He secured nega-
tive results but his method was weak. His birds were free to
approach all of the boxes in one of which there was always food.
There was no strong incentive for the bird to seek the correct
box first because it was ultimately rewarded with food regardless
of its first approach. Even had the results been positive
nothing definite could have been asserted about the detail vision
of the birds on account of the lack of any system of control.

68 HAROLD C. BINGHAM

Porter finds that the sparrow and cowbird can discriminate
three horizontal Hnes or a black diamond from a blank card or
from each other. No more weight can be attached to these
results than could be given to his efforts to study form percep-
tion. On account of the widely differing method and apparatus
a comparison of Porter's results with my own would be quite
out of place.

The chick, then appears inferior to the crow in the matter
of form perception. Where the crow appears to react by means
of a relative form difference, the chick reacts to a circle-triangle
situation on the basis of a specific form, or shape, difference.
Responses to this specific form relation are much less readily
acquired by the chick than responses to size differences.

CHAPTER VII
Flicker Perception

Inasmuch as the preceding studies with the chick indicated
unequal visual sensitiveness to the spatial details of size and form,
there arose the question of the importance of the role played by
moving stimuli in the chick's vision. Movement, indicating dan-
ger or food, certainly is of great daily importance to the seeing
animal. The remarkable ability of birds to discover worms has
been plausibly attributed to the visual perception of movement.
Might not the chick, then, demonstrate greater perceptual ability
if a problem were presented demanding reactions to moving
stimuli instead of stationary spatial differences?

To test the chick's sensitiveness to movement, the same dark
room-discrimination method was employed. The stimuli pre-
sented were two 6 cm. circular openings illuminated as in the pre-
ceding experiments. The sources of these illuminations, however,
were attached to an interrupting device by means of which the
lights could be made to flash at different, though regular, intervals.
The positive rate was adopted as once per second. The negative
stimulus was a similar illumination flashing more rapidly.

To produce these interruptions a rotating disc of wood, one
inch thick and four inches in diameter, was used. Bearing upon
the circumference of the disc were two pairs of brass points, each
pair being in the circuit of one of the source lamps. When one of
these pairs of points bore upon a conducting surface, the circuit
was complete ; when it bore upon a non-conductor, the circuit was
broken. The interruptions of the sources of illumination, then,
depended upon an alternation, on the circumference of the disc,
between a conductor and an insulator.

The wooden surface itself provided the insulation. For the
conduction, thin bands of brass were firmly fitted and screwed to
the circumference of the disc. Thus, a brass band extending half
way around the circumference of the rotating disc would provide
a complete circuit half of the time. In other words, the lamp de-
pending for its illumination upon this band of brass would flash

70

HAROLD C. BINGHAM

off and on at regular intervals and the on interval would equal the
off interv^al.

The brass points were grouped so that one pair bore upon the
left, the other upon the right edge of the circumference of the disc.
By attaching to the right edge, for example, a single band of brass
equal to half the circumference of the disc, and to the left edge two
bands of brass each equal to one-fourth of the circumference, the

A; or or

Kevoluing Due
Non Conductor

J)ouble Syy/fch

Fig. 7

Flicker apparatus. Schematic diagram showing how flicker stimuli were
produced. The lamps were mounted on the movable carriages in the same source
compartments described in Chapter III. By means of the double switch shown
in the wiring scheme it was possible to alternate between the two lamps the slow
and rapid flashes. The revolving disk is illustrated in this figure as much too
large when compared with the illustration of the motor. The worm gear reducer,
described in the text, is not shown in the illustration.

lamp connected through the right pair of points would have on
and off intervals twice as long as the intervals of the other lamp.
But, for any period of time, the total of the 07i or off time of the
one lamp would equal the same of the other lamp. Similarly, the
on and off flashes of the one lamp could be divided so that the
intervals were one-third or one-fourth of the intervals of the other
lamp.

VISUAL PERCEPTION OF THE CHICK 71

A double switch was inserted in the connections so that the
relative locations of the slow and rapid flashes could be reversed.
The disc was driven by an "A.C." motor having a commercial
rating of one-eighth horse power and twelve hundred revolutions
per minute. The excessive speed was cut down by means of a
worm gear reducer so that the rotation rate of the disc in most
of the training tests was approximately once per second. The
lamp interrupted by a single insulation, therefore, flashed on and
off once every second. The rate was noted by recording with a
stop watch the time elapsing during 50 flashes. Practically no
variations were observable during the experiment.

Three driving pulleys of different sizes on the motor and three
on the reducer made possible nine different speeds ranging between
50 revolutions of the disc in 162 seconds and 50 revolutions in
39 seconds. The rate of flashing, how^ever, was increased two or
more times depending upon the number of breaks on the circum-
ference of the disc.

There is a definite objection to this method of measuring the
interruptions, but for the present purpose, rough tests indicate
that it may be ignored. Since each on interval heats the filament
of the illuminating lamp, there is a tendency for the illuminated in-
terval to last longer than the dark interval, and as the alternating
intervals become shorter there would be a tendency for the total
time of illumination to become greater than the total time of
darkness. Yet it seems justifiable to ignore this possibility since
the number of alternations is unchanged. Moreover, the glow
of a filament after the "flash-off" seemed to be completely swal-
lowed up by the opal flashed glass which served as a diffusing sur-
face. It was also found that the tungsten lamps responded
quite promptly to the off and on contacts and more promptly than
the carbon lamps.

The training was begun with a one-three discrimination differ-
ence. The chick was trained to choose the slower rate. As in
the spatial tests, a visual complex consisting of greater brightness
added to the slower stimulus was presented at first and gradually
changed to the single detail of flicker. Control of the luminous
flux was gradually introduced by means of the upper illumination,
and, in a few tests, by an additional 2 c.p. lamp located at the
middle of the rear of the experiment box. Further controls were
made after the habit became established, by irregularly varying

72 HAROLD C. BINGHAM

the source distances within a maximum of 40 cm., and by irregu-
larly changing the shifter from left to right.

Chick 27 learned the problem so much more rapidly than the
other subjects that it was evidently ready for the one-two relation
before the others had acquired the one-three habit. In view of
the possibility of the inequality in the two stimuli of the dark-
light alternations, it was deemed desirable to reduce the difference
rates towards the threshold. This would not only indicate the
chick's threshold of difference, but also make less the possibility
of different light-dark ratios serving as the basis of discrimination.
Since early experience had emphasized the importance of concen-
trating the work as much as possible while the chicks were in
satisfactory condition, further work with the other chicks was
abandoned and the mechanism was changed for one-two dis-
crimination. The change seems to have been fortunate, for the
other chicks dropped off shortly, and chick 27 did not last long
enough to complete more than an elementary phase of this par-
ticular task.

The details of the learning of this task by chick 27 are furnished
in Chapter VIII where tables 15 and 16 present quantitatively
several phases of behavior in addition to the criterion of correct
choice per series of ten tests. Table 10 presents the discrimination
conditions, reactions of the chick, and the controls employed.

To determine the specific rates of the two stimuli, the method
was adopted of recording with a stopwatch the time elapsing
during fifty flashes of either stimulus. Thus the time relation is
given as 1-3 or 39" - 13" for the first twenty series. In series
21 the same relative rate was continued, 1-3, but the specific rates
were made slower, the time for 50 flashes being respectively 54
and 18 seconds. The next change, introduced in series 30, January
10th, results in both relative and specific variation. The flicker
relation becomes 1-2 and the respective periods for 50 flashes
are 36 and 18 seconds. Slight changes were made in series 31,
and in series 32 the effect of substituting without additional train-
ing, 80-40 and 20-10 periods was observed.

At the close of series 20, the controls had become rigorous
enough to indicate that the flicker difference, or some aspect of it,
was really perceived. On January 8, at the close of the 26th series,
an unusually severe control was introduced. The upper lamp was
changed from a 4 c.p. to a 16 c.p. illumination, and in addition,

No. 27.

TABLE 10

Flicker Perception

Hatched: December 14, 1916.

Sex: 9

B
1
2

3
4

5

6

7

8

9
10
11
12
13
14

15
16
17
18
19
20

21

22

23
24
25

26 (1)

27 (2)

28 (3)

29 (4)

30 (5)

31 (6)

♦32
34

(7)
(8)
(9)

35 (10)

36 (11)

37 (12)

38 (13)

39 (14)

40 (15)

TIME RELATION

1-3 = 39" - 13'
for 50 flashes

1-3 = 54
for 50

'-18"
flashes

1-2 = 36" - 18"
for 50 flashes

1-2 = 40" - 20"
for 50 flashes

+ RE-

TIME

SPONSES

-

-

5

70"

6

91

9

72

5

15

7

3

6

3

7

4

7

7

4

?

9

16

8

15

5

23

8

17

7

15

9

11

6

12

5

8

8

22

10

23

10

65

7

14

6

2,i

10

7

8

53

10

44

10

51

8

85

8

16

10

24

7

14

10

22

9

11

9

22

9

10

9

8

10

10

10

13

8

5

9

4

Preliminary; no regular pro-
cedure.

Ibid.

Ibid.

Ibid.

+ source 50; — , 250

+ , 70; — , 230. Time taken from
start to choice. Weak shock.

+ , 80; -, 220.

+ , 100
+ ,120
+ ,130

-,200.
- , 200.
-,200.

Upper

Upper

+ , 160; -,200

Irregular variation between 150

and 200.
Ibid.

Right, 160; left, 200.
Right, 200; left, 160.
Right, 160; left, 200.
Ibid.
Right, 200; left, 160,

illumination.
Ibid. First 5 choices correct;

time 12".
Irregular variation.

illumination,
Ibid.
Ibid.

Right, 160; left, 200.
Right, 200; left, 160.
Upper illumination increased

from 4 to 16 c.p. 4 c.p.

lamp at top of middle of

rear end of discrimination

box. Errors in tests 3 and

10.
Ibid.

Rear light removed
Ibid. Errors in 3, 6, 7.

Ibid.

Ibid.

Ibid.

Ibid.

Ibid.

Ibid.

Ibid.

Ibid. Physically weak.

Ibid.

1915
12/21

12/22
12/23

12/25
12/27

12/29

12/29

12/29

12/30

12/31

12/31

1/1/16

1/1

1/1

1/1

1/3

1/3
1/3

1/4
1/4
1/5

1/5
1/6

1/6
1/7
1/7
1/8
1/8

1/8

1/10

1/10

1/12

1/13
1/14
1/15
1/17
1/17
1/18
1/20
1/20

' Tests for specific or relative stimulus difi^erence; see table 17, page 99, and accompany-
ing text. The tests of series 32 (7) for evidence of relativity or specificity were dis-
tributed among the tests of subsequent series.

73

74 HAROLD C. BINGHAM

a 4 c.p. lamp was placed at the rear of the reaction chamber.
This combination of illuminations made the experiment box quite
light. The novelty interfered considerably as is indicated by the
rise in the chick's time for choosing. Out of the entire series,
however, the chick made eight correct choices.

Although the ultimate choices were not seriously affected by
this introduction of novel conditions, the behavior of the chick
was considerably modified. In the now relatively light reaction
chamber there was considerable pecking and searching for food.
Two or three attempts to jump out of the box were made. The
error in test 3 apparently resulted from this pecking and meander-
ing which continued longer than in the two preceding tests.
Finally, the chick seemed completely to lose its orientation, for
it searched the rear portion of the reaction chamber as much as the
forward portion and entered the starting chamber once. The
entrance into the wrong compartment was very similar to the
entrance into the starting chamber. It did not appear as a
definite choice and the chick was not shocked. In the beginning
of the test, two careful approaches were made to the + compart-
ment and the — compartment was apparently ignored. Then
followed this random wandering, jumping, and pecking during
which the chick seemed to loose its orientation.

The choices and behavior of chick 27 indicate that it was able
to discriminate in this series of January 8th. The error in the
last test of the series was apparently due to carelessness. The
ten tests were made within forty-five minutes and only two hours
earlier a series had been completed. The chick was "full" and
probably had no vivid memory of punishment on account of its
high percentage of correctness in past records. Carelesssness
would be wholly natural at this stage of the work.

Even though it be conceded that the discrimination was satis-
factory, there is another point in the behavior which I regard as
having greater significance. In the detailed records (unpublished)
of behavior in this series, it appears that chick 27 made an un-
usually large number of approaches to the edge of the electric
compartments before venturing in. In addition to this, it "peeked
over" towards the exit frequently tossing the head up as though
looking at the top of, or over, the door. This suggested that the
chick was relying upon the flux of light, for under the cross bar
over the forward end of the experiment box, the additional illumi-

VISUAL PERCEPTION OF THE CHICK 75

nations left a faint shadow which was regularly interrupted by
the flashing stimuli. This resulted in a slight but regular alterna-
tion of light and shadow on the side of the electric compartments
immediately above the exits. The interruption on each side, of
course, depends upon the rate of its respective source. This
indicates that the interruption of the flux was the basis of the
flicker discrimination.

It thus appears that the chick readily discriminates, on some
basis, at least a one-two flicker difference. These preliminary
results convince me that here is a most promising field of investi-
gation, especially with birds. Interrupted illuminations seem to
have high stimulating value, making it relatively easy to direct
the attention of the subject to the movement detail. With such
stimuli the amount of aimless wandering seems to be considerably
less than that which occurs with size and form stimuli.

It is hazardous to compare results obtained in this study from
spatial and moving stimuli on account of the improvement in
technique which clearly occurred during early stages of the inves-
tigation. Since flicker perception represents the last stage of the
study it is possible, yet doubtful, that indications of increased re-
sponsiveness to moving stimuli are due to improvement in experi-
mental technique. Possibility of individual differences should not
be overlooked because objective data on flicker perception have
been obtained from a single subject. Moreover, it is not possible
to analyze satisfactorily the visual factors that were operative in
the intermittent stimuli which were discriminated. On the other
hand, subjective judgments support these preliminary data. It
seems probable that we shall find the chick, and possibly the birds,
more responsive to movement than to spatial factors. At any
rate, these results indicate further possibilities of experimentation
in animal behavior which ought to receive consideration.

CHAPTER VIII

Problem Learning

1. Analysis of the problem

The problem which the chick is expected to solve in the experi-
ments that have been described in the preceding chapters may be
regarded, for present purposes as divided into two separate tasks.
The first task, learning the experiment box, is really a maze prob-
lem. The experiment box with its entrance box, discrimination
chamber, electric compartments, exits, and nest boxes — is only
one of various types of maze. The types of maze, commonly re-
garded as such, provide, more or less frequently, choices be-
tween open and blind passage ways. These open or closed
passages are also present in the experiment box, but the two
courses are permanent in the maze proper whereas the corre-
sponding courses in the experiment box are interchangeable.

The principle of interchanging the open and closed passages of
the experiment box demands the simultaneous display of two dif-
ferent stimuli, one signifying ''this way out" and the other "exit
closed." With the introduction of these two differentiating stim-
uli, there appears for the animal the second task of "interpreting
the signs." Before the experimenter can study the animal's
ability to interpret these signs — the power of distinguishing, for
example, a 5 cm. square from a 3 cm. square or a circle from a tri-
angle— it is necessary that the animal should have learned the
principle of the experiment box or maze problem. The solution
of this maze problem provides opportunity for noting the animal's
method of forming habits.

As a matter of economy, the experimenter naturally presents the
signs of "exit" and "closed" in appropriate relation to the course
which the animal is required to follow in learning the initial maze task.
The positive ( + ) sign meaning "exit" is displayed at the begin-
ning of the training series in significant connection with the course
prescribed for the animal to reach the nest box. Similarly, the
negative ( — ) sign meaning "closed" appears contiguously with the
forbidden nest box. With the exception of two preliminary or

VISUAL PERCEPTION OF THE CHICK 77

preference series, such a procedure was followed while the chick's
learning was being studied. The two phases of the general dis-
crimination problem, therefore, cannot be sharply separated, but
throughout the present chapter they will be regarded as distinct,
and referred to as the primary and secondary tasks in the forma-
tion of the discrimination habit.

2. Individual characteristics

On December 21, work was started upon a problem in flicker
discrimination with five chicks belonging to the fourth group.
The chicks were at that time approximately one week old. The
present remarks will be confined chiefly to three individuals,
chicks 24 9 , 25 9 , and 27 9 . The other two chicks shortly suc-
cumbed to the confinement of the laboratory. No. 24 was char-
acterized in my notes as rugged and next in size to No. 25. No.
25, a White Plymouth Rock, remained from the first robust and
oversized.^ No. 27 was somewhat undersized. Chicks 24 and
27 were Barred Plymouth Rocks.

On the first day each of the chicks was allowed to escape ten
times at either exit. This was the first preliminary series. The
positive stimulus was displayed first at the left, and thereafter
alternately at the right and the left. Six tests were made in the
morning and four in the afternoon. During the morning tests the
exits were wide open. In the seventh test, the doors were nearly
closed and thereafter were completely closed. In the eighth,
ninth, and tenth tests and in the series following, the door of the
exit was opened as soon as the chick had chosen the compartment.

On December 22, the second preliminary series was similarly
started. Four tests were given in the morning, three in the
afternoon, and the remaining three in the morning of the follow-
ing day. In these two preliminary series, no preference for either
of the stimuli was noticeable. Table 11 summarizes the choices
with reference to the stimuli which, in subsequent training,
became plus ( + ) and minus ( — ) signs. The table Indicates
that there was no preference for either stimulus.

Certain right or left preferences, on the other hand, are to
be found. Table 12 summarizes these tendencies. No. 27
shows a distinct preference for the left side but chicks 24 and
25 manifest no preference.

^In comparison with the chicks of the group.

78

HAROLD C. BINGHAM

In view of the fact that No. 27, in the secondary task, shows
learning superiority over chicks 24 and 25, these initial pref-
erence differences are suggestive. Some of the peculiar types
of native behavior manifested by these three chicks appear
in table 13. Reference to the table indicates a close similarity
between the behavior of chicks 24 and 25 which differs some-

TABLE 11

STIMULUS CHOSEN

CHICK

+ Stimulus

— Stimulus

15

11

9

23

11

9

24

9

11

25

10

10

27

12

8

what from the behavior of No. 27. The average time of
escape for chicks 24 and 25 during these preliminary series
is respectively 8.0 and 8.6 seconds; for No. 27, it is 21.7 seconds
or two and one half times as great. Further similarities and
differences may be noted in the tendencies of chicks 24 and 25
to waver between choosing the right and left sides, and the
tendency of No. 27 to stick persistently, after the first two
tests, to left choices with only two variations appearing in tests
10 and 13.

TABLE 12

CHICK

RIGHT CHOICES

LEFT CHOICES

15

13

7

23

15

5

24

10

10

25

10

10

27

4

16

On the other hand, table 13 suggests close similarity among
all three chicks with reference to the stimuli. The letter
following the number of the test indicates the side on which
the + stimulus appeared. The columns headed "choices"
show which stimulus and which side was chosen in any test
by each individual. As three times is the greatest number of
successive 4- or — choices, it is quite evident that neither of

VISUAL PERCEPTION OF THE CHICK

79

the stimuli was an original factor determining the choosing.
Thus, among the individual native tendencies that may have
existed with these different chicks as influences in forming
the habits demanded by this problem, the tendency of right
or left choosing appears to be most pronounced.

TABLE 13
Preliminary Series

TESTS

NO. 24 9

NO. 25 9

NO. 27 9

DATE

Choices

Time

Choices

Time

Choices

Time

1 1

2 r

3 1

4 r

5 1

6 r

+ 1
+ 1

24

22
9
9

5
4

+ 1

- 1

+ 1

- 1

+ 1

+ r

6

25
4
4

5
5

+ r
+ 1

+ 1

24
61
33
12
9
5

12/21

a.m.

7 1
■ 8 r

9 1
10 r

+ r

4
6

7
3

— r

+ r
+ 1
+ r

5
31

9
10

+ 1

+ 1
+ r

10
9
6

25

12/21
p.m.

11 r

12 1

13 r

14 1

+ 1

— r
+ r

— r

6

13
6
4

— 1

— r

+ r

— r

5
10

9
10

+ 1
+ r
+ 1

7
26
33
22

12/22
a.m.

15 r

16 1

17 r

- 1

+ 1

+ r

3

7

11

+ r

— r

- 1

7
9
4

+ 1

12
51
11

12/22
p.m.

18 1

19 r

20 1

+ 1
+ r
— r

5
8
8

+ 1

- 1

— r

9

11

7

+ 1
+ 1

25
26
27

12/23
a.m.

Average
time

8.0

8.6

21.7

With the twentieth test the two preliminary series ended
and thereafter the training series began. At this point the
chicks, instead of being allowed to escape where they chose,
were required to escape from the compartment where the +
stimulus was displayed. The animals were aided in releasing
the tripping device, because they lacked sufficient weight; but
the experimenter could not cause a release of the exit door
until the chick stood in the proper place, when a pull on the
silk line released the tension of the lifting spring so that the

80 HAROLD C. BINGHAM

chick's weight released the trip. The chicks readily learned
to approach close to the door of the exit which brought them
upon the tripping device. As the weight of the growing chicks
increased the exit automatically opened when they stepped
on the tripping device. The electric shock was not intro-
duced until the fourth training series.

In the first two series of training, the visual stimuli differed
with respect only to flicker. This was a period of training
during which the chick's task was chiefly to learn the primary
problem. Visual discrimination was encouraged at the begin-
ning of the third series by presenting stimuli differing with
respect to several factors. This third series may be regarded
as transitional between the primary and secondary tasks. It is
the last series during which the time record was made from
start to exit. In subsequent series the time record represents
the number of seconds elapsing between entrance to the discrim-
ination chamber and choice of a compartment. If the choice
happened to be wrong (— ), therefore, considerable time
might subsequently elapse beteeen the choice and the escape,
whereas the time elapsing between choice of the + compart-
ment and escape was regularly measurable only in fractions of
a second. The maze problem, it was assumed, was sufficiently
learned at the end of the third series to permit introduction of
the secondary task of discrimination and correct choosing.

For further comparison of the characteristic? of the chicks
during the preliminary series, table 14 is presented. This
table indicates the number of times right or left sides and +
or — stimuli were successively chosen throughout the first two
preference (preliminary) series. Frequency of change is greater
as the number of successive choices of the same side or the
same stimulus is smaller. The total number of changes is
indicated by the number in the first column which stands on
the same line with the last record in any other column. For
convenience the total number of changes in choice of sides or
stimuli is indicated at the bottom of each column. Persistence
in choosing is assumed to be greater as the number of recorded
changes is smaller. In other words, the fewer the changes
recorded for any individual, the greater is the indication of its
persistence in following up a particular fine of response.

VISUAL PERCEPTION OF THE CHICK

81

TABLE 14

Initial Preferences

Computed from records in the two preliminary series

FREQUENCY OF CHANGE

NO. 24 9

NO. 25 ?

NO. 27 9

Side

Stimulus

Side

Stimulus

Side

Stimulus

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

1 2
r 1
1 3
r 3
1 2
r 3
1 2
r 1
1 1
r 2

+ 1

- 3
+ 1

- 2
+ 1

- 2
+ 1

- 1
+ 1

- 2
+ 4

- 1

1 5
r 3
1 1
r 1
1 1
r 5
1 3
r 1

+ 1

- 1
+ 1

- 1
+ 2

- 1
+ 3

- 2
+ 1

- 1
+ 1

- 2
+ 1

- 2

r 2
1 7
r 1
1 2
r 1
1 7

- 1

+ 2

- 1
+ 1

- 1
+ 1

- 1
+ 2

- 1
+ 3

- 1
+ 1
-- 1
+ 1

- 1
+ 1

Total number of changes

9

11

7

13

5

15

TABLE 14a

Initial Preferences

Computed from records in first three training series

PREQXIENCY OF CHANGE

NO. 24

NO. 25

NO. 27

Side

Stimulus

Side

Stimulus

Side

Stimulus

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

r 2
1 1
r 3
1 1
r 2
1 2
r 2
1 2
r 2
1 3
r 2
1 1
r 1
1 1
r 4
1 1

+ 1

- 3

+ 1

+ 3
+ 1
+ 2
+ 1
+ 5
+ 1
+ 2

r 2
1 1
r 3
1 2
r 4
1 2
r 2
1 1
r 1
1 1
r 1
1 1
r 4
1 1
r 1
1 1
r 1
1 1

+ 1

- 3
+ 1

- 2
+ 2

- 6
+ 2

- 2
+ 1

- 1
+ 1

- 1
+ 1

- 1
+ 5

1 1
r 3
1 4
r 1
1 6
r 2
1 1
r 6
1 2
r 1
1 1
r 1
1 1

- 2
+ 1

- 2
+ 1

- 1

+ 5

- 2
+ 2

- 1
+ 5

- 1
+ 7

Total number of changes

15

16

17

14

12

11

82 HAROLD C. BINGHAM

Table 14a is similarly presented to show evidence of pref-
erences during the first three training series. Perhaps the
outstanding feature of the behavior represented In table 14a Is
the change in preference of No. 27 from side to stimulus. This
change Is noticeable at the beginning of the third training
series when the two flicker stimuli were made more conspicu-
ously different by the addition of other visual differences.
With the introduction of a discrlmlnable visual complex there
occurred a noticeable change in the behavior of No. 27, Indicated
by six side changes and only three stimulus changes, whereas
the behavior of No. 24 and No. 25 did not noticeably change. As
indicated in tables 14 and 14a, No. 27 made comparatively
few changes In choice of sides and many changes In choice of
stimuli throughout the two preliminary series and the first two
training series. From the beginning of the third training
series this relative preference for side suddenly ceased and
persistence in choosing the + stimulus became dominant.
In the last 13 choices recorded for No. 27 in table 14a, only
one stimulus choice is — (wrong). ^

3. Quantification of the learning process

In learning the primary maze task there are certain details
of behavior which are fairly measurable. The possibility of
quantitatively measuring these constant evidences of learning
was suggested when certain gradual changes were observed in
the chick's behavior towards the exit. During the prelim-
inary series, while the exits were open, there is little reason
for suspecting that the stimuli were of any significance to the
chick. After the exits had been closed in these preliminary
series, the stimuli became a factor attracting the chicks to the
electric compartments but the exit door was opened as soon as
the compartment was entered. The new requirement in the
training series was to go Into the compartment towards one of
the stimuli, to turn outwards at the furthermost end of the
electric compartment, and to step upon the tripping device
close to the unopened exit In a dark wall. The door of the
next exit was discernable during preliminary series from the
other portions of the wall, but much less clearly than in daylight
vision.

^ Two + choices occurred at the end of the preceding series.

VISUAL PERCEPTION OF THE CHICK 83

A significant change in this learning process occurred when
the chick, moving in the direction of the stimulus, turned
aside and approached the exit with that peculiar side-to-side-
head-bobbing-and-neck-stretching behavior that is character-
istic of a curious old hen. Pushing close to the exit, the chick
would squat lower and lower with repeated upward bobs of
the head and stretching of the neck as though in anticipation
of the upward movement of the exit door. As the door
opened, admitting to the relatively dark experiment box a
cheerful flood of light, the chick would commonly express
its satisfaction by means of the familiar hovering twitter.
Where food was a rewarding motive, the hovering twitter was
often replaced by a clearly distinguishable feeding twitter,
and frequently the expressed satisfaction was a combination
of the two types of vocalization.

From the detailed records of behavior in each individual
test,3 table 15 has been computed as a measurement of the
learning process in relation to the electric compartments. The
table presents quantitatively the changes in the behavior of the
chicks with reference to the + responses, the outside corners,
the stimuli, and the inside corners. The first column merely
represents the series of ten tests to which any of the quanti-
tative measurements belong. The last column indicates the
date on which the series was made. The table further presents
for comparison the records of chicks 24, 25, and 27. The four
columns arranged under each individual chick are arranged to
show in any series (1) how many correct ( + ) choices were
made, (2) how many times the outside corners were approached,
(3) how many times the stimuli were tried as a place of possible
exit, and (4) how many times the inside corners were similarly
tried. For my method of collecting these data, the reader is
referred to Chapter IV, pages 37 and 38.

1. If the chick profits by preceding experiences in this maze
task, tendencies toward certain definite numerical limits should
appear in each of the four columns presented in table 15. The
number of correct choices, if there be no preference for either
stimulus, should start at approximately five. Advance in the
series should ultimately bring an increase in this number of
correct choices which approaches ten as a limit.

^ For method of recording this behavior see page 36.

84

HAROLD C. BINGHAM

Exclusive of series 3, as reported in table 15, there is very
little evidence of learning before series 10. The high record in

TABLE 15

Measurement of Learning in terms of Correct (+) Responses, Trials at Outside Corners,

Trials at Stimuli, and Trials at Inside Corners

NO.

'4 9

NO. i

5 9

SERIES

DATE

Correct

Outside

Stimuli

Inside

Correct

Outside

Stimuli

Inside

1

4

26

23

6

4

37

22

8

12/23-24

2

5

25

20

7

3

18

11

5

12/25

3

8

15

11

6

7

26

12

4

12/27

4

5

17

6

4

5

18

9

1

12/28

5

5

31

10

4

6

22

18

3

12/29

6

3

25

5

1

8

15

15

5

12/29

7

6

29

11

3

5

18

13

1

12/30

8

5

26

5

3

3

19

1

1

12/30

9

6

17

6

5

5

21

6

1

12/31

10

5

14

1

3

8

14

4

2

12/31

11*

6

12

2

3

3

24

0

3

1/1

12

6

22

1

3

X

0. 27. 9

Average

1

5

22

21

6

4.3

28.3

22.6

6.66

12/23-24

2

7

25

8

6

5.0

22.6

13.0

6.00

12/25

3

9

20

9

6

8.0

20.3

10.6

5.33

12/27

4

5

19

5

3

5.0

18.0

6.6

2.66

12/28

5

7

20

6

4

6.0

24.3

11.3

3.66

12/29

6

6

19

5

1

5.6

19.6

8.3

2.33

12/29

7

7

16

6

3

6.0

21.0

10.0

2.33

12/30

8

7

18

4

2

5.0

21.0

3.3

2.00

12/30

9

4

20

5

3

5,0

19.3

5.6

3.00

12/31

10

9

11

2

4

7.3

13.0

2.3

3.00

12/31

11*

8

14

3

4

5.6

16.6

1.6

3.33

1/1

12

5

19

2

1

5.5

20.5

1.5

2.00

1/1

13

8

14

3

8

1/1

14

7

16

0

3

1/1

15

9

12

0

3

1/3

16

6

17

2

4

1/3

17

5

20

0

2

1/3

18

8

14

0

0

1/4

19

10

12

0

3

1/4

20

10

11

0

0

1/5

21

7

14

1

1

1/5

22

6

14

0

0

1/6

23

10

11

1

2

1/6

24

8

10

0

0

1/7

25

10

10

0

0

1/7

26

10

10

0

0

1/8

* Brightness reversals at beginning of series 11.

series 3 is probably due to a combination of greater relative
brightness with the + stimulus for the purpose of hurrying
the discrimination. Although brightness reversal was not

VISUAL PERCEPTION OF THE CHICK - 85

introduced before the beginning of series 11, it is improbable
that the brightness factor served as a basis of discrimination
in more than two or three series after the third. The bright-
ness difference was rapidly reduced and in view of the bright-
ness threshold as determined by Lashley (reference 21 : Chapter
2), this factor would have been nil after series 5. The maximum
number of correct choices in a series is first reached by chick
27 in series 19. Additional comment upon the appearance of
discrimination evidence will be postponed for discussion in
connection with the shock as a motive in the learning.

2. The number of correct choices offers a measurement of
progress in forming the discrimination habit or secondary task.
Tn the primary or maze task a more reliable unit of measurement
to denote progress is the elimination of useless movements.
Change in the number of trials at the outside corners offers
one means of determining improvement In the experiment
box. To indicate improvement, the number of trials at the
outside corners must begin relatively high and decrease with
ten as the minimal limit. The larger numbers in the early
series may be due either to the chick's failure to step upon the
proper portion of the tripping device, even though the approach
is the -f exit (outside corner), or to choices of the — compart-
ment and consequent trials at the — exit.

Improvement with respect to outside corners Is generally
noticeable from the first. While each individual has "bad
days," the table Indicates a general tendency towards decrease.
The average for the three chicks suggests that the curve repre-
senting the number of outside corner trials would tend to
become a fairly straight line gradually descending towards the
limit If a sufficiently large number of individuals were tested.
This limit is reached by chick 27 for the first time in series 24
and maintained throughout three successive series.

3. Similarly, the number of trials at the stimuli must be
relatively large at the beginning. As the chick learns Its task,
however, the unsuccessful attempts at the stimulus decrease
with considerable rapidity. Chick 25 first eliminates all
attempts at the stimulus In series 11. Chick 27 next reaches
the same stage in series 14. It Is Interesting to note that
perfection In this particular negative task is attained more
rapidly than In the positive task of outside corner trials.

86 HAROLD C. BIXGHAM

4. The remaining factor in this primary learning process is
the number of approaches to the inside corner. The minimal
limit denoting perfection in this task, like the preceding, is zero.
This process of eliminating inside corner trials is also a negative
task and perfection, again, is approached relatively early.
Several of the first ten series contain only a single inside trial.
Absolute perfection, however, is not reached by any chick
before series 18.

The effect of the preliminary series is suggested when one
compares the records of inside approaches with those of outside
and stimuli approaches. The inside trials in series 1 are slightly
less than one fourth the outside trials and one third the stimulus
approaches. The chicks, it would seem, had acquired in the
preliminary series a tendency to go to the outside rather than
the inside corners. Turning to the averages, it appears that
the outside corners are approached, on the whole, a greater
number of times than the stimuli and the inside corners com-
bined. This correctly indicates that the chick, in turning
from an outside corner, often went to some other part of the
box than my list here indicates. Often the turn was towards
the rear of the experiment box followed by a return. Some-
times it was a fairly direct course from one outside corner to
the other. Changes from outside corners to other parts followed
failure to make an adequate adjustment to the tripping device
at the first exit approach and the chicks frequently made
several exit attempts before effecting the release of the exit
door.

The decline in the number of stimulus approaches is the
most abrupt change that appears in any of the four criteria.
This is partially due to the fact that the number starts nearly
as high as that for outside approaches, but the latter cannot
go below ten whereas the stimulus trials have zero for a
limit. The high beginning in number of stimulus trials is
probably due to the fact that the stimuli are the attracting
factors calling the animal to the exit end of the experiment
box. Efforts directed toward the stimulus are therefore more
to be expected at first than efforts at the inside corners. The
most significant feature in these stimulus records, however, is
the abrupt and rather permanent decreases that occur.

VISUAL PERCEPTION OF THE CHICK

87

More suggestive, perhaps, than these numbers representing
total approaches throughout a series, is a measurement based

TABLE 16

Measurement of Learning hi Terms of Correct (+) Responses, First Trials at Outside Corners,

First Trials at Stimuli, and First Trials at Inside Corners of Electric Compartments

NO. 24 9

NO. 25 9

SERIES

Correct

Outside

Stimuli

Inside

Correct

Outside

Stimuli

Inside

1

4

2

6

2

4

6

4

0

2

5

4

6

0

3

10

0

0

3

8

5

5

0

7

5

5

0

4

5

9

1

0

5

6

4

0

5

5

8

2

0

6

2

8

0

6

3

7

3

0

8

1

9

0

7

6

4

6

0

5

0

10

0

8

5

8

2

0

3

10

0

0

9

6

8

1

1

5

9

1

0

10

5

7

1

2

8

8

1

1

11

6

9

1

0

3

10

0

0

12

6

10

0

0

No. 27. 9

Average

1

5

4

5

1

4.3

4.0

5.0

1.0

2

7

10

0

0

5.0

8.0

2.0

0.0

3

9

7

3

0

8.0

5.6

4.3

0.0

4

5

9

1

0

5.0

8.0

2.0

0.0

5.

7

9

1

0

6.0

6.3

i.c,

0.0

6

6

9

1

0

5.6

5.6

4.3

0.0

7

7

7

3

0

6.0

3.6

6.3

0.0

8

7

9

1

0

5.0

9.0

1.0

0.0

9

4

10

0

0

5.0

9.0

0.6

0.3

10

9

8

1

1

7.3

7.6

1.0

1.3

11

8

10

0

0

5.6

9.6

0.3

0.0

12

5

10

0

0

5.5

10.0

0.0

0.0

13

8

6

0

4

14

7

9

0

1

15

9

9

0

1

16

6

9

0

1

17

5

8

0

2

18

8

10

0

0

19

10

7

0

3

20

10

10

0

0

21

7

10

0

0

22

6

10

0

0

23

10

8

0

2

24

10

0

0

0

25

10

0

0

0

26

10

0

0

0

upon first approaches to these various departments. In
table 16 appear measurements computed upon such a basis.
At the beginning, these first approaches suggest, the stimuli

88 HAROLD C. BINGHAM

and outside corners are close rivals for the majority of trials.
Ignoring certain fluctuations, there is a tendency for the outside
corners to gain the supremacy after the first series. At no
time are the first choices of inside corners at all prominent.
Some of the fluctuations which are noticeable in the table will
be discussed in connection with the effects of punishment
upon the learning process.

4. Relatio7i of punishment to learning

Reflections by certain observers upon the method of punish-
ment led me casually to note its significance in the learning
process. In general, my judgment concerning the method is
that it may be made most pernicious in animal training. On
the other hand, it is also a valuable supplement. Whether
pernicious or meritorious depends upon the intelligence with
which it is applied. If the experimenter is careless in its appli-
cation the results of previous training, may be swept away in a
flash. If he shows wisdom in its application he can certainly
increase the efficiency of his technique.

My lesson on the pernicious possibilities of punishment
came early with the first group of chicks. In my eagerness for
results I endeavored to hurry the discrimination habit by
severe^ and repeated shocking for wrong choices. The discrim-
ination was easy, as subsequent results proved, but the number
and severity of shocks were excessive. The result was inevi-
table. The most sensitive chicks became so fearful of the
electric compartments that they only cowered and chirred in
the rear of the discrimination chamber. Thus the subjects
which continued to work were only the less sensitive animals.
It was not a question of the difificultness of the discrimination
problem, therefore, but a lack of intelligence in the application
of the electric shock that brought disastrous results.

Subsequent observations have encouraged the belief that
the value of the shock cannot be determined in terms of ease

^ The term "severe" should not be interpreted as "painful." The severest shock
that could be administered is little more for the human hand than a tickle. The
outstanding feature of the shock as a means of punishment is probably its sudden-
ness and unexpectedness. It is very probable that a chick which chirrs, flees from
the shock, and cowers in a far corner is comparable to a chick which behaves
similarly when a flying hawk suddenly appears. It is not always pain, no doubt,
that calls forth escape behavior.

VISUAL PERCEPTION OF THE CHICK 89

or difficultness of discrimination, but upon the basis of an
optimum dependent upon the sensitiveness of the animal.
Experimental data can determine this optimal punishment
within the limits of the distribution of sensitiveness in any
type of animal. One should expect this optimum to vary not
only with the different types of animals, but also with different
individuals and at various stages of individual development
and progress.

In general, the chick tends to acquire a fairly definite succes-
sion of acts in arriving at the appropriate conclusion of the
series. In this series ther^ may be a choice which involves
entrance, for instance, into the left compartment. If the left
compartment happens to be +, the trial at the left exit may
culminate the series. If the left compartment, on the contrary,
is — , the series of acts must be continued until the right exit
is tried. A certain definite "set" in serial acts tends to occur,
making certain errors a natural consequence.

In terms of neural organization, an integration occurs of
which a portion must be disintegrated and another portion
must be retained and further stamped in. That portion of the
"set" w^hich corresponds to the error must be eliminated and.
If possible, without destroying the appropriate portion of the
"set." Thus, the way to eliminate the error portion of the
series is to shake up the "set" at the place corresponding to
the error, but without shaking up the appropriate portions of
the "set." A shock, then, just severe enough to have only
an immediate effect is the desirable degree of punishment.
Its effect should not spread beyond the error portion of the
neural organization, and yet it should be efficient at that
particular point. An excessive shock may uproot the entire
"set" and destroy all progress that previous training has
fostered. The optimal punishment, then, should be just
enough to disintegrate the undeslred act (or acts) of the series
but not enough to extend beyond that one element in the series
to other valuable acts.

My method of training chicks of the early groups was merely
to adopt what appeared to be a satisfactory shock (tickle)
and accept the chance results it brought. With the three
chicks discussed in this paper, particular attention was given
to the effect of the shock. It appeared that the heavier chicks,
24 and 25, reacted more strenuously than the lightest. No. 27.

90 HAROLD C. BINGHAM

The present observations are not presented as in any way
contributing to the problem of the relative merits of punish-
ment and reward. They are presented for the sole purpose of
showing that the problem can not be disposed of after the
dogmatic method of some writers nor on the basis of the super-
ficial study of certain experimenters. The problem is intro-
duced in this connection because of its obvious relation to the
learning process.

The shock, in my study, was introduced for the first time
at the beginning of series 4. The coil was set at 6 and connected
with a Columbia Dry Cell No. 6. The felt pad on the floor
of the entrance box was water soaked and the floor of the
discrimination chamber was covered with black dry cardboard.
Under these conditions, my records for No. 24 in test 4 and
for No. 25 in test 5 of this series contain this interpretation:
"shock seemed to tickle;" for No. 27 in test 5, where there
were three possibilities of shock, "no response."

The coil was now set at 5 and the floor conditions were left
unchanged. Under this new arrangement. No. 27 made no
perceptible response to the closing of the electric circuit.

In the beginning of series 6, the cardboard on the floor of
the discrimination chamber was water soaked. Following this
change. No. 27 made no obvious response in test 3; in test 4
it turned back immediately upon the administration of the
shock, but there was no violent response. The choices of
No. 25 were fortunate and it escaped with very little punish-
ment in this series. No. 24 was shocked every time an approach
was made into the wrong exit, the total number of shocks
being ten for seven wrong choices.

The behavior of the three chicks clearly indicates that No.
27 was least sensitive to this particular shock. It now remains
to note whether or not there is evidence of a correlation between
the responses to shocks and the disturbances in the methods
of responding to the problem; and furthermore, to determine
if possible the relative effects of such disturbances.

Responses to the electric current were so slight that little
or no effects in series 4 or 5 may with reason be expected. In
test 6, the shock having been increased, possible effects may
be noted. The number of correct choices by No. 24 suddenly
drops to 3. That this change is due to a "shaking up" is
suggested by radically different types of behavior appearing

VISUAL PERCEPTION OF THE CHICK 91

in each individual test after the third. The individual records
in detail cannot well be presented, but reference to them
convinces me that these changes in behavior are related to
the occurrence of the shock. The "shaking up" in each test
where wrong choices occur clearly affects subsequent behavior.
No extraordinary results for No. 24 appear in table 5 prior
to series 10 in which the number of stimuli approaches drops
off abruptly. In series 9 and 10, as presented in table 6, a
change appears with reference to first approaches to inside
corners. The single first inside approach of series 9 occurs in
test 9 following a shock for a wrong choice. My notes in this
connection read:

No. 24 is becoming wary of shocks; has not reached stage, however, of No. 27.
, No. 24 is very sensitive; jumps and screams sUghtly at each punishment; continues

into compartment with sudden haste and jumps when shocked No.

24 is beginning to appreciate that shock 'means something;' No. 27, in addition,
seems to be coming to an understanding of what the shock means.

In connection with test 10, in which an error also occurs,
the following observations are reported:

Approached — compartment very cautiously; entered cautiously after inspect-
ing compartment from two points of view, — one near the partition, the other
near the right side of the experiment box; stopped on receiving very brief shock
(did not jump or scream) ; same touch of key was repeated (no jump or scream) ;
turned out from right side; moved deUberately but continuously in a semi-circle
and entered — compartment near partition; started toward left front corner
(inside), but stopped 6 or 8 cm. from the beginning of the electric wires and
retraced the distance entered; turned sharply around end of partition, entered
+ compartment, and crossed diagonally to + exit.

In series 10, No. 24 made two first approaches to the inside
corners. One of these responses occurs in test 3 following
a shock, the other in test 5 where the inside corner of the com-
partment was first approached but followed by a retreat, a —
choice, and a shock. During the latter half of series 10, It
was observed that the chick became "hasty" in making choices.

Turning now to the tabulated records of No. 25, table 5
shows for series 8 a sudden drop in the number of stimulus
approaches. Here, it appears, is the point where the severity
of the shock succeeded in taking the chick's attention away
from the stimulus. The punishment, as that for No. 24 In
series 9, tears up a "set" by commanding attention that was

92 HAROLD C. BINGHAM

formerly given to other factors. That the effects of the punish-
ment in the training of No. 25 were too far-reaching, is indicated
by the number of + responses in series 8 to 11 inclusive. Series
10 is the only one where a high percentage of correctness ap-
pears, but my notes suggest that the record is misleading.
The notes on this point read: "The high + record is not con-
firmed by the behavior . . . . ; takes either compartment
hastily without trying to get a cue." The time record confirms
the notes. For the series, the average time for choosing was
3.4 seconds. The maximum time in any one test was 5 seconds.

For chicks 24 and 25, therefore^ the behavior indicates that
the effects of the shock were too far-reaching. There occurs
such a shake-up in the responses that the chicks get even
further from the solution of the discrimination problem than
they were several series earlier. For No. 27, on the contrary,'
the shaking-up does not appear to have such negative results.

For No. 27 it is difficult to point to any abrupt, changes in
the measurements provided by tables 15 and 16. In general
there seems to have been a fairly regular development of the
discrimination habit. Instead of taking suggestive numbers
from the tables, therefore, I shall refer chiefly to my notes
and details of behavior.

The first suggestive change in the behavior of No. 27 occurred
in series 9. The following note appears in connection with
test 8 in which there is an error and approximately three times
as much activity recorded as in any other single test of the
series:

Behavior indicates that No. 27 has become upset by punishment; it did not
know what to make of the two shocks this time. For some inexplicable reason it
seemed to want to go to — exit, but after shock, was wary of both compart-
ments; frequently crowded up close but would not risk stepping in. Seems to
be a new stage in learning; recognizes that danger (shock) is connected with these
escape compartments. No other chick has reached this stage.

For test 9, the notes read :

Very careful; behavior bears out above suggestion regarding shock; approached
electric compartment very carefully; stopped the instant the shock came, turned
out, and went directly to the + exit. Suggests that shock has become a sign that
'this is wrong compartment; other one is right.'

The notes in connection with test 10 are similar:

VISUAL PERCEPTION OF THE CHICK 93

Again confirms above notes; each movement was "deliberate;" careful looking
in direction of stimuli; inspected — compartment thoroughly from two positions
close to wires, one being near partition, other near outside of box: + compart-
ment next inspected and entered very short distance; chick stopped and looked
carefully again, then continued slowly but steadily to the exit where it escaped.
The first evidence that No. 27, or any other chick, considers stimuli in choosing;
doubtful if chick understands problem, but it is approaching a solution.

These observations occur at the end of a series in which the
chick made the fewest correct choices during its entire period
of training.

In series 10, chick 27 made 9 correct choices. An error
occurred in test 2. A note in explanation of this error says:

See notes for series 9 (given above); confirms observations there recorded.
Concerning the series, the notes read: The same carefulness and deliberation
appear throughout this series. The bird has evidently discovered some factor in
the last thirteen tests which enables it to respond appropriately.

In connection with series 11, it was observed:

The behavior indicates that it recognizes the stimuli, or some accompaniment)
as representing a signal for choosing; the recognition, as such, however, does not
imply that the ''signs" are fully comprehended; the discrimination habit is started
and it must now be "worn in" by repetition. The problem seems to be recognized
and the chick is putting forth an effort to solve it. The chick is in a very inter-
esting stage of learning.

An interesting comment comparing the stages of learning
attained by chicks 24 and 27 appears in my notes at the close
of series 11.

No. 24 seems to have reached a stage attained by No. 27 at the close of series
9. There is now a tendency to turn back after a shock instead of continuing to
exit; the shock is beginning to have a definite meaning. The significance of the
shock, however, is not so great for No. 24 as it was for No. 27 in the last three
tests of series 9. No. 24 does not always turn back on receiving the shock. No.
27 has not failed to turn back promptly since test 8, series 9. This is probably
due to the fact that the same intensity of shock affects No. 24 more than No. 27.
The shock seems to be optimal for No. 27 but too severe for No. 24.

The low number of correct choices by No. 27 in series 12 is
probably due to the introduction of new brightness conditions
for control purposes. Errors occurred in tests 1 and 7-10.
These new features were added only in tests 7-9. It is question-
able, also, whether an error should be charged in test 10.
After comparing both stimuli, the chick went to the rear right

94 HAROLD C. BINGHAM

corner of the discrimination chamber. From this point it
advanced to a position close to the — compartment, and in
turning, stepped upon the entrance edge of the electric wires,
but stepped off immediately and went directly to the + com-
partment which was promptly chosen. An error recorded for
test 6, series 14, is likewise questionable.

The records of No. 27 in series 12-14 were made in a single
day (Saturday). The chick was allowed to rest over Sunday.
On Monday, series 15-17 were completed. In connection
with the 15th series it was noted that "the characteristic
behavior of Saturday persisted over Sunday." In series 16
the chick seemed to become careless prior to the last three
tests, in which the choices were carefully and correctly made.
Preceding shocks for wrong choices, it appeared, had encouraged
greater care which, it was hoped, would continue throughout
series 17. During the first five tests carefulness did continue
and the responses were correct except in test 1. But during
the last half of the series, four wrong choices occurred.

In series 17, the chick began for the first time to make the
feeding-hovering twitter during the choosing period prior to
its entrance into the electric compartment.

5. Summary

Natively, No. 27 was found to differ from chicks 24 and 25
in manifesting a persistent side preference and in responding
more slowly. Throughout the first three training series, No.
27 maintains an average time for -f choices practically equal
to that of the other chicks. In series 2 and 3, the — choices
of No. 27 are notably low and the average time involved when
these errors occur is much higher than that for the other chicks.
On the whole, it seems that in No. 27 there was a relatively
high persistence combined with a relatively low variability.
These characteristics seemed to remain at a fairly constant
level throughout the period of experimentation. Alternation
between moderate persistence and high variability was charac-
teristic of chicks 24 and 25. The former type of behavior
probably predominated in No. 24, the latter in No. 25. Under
the conditions of this experiment, No. 27 was the most teachable
subject, although in certain phases of the tasks presented it
it did not always excel.

VISUAL PERCEPTION OF THE CHICK 95

The maze habit was found to be measurable with reference
to approaches to (1) outside corners, (2) inside corners, and
(3) stimuli. Measurability of the discrimination habit was
based chiefly upon + and — choices, but certain quahtative
criteria were found to be highly suggestive of stages of learning.

The method of punishment for wrong choices was found to
be both valuable and pernicious. If intelligently used, it
seems, on theoretical grounds at least, to be highly valuable;
if carelessly applied, punishment can radically interfere with
learning. It is my conclusion, although the present study
does not prove it, that the optimal shock is dependent more
upon the sensitiveness of the animal than upon the ease or
difficultness of discrimination.

For adequate descriptions of animal behavior in problems
of discrimination, these results indicate that finer criteria than
right and wrong records should be seriously considered. Detailed
analysis of the behavior of No. 27 reveals, at least in one place,
evidence of discrimination that is not apparent from gross
records of right or wrong choices. Furthermore, the common
practice of presenting to many or to all subjects precisely
identical experimental conditions should be critically considered.
Evidently there is an optimal degree of punishment, for example,
which varies with individuals. It is desirable at least that
mass results be checked with intensive individual observations.

CHAPTER IX

Ideational Behavior

The question whether the chick can react on the general
basis of triangularity-circularity, largeness-smallness, or rapidity-
slowness was suggested primarily by the different positions
regarding ideation taken by Thorndike and Cole. Their
divergent conclusions concerning the evidence of ideas furnished
by infra-human behavior have stimulated considerable subse-
quent work. The possibility of my study furnishing data for
this problem is suggested in the question : does the chick discrim-
inate on the basis of specific stimuli, or on the basis of relative
stimulus difference?

Aside from its ideational bearing, this question has consid-
erable significance also for infra-human threshold experimen-
tation. If it be found possible to train an animal consistently
to choose on the basis of relative stimulus difference, it would
seem desirable to present, interspersed with the training stim-
ulus setting, other pairs of stimuli differing proportionately
by greater and less amounts both above and below the training
pair. Such a procedure would tend to distribute the effect
of training more evenly through a sequence of daily series.
In threshold studies it would make possible a more constant
vividness, in the awareness of the subject, of the problem
which is presented. In the end, the experimenter could be
more certain that the failure of his subject to respond appro-
priately was due to the threshold limit and not to indifference
to the demands of the problem.

Because the material that has been gathered on this question
of relativity or specificity of chick reactions is wholly incidental
to the primary tasks that have been discussed in the preceding
chapters, no attempt will be made to present it in detail. The
result of these incidental tests with forms has been presented
in Chapter VI. The failure of the chick to respond correctly
when the triangle was inverted suggests a negative answer
to this question of relative form difference. Negative results
are also in evidence when the size of the triangle was changed.

VISUAL PERCEPTION OF THE CHICK 97

In similar tests on relative size difference, however, the
results are more positive. These incidental tests were as
easily made with sizes as with forms. After the chick had
been trained to react appropriately to the o 28 + — o 12 +
stimuli, it w^as confronted, without further training, with two
different circles the sizes of which bore the same relation to
each other as that of the former pair. This test was made
with two different pairs of circles. One pair was larger, the
other smaller, than the training pair of circles, but the relative
difference was a constant.

The diameters of the circles used ino 28+ — o 12+ discrimi-
nation are respectively 6 and 4 cm. The standard is thus three
halves of the variable. A habit having been established with
the 6-4 circles, these stimuli were replaced, during five tests,
with 4-3 circles; during five more tests, with 9-6 circles. Nearly
all of the chicks mentioned in Chapter V were tested in this
manner.

The 9 cm. circles used in the tests with the larger pair of
circles was cut from black card board because regulation plates
of that size were not available. The circular opening was
neatly made by boring with a bit through two thin strips of
board placed on each side of the card board. In the dark
room human observers could hardly tell that it was a substi-
tute. Since there was no training with this cardboard plate,
the substitution could have had no influence.

Not all chicks responded similarly to these novel size rela-
tions, but evidently those which were best trained responded
most positively to the relative stimulus difference. With all
of the subjects tested, the total positive reactions to the larger
stimulus considerably exceeded the total choices of the smaller
circle. Chick 21 responded without an error to both the
9-6 and 4-3 situations. Several other chicks responded
perfectly at least to one of the novel pairs of stimuli. On the
whole, the 9-6 test resulted in more correct reactions than the
4-3 test relation.

In brief, these tests with new pairs of circles having approx^
imately proportional size differences indicate that a chick
trained to choose a 6 cm. circle and to reject a 4 cm. circle can
choose, without further training, the 4 cm. circle when presented

98 HAROLD C. BINGHAM

with a 3 cm. circle. Likewise, it can choose a 9 cm. circle
•circle when presented with a 6 cm. circle. In the one test,
what was formerly the sign for rejection is accepted as a positive
sign; in the other, what was the positive sign is rejected as the
-shock sign. Choices, it would thus seem, can be made on the
basis of relative stimulus difference.

In the preceding paragraph, it has been said that the chick
€an choose on the basis of relative size difference. In my
preliminary account (Chapter II: 2, p. 113), the statement is
made that the chick will choose on this basis. Obviously,
my earlier assertion is too loose, for all chicks did not react
unmistakably in this manner. It would seem wholly safe, in
view of the number that did respond positively, to say that
the chick can react on the basis of size relationship. Whether
it will so respond probably depends somewhat upon the degree
of previous training.

Coburn's investigation of the crow's reactions to relative
stimulus difference is hardly comparable with my work on the
chicks. Before the relative difference tests were made, the
crows had been trained with 5-2, 9-5, 5-3, 3-2, and 6-Z combi-
nations. On August 25th, the size training ended with the
6-3 combination. On the following day, "relative reactions"
to the 3-2 combination were tested with positive results.
This was followed by 6-4 training and positive results for the
"9-6 "relative reactions." Again the 6-4 training was followed
by positive results for the 3-2 "relative reactions."

In comparison with the preliminary training of my chicks,
the crows had received a wide variety of experience with dif-
ferent circle combinations before the tests for relative reactions
were made. The crows had actually received training with
the 3-2 combination which makes me inclined to question the
propriety of Coburn's conclusion:

The results of these experiments indicate fairly clearly the relativity of the

crows' reactions For example, on August 24th and 25th, when the

3 centimeter circle was presented with the 5 centimeter circle, the crow reacted
to the 3 centimeter circle thirty-seven times negatively and three times posi-
tively. On August 26, the 3 centimeter circle, displayed with the 2 centimeter
circle, was reacted to positively in every case.

Coburn seems to overlook the fact that this same crow had
been specifically trained on the 3-2 relation from August 12th
to 17th.

VISUAL PERCEPTION OF THE CHICK

99

Furthermore, Coburn's 9-6 relative reaction tests had been
closely approximated in training when the 9-5 combination had
been employed. For this reason his additional comment again
seems questionable: "The results for crow No. 1 with the 6
centimeter circle when displayed with the 4 centimeter and the
9 centimeter circles, on August 26th and 27th were almost as
decisive." While there had been no specific training on the 9-6
combination as such, there had been, aside from its close ap-

TABLE 17

Reactions to Relative Flicker Difference

Chick 27 9

DATE

STIMULI

NUMBER OF

TESTS

NUMBER
CORRECT

CONTROL SERIES 7

1/12

40"

- 20" (50 flashes)

5

5

20

- 10 " "

1

1

1 (faster)

40

- 20 '* "

5

5

80

- 40 " "

1

1

1 (slower)

1/13

40

- 20 " • "

5

5

80

- 40 " "

1

0

2 (slower)

40

- 20 " "

5

4

20

- 10 " "

1

0

2 (faster)

1/14

40

- 20 " "

5

4

80

- 40 " "

1

0

3 (slower)

40

- 20 " "

5

5

20

- 10 " "

1

0

3 (faster)

1/15

40

- 20 " "

5

4

20

- 10 " "

1

0

4 (faster)

40

- 20 "

5

5

80

- 40 " "

1

1

4 (slower)

20

- 10 " "

1

1

5 (faster)

80

- 40 " "

1

1

5 (slower)

proximation, frequent changes from one relation to another which
might have had the effect of a "relative difference preparation."
It seems probable that Coburn's conclusions are correct but
they are hardly justified on the basis of his experimental
method alone.

My results on the reactions of the chick to relative rates of
flicker are less positive than the results from the size tests.
The method in the flicker study was slightly changed. As
table 17 indicates, the chick's training was continued daily
while two relativity tests were inserted in each series consisting
both of faster and slower rates of flashing. The first day the
chick chose the slower rate regularly in all training tests and in
both of the relative flicker tests. On the second and third days

100 HAROLD C. BINGHAM

it responded negatively in both of the relative stimulus tests.
On the fourth day, four of the relative reaction tests were made,
of which three were correct. The final result for these special
tests was five correct and five wrong reactions. In the faster
relative combination, there were only two correct choices; in
the slower, there were three positive responses out of five.
In each of the first relative flicker tests the responses were
correct.

The results of the tests on the relativity of the chick's choices
is thus different for each of three kinds of stimuli. With forms,
it was so difficult to train subjects that the results are rather
meager, but as far as they go they seem to be negative. With
sizes the results are fairly positive. With flicker perception,
again, the data are too meager to warrant a conclusion on the
relativity problem. In general, it seems that with certain
stimuli the chick can respond on the basis of relative stimulus
difference.

CHAPTER X

General Conclusions and Summary

The results of the preceding study can leave no doubt as to
the relative importance of size and form in the chick's visual
life. It has been found that the number of tests necessary
for the formation of a discrimination habit is much greater for
forms than for sizes. Moreover, "there was little difficulty
in finding chicks that could readily discriminate sizes, but
subjects that could discriminate forms, even after long training,
seemed to be less plentiful. This study has also revealed
the fact that even after perfect reactions to the circle-triangle
are established, the form element as such plays no part in the
discrimination.

The relative discriminative value of size and form was further
tested by removing from the visual complex all but one of
these factors and comparing the effect of different eliminations
upon the chick's choice. The remainder of the complex,
exclusive of size and form, was treated as a single factor and is
spoken of as "brightness and general illumination." In table
18 the result of this special test is summarized from the records
of chicks 15, 16, 17, 18, 19, 20, and 21. The averages of this
table indicate that size was first in importance. Correct
choices (R) on the basis of size difference alone, series 14, are
S6 per cent of the total chances. The complex remainder
called "brightness and general illumination" stands next, series
15, with nearly 70 per cent of correct choices. The average
record is lowest when form provided the only discriminable
basis, series 12, with correct and wrong choices nearly equal. It
thus appears that the relative value of special factors in the
chick's vision, arranged in the order of their importance, are
size, the complex of brightness and general illumination, and
form.

No certain comparison between the role of flicker and that
of size and form can be made on the basis of the present study.
However, on the strength of casual observation, it appears

102

HAROLD C. BINGHAM

that flicker stimuli have a greater stimulating value for the
chick than spatial stimuli. The perception of movement offers
a most promising field for subsequent study of vision in animals.

Relative

TABLE 18
Values of Visua

' Factors

DATE

SUBJECTS

CONDITION OF
DISCRIMINATION

15

16

17

18

20

21

R

W

T

R

w

T

R

w

T

R

w

T

R

w

T

R

\v

T

R

w

T

A 28+
brighter
and in
lighter com-
partment
than 07+

1

Feb. 20

10

0

34

8

2

40

9

1

16

8

2

40

7

3

40

9

1

10

8^

30

2

22

10

0

8

8

2

10

9

1

10

10

0

5

7

3

ii

9

1

18

8f

U

14

3

24

9

1

8

8

2

21

10

0

7

10

0

8

8

2

12

10

0

5

9

1

lOi

4

26

10

0

3

6

4

35

10

0

6

9

1

4

9

1

10

10

0

3

9

1

lOi

5

27

8

2

11

7
9

3

1

24

7

9
10

1

0

22
6

7
10

3
0

16

4

9
10

1
0

21

7

6
10

4
0

12
3

7!
9§

2^

1
3

171

6

28

9

1

7

51

7

29

9

1

14

9

1

4

10

0

5

9

1

6

10

0

8

10

0

2

n

1
2

61

8

29

10

0

3

9

1

9

10

0

2

n

^

4f

9

Mar. 1

10

0

5

10

0

20

10

0

7

10

0

10!

10

1

10

0

12

10

0

12

Form and size

11

1-2

8

2

8

9

1

22

6

4

4

8

2

10

10

0

4

9

1

2

81

If

8f

Form

12

2

3

3

—

3

3

—

3

3

—

2

4

—

4

2

—

4

2

—

31

2|

All factors
combined...

13

4

10

0

6

9

1

45

10

0

4

7

3

36

10

0

4

10

0

3

n

2
3

161

Size

14

4

9

1

7

9

1

30

10

0

3

8

3

4

9

1

3

7

3

6

8§

u

8t

Brightness
and general
illumina-
tion

15

5

5

5

9

6

4

26

9

1

5

7

3

3

7

3

7

7

3

5

6|

3^

9§

It seems probable that we shall find in birds considerable
acuity for moving stimuli.

In studies of the chick's vision, prior to the investigation
of spatial perception which has been reported in the preceding

VISUAL PERCEPTION OF THE CHICK 103

pages, it is quite probable that the reactions observed were
made, not to the visual details which the experimenters sought
to present to the chick, but to unsuspected visual complexes ,
similar to the "complex remainder" which is here designated
as "brightness and general illumination." It is desirable that
subsequent studies further analyze our visual complex and
determine the visual acuity of the chick with each element as
has been done with size and form.

When presented with a standard circle 6 cm. in diameter,
it has been found that a variable circle 4.5 cm. in diameter
can be discriminated by the chick. If one were to increase
the size of the variable by increments no greater than one
millimeter, it is probable that the chick would be able to
discriminate a larger variable than the 4.5 cm. circle. The
threshold of difference with a standard 6 cm. circle, therefore,
is certainly one-fourth and perhaps less. In the perception
of size the ability of the chick is evidently inferior to that of
the crow.

Earlier experiments on the chick's perception of forms have
failed to eliminate all discriminable possibilities other than the
factor of form. The chick can discriminate between circles
and triangles and circles and squares which are equal in area,
but, with the conditions as described in this study, none of
the subjects were able to discriminate between visual stimuli
on the basis of form alone. Reactions to visual stimuli which
have been interpreted by observers as indicating form discrimi-
nation have probably been made on the basis of unequal
stimulation of different parts of the retina. If local inequality
of excitations be the basis of these reactions, then the apparent
discrimination of form by the chick is, in reality, a keen percep-
tion of size differences. In the perception of form, also, the
ability of the crow seems to surpass the ability of the chick.

The chick evidently has a keener perception of flicker than
of form stimuli. The present experiment indicates that it
perceives at least a two to one flicker relation. How much
lower the threshold may be, this study has not indicated.

The discrimination method employed in the present investi-
gation furnishes a means of recording the learning process of
the chick. An essential part of the discrimination habit is a
maze problem. The formation of this maze habit, it was
found, can be quantitatively measured in terms of approaches

104 HAROLD C. BINGHAM

to outside and inside corners and to stimuli. The most reliable
measurement of the discrimination habit itself is the number of
correct choices. In addition to these quantitative measurements,
certain qualitative criteria were found to be highly suggestive
of stages of learning.

The method of punishment was found to be both valuable
and pernicious. If intelligently used, it seems to be highly
valuable; if carelessly applied, punishment can radically inter-
fere with learning. The maximum value of punishment in
chick work evidently depends upon the alertness of the experi-
menter in selecting an optimal punishment for each subject.

A chick can acquire a perfect circle-triangle or circle-square
habit, but control tests indicate that it has no general idea of
circularity In contrast with triangularity or squareness. On
the other hand, a large-small trained chick can react positively
to the larger of two stimuli even though this particular stimuli
had been the shock sign of the training combination. Also,
the positive stimulus of the training pair, when presented with
a larger circle, may call forth negative reactions. Positive
results on the question of relativity were not secured from the
flicker experiments.

The order of importance of the visual factors which have been
studied, It would seem, Is size or flicker, the brightness and
general Illumination complex, and form.

Bibliography

A selected bibliography on the origin and history of the domestic fowl appears
on page 5. A complete bibliography of experimental work with the domestic
fowl to 1916 appears on page 20.

EDITORIAL ANNOUNCEMENT

The next number will complete volume 4 of the Behavior Mono-
graph Series. Hereafter the series will appear under the title COM-
PARATIVE PSYCHOLOGY MONOGRAPHS. The new series will
be under the editorship of Professor Walter S. Hunter of the Univer-
sity of Kansas, with the co-operation of Professor H. A. Carr of the
University of Chicago, Professor S. J. Holmes of the University of
CaHfornia, Professor K. S. Lashley of the University of Minnesota,
and Dr. Robert M. Yerkes of the National Research Council.

Manuscripts may be sent to Professor Hunter at Lawrence, Kansas,
or to any other member of the editorial board.

The new monograph series will be broader in scope than the old
one and, in addition to studies in animal behavior, will publish work
in human psychology conducted from the comparative point of view.

Comparative Psychology Monographs

Volume 1

1. Studies of the reliability of the problem-box and the maze. By
W. T. Heron and W. S. Hunter.

Price per volume of 500 pages, net postpaid, $5.00, United States; $5.25,
Canada; $5.50, other countries.

Published by
WILLIAMS & WILKINS COMPANY

Mt. Royal and Guilford Avenues
Baltimore, U. S. A.

The Behavior Monographs

Edited by JOHN B. WATSON
The Johns Hopkins University, Baltimore, Md.

VOLUME 1

No. 1 The development of certain instincts and habits in chicks. By Ferderick S. Breed.

Pp. iv + 78, $1.00, postpaid.
No. 2 Methods of studjdng vision in animals. ByRobertM.YerkesandjJohn B.Watson.

Pp. iv + 90, $1.25, postpaid.
No. 3 An experimental study on the death-feigning of Belostoma (-Zaitha Aucct.) flumin-

eum Say and Nepa apiculata Uhler. By Henry H. P. Severin and Harry C.

Severin. Pp. iii + 47, $.65, postpaid.
No. 4 The biology of Physa. By Jean Dawson. Pp. iii + 120, $1.50, postpaid.
No. 5 The function of the vibrissae in the behavior of the white rat. By Stella Burnham

Vincent. Pp. iv + 81, $1.15, postpaid.

VOLUME 2

No. 6 The delayed reaction in animals and children. By Walter S. Hunter. Pp. v + 86,

$1.15, postpaid.
No. 7 The Canada porcupine: a study of the learning process. By Leroy Walter

Sackett. Pp. iii + 84, $1.15, postpaid.
No. 8 Audition and habit formation in the dog. By Harry Miles Johnson. Pp.

iv + 78, $1.00, postpaid.
No. 9 Habit formation in a strain of albino rats of less than normal brain weight. By

Gardner Cheney Basset. Pp. iv + 46, $.65, postpaid.
No. 10 Distribution of effort in learning in the white rat. By John Linck Ulrich. Pp.

iii + 51, $.65, postpaid.
No. 11 The effect of age on habit formation in the albino rat. By Helen B. Hubbert.

Pp. v + 55, $.65, postpaid.

VOLUME 3

No. 12 The mental life of monkeys and apes: a study of ideational behavior. By Robert
M. Yerkes. Pp. iv + 145, $1.50, postpaid.

No. 13 A study of perseverance reactions in primates and rodents. By G. V. Hamilton.
Pp. iv + 65. $.75, postpaid.

No. 14 An analysis of the learning process in the snail, physa gyrina Say. By Elizabeth
Lockwood Thompson. Pp. iii + 97, $1.25, postpaid.

No. 15 The effect of length of blind alleys on maze learning: An experiment on twenty-
four white rats. By Joseph Peterson. Pp. iii + 53, $.75, postpaid.

No. 16 An experimental study of transfer of response in the white rat. By Harry H.
Wylie. Pp. iii + 66, $1.00, postpaid.

VOLUME 4

No. 17 Transfer of training in the white rats upon various series of mazes. By Rutledgc

T. Wiltbank. Pp. iii + 65, $1.00, postpaid.
No. 18 Redintegration in the albino rat. By William Brockbank, A.M. Pp. iii + 66,

$1.00, postpaid.
No. 19 Discrimination of light of different wave-lengths by fish. By Cora D. Reeves.

Pp. iv -f 106, $1.25, postpaid.
No. 20 Visual perception of the chick. By Harold C. Bingham. Pp. 00 + 000, $1.00,

postpaid.

Price $5.00 per volume, net postpaid

May be ordered from

WILLIAMS & WILKINS COMPANY

Mount Royal and Guilford Avenues

Baltimore, U. S. A.

Behavior MonogI^aphs

Volume 4, Number 5, 1922 Serial Number 2 1

Edited by
WALTER S. HUNTER

University of Kansas

Heredity of Wildness and Savageness in

Mice

BY
CHARLES A. COBURN

Published by
WILLIAMS & WILKINS COMPANY
Mount Royal and Guilford Avenues
Baltimore, U. S. A.

The Journal of Comparative Psychology will accept for publication, studies which
contribute to our knowledge of mental function and behavior in any organism.

Manuscripts may be sent to Knight Dunlap, The Johns Hopkins University, Baltimore,
Md.; to Robert M. Yerkes, 1701 Massachusetts Ave., N. W., Washington, D, C; or to any
other member of the editorial board.

Twenty-five reprints of articles will be furnished gratis to contributors when ordered in
advance. An order slip, with table of cost of additional reprints, is sent with proof.

The Journal of Comparative Psychology is published every other month, appearing
February, April, June, August, October and December. At present, one volume of approxi-
mately 500 pages is issued a year. Subscription is by the volume only, and not by the
year. The Comparative Psychology Monographs are issued irregularly.

Correspondence concerning business matters should be addressed to Williams &
WiLKiNS Company, Publishers of Scientific Journals and Books, Mount Royal and Guil-
ford Avenues, Baltimore, U. S. A.

Subscription Price: Current Volume: Volume II, United States and dependencies,
Mexico, Cuba, S5.00; Canada, $5.25; other countries, $5.50. These rates apply also to the
Comparative Psychology Monographs. Prices are net, postpaid. Foreign subscriptions to
the current volume accepted at par of exchange.

Back Volume: Vol. I: Price, $6.00, United States, Mexico, Cuba; $6.25, Canada; $6.50
other countries. Prices are net, postpaid.

Back Volumes: Complete sets of Psychobiology (Volumes I and II), of the Journal oj
Animal Behavior (Volumes I to VII, inclusive) and of the Behavior Monographs (volumes
I to IV), may be obtained from the publishers of the Journal of Comparative Psychology
at the price of $5.00, United States and dependencies, Mexico and Cuba; Canada, $5.25; other
countries, $5.50 per volume. Prices are net, postpaid. Foreign subscriptions to the back vol-
umes accepted at the current rate of exchange.

Single Volumes, when available, may be obtained at $5.00 a volume. The complete
sets are not broken to supply single volumes.

Subscriptions are received at the following addresses:

For Argentina and Uruguay: Beutelspacher & Cia., Sarmiento 815, Buenos Aires,
Argentina.

For Australia: Angus & Robertson, Limited, Sidney; Stirling & Co., 317 Collins St.,
Melbourne.

For Belgium: Henri Lamertin, 68 Rue Coudenberg, Bruxelles.

For the British Empire, except Australia and Canada: Order from your own agent,
or from any British bookseller, or order direct from the publishers.

For Canada: Wm. Dawson and Sons, Ltd., 87 Queen Street, East, Toronto, Canada.

For Denmark: H. Hagerup's Boghandel, Gothersgade 30, Kovenhavn.

For France: Emile Bougault, 48, Rue des Ecoles, Paris.

For Germany: R. Friedlander & Sohn, Buchhandiung, Carlstrasse 11, Berlin, NW. 6.

For Holland : Scheltema & Holkema, Rokin 74r-76, Amsterdam, Holland.

For Japan and Korea: Maruzen Company, Ltd. (Maruzen-Kabushiki-Kaisha), 11 to 15
Nihonbashi Tori-Sanchome, Tokyo; Fukuoka, Osaka, Kyoto, and Sendai, Japan.

For Spain: Ruiz Hermanos, Plaza de Santa Ana, 13, Madrid.

For the United States and other countries except as above: Williams & Wil-
KiNS. Company, Mount Royal and Guilford Avenues, Baltimore, U. S. A.

No claims for copies lost in the mails will be allowed unless such claims are received
within 30 days (90 days, foreign) of the date of issue. Claimants must state that the publi-
cation was not delivered at their recorded address. The publishers will not be responsible
for loss due to change of address unless notification is received at least one week in advance
of issue.

Behavior Monographs

Volume 4, Number 5, 1922 Serial Number 2 1

Edited by
WALTER S. HUNTER

• University of Kansas

Heredity of Wildness and Savageness in

Mice

BY

CHARLES A. COBURN

Published by
WILLIAMS & WILKINS COMPANY
Mount Royal and Guilford Avenues
Baltimore. U. S. A.

CONTENTS

PAGE

INTRODUCTION 1

MATERIAL AND METHODS 2

EXPERIMENTAL RESULTS 14

A. General comparison of inheritance in the hybrid generations of both

series 14

B. Effect of age, frequency and number of tests on the lowering of the

grade of wildness and savageness in the successive tests 23

C. Results from crossing wild females with tame males in comparison

with results from crosses of tame females with wild males' 38

D. General summary of individual inheritance and the results of selec-

tive breeding in the hybrid generations of both series 45

OBSERVATIONS ON "SINGING" IN MICE 63

SUMMARY AND CONCLUSIONS 66

BIBLIOGRAPHY -. 68

The writer wishes to acknowledge his indebtedness to Professor Robert M.
Yerkes, without whose valuable suggestions and aid this study would not have
been possible, and to Viola Hunter Coburn for assistance in the compilation of
the tables.

HEREDITY OF WILDNESS AND SAVAGENESS IN

MICE

INTRODUCTION

For many years, especially since the appearance of Galton's
work on the inheritance of mental abilities (1869), students of
heredity have recognized the possibility that mental character-
istics are inherited as well as the physical and, perhaps, in a
somewhat similar way. The later studies of Galton (1897),
and those of Woods (1906), Pearson (1903), Ellis (1904), Bentley
(1909), Davenport (1896), Goddard (1911), Conklin (1915),
and others tend to substantiate the hypothesis of the inherit-
ance of mental traits.

All of these studies, however, pertain to the heredity of mental
or moral characteristics in mankind, and the data from which
the studies were made were obtained from various historical
and observational sources. In 1910 Professor Robert M.
Yerkes began an investigation of the hereditability of savage-
ness, wildness, and timidity in rats (1913). So far as the writer
has been able to ascertain this was the first study in mental
heredity to be made in a laboratory where the conditions under
which the subjects lived could be controlled to some extent.
The rats used in this study numbered 300 and consisted of
wild rats, tame rats, first generation hybrids obtained from
crosses of the wild w4th the tame rats, and the second genera-
tion hybrids procured by crossing the first generation among
themselves. The investigation, though incomplete, proved
conclusively that these three characteristics, wildness, savage-
ness, and timidity are heritable in rats.

In order to gain more definite information concerning the
modes of transmission' of such behavior-complexes as the above,
the writer, at the suggestion of Professor Yerkes, undertook the
investigation described in this paper. Owing to the limited
space in the Harvard Psychological Laboratory mice were
used instead of rats. The study was begun in December, 1911,
and continued without interruption until May, 1914. The

2 CHARLES A. COBURN

number of mice used was about 1300 and consisted of wild
mice, tame mice, and hybrids of the first, second, and third
generation.

Professor Yerkes mentioned in the report of his investiga-
tion the difficulty he experienced in always being sure he was
distinguishing between wildness and timidity in rats. He
says: "It is indeed extremely doubtful whether it can with
sufficient certainty be distinguished from wildness to render
measurements significant." The characteristics of rats are
so different from those of mice that it would be all the more
difficult to make this distinction in the case of the latter. For
this reason timidity was omitted in this study and only the two
behavior-complexes, wildness and savageness, were investigated.

MATERIAL AND METHODS

From the beginning of the work in December, 1911, until
June 7, 1913, the mice were kept in one of the large, well-venti-
lated rooms of the Harvard Psychological Laboratory. Upon
the latter date the mice were taken to the Franklin Field Station
at Franklin, New Hampshire, where the excellent facilities
allowed even greater progress during the summer than in the
quarters at Harvard although in no way were the methods of
the experiment changed. Upon the return from the Franklin
Field Station in October, 1913, the study occupied one of the
rooms of the newly-completed Harvard Laboratory of Animal
Psychology where it remained until the completion in May,
1914. This room was well lighted through the glass ceiling
but possessed no direct connection with the outside, hence had
to be ventilated by a system of fans. This, however, seemed to
have no effect upon the work as the mice throve equally as
well here as in the former quarters.

In the beginning the mice were kept in the large cages used
by Professor Yerkes in his work with the dancing mice (1907)
from the report of which a description may be obtained. Later,
when the large number of mice compelled the use of more cages,
the type of cage used by Professor William E. Castle and his
associates at the Bussey Institute, Forest Hills, Mass., was
adopted. A description of this cage, to the knowledge of the
writer, has never been printed. It is made entirely of woven
wire of one-quarter inch mesh and is, in form, a truncated rec-

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE 3

tangular parallelepipedon, the truncation beginning at a line
in the top 8 cm. from the back side and extending to a Hne
in the front side 8 cm. from the bottom. The bottom of the
cage is 30 cm. long by 21 cm. wide; the back, 21 cm. wide by
21 cm. high; the portion of the top parallel with the bottom
and the portion of the front parallel with the back are each
8 cm. by 21 cm. The slanting surface comprising the remain-
ing part of the top and front is 26 cm. long by 21 cm. wide. The
two remaining sides of the cage are five sided, the measurements
of each being 30 cm. by 21 cm. by 8 cm. by 26 cm. by 8 cm.
In the slanting surface, 8 cm. from the top and bottom and
5.5 cm. from each side, is the door 10 cm. by 10 cm. The lid
is 12 cm. by 12 cm. hinged at the top.

The very convenient watering system in use at the Bussey
Institute was also adopted. It consists of a 10 or 12 ounce
bottle fitted with a perforated rubber cork through which is
passed a small glass tube bent at an angle of about 120 degrees
so that when the bottle is placed upon the slanting surface
of the cage the tube will extend down through the wire of the
fore part of the cage to within easy reaching distance from the
floor of the cage. The lower end of this tube is slightly drawn
together so the water will not flow of its own accord. The
mice soon learn to drink drop by drop from the tube. Thus
there is a constant supply of clean water without the evil effects
of a wet cage as when the water is placed in a receptacle on the
floor or side of the cage. The bottle serves also as a weight
on the lid of the cage making it impossible for the mice to escape
in this way.

When in use these cages were placed upon wooden shelves
w^hich were covered with sufficient sawdust to well cover the
bottom of the cage. This sawdust would remain on the shelf
when the cage was raised and so could be replaced with fresh
sawdust without greatly disturbing the mice.

Shredded tissue paper was used as bedding for the mice and,
since it was the purpose of the experiment to minimize as much
as possible the effect of the presence of the experimenter while
cleaning cages, feeding or testing the mice, there was always
sufficient paper in the cage to enable the mice to hide in it if
they so wished. Until all the tests were made the mice were
seldom handled or otherwise greatly disturbed except when
being tested.

4 CHARLES A. COBURN

It was found that the Bussey Institute cage was very much
more adaptable for the breeders than for testing purposes
hence when the mice were weaned they were separated accord-
ing to sex and placed in the larger cages and the subsequent
tests made from these. Only occasionally, when the number
of mice being tested was too great for the supply of large cages
were the smaller cages used in this way.

The food for the mice consisted of cracked corn, oats, sun-
flower seed, bread moistened with milk, and occasionally a
bit of lettuce or other green food. In the beginning some diffi-
culty was experienced in so regulating the food to keep the
mice, especially the breeders, in the best condition. The first
few months' experience, however, was sufficient to obviate
this difficulty.

The method of testing used by Professor Yerkes was
followed in this study as closely as was possible. The mice
were classified according to the six grades, namely, 0, 1, 2, 3,
4, 5. The grade 0 indicates the absence of all evidences of
wildness or savageness, and the grade 5 indicates the presence
of all the various signs of wildness or savageness in the greatest
number and intensity. After a few mice were tested definite
standards which represented the various grades were fixed
and there is fairly definite evidence to indicate that these
standards were adhered to throughout the investigation. Every
few weeks an additional test was made of several mice which
had been tested a day or two previous along with so great a
number that it would be impossible to remember the grade
of any mouse. In making these check tests the mice were
taken at random from the different cages and after testing
them a comparison would be made of the results of this test
and the one immediately preceding. The grading varjed
occasionally but rarely more than one grade. The results
given below were chosen at random and are typical of such
tests.

The mice were weaned, given their initial test and numbered
when they were about five or six weeks old. The age varied
somewhat in order to allow several litters to be grouped and
thus facilitate the work. At this age the mice are not full-
grown but they have full use of all their organs of sense and
are very active.

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

N-IMBER OF ilOUSE

DATE

GRADE OF WILDNESS

GRADE OF SAVAGENESS

512 9

April 9
April 11

5
5

5

5

517d^

April 9
April 11

5

5

3

4

990 cf

January 20
January 22

3
3

3
4

1002 9

January 20
January 22

4
4

4

4

1008 d^

January 20
January 22

1
1

1
0

1040d^

January 20
January 22

3

4

2
2

In order to avoid any effect which a change of surroundings
and a possible change of temperature or light may have upon
the behavior of the mice, all testing was done in the room occu-
pied by the mice. The cage containing the individuals to be
be examined was placed upon the testing table which was
located, as a rule, in one corner of the room. Upon the table
were also placed paper, pencil, paper punch for marking the
mice, and a pair of placental forceps over the jaws of which
had been placed soft rubber tubing in order to lessen the pres-
sure and also the chances of injuring the tail of the mouse which
required its use in catching. The gloves worn by the experi-
menter while testing were of kid of medium thickness.

Removing the mouse from the nest was done, when the
mouse was hiding in the nest, by gently pushing back with
the left hand the shredded tissue paper of the nest until the
tail was exposed and then, grasping the tail between the thumb
and forefinger, the mouse would be equally as gently pulled
from the nest and at the same time placed upon the palm of
the hand by turning the palm upward. In this way the mouse
could often be removed without greatly disturbing it. When
the smaller Bussey Institute cage was used, since the door was
so small, often the tail when exposed had to be grasped by the
placental forceps and thus the mouse pulled from the nest and
placed upon the palm of the hand. When the mouse to be
removed was not hiding in the nest the forceps were used to

6 CHARLES A. COBURN

grasp the tail and hold it only until it could be grasped by the
fingers of the left hand whereupon the removal would proceed
as usual. It is to be noted that the forceps were used as little
as possible since the chance of unconsciously using unnecessary
pressure was much greater with the forceps than with the fin-
gers. The mouse was allowed to remain on the palm of the
hand for a moment while a definite impression of its behavior
was obtained. Then with the tail between the thumb and
forefinger of the right hand and the mouse on the palm of the
left or being gently held in the left hand the sex would be de-
termined. This done the mouse would be returned to the
position on the palm of the right hand and one of the fingers
of the left hand or a pencil would be moved fairly rapidly up
and down just before the mouse, the experimenter at the same
time making a clicking sound with his tongue and teeth. The
object of this was to get the behavior of the mouse when excited
if such was possible. This completed the testing observations
in the case of the initial tests and, the judgment being made,
the result was thereupon entered. In the subsequent tests
the behavior of the mouse when released was noted before
making the judgment. The mouse was then numbered and
placed in a cage with others of like sex and, as nearly as possible,
like age in order to prevent promiscuous breeding and the
killing of the younger mice by the older.

The system of numbering was identical with that in use at
the Bussey Institute. It consisted in making certain notches
or holes in the ears of the mouse. When viewed from behind
the number indicated on the right ear represented the units
and that on the left ear the tens of any particular number.
Absence of any mark indicated zero. 1, 2, 3, etc. were re-
spectively designated as follows: hole in the upper side of ear,
hole in outer side, hole in lower side, notch in upper edge, notch
in outer edge, notch in lower edge, notches in upper and outer
edges, notches in outer and lower edges, notches in upper and
lower edges. Ten was indicated by a hole in the upper side
of left ear and nothing in the right ear, 22 by a hole in the outer
side of both left and right ears, etc. The wounds thus made
were usually observed to be completely healed in a very few
days. By such combinations any number up to and including
99 could be indicated. As the different hundreds were desig-

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

nated on the outside of the cages and a description of each
mouse was taken at the time of the first test, no difficulty was
ever experienced in determining the identity of any mouse
which had escaped from its cage. It may be mentioned that
it w^as necessary to make the rooms occupied by the mice
rat proof as well as mouse proof. In the beginning sufficient
caution was not taken in this respect the result being that
many mice were killed by rats and others lost by escaping from
the room after they had by some chance escaped from their
cage.

The following Is the record of the initial test of one litter.
The records of the subsequent tests were kept similarly ex-
cept the omission of the description of the mice and the num-
ber of their parents.

June IS, 1913. F3's from F2 409 9 and F2 440 d'. Born April 29, 1913

NUMBER or MOUSE

DESCRIPTION

WILDNESS

SAVAGENESS

5819

582 9

583 d'

584 c^

585 9

Chocolate, black eyes

Gray

Chocolate, red eyes

Chocolate, red eyes

Chocolate, black eyes

4
3
3

4
4

3
3
3
4

5

The types of behavior which were considered the chief indi-
cations of wildness In mice and were relied upon as a basis for
grading were as follows.

1. Hiding from view In the cage just before the test or im-
mediately after, and attempts to hide from view in the hand
during the test. A wild mouse, though It may be feeding in
the open in the cage while the experimenter is moving about
the room or standing before the cage, will scurry to the back
of the cage and hide in the nest at the first move of the experi-
menter to open the door of the cage. A tame mouse, on the
other hand, will continue eating and is easily caught, in fact, it
will rarely make any attempt to prevent it. When the wild
mouse has been caught it is continually trying to find some
crevice between the fingers in which it may hide. If it succeeds
in covering its head it will often remain perfectly quiet as long
as it Is allowed to maintain this position.

2. Running and jumping excitedly about the cage just before
being caught and attempts to escape from the hand after being

8 CHARLES A. COBURN

caught. Often the wildness is sufficiently great that the mouse
apparently forgets to hide in the nest at the first move to open
the door of the cage, or it may have hidden there but at the
first touch of the hand on the outside of the nest it will jump
from the nest and begin to run frantically about the cage arid
finally try to hide in the upper far corner of the cage or it may
continue to run about the cage until caught. When one with
such wildness is caught it continues to try to escape, jumping
about the hand and trying energetically to pull its tail loose.
After some time it may become quiet but will immediately
resume the attempts to escape at the slightest move of the ex-
perimenter, especially if the pressure of the fingers on its tail
is somewhat lessened.

3. Squeaking. This was rarely observed except in very
young mice or older mice when they were being graded for
the first time. However it is included as an indication of
wildness since it was never observed in tame mice.

4. Urination and defecation. Both of these are associated
with the greater degrees of wildness. They were used in grading
only in a general way.

5. Jumping from the hand when released and immediately
hiding in the nest or excitedly running about the cage. When
a very wild mouse is being put back in the cage it will imme-
diately jump from the hand the instant the pressure on its tail
is released. It usually jumps straight ahead no matter whether
that is toward the top, side or front of the cage, or directly
through the door of the cage toward the experimenter. If it
alights in the cage it generally dashes toward the nest and hides.
At other times it may run about the cage but sooner or later
will seek the nest. A tame mouse will never jump in this way.
It will go to the edge of the hand and look down. If the dis-
stance is very great it will try the other side and will not of its
own accord leave the hand unless it can reach the floor with
its fore feet when it hangs over the side of the hand. When it
is safely back in the cage it makes no attempt to hide.

Professor Yerkes found that there are two kinds of savage-
ness in rats. These he termed defensive and offensive. Only
the former is found in mice. The most savage mouse will
never attack the observer nor make the slightest move to do
so. It will only bite when it is caught and unable to get loose.

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE 9

Biting in this case is not, however, always a sign of savageness.
Quoting from the writer's notes; "Biting when just caught is
not always a sign of savageness for tame individuals will bite
if hurt." This is corroborated by Miss Abbie E. C. Lathrop,
Granby, Mass., who has had many years' experience in breed-
ing and handling mice. The following statements are taken
from one of her letters. "Every one who comes here and tries
to handle mice gets bitten even by tame mice, because of grasp-
ing the tail too tightly. Some mice are cross even when not
hurt, but gentle mice resent pain and bite."

The chief indications of savageness are biting and squeaking.
A mouse possessing the greatest degree of savageness may
indicate this by biting in either of two ways depending largely,
it seems, upon the degree of wildness which accompanies it.
Very often a very savage mouse while being pulled from the
nest and before the hand has touched its body will swing about
and sink its teeth into the glove. This act is the nearest ap-
proach to the offensive savageness of the rat. If there is a
great degree of wildness present which will cause it to make
violent attempts to free itself, everytime that its feet touch a
part of the glove in its frightened jumping about the teeth
are vigorously used. Often in the instant before the next
jump two or three bites will be made in the glove, the mouse
trying, as it were, to find some vulnerable spot. If the glove
has even a small hole in it in an accessible spot the mouse of
this type is very apt to find it and use its teeth with considerable
effect.

On the other hand, if there is a much less degree of wildness
present the mouse will often seize the glove with its teeth
and tenaciously hold on until compelled to let go by the experi-
menter whereupon it will immediately fasten its teeth in another
part of the glove. The mouse with no savageness will make
no attempt to bite even when teased unless the tail is held too
tightly or it is otherwise hurt. However, many of the non-
savage mice made no attempt to bite even when being numbered.

Between these two extremes of savageness there are many
types of behavior which were considered as representing
different grades of savageness as, for instance, the mouse with a
low degree of savageness will, perhaps, use its teeth but once
or twice during the test and then only in an indifferent way,

10 CHARLES A. COBURN

or it may not bite at all until it is teased whereupon it will do
so vigorously for an instant or two.

Squeaking is an indication of a fairly high degree of savage-
ness unless the mouse is hurt, in which case it may be observed
where there is little or no savageness.

Although the types of behavior mentioned above are the
chief indications of wildness and savageness in mice, a great
many others which are very significant often appear. The
range of individuality of mice, to the experienced observer, is
very great. In spite of these many indications and the differ-
ent degrees of intensity with which they appear in the behavior
of mice during a test, very rarely did the writer, after he had
had some experience, have difficulty in forming a judgment of
the grade of wildness or savageness observed.

A description of the wild and tame mice used in this study
is given in table 1.

The observations and tests made on the tame mice from the
Bussey Institute and Miss Abbie E. C. Lathrop for some time
after they arrived at the laboratory indicated in no case a grade
of wildness or savageness above 0. The mice from the Bussey
Institute were from a strain which had been in use there for
some time for various color and structural studies and were
from stock which had originally been obtained from Miss
Lathrop. After they had been used in the present study for
about a year, it was quite accidently learned that there was
some possibility of their possessing some wild blood as mice
from this strain had at various times been crossed with wild
in the color studies. In order to secure more definite informa-
tion a mating was made from these mice and the offspring tested.
The results are shown in table 2.

The table gives in the upper left hand corner (a) the average
number of tests; {b) the range of tests; (c) the average age at
the time of the first test; (d) the average age for the last test;
immediately below, (e) the number of mice and sex; to the
right of this, (/) the average grade attained in the first test for
wildness; (g) the average of the first and second tests; {h) the
average of all the tests; (i) the average of the third, fourth and
fifth tests; (j) the average grade attained in the last test. Im-
mediately to the right of these averages is presented the dis-
tribution of the mice in the grades 0 to 5. The averages for

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

11

the tests for savageness and the distribution of the mice in
grades for savageness are presented in Hke manner.

TABLE 1
Parental Generation

MOUSE

DESCRIPTION

WHERE OBTAINED

19

Chocolate, tame

Bussey Institute, Forest Hills, Mass.

2c?

Chocolate, tame

Bussey Institute, Forest Hills, Mass.

6c?

Chocolate, tame

Bussey Institute, Forest Hills, Mass.

7c?

Chocolate, tame

Bussey Institute, Forest Hills, Mass.

5019

Chocolate, tame

Miss Abbie E. C. Lathrop, Granby, Mass.

502 9

Chocolate, tame

Miss Abbie E. C. Lathrop, Granby, Mass.

503 9

Chocolate, tame

Miss Abbie E. C. Lathrop, Granby, Mass. .

504 c?

Chocolate, tame

Miss Abbie E. C. Lathrop, Granby, Mass.

522 9

Cream, tame

Miss Abbie E. C. Lathrop, Granby, Mass.

523 9

White, tame

Miss Abbie E. C. Lathrop, Granby, Mass.

524 9

White, tame

Miss Abbie E. C. Lathrop, Granby, Mass.

525 9

Black, tame

Miss Abbie E. C. Lathrop, Granby, Mass.

527 9

White, tame

Miss Abbie E. C. Lathrop, Granby, Mass.

528 9

White, tame

Miss Abbie E. C. Lathrop, Granby, Mass.

529 9

Chocolate, tame

Raised in laboratory, parents, 503 9 and 504 c?

530c?

Chocolate, tame

Raised in laboratory, parents, 503 9 and 504 c?

5319

Chocolate, tame

Raised in laboratory, parents, 503 9 and 504c?

532 9

Chocolate, tame

Raised in laboratory, parents, 503 9 and 504c?

533 c?

Chocolate, tame

Raised in laboratory, parents, 503 9 and 504 c?

534 c?

White, tame

Miss Abbie E. C. Lathrop, Granby, Mass.

652 9

White, tame

Miss Abbie E. C. Lathrop, Granby, Mass.

39

Gray, wild

Dwelling-house, Cambridge, Mass.

49

Gray, wild

Dwelling-house, Cambridge, Mass.

59

Gray, (Singer) wild

Dwelling-house, Chestnut Hill, Mass.

8c?

Gray, wild

Dwelling-house, Cambridge, Mass.

225 9

Gray, wild

Emerson Hall, Cambridge, Mass.

346 c?

Gray, wild

Morgan Memorial, Boston, Mass.

348 9

Gray, wild

Morgan Memorial, Boston, Mass.

472 c?

Gray, wild

Dwelling-house, Brooklyn, New York

474c?

Gray, wild

Dwelling-house, Brooklyn, New York.

475 9

Gray, (Singer), wild

Dwelling-house, Brooklyn, New York

510c?

Gray, wild

Raised in laboratory, parents, 225 9 and 350c?

5119

Gray, wild

Raised in laboratory, parents, 225 9 and 350 cf

512 9

Gray, wild

Raised in laboratory, parents, 225 9 and 350c?

513 9

Gray, wild

Raised in laboratory, parents, 225 9 and 3500^

514 9

Gray, wild

Raised in laboratory, parents, 225 9 and 350c?

515c?

Gray, wild

' Raised in laboratory, parents, 225 9 and 350c?

516 9

Gray, wild

Raised in laboratory, parents, 475 9 and 4740?

517c?

Gray, wild

Raised in laboratory, parents, 475 9 and 474c?

518c?

Gray, wild

Raised in laboratory, parents, 475 9 and 474c?

519 9

Gray, wild

Raised in laboratory, parents, 475 9 and 474 c?

520c?

Gray, wild

Raised in laboratory, parents, 475 9 and 474c?

521c?

Gray, wild

Raised in laboratory, parents, 475 9 and 474c?

350 c?

Gray, wild

Emerson Hall, Cambridge, Mass.

As is readily seen from the table there was undoubtedly some
trace of wild blood in these mice, but the amount was so small
that all evidence of its presence was effaced by the time the

12

CHARLES A. COBURN

mice were 110 days old. The wildness and savageness had
evidently been counteracted by the taming effect of the daily
presence of the experimenter while feeding and cleaning the
cages, and the handling during the five tests. The degree of
savageness was only about half as great as that of wildness
and did not persist as long, since after the second test the grade
for savageness was in no case above 0. Another point which
should be mentioned here but which is not shown in the table
is the difference in the degree of wildness and savageness of
the males and females. If the average of all the five tests is
used as a criterion, the females were about twice as wild and
savage as the males since the average grades for wildness and

TABLE 2
Summary of Results for One Litter of Offspring of Mice from Bussey Institute

Average number tests, 5

Range of tests, 5 First test

Average age in days :

First test, 46 Average 1 and 2.
Last test, 110

Average all

5 mice

(3 males and Average 3, 4 and 5 . . .
2 females)

Last test

WILDNESS

SAVAGENESS

Distribution of mice in
grades

Distribution of mice in
grades

Av.

0

1

2

3

4

5

Av.

0

1

2

3

4

S

2.6

1

1

2

1

1.8

2

2

1

2.1

4

1

1.1

4

1

1.01

4

1

0.44

4

1

0.53

3

2

0.0

5

0.0

5

0.0

5

savageness, respectively, were 1.8 and 0.6 for the females in
contrast with 0.73 and 0.33 for the males.

As a result of this evidence of impurity in the tame mice,
2 males and 10 females were obtained from Miss Lathrop.
Before these were used, however, a mating was made from them
and the offspring tested as in the previous instance. There
was not a mouse in any of these tests which received a grade
above 0 in either wildness or savageness. The proof of the
purity of the tameness and non-savageness of these young
mice (table 1, nos. 529 to 533 inclusive), and their parents was
considered so conclusive that there was no hesitation in assum-
ing an equal purity in the. remaining 10,

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE 13

The 1 1 wild mice whicii were captured and used for breeding
purposes in this study were, with the exception of mouse 472 cf ,
full-grown when caught. This mouse was judged from 6 to
8 weeks old. All of these mice were tested several times for
wildness and savageness and in each test manifested the highest
grades. The last test occurred when the time of captivity
varied from 1 to 6 months. Each received a grade of 5 for
wildness and savageness except two (nos. 225 9 and 474 c^),
which were graded 4 in savageness. Although it was neces-
sary to handle these mice when weaning their young or chang-
ing their mates, there was during this handling no observation
of any behavior which would tend to indicate that their wild-
ness or savageness had decreased to any appreciable degree
even after they had been in captivity from one to two years.
However, when not being handled there was a very noticeable
difference in the degree of wildness. Instead of scurrying to
their nest as they did when the experimenter entered the room
during their early captivity, later they would, as a rule, re-
main in the open and continue eating, but it was never possible
to persuade them to come to the front of the cage to secure a
bit of food from the experimenter's fingers.

The 12 wild mice, nos. 510 to 521 inclusive, were raised in
the laboratory and were tested five or six times in the same
manner and at regular intervals as the mice of table 2. A
summary of the results of these tests are presented in tables
15 and 16. These mice were chiefly used in matings with the
tame mice obtained from Miss Lathrop.

Although no more crosses were made between the mice from
the Bussey Institute and the wild individuals after the dis-
covery of the apparent trace of wild blood, the series begun
by them was continued and the results will be presented in
this study. These results, however, will be summarized in
separate tables except in such cases where the information to
be noted is equally significant in each series.

To avoid any confusion which might otherwise arise the
series from the crosses of Bussey Institute mice with wild mice
will be termed Series A, and the first, second, and third hybrid
generations of this series will be designated, Fia, F2a, and Fsa,
respectively. Likewise the series from crossing the tame mice
from Miss Lathrop with wild mice will be termed Series B,
and the hybrid generations. Fib, F2b, and Fab.

14 CHARLES A. COBURN

The first generation hybrids In each series were obtained
by crossing wild females with tame males and tame females
with wild males. The second generations of hybrids were
obtained by crossing the first generation hybrids among them-
selves without selection as to wlldness or savageness, except
in the cases where selective breeding was tried, the results of
which will be presented in separate tables. Likewise by cross-
ing the second generation hybrids among themselves the third
generation of hybrids were obtained.

It was aimed to test each individual five times. However,
there was a number which because of death or other reason
did not receive that number. Those that received but one
test have not been considered in the results. On the other
hand, for reasons which will be discussed later, some mice
received six to eight tests.

Deducting those Individuals used in experiments to deter-
mine the effect of age on the lowering of grades of wlldness
and savageness in the successive tests, the average age for the
first test was 38.8 days, and for the fifth test, 92.8 days. The
average time between the first and second tests was 17 days;
between the second and third, 15 days; between the third and
fourth and between the fourth and fifth, 11 days each.

EXPERIMENTAL RESULTS

A. General comparison of inheritance in the hybrid generations

of hath series

In table 3 Is presented a general summary of the results for
the three generations of both series. A comparison of the
averages in each of the generations shows a very great simi-
larity, the difference in any case being so small as to be practi-
cally negligible. However, the difference in the grades of
wlldness and savageness attained in the first and last test is
significant. This lowering of the grades is gradual in each
generation but Is greatest in both wlldness and savageness in
the F2's. The model grade, i.e., the grade which is the most
frequent, is either 3 or 4 for wlldness In each generation, and
for savageness ranges from 3 to 5.

In tables 4 and 5 the results of the series are presented sepa-
rately. The similarities of the averages in the generations so

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

15

apparent in table 3 naturally maintain to some extent in the
results of each series. There are, however, some differences

TABLE 3
Summary of Results for the Three Generations of Both Series

Average number tests, 4.46
Range of

tests, 2-5 First test

Average age in days:

First test Average 1 and 2 ,

35.27

Last test,

79.36

294 Fi

Males and

females

Average all

* (285)
Average 3, 4 and

5

Last test

Average number tests, 4.70
Range of

tests, 2-5 First test

Average age in days:

First test Average 1 and 2

40.99
Last test, 92.39

Average all .... .

410 Fo

Males and

females

*(405)

Average 3, 4 and

5

Last test

Average number

tests, 4.83 First test

Range of tests, 2-5
Average age Average 1 and 2

in days:

First test, 38.09
Last test,

5.71

164 Fs

Males and
females

Average all

*(162)

Average 3, 4 and
5

Last test

Distribution of mice in grades:

Av.

3.95
3.79
3.49

3.25
3.16

3.89

3.73
3.50

3.02
2.81

3.60

3.80

3.48

3.25
2.98

0 12 3 4 5

128

117

114

87
192

164
142

142
163
135

181

201
159

114
89

102

110
19

SAVAGENESS

Distribution of mice in grades:

Av. 0 12 3

4.01

3.74
3.13

2.63
2.45

4.09

3.72
3.26

2.63
2.13

4.07

3.80

3.40

3.13
2.91

18
101

85

71

126

92

74
127

107
106

101
131

75

128

154

117

103
83
26

20

23

170

152

45

40
39

82

53

28

30
41

* The number of mice that received three or more tests.

to be noted in a comparison of the two series. The averages
of the first test and of the first and second tests for wildness in
the Fi and also in the F2 of both series are practically equal,

16

CHARLES A. COBURN

but the remaining averages in these two generations of Series
B are in each case higher than those of Series A. The summary
of the tests for savageness in these two generations indicates

TABLE 4
Summary of Results for the Three Generations of Series A

Average number tests, 4.13
Range of

tests, 2-5 First test

Average age in days:

First test Average 1 and 2 . . .

31.22
Last test, 75.61

Average all

93 Fia (90)

Males and Average, 3, 4 and 5
females

Last test

Average number tests, 4.40
Range of

tests, 2-5 First test

Average age in days:

First test, Average 1 and 2 . . .

41.66
Last test, 98.72

Average all

177 F2a (173)

Males and Average 3, 4 and 5.
females

Last test

Average number tests, 4.88
Range of

tests, 3-5 First test

Average age in days:

First test, Average 1 and 2 . . .

36.76
Last test, 93.52

Average all

114 Fsa (114)

Males and Average 3, 4 and 5.
females

Last test

WILDNESS

Distribution of mice in
grades:

Av. 0 1 2 3 4 5

3.<
3.72

3.31
2.92

2.75

3.90

3.72
3.05

2.52
2.39

3.44

3.41
3.23

3.10
2.79

SAVAGENESS

Distribution of mice in
grades:

Av. 0 1 2 3 4 S

3.97

3.58
2.96
2.38
2.04

3.94

3.46
2.64

1.97
1.67

3.63
3.26

3.01

2.78

that the mice of Series B attained a higher average grade in
the first test and maintained a greater degree throughout all
the tests. Likewise the Fsb averaged higher grades in both

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

17

wlldness and savageness than the Fsa. If the assumption
that there was some wild blood in the tame parents of Series

TABLE S
Summary of Results for the Three Generations of Scries B

Average number tests, 4.62
Range of tests,

2-5 First test

Average age in days:

First test, Average 1 and 2 . .

37.18
Last test, 81.09

Average all

201 Fib (195)

Males and Average 3, 4 and 5
females

Last test

Average number tests, 4.95
Range of tests,

2-5 First test

Average age in days :

First test. Average 1 and 2. . .

40.66
Last test, 87.15

Average all

233 Fab (232)

jNIales and Average 3, 4 and 5
females

Last test

Average number tests, 4.70
Range of

tests, 2-5 First test

Average age in days:

First test. Average 1 and 2. . .

41.12
Last test, 77.62

Average all

50 Fsb (48)

Males and Average 3, 4 and 5 .
females

Last test

Distribution of mice in
grades:

Av.

3.94

3.83
3.56

3.36
3.37

3.

3.75
3.50
■i .62
3.12

3.98

4.69

4.07

3.62
42

111
99

101

111
113

87
63

29

3i
24

SAVAGENESS

Distribution of mice in
grades:

Av.

4.02

3.82
3.20

2.73
2.63

4.21

3.93
3.40
3.03

62

845
3882

92045 58
33113560

78

69

22

18
22

11

53

27

74059
6304655

2.48492031
4.60

4.19

3.74

3.41
3.20

112

117
41
37
35

34

23
12

11
91 15

A is correct, these results are directly opposed to the results
Professor Yerkes obtained with the rats.^

^ Robert M. Yerkes, The Heredity of Savageness and Wildness in Rats. The
Journal of Animal Behavior, iii. No. 4, page 293.

18 CHARLES A. COBURN

Of the five averages included In the tables, the average of
all the tests and the average of the third, fourth, and fifth tests
seem to present the most accurate estimates of the degree of
wildness or savageness of the mice, because the number of
tests included in these averages tends to counteract the effect
of any extreme grade of wildness and savageness or tameness
and non-savageness that may have been received in an occa-
sional test due to the influence of some extraordinary conditions
upon the mouse or the experimenter or both at the time of the
test. Of these two averages, that of the third, fourth and fifth
tests is probably the more reliable measurement of the traits
in question. This is because there is included in the average
of all tests the results of the first two tests which are consid-
ered to give an exaggerated estimate of the wildness, and per-
haps also savageness, possessed by the mouse on account of
the possible effect, in the case of the first test, of the experience
of being handled for the first time ; and, in the case of the second
test, because of any effect which may remain from the more
or less painful experience of being numbered immediately after
the first test is made. The fact that the lowering of the grades
of wildness is greater between the second and third tests than
between any other two successive tests seems to support this
statement (table 12, page 23).

Using the average of all tests and the average of the third,
fourth and fifth tests as criteria, the arrangement of the genera-
tions of Series A with respect to grades attained in wildness
and savageness is (1) F3, (2) Fi, (3) F2, where one represents
the highest grade and three the lowest grade. This order of
arrangement is maintained when the degree of wildness of the
generations of Series B is considered, but the order of this series
with respect to savageness is (1) F3, (2) F2, (3) Fi.

The modal grade for wildness is either 3 or 4 in all the genera-
tions of both series except F2a where it ranges from 2 to 4.
The modal grade for savageness is either 3 or 4 in Fia and either
4 or 5 in Fsb. In the case of Fsa, Fib, and Fab, it ranges from
3 to 5, and in F^sl the range is from 0 to 5.

In tables 6, 7, and 8, the results for the males and females
of each generation of Series A are presented separately, and in
tables 9, 10, and 11, those of Series B are shown.

The averages for the males and females of Fia (table 6) in-
dicate but little difference in the grades of wildness and savage-

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

19

ness attained by each. The average grades for wildness at-
tained by the males in the first test and in the last test are
almost identical with those of the females. But in the case
of the averages for wildness in all tests and in third, fourth,
and fifth tests, those of the males are a fraction of a grade higher
than those of the females. While the averages of the first
tests for savageness are somew^hat higher for the females than
for the males, the average grades attained in the last tests are

■ TABLE 6

Summary of Results for First Generation Hybrids, Fi, of Series A

Average number tests,
Range of tests, 2-5
Average age in days :

First

test, 31.60

Last test, 75.63

41 Fia
Males

Average number tests,
Range of tests, 3-5
Average age in days:

First

test, 30.92

Last test, 75.59

52Fia
Females

3.82

First test

Average 1 and 2. .

Average all

(38)

Average 3, 4 and 5

Last test

4.19

First test

Average 1 and 2. . .

Average all

Average 3, 4 and 5
Last test

Distribution of mice
in grades:

Av. 0 1 2 3 4 5

4.0

3.74
3.33

2.94

2.75

3.98

3.70
3.15

2.92

2.75

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1 2 3 4 5

3.80
3.45

2.98

2.52
2.14

4.11

3.68

2.11

2.36
1.96

16

16

just the reverse. The grades representing the averages of all
the tests are practically equal, 2.98 for the males and 2.99 for
the females. The males in the third, fourth and fifth tests for
savageness received, as an average, 0.16 grade higher than the
females.

The modal grade for wildness of the males and females of
Fia is either 3 or 4 in each case. The mode for savageness of
the males ranges between 2 and 4, and of the females, between
1 and 4.

20

CHARLES A. COBURN

TABLE 7
Summary of Results for Second Generation Hybrids, F^, of Series A

Average number tests,
Range of tests, 2-5
Average age in days:

First
test, 42.00

Last test, 99.21

82F2a
Males

4.45
First test.

Average 1 and 2. . .

Average all

(79)

Average 3, 4 and 5

Last test

Average number tests,
Range of tests, 2-5
Average age in days:
First test, 41.38
Last test, 99.27

95 F2a females

4.36
First test.

Average 1 and 2 .

Average all

(94)

Average 3, 4 and 5 .

Last test

Distribution of mice
in grades:

Av.

4.00

3.67

2.93

2.31
2.10

3.82

3.76

3.17

2.70
2.64

SAVAGENESS

Distribution of mice
in grades;

0

1

2

3

4

5

Av.

0

1

2

3

4

1

21

36

24

4.46

1

15

37

2

23

40

17

3.44

1

9

22

38

22

42

16

2

2.54

11

27

31

12

10

16

17

22
22

29
21

11
11

1
1

1.79
1.34

9

35

19
12

23
14

25
15

2

5

2

31

44

18

4.07

1

12

24

29

21

50

24

3.63

1

1

6

25

39

9

54

30

2

2.83

1

6

28

37

20

1
2

3
8

27
34

45
31

16
15

2
4

2.19
2.00

3
17

22
21

28
22

27
21

12
11

29

12

1

1
1

29

23

3

2
3

TABLE 8
Summary of Results for Third Generatio7i Hybrids, Fs, of Series A

Average number tests.
Range of tests, 3-5
Average age in days :

First

test, 35.06

Last test, 89.74

62 Fsa

Males

Average number tests.
Range of tests, 3-5
Average age in days:

First
test, 38.78

Last test, 98.03

52 Fsa
Females

4.80
First test

Average 1 and 2. . .

Average all

Average 3, 4 and 5
Last test

4.98
First test.

Average 1 and 2. . .

Average all

(50)

Average 3, 4 and 5

Last test

WILDNESS

Distribution of mice
in grades:

Av.

3.38

3.30
3.07
3.08
2.69

3.51

3.54

3.30

3.15
2.92

2

3

5

26

5

22

16

24

17

20

12

19

2

25

16

8

20

9

3

18
20

SAVAGENESS

Distribution of mice
in grades:

3.95

3.66
2.99
2.69
2.48

3.71

3.60

3.46

3.37
2.96

2

3

4

19

4

13

13

22

14

21

11

13

3

12

3

19

6

20

6

7

14
16

HEREDITY OF WILDNESS AND SAVAGENESS IN ^NIICE

21

The averages for the females of Foa and Fsa are higher In
grade than those of the males in every instance except the
averages for the first tests for wildness and savageness in both
generations and the average of the first and second tests for
savageness in the Fsa. The modal grades in these two genera-
tions range as follows: for males in wildness and savageness
respectively; F2a, 2-4 and 0-4; Fsa, 3-4 and 0-5 ; for the females,
F2a, 2-4 and 2-4; Fsa, 3-4 and 3-5.

TABLE 9
Siimviary of Results for First Generation Hybrids, Fi, of Series B

Average number tests,
Range of tests, 2-5
Average age in days:

First
test, 36.76

Last test, 77.09

102 Fib

Males

Average number tests.
Range of tests, 3-5
Average age in days:

First
test, 37.43

Last test, 72.08

99 Fib

Females

4.49

First test

Average 1 and 2. . . .

Average all

(96)

Average 3, 4 and 5 .

Last test

4.77

First test

Average 1 and 2. . . .

Average all

Average 3, 4 and 5 .
Last test

WILDNESS

Di^

tribulion of mice
in grades:

Av.

3.57

3.56

3.30

3.13
3.04

4.02

3.93

3 .73

3.59
3.71

SAVAGENESS

Distribution of mice
in grades:

Av.

3.96

3.78

3.07

2.53
2.50

4.10

3.84

3.30

2.91
2.65

In the three generations of Series B (tables 9, 10 and 11), the
females attained higher grades in every instance except the
average of the first tests for savageness In Fsb. The modal
grades of the averages of this series likewise indicate that the
females are generally more wild and savage than the males.

From the results as shown in table 3 it seems fairly clear
that the inheritance of wildness and savageness in mice is
Mendelian of the "blending" or multiple factor type. The
first section of this table shows that the wildness and savageness
of the Fi's vary about the intermediate grade 3 with some

22

CHARLES A. CDBURN

TABLE 10
Summary of Results for Second Generation Hybrids, F^, of Series B

Average number tests,
Range of tests, 3-5
Average age in days:

First
test, 41.95

Last test, 87.64

115 Fab Males

Average number tests,
Range of tests, 2-5
Average age in days:

First

test, 39.40

Last test, 86.67

118 Fab
Females

4.97
First test.

Average 1 and 2. . .

Average all

Average 3, 4 and 5
Last test

4.93
First test.

Average 1 and 2. . . .

Average all

(117)

Average 3, 4 and 5 .

Last test

WILDNESS

Distribution of mice
in grades:

3.86

3.61

3.12
2.93

3.93

3.90

3.

3.53
3.31

27

53

29

SAVAGENESS

Distribution of mice
in grades:

Av.

4.06

3.95

3.24
2.71
2.00

4.35

4.22

3.70

3.55
2.94

27

2>3,

51

TABLE 11
Summary of Results for Third Generation Hybrids, Fz, of Series B

Average number tests, 4.65

Range of tests, 2-5
Average age in days:

First
test, 41.13

Last test, 80.17

23 Fsb
Males

First test.

Average 1 and 2. . .

Average all

(22)

Average 3, 4 and 5

Last test

Average number tests, 4.74
Range of tests, 2-5 First test
Average age in days:
First

test, 41.11
Last

test, 83.00
27 Fab
Females

Average 1 and 2.

Average all

(26)

Average 3, 4 and 5

Last test

Distribution of mice
in grades:

Distribution of mice
in grades:

Av.

0

1

2

3

4

5

Av.

0

1

2

3

4

3.78

7

14

2

4.65

1

6

3.57

1

5

14

3

4.00

3

9

3.44

3

12

6

2

3.45

5

6

7

3.34
3.08

1

2
1

2
5

8
8

5
4

5
4

3.04
2.65

4

3

4

7
4

2
2

4
1

4.14

4

15

8

4.55

3

6

3.77

3

19

5

4.16

1

14

3.82

8

18

1

3.90

1

5

14

3.85
3.70

1

7
10

18
11

1

5

3.71
3.62

2

4

1

3
9

14
8

SAVAGENESS

16

11

5

6

8

18

12

7

5
7

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

23

individuals possessing a grade of wildness and savageness or
tameness and non-savageness equal to that of the parental
generation. The second section of the table indicates that the
degree of wildness and savageness of the F2's varies about the
same intermediate grade and that this variability is also in
accordance with Mendelian expectation in being greater than
that of the Fi's. The number of cases studied, however, is
scarcely large enough to give sufhcient basis for an estimate of
the number of factors involved.

B. Effect of age, frequency and 7iumber of tests on the lowering
of the grade of wildness and savageness in successive

tests

In table 12 are given the average results of the successive
tests of all hybrids of the three generations that received the
regular tests, and also the amount of difference in the grades
of wildness and savageness in the successive tests.

TABLE 12
Summary of Results for Regular Tests of All Hybrid Generations

NUMBER MICE

NUMBER TEST

AVERAGE GRADE FOR

DIFFERENCE IN GRADE

Wildness

Savageness

Wildness

Savageness

868
850
850
797
645

First

Second

Third

Fourth

Fifth

3.87
3.60
3.30
3.11
2.95

4.06
3.56
3.10
2.65
2.21

0.27
0.30
0.19
0.16

0.50
0.46
0.45
0.44

It is readily seen by this table that there is a fairly gradual
lessening in the grades of wildness and savageness from the
first test to the last. The greatest amount of decrease in the
grade of wildness occurs between the second and third tests.
With this exception the amount of decrease is greatest between
the first and second tests and gradually diminishes with the
successive tests.

In the attempt to determine the cause or causes for this
lowering of the grades of wildness and savageness with the
repetition of the tests, three different methods were used as
follows :

1. Individuals were allowed to remain untested, unnumbered,
and without being handled, in the cage in which they were

24

CHARLES A. COBURN

born, (the parents usually being removed), until they had
attained the age at which the fifth test would ordinarily be
given. Then the five tests were given in the usual way.

This method was used with 9 individuals of Fia, and with
48 individuals of F2a. The results are given in tables 13 and

TABLE 13
Summary of Resitlts of Tests of Old Mice of First Generation Hybrids Fi, of Series A

Average number tests, 4.00

Range of tests, 4 First test

Average age in days:

First
test, 86.55 Average 1 and 2

Last test, 139.00

Average all

9 (old) Fia

Males and Average 3, 4 and 5.

Females Last test

Average number tests, 4.00

Range of tests, 4 First test

Average age in days :
First

test, 88.33 Average 1 and 2. . .
Last test, 141.00

Average all

3 (old) Fia

Males Average 3, 4 and 5
Last test

Average number tests, 4.00

Range of tests, 4 First test

Average age in days:

First test, 85 .66 Average 1 and 2

Last test, 138.00

Average all

6 (old) Fia

Females Average 3, 4 and 5.
Last test

WILDNF.SS

Distribution of mice
in grades:

Av. 0 1 2

4.44

4.11
3.30

2.61

2.44

4.66

4.16

3 .33

2.50
2.33

4.33
4.08

2.66
2.50

S.\VAGENESS

Distribution of mice
in grades:

Av.

44.44

33.61

2.80

2.00
2.00

4.66

3.83

3.00

2.16
1 .33

4.33

3.50

2.70

1.92
2.33

5 1
1

14. A comparison of the first part of table 13 with the first
part of table 4 shows that while the older mice average higher
in wildness and savageness in the first test (4.44 and 4.44,
respectively, to 3.98 and 3.97), they grade lower in the fifth
test (2.44 and 2.00 to 2.75 and 2.04). When the results of
the males and the females are separated the above statement

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

25

remains true with the exception of the grade for savageness of
the old Fia females which exceeds that of the Fia females of
ordinary age as shown in table 6. The results of the 48
F2a's of table 14 when compared with the second part of table
4 and with table 7 indicate that the greater age has caused a

TABLE 14
Summary of Rcsidls of Tests of Old Alice of Second Generation Hybrids, Fo, of Series A

Average number tests, 3.98

Range of tests, 3-4 First test

Average age in days:
First

test, 85.39 Average 1 and 2

Last test, 135.41

Average all

48 (old) F.a Males

and Females Average 3, 4 and 5 .
Last test

Average number tests, 4.00

Range of tests, 4 First test

Average age in days:

First test, 85.19 Average 1 and 2. . . .
Last test, 134.88

Average all

26 (old) Fsa

Males Average 3, 4 and 5 .
Last test

Average number tests, 3.95

Range of tests, 3-4 First test

Average age in days:
First

test, 85.63 Average 1 and 2. . . .

Last test, 136.04

Average all

22 (old) Foa

Females Average 3, 4 and 5 .
Last test

Distribution of mice
in grades:

Av. 0 1 2 3 4 S

3.79
3.50

2.94

2.38
2.04

3.65

3.17

2.62

2.07
1.69

3.95

3.88

3.33

2.76
2.45

14

12

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1 2 3 4 5

3.68

3.19

2.43

1.64
1.35

3.65

3.01

2.18

1.34
1.07

3.72

3.41

2.73

2.04
1.68

11

decrease in the grades of wildness and savageness of all the
tests.

2. The number of tests was increased to eight, and the grades
of the last three tests noted.

The individuals used in this experiment numbered 17, and
were of the F2a. The results are given in table 15. The

26

CHARLES A. COBURN

results of the first five tests are essentially the same as those
given for the F2a'sin tables 4 and 7, showing that these 17 indi-
viduals were normal. The significance of the table is merely

TABLE 15

Summary of Results for Seventeen Individuals of Second Generation Hybrids, Series A
when the number of Tests is increased to Eight

Average number tests, 8
Range of tests, 8
Average age in days:
First test, 57.88

Fifth test, 121.82
Eighth test, 167.82
17 Foa males and
females

Average number tests, 8
Range of tests, 8
Average age in days:
First test, 57.54
Fifth test, 121.77
Eighth test, 167.77
7 F2a Males

First test .

Average 1
and 2. . .

Average first 5
Average 3,4,

and 5

Fifth test

Average 6, 7,

and 8

Eighth test

Average number tests, 8
Range of tests, 8
Average age in days:
First test, 57.92
Fifth test, 121.98
Eighth test, 167.97
10 Foa females

First test

Average 1 and 2

Average first 5
Average 3,

4 and 5

Fifth test

Average 6,

7 and 8

Eighth test . . . .

First test

Average 1 and 2

Average first 5 .
Average 3,

4 and 5

Fifth test

Average 6,

7 and 8

Eighth test

Distribution of mice
in grades:

Av. 0 1

3.29

3.52

3.06

2.74
2.82

1.64
1.47

3.58

3.57

2.19

2.30
1.85

1.22
1.00

3.10
3.50

2.82

3.06
3.60

2.13
2.10

1

2

3

—

3

8
8

1

13

2
2

3

5

10

5

8
7

6

3

2

5

1

3
3

1

5

2
2

2
4

3
1

6

4

1

1

2

5
5
8

1
2

7
4

2

5
3

2
4

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1

3.38

2.59

2.00
2.00

1.30
1.00

4.33

3.47

1.39

1.30
1.14

0.40
0.42

3.70

3.60

2.60

2.56
2.60

1.96
1.40

to show that the decrease in the grades of wildness and savage-
ness continues with the successive tests when the number of
tests is increased to eight.

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE 27

3. Litters were equally divided and the individuals of
one group were tested every day or every two days until the
five tests were given, while the individuals of the other group
were tested at the usual times.

The division of the mice in this experiment depended entirely
upon the order in which the mice were caught for their first
test. The first mouse was put in the cage of those to be tested
at the usual times and the second was put into the cage of those
to be tested daily. Hence, each litter was divided into two
practically equal groups which, having the same ancestry,
should form a basis for a fairly accurate estimate of the effect
of frequency of tests upon the lowering of the grades of wild-
ness and savageness.

The first mice to be used in this experiment were the 12 wild
mice raised in the laboratory and later used as wild parental
stock (table 1). There were two litters of six mice each. One
litter had four males and two females and the other two males
and four females. They were therefore not divided according
to chance, as has just been mentioned but into two groups of
three males and three females each. The results of the group
with the usual time intervening between the tests (Group 1)
are given in table 16, and those receiving daily tests (Group
2) in table 17. In the case of wildness, when the results of the
males and females are considered together as are shown in the
first part of the tables, the averages are identical with the ex-
ception of that for the fifth test which is 0.34 grade greater
for the mice tested daily. The averages of the first test for
savageness are also the same for both groups, and the remain-
ing compare very favorably, the difference being in no case
greater than 0.25 grade. When the results for the males and
females are considered separately the females of Group 2 are
seen to grade, as a rule, slightly higher than those of Group 1.

The sixth and seventh tests of the mice of Group 2 were given
at the same time the mice of Group 1 were receiving their fourth
and fifth tests. Assuming that the three factors, namely,
age, presence of the experimenter while feeding and testing, and
the handling during the tests, each have a certain effect on the
lowering of the grades in successive tests; a comparison of the
averages of the fourth and the fifth tests of the mice of Group
1 with the averages of the sixth and seventh tests of those of

28

CHARLES A. COBURN

Group 2 should give the measure of the effect of the two extra
tests, since the ages of both groups are the same and hence, the
effect of the presence of the experimenter would likewise be
equal. In the case of the males this effect seems to be indicated
in both wildness and savageness as the average grades for the

TABLE 15
Summary of Results of Group 1, Parental Generation, Wild, Tested at Usual Times

Average number tests, 5 '

Range of tests, 5 First test

Average age in days:

First test, 37 Average 1 and 2 . . . .

Fourth test, 70

Last test, 88 Average all

6 P (Wild)

males and Average 3, 4 and 5. . .

females Average 4 and 5 . . . .

Last test

Average number tests, 5

Range of tests, 5 First test

Average age in days:

First test, 37 Average 1 and 2 . . . .

Fourth test, 70

Last test, 88 Average all

3 P (Wild)

Males Average 3, 4 and 5. .
Average 4 and 5 . . . .
Last test

Average number tests, 5

Range of tests, 5 First test

Average age in days:

First test, 37 Average 1 and 2

Fourth test, 70

Last test, 88 Average all

3 P (Wild)

females Average 3, 4 and 5. .
Average 4 and 5 . . . .
Last test

WILDNESS

S.WAGENESS

Distribi^ion of mice
in grades:

Distribution of mice
in grades:

Av.

0

1

2

3

4

5

Av.

0

1

2

3

4

5

5.0

6

4.5

1

1

4

5.0

6

4.58

2

4

4.83

6

4.23

4

2

4.66

4.58
4.16

1

1
1

3

5
5
2

4.0

3.83
3.0

1

2
1
4

3

4

1

1
1

5.0

3

4.66

1

2

5.0

3

4.66

1

2

4.86

3

4.2

2

1

4.77
4.66
4.33

1

1
1

2
3

1

3.88
3.66
2.66

1

1

1

2

1
2

1

5.0

3

4.33

1

2

5.0

3

4.5

1

2

4.8

3

4.36

2

1

4.66

4.5

4.0

1

1
2

2
2

1

4.11

4.0

3.33

1

2

2
2
1

1

sixth and the seventh tests of Group 2 are, respectively, 0.66
and 0.33 of a grade lower than the averages for the fourth
and fifth tests of Group 1. In the case of the females, however,
this does not maintain inasmuch as the averages for wildness
are identical in each group and the average for savageness in
the sixth and seventh tests of Group 2 is 0.5 of a grade higher

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

29

than that for the fourth and fifth tests of Group 1. The differ-
ences in the grades of these wild individuals are so slight and
the number of mice so small that these results cannot be con-

TABLE 17
Summary of Results of Group 2, Parental Generation, Wild, Tested Daily

Average number tests,
Range of tests, 7
Average age in days:
First test, 37
Fifth test, 41
Sixth test, 70
Seventh test, 88

6 P (Wild)
males and
females

Average number tests.
Range of tests, 7
Average age in days:
First test, 37
Fifth test, 41
Sixth test, 70
Seventh test, 88

3 P (Wild)
Males

Average number tests.
Range of tests, 7
Average age in days:
First test, 37
Fifth test, 41
Sixth test, 70
Seventh test, 88

3 P (Wild)
Females

7_

First test

Average 2 and 1 . . . .

Average first 5

Average 3, 4 and 5. .

Fifth test

Average 6 and 7 . . . .

7_

First test

Average 1 and 2 . . . .

Average first 5

Average 3, 4 and 5. .

Fifth test

Average 6 and 7

7_

First test

Average 1 and 2 . . . .

Average first 5

Average 3, 4 and 5 . .

Fifth test

Average 6 and 7 . . . .

WILDNESS

Distribution of mice
in grades:

Av. 0 1

5.0
5.0

4.83
4.66

4.5
4.25

5.0

5.0

4.66

4.44

4.0
4.0

5.0
5.0
5.0
5.0

5.0

4.5

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1 2 3 4 5

4.5
4.33

4.11
4.2

4.0
3.91

4.0

3.66

3.86

4.0

4.0

5.0
5.0

4.53

4.22

4.0
4.5

sidered of very great value. The chief significance of tables 16
and 17 is to give an example of the reactions of wild mice to
the conditions of the experiment and thus furnish a basis for
the more accurate judgment of the hybrids.

30

CHARLES A. COBURN

Mice from Fib were next used in this study. 14 individuals
comprised Group 1 and 16 Group 2. The results are given
in tables 18 and 19. From F2b four groups were used. The
mice forming two of these groups were very much younger than
those comprising the other two groups of Fob and also the two

TABLE 18

Summary of Results of Group 1, First Generation Hybrids, Fi, of Series B, Tested at the

Usual Times

Average number tests, 5

Range of tests, 5 First test

Average age in days:
First

test, 52.00 Average 1 and 2 . .

Last test, 82.28

Average all

14 Fib males

and females Average 3, 4 and 5
Last test

Average number tests, 5

Range of tests, 5 First test

Average age in days:
First

test, 36.00 Average 1 and 2 . .

Last test, 66

Average all

3 Fib Males Average 3, 4 and 5
Last test

Average number tests, 5

Range of tests, 5 First test

Average age in days:
First

test, 56.36 Average 1 and 2 . .

Last test, 86.76

Average all

11 Fib females Average 3, 4 and 5
Last test

Distribution of mice
in grades:

Av.

4.14

4.35

4.05

3.88
4.07

4.35

4.66

4.26
4.0

4.33

4.09

4.22

4.0

3.84

4.0

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1

4.57

4.53

3.94

3.54
3.21

4.33

4.66

3

3.22

3.0

4.63

4.50

3.98
3.63
3.27

groups of Fib. There were 37 of these younger mice, 19 in Group
1 and 18 in Group 2. The results are presented in tables 20
and 21. Group 1 of the older mice contained 17 individuals,
while 19 mice comprised Group 2. The results of the tests of
these are given in tables 22 and 23.

HEREDITY OF WILD NESS AND SAVAGENESS IN MICE

31

TABLE 19
Summary of Results of Group 2, First Generation Hybrids, Fi, of Series B, TesledDaily

Average number tests,
Range of

tests, 4-6
Average age in days :
First

test, 54.0
Fifth test, 58.93
Sixth
test, 86.21

16 Fib Males

and
Females

Average number tests,
Range of

tests, 4-6
Average age in days:
First

test, 43.4
Fifth test, 48.2
Sixth

test, 76.66
5 Fib Males

Average number tests,
Range of tests, 6
Average age in days:

First

test, 58.81

Fifth test, 63.81

Sixth
test, 88.81

11 Fib

Females

5.86

First test

Average 1 and 2 . . . .

Average first 5

Average 3, 4 and 5 . .

Fifth test

(15)
Sixth test

(14)

5.4

First test

Average 1 and 2 . . . .

Average first 5

Average 3, 4 and 5 . .
Fifth test

(4)
Sixth test

(3)

6

First test

Average 1 and 2 . . . .

Average first 5

Average 3, 4 and 5. .

Fifth test

Sixth test

WILDNESS

Distribution of mice
in grades:

Av. 0 1 2 3 4 5

4.62
4.22

4.06
3

3.42
3.78

4.8
4.2

3.57
3.0

4.0

4.54

4.22

4.14

4.09
4.09
3.72

11

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1 2 3 4 5

4.62
4.35

3.87
3.58

3.06

3.78

4.4
3.8

3.6

3.07

2.0

4.0

4.72

4.59

4.07

3.72
3.54
3.72

11

The sixth test of Group 2 of each set was given at the same
time the mice of the corresponding Group 1 received their
fifth test. It is to be noted that while the average grade for

32

CHARLES A. COBURN

the sixth test is greater in each case than that for the fifth
test of the same group, (First section of tables 19, 21, and 23),
due, without doubt, to the absence of any handling by the ex-
perimenter during the 27 or 28 days intervening since the fifth
test, yet only in the case of the younger mice of F2b is the aver-

TABLE 20

Summary of Results of Group 1, Second Generation Hybrids, Fo, of Series B, Tested

at Usual Times

Distribution of mice
in grades:

Average number tests, 4.89
Range of

tests, 3-5 First test.

Average age in days:
First

test, 34.52
Last test, 65.15

19 Fab Males
and Females

Average 1 and 2
Average all

Average 3, 4 and 5. .
Last test

Average number tests, 5
Range of tests, 5 First test

Average age in days:

First

test, 34.4

Last test, 66.3

Average 1 and 2

9 Fab Males

Average all

Average 3, 4 and 5 .
Last test

Average number tests, 4.77
Range of

tests, 3-5 First test.

Average age in days :
First

test, 34.64
Last test, 62.22

Av.

lOFgb
Females

Average 1 and 2 . .
Average all ,

Average 3, 4 and 5.
Last test

4.0

3.8f
3.18

2.69

2.42

3.7

3.45

2.98
2.66
2.4

4.33

4.37

3.41

2.72
2.44

10

SAVAGENESS

Distribution of mice
in grades:

Av.

3.94

4.18

3.18

2.49
1.78

3.8

3.9

3.12

2.6

2.3

4.11

4.5

3.25

2.36
1.22

0 12 3 4 5

12

age grade for wildness in the sixth test of Group 2 greater than
that for wildness in the fifth test of the corresponding Group
1 (table 20). The average grade for savageness for the sixth
tests of each Group 2 is greater than that for the fifth test of
the corresponding Group 1.

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

33

The explanation for the fact that the sixth test for wildness
of the older mice in the second groups did not follow that of
the younger mice in being higher in grade than the fifth test of
the corresponding first groups may be found in the difference
observed in the general type of reaction to handling as the

TABLE 21
Summary of Results of Group 2, Second Generation Hybrids, Ft, of Series B,Tested Daily

Average number tests, 5.94

Range of tests, 5-6 First test

Average age in days:

First test, 35.11 Average 1 and 2 . .
Fifth test, 39.88

Sixth test, 66.64 Average first 5. . . .
18 Fob INIales Average 3, 4 and

and Females 5

Fifth test

Sixth test

Average number tests, 5.88

Range of tests, 5-6 First test

Average age in days:

First test, 35.33 Average 1 and 2 . .

Fifth test, 40.33

Sixth test, 66.00 Average first 5 . . . .

9 Fab Average 3, 4 and

Males 5

Fifth test

Sixth test

Average number tests, 6

Range of tests, 6 First test

Average age in days:

First test, 34.88 Average 1 and 2..

Fifth test, 39.44

Sixth test, 67.22 Average first 5. . . .

9 F2b Average 3, 4 and

Females 5

Fifth test

Sixth test

Distribution of mice
in grades:

Av. 0 1 2 3 4

4.05

3.66

3.45

3.31
3.05
3.35

4.11

3 .33

3.06

2.88
2.77
3.12

4.0
4.0

3.84

3.74
3.33
3.55

SAVAGENESS

Distribution of mice
in grades:

Av.

4.44

4.05

3.52

3.16
2.77
3.82

4.44

3.83

3.2

2.77
2.22
3.0

4.48
4.27

3.55
3.33
4.55

0 1

11

hybrids grew older. There seemed to be a lesser tendency to
violent reactions of either wildness or savageness with a some-
what more agressiveness in the latter. With age they became,
as it were, more deliberate in their movements.

In order to compare more easily the results included in
tables 18 to 23 inclusive, the more important points of these

34

CHARLES A. COBURN

have been brought together in table 24. The headings at the
top show the different hybrid generations and the groups.
The headings at the side indicate (a) the table in which the
complete results of a group may be found, (b) the number of
mice in each groups, (c) the average age in days of the group at
the first test, the fifth test, and the difference between these

TABLE 22

Summary of Results of Group 1, Second Generation Hybrids, Fo, of Series B, Tested at

Usual Times

Average number tests
Range of tests, 5
Average age in days :
F rst test, 59.82
Last test, 91.35

17 Fjb Males
and Females

5
First test

4.0

4.14

3.84

3.64
3.64

4.0

4.14

3.85

3.66
3.57

4.0

4.15

3.84

3.63
3.7

Average 1 and 2. . .

Average all

Average 3, 4 and
5

Last test

Average number tests,
Range of tests, 5
Average age in days:
First test, 59.57
Last test, 91.14

7F2b
Males

5
First test . . ■

Average 1 and 2. . .

Average all

Average 3, 4 and
5

Last test

Average number tests.
Range of tests, 5
Average age in days :
First test, 60.0
Last test 91.5

10 Fab
Females

5
First test

Average 1 and 2. . .

Average all

Average 3, 4 and
5

Last test

Distribution of mice
in grades:

Av. 0 1 2 3 4 5

SAVAGENESS

Distribution of mice
in grades:

Av.

4.05

4.35
3.49

2.92

2.58

4.0

4.42

3.42

2.76
1.85

4.1

4.3

3.6

3.13
3.1

0 1

ages, or in other words, the number of days required for the
tests in each group, {d) the average grade of wildness for the
first test, for the fifth test, the difference between the first and
fifth, the average of the first and second tests for wildness, the
average for wildness as shown by the third, fourth, and fifth
tests, and the difference between these two averages. Imme-

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

35

diately below this, the results for savageness are treated in a like
manner. Following this are given, (/) the difference between
the number of days required for the tests in Group 1 and in
Group 2, (g) the difference in decrease in the average grade of
wildness of Group 1 and Group 2 as shown by the difference

TABLE 23
Summary of Results of Crovp 2, Second Generation Hybrids, Fo, of Series B, Tested Daily

Average number tests,
Range of tests, 6
Average age in days:
First test, 59.42
Fifth test, 64.42
Sixth test, 91.0
19 F.,b Males
and Females

Average number tests,
Range of tests, 6
Average age in days :
First test, 59.3
Fifth test, 64.3
Sixth test, 90.9
10 Fob
Males

Average number tests.
Range of tests, 6
Average age in days:
First test, 59.55
Fifth test, 64.55
Sixth test, 91.11
9 Fab Females

First test

Average 1 and 2 . .

Average first 5. . . .
Average 3, 4 and

5

Fifth test

Sixth test

6
First test

Average 1 and 2 . .

Average first 5. . . .
Average 3, 4 and

5

Fifth test

Sixth test

6
First test

Average 1 and 2 . .

Average first 5. . . .
Average 3, 4 and

5

Fifth test

Sixth test

WILDNESS

Distribution of mice
in grades:

Av. 0 1

4.52
4.18

3.89

3.7

3.31

3.63

4.7

4.05

3.76

3.53

3.3

3.6

4.33

4.27

4.04

3.88
3.33
3.66

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1

4.26
4.15

3.46

2.94
2.68
3.47

4.5

4.05

3.4

2.93

3.1

3.4

4.0

4.22

3.46

2.96
2.22
3.55

in the average grades of tests one and two and of tests three,
four and five, {h) the rate of decrease per day as obtained by
dividing the fraction of a grade found in (g) by the number of
days obtained in (/), Following this in (i) the same values
are found for savageness.

36

CHARLES A. COBURN

The decrease in both wildness and savageness for hybrids
Fib (tables 18 and 19) compares very favorably with the older
mice of F2b (tables 22 and 23), except that the latter show a

TABLE 24
General Summary of Tables 18, 19, 20, 21, 22, and 23

1

HYBRIDS Fib

2

HYBRIDS F2b

3

HYBRIDS F2b

Group
1

Group 2

Group

1

Group 2

Group
1

Group 2

Table

18
14

82.28
52.00
30.28

4.14
4.07
0.07

4.35
3.88
0.47

4.57
3.21
1.36

4.53
3.54
0.99

19
16

58.93

54.00

4.93

4.62
3.42
1.20

4.22
3.93
0.29

4.62
3.06
1.56

4.35
3.53
0.82

25.35

0.18
0.0071

0.17
0.0067

20
19

65.15
34.52
30.63

4.0

2.42
1.58

3.89
2.69
1.20

3.94
1.78
2.16

4.18
2.49
1.69

21
18

39.88
35.11

4.77

4.05
3.05
1.0

3.66
3.31
0.35

4.44
2.77
1.67

4.05
3.16
0.89

25.86

0.85
0.0328

0.80
0.0309

22
17

91.35
59.82
31.53

4.0

3.64

0.36

4.14
3.64
0.50

4.05
2.58
1.47

4.35
2.92

1.43

23

Number of mice

19

Average age in days:

Fifth test

64.42

First test

59.42

Difference

5.0

Wildness, average grade:

First test

4.52

Fifth test

3.31

Difference

1.21

Tests 1 and 2

4.18

Tests 3, 4 and 5

3.7

Difference

0.48

Savageness, average grade:

First test

4.26

Fifth test

2.68

Difference

1.58

Tests 1 and 2

4.15

Tests 3, 4 and 5

2.94

Difference

1.21

Difference in number days re-
quired for tests in Group 1 and
Group 2

26.53

Difierence in decrease in average
grade of tGroup 1 and Group 2,
as shown by tests 1 and 2,
and 3, 4, and 5 :

Wildness

0.02

Rate of decrease per day .
Savageness

0.00075
0.22

Rate of decrease per day .

0.00829

slightly greater decrease than the former. Between these
two, however, and the younger mice of F2b (tables 20 and 21)
there is a striking difference. Considering the difference be-
tween the first and last tests for wildness and savageness in

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE 37

each of the three sets of groups (numbered for sake of conveni-
ence, 1,2, and 3), it is to be noted that this difference in Group
2 of the first and third set is greater than the difference in the
corresponding Group 1, while in the second set the reverse is
true. In other words, with the older mice, whether they are
from Fib or Fob, the decrease in grade of wildness and savage-
ness is greater when the mice are tested every day than when
they are tested at intervals of six or eight days, but with the
younger mice the daily testing shows the lesser decrease.

When, however, the decreases are considered from the dif-
ferences of the averages of tests one and of tests three, four
and five, the results are uniform in all three sets of groups,
that is, the results in the first and third set are reversed and
in each set the tests given at intervals of 6 or 8 days show a
greater decrease in the grades of wildness and savageness than
in the case of the daily testing. It is the opinion of the writer
that the difference between the average of the first and second
tests and the third, fourth and fifth tests forms the more ac-
curate basis for the judgment of the rate of lowering of the
grades of wildness and savageness, since the effect of an error
in testing is thus minimized.

Assuming that a test has an equal effect in lowering the grade
of wildness or savageness of the mice in Group 1 and Group
2, from the results in table 24 it is possible to obtain an approxi-
mate estimate of the effect of age plus the effect of the presence
of the experimenter in the room while feeding, cleaning cages
and testing other mice. For instance, in the case of Group
1, Fib, 0.47 grade represents the lessening of wildness due to
the effect of five testings, plus 30.28 days of age and the pres-
ence of the experimenter during that time. From Group 2,
Fib, it is observed that 0.29 grade represents the lessening of
wildness due to the effect of five testings, plus 4.93 days of age
and the effect of the presence of the experimenter. Subtract-
ing the results of Group 2 from those of Group 1, there is a
remainder of 0.18 grade which represents the lessening of the
wildness due to the effects of 25.35 days of age and the presence
of the experimenter during that time. In the same way the
lessening in savageness for the same reasons is found to be 0.85
grade. By dividing the amount of decrease by the number
of days are obtained the rates of decrease per day which in

38

CHARLES A. COBURN

this case are 0.0071 grade for wildness and 0.0067 grade for
savageness. These represent the grade-lowering effect of
age and the presence of the experimenter exerted each day on
the mice of Fib when they receive their first test at the average
age of 53 days. Likewise the rates per day for wildness and
savageness of the younger mice of Fb are, respectively, 0.0328
grade and 0.0309 grade, and for the older mice of F2b, 0.00075
grade and 0.00829 grade. It would thus seem that age and
the presence of the experimenter have an effect in lessening the
grade of wildness and savageness of the mice inversely in pro-
portion to the age of the mouse, providing the first test is given
when the mice are between 33 and 60 days old. The rates ac-
cording to sex are as follows:

WILDNESS

RATE OF DECREASE

PER DAY

SAVAGENESS

RATE OF DECREASE

PER DAY

Fib
Males

0.00833
0.00976

0.0126
0.0603

0.00225
0.0049

0 0281

Females

0 0000

F2b (Young)
Males

0 00892

Females

0 0616

F2b (Old)
Males

0 020 S

Females

-0 00339

From these results it is seen that the females had a higher
rate of decrease per day for both wildness and savageness ex-
cept in savageness of the Fib and the F2b (Old), in which cases
it was zero for the former and a negative quantity (0.00339)
for the latter.

C. Results from crossing wild females with tame males in com-
parison with the results from crosses of tame females
with wild males

The hybrids used by Professor Yerkes in his study were
all obtained by crossing tame female rats with wild male rats.
It was to neutralize the effect due to the possibility of the par-
ent of one sex exerting a stronger hereditary and environmental
influence on the offspring than the parent of the opposite sex
that the first generation hybrids of both series of this study were
obtained about equally from wild females crossed with tame

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

39

males and from tame females crossed with wild males. For
the same reason both parents were allowed to remain, with
few exceptions, in the cage with the young until they were
weaned. It is proposed, in this section of the paper, to present
for closer consideration groups of hybrid individuals of the

TABLE 25

Summary of Results of First Generation Hybrids, Fi, of Series A,fromMatings of Parental
Generation, Tame Female with Wild Male

Average number tests, 4.06
Range of tests, 3-5 First test

Average age in days:

First test, 43.14

Last test, 63.0

Average 1 and 2.

45 Fia Males
and Females

Average all

Average 3, 4 and

5

Last test

Average number tests, 4.17
Range of tests, 3-5 First test

Average age in days:

First test, 32.11 Average 1 and 2.
Last test, 56.2

Average all

17 Fia Average 3, 4 and

^lales 5

Last test

Average number tests 4.0

Range of tests, 3-5
Average age in days:
First test, 49.85
Last test, 67.25

28 Fia
Females

First test .

Average 1 and 2. .

Average all

Average 3, 4 and

5

Last test

Distribution of mice
in grades:

Av. 0 1

4.02

3.61

3.15

2.70
2.44

3.

3.47

2.87

2.32
2.29

4.10
3.69
3.33

2.96

2.53

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1

4.04

3.55

2.79

2.06
1.75

4.0

3.61

2.80

2.05
1.2

4.07

3.51

2.79

2.07
1.25

17

10

710

first, second and third generations obtained from a cross of
wild females with tame males and other groups obtained from a
cross of tame females with wild males. In most cases the individ-
uals of the F2 were offspring of the Fi mice represented in the
tables, likewise most of the F3 mice were obtained from
matings of the F2.

40

CHARLES A. COBURN

The results of tests of the group of first generation hybrids
of Series A from matlngs of tame females with wild males are
presented in table 25, and from matlngs of wild females with
tame males In table 26, while those of Series B are to be found
in tables 27 and 28, respectively. In the same order the results
from the second generation hybrids are given In tables 29,

TABLE 26

Summary of Results of First Generation Hybrids, Fi, of Series A, from Matings of Parefital
Generation, Wild Female liith Tame Male

Average number tests,
Range of tests, 4-5
Average age in days:
First test, 35.84
Last test, 62.38

50 Fia Males
and Females

Average number tests,
Range of tests, 4-5
Average age in days:
First test, 3L53
Last test, 62.16

23 Fia Males

Average number tests,
Range of tests, 4-5
Average age in days:
First test, 34.44
Last test, 62.57

27 Fia Females

4.52
First test .

Average 1 and 2.

Average all

Average 3, 4 and

5

Last test

4.52
First test .

Average 1 and 2.

Average all

Average 3, 4 and

5

Last test

4.51
First test.

Average 1 and 2 . .

Average all

Average 3, 4 and

5

Last test

WILDNESS

Distribution of mice
in grades:

Av. 0 1 2 3 4 5

4.0

3.91

3.36

2.93
3.02

4.08

4.04

3.48

3.03
3.17

3.92
3.79
3.27

2.85

2

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1 2 3 4 5

4.0

3.63

2.96

2.42
2.07

3.65
3.34
2.81

2.39

2.25

4.29

3.87

3.08

2.45
1.92

30, 31, and 32; and those from the third generation hybrids in
tables ?)d), 34, 35, and 36.

The differences in the results of the tests of these mice are so
slight that no attempt will be made to discuss each table sepa-
rately. The essential points from all the tables have been
brought together and are presented in table 37. In the first

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

41

column of this table will be found the average grade for all
the tests and the average grade for the third, fourth, and fifth
tests for wildness and savageness of the first, second, and third
generation hybrids of Series A from the matings of tame fe-
males with wild males. In the second column are given the

TABLE 27

Summary of Results of First Generation Hybrids, Fi, of Series B,from Matings of Parental
Generation, Tame Female with Wild Male

Average number tests,
Range of tests, 5
Average age in days:
First test, 36.26
Last test, 73.26

30 Fib Males
and Females

Average number tests,
Range of tests, 5
Average age in days :
First test, 37
Last test, 73.0

6 Fib Males

Average number tests.
Range of tests, 5
Average age in days:
First test, 39.36
Last test 80.0

24 Fia females

First test

Average 1 and 2 . .

Average all

Average 3, 4 and

5

Last test

5
First test

Average 1 and 2

Average all

Average 3, 4 and

5

Last test

5
First test

Average 1 and 2. .

Average all

Average 3, 4 and

5

Last test

Distribution of mice
in grades:

Av.

4.43

4.13

3.54

3.14
3.0

4.33

3.91

3.13

2.61
3.0

4.86

4.18

3.64

3.25
3.27

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1

4.2

4.01

2.8

2.21
1.96

3.66

3.58

2.33

1.50
1.66

4.72

4.12

3.08

2.38
2.22

9 5

sexes which attained the higher average grade in these tests.
In the third and fourth columns are presented the like values
for Series B of the same matings, while in the fifth and sixth,
and the seventh and eighth columns are given these values
for Series A and Series B, respectively, from matings of wild
females with tame males.

42

CHARLES A. COBURN

From this table it is readily seen that the mice of Series A,
tame female by wild male, became generally less wild and less
savage in succeeding generations according to the averages
of all the tests and the averages of the third, fourth, and fifth
tests, the only exception being the average of all the tests for
savageness of the F2 hybrids, which grade is greater than the

TABLE 28

Summary of Results of First Generation Hybrids, Fi, of Series B, from Matings of Parental
Generation, Wild Female with Tame Male

Average number tests,
Range of tests, 4—5
Average age in days:
First test, 39.76
Last test, 104.52

26 Fib Males
and Females

Average number tests,
Range of tests, 4—5
Average age in days:
First test, 37.70
Last test, 101.88

17 Fib
Males

Average number tests.
Range of tests, 4—5
Average age in days :
First test, 32.77
Last test, 107.48

9 Fib Females

4.65
First test .

Average 1 and 2. .

Average all

Average 3, 4 and

5

Last test

4.53
First test .

Average 1 and 2. .

Average all

Average 3, 4 and

5

Last test

First test

Average 1 and 2 . .

Average all

Average 3, 4 and

5

Last test

Distribution of mice
in grades:

Av. 0 1 2 3 4

3.76
3.61

3.35

3.16
3.22

3.71

3.14

2.93

2.74
2.76

3

4.44

4.09

3.84
4.11

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1

4.07
3.94

3.25

2.73
2.47

3.88

3.73

2.93

2.3
1.44

4.44

4.33

3.81

3.46
3.62

like average of the Fi or the F3. The females of this series
were, as a rule, more savage than the males, and were also more
wild in the first and second generations, the males grading
higher in wildness in the third generation.

The results of the crosses of Series B, tame female by wild
male, are directly opposed to those of Series A, the mice in this

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

43

case attaining, without exception, increasingly higher average
grades of wildness and savageness in the successive generations.
In this series, again, the females grade higher in wildness
and savageness in the first two generations while the males of
the third generation were both more wild and more savage
than the females.

TABLE 29

Summary of Results of Second Generation Hybrids, F2, of Series A, from Matings of Parental
Generation, Tame Female with Wild Male

Average number tests, 4.39

Range of tests, 3-5 First test

Average age in days :

First test, 47.78 Average 1 and 2. . .
Last test, 115.22

Average all

41 Foa IMales Average 3, 5 and

and Females 5

Last test

Average number tests, 4.35

Range of tests, 4—5 First test

Average age in days :

First test, 50.30 Average 1 and 2. .

Last test, 107.25

Average all

20 Foa Males Average 3, 4 and

5

Last test

Average number tests, 4.42

Range of tests, 3-5 First test

Average age in days:

First test, 45.38 Average 1 and 2. .

Last test, 120.45

Average all

21 F2a Females Average 3, 4 and

5

Last test

Distribution of mice
in grades:

Av.

3.51

3.35

2.92

2.57
2.48

3.35
3.17
2.73

2.36

2.2

3.66

3.52

3.10

2.76
2.76

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1 2 3 4 5

4.12
3.47

3.21

1.96
1.38

4.2

3.4

2.39

1.53
1.14

4.04

3.54

3.97

2.37
2.45

The results of both Series A and Series B of the hybrids,
wild female by tame male, seem to show, as a rule, that the
grades attained in wildness and savageness in the first genera-
tion are lowered in the second generation and then, in the
third, raised to somewhat near that attained in the first. The
exceptions to this rule are the averages for all the tests for

44

CHARLES A. COBURN

wildness in Series A, which show a decrease in wildness in the
successive generations, and the averages of the third, fourth
and fifth tests for savageness in Series B, which indicate a sHght
increase in savageness in each succeeding generation.

The grades in wildness and savageness attained by the
females of Series B are generally in excess of those attained by

TABLE 30

Summary of Results of Second Generation Hybrids, F2, of Series A ,from Matings of Parental
Generation, Wild Female with Tame Male

Average number tests,
Range of tests, 3-5
Average age in days:
First test, 38.07
Last test, 104.2

39 Faa Males
and Females

Average number tests,
Range of tests, 4-5
Average age in days:
First test, 3L53
Last test, 97.0

13 Foa
Males

Average number tests.
Range of tests, 3-5
Average age in days:
First test, 41.34
Last test, 100.68

26F2a
Females

4.10_ ,

First test .

Average 1 and 2. .

Average all

Average 3, 4 and

5

Last test

4.30 _

First test .

Average 1 and 2. .

Average all

Average 3, 4 and

5

Last test

First test

Average 1 and 2.

Average all

Average 3, 4 and

5

Last test

DJs

tribution of mice
in grades:

Av.

3.96

3.84

3.1

2.4
2.43

4.0

3.8

2.92

2.16
2.46

3.96
3.86
3.2

2.53
2.42

0 12 3 4 5

SAVAGENESS

Distribution of mice
in grades:

Av. 0

3.97

3.52

2.76

2.04
0.86

4.23

3.76

2.83

2.03
1.25

3.84

3.4

2.73

2.05
0.72

the males. In Series A, however, the males graded higher in
wildness in the first generation and higher in savageness in the
third generation. In the case of the averages for savageness
of Fi, Series A, tame female by wild male, and of F2, Series B,
tame female by wild male, and also in the case of the averages
for wildness of F3, Series B, wild female by tame male, the males

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

45

attained the higher grades, according to the averages of all the
grades, and the females, when the averages of the last three
tests are considered, however,, the females are assumed to be
the more savage or wild, as the case may be, since the height
of the grades of the females, though not so great in the first
two tests as that of the males, was more nearly maintained
throughout the five tests.

TABLE 31

Summary of Results of Second Generation Hybrids, Fo, of Series B, from Matings of Parental
Generation, Tame Female with Wild Male

Average number tests,
Range of tests, 4-5
Average age in days:
First test, 38.01
Last test, 85.82

66 Fob Males
and Females

Average number tests,
Range of tests, 5
Average age in days:
First test, 38.36
Last test, 88.36

30 Fob
IMales

Average number tests,
Range of tests, 4-5
Average age in days:
First test, 37.72
Last test, 83.58

36 Fob
Females

4.97

First test

Average 1 and 2 . .

Average all

Average 3, 4 and

5

Last test

5

First test

Average 1 and 2. .

Average all

Average 3, 4 and

5

Last test

4.91 _

First test

Average 1 and 2 . .

Average all

Average 3, 4 and

5

Last test

Distribution of mice
in grades:

Av.

4.21

3.99

3.72

3.54
3.31

4.2

3.86

3.64

3.48
2,. 32,

4.22

4.09

3.79

3.59
3.30

0 1

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1

4.63

4.42

3.90

3.55
2.96

4.63
4.43
3.79

3.36

2.7

4.63

4.98

3.99

3.69
3.2

D. General summary of individual inheritance ajid the results
of selective breeding in the hybrid generations of both

series

In this section of the study the attempt is made to indicate
the differences in the inheritance of individual mice rather than

46

CHARLES A. COBURN

that of groups of mice as in the preceding sections. The results
of the inheritance of the offspring from selected matings are
included in the results of the general matings in order to give
better opportunity for comparison than could otherwise be
obtained.

TABLE 32

Summary of Results of Second Generation Hybrids, F2, of Series B,from Matings of Parental
Generation, Wild Female with Tame Male

Average number tests, 5
Range of tests, 5
Average age in days :
First test, 39.96
Last test, 80.23

25 Fab Males
and Females

Average number tests, 5
Range of tests, 5
Average age in days:
First test, 38.61
Last test, 77.15

13F2b
Males

Average number tests, 5
Range of tests, 5
Average age in days :
First test, 41.41
Last test, 87.91

12 Fab
Females

First test

Average 1 and 2. .

Average all

Average 3,

4 and 5

Last test

First test

Average 1 and 2. .

Average all

Average 3,

4 and 5

Last test

First test

Average 1 and 2 . .

Average all

Average 3, 4

and 5

Last test

Distribution of mice
in grades:

Av.

3.28

3.4

2.92

2.6
2.28

3.3

3.21

2.66

2.3
1.76

3.25

3.62

3.2

2.91
2.83

0 1

SAVAGENESS

Distribution of mice
in grades:

Av.

3.36

3.84

3.2

2.77
2.04

3.84

3.8

2.95

2.38
1.46

3.66

3.87

3.46

3.19
2.66

To show to best advantage the individual inheritance, the
results of the averages of the third, fourth, and fifth tests have
been arbitrarily chosen as the criteria and the six grades, 0 to
5, have been divided into two equal parts. All mice which
received as an average of their third, fourth and fifth tests,
grades not exceeding 2.4 are considered tame and non-savage.
Those whose averages of these tests were equal to or between

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

47

2.5 and 5, are considered wild and savage. It is thus seen by
this arrangement that four distinct groups are possible, i.e.,
a mouse would be found in one of the four classes, viz: wild
and savage (w-s), wild and non-savage (w-n), tame and sav-
age (t-s), tame and non-savage (t-n). Since the averages of
the tests of the wild parental stock were greater than 2.5 grade

TABLE 33

Summary of Results of Third Generation Hybrids, F3, of Series A, from Malings of Parental
Generation, Tame Female with Wild Male

Average number tests, 5
Range of tests, 5
Average age in days:
First test, 38.0
Last test, 112.0

6 Fsa Males
and Females

Average number tests, 5
Range of tests, 5
Average age in days:
First test, 38
Last test, 97

3 Fsa
Males

Average number tests, 5
Range of tests, 5
Average age in days :
First test 38
Last test, 127

3 Fsa
Females

First test

Average 1 and 2.

Average all. .
Average 3, 4

and 5

Last test

First test

Average 1 and 2

Average all . .
Average 3, 4

and 5

Last test

Distribution of mice
in grades:

First test

Average 1 and 2.

Average a 1 . .
Average 3, 4

and 5
Last test . . . .

Av.

0

1

2

3

4

3.83

1

5

3.16

1

2

3

2.6

3

3

2.22
1.83

1

1
1

2
2

3
2

3.66

1

2

2.83

1

1

1

2.66

2

1

2.55
2.0

1

1
1

2

1

4.0

3

3.5

1

2

2.53

1

2

1.88
1.66

1

1

1
1

1
1

SAVAGENESS

Distribution of mice
in grades:

Av.

3.83
3.41

2.4

1.72
1.66

3.66

3.16

2.26

1.66
2.0

4.0

3.66

2.53

1.77
1.33

1 2

2

1

they are classed as wild and savage, and similarly the tame
parental stock are tame and non-savage inasmuch as their aver-
ages never exceeded 2.4 grade.

From the matings of the parental stock and of the first and
second hybrid generations in the general and selected breeding
eleven different sets are to be found. These include all the

48

CHARLES A. COBURN

different matings used in this study. In addition to the sym-
bols they are designated in the tables by the Roman numerals
which precede them in the following list.

I. Female wild and savage X male wild and savage ( 9 w-s X d^ w-s)
II. Female wild and savage X male wild and non-savage ( 9 w-s X cf w-n)
III. Female wild and non-savage X male wild and savage ( 9 w-n X cf w-s)

TABLE 34

Summary oj Results of Third Generation Hybrids, Fs, of Series A, from Matings of Parental

Generation, Wild Female ■with Tame Male

Average number tests, 4.21
Range of tests, 3-5 First test

Average age in days :

First test, 41.64

Last test, 103.22

Average 1 and 2.

14 Fsa Males
and Females

Average all . .
Average 3, 4

and 5

Last test

Average number tests, 3.75
Range of tests, 3-5 First test

Average age in days :

First test, 35.12

Last test, 98.66

Average 1 and 2.

8F3a
Males

Average all
Average 3,

4 and 5. . .
Last test

Distribution of mice
in grades:

Average number tests, 4.83

Range of tests, 4—5 First test

Average age in days:

First test, 48.66 Average 1 and 2 .
Last test, 105.5

Average all

6 Fsa Average 3,

Females 4 and 5

Last test

Av.

0

1

2

3

4

5

Av.

0

1

2

3.5

2

4

5

3

3.92

3.25

2

5

5

2

3.6

1

2.89

7

4

3

3.05

1

2

2.58
2.64

1

3

6

3

2
4

5
4

2.54
2.66

1
2

1

3
2

3.0

2

3

2

1

4.12

2.81

2

3

2

1

3.75

1

2.56

4

4

3.26

1

2.28
2.62

1

5

1
3

2

2

2

2.71
2.33

1
1

1
1

4.16

1

3

2

3.66

3.83

2

3

1

3.41

3.24

3

3

2.82

2

2.82
2.66

1
2

1

1
2

3
2

2.41
2.83

1

1

2
1

SAVAGENESS

Distribution of mice
in grades:

IV. Female wild and savage X male tame and non-savage ( 9 w-s X d^ t-n)
V. Female tame and non-savage X male wild and savage ( 9 t-n X c? w-s)
VI. Female wild and non-savage X male wild and non-savage ( 9 w-n X & w-n)
VII. Female wild and non-savage X male tame and non-savage ( 9 w-n X cf t-n)
VIII. Female tame and non-savage X male wild and non-savage ( 9 t-n X c?' w-n)
IX. Female tame and non-savage X male tame and non-savage ( 9 t-n X c? t-n)
X. Female tame and non-savage X male tame and savage ( 9 t-n X cf t-s)
XI. Female wild and savage X male tame and savage ( 9 w-s X c^ t-s)

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

49

A general summary of the results of the three hybrid genera-
tions, according to the above arrangements, is presented in
table 38. The first column of this table, beginning at the top,
gives the following values for Series A of the first hybrid genera-
tion: (a) number of litters, (b) total number of mice, (c) num-
ber of mice which were wild and savage, (d) number which were

TABLE 35

Summary of Results of Third Generation Hybrids , Fs, of Series B,fromMatings of Parental
Generation, Tame Female with Wild Male

Average number tests,
Range of tests, 4-5
Average age in days:
First test, 44.09
Last test, 87.12

11 Fsb Males
and Females

Average number tests,
Range of tests, 4-5
Average age in days:
First test, 44.6
Last test, 86.25

5 Fab
Males

Average number tests.
Range of tests, 4-5
Average age in days:
First test, 43.66
Last test, 88.0

6 Fab
Females

4.62

First test

Average 1 and 2 . .

Average all

Average 3,
4 and 5
Last test

4.8 _

First test

Average 1 and 2. .

Average all

Average 3,

4 and 5

Last test

4.5 _

First test

Average 1 and 2 .

Average all

Average 3,

4 and 5

Last test

Distribution of mice
in grades:

Av.

4.0

3.77

3.9

4.0

3.72

4.0

4.08

4.2
3.8

4.0

3.75

3.74

3.73
3.66

0 1

SAVAGENESS

Distribution of mice
in grades:

Av. 0 1 2 3 4 S

5.0
4.5

4.37

4.27
4.5

5.0

4.6

4.62

4.64
4.75

5.0

4.41

4.14

3.93
4.25

wild and non-savage, (e) number which were tame and savage,
(f) number which were tame and non-savage. Immediately
below this are the number for the males which in turn are fol-
lowed by the numbers for the females. Following these values
are the numbers of mice of this series which, according to the
foregoing classification were, respectively, wild, tame, savage.

50

CHARLES A. COBURN

and non-savage. The number of mice which were consid-
ered wild is found by adding the number of wild and savage
and the number of wild and non-savage found in the first divi-
sion of the table. Likewise the total number of tame mice is
the sum of the tame-savage and the tame-non-savage. The

TABLE 36

Summary of Results of Third Generation Hybrids, F^, of Series B, from Matings of Parental
Generation, Wild Female with Tame Male

Average number tests, 4.64

Range of tests, 3-5 First test

Average age in days:

First test, 42.35 Average 1 and 2
Last test, 82.63

Average all ... .
14 Fsb Males Average' 3, 4 and

and Females 5

Last test

Average number tests, 4.54

Range of tests, 3-5 First test

Average age in days :

First test, 42.09 Average 1 and 2
Last test, 76.0

Average all ... .
1 1 Fsb Average 3, 4 and

Males 5."

Last test

Average number tests, 5

Range of tests, 5 First test

Average age in days:

First test, 34. Average 1 and 2

Last test, 100.33

Average all

3 Fab Average 3, 4 and

Females 5

Last test

WILDNESS

SAVAGENESS

Distribution of mice
in grades:

Distribution of mice
in grades:

Av.

0

1

2

3

4

5

Av.

0

1

2

3

4

5

3.64

7

5

2

4.5

2

3

9

3.59

1

4

6

•3

4.25

2

5

7

3.46

2

5

7

3.43

3

3

6

2

3.36
3.14

1

1

2
3

4

5

5
2

2
3

2.81
2.63

2

3
3

3
3

2
2

4
1

2
3

3.63

6

3

2

4.54

1

3

7

3.95

1

3

6

1

4.19

2

4

5

3.52

2

5

4

3.32

3

2

5

1

3.2
2.9

1

1

2
3

3
4

3

1

2
2

2.62
2.63

2

3
2

2
3

2
1

2

2
3

3.66

1

2

4.33

1

2

4.33

1

2

4.5

1

2

3.26

3

3.8

1

1

1

3.88
4.0

1
1

2
1

1

3.23
3.66

1

1

1

2

1

sum of the wild-savage and the tame-savage equals the total
number which were savage, while the sum of the wild and non-
savage mice and the tame and non-savage gives the number
which were non-savage. In the second column are given the
percentages which indicate what part of the total number of
mice are the various numbers. The third and fourth column

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

51

give these values for Series B of the Fi hybrids. Then follow-
in order similar values for Series A and B of the F2 and F3.

The chief value of this table is to show the similarity in the
results of the three generations. With the exception of the
results of the males and females (first section, table 38) and
of the males of F2a, the percentages for the four classes in each
division of the table range from highest to lowest in the fol-
lowing order: wild-savage, wild-non-savage, tame-non-savage,

TABLE 37
General Summary of Tables 25 to 36 Inclusive

TAME FEMALE X WILD MALE

WILD FEMALE

X TAME MALE

Series A

series B

Series A

Series B

Aver-
ages

Sex testing
highest

Aver-
ages

Sex testing
highest

Aver-
ages

3.36
2.93

2.96

2.42

3.10
2.40

2.76
2.04

2.89

2.58

3.05
2.54

Sex testing
highest

Aver-
ages

3.35
3.16

3.25
2.73

2.92
2.60

3.2

2.77

3.46
3.36

3.43
2.81

Sex testing
highest

Hybrids Fi
Wildness
All tests

3.15
2.7

2.79
2.06

2.92
2.57

3.21
1.96

2.6

2.22

2.4
1.72

Female
Female

Male
Female

Female
Female

Female
Female

Male
Male

Female
Female

3.54
3.14

2.8
2.21

3.72
3.54

3.9

3.55

3.9
4.0

4.37
4.27

Female
Female

Female
Female

Female
Female

Female
Female

Male
Male

Male
Male

Male
Male

Female
Female

Female
Female

Male
Female

Female
Female

Male
Male

Female

Tests, 3, 4, 5

Savageness
All tests

Female
Female

Tests, 3, 4, 5

Hybrids F^
Wildness

All tests

Female
Female

Tests, 3, 4, 5

Savageness
All tests

Female
Female

Tests, 3, 4, 5

Hybrids F3
Wildness
All tests

Female
Male

Tests, 3, 4, 5

Savageness
All tests

Female
Female

Tests, 3, 4, 5

Female

tame-savage. The order for the males and females of F2a
is tame-non-savage, wild-savage, wild-non-savage, tame-sav-
age. This change from the order of the remaining hybrids
is caused by the great number of tame-non-savage males which
is almost twice as great as the combined number of wild-savage
and wild-non-savage. The percentage of tame-savage remains
the smallest among the males but the numbers of wild-savage
and wild-non-savage are the same, each being 24. It is to be
noted that the percent of wild-savage females in each group

52

CHARLES A. COBURN

is much in excess of that of the males, especially in the case of
the Fsb where they are, respectively, 0.469 and 0.224.

In table 39 appear the results of the Fi hybrids divided ac-
cording to the matings of the parental stock, i.e., wild female
and tame male, or tame female and wild male. With the ex-
ception of the results from the matings of tame female with wild
male in Series A, the percentages follow the order noted in the

TABLE 38
General Summary of the Individual Inheritances of the Three Hybrid Generations

Litters

Mice total.

Males and
Females

Males.

Females.

Males and

females .

w-s
w-n
t-s
t-n

w-s
w-n
t-s
t-n

w-s
w-n
t-s
t-n

Wild
Tame
Sav.
Non-
sav.

Fl HYBRIDS

Series A

0.454
0.323
0.051
0.171

0.161
0.111
0.04
0.09

0.292
0.212
0.01
0.08

0.777
0.222
0.505
0.494

Series B

50
216

139 0.643

0.25

0.013

0.092

510.236

29
1
2

193

23

142

74

0.115
0.009
0.083

0.407
0.134
0.004
0.009

0.893
0.105
0.656
0.342

Fz HYBRIDS

Series A

44
212

70
56
14

72

24

24

6

46

46

32

8

26

126
86
84

128

0.33
0.264
0.066
0.339

0.113
0.113
0.028
0.216

0.216
0.15
0.037
0.122

0.594
0.405
0.396
0.603

Series B

52
269

169
62

7
31

71

4
26

29
3

5

231
38

176
93

0.628
0.23
0.022
0.115

0.264
0.122
0.014
0.097

0.364
0.107
0.011
0.018

0.858
0.137
0.65
0.345

Fs HYBRIDS

Series A

25
114

64
19
13
18

30

13

7

12

34
6
6
6

83
31

77
37

0.561
0.165
0.113
0.157

0.263
0.113
0.061
0.105

0.298
0.052
0.052
0.052

0.726
0.270
0.674
0.322

Series B

0.693
0.224
0.02
0.061

0.224
0.142
0.02
0.061

0.469
0.081

0.917
0.081
0.713
0.285

Fl generation in table 38, namely, wild-savage , wild-non-sav-
age, tame-non-savage, tame-savage. In the case of this ex-
ception the non-savageness of the tame female seems to have
exerted a greater influence over the offspring than usual, es-
pecially with the females. The wild-non-savage class is the
greater among the females and the tame-non-savage among the
males. In the combined results of the males and females the
wild-non-savage mice are in greater numbers, the wild-savage

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

53

and tame-non-savage are equal and the tame-savage are, as
usual, the least in numbers. This apparent influence of the
tame female over the offspring is not shown in the like mating
in Series B, since only 0.08 of the total number of mice (sum of
t-s and t-n) are found in the tame classes and 0.328 (sum of
w-n and t-n) are in the non-savage classes. These percentages

TABLE 39

Summary of Matings of Parental Generation and Results of Indimdtial Inheritance of Fi

Hybrids, Series A and B

P MATING

Series A

Series B

IV

9 W-S

X
cf T-N

V
9 T-N

X
(? W-S

IV
9 W-S

X
cTT-N

V

9 T-N

X

cf W-S

3

9

54

32

17

1

4

12
8
1

2

20
9

2

49

5

33

21

C

.a

B

8
45

13

15

4

13

4
3
3

7

9

12

1

6

28
17
17
28

c
o

s

3

E

3

Litters

w-s
w-n
t-s
t-n

w-s
w-n
t-s
t-n

w-s
w-n
t-s
t-n

Wild

Tame

Sav.

Non-

sav.

0.592
0.314
0.018
0.074

0.222
0.148
0.018
0.037

0.37
0.166

0.037

0.906
0.259
0.777
0.388

0.288
0.333
0.088
0.288

0.088
0.066
0.066
0.155

0.20
0.266
0.022
0.133

0.621
0.376
0.376
0.621

13

55

33

12

1

9

11
5
1
9

22

7

45
10
34
21

0.60
0.218
0.018
0.163

0.20
0.091
0.018
0.163

0.48
0.127

0.818
0.181
0.618
0.381

37
161

106

42

2

11

40

20

1

9

66

22
1
2

148
13

108
53

Mice total

Males and females <

0.658
0.26

Males '

Females '

Males and females \

0.012
0.068

0.249
0.124
0.006
0.056

0.409
0.136
0.006
0.012

0.918

0.08

0.67

0.328

are in each case higher in the opposite mating (wild female with
tame male).

The results of the matings of the mice of Series A, Fi, are
presented in table 40, and those of Series B in table 41. The
division of the number of mice of any generation into various
groups according to the grades of wildness and savageness of

54

CHARLES A. COBURN

their parents lessens the value of each group, as far as mere
numbers are concerned, in proportion as the numbers compris-
ing each group are lessened. In this case, however, the loss
seems to be more than counteracted by the value of the demon-
strated tendency that a difference in the grade of wildness and
savageness of the hybrid parents seems to make some, though

TABLE 40

Summary

of Matings of Fi

Hybrids,

Seri

es A , and Results of Individual InJtcrilajices

of Fi Hybrids, Series A

Fl MATINGS

I

II

III

IV

VI

VII

XI

9 W-S

9 W-S

9 W-N

9 W-S

9 W-N

9 W-N

9 W-S

X

X

X

X

X

X

X

cf W-S

cf W-N

cf W-S

cfT-N

cf W-N

cf T-N

cf T-s

i-t

C

S

C

S

c

u<

*S

(U

C

t-i

c

fc

c

XJ

^

^

V

^

XI

J3

4)

J2

u

B

o

H

^

E

^

E

^

E

^

E

O

E

3

QJ

3

o

3

OJ

3

O

3

OJ

3

£

3

Ui

^

f^

'Z.

IX,

^

Oh

'Z

Ch

:?

Ph

12;

CL,

a:

(2

Litters

7

8

7

7

8

4

3

Mice total

35

39

39

30

Zi

23

13

Males
and <
females..

w-s

17

0.485

6

0.153

13

0.333

8

0.266

13

0.393

8

0.347

5

0.383

w-n

4

0.114

18

0.461

7

0.179

7

0.233

9

0.272

6

0.26

6

0.461

t-s

3

0.085

4

0.102

5

0.166

1

0.03

t-n

11

0.314

15

0.383

15

0.383

10

0.333

10

0.303

9

0.391

2

0.153

(

w-s

8

0.228

1

0.025

2

0.051

3

0.10

7

0.212

3

0.13

Males ■■■ -,

w-n
t-s

3
1

0.085
0.028

7

0.179

2

2

0.051
0.051

4
3

0.133
0.10

4

0.121

3

0.13

1

0.076

.

t-n

5

0.143

9

0.23

10

0.256

7

0.233

8

0.242

5

0.217

2

0.153

r

w-s

9

0.257

5

0.128

11

0.282

5

0.166

6

0.181

5

0.217

5

0.383

Females . . -

w-n

1

0.028

11

0.282

5

0.128

3

0.10

4

0.121

3

0.13

5

0.383

t-s

2

0.057

2

0.051

2

0.066

2

0.06

t-n

6

0.171

6

0.153

5

0.128

3

0.10

2

0.06

4

0.173

Wild

21

0.599

24

0.614

20

0.512

15

0.499

22

0.665

14

0.607

11

0.844

Males

Tame

14

0.399

15

0.383

19

0.485

15

0.499

11

0.333

9

0.391

2

0.153

and

Sav.

20

0.57

6

0.153

17

0.435

13

0.432

14

0.423

8

0.347

5

0.383

females..

Non-

^

sav.

15

0.428

11

0.844

22

0.562

17

0.566

19

0.575

15

0.651

8

0.614

not a striking difference, in the grade of wildness and savage-
ness of the offspring. If the results given in table 40 are con-
sidered, it would seem fairly reasonable to expect more of the
mice from Mating I. ( 9 W-S X cf W-S) to be wild and savage
than tame and non-savage since the parents both graded wild
and savage. This seems to be the case, at least to some ex-
tent, since the percentage of wald mice and the percentage of

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

55

savage mice are in excess, although in this particular instance
one third of the offspring were considered tame and non-savage.
In Mating 11(9 W-S X cf W-N) and Mating III ( 9 W-N
X cf W-S) wherein one parent in each mating was non-savage
the majority of the offspring, according to the above statement,
should be wild since each parent was wild, while the number

TABLE 41

Summary of Maiings of F

1 Hybrids,

Series B,

and Results a

/ Individual Inheritances

of Fi Hybrids, Series B

Fi MATINGS

I

11

IV

VI

VII

VIII

9 W-S

9 W-S

9 W-S

9 W-N

9 W-N

9 T-N

X

X

X

X

X

X

cf W-S

c? W-N

cf T-N •

cf W-N

cfT-N

c? W-N

i-i

_^

^

■*->

^

^

Lh

^

Ui

-->

U

^

i>

a

C

a

c

C

c

^

J3

XI

<u

B

u

s

'-'

E

^

E

E

^

E

'J

3

QJ

3

QJ

OJ

3

L4

3

OJ

3

Lh

■z,

h

^

Ph

1

7
4

Oi

5
22

8

^

Z

fu

12;

(li

Litters

W-S

28
147

110

0.748

7
34

17

0.50

0.571

0.363

10

53

25

0.471

1
6

5

Mice total

0.833

Males and females <

w-n
t-s

21

4

0.142
0.027

5
3

0.147
0.088

2

0.285

11

0.50

22

0.415

1

0.166

,

t-n

12

0.081

9

0.264

1

0.142

3

0.136

6

0.113

W-S

44

0.299

6

0.176

2

0.285

4

0.181

13

0.245

2

0.333

Males <

w-n

13

0.088

1

0.029

2

0.285

6

0.272

10

0.188

1

0.166

t-s

2

0.013

2

0.058

t-n

9

0.06

8

0.235

1

0.142

3

0.136

5

0.094

f

w-s

66

0.449

11

0.323

2

0.285

4

0.181

12

0.226

3

0.50

Females <

w-n
t-s

8
2

0.054
0.013

4

1

0.117
0.029

5

0.227

12

0.226

.

t-n

3

0.02

1

0.029

1

0.018

r

Wild

131

0.89

22

0.647

6

0.856

19

0.863

47

0.886

6

0.999

Males

Males and females

Tame

16

0.108

12

0.325

1

0.142

3

0.136

6

0.113

Sav.

114

0.775

20

0.588

4

0.571

8

0.363

25

0.471

5

0.833

Non-

33

0.223 14

0.411

3

0.427

14

0.636

28

0.528

1

0.166

sav.

of savage and non-savage should be about equally divided.
From the last section of table 40 it is seen that the number
of wild does exceed the number of tame but in the case of Mating
II by only one mouse. The savage and non-savage are fairly
well divided (17 and 22 respectively) in Mating III but in
Mating II the non-savage are over five times as numerous as
the savage.

56

CHARLES A. COBURN

W s

00

oo to

CN CN

oooo

t^ CV]

CN CO

Tt »*

t^ CN Cs

Z ^

JU3D J3J

to Tf

Tf CN

ID Tt Os

XHXH

o o

o o

o o

ooo

J3quin|>;

Ol !>.

Tf ro

<-0 CN

^ ^

Th cot'-

Tf i/^ CM r^

lO 1^ t^

00 lO lO

Os OS O CN

T-l CO Tj* O

CO oo

CN CO CO

Tt ■* liO ■*

2^ ^;

t^O-i'rt

CN T-H ^H

Tt OO

t^ C-) 00 T-l

S<HXH

o oc o

o o o

ooo

OOOO

aaquinx

l:^ 00

CN

CO fO CO

CN T-< T-H

•^ t^ rt< 'ti
CN CN

^

vO—i

, 1 ,_,

^

t^ O '-H

^

ro 00 CN ON

00 00

lO OS Ov

•.-H 00 CN t--

^ ^

O'-H OO

rt< oo

00 i-lt^ CN

-^

ooo o

o o

o oo

OOOO

ij

J9quin^\[

CN '-1

t— CN »-l T-l

CN CN

lO ^ ^

O^CN00 CO

£-<

tC

iTiin o

>n

lO lO lO

O "0 lO to

^ ^

CN CM lO

CN

CN CN CN

l/^ CN CN t^

^

o 'b

oo o

o

OO o

OOO

jaquinx

■rt Ti<

^T-t CN

-

^^ ^

CN i-l '-H CO

vO fO

o

CO

O CO OS

O ro

^

CO

SO CO OS

s

^ ^S

5U3D J3J

O ro

o

CO

SO CO OS

"2

c

>^X^
o ^b

O O

o

o

OO O

•»(

jaquin^

T-H (V)

CN i-(

CN

-

<M r-H CO

H

■^

ro O

o

CO

CO O OS

S

m \o

■o

CO

CO so OS

•"S"

^ c^

ro O

o

CO

CO so OS

'^

fc<

>HX^
o 'b

o o

o

o

OOO

jgquinsj

^ CO

■rt CN

CN

^

tH CN CO

-^

•^ rsi to t--

O to CN

to VO lO lO

SO CS OS Os

■^

ro 00 O ^H

ro 0\ »0

0\ Os VO VO

■rt 00 Os OS

^

tf ;z;

rj< CN O CN

CN '-H «-l

'-'OOO

t^Ol TjH ^

^

'S
5

^^XH
Of lb

oooo

oo o

oooo

OOOO

jaquin^

00 VO

O ro CO O
CN '-I '-I

■^ 0\ t^

OS Tj^ coco

CO CO CO CO
CO T-< CN CN

"•r

liO

iT)

O O

^ ^

o

CN

t~~

o o

^'U

o 'b

^

o

o

^ T-H

Co

J3quin|(j

■rt -^

■*

-^

CO

^ '^

r^

o

O

o o

f^

c/5 ;z;

5U33 HJ

o

CO

CN

o o

rC'

G^X^

1-H

o

o

T^^ ^

!<;

o lb

jaquin^;

T-HIO

ir>

'^

^— 1

lO lO

s-

O ro

CO CO

CO

so CO so CO

«i

O ro

CO CO

CO

so CO so CO

bo

^ ^

1U3D J3J

O ro

CO CO

CO

SO CO O CO

'^

o 'b

o o

o o

o

OOOO

jaquitiNj

■r-1 CO

CN r-H

^

^

CN »-l CN ■rt

'^

?-^

>

t

'T ? c. c

'? r ^ a

f ? « n

2s>?

1

^ &*^^

^^ f^ Z Z

^H^l

^

12

"rt

*■ s^

^ V-

C a

C a

tAS

^■rt

<u

'^ rt

^ s

1)

a

s s

•1-J u

•rt^

cS

6

rt4i

^-!

^

<^

1»H

u.

^

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

57

In Mating IV ( 9 W-S X d" T-N) in which case the male
parent is tame and non-savage the percentages for wildness
and tameness are identical, while those for savageness and non-
savageness are 0.432 and 0.566 respectively. This seems to
indicate that the hereditary influence of the parents was about
equally divided between wildness and tameness and between
savageness and non-savageness.

TABLE 43

Summary of Malings of F2 Hybrids, Scries B, and Rcsidls of Individual Inheritances

of F3 Hybrids, Series B

F2 MATINGS

I

9 W-S

X
cf W-S

II

9 W-S

X
cf W-N

IX

9 T-N

X

cfT-N

Number

Per cent

Number

Per cent

Number

Per cent

Litters

w-s
w-n
t-s
t-n

w-s
w-n
t-s
t-n

w-s
w-n
t-s
t-n

Wild

Tame

Sav.

Non-sav.

10
41

27

10

1

3

9

7
1
3

18
3

37

4

28

13

0.658
0.243
0.034
0.073

0.219
0.17
0.024
0.073

0.439
0.073

0.901
0.097
0.682
0.316

1

2
2

2

2
2

1.00

1.00

1.00
1.00

2
6

5
1

2

3
1

6

5
1

Mice total

Males and females <

0.833
0.166

Males ■

Females ■

Males and females <

0.333

0.50
0.166

0.999

0.833
0.166

In Mating VI ( 9 W-N X cf W-N) the wildness percentage
is 0.665 and the non-savage 0.575 in comparison with 0.333 for
tameness and 0.432 for savageness. The effect of the grading
in the parents seems to be here also quite apparent.

The percentage for non-savageness in Mating VII ( 9 W-N
X cf T-N) is 0.651 in comparison with 0.347 for savageness.
While this seems to show the effect of the non-savageness of
the parents, there was no equalization of wildness and tameness
as the percentages for these are 0.607 and 0.391 respectively.

58 CHARLES A. COBURN

In the results of the tests of the 13 mice from Mating XI
( 9 W-S X cf T-S) there is a contradiction of the results in
the preceding matings because, though both parents were
savage, 8 of the offspring were non-savage and 5 savage.

From the last division of table 41 similar results to those
just noted in the preceding table may be seen to exist fairly
regularly throughout the matings from which came the F2b
hybrids. The exceptions in this table are found in Matings
IV, VII, and VIII. In Mating IV ( 9 W-S X cf T-N) the
savage and non-savage ofifspring are rather evenly proportioned
but the Wild exceed the tame by 6 to 1 . In Mating VI I ( 9 W-N
X (^ T-N) the non-savage are slightly more numerous than
the savage but, as in Mating IV, the wild are almost 8 times
greater in numbers than the tame. Likewise the 6 mice from
Mating VIII (9 T-N X c^ W-N) show practically opposite
results of inheritance since all 6 were in the w41d class and the
proportion of non-savage to savage was 1 to 5.

In tables 42 and 43 the summaries of the tests of the Fsa
and Fsb hybrids are presented. Evidence that the grade of
wildness and savageness existent in the parent tends to pro-
duce a similar grade in the offspring is to be found in the re-
sults of Mating I of both tables, in the results for wildness and
tameness in Mating II of each table, and in Mating III of
table 42, in the results for savageness and non-savageness
in Mating IV, and also in the complete results of Mating VI
and VII of this same table.

The exceptions in the case of Mating IX ( 9 T-N X cf T-N)
of both series and of Mating X ( 9 T-N X cf T-S) of Series A
(table 42) are interesting inasmuch as the number of wild in
the offspring is in excess of the number of tame although both
parents were in each case classed as tame. The number of
savage individuals exceeds that of the non-savage in Mating IX
although both parents were non-savage. This is likewise true
in Mating VIII ( 9 T-N X d^ W-N).

Mating I and Mating II are the only ones which appear
throughout both series of the two hybrid generations. The
results of these matings are presented in table 44 in order to
enable a more comprehensive comparison. The blending
heritable tendency that has been mentioned above may be
more clearly noted in this table. It is rather well defined

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

59

^

fen

CO

Kl

Co

^

CQ

^

•^

s

^

t-5

PQ

^

fH

CO

fcl

fc.

f>

ti,

s

^ fe<

CO

Co

•^^

CO

o

o

o o

M

JU33 J5(J

o

o o

2

CG

»-l

•^ -^

i3quin|.j

^ <N

CN

CN

CN CN

s

fl

^

ffi

o

O

o

s s

<

jnao laj

o

00

CN

o o

X

1

1/3

>— 1

o

o

■rt ■rH

jaquinivi

rt ir:

lO

Tf

-

IT) lO

t^ 00 -^

so OS OOIO

r'O t~^ OS OS

l:^ CN 00 ■rt

O ■* 00 o

1^ r^ lO ro

r~i -H cs cN

so CO lO ■<*

M

1U33 J3 J

"^ T-, O (N

^ O O CN

ro T^OO

O

r2

CO

o ooo

OOOO

OOOO

OOOO

jaquin^

t^ lO ro OS

so ^CN CO

•rH^ rt T-H

CN CN O •*
CN rt CN rt

<

J2

>>

ro T-^ ro

lO OS

CO r-1 ro

Tt* CO c^ ^

ir> o 00

CN t^ ro

CN 00 lO

rt 00 LO •*

t^

<;

5U33 J3 J

^ Tf< ro

O ■^ CN

rt CN '-H

SO CO rt 00

Ol

OO O

OO O

OO o

OOOO

jaquin^

00 C>

o 00 "0

tH t^ OS

iri rt so

^ lO SO CO
CN ■rt ro

OD CO -^ ro

OS rt< PO

Osro

rt t^ ro SO

iri TjH (rq t-~

T-H 1~~ CN f^

ro t^

m

}U33 J3<I

VO CS O O

CN ■^ O O

^ O

Os O so CO

5?

CO

OO OO

OOOO

OO

OOOO

J9quinj.j

t^ O ^ fO

OS t^ -^ ro

CO (^

t^ •^ 00 CO
CO CN rt

so fO

ro ro

<^

SO CO SO CO

^

so ro

ro ro

ro

SO CO SO CO

fa

<

}U33 J3J

so ro

CO ro

ro

SO CO so CO

X

m

1

CO

o o

O O

O

OOOO

jsqiunjyj

i-H CC

CN ■^

^

^

CN ■rt CN ■rt

GC r^j t^ •--(

OS OO CO

Os ■* ro

00 to CO

tJ< Tt< tN 00

OS OO ^ so

■rt lO rt <M

0\ OO- CN

«

jaao J3 J

t^ ^ OO

CN O O O

^ OOO

00 T-lt^ CN

o
•z.

T3

OO O O

OOOO

OOOO

OOOO

jsquin^

00 t^

O '-^ ■* <-<i

•^ ro csi Os

so OO CN ro

■rt so Tt< CO

H

CN -^

^ CN ■rt

•* T^

so

CO ■rt rt CO

<

s

^-H T— 1

LO •* lO ■^

00 in 00 ro

r^ OO 1^ rt

C3s Os 00

CO »^ OO -^

CS 00 CN T^H

lO CN lO J^

OS OS t^ CN

<

}U33 J3<I

^ ^ OfO

CS O O ^

CS O O rt

lO CO lO -^

O O O o

OOOO

OOOO

OOOO

a3quin^sj;

i-^ •rl< C-j '-I

CO fO ^ lO

OS rt CN so

rt tHOiO
CM rt C^ rt

T ? =« c

T ? c« c

=f ? c« C

s: ^^ J.

& ^ -til -M

^ ^.tL J.

^

^

^ s^

tn

OJ

g

S

o

«■

-a
c

o-

<u

(U

*j _^-

rt

rt

s

a

§

<5

§

^

S

1

60

CHARLES A. COBURN

throughout the Series of Mating I but is noted only in the
inheritance of wildness and tameness in Mating II, the propor-
tion of savage and non-savage varying greatly in each case.

On the following pages there are presented three repre-
sentative genealogical charts which give the number of lit-
ters from each mating, the classification of the mice and
the numbers of the individual breeders.

w-s

7

1 Litter

W-S

5_

6 Litters

Fla

w_8 w-n

t-B trn_
9 1<? 2S 1?

1^

w-8 W-n t-8 t-n

F2a 2^ 2<f 1? 5^5? "-s w-n t-a t-n

t:A\ '^-'^

w-8 W-n t-s t-n

(Killed before teating)

1

1 Litter

w-B w-n t-8 t-n
41^

w-8 W-n t-s t-n
2S 1^ 1?

FJa

w-s w-n t-8 t-n
iS 12 5<3' 12 2?

1

1 Litter

w-B W-n t-8 t-n
2^ 3$

w-8 W-n
2^ 2$

t-8 t-a
1?

Chart 1.

A closer study of these charts reveals instances of inheritance
which suggest the possibility of some slight segregation of the
behavior-complexes under consideration. An example of this
may be found in charts 1 and 2, graphically illustrated as follows:

Fia

9 W-N X c?T-N

9 W-N X d^W-S

9W-SX c?W-S

F2a

Fsa

W-N

9 429-

W-N
-cf438-

T-N

-9 351

w-s w-n t-s t-n
2 1

w-s w-n t-s t-n

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE

61

r-^r

^Ch^

r'or

u

62

CHARLES A. COBURN

As is seen here 9 429 came from a 9 W-N X cf T-N mating,
cf 438 from a 9 W-N X d' W-S mating, and 9 351 from a
9 W-S X cf W-S mating. From the mating of 9 429 (W-N)
with cf 438 (W-N) there were 2 w-n and 1 t-n mice, while from
the mating of 9 351 (T-N) with cf 438 (W-N) all the five
mice were w-s. From the blending type of inheritance alone
we would be led to expect a higher grading from the former

P2b

F3b

W-S

?

I
225

T-N
<?

504

W-3

9
1

T-M
1

1

343

I

W-8 "-8

Fib 2<ri«

677
I

1 Litter
' — v=s-

5<?

1 Litter

I ^

w-8 W-n t-8

2<?

tr=n
679

w-8

4<?g2

I I

i26 629

M

5 Litters

5 Litters

I ,

w-n x-e

2$

Tin—
26 1$ 7<? 1?

)?.

395 39A(IJied before

tested)

42^-

2^ 2(? 12

t-B

1 Litter

2(?' 12
(Died before

tested)

w-8 w-n

3d"

(1 died after let. teat)

Chart 3.

t-n

mating than from the latter since both parents of the former
were W-N, while the female in the latter was a T-N. To find
a possible explanation we must go back to the preceding gen-
eration. Here we find in the case of the former mating an
ancestry consisting of two female W-N, one male T-N and the
other W-S, while in the case of the latter mating one female
of the preceding generation was a W-N and the other female,
and the two males were each W-S. Other cases are to be found

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE 63

where wildness seems to behave in a similar manner, as for
instance, in chart 2 (F2a, 9 420 with c? 377) and in chart 3
(Fab, 9 894 with d^ 893).

While the segregation of the various determining factors of
savageness or wildness and their behavior as a recessive or,
as it seems in some instances, a dominant, apparently explains
the inheritance in these various crosses, the number of matings
giving such results is comparatively few, hence, a statement
more definite than a mere suggestion of a possibility must
necessarily be left until a further study in this particular field
has been made.

OBSERVATIONS ON "SINGING" IN MICE^

In the literature of animal behavior appear several refer-
ences, to the production by mice of sounds of musical quality.

The "singing" of mice is described variously by different
writers. Lee (1878) states that it consists of a series of chirps
at the rate of three or four per second. At the beginning of
the series, the chirps are low but gradually they become louder.
The "song" of one mouse this author likens to the sweet and
varied warbling of a canary. Every note was "clear and dis-
tinct."

In referring to the same phenomenon, the naturalist Brehm
(1896), attributes the following descriptions to various observ-
ers. One informant states that the "song" is an irregular
mixture of chirps and trills with here and there a snarling,
smacking sound followed by a low murmur. Another describes
it as a twitter which is a mixture of long drawn squeaking and
piping sounds which may be heard at a distance of twenty
paces.

One observer noted the phenomenon only in the case of a
female mouse while giving birth to young, while another ob-
server states that only the male "sings."

The majority of those who have heard "singing" in mice
have assumed that it is due to a diseased condition of the lungs
or of the vocal organs, but conditions so diverse as pregnancy
and parasites in the liver have also been suggested as causes.

- Revised note which appeared under heading of Singing Mice, Journal of
Animal Behavior, 1912, ii, No. 5, pp. 363-366.

64 CHARLES A. COBURN

The writer desires to add, to the observations already re-
ported, an additional record of "singing" mice. About the
first of December, 1911, while working one evening in his study,
he heard a series of sounds which seemed to come from above
the ceiling. At the time, they were thought to resemble the
soft chirping of a bird.

Shortly afterward, some wild mice were needed for breeding
experiments and, by means of a trap, two mice, a male and a
female, were captured in the room.

These animals, while being taken to the Harvard Psychologi-
cal Laboratory, produced sounds like those previously heard
in the room and they continued to do so at intervals after
being placed in a laboratory cage. A few days after their
capture, the male escaped. Inasmuch as the writer had not
separated the two mice, placed each in a separate cage in a
different part of the room, it is not possible to state whether
the sounds were produced by both the mice or only by the
female which remained and continued to "sing."

The female was mated with a tame mouse and produced,
during the period of observation, five litters, thirty-three indi-
viduals. None of these offspring produced unusual sounds,
nor did "singing" appear in the second or the third generations
obtained from these hybrids.

The "singing" individual, so far as could be ascertained,
was a common house mouse (Mus musculus). She was some-
what larger than the ordinary wild female, but no other exter-
nal peculiarities were noted. She was extremely active and
savage and her mate always bore the marks of her teeth. An
attempt to mate her with a second tame male resulted in the
death of the latter.

No definite time for "singing" was noted, except three or
four days before and for six or seven days after the birth of a
litter. It was observed, also, that the individual "sang"
sometimes when frightened.

The sound is best described as a rapid whole-toned trill
involving the tones C and D as is indicated below.

1^ T^ lJ '^ --^U^^

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE 65

The quality of the tone resembled somewhat that of a fife
or flute, but each tone ended with a slight throaty click. The
tones were uttered at the rate of four or five per second in
groups of varying size. Sometimes a group occupied one
second, sometimes as long as ten seconds. As a rule, the tones
of a group were not clear and distinct but, instead, were uttered
so rapidly as to seem connected. The throaty click was more
noticeable in the case of the last tone of a group. Often the
"singing" would be continued for a period of ten or fifteen
minutes with rests between groups.

The sound could readily be heard at a distance of fifteen or
twenty feet, but it was difficult to localize it. The individual
"sang" little during June, 1912, and it was not heard after
July 1st, 1912. She died in August, apparently of old age.

During May, 1912, "singing" was again heard in the room
in which the "singing" mice had earlier been captured, but
efforts to capture the "singer" failed.

In January, 1913, a "singing" mouse was captured in the
home of an Italian family in Brooklyn, New York. It was
brought to the Laboratory and appears in the study described
in this paper as female no. 475. There was captured with it
a male, mouse no. 474, which, however, did not "sing." These
two mice were mated and produced six of the wild individuals
which were raised and tested and later used as breeders in
Series B of the study in wildness and savageness inheritance.
This "singer" was mated with a tame male, and again with a
F2 hybrid male but no "singer" appeared in these two litters,
nor among the first, second, or third generation hybrids from
the six wild offspring mated with tame individuals.

Later another "singer" was found on a farm in Michigan.
It was sent to the Laboratory and proved to be a female also.
No attempt was made at breeding this individual since it
arrived just before the end of the study and there would not
have been sufficient time for any observations to be made on
the offspring.

The "singing" of these last two individuals was very much
like that of the first, except that the mouse from Michigan
did not "sing" as long at any time nor as frequently as did the
other two.

The writer received several letters from different persons
in various parts of the United States giving their observations

66 CHARLES A. COBURN

on this "singing" phenomenon in mice. In each case the
observations had been merely casual and the reports gave no
new information on the subject.

The results of this investigation, though by no means com-
plete, indicate fairly definitely that this "singing" characteris-
tic in mice is not heritable. From the fact that no males were
captured that were definitely known to possess the characteris-
tic, and the phenomenon in two of the three individuals observed
was more often noted just before and after the birth of young,
it seems quite possible that it is a characteristic peculiar to
females and, perchance, due to a diseased condition or otherwise
structural defect of the vocal organs, caused or, at least, accen-
tuated by the birth of young.

SUMMARY AND CONCLUSIONS

1. When the grades of wildness and savageness attained by
mice of each hybrid generation are considered collectively,
the lowering of the average grades of wildness and savageness
in the successive tests is gradual in each generation but is great-
est in both wildness and savageness in the second hybrid gen-
eration. The greatest amount of difference in the grade of
wildness for any two consecutive tests occurs between the second
and third.

2. The mice of Series B, as a whole, graded higher than those
of Series A in both wildness and savageness. This result does
not indicate the presence of wild blood in the tame parents of
the hybrids of Series A, and hence is directly opposed to the
results Professor Yerkes obtained with the rats.

3. Savageness is more usually associated with wildness than
with tameness.

4. The females, as a rule, attained higher grades in wildness
and savageness than the males.

5. From the results of the tests of the first and second hybrid
generations of Series A, it was found that, generally speaking,
the older the mice the lesser the average grades in wildness
and savageness attained by them in the five tests.

6. The decrease in the grades of wildness and savageness
continues with the successive tests when the number of tests
is increased to eight.

7. Age, frequency of tests, number of tests, and the pres-
ence of the experimenter in the room, each has a certain effect

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE 67

in the lowering of the grade of wildness and savageness in
the hybrids.

8. The hybrids of Series A from the matings of parental
generation tame female with wild male became generally less
wild and less savage in the succeeding generations, while the
hybrids of Series B of this mating attained an increasingly
higher average grade of wildness and savageness in the suc-
cessive generations. The females of both series from this mating
were more wild and savage in the first and second generations
and the females of Series A more savage in the third generation
than the males, the males of both series being more wild than
the females in the third generation, while the males of Series
B were also more savage than the females in this generation.

9. The average grades attained in wildness and savageness
in the first generation hybrids of both series, from matings
of parental generation wild female with tame male, were, as a
rule, lowered in the second generation and then, in the third
generation, raised almost to the height of the grades obtained
in the first. The grades received by the females of both series
from this mating are in excess of those obtained by the males
except in the case of the males of Series A, which graded higher
in wildness in the first generation and higher in savageness in
the third generation than the females of Series A.

10. Under the conditions existing in this study it is very
doubtful whether there was much, if any, difference in the
inheritance of wildness and savageness by the hybrids due to
the fact that one type of parental mating was wild female and
tame male and the other, tame female and wild male.

11. When the individual mice are separated into the four
classes; wild-savage, wild-non-savage, tame-savage, tame-non-
savage, according to the average of the third, fourth and fifth
tests for wildness or savageness being equal to or above 2.5
grade or below this, the wild and savage class, as a rule, rep-
resented the greatest number of mice, the wild and non-savage
class the next highest, the tame and non-savage the next, and
the tame and savage class the least number. The one excep-
tion to this is in the second generation hybrids of Series A, for
which the order was as follows: (1) tame-non-savage, (2)
wild-savage, (3) wild-non-savage, (4) tame-savage, where 1
represents the greatest number of mice and 4 the lowest.

68 CHARLES A. COBURN

12. The number of wild and savage females in each hybrid
generation is always greater than the number of wild" and sav-
age males.

13. In the experiments in selective breeding it was generally
to be noted that there was a tendency for the degree of wildness
or savageness possessed by the parents to be reproduced in the
offspring, that is, if both parents were wild, (the average of
their third, fourth and fifth tests for wildness being equal to or
exceeding 2.5 grade), in the majority of cases it would be found
that the number of wild offspring was greater than that of tame,
or if the parents were tame the greater number of offspring
would be tame. The occasional isolated case in which this
tendency did not maintain and where the results seem to indi-
cate some particular combination of hereditary units is very
interesting. It confirms the conviction of the writer that much
more work is needed in this field before the precise manner of the
inheritance of wildness and savageness in mice can be definitely
stated. However, from this study it seems that the two be-
havior-complexes in question are the result of several different
inheritance factors which follow the Mendelian rules.

14. Three female wild mice were found which possessed the
"singing" property. No male was obtained which had this
characteristic. This "singing" did not appear in any of the
offspring to the third generation. It is evidently not inheri-
table, and quite possibly is a characteristic peculiar to females.
The phenomenon seems to be associated in some way with the
birth of young, and probably is caused by some structural defect
or a diseased condition of the vocal organs.

BIBLIOGRAPHY

G ALTON, Francis.

1869. Hereditary genius, an inquiry into its laws and consequences.
London, pp. 390.

1875. A theory of heredity. Contcrnp. Review, xxvii, pp. 80-95.
Lee, Henry.

1878. Singing mice. Popular Science Monthly, xiv, pp. 102-105.
G ALTON, Francis.

1889. Natural inheritance. London, pp. 259.
Ball, W. P.

1890. Are the effects of use and disuse inherited. London, pp. 156.
Collins, F. H.

1891. The diminution of the jaw in the civilized races, and effect of disuse.
London, pp. 16.

HEREDITY OF WTLDNESS AND SAVAGENESS IN MICE 69

Packard, A. S.

1893. An attempt to distinguish between congenital and acquired char-
acters. Proc. Amer. Phil. Soc, xxxi, pp. 139-192.
Baldwin, J. ]\I.

1896. Physical and social heredity. Amer. Nature, xxx, pp. 422-428.
Breigi, a. E.

1896. The life of animals. Chicago, p. 338.
Davenport, C. B.

1896. Experimental morphology. New York, pp. 280. (Second Vol-
ume, 1899.)

G ALTON, Francis.

1897. The average contribution of each several ancestor to the total herit-
age of the offspring. Proc. Roy. Soc, London, Ixi, pp. 401-413.

Shuffeldt, R. W.

1897. Natural history of the United States, p. 420.
Davenport, C. B.

1899. Statistical methods with special reference to biological variation.
New York and London, pp. 148.

Hamilton, D. J.

1900. On heredity in disease. Scot. Med.b' Surg. Journal, 'v, pp. 289-303.
Irelantd, W. W.

1900. Degeneration, a study in anthropology. Intcrnat. Monthly, i,
pp. 235-279.

Butler, A. W.

1901. A notable factor of social degeneration. Science, xiv, pp. 444-453.
Darbishire, A. D.

1902. Note on results of crossing Japanese waltzing mice with European
albino races. Biomctrika, ii, pp. 101-104.

Blanchard, N.

1903. On inheritance (grandparent and offspring) in thoroughbred race
horses. Biomctrika, ii, pp. 229-233.

Castle, W. E.

1903. Mendel's law of heredity. Proc. Amer. Acad., xxxviii, pp. 535-548.

Science, xviii, pp. 396-406.
1903. The heredity of Angora coat in mammals. Science, xviii, pp. 760-761.
1903. Heredity of sex. Bidl. Museum Comp. Zool. Harvard, xi, no. 4,

pp. 189-218.
1903. The laws of heredity of Gal ton and Mendel and some laws governing
race improvement by selection. Proc. Amer. Acad., xxxix, no. 8,
pp. 223-242.
Castle, W. E., and Allen, G. M.

1903. The heredity of albinism. Proc. Amer. Acad., xxxviii, pp. 603-622.
Darbishire, A. D.

1903. Note on results of crossing Japanese waltzing mice with European
albino races. Second Report, Biomctrika, ii, pp. 165-173. Third
Report, Biomctrika, ii, pp. 282-285.
Doncaster, L.

1903. Experiments of hybridization, with especial reference to effect of
conditions on dominance. Phil. Trans., cxcvi, pp. 119-173.

70 CHARLES A. COBURN

Lee, Alice.

1903. On inheritance (g. grandparents and offspring) in thoroughbred
race horses. Biometrika, ii, pp. 234-240.
Pearson, Karl.

1903. On inheritance of mental and moral characters in man, and its com-
parison with the inheritance of physical characters. Trans. Anthropo-
logical Inst, of Gr. Britain and Ireland, pp. 179-237.

Ellis, Havelock.

1904. A study of British genius. London, pp. 300.
G ALTON, Francis.

1904. Distribution of successes and of natural ability among the kinsfolk
of Fellows of Royal Society. Nature, Ixx, pp. 354-357.
Washburn, M. F.

1904. A factor in mental development. Philos. Rev., xiii, 622-636.
Castle, W. E.

1905. Recent discoveries in heredity and their bearing on animal breeding.
Pop. Sci. Monthly, Ixvii, 193.

1905. Heredity of coat characters in guinea pigs and rabbits. Carnegie
Inst. Publication, no. 23, pp. 1-78.
Darbishire, a. D.

1905. On the supposed antagonism of Mendilian to biometric theories of
heredity. Mem. and Proc. of Manchester Lit. and Phil. Soc, xlix,
no. 6.
Davenport, C. B.

1905. Evolution without mutation. Jour, of E.vper. Zool., ii, pp. 137-143.
Ingersoll, E.

1906. Life of mammals. New York, p. 429.
Woods, F. A.

1906. Mental and moral heredity in royalty. New York, pp. 312.

1906. Family ability. Nature, Ixxiv, pp. 97-98.
Yerkes, R. M,

1907. The dancing mouse; a study in animal behavior. New York, pp. 290.
Bentley, M.

1909. Mental inheritance. Popular Science, Ixxv, pp.. 458-468.
Doncaster, L.

1910. Heredity in light of recent research. University Press, Cambridge,
pp. 140.

Davenport, C. B.

1911. The study of human heredity. Cold Spring Harbor, New York,
pp. 17. (Joint authors: Laughlin, H. H., Weeks, D. F., Johnstone,
E. R., and Goddard, H. H.)

Sumner, F. B.

1911. Effects of temperature on growing mice and the persistence of such
effects on subsequent generations. Amer. Nat., xlv, pp. 90-98.

Coburn, C. a.

1912. Singing mice. Jr. Animal Behavior, ii, no. 5, pp. 364-366.
Phillips, J. C.

1912. Note on wildness in ducklings. //'. Animal Behavior, ii, no. 5,

pp. 363-364.
1912. Size inheritance in ducks. /;-. Exp. Zoology, xii.

HEREDITY OF WILDNESS AND SAVAGENESS IN MICE 71

Emerson, R. A., and East, E. M.

1913. The inheritance of qualitative characters in maize. Res. Bull. 2,
Agr. Exp. Sta. Neb.
Morgan, T. H.

1913. Heredity and sex. Columbia Univ. Press, New York, pp. 282.
Yerkes, R. M.

1913. The heredity of savageness and wildness in rats. Jr. Animal Behavior,
iii, no. 4, pp. 286-296.

MacDowell, E. C.

1914. Size inheritance in rabbits. Carnegie Inst. Wash. Pub. 196.
Phillips, J. C.

1914. A further study of size in ducks. Jr. Exp. Zool., xvi.
CONKLIN, E. G.

1915. Heredity and environment in development of men. Princeton Univ.
Press, Princeton, pp. 533.

Castle, W. E.

1916. Genetics and eugenics. Harvard University Press.
Morgan, T. H.

1919. The physical basis of heredity. J. B. Lippincott Co., Philadelphia,
pp. 300.

EDITORIAL ANNOUNCEMENT

This number completes volume 4 of the Behavior Monograph Series.
Hereafter the series will appear under the title COMPARATIVE
PSYCHOLOGY MONOGRAPHS. The new series will be under the
editorship of Professor Walter S. Hunter of the University of Kansas,
with the co-operation of Professor H. A. Carr of the University of
Chicago, Professor S. J. Holmes of the University of California,
Professor K. S. Lashley of the University of Minnesota, and
Dr. Robert M. Yerkes of the National Research Council.

Manuscripts may be sent to Professor Hunter at Lawrence, Kansas,
or to any other member of the editorial board.

The new monograph series will be broader in scope than the old
one and, in addition to studies in animal behavior, will publish work
in human psychology conducted from the comparative point of view.

Comparative Psychology Monographs

Volume 1

1. Studies of the rehability of the problem-box and the maze. By
W. T. Heron and W. S. Hunter.

2. A Behavioristic Study of the Activity of the Rat. By Curt P.
Richter.

Price per volume of 400 pages, net postpaid, $5.00, United States; S5.25,
Canada; $5.50, other countries.

Published by

WILLIAMS & WILKINS COMPANY

Mt. Royal and Guilford Avenues

Baltimore, U. S. A.

The Behavior Monographs

Edited by WALTER S. HUNTER
University of Kansas

VOLUME 1

No. 1 The development of certain instincts and habits in chicks. By Ferderick S. Breed.

Pp. iv + 78, $L00, postpaid.
No. 2 Methods of studying vision in animals. ByRobertM.Yerkes and John B.Watson.

Pp. iv + 90, $1.25, postpaid.
No. 3 An experimental study on the death-feigning of Belostoma (-Zaitha Aucct.) flumin-

eum Say and Nepa apiculata Uhler. By Henry H. P. Severin and Harry C.

Severin. Pp. iii + 47, $.65, postpaid.
No. 4 The biology of Physa. By Jean Dawson. Pp. iii + 120, $1.50, postpaid.
No. 5 The function of the vibrissae in the behavior of the white rat. By Stella Burnham

Vincent. Pp. iv + 81, $1.15, postpaid.

VOLUME 2

No. 6 The delayed reaction in animals and children. By Walter S. Hunter. Pp. v + 86,

$1.15, postpaid.
No. 7 The Canada porcupine: a study of the learning process. By Leroy Walter

Sackett. Pp. iii + 84, $1.15, postpaid.
No. 8 Audition and habit formation in the dog. By Harry Miles Johnson. Pp.

iv + 78, $1.00, postpaid.
No. 9 Habit formation in a strain of albino rats of less than normal brain weight. By

Gardner Cheney Basset. Pp. iv + 46, $.65, postpaid.
No. 10 Distribution of effort in learning in the white rat. By John Linck Ulrich. Pp.

iii + 51, $.65, postpaid.
No. 11 The effect of age on habit formation in the albino rat. By Helen B. Hubbert,

Pp. v + 55, $.65, postpaid.

VOLUME 3

No. 12 The mental life of monkeys and apes: a study of ideational behavior. By Robert
M. Yerkes. Pp. iv + 145, $1.50, postpaid.

No. 13 A study of perseverance reactions in primates and rodents. By G. V. Hamilton.
Pp. iv + 65. $.75, postpaid.

No. 14 An analysis of the learning process in the snail, physa gyrina Say. By Elizabeth
Lockwood Thompson. Pp. iii + 97, $1.25, postpaid.

No. 15 The effect of length of blind alleys on maze learning: An experiment on twenty-
four white rats. By Joseph Peterson. Pp. iii + 53, $.75, postpaid.

No. 16 An experimental study of transfer of response in the white rat. By Harry H.
Wylie. Pp. iii + 66, $1.00, postpaid.

VOLUME 4

No. 17 Transfer of training in white rats upon various series of mazes. By Rutledge

T. Wiltbank. Pp. iii + 65, $1.00, postpaid.
No. 18 Redintegration in the allaino rat. By William Brockbank, A.M. Pp. iii + 66,

$1.00, postpaid.
No. 19 Discrimination of light of different wave-lengths by fish. By Cora D. Reeves.

Pp. iv + 106, $1.25, postpaid.
No. 20 Visual perception of the chick. By Harold C. Bingham. Pp. vi -f 104, $1.50,

postpaid.
No. 21 Heredity of wildness and savageness in mice. By Charles A. Coburn. Pp. iv +

71, $1.00, postpaid.

WILLIAMS & WILKINS COMPANY

Mount Royal and Guilford Avenues
Baltimore, U. S. A.

3 2044 106 181 886

?V T,

►■^■- •■"W«>4r ;,

.^5r-

t'^i^'^

>i\^

■^H^^-

^^•^-vJ

•^^*' -OiN