Pd.
um T:

.53; as A {WOW}! I}?

h3g3 8!...“
.vi’ 1- n‘hyl

Luna. 9:?!

9.x” Pun“.

«Pu. Dan
at 1v... Thur”
town“)? 6...!“— Y;
a»: 35m. 3
H...» mm 5.
erg at. 3 MM“. 8
:9.“ nan“ E

,. ”Mu.” uh... «Wu 4

n. .3 .3. .5. Cu
If...” h“, .\ 4'!
“NH 7th” n «flan.
warm Pl... rwhu.
min, y .67..
m k. a. .‘M. P
may: HM.» an}
P7... .. J 3 wt
Poll. ".11.“. r.
“Hum in» “In. , .

‘1!
Sfl v. 1 «Kb
?OPJII p; In.” “0.9
an 5...
file...“
”KIM
an“
.9... ..
Sufi.
E
in.»

:
:

_
y

2

_.
__
_

. E
__
. _:

 

~__:___:_.t_:_:_Z

AM G01?

7

5

‘

L:

7

BY
JEL

'e

71‘

ROBERT

This is to certify that the

thesis entitled
Learning of a differential

reSponse as a function of
stimulus-re3ponse asynchrmnism

presented by

Robert W. Coy

has been accepted towards fulfillment

of the requirements for

* if . AJ__ __-_degree irtifi§jl§£0l9 FY

 

 

LEARNING OF A DIFFERENTIAL RESPONSE AS A FUNCTION OF
STIMULUS-RESPONSE ASYCHRONISM

by
Robert William.Goy

A.THESIS

Submitted to the Graduate School of Michigan
State College of Agriculture and Applied
Science in partial fulfilment of the
requirements for the degree of

MASTER.OF ARTS

Department of Psychology
1948

ACKNOHLEDGEMENTS

The writer wishes to empress his deep appreciation to Dr.
M3 R. Denny for his willing guidance in the work reported in this
paper, and for his constructive criticism and complete patience
in the preparation of this manuscript. The author also wishes to
acknowledge the courtesy of Dr. H. R. Hunt, of the Department of
Zoology, in providing Space in his animal laboratory so that the
present study might take place. To the members of the Department
of Psychology whose patient discussions have aided in clearing a
visible path for this paper, and to all other friends whose
assistance in the course of this work has been of great value,

the writer also wishes to extend his gratitude and deep appreciation.

20 804.5

TABLE Or CONTENTS

Introduction ‘
Theoretical Background
Experimental Technique

I. Apparatus

II. Subjects

III.Procedure
Results

Table I

Fig. 1

Table II

Figure 2

Figure 5

Table III

Figure h

Figure 5
Discussion

Table IV

Figure 6

Figure 7

Table V

CONCLUSIONS

Summary
Reference

Appendix

nu
h6
1+7

INTRODUCTION
The present study is based on an account of the
temporal aSpects of learning in a reinforcement situation

as presented by Hull (5) in his book, "Principles of

 

IBehavior". The variable of time enters into Hull's de-
scription of learning in essentially three ways: one, the
delay of reward; two, frequency of the trials; three,
stimulus reSponse asynchronism.

The present investigation is concerned with the latter
role of the time variable, its former two functions in
this learning situation being held relatively constant.

The subject of trace conditioning has never been fully
eXploited in a systematic manner within any of the exist-
ing psychological systems, although concepts such as stim-
ulus trace, perseveration, and mediational events in gen-
eral have long been incorporated into most systems as
.necessary concepts for an adequate representation of vari-
ous behavioral phenomena. More comprehensive work is felt
to be immediately necessary in order to correct and preperly
extend current uses of this type of concept.

Evidence for the application of stimulus trace to
learning processes usually considered to be on a higher
level than simple conditioning is particularly lacking,
and those systems utilizing the concept in a description
of this kind of behavior have been rebuked for being too
molecular, unnecessarily complicated, or speculative. Those

systems distrusting the concept in this usage have generally

2

preferred labels of symbolic functions, representative
factors or association Spans. The present paper was
therefore designed to test the role of stimulus trace
in differential reSponse learning as this concept has

been formulated by one of the major current behavior

theorists.

THEORETICAL BACKGROUND

Probably the most thorough going and testable form-
ulations of stimulus trace in the field today is that
advanced by Hull. The concept itself enters his theory
early in his first postulate, although at this time its
relationship to learning is not stated nor is its mathe-
matical formulation in this particular capacity given un-
til the fourth postulate.

Hull states that both the limit of habit strength and
the rate of acquisition of habit strength are a function of
the magnitude of the stimuhmatrace at the time of occur-
ence of the response, other variables such as the delay
of reinforcement being equal. He states,

"Numerous eXperiments have shown that the gradient of
reinforcement remaining constant the most favorable tem-
poral arrangement for the delivery of the conditionfland
unconditioned stimuli is to have the latter follow the
former by something less than a half second. But as the
asynchronism of the onset of the two stimuli deviates from
thiscptimal relationship in either direction, there is a
falling off in the habit strength which will result from a
given quality and number and reinforcements, the rate of
decline in each direction probably being simple decay
function of the nature and extent of stimulus asynchronism".
(Hull, 5, p. 176).

The fourth postulate shows more clearly perhaps than
the foregoing paragraph the true role of stimulus trace in
a reinforcement situation as well as its relation to other
. variables. This postulate is as follows:

"Whenever an effector activity (r-R) and a race tor
activity (S-s) occur in close temporal contiguity (s r) and
this sCr is closely and consistently associated with the
diminution fif need (G) there will result an increment to a
tendency (s r) for that afferent impulse on later occasions
to evoke that reaction. The increments from successive
reinforcements summate in a manner which yiklds a combined

h

-habit strength (SHR) which is a simple positive growth
function of the number of reinforcements (N). The upper
limit (M) of this curve of habit growth is the product of
(l) a positive growth function of the magnitude of need
reduction which is involved in primary or associated with
secondar , reinforcement; (2) a ne ative function of the
delay (t in reinforcement; and (3 (a) a negative growth
function of the degree of asynchronism (t') of S and B
when both are of brief duration, or (b) in case the action
of S is prolonged so as to overlap the be inning of R, a
negatfve growth function of the duration t") of the con-
tinuous action of S on the receptor when R begins."

The relationships given in the above postulate are
expressed mathematically in the following manner:

SHE : M(1-e’kG)o‘Jta‘Ufl(1-3'1N)
where,

M’

The absolute physiological limit of habit strength
attainable under optimal conditions of learning
with unlimited reinforcement.

e : lO, the base of common logarithmns

G : the amount of need reduction as measured by objective
criteria such as size of reward

t : the delay in reinforcement

TR : the time of occurence of R
Té the time of onset of S

t' : TR-Té-hh, where S and R are of brief duration (.ha
has been empirically determined as the time of max-
imum recruitment).

N : number of reinforcements

k, i, u, and J : empirical constants.

In this particular problem, the size of reward is kept
constant for all experimental groups and thus affectsall
the acquisition curves in the same manner; the delay in
reinforcement is assumed to be zero so that this term in

the equation (e-Jt) remains Optimal and constant, and since

for any measure of direct comparison that may be employed
the number of reinforcements will be kept constant for

all groups, this term in the equation also has a constant
expression in the determination of the measures involved.
The fourth term in the equation (e’ut') is the one with
which we are directly concerned, and the major variable of
the present study. The empirical constant (u) is determined
from the curve fitted to the data and in this manner ex-
presses the Operation of the other variables.

The original formulation of stimulus trace has been
extended to include the effects of conditions such as the
frequency of presentation of S and the ordinal position
of S in a given series, (Reynolds, 5, p. 17), but these
factors are also uniform for the groups involved.

It may be well to give a verbal definition of the con-
cepts expressed in the above formulation which are important
to us. For this purpose Hilgard supplies a very good defi-
nition of stimulus trace as follows:

"Stimuli impinging upon a receptor give rise to afferent
neural impulses which rise quickly to a maximum intensity
and then diminish gradually. After the termination of the
stimulus, the activity of the afferent neural impulse con-
tinues in the central nervous tissue for some seconds."
(Hilgard, 2, p. 81).

In conjunction with the above, Hull's imposed condi—
tion for the generation of trace conditioned reaponses is
that S be of brief duration. This is defined by Hull as
an amount of time less than the amount of time required for
maximum receptor discharge. The time for maximum receptor

discharge has been determined behaviorally by recent invest-

igators and found to be about hSO msc. Thus, Operationally,

a brief stimulus is one whose duration is less than LEO
msc.

The decided parallel, if not covariation, of intra-
neural phenomena and the empirical results of trace types of
conditioning,insofar as temporal factors were ccncerned,led
Hull to hypothesize a neuralogical basis for stimulus trace.
This hypothesis states: Other things equal, the increment
to the strength of a receptor-effector connection resulting
from a reinforcement is an increasing function of the fre-
quency of the associated receptor discharge, or the inten-
sity of the resulting afferent impulse.

A fairly recent study by Kimble (h) however disclaims
the neuralogical basis of stimulus trace but asserts the
correctness of the mathematical treatment of the concept.
The neuralogical basis of this concept is however completely
unnecessary insofar as its descriptive use in behavior is
concerned. Kimble investigated time intervals of 100, 200,
225, 250, 500 and too msc. using an eye-wink reflex elicit-
able by an unconditioned stimulus of a puff of air and a
conditioned stimulus of a light source of .55 millilamberts
and a duration of 1500 msc. The time intervals were meas-
ured from the onset of the conditioned stimulus. Periodic
test trills were given during the conditioning process so
that the latency of the reSponse in question would not ob-
scure the data on acquisition. The results of the study
clearly show that the hOO msc. interval was the Optimal of

those used. These results are in essential agreement with

o

the previous study of Wolfle (9) in which she conditioned
finger withdrawal reSponse originally elecited by an elec-
tric shock to a conditioned stimulus of a sharp click. Of
the time intervals investigated an interval of 500 msc.
yeilded Optimal results. However from an extrapolation
Than thdsedata Hull places the Optimal interval at th msc.

Another study by Reynolds (5), using the eye-wink re-
flex to a puff of air and a conditioned stimulus of a click
of 50 msc. duration, investigated intervals of 250, h50,
1150 and 2250 msc also placed the Optimal interval around
h50 msc. The Heynold's study also investigated the effect
of massing the training trials and gives clear-cut evidence
as to the deleterious effect of this. Under conditions of
massed presentation; with only 10 to 20 records between
trials, both the rate of acquisition and the maximum level
of conditioning attainable under the 90 reinforced trials
given were considerably reduced.

The results of conditioning data in general then would
tend to confirm Hull's hypothesis Of an anterior and poster-
ior stimulus asynchronism gradient both Of which are simple
decay functions of an optimal interval ofestimulus asynchron-
ism around th msc. for conditioning. The investigations
also confirm the hypothesis that the limit of fall of this
gradient is substantially above 0 (around 20%) in the case
Of both visdral and auditory receptors.

Of those studies investigating stimulus trace in a

trial-andeerror situation, the studies of Warner (7) and

Wilson (8) are most prominent. Warner's study with white
rats consisted Of a reSponse of jumping a low fence for-
merly elicited by an electric shock becoming conditioned

to a cue-stimulus of a buzzer which lasted for one second.
Warner used intervals of one second, 10 secondg 20 seconds,
and 50 seconds between the cue-stimulus and the shock. The
time between trials was one minute for the one second group,
one minute and 20 seconds for the 10 second group, one min-
ute and ho seconds for the 20 second group, 2 minutes and
10 seconds for the 50 second group. The animals jumped a
low fence in the middle of a box from one half of the floor
which was charged with an electric current to the other
half which was not charged, but which became charged on

the next trial. The animals were given 50 trials a day.
Warner found learning in at least a few animals in all Of
the groups except the one with an interval of 50 seconds.
It is significant however, that the number of trials requir-
ed by the few animals in the 20 second group that did meet
the criterion is not significantly larger than the number
of trials required by the animals, in either the 1 sec. or
10 sec. groups. Furthermore, it is conceivable that the
criterion of six consecutive crossings could have been met
purely on the bases of chance behavior when the nature of
the apparatus is considered, eSpecially in the 20 second
group where such a large number of trials was given. And
in fact, an examination of the data Warner presents shows

very little Of the gradual accumulation of a reSponse tend-

ency so characteristic of learning in any except the I
second group. In addition, none of the animals, in any
of the groups other than the 1 sec. group, repeated the
correct response after the criterion had been met in any
consistent manner, but instead showed equally strong ten-
dencies to other kinds of escape and extraneous behavior
patterns. Hull attributes whatever learning may be said
to have been generated to a type of conditioning called
"cyclic-phase" conditioning. Because of the exactness of
the time intervals between trials which Warner maintained,
and the relatively constant rate of return equilibrium
of body tissues affected by shock, it is conceivable that
the animals under study were reSponding to some point on
the gradient of return to equilibrium rather than to the
buzzer itself. This hypothesis is in fact supported by the
results Of Warner's test trials. On the day after the an-
imals had reached the criterion, they were again placed in
the box and all Operations that had previously been performed
by the experimenter were repeated, and all conditions of the
experiment repeated with the exception that both the buz-
zer and the shock were omitted. In no case, under these
conditions, were the animals observed to cross the fence.
Warner attributes this to the absence of the buzzer.
However, such results are equally as predictable on the basis
of cyclic phase conditioning as may be seen in the following
analysis. 0

The data show that many of the animals could have attained

the criterion of six consecutive crossing purely on the basis

10

of chance, and further more, in those cases where the
frequency of correct reSponses was such as to indicate
learning, none Of the animals ever reSponded on a particular
day until they had received at least one electric shock indi-
cating that the shock and the gradual return to equilibrium
may well have been the stimulus evoking the reaponse. Thus
Warner's own test trials tend to validate Hull's explanation.
The Wilson study,called, "Symbolic Behavior in the
White Hat", (8), was presumably investigating delayed reSponse.
However, the familar experiments in delayed reSponse never take
place during the learning of the reSponse as Wilson's study
does, but always investigate the effects of delaying an already
learned reSponse to a given stimulus. To introduce varying time
intervals between the cue-stimuli and the differential reaponse,
Wilson used runways of three different lengths; an 8 inch alley,
a 2h inch alley and a 60 inch alley. He does hot record any
average times for traversing those distances, although 500 msc.,
1500 msc and hOOO msc. would seem to be adequate approximations.
The cue-stimuli consisted of (a) a forced right turn or (b).a"
forced left turn in a portion Of an H-shaped maze designated as
the stimulus chamber. The differently lengthed alleys were intro-
duced as the cross bar Of the H between the stimulus chamber and
the other arm Of the H was designated as the reSponse chamber.
The animal was required to learn a right turning reSponse in
the reSponse chamber following a forced right turning reSponse

in the:3timulus chamber with one Of the above mentioned time

intervals intervening between stimulus and reaponse period.

The animals were reinforced on each correct reSponse with "a

11

small nibble Of food". The criterion of mastery was 52 correct
choices in he consecutive'frials, 20 trials per day being

given. A chance sequence of the forced turning reSponses

was used. The goal boxes were placed at the ends of both

wings of the reaponse chamber, but on the occasions when all
animals made the wrong reSponse the door to the goal-box was
locked and early in training the animal was given a shock instead
of food ( a practice soon discontinued because of its disruptive
effects ). Acquisition curves are not presented in the paper,
but it is stated that animals were given as many as 1000 trials
in which to meet the criterion. The average for the 500 msc.
group was 716 trials, 792 for the 1500 msc group and 87h for

the h000 msc. group. However, in the latter groupoonly two
animals out of 11 used reached the criterion. This is equiva-
lent to 20% mastery whereas, in the 1500 msc. group 60% masterymms
reached, and in the 500 msc group 100% mastery was reached in
considerably less trials. The difficulty with the data presented
in this way is that they do not show the percentage of mastery
after a given number of reinforcements which is a determinant

of considerable importance in the approach utilised by the pre-
sent paper. Doubtless had this been done, a true gradient fall-
ing as low as 10% for the h,000 msc. group would have been ob-
tained. Even as presented however, the results offer no
disturbing departure from those Obtained in trace condition-

ing, although it must be remembered that the relationship

is here represented by data from only three groups with con-

siderable overlap from group to group. The actual limits

for the posterior asynchronism gradient are not determinate,

12
because several factors entered into the situation Operating
in such a way as to reduce both the rat. and maximum of
learning markedly. These factors were; 1) massing of trials,
2) the presence of the disruption due to shock in the initial
trials, 5) a delay in reinforcement due to the construction
of the apparatus that may have been as long as two seconds,
h) the fact that reinforcement occurred on incorrect as well
as correct reaponses due to the fact that the arms of the H
containing the goal chambers were identically constructed,
(see Denny, 1). Moreover, the data in the h,000 msc. group
are particularly obscured by the use of a runway 60 inches
long as a means of introducing the delay. At the beginning
of learning, the time in which this distance is traversed
may be greater than ten seconds, eSpecially since no goal
association had been built up by a period of preliminary
training. With these unfavorable conditions for learning, a
very large number of reinforcements are required to bring the
reaponse to the observable level - above threshold. According
to this concept, a habit must develop to a certain degree of
strength called the reSponse threshold in order to overcome
slight fluctuations in strength which reduce the reSponse
tendency to the degree where the habit does not make an
empirical showing, (Hull, 3).

It may be argued that on the basis of conditioning data
the posterior asynchronism gradient reaches its lower limit
or fall at about three seconds, and that time intervals

longer than this do not yield appreciable differences. It

13

is this very argument, however, which this paper attempts

to IhOW may not be a valid form of argument in so far as trace
learning is concerned. The facgfthat the asynchronism gradients
have neve£:§dequately worked out for differential response:

learning.

EXPERIMENTAL TECHNIQUE
I. Apparatus.

The apparatus used in this experiment is best divided
in two parts for the purpose of a clear presentation. The
first part or the apparatus to be discussed is that of the
simple T-Mlze. This part or the apparatus was built in
four sections.

.1. The Simple TuMaze.
l. The Starting Box.

The starting box was 9” long, 6' wide with walls and
floor or B/h' plybwood painted white. The box had a hinged
tOp made or 1/8” plydwood 6' wide, 7” long for the distal
portion or the cover. That portion or the cover, proximal
to the exit door or the starting box consisted of 2” or i"
mesh hardware cloth so that the rat's behavior in the
starting box could be observed. The cover to the starting
box also supported the stimuli, ie. the lights (three, 3
watt, G. E. Neon Lamps), and a buzzer obtained from.a small
commercial electric scalp vibrator. This type or buzzer
was used because of the moderate intensity or the buzzer.

The exit door of the starting box was 3" wide, L” high
and attached by a spring to the tap at the starting box
directly above. The door was held in a closed position by
a latch which was Opened electrically by a 2 lb. solenoid.
Thus when the latch was removed the exit door was raised
by the spring and allowed the rat to enter the second por-

tion of the maze.

0‘

15

2. The choice point of the maze.

This portion of the maze was built of 3/5” ply-wood.
The interior of it was painted a uniform.gray with an ap-
parent brightness approximately half way between the cone
trasting brightnesses of the two and boxes. At either end
of the choice point section gray woolen curtain were at-
tached to the sides of the choice box.and suspended from.

a cross bar one inch from.either end of the choice point.
The first three inches above the floor was:the first point
of attachment of the curtains to the side walls of the
alley so that the animal could gain entrance to the end

box by lifting the free flap of the curtain with his nose.
The walls of the choice box (and similarly the two curtains)
were nine inches high, and the whole thing was covered by

a removable piece of hardware cloth of iP mesh.

3. The negative End Box.

The negative end box was an alloy 5" wide and 18” long
with walls and floor composed of B/L' plybwood, and having
its entire interior painted with flat black house paint.
The tap of the negative end box was covered with removable
i” mesh hardware cloth. The entrance to the box, ie. the
end proximal to the curtain in the choice point contained
a vertical sliding door, one side of which (that side that
,might possibly have been visible to the rat inside the
choice point) was painted gray, and the other side of which

(the side visible from the inside of the negative end box
when the door was closed) was painted black. This door was
always open at the beginning of each trial, and closed imp
mediately after the animal entered the box.

A. The Positive Ind Box.

The positive and box or goal box had dimensions ident-
ical with that of the negative end box, but the internal ap-
pearance of the box.was made completely different. The
walls of this box were painted flat white. The floor of the
box was covered with a 5' thickness of standard soundproof-
ing material, the first 9' of which had been sanded to form
an uphill grade. At the distal end of the box a small crys-
tal food dish similiar to a common furniture coaster was
placed. The top of this box was left uncovered. The proxi-
mal end of the box contained a vertical sliding door ident-
ical with the door in the negative box with the exception
that the internal surface was painted white.

B. Electrical Synchronization; the second part of the ap-
paratus.

The second part of the apparatus consists of the elec-
trical wiring of the stimuli and their synchronization with
the electrical solenoid which Opened the exit door of the
starting box. Two electrodes, spaced about one inch apart
measured vertically and of the same length, were held sta-
tionary in such qhay that the surface of an ordinary piece
of graph paper stretched tightly over a revolving drum on an\

.17

electrically driven kymograph.moved beneath them, At regular
intervals a shot was cut in the paper so that one or both
(as thn case might be) of the electrodes made contact with
the metal surface of the drum which was charged with an electric
current. The upper of the two electrodes (on the vertical
axis) was wired to a double knife edge switch and carried
'the current from.the drum through the switch to either the
‘buzzer or the lights as the case might be. The lower of the
two electrodes was connected to the solenoid and made con-
tact with the drum.either concurrently with the stimulus _
electrode, 250, 1,250, 2,000 or h,000.msc. after contact was
.made by the upper electrode. The time intervals were care-
fully controlled by spacing the slots along the horizontal
axis of the graph paper. In the case of the simultaneous
contact with the drum, the upper and lower electrodes used
the same slot. The slot for the stimulus electrode was i
got an inch long, a linear distance equal to 250 msc. ac-
cording to the speed of revolution of the drum. Thus both
the light and the buzzer had a duration of 250 msc., an
amount of time considerably lower than the empirically de-
termined maximum.recruitment time as it is defined behav-
iorally.
II. Subjects.

The subjects for the experiment were 27 male albino

rats from.the new colony of the department of psychology

18

of Michigan State College. They ranged between the ages of
80 to 110 days. They were divided into five groups in the
followingvmanner:

Group I, a simultaneous group.contained six animals.
Group II, with a time interval of 250 msc. was composed of
five animals at the outset of the experiment although one
animal of this group died in the training process. Group
III, with an Sc-Su interval of 1,250 msc. contained six
animals. Group IV, with a time interval of 2,000 msc. con-
tained four animals. And Group V, with a time interval of
h,000.msc. contained six animals.*

Control in the selection of animals was accomplished
in the following manner. Any animals that in the course of
preliminary training failed to accomplish the task within a
range of from..6 seconds to 1.8 seconds (a range of plus and
minus one standard deviation around the mean time for ac-
complishing the task) were discarded.

III. Procedure.

Each subject was given twenty preliminary training
trials at the rate of four trials per day, with ten minutes
between individual trials. The apparatus used for the pre-
liminary training consisted Of the starting box of the T-
Maze placed in Juxtaposition with the positive and box. The

 

* It should be noted at this time however that to be con-
sistant with the preceeding formulation of the Sc-Su inter-
val, the 250 msc. group should be considered a simultaneous.
group since R occurred simultaneously with the offset of So.
All following groups should have their time values decreased
by 250 use. This will be done later in the discussion.

animals were placed in the starting box and the door Opened
so that the animal might run to the distal end of the goal
box where he received i gram Standard Purina Dog Chow as
reinforcement. The purpose of this preliminary training was
twofold: (1) TO condition the animal to respond immediately
to the opening of the exit door in the starting box, so that
in the actual problem the door would play a role equivalent
to that Of an unconditioned stimulus (Su). (2) To build up
secondary reinforcement in the goal box.

The major experimental groups were then divided into
two sub-groups, each of which contained half of the animals
in the original group. One Of the sub-groups was run in
the morning or early afternoon and the other was run in the
evening between the hours of seven and nine P. M; Further,
for one Of the sub-groups the goal box formed the right arm
of the T on trials in which the buzzer was used as the cue
stimulus, and the left arm of the T for the other. The ex-
perimental procedure was then the same for all animals with
the exception of the interpelation of varying degrees Of time
between the onset of the cue stimuli and the Opening of the
door of the starting box. The rats were required to respond
consistently to the right (or left depending on the sub-group)
when the buzzer went off concurrently with or slightly before

the Opening of the exit door Of the starting box. When the

G n
1'“ ‘a

light was used the response demanded was Opposite to that

demnded when a buzzer trial was given.

The rats were given

six trials a day, three with the buzzer and three with the

lieurt, following the order of presentation provided for by

the following table:

First week

2421.13.15.
§BLBBL
gLBBLB
Fourth gLLBBL
Week ELBLLB
gsBLLB
MBLLBL
'S‘I'TWT

Third Ieek

2: tr r' o: ts r! r4 Kn

ha fl:

pa lg

KO I"!

Second

‘Week

The animals received 1/3 gram.Purina dog food as rein-

forcement for correct responses.

When the animal made an

incorrect response and entered the negative and box, he was

retained there far 20 seconds, a period of time approximately

equal to the amount of time required by the animals to con-

sume 1/3 gram.of food. The criterion of response used was

the animal's (with his nose) pushing against one of the cur-

tains in the choice point.

This criterion was adopted be-

fore training because it was realized that after a certain

21

amount Of training, the animals would refuse to enter the
negative end box. In the case of an incorrect response, as
soon as the animal had nosed the wrong curtain, the vertical
sliding doors to the starting box and the positive and box
were closed, thus eliminating the use of a correction tech-
nique. In such a case, if the animal refused to enter the
negative and box within thirty seconds after leaving the
starting box, he was removed from the choice point and an
incorrect trial scored.

For the ten.minutes between trials, the animals were
kept in small wooden carrying cages. The purpose of this
was to remove any delayed effects of secondary reinforcement
that the home cage might have provided, and alternation due
to reactive inhibition.

The amount of time that animals remained in the start-
ing box before the presentation of either stimulus varied
randomly within a small range Of from.twO to ten seconds
depending for the most part upon how quickly the animal ori-
ented itself toward the exit door. At the end of the exper-
imental period, the animals were removed to their home cage
and fed a standard amount of the same food they had received
for reward in no greater quanity than 9 grams and no less
than 8 grams.“

The order in which the animals were run was varied con-
stantly so that no animal followed the same animal it had
followed on the previous trial. This was done to eliminate

the possibility Of tracking.

RESULTS

The conventional form of acquisition curve for all
five of the experimental groups is given in Fig.1, the.
data for these curves being in concise form in Table I.
‘It_will be Observed that, for the most part, all the curves
plotted are negatively accelerated and indicate a simple
positive growth function. However, the curves for all
groups except the simultaneous group show a positive accel-
eration at the beginning of learning and tend to be S-
Ihaped or ogival in appearance. It will also be noted
that the greater the time interval interpolated between
Sc and Su, the greater the slanting of the S-shaped curve.
This fact we may interpret as indicating the relatively
slow accumulation of habit strength in the groups with
the longer time intervals. This seems to be the .most
valid explanation because of the fact that fluctuations
in the strength Of a.habit in the beginning Of habit
strength generation cause the habit in question to be
obscured fram.Observation. (For the theoretical formula-
tion or behavioral oscillation see Hull,3, p. 289). _

.An examination Of Table I, reading down the vertical,
columns, clearly indicates the general trend to a lower
empirical probability of response evocation, within the
limits of an_equivalent number of trials, as the time
interval increases from zero to 4,000 msc. This obser-

vation is illustrated graphically in Fig. 2 derived

I,

90
0/.»

.mosc one no wdwasonw on» go awesome ma mane .macwnp no Henson came on» ad
soanopane one conceon >H can HHH museum use .HH and H masoao use» sofipcoaaqaa msoodonno one meanneo
pa nwsonpac .mawflno meansoomeoo .ooo .4m .om .ma nevus macaw some you upwuoaom punch no mecca ensue

haobapcaon s ma pH access mane aH .mqasaoo awesome» one used coon mapspauona umoa ma cane» chops any

ooa N.>m ¢.¢m o.mh A.Hc 5.0m b.mm o.mH m.ha N.ma o.ma o.HH N.oa m.m b mbomw

00H m.mm 4.mw 5.05 h.am H.¢m o.wH m.ba o.mH m.NH m.oa mb.o >H mDOmO

ooa m.b¢ o.mm w.mb «.mm w.mm o.mH m.nH 0.0H m.ma b.0d m.m HHH mbomc

00H m.Hm n.05 o.mo w.Nm o.wa w.oa m.ma m.HH m.m HH mbomw

00H m.hm 4.0m «.mh 5.0m o.mH m.ua «.md o.ma N.0H H mDome
mqumB mndeH M>Habofimzoo ma 2H

HPHBDommzoo mH 2H Roammoo.fi fib4mflh< mmmzommmm Bommmoo ho mmmfidz mo4mm>4 .H HAMQB

0/9 Caveat in I8 Boaseadtdue trials

.5? ,4.

’0' ‘ , .’. J
. ’g’

e. I e
I
/.- Ill /'
so {,3 ./
11-. / X
1.. /
470 ,4"! _ /. 6"”? ’ “3“"

 

'8 36 5'! 72. 90 108 :26
Number ofv TvLmLs ‘9

Figure 1, above, is plotted from values obtained in

Table I, and has for its Ordinate values the %correct
responses in a block of 18 consecutive trials.

2E5

from.values Obtained in the eleventh column of Table I.
The above method Of presentation Of the data, howb
ever, is not entirely consistent with a reinforcement the-
ory Of learning in which the number of reinforcements is
considered a more important determinate of learning than
the number Of trials. With this in mind, the data have
been re-arranged in Table II. There it may be observed
that the same general relationships hold true. It will
be recalled that in the section Of this paper dealing)
with the theoretical formulation of stimulus trace, it
was stated that the effects Of the time intervals are
most clearly Observed when the number of reinforcements
is the same, other factors such as size and delay of
reward being equal. For the purpose of demonstrating
this principal, the probabilities of response for the
five groups after forty-eight reinforcements were used
as the ordinate values in Fig. _3, andwere plotted agairet
the apprOpriate time intervals. _In Fig. 3 we clearly
see the effects Of increasing the Sc-Su interval upon
the amount of habit strength.attainable with.a given
number of reinforcements.. For the 4,000 msc..group, the
strength Of the habit provides a probability of response
of .764, or 26.4% above the level of chance expectancy.
Whereas, Group I has already attained the level of 10q%

performance.

26

Amanda case can nH ones on has pH
noaaeam wean: newnoapm passe no nofipfimwsuou on» no cassava confines when a manomonmon mans» chops one

oo.H cam.
oo.H
oo.H
oo.d

405.
Ham.
0mm.
0mm.

oo.a

moo.
msc.
mow.
0:5.
nmh.

05m.
000.
000.
msc.
emu.

maz Eamomogmm NH mug
mmahfi mfimZommmm ho MBHHHmdmomm madmflhfl

new.
mom.
mmm.
04m.
mum.

O.NH

H.ma
o.NH
o.NH
o.NH

u.ma
m.4a
m.nH
o.ma
o.NH

«.5H
m.wa
m.¢H
N.©H
m.mH

w.om
H.0N
m.oa
m.ua
0.0H

mBZMEHQKOMZHmm NH ZH¢6 OB
QHMHDGHM manmB ho mfimSDz Hade>4

.H canoe mu

.mudoaoonoudwon “chapsoomaoo pony whammoooam some women mwansonm

o.NN b
m.am
h.NN HHH mDomc

mDomc
bH Abomc

N.NN HH mboma

n.0N H mDomo

.HH mumda

 

27

35‘

86

Bo

 

95

       

0 to 4

‘rfinc-e

Fig.2 has as its ordinate values er
taken from.column 11 in Table 1., p centages

 

 

 

0' u 3.0 3.0 1:0

Fig. 3 is drawn from the average probabilities
of response provided in Table II, column 10.

28

Hull's formulation of habit strength generation under
.conditions of stimulus trace also asserts that the rate_
of generation will be reduced as the time interval becomes
longer, having its limit appreciably above zero. For the
purpose of showing the effect of the time intervals upon
the rates of learning, the curves in Fig. 4 were drawn.
These curves were based upon the cumulative data for each
group shown in Table III. From the values listed in the
table, it may be easily observed that the rate of increase
of any given curve constantly decreases in a.manner resembflng
a simple positive growth function, the increment to the
‘cumulative percentage constantly decreasing as a function
of the number of reinforcements attained.~ However, at any
given number of reinforcements the increases in the cumula-
tive percentages of the various groups is less as the ‘
thme’interval of the group is larger. For example, at
forty-eight reinforcements, Group I has a cumulative per-
centage increase of 2.7% over the cumulative percentage
recorded for thirty-six reinforcements. For Group I’howb
ever, the difference between the percentages reached at
the same number of reinforcements is only 1.9%.L It must
remembered however that these reinfOrcement values for
Group V occur relatively near the beginning of learning
where the rate of acquisition is nearly maximal, whereas
the.same numbers of reinforcements place Group I near

the end of learning where the rate is normally low.

29

TABLE III. THE RATIO or THE CUMULATIVE TOTAL or REINE‘OBCENEENTS
TO THE CUMULATIVE TOTAL or TRIALS (N/T)

N1 N2 N3 N a N 5 N6 N7 N 8

GROUP I .586 .658 .69h .753‘ .792 .820 .842 .860
GROUP II .535 .600 .614 .698 .7u2: .776 .800 .822

012002 III 2539 .561. .627 .678 .725 .759: .786 .813
0mm: xv .570 .583 .603 .658 .705 .7a0 .769 .793
GROUP v .545 .562 .600 .631. .675 .712 .741. .768

The above table gives the values of the ratio of & to T where N
increases consfiantly as learning progresses in “11?! following manner:
N1 equals 12, equals 12; N2 equals N1 plus 12, equals 21. and
Trepresents the average cumulative total number of trials.

3 These values are close to the N/T ratio for the various groups
at the time the criterion was reached. No values are asterisked
for IV and V because they fall more accurately between the cate-
gories used than within them.

 

W

I
106

 

W‘r-év—k

Reinhvcemewks -->

L I
‘18 60

I
36

 

Vl-

 

qfa‘aauo’) 0/0 “‘13”de O

50

Figure. 4 .

30

31

The curves in Fig. 4 have a particular value in so
far as they express clearly the differential effects of a
time interval upon the maximum.cumulative percentage at-
tainable at any given number of reinforcements. They
are of even more theoretical interest in so far as they
show a differential effect of the time interval upon the
cumulative percentages at the time the criterion was met,
regardless of the number of reinforcements. In this re-
spect, the following percentages were reached: Group I,
75.8%; Group II, 73.8%; Group III, 72.1%; Group IV, 71.%;
Group V, 71.2%. The reliability of these relatively small
~differences is attested to by the smoothness of the curves
when the data are handled in the above manner. These 1
differences are of some theoretical importance since the
maximum.habit strength attainable (M),-that is to say the
upper physiological limit is also theoretically reduced by
interpolating larger time intervals, and in the absence of
any direct measure of habit strength, these values may
have some significance. The curves are in no way to be
confused, however, with a true.measure of habit strength, for
to regard them as'such would demand postulating that habit
strength was a function of the ratio of reinforced to non-
reinforced trials, which the theory underlying this study

does not prOpose to be the case.

32

 

4.50.4 e\o max. Q1». €666.03.
«43>... gala-2 05955.5 om.

time in seconds

Figure 5-

so
03

Another measure of the rate of learning is seen in
.Fig. 5. In this method, an arbitrary value well above
the maximum number of trials required by the slowest
group is selected, and the average of the total number of
trials to reach.a given level of performance for each
group is then subtracted from.this value. In the present
case, the largest number of trials was required by Group
‘V, and the average of the group was found to be 111, so .
an arbitrary value of 150 was selected. The level of per-
formance was 75.3% as measured by the cumulative data.
This level was selected because it was the level of perf-oflr'iflll“cc
of the zero delay group at the time the criterion was
reached. This curve is not based entirely upon empirically
obtained data, although.the assumptions upon which the
extrapolations were.made are reasonably tenable. In this
method also Group I is seen to have the fastest rate of

rise, and the other four groups slower rates in accordance

with the length of the time interval.

' DISCUSSION

In the original introduction to the concept of stimu-
lus trace, it was stated that the magnitude of the stimu-
lus trace decreased as a simple decay function of the time
interval between the offset of So and the occurrence of R.
In the foregoing presentations of the results,the time,
vintervals were measured from.the onset of So. This method
ofpresentation does not obscure the relationship expressed
between groups II, III, IV, and‘V, since any re-arrangient
of the data for these groups would simply involve a sub-
h traction of a constant value of 250 msc. for the thme axis.
It may however present an untrue picture of the relation-
ship of Group I to t‘e other groups, and furthermore cre-
ate a false impression 6f a serious discrepancy between the
.results of this investigation and the previous studies on
the acquisition of trace conditioned responses. for instance,
in the studies of Wolfle only small differences were observed
between a simultaneous group, equivalent to this investigations
250 msc. group, and time intervals of about .8 seconds which
would correspond to the present 1,000 msc. group. The dif-
ference in character of the problem.involved.in.the present
form of simultaneous group is implied.by the fact that the
conventional acquisition curve for this group closely resem-
bled a simple positive growth function, whereas the acquisi-
‘don.curves for Groups II, III, IV, and‘V were alike in being

better approximated by S-shaped curves.

TABLE IV

'5 E
GROUP I 63.0 A6.3
GROUP II' 82.3 63.8
GROUP III 86.0 66.0
GROUP Iv 96.0 71.5
GROUP v 111 81.0

In the above table T'stands for the average total number of
trials required by each group to reach the iterion of 18
consecutive correct responses. The symbol stands for the
average total number of reinforcements required to reach the
same criterion of performance.

35

OWN“

36

GP

Ffauwe. 6

3

(h
0‘

 

\ Toto»). Nmbcv £ Reta?
~a
Ln

§

‘8
0

Figure. 7

(9
O

\1
0

 

'TBthL.hhumchc%VTV@nls

8

 

0 w 100 “so .
3:.5UL “chewed. w. secs.

Figs. 8' and '7 above were designed to
illustrate the marked d fference(in
terms of a general trend between the
sire group and the other four.

37

Similarly, the fact that the increase in rate of the
cumulative‘percentage acquisition curves was not con-
sistently greatly different for Groups II_and III, as
'was mentioned before, is probably attibutable to the
facts outlined above, and does not constitute a depar-
ture from.theoretical expectations. The anomalous re-
lationship Of Group I to the other groups, with respect
to stimmlus trace, is made quite clear in Figs. 6 and 7,
as well as in Table IV.

'. The serious departure from.conditioning data that
the present study provides is the fact that groups with
the time interval as long as 3,750 msc. reached a level _
of learning far higher than that anticipated on the basis
of the conditioning data. For the most part, this dif—_
ference is largely attributable to more adequate control
I of the relationships between drive and reinforcement..A
learning situation in which reinforcement is delivered
by the avoidance of a negative stimulus represents a
situation,of relative instability in so far as control
of these factors is concerned. The fact that in a
differential response situation with.unlimited reinforce-
ments, two or more responses are being learned in a
lunified manner is a possible factor contributing to these
differences, and as such.may not be completely ignored.

This study departs considerably from the results
obtained by'Wilson previously reviewed in this article.

‘

 

.. -*.___.._ _._—__ _......__.
. I

 

 

Wilson obtained evidence of only 20% learning in a 4,000
msc. group, whereas in the present study, all the animals
in the 4,000 msc. group reached the criterion of mastery.
The most plausible explanation for this difference is the
factor of inadequate Control over the time interval in the
WilsOn study. .As was pointed out in the section dealing
with theoretical background, the interpolation of long
runways is not comparable to the interpolation of strictly
controlled time intervals because of the shortening of the
time required to traverse the distance as learning progresseh
For this reason, Wilson's 60 inch alley probably has an

effect on learning much the same as that of a time interval

of 8 seconds or more. The mean number of trials for each

group in the Wilson study to reach.the criterion of mastery'

was 716, 792, and 874 for the 500, 1500, and 4000 msc.'groups
respectively. This indicates that the mean number of trials
for Wilson's longest group, was almost nine thmes as large
as_the mean number required by the present longest delay,
group. 'This discrepancy is attributable to the fact that

the factors of size of reward, delay of reward, and massing

of trials, were carefully controlled in the present ,tudy
at very nearly their Optimal values for learning.

Another factor contributing to the rapid speed of
learnifig by the animals in the present study was the con-
finement of he Operation of secondary reinforcement to only
one side of the maze for a given response. Secondary_rein-
forcement is acquired by the stimulus objects in the vicinity

of the goal, and this secondary reinforcement is capable of

0'.

Operating in a manner equivalent to primary reinforcement,
providing that periodic primary reinforcement is provided.
Thus, in the case of identical end b6Xes, the wrong re-
sponse to a given stimulus is reinforced almost as much-
as the correct response to the same stimulus with the re-
sult that considerably more primary reinforcements are
necessary for learning. The above eXplanation is supported
by findings in a study by Denny (l), in which it Was repor-
ted that animals learning a maze only reinforced in one
particular kind of a goal box-of markedly different appear-
ance from the negative end box learn‘d ronsiderably faster
than animals learning the same maze with.the exception that
both end boxes were identical thus eliminating differential
secondary reinforcement.

.A recent study, performed concurrently but indepen-
dently of this study, reports results more in agreement
with'Wilson than those of this paper. This study was per-
formed by Smith (6), at the University of Iowa, and he has
'kindly provided the author with a table summarising his ‘H
results in regard to number of trials. The data are found
here in fable‘v. In the Smith study, the different cue
stimuli consisted of running, on one trial, through a 24
inch.alley the interior of which.was painted white, and
on another trial, running through a 24 inch alley painted
black. Thus, the stimuli involved were both mediated by
the same sense organ. .After the animals had run through

40

TmBLE V
Sc - SU Interval 0 msc. 600 750‘ 2,000 5,000
Type of Rat Hooded Albino Hooded Albino Hooded Hooded Albino
70 150 250 900 580 £20 1500*
120 70 330 900 200 lth 1500*
150 100 190 610* 12h0* 1L80* 2100*
to 220 380 1050 1h00 2100*
to 70 120 520 1400- 2100*

70 290 180
30 160 140
90 160 270
70 160 1L0
90 290 280

Median 70 160 220 900 580 1400 2100*

The above table gives the number of trials required by each animal
to reach the criterion, and the median number for each experimental
group.

Asterisks indicate that animal had not reached the criterion and
running was discontinued at the end of the given number of trials.
(From.date supplied by M. Smith, University of Iowa, Iowa City,
19h8 obtained in an unpublished study.)

41

either one or the other of the stimulus alleys, which

. were presented in a random.order, they were confined in

a neutral delay chamber for intervals of time varying in

the following manner: a zero delay group, measured from

the time of offset of So, a .75 sec. group, a 2 sec. group,
and a 5 sec. group. ( These are the time intervals for the
.Albino subjects only, one of the Hooded groups being run.under

conditions of delay of .6 sec.) In this study, the zero

second group required.mere than twice the number of trials
required by the 250 msc. group of the present study which
was its equivalent. Further, it required one-and-aphalf
times as many trials as the 4,000 msc. group of the present
study. Smith found that only sq% of the animals in the

5 sec. group, and the e were all Hooded animals, could master
the problem. 0f the Albinos used in a similar 5 sec. group,
none, or Q%, reached his criterion of 18 correct out of 20
consecutive trials. This would indicate that the limit of
fall of habit strength is very low, but appreciably above
zero, when the Sc-Su interval is about five seconds. These
results are fairly well in line with the results obtained
from conditioning experiments. However, in the present
study, extrapolations of the curves based on average total
number of responses and average total number of reinforce-
ments, were made. The time interval which is to be expected
to require an impossible or extremely large number of trials

or reinforcements according to these extrapolations is

around ten seconds. These extrapolations were based on
theoretical expectations, since the four points obtained
from.the present study are represented almost as well oy . _
a straight line as by a positively accelerated growth curve.
The differences between the Smith data, and those of the
present study, are largely attributable to two factors;
one, the fact that secondary reinforcement was allowed to
‘ Operate as a reinforcer of incorrect as well as for correct
responses because of the identical construction of the end
boxes in the Smith report, a factor previously discussed
in relation to Wilson, and two, the fact that the present.
study employed two sense modalities for the mediation of the
cue stimuli whereas the Smith study used only the visual
sense. Thus, in the case where only one sense modality is
employed, there is a greater possibility of stimulus gen-
eralisation which would tend to impede the accumulation of
habit strength for differential responses since the So for
one response is on the generalisation gradient of the So
for the other response. A number of investigators have
shown that the generalisation gradient for responses
involving two sense modalities is remarkably steeper than
the gradients for responses evoked by stimulation of only
one sense modality. .

Another difference between the two studies is the de-

cided difference in the duration of the cue stimuli. It

43

will be recalled that stimulation as employed by
Smith involved running either a black or White alley
24 inches in length. This means that the duration of
the stimuli is at least 1,000 msc., or four times as
long as that used in this study, and considerably .
above the maximum recruitment period defined behavior-
ally. This does not, according to Hull‘s original
formulation, generate true trace reactions, and because
of this, direct comparisons of the present report with
Smith‘s study may involve some contradiction.
~Comparison of these results with those obtained by_
Warner does not seem to be particularly profitable since
warner's study has already been dismissed as being rep-
resentative of oyclicephase conditioning, rather than.
trace conditioning.' warner's study was rather thoroughly
discussed in the section of this paper dealing with the
theoretical background, and the reader is referred to

that discussion if further clarification is desired.

SUMMARY AND CONCLUSIONS
1. five groups of animals were trained in e reinforce-
ment situation at five different Sc-Su intervals of O,
250, 1250, 2000, and 4000 mac. respectively, primarily
for the purpose of investigating the relationship between
the level of learning and rate of learning and the Sc-Su
interval.
2. The results indicate that, for the range of intervals
used in the present study, the level of learning attainable
within a given number of reinforcements is simple decay
function of the magnitude of the stimulus trace at the
time of occurrence of R, assuming the magnitude of stimulus
trace bears a direct relationship to the amount of time
that has elapsed since the occurence of S.
3. The limit of fall of maximum learning is considerably
higher in s reinforcement situation in which the role of
secondary reinforcement is made Optimal then in a cone
ditioning situation.
h. In.differentiel response learning, the limit of fall
of the maximum.atteinable under a given number of rein,
forcements appears to raised for any given time interval
when two sense modalities are used.
5. When the Offset of So is contiguous with the occurence
of R, acquisition occurs at a slightly faster rate and to
a higher maximum than for intervals greater_than one second.
6. When the onset of a given So, which has a duration less

then the amount Of time required for maximum recruitment,

45

is made contiguous with the occurence of R, the maximum
level of learning is considerably higher and is attained
considerably faster than is possible for habits generated
under the conditions of trace imposed by this study.

7. A cue stimulus of 250 msc. having its onset contig-

uous with the occurence of R appears to bear an anomalous
relationship to the data of trace generated differential
response learning. That is to say, that habits generated
under this condition will probably be better expressed by
different functions than those formulated for trace generated
habits.

8. In general, the acquisition of a differential response
obeys the same general relationships provided for by Hull's
present mathematical formulation of stimulus trace, and

no concepts such as symbolic behavior, representative
factors, or association span need be devised for an adequate

account.

9.

10.

REFERENCES

Denny, M. R. The effect of using differential and boxes
in a simple T-Maze learning situation. J. exp. Psychol.,
191.8, 38, 255— 2.49.

Hilgard, E. R. Theories of learning. New York: Appleton-
Century-Crofts, Inc., l9h8.

Hull, 0. L. Principles of behavior. New York: D.
Appleton-Century, l9h0.

Kimble, G. A. Conditioning as a function of the time be-
tween conditioned and unconditioned stimuli. J. exp.
Psychol., 1947, 37, I*t5.

Reynolds, B. The acquisition of a trace conditioned re-
sponse as a function of the magnitude of the trace. J.
exp. Psychol., l9h5, 35,:5-30.

Smith, M. Unpublished doctoral thesis (Title 11111700395).
University of Iowa, Iowa City, l9h8.

Warner, L. The association span of the white rat. Ped.
sem. and J. genet. Psychol., 1932, 41. 57-90.

Wilson, M; O. Symbolic behavior in the white rat. J.
comp. Psychol., l93h, 18,'z9-49.

‘WOlfle, H. M. Time factors in conditioning finger with-

drawal. J. gen. Psychol., 1930, a, 372-378.

welfle, H. M. Conditioning as a function of the interval
between the conditioned and the orginal stimulus. J.
gen. Psychol., 1932, 7, 80-103.

APPENDIX

e.

GROUP I

Trials

Correct
1 2 3 a 5 6 7 8 9 10 11 12 13 11 15 16 17 18 Total
Animal#l.lOOllllllllOlOlllO 13
101111101111010110 13
010111111111111111 16
111111 6
Animal#2.lllOOOllOllOllOlOO 10
001101101100110111 11
110011111111111111 16
111111 6
Animal#3.101 00011110110101 10
110110111111111011 15
11111111111111118
Animal#4.lOlOOlllOOlllOllOl 11
110010010111111110 12
011011111111111111 16
111111 6
Animal#5.0000llOllOlOllOllO 9
100111111111111110 15
111011101110110001 12
111111111111111111 18
111111161546.010001100101011001 8
011011011011110110 12
011110011110101101 12
010101111111111111 15

111111

The above tabulations represent the behavior of the
animals in Group I recorded in rows of 18 consecutive trials
given at the rate of six trials per day. A.number "l" in
the above table represents a correct response, and the "0”
represents an incorrect response. The following pages con-
tain simililr tables for Groups II, III, IV and V. If the
reader wishes to know the order of presentation for the
sthuli, he is referred to the section in the report on
experimental technique, Part III, procedure.

GROUP II

Trials

Correct

1 2 3 A 5 6 7 8 9 10 11 12 13 lb 15 16 l7 18 Total

1111

1011

1001

0111

0101..

1101

0011
1011..
0011
1011
01111
00111
11011
01111
11101..
10111

Animal #2.

1111

1101

0111..

0111

1111

01.1.1

111.1

0011

0101
1011
0111
0101
1011
0111
0111
1111
0001
111...].

Animal #3.

10

10
1k
17

6

1111

1111

1111

0001

1111

0011

1111

1101

0111
1011
0011
001.].
01111
11111..
00011..
01011
10101
10111

Animal #h.

GROUP III

Trials
1 2 3 4 5 6 7 8 9 10 ll 12 13 1h 15 16 17 18 Total

Correct

10
1h
18

1011

0011

0111

0111

1001

1111

Animal #1.

0111

1011

1101

0101

1011

0111

1111
1101
1011
0111
0011
0111
1011
1111
1111
1111

Animal #3.

101111

001111

000011

101111

011011

100111..
001111
000101
010111
010111
001111
000111

10
12
17
12

1011

0011

0111

1001

0111..

1001

10111..

01111

11111
00111
11001
00011
01111
1111.].
11011
10011
10111...
00111

Animal #6.

GROUP IV

Trials
1 2 3 h 5 6 7 8 9 10 ll 12 13 14 15 16 17 18 Total

Correct

Animal #1.

80919’02
.I. 11.1

10011
11011
00111
01001
10111
01111
101111

000111..

101.101..
0101.11.
1.1.1111
010011
001111..
111011..
01011....
000111..
011111
11000.].

Animal #2.

11111

1001......

00111

1.11.1.1

01.101

101.11..

0011...].
11.111...
1001.1.
01.0.1.1
111.111..
01.1.1.1.
10111..
01101..
1.0111...
1.1.101.

11111111 #3.

01111

11111..

011111..

111111..

011111...
100011..
100101
011111.
101111..
011111..
0.1.1111
00011.1
110111..
111.011..

Animal #1..

GROUP V

Trials
1 2 3 4 5 6 7 8 9 10 ll 12 13 11 15 16 17 18 Total

Correct

9981438
111

010001..
111111
101111
000111..
111111
000111..
000111
111011

011111..
000101
111111
100101..
111001..
001111
100111..
111111...

mun 3mm
011111
100111..
111011..
111011
100011...
011111
011101
101111

001111
110111..
101011
000111..
010111
100111..
001011
111101

Animal #2.

0015678
1111111

110111
100111..
0.11111
19.1101
111011
011111
010111

111011

001111
001.111..
11.11.11.
11.01.11.
1.101011.
1.101101
00101.11
0101.111.
0011111.
1111111.

Animal #3 .

9 unun.
101111
010101
001011
101011
011111.
111111..
011101
101011

110011
010111
001111
111111
0001111
1100101
0111111
1001011
1111111
0001011

M

991:62
111.1

11111..
00011..
01111
11111..
10101..
10011
010111..

111111..

001101
000011
111011..
100111
010111
010011
111111..
001101
101011..
001111..

Animal #5.

013L502
111111.

0111......
11.101
000.11
0111...].
11011
00101
101111

001.111.

111.101.
11111.1.
1.01111.
01.0111.
1101.11
010011
1.01101
11.1111
011.111.
111.011..

Animal #6.

i“: 1 "E '1
«J5 ' ;

35%.; .5 .425.

 

' \-
O

L
U

1..

h

I

0::

3:-

_ 1'

 

 

 

 

 

 

I'ITUIYIl HI! BmITI'ES
6 9 8 5

J

S
R
F.
w
N
U
F.
T

GWHIISMII ill
3 1293 0306

Ill

 

MICHI

.5 I ’lldll'llb 1