Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
The Effects Of Prior Part-Experiences On Visual Form Perception In The Albino Rat
(USC Thesis Other)
The Effects Of Prior Part-Experiences On Visual Form Perception In The Albino Rat
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
THE EFFECTS OF PRIOR PART-EXPERIENCES ON
VISUAL FORM PERCEPTION IN THE ALBINO RAT
by
Frieda Bomston Libaw
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Psychology)
June 1961
UNIVERSITY O F SOUTHERN CAUFORNIA
GRADUATE SC H O O L
U N IV ER SITY PA RK
LOS A N G ELE S 7 . C A L IF O R N IA
This dissertation, written by
...... SrjUdA. .BQmatfco.Jvlbaw..........
under the direction of h.AV..Dissertation Com
mittee, and approved by all its members, has
been presented to and accepted by the Dean of
the Graduate School, in partial fulfillment of
requirements for the degree of
D O C T O R O F P H I L O S O P H Y
Duin
Date JunA, 1961
.TAT! DI
tairman
..
ACKNOWLEDGMENTS
It is impossible to acknowledge all the many in
direct sources of assistance that play so large a part in
the completion of any work. I should, however, like to
express my heartfelt gratitude to those who were most
directly helpful:
1. To my doctoral committee and to Dr. Neil
Warren, for supporting my application for a
Public Health Service Research Fellowship.
Without its financial assistance, this study
would have been far more difficult to execute.
2. To Dr. Philip Merrifield, who not only helped
immeasurably in the selection of appropriate
statistical tests of the hypotheses but who
also helped explain an "odd" result by tracking
down a few small errors in computation. He is,
however, in no way to be considered responsible
for any errors that may remain.
3. To Larry Flood and C. T. Thomas for their
meticulous design and construction of the
experimental apparatus to difficult specifica
tions.
4. To C. T. Thomas for his aesthetic and accurate
drawings of the experimental equipment.
3. To George Kevorkian for his construction of the
Normal Living cages.
6. To Saul Blackman for his photographs of the
visual stimuli.
7. To Marilyn Kurzweil for her friendship as well
as for her efficient secretarial services,
often at considerable inconvenience to herself.
ii
8. To Louis A. Gayle for invaluable assistance in
the care of the makeshift laboratory in my
garage and of its inhabitants, including the
experimenter, who often had reason to be grate
ful for his encouragement and dependability.
9. To mv husband Bill, not only for his photo
graphs of the equipment, but for his affection
and patience during some rather trying times
in tne conduct of the experiment and writing
up its results.
10. To my parents, Abram and Jennie, to whom I
should like to dedicate whatever may be of
value in this work. For it was they who did
much to instill in me the profound respect for
learning which has sustained me through the
many long and difficult years in which I have
doggedly pursued my academic career.
11. And, finally, to my daughter Marva, who shared
those years with me and grew into charming
young womanhood despite them.
iii
TABLE OF CONTENTS
ACKNOWLEDGMENT S
Page
ii
LIST OF TABLES vi
LIST OF ILLUSTRATIONS vii
Chapter
INTRODUCTION 1
II.
III.
IV.
V.
Hebb's Theory of Visual Form
Perception
The Test of Hebb's Theory of
Perceptual Development
Organization of the Remainder
of the Thesis
REVIEW OF THE LITERATURE.................. 13
METHOD..................................... 24
Experimental Design
Apparatus
Procedure
RESULTS ................................. 90
Descriptive Findings
Analytic Findings; Hypothesis-
testing Statistics
DISCUSSION AND INTERPRETATION OF THE
FINDINGS OF THE STUDY................... 119
Assessment of the Experiment
and Its Relationship to
Hebb's Hypothesis
Reappraisal of the Data and
Formulation of Alternative
Hypotheses
iv
Chapter Page
VI. SUMMARY, CONCLUSIONS, AND SUGGESTIONS
FOR FURTHER RESEARCH .................... 155
Summary
Conclusions
Suggestions for Further Research
LIST OF REFERENCES............................... 164
v
LIST OF TABLES
Table Page
1. Distribution of Sexes in the Treatment
Groups..................................... 92
2. Distribution of Acquisition Scores for
Males and Females........................... 94
3. Analysis of Variance in a Two-Way Classifica
tion without Replications: Litter vs.
Treatment Groups........................... 97
4. Distribution of Three Acquisition Scores for
Individual Treatment Groups ................ 99
5. Rank Order Correlations of Two Measures of
Central Value and Two Measures of Disper
sion in Three Acquisition Scores ........... 101
6. Distribution of Three Acquisition Scores
for Composite Treatment Groups ............ 103
7. Rank Order Correlations of Three Acquisition
Scores with Each Other.......................107
8. Rank Total of Three Acquisition Scores for
Individual Treatment Groups ................ Ill
9. Two-Way Analysis of Variance by Ranks for
Individual Treatment Groups ................ 113
10. Rank Totals of Three Acquisition Scores for
Composite Treatment Groups ................ 114
11. Two-Way Analysis of Variance by Ranks for
Composite Treatment Groups ................ 116
12. Values of T from Wilcoxin's Matched-Pairs
Signed-Ranks Test of Differences between
Treatment Groups ........................... 132
13. Backward Learning Scores for All Groups
Using Mean Daily Error per Block of
Ten Trials................................... 135
14. Choice of Stimulus in First Discrimination-
Learning T r i a l............................... 140
vi
LIST OF ILLUSTRATIONS
Plate Page
I. Experimental Living Cage, Breeding Cage,
Normal C a g e ............................. 37
II. Construction of the Experimental Living
Cage............ 41
III. The Line Stimulus......................... 47
IV. The Two-Lines-and-Included-Angle Stimulus . 49
V. The Discrimination Stimuli................. 51
VI. The Non-Pattemed Light Stimulus........... 53
VII. The Feeding Apparatus.......... 58
VIII. The Discrimination Apparatus ............... 62
IX. Construction of Discrimination Apparatus . . 64
X. Details of Stimulus Assembly ............... 67
vii
CHAPTER I
INTRODUCTION
An understanding of the nature of perception must
be regarded as one of the major tasks confronting a
science of behavior. Acknowledgement of the importance of
this task may be found in the fact that almost all of the
early "systems builders" in psychology felt it incumbent
upon them to present some explanation of perceptual expe
rience consistent with their general theoretical formula
tions (Boring, 1950). Although psychology has largely
outgrown its division into schools, it has by no means
settled all its controversies about the nature of percep
tion (Helson, 1951). New theories continue to arise
(Allport, 1955; Bartley, 1958; Solley and Murphy, 1960).
Few of these newer theories claim to systematize all the
data which have been classified under the rubric of per
ception. But even within each One's admittedly restricted
sphere there has been little systematic exploration.
The present investigation attempts to deal system
atically with the implications of one small but significant
aspect of a fairly recent theoretical contribution to the
psychology of perception (Hebb, 1949). It is an effort to
test empirically certain deductions made from Hebb's theory
of the development of the visual perception of geometrical
configurations.
Hebb's Theory of Visual Form Perception
One of the basic postulates of Hebb's theory is that
the ability to perceive organized visual patterns is the
"result of a long learning process" (Hebb, 1949). This is
not to say that he regards all organization of perception
as due to a learning process. He concedes that some per*
ceptual organization is innately determined and independent
of learning. But, from a reinterpretation of some of his
early experiments he concludes that, at least in rats, the
ability to discriminate brightness differences, to dis
tinguish figure from ground, to gauge depth, as well as
the ability to focus on an object and to follow its edges
or contours, are all part of the native equipment of the
organism. These abilities, however, only enable the
organism to perceive a primitive visual "unity" (Hebb,
1949). Unity of a visual figure is, in Hebb's system,
immediately perceived. It is simply the result of the
organism's ability to distinguish an amorphous whole as
separate and distinct from its background. There is, how
ever, another aspect of form perception which Hebb calls
the "identity" of a figure (Hebb, 1949). Identity, in
contrast to unity, is more than simple differentiation of
3
figure from ground. It Involves both generalization and
the perception of equivalence. It is not merely the thing-
character of an object in the field but its sameness and/or
difference from other objects; or, as Hebb defines it, "the
properties of association inherent in a perception— dis
tinct from its meaning" (Hebb, 1949). While the perception
of unity is regarded as the result of native endowment, it
is the perception of identity which is thought to involve
learning.
A second postulate of Hebb's system is that percep
tual learning develops through the gradual integration of
parts into wholes by an additive process (Hebb, 1949).
Hebb explains the summation of parts into perceptual wholes
by constructing a theoretical model of the central nervous
system. This model postulates the existence or development
of structures which are correlative to perceptual func
tions. The first of these theoretical constructs is the
"cell-assembly." This is Hebb's neurological correlate of
a "simple" perception. His definition of a cell-assembly
is:
. . . a diffuse structure comprising cells in the
cortex and diencephalon (and also, perhaps, in the
basal ganglia of the cerebrum), capable of acting
briefly as a closed system, and usually having a
specific motor facilitation. (Hebb, 1949)
The development of this diffuse neurological structure is
thought to occur in the following manner:
When an axon of cell A Is near enough to excite
cell fr an? repeatedly or"*persistently takes part
in firing it. some growth process or metabolic
cEanae takes place in one or both cells such that
A* self iHencras"one oTTtEe ceITa“TIrrngT“ii---
IncreaaeqT (HeEF. "TOST1-------------------
A visual cell-assembly, then, might be schematized, as
being composed of particular retinal cells, cells in the
visual cortex, the association areas, and the tracts lead
ing to them as well as cells stimulated by the ocular mus
culature. All of these, through frequent repeated stimu
lation, have a mutually facilitative effect in their firing.
The perception of a "complex" form or, in Hebb's
terms, a "superordinate perception," such as a triangle or
square, requires the integration of several cell-assemblles
into a phase sequence. His schematic of the integration of
the several parts of a figure into a distinctive whole is:
Activity in an assembly a, aroused by fixation
on an angle A of a triangle can occur independ
ently of b or c. When A, B, or C are looked at
successively in any order but in a short period of
time, activity may continue by reverberation in
two of the structures while the third is sensorily
aroused. In these circumstances, conceivably,
there is a frequent activity in the three assem
blies a, b, and c at the same time. These lie
Interlaced with each other in what is grossly the
same tissue of the cerebrum . . . the simultaneous
activity would result in an integration of the
three systems. (Hebb, 1949)
If these several activities coexist and can be aroused in
any order:
^Hebb's italics.
It is a reasonable inference that two of these
determinate actions simultaneously would have a
determinate effect, tending to excite specific
transmission units, and that these units would
tend to organize in the same way that the earlier-
established units were organized. Activity in a
superordinate structure (in this case, t^) ls~~then
best defined as being whatever determinate,
organized activity results from repeated activity
in the earlier-developed structures giving rise
to ItT ( B ET f i W j y------------- ------ ----
From this there develops a temporal sequence of arousal
which also involves motor activity because, in the develop
ment of the cell-assembly underlying a visual perception,
there are always changes in eye fixation. At least one
motor component must become liminal before the next assem
bly is aroused because the eye must move to fixate on the
next stimulus which, in the case of a triangle, would be
one of its angles. When the transmission unit, or assembly
jt, becomes organized, its activity intervenes between the
subordinate assemblies but it does not supersede them. The
sequence, in terms of assemblies, becomes something like
this: a-b-t-c-a-t-c-t-b-a-c-t-a, etc. Behaviorally this
corresponds to an alternation of perceiving first one angle
then another, then the total triangle and then another
angle, etc. Hebb's position may now be summed up in his
own words:
. . . the integration of t, the basis of perceiving
a distinctive total figure, essentially involves a
sequence of cortical events with motor components.
. . . In the early stages of perceptual development,
2Hebb's italics.
t_ might be excited only after repeated activations
of a, b, and c, but later might be aroused follow
ing sensory activity of a alone, so that the
triangle would be recognTzed with a single glance.
(Hebb, 1949)
It Is this view of neural activity which prompts
Hebb to state:
A triangle then Is a complex unity In perception,
not primitive. As a whole, it becomes distinctive
and recognizable only after a prolonged learning
period. . . . Similar triangles of unequal size, a
solid and an outline triangle, even a solid plane
triangle and one circumscribed with a circle con
tain a number of identical elements— lines and
angles with the same orientation. . . . Lines and
angles. then, can be treated as perceptual elements,
not fully innate in perception, but partly so. and '
likely to be learned before more complex patterns
are'. (H5bE7 1949)3--------------------- -------
The Test of Hebb's Theory of
Perceptual Development
Basic Assumptions and Propositions
Underlying the Hypothesis
Hebb's neurological model-building is essentially
an attempt to explain how perceptual elements become inte
grated into a single, determinate, perceptual pattern.
The "correctness" of his physiology, however, is far from
crucial for any empirical study of his perceptual theory.
The behavioral implications of this theory can be tested
in the laboratory without reference to any model of physio
logical functioning. In point of fact, it would be
patently impossible in our present stage of knowledge to
3Italics mine, FBL.
7
attempt a direct validation of Hebb's neurological
theories. They are scientifically good theories primarily
because at this time their behavioral implications are
available for empirical study. It should be emphasized,
therefore, that in this sense, any physiological specula
tions are only indirectly relevant to the present investi
gation, for it is Hebb'8 behavioral theories, irrespective
of their neurological substrate, which have been translated
into operational terms and empirically tested.
Assuming a strictly behavioral stance, if we accept
Hebb's proposition that the perception of a whole is some
thing which must be learned and if it is learned only after
its component elements have been learned, then it follows
that the amount of learning required to perceive it should
vary Inversely with the number of component elements pre
viously experienced. If we take a triangle as an example
of a perceptual whole and if we accept Hebb's definition of
lines and angles with the same orientation as the tri
angle's as being identical, component, perceptual elements
of that triangle, then it is possible to define a variable
which proceeds from a region of no identical elements to a
region which includes all of the Identical elements and
whose Intervening steps follow in ordinal, "add-a-part"
sequence. The fewer the lines and angles with which an
animal had prior commerce, the more learning it would still
have to do in order to perceive the total triangle; the
8
more identical elements it had previously encountered, the
less learning would remain to be done.
This statement of inverse relationship constituted
the basic hypothesis of the present investigation. The
experimental hypotheses and, consequently, the experimental
design were directly derived from it.
Summary of the Experimental Design
In order to test this major hypothesis it was neces
sary to devise some means of independently manipulating the
several aspects of the postulated variable and determine
their differential effect on ability to perceive a total
visual pattern. To this end a rough scale of part-experi
ences was constructed.
Because Hebb uses it as his illustrative example
(Hebb, 1949), an equilateral triangle was selected as the
prototype of a complex whole in perception. The perceptual
elements of which it is composed were broken down into five
divisions:
I* Darkness. which was considered equivalent to
sub-zero number of elements because it elimi
nated not only experience with visual pattern
but prevented all visual experience.
2. Non-patteraed light. which was considered
equivalent to zero number of elements because,
although it permitted visual stimulation with
light, It provided no experience with patterned
visual fields.
3. Straight line, which was considered equivalent
to lowest number of perceptual elements because
since It was of the same length and orientation
It could be regarded as one side of an equilat
eral triangle and because It provided a simple
visual pattern by sharply dividing the percep
tual field into figure and ground.
4. Two-lines-and-included-angle, which was consid
ered equivalent to intermediate number of
perceptual elements because it simulated two
sides and one angle of an equilateral triangle
and added complexity to the patterning of the
visual field by providing a nodal fixation
point from which the eye could move in two
directions (Hebb, 1949).
5. Equilateral triangle, which was considered
equivalent to the total number of perceptual
elements by definition.
6. Normal visual environment. which was equivalent
to an indeterminately greater number of percep
tual elements than those needed for the percep
tion of an equilateral triangle.
Such a scale appeared to meet the requirements of
10
ordinal measurement (Stevens, 1951). Each step Incorpo
rated the elements of the step which preceded It but added
new elements which, from the standpoint of Hebb's theory,
would be regarded as determinants of learning to perceive
an equilateral triangle. There is at least face validity
for the assertion that it ranged the relevant part-experi
ences along a dimension of lesser to greater amounts of
the perceptual elements which allegedly precede the ability
to perceive this particular complex visual pattern.
The manipulation of the independent variable
called for exposing groups of animals to a controlled
visual environment limited to a single one of the types of
visual experience represented by each of these scale
divisions. Experimental animals, therefore, were divided
into six groups, each of which experienced, from birth to
adulthood, a visual environment entirely limited to the
region of visual experiences defined by one division of
this ordinal scale. One group of animals thus experienced
no visual stimulation at all since they were raised com
pletely in the dark. A second group received stimulation
only from a diffuse, non-patterned light. A third group
had all of its visual experience limited to a single,
straight line. A fourth group could view only two lines
and an included angle of 60 degrees. A fifth group had
its visual experience limited to a view of an equilateral
triangle, while the sixth and last group was able to
11
obtain a relatively unobstructed view of the laboratory
environment through the sides and top of an open-mesh cage.
It was assumed that differences in ability to per
ceive an equilateral triangle which resulted from this
differential prior experience would be evidenced in differ
ences in ability to learn to distinguish the triangle from
a circle. Discrimination learning, therefore, was accepted
as a suitable dependent variable for the study.
Organization of the Remainder of the Thesis
This introductory chapter has presented an inter
pretation of Hebb's theory of the development of the
ability to perceive a whole visual form and has outlined
the manner in which the theory could be empirically tested.
The remainder of the thesis is arranged so that this
introductory section is followed by an evaluative review of
the relevant literature. This review is succeeded by a
three-part account of the method used in carrying out the
proposed empirical test of the theory. The first part
consists of a detailed presentation of the experimental
design. The second and third parts contain, respectively,
the descriptions of the apparatus and the procedures used
in translating the design into the realities of the labora
tory. The findings of the study are then stated. A
discussion of the conclusions that may be drawn from them
12
and their theoretical implications follows. The thesis is
terminated with a summary of the entire study and some
suggestions for further research.
CHAPTER II
REVIEW OF THE LITERATURE
Although there now exists a considerable body of
published research which acknowledges its source in Hebbian
theory, none of it has apparently been directed toward
isolating the effects of prior part-experiences on form
perception, even though it would seem that such a test
would be crucial to Hebb's position on perceptual develop
ment. Most of the published work seems to have concen
trated on two much broader aspects of the effects of in
fantile experience on adult behavior, the effects of
sensory deprivation and the effects of early enrichment of
the perceptual environment.
While those workers who have conducted both types of
experiments feel that their studies derive from Hebb's
theory, it might be well to recognize at the outset that
there is a logical difference between them. Deprivation
of sensory experience would, if the present interpretation
of Hebb is correct, presumably prevent the formation of
cell-assemblies and phase sequences which are the basis for
certain types of perceptual behavior. An organism reared
in the dark and deprived of all visual experiences simply
13
14
would noC have had the opportunity to form the neural con
nections that ostensibly can only result from repeated
visual stimulation. From Hebb's point of view It should
be impossible for such an animal to have formed cell-
assemblles in the visual area at all. On this basis it
is a simple and logical deduction from the theory that an
organism so deprived should require more trials to learn
to solve a form-discrimination problem, for example, than
one who has been reared in a normal, illuminated environ
ment and has thereby acquired the necessary cell-assemblies
and phase sequences to make perception possible. The two
organisms would, so to speak, be entering a race from two
different starting lines. The light-deprived animal having
much farther to run, it is understandable that it should
take longer to reach the goal line.
The logic of the enrichment studies, however, is not
so clear-cut and its derivation from Hebb's hypotheses is
certainly not as direct. For while Hebb's claim can be
interpreted to mean that a given type and amount of sensory
stimulation is both necessary and sufficient for the forma
tion of the particular cell-assemblies and phase sequences
that enter into the perception of a particular type of
visual form, and it would follow that deprivation of this
kind of sensory experience would hinder perception, it does
not necessarily follow that an animal given a wider and
more varied experience should be superior in this same,
15
particular performance. Hebb is, perhaps, not too specific
in stating the necessary conditions for the perception of a
triangle, but he does indicate that the opportunity to view
a triangle over a period of time is sufficient for the
formation of the cell-assemblies and phase sequences which
make its perception as an identity possible. There is,
therefore, little reason to believe that an animal which
has received additional experiences, perhaps with squares,
circles, striations, or any other visual perceptual forms,
should be better able to perceive a triangle than an animal
whose perceptual training has been limited to the viewing
of triangles alone. On the contrary, Hebb is quite spe
cific in attributing the experimental facts of pattern
equivalence to identical elements and, to paraphrase
Gertrude Stein, a triangle is a triangle is a triangle.
Once you can perceive it you can perceive it. Being also
able to perceive a circle may give additional facility to
the animal but it could hardly help him to see a triangle
better.
The fact that many studies which have provided en
riched environments (Forgays and Forgays, 1952; Forgus,
1955; Gibson and Walk, 1956; Gibson, Walk, Pick and Tighe,
1958; Griffiths and Stringer, 1952; Hebb, 1947; Hymovitch,
1952) do show that animals so reared are superior in the
performance of certain tasks over their normal cage con
trols, requires either an additional postulate in Hebb's
16
theory or, what is more likely, a more careful analysis of
the requirements of the task so that a more carefully con
trolled, more directly relevant type of experience can be
provided and its effect ascertained.
In any case, these studies are not directly perti
nent to the present investigation since none of them, by
their very attempts to provide "enriched" environments,
have been concerned with defining the part-experiences
which are either necessary and/or sufficient to the percep
tion of visual form.
The findings of the deprivation studies, however,
are somewhat more relevant to the present problem. Studies
of this type include the early study by Hebb (1937) where
some of the data obtained were later compared to data de
rived from a group of rats trained by Lashley which indi
cated that dark-reared rats required six times the number
of trials required by normal cage-reared animals to learn
a form-discrimination on the Lashley jumping stand (Hebb,
1949).
Forgus (1934) reared three groups of rats under
varying conditions of both visual and proprioceptive
stimulation. One group was reared in a large black plywood
box and allowed to view, and to manipulate many white ob
jects such as blocks, alleys, tunnels, elevated platforms
and inclined planes. A second group was reared in the
same type of box but was allowed to live only in the center
of it in a glass cage through which it could view the
same objects. The visual stimulation for the first two
groups, then, was comparable, but the proprioceptive cues
provided by an opportunity to handle the objects was
different. The second group had a minimum of propriocep
tive experience. The third group was subjected to an
environment in which both visual and proprioceptive expe
rience was kept at a minimum. This last group was decid
edly inferior in learning an elevated maze, in learning to
make a visual discrimination and in a test of spatial
problem-solving.
Hymovitch (1952) found that rats which had been
reared in stovepipe cages and in closed activity wheels,
both providing a minimum of visual experience, were infe
rior in their performance on a closed field test to those
animals which had been reared in a free environment or in
mesh cages where the visual experience was, presumably,
normal.
Riesen (1947), working with primates, has demon
strated that chimps reared in the dark for the first 16
months of life showed a complete absence of obstacle avoid
ance on the basis of visual cues. In another of his
experiments (1950), he reared one chimp in complete dark
ness, another in diffused light, and a third in a normally
illuminated room. The blink reaction took 15 days to
appear in the first, 6 days in the second. Steady fixation
18
took 30 days for the first and 13 days for the second.
Trials to learn a conspicuous visual stimulus associated
with shock took 2 trials in one day for the normal animal,
13 day8 of 2 trials per day for the diffused light animal
and 13 days of 2 trials per day for the dark-reared animal.
It should be noted that ophthalmoscopic examination re
vealed evidence of retinal degeneration in the totally
deprived animal but no ocular abnormality in the other two.
Despite this ocular disturbance, however, the dark-reared
animal required only 4 additional trials compared to the
animal reared in diffused light. These two animals were
far more alike in their learning trials than either was to
the light-reared one. This finding is completely in line
with predictions made from Hebb's theory and with the
hypothesis of the present study, though it should be noted
that the higher we go on the phylogenetic scale the more
difficulty an animal apparently encounters in overcoming
the effects of infantile deprivation. The discrepancies
in variously-reared rats might not be as great as that
observed in the primates, including man.(Riesen, 1950).1
^These conclusions corroborate Hebb's theoretical
position with respect to phylogenetic differences in per
ceptual learning. He states tnat higher forms require
more time for early learning but are able to learn more
complex relationships later on. The lower forms, such as
the rat, although they are never able to master the com
plexities possible to the primates, go through essentially
the same learning process and, "Given a really new and
unfamiliar set of sensations to be associated with motor
responses, selectively, the first definite and clearcut
19
Beach and Jaynes (1954), who reviewed many studies
on the effects of early deprivation in a variety of
species, conclude that:
The evidence is fragmentary and at points incon
clusive, but it does suggest that in at least some
species of fishes, birds, lower mammals, and pri
mates, absence of the normal amount of visual
stimulation during the developmental period may
result in inability to respond adaptively to visual
cues when such cues first become available to the
individual. The defects do not appear to be
permanent but seem to disappear with increasing
experience in visually directed responses.
(Beach and Jaynes, 1954)
For the most part, those studies which investigate
the effects of visual deprivation indicate that some sort
of prior visual experience is required for an animal to be
able to utilize visual cues.
Only one study encountered in the literature failed
to show a significant difference in the expected direction
between rats reared in the dark and those reared in a
normal cage environment when they were later tested on a
visual form-di8crimination task (Gibson, Walk and Tighe,
1959). This study, however, suffered from a grievous
methodological flaw. All of the rats in the study, both
those reared in the dark and those reared in the light,
were subjected to an open-field test of emotionality and a
association appears sooner in rat than man" (D. 0. Hebb,
The Organization of Behavior, John Wiley and Sons, New
York, 1949). The rat, in other words, may never be able
to form as highly complex or well-integrated cell-assem-
blies and phase sequences as are possible for man but what
he can form, he will form sooner.
20
test of depth discrimination prior to their training on the
form-discrlmination problem. Both these tests were con
ducted in an illuminated environment. Although they
required very little time, their effects on the subsequent
performance of the animals cannot be determined. Even more
troublesome, however, is the prolonged exposure to pat
terned light which accompanied the training on the form-
discrimination problem itself. All of the animals were
kept in the light during the learning trials. The dark-
reared animals were removed to normal mesh cages for the
entire time that discrimination training was carried on.
Since this training proceeded at 10 trials per day and the
dark-reared group averaged 170.0 trials to learn the tri
angle versus circle discrimination, the animals were in the
normal cage environment, in the light, for approximately 17
days. It is quite possible, and, indeed, highly probable,
that there was sufficient visual stimulation during this
time to obscure whatever differences might have initially
existed between these dark-reared animals and the normal
cage controls.
One recent study, however, used the technique of
rearing animals in the dark and used a mode of testing
similar to the one which was used in the present inquiry.
Michels, Bevan and Strasel (1958) reared their animals in
darkness, not to test the effects of deprivation per se.
but as a means of controlling visual experience prior to
21
the presentation of several discrimination problems. Their
findings also have some relevance for the present study.
They used male rats and divided them, at birth, into four
groups. Two of these groups were reared in the dark.
Their only visual experience occurred during the discrim
ination test periods in an apparatus that was kept dark
except for the illuminated stimulus panel. The other two
groups, tested in the same apparatus, were kept under
normal laboratory conditions. They experienced a period
of familiarization with the test apparatus, which was kept
entirely dark during this time. After the period of
familiarization, one of the light-reared groups and one of
the dark-reared groups were presented three problems re
quiring a discrimination of magnitude of brightness, of
size, and of height. The presentation of these discrim
ination problems was done in a latin-square sequence and
was followed by a form-discrimination problem. The form-
discrimination problem involved the differentiation of a
heart-shaped form presented in normal orientation and a
rectangle topped by an equilateral triangle.
The light-reared group showed uniform proficiency
on all four tasks. They were also significantly superior
to the dark-reared animals. But one of the most interest
ing results of the experiment was the effect of ordinal
position in the presentation of the stimuli. This was
significant only for the dark-reared animals, not for those
22
reared in the light. The deprived group showed progessive
improvement in the series even though their performance on
the form problem was still significantly poorer than that
of the light-reared group.
Since the stimuli for the size problem and the
height problem (but not the brightness problem) involved
the presentation of complete forms (triangles in the first
instance and rectangles in the second), the animals did
have an opportunity to view patterns. Presumably, from
Hebb's point of view, the process of cell-assembly and
phase sequence formation could have begun through an expo
sure to these stimuli. Since lines and angles occur both
in triangles and rectangles, it is conceivable that the
cell-assemblies formed through experience with one set of
perceptual elements would have a facilitative effect on the
learning of the other, regardless of which was presented
first. This might account for the progressive improvement
noted. The fact that the dark-reared group was still
inferior to the light-reared group on the form-discrimina-
tion problem might be interpreted as due to differences in
the visual experience obtained exclusively during the
training procedures contrasted to that obtained by the
normal cage group prior to as well as during the training
period. In themselves, the training procedures may not
have provided sufficient visual experience for the
formation of phase sequences required for almost immediate
23
perception of the discriminations of the forms' identity.
In addition, it is well to note that the triangles and
rectangles used in the experiment which were part of the
animals' prior visual experience are, in Hebbian terms,
themselves complex forms. The experiment, therefore,
while in some ways offering suggestive, supportive evidence
for Hebbian theory, does not deal directly with the effects
of prior perceptual elements on consequent ability to per
ceive total, complex forms.
No study, in fact, could be found in the literature
which attempted to ascertain these effects. Yet a direct
test of Hebb's theory of the development of the perception
of identity of visual forms would seem to require their
empirical demonstration. It was in an effort to provide a
setting in which such an empirical demonstration could
occur that the present study was devised.
CHAPTER III
METHOD
Experimental Design
The basic hypothesis of the present study was that
an inverse relationship would be obtained between the
amount of learning required to perceive a total visual
form and the number of component elements of that form
which had been previously experienced.
The Independent Variable
By implication, the hypothesis required a manipula
tion of prior experience in such a manner as to vary the
number of those perceptual elements relevant to a given
total visual form. To serve this end, an ordinal scale of
perceptual elements was constructed in which the scale
regions were deemed to be relevant part-experiences enter
ing into the subsequent ability to learn to perceive a
total equilateral triangle of specified dimensions.^
*Cf. "Provision for Visual Stimulation," pp. 43-54,
for exact descriptions of the stimulus patterns used in the
laboratory. The specifications given there are, in effect,
the operational definitions of both the "total visual form"
and the "perceptual elements" and/or "visual part-experi
ences" used in this study.
24
25
The six regions of the scale have already been out
lined in the introductory chapter. These six regions,
ranging from least to greatest number of perceptual ele
ments, were: dark, non-patterned light, line, two-lines-
and-included-angle, triangle, and normal cage environment.2
Ideally, the manipulation of the independent vari
able would call for each subject to experience all of the
stimulus conditions implied by the scale divisions. If
the subject could gradually have his part-experiences in
creased and his perceptual ability measured at the same
time, he could serve as a gauge and we could watch the
required learning decrease as the perceptual elements were
increased! Such a neat manipulation of variables, alas,
was not possible since animals, even lowly white rats,
steadfastly refuse to act like Inanimate gauges. They
^Strictly speaking, both the dark and normal cage
conditions would not have been necessary to the logic of
the design of an experiment devised to test the basic
hypothesis of the present investigation. Neither the com
plete absence of visual stimulation nor a surfeit of non-
specifiable visual stimuli could serve to elucidate the
relationship between amount of prior part-experlence and
later ability to learn to perceive form. The four other
scale divisions would have served this purpose. However,
since most of the experimental literature related to Hebb*-s
perceptual theories contrasts the perceptual abilities of
dark-reared and normally-reared animals, the inclusion of
these two scale regions served, in effect, as a replication
of those studies under the modified stimulus conditions of
the present experiment. In addition, their inclusion made
it possible to gather information which would permit the
formulation of a wider range of alternate hypotheses for
further investigation if the present hypothesis derived
from Hebb's theory should not find support in the data.
26
persist in being modified by their experiences. In order
to arrive at a reasonable facsimile of the ideal manipula
tion of the independent variable, therefore, it was neces
sary to take groups of animals and submit each group to a
single one of the stimulus conditions implied by the scale.
Because genetic factors over which we have no direct
control might have a bearing on an animal's sensitivity to
visual stimulation and/or later learning ability, it was
also considered necessary to devise some means of indi
rectly controlling or eliminating the effects of genetic
differences. Ideally, this problem might have been solved
in one of two ways. If it were possible to obtain a random
sample of sufficient size, we might be justified in assum
ing that genetic effects were randomly distributed and,
therefore, without any biasing effects on the results. Or,
we could have a population composed entirely of identical
"twins." Again, neither of these approaches was possible
and some alternative practical solution to the problem had
to be found. The split-litter technique was adopted.
This resulted in a Treatments x Litters design in
which each of the six animals from a single litter was
randomly assigned to one of the stimulus conditions. There
was a total of nine such litters distributed across the six
treatment conditions. Since each of the treatment groups
represented, in effect, one of the scale divisions, com
parisons of the groups made it possible to ascertain the
27
relative effects of the part*experiences they represented
and thus constituted an acceptable means of experimental
manipulation of the independent variable.
The Dependent Variable
Measurement of the dependent variable implied by
the hypothesis again required a pragmatic solution which
fell short of the ideal. No direct measure of an animal's
ability to perceive a total form is as yet available. The
substitute measures adopted by the present investigator
have been widely used by others as an indirect means of
measuring perception. They were all measures of an
animal's ability to learn to solve a discrimination problem
by responding "positively" to a triangle and "negatively"
to a circle of equal brightness. Three measures were
adopted: (1) trials required to learn to discriminate
between the two visual forms correctly in eighteen in
stances out of twenty consecutive attempts; (2) the total
number of errors made in reaching the same criterion; and
(3) the total time of exposure to the two simultaneously
presented stimuli before the same criterion was reached.
This resulted in three acquisition scores for each subject:
(1) number of trials to criterion; (2) total number of
errors; and (3) total exposure time.
Cf. pp. 83 ff. for detailed description of proce
dures used in obtaining these measurements.
28
Using a measure of discrimination learning as a
means of determining an animal's ability to perceive a
visual pattern has particular justification in this study
beyond the mere fact of the precedent set by other inves
tigators. In the earlier discussion of Hebb's theory it
was pointed out that he makes a distinction between the
perception of primitive visual unity, which he feels is
unlearned, and the perception of the identity of a visual
figure which requires learning.^ In testing an hypothesis
based on his theory of perceptual learning, therefore, it
would seem that attention should be focused on the percep
tion of visual identity. Since the identity of a perceived
object involves a recognition of its sameness and/or dif
ference from other objects in the visual field, it would
seem that requiring an animal to respond differently to
two different patterns presented simultaneously (the usual
situation in a discrimination problem and the one adopted
in the present study) would be one way to assure that the
animal was Indeed being required to perceive their
identity.
Basic Assumptions and Presuppositions
Underlying the Experimental Design
It should be noted that when pattern discrimination
is studied by the method used in the present study, the
^Cf. Chapter I, pp. 2 ff.
29
animal is asked, in effect, to do at least three things.
It is asked: (1) to perceive the differences between two
visual forms; (2) to learn to associate reward with one of
these forms and non-reward with the other; and (3) to learn
a "correct" motor response to the stimulus which, in this
instance, was to push open the door on which the "positive"
stimulus was mounted. These three tasks were confounded
and it is difficult to determine whether the ability of the
animals to perceive differences (the important aspect of
the problem and the one which was presumably being measured
as the dependent variable of the study) was masked by the
necessity of learning one or both of the other two tasks
and that, perhaps, the difference between the triangle and
circle was immediately perceived. There is no way of
determining if some unknown aspect of their prior experi
ence created differences in the subjects which either made
association with reward or mastery of the motor task dif
ferentially difficult. The use of discrimination measures
as ordinarily used always makes the implicit assumption
that either no differences in ability to learn certain
irrelevant tasks are to be found among the subjects or that
they are randomly distributed so that they do not introduce
any systematic distortion of the measurement of the depend
ent variable. Some such assumption must be made if the
measured differences are to be justifiably attributed to
differences in perceptual ability and, in turn, to the
30
differential effects of the manipulated variable.
Another assumption which was made in designing the
experiment was that the method of manipulating the inde
pendent variable was Indeed a valid means of administering
the kind of part-experiences Hebb considers relevant to
perceptual learning.
A third assumption, obvious but important for the
interpretation of the data, was that the split-litter tech
nique did actually equate all the groups for any constitu
tional factors which might influence the results. If any
differences did exist, it was simply assumed that these
were "differences that made no difference."
The Experimental Hypotheses
Once these assumptions are made and the validity of
the measures of both the independent and dependent variable
are granted, it becomes possible to rephrase the general
hypothesis of the study into the experimental hypotheses
which were to be put to test in the laboratory. These
hypotheses were:
1. Significant differences would be obtained in the
rank totals of the six groups on the measures of
discrimination learning. Stated in the form of
a null hypothesis, the six different stimulus
conditions would have no differential effect on
the rank totals obtained by the groups on a
measure of discrimination learning. The sta
tistical test would be by means of a Friedman
nonparametric two-way analysis of variance by
ranks and the null hypothesis would be rejected
if the result obtained from this analysis were
to have a probability associated with its
occurrence equal to or less than o< - .05.
If significant differences in rank totals
occurred, they would be such that further
analysis by intergroup comparisons would reveal
that the groups would be ranked on the basis of
low to high scores on the measures of discrim
ination learning in an inverse order to the
amount of their prior visual experience. That
is, from lowest scores (best discrimination
learning or least amount of additional learning
necessary to make the discrimination) to highest
scores (poorest discrimination learning or
greatest amount of additional learning neces
sary), the six groups would fall in the follow
ing order: Normal, Triangle, Two-Lines-and-
Included-Angle, Straight Line, Non-Pattemed
Light, and Dark. The null hypothesis stated
that intergroup comparisons would not demon
strate this predicted inverse relationship.
Since the direction of difference was predicted,
the statistical tests would be one-tailed tests
of significance of differences in correlated
measures. The null hypothesis would be accepted
if any of the observed statistically significant
differences among the six groups were contrary
to the predicted direction of difference; it
would be rejected if statistically significant
differences in the expected direction were ob
tained in the four part-experience groups, i.e.,
the Non-Pattemed Light, Straight Line, Two-
Lines -and-Included-Angle, and Triangle groups.
A probability associated with the occurrence of
one-tailed tests of difference scores would have
to be equal to or less thano<« .05 in order to
be considered significant.
Alternate Experimental Hypotheses
Since it was considered a likely possibility that
the measure of discrimination learning used in the study
might prove too "crude" to pick up any differences that
might exist in so "fine" a gradation of part-experiences,
it was felt that essentially the same experimental hypoth
eses might be applied to a "coarser" scaling of the inde
pendent variable. This would be accomplished by combining
the stimulus groups so that, instead of six groups of nine
animals each, there would be three groups of eighteen
33
animals each.
There seemed to be adequate theoretical justifica
tion for combining the groups to make coarser groupings.
According to the interpretation of Hebb's theory given in
Chapter I, there should be no essential difference in the
performance of dark-reared animals and those reared in
non-pattemed light. Neither group having experienced any
patterned visual stimulation, neither could be considered
to have had any prior commerce with the elements of visual
form perception. Likewise, as long as a triangle was used
in the discrimination problem, no difference in the per
formance of animals who had an opportunity to view the
triangle and those reared in a normal environment should
be expected. Similarity in the performance of the Line
and Two-Lines-and-Included-Angle might also be reasonable
if both are thought of as providing patterned visual
stimulation but not more than part-experience toward the
perception of a total triangle. However, if Hebb's theory
is correct, these three groups should differ significantly
from one another on measures of discrimination learning.
Most additional learning (highest scores) should be needed
by the combined Dark and Non-Pattemed Light groups; inter
mediate amounts of additional learning should be needed by
the combined Line and Two-Lines-and-Included-Angle group;
least amount of additional learning (lowest scores) should
be required by the Triangle and Normal Cage groups when
34
combined into one.
It would be redundant to repeat the statements of
the specific experimental and null hypotheses or the sta
tistical tests which would be used. They require only
minimal rephrasing of those already stated to make them
applicable to the combined groups.
The actual execution of the experimental design,
the detailed operations that were conducted in the labora
tory, and consequently the operational definitions of the
variables and the terms used in the study should be clari
fied in the following two sections of this chapter, which
are devoted to a detailed description of the apparatus
and procedures.
Apparatus
From the time of their birth, the subjects of the
experiment had to be maintained in a visual environment
which could be carefully controlled. Adequate manipulation
of the independent variable necessitated equipment that
would enable the experimenter to keep the experimental
animals in complete darkness and/or to govern the amount
and the patterning of any visual stimuli they were per
mitted to view. The collection of data for the measurement
of the dependent variable also required apparatus in which
the amount and patterning of light was carefully regulated.
The successful execution of the entire experimental design,
35
as a matter of fact, depended heavily on the adequacy of
the equipment and apparatus, most of which was not commer
cially available and which, therefore, had to be specially
designed and built.
Specifically, the equipment and apparatus included
the following: breeding cages; two types of living cages,
those in which the experimental animals were kept and those
which housed the normal controls; feeding apparatus, to
permit maintenance of the deprivation schedule during the
period in which subjects were undergoing discrimination
training; the discrimination apparatus in which the train
ing occurred; and the simple device in which the animals
were transported from one apparatus to another so as to
minimize any extraneous exposure to light.
Breeding Cages
One of the cages in which the animals were bred, in
which parturition occurred, and in which the first nine
days of postnatal life were spent, is photographed in
Plate I. It was a stainless steel cage, 10 Inches wide and
16 inches long. At the rear, 5 inches of the cage was
"roofed" with solid metal and for this portion of the cage,
the height was 10 inches. From this "roof" a hinged 1/2-
inch by 1-inch wire mesh cover was attached and sloped
down to the front "wall" which was 7 inches high. This
gave a 3-inch slope in 11 inches so that a water bottle,
PLATE I
TOP: EXPERIMENTAL LIVING CAGE
BOTTOM: LEFT--BREEDING CAGE
RIGHT--NORMAL CAGE
38
placed in a groove in the wire mesh, could be tilted at an
angle sufficient to allow the free flow of water through
the water tube.
Two sides, the back wall, and the floor were solid
sheets of metal. The front wall had a 4-inch by 6-inch
opening in the metal over which a 1-inch by 1-inch wire
mesh removable food bin was mounted.
While light could freely enter these cages, the
animals could only observe activity in the laboratory
through the mesh top of the cage and through the front
mesh food bin, since the rest of the cage consisted of
solid construction.
These cages were obtained through the courtesy of
the University of Southern California Biochemistry Depart
ment, and it is not known whether they had been purchased
from some conmercial source or whether they had been
specially constructed for the department's use.
Living Cages
There were two types of living cages: those for the
normal control animals and those designed for the experi
mental animals.
1. Normal Cages. A photograph of a normal living
cage appears in the lower right-hand portion of Plate I.
These cages were constructed of 1/2-inch by 1-inch stain
less steel wire mesh. They were 7 inches wide and 15
39
inches long and 7 inches over-all in height, but the floor,
constructed of the same mesh, was raised one inch, leaving
the actual height of the living space at 6 inches. Within
the cage, suspended from the top, was a wire basket 6
inches long, 3 inches wide, and 4 inches deep, divided into
two compartments, one of which accommodated a water bottle
and the other serving as a food bin. The front of the cage
had a 4-inch by 6-inch access door of hinged mesh held by a
spring clip. This door covered a 3-inch by 5-inch opening
in the cage.
The cages were placed on stainless steel trays in
which the excrement was collected. They could be cleaned
simply by lifting them from their trays, cleaning the trays
and replacing the cages. This procedure did not require
that the animals themselves be handled.
No special stimuli were mounted on the cage walls,
but their construction permitted maximum visibility of the
laboratory because of the openness of the mesh.
2. Experimental Cages. A photograph of the experi
mental cage appears in Plate I. Details of its construc
tion are shown in Plate II. These experimental cages were
lightproof boxes constructed of 3/4-inch lumber. The outer
dimensions of these cages were 11 inches by 18 inches by
10 inches.
PLATE II
CONSTRUCT ION OF THE EXPERIMENTAL
LIVING CAGE
42
a. Provision for Ventilation, Feeding and
Watering: On both of the long sides of the box, ventila
tion ducts were constructed that permitted the passage of
air but were baffled to prevent the entrance of light into
the cage. Detail A on Plate II shows the construction of
the light trap. Air circulation was maximized by having
the ventilating duct built toward the bottom of the cage
from the right side and toward the top of the cage on the
left. On the front of the cage a water bottle was mounted.
A metal drinking tube was permanently inserted. To this a
black rubber tube led to the bottle which was held by a
spring clip. The bottle could be removed for refilling by
removing it from the rubber tubing. A clamp placed on the
tubing prevented the admission of light when the water
supply was being renewed.
Food was supplied by means of a tubular food reser
voir with an opening at the top of the cage. The tube was
an aluminum cylinder 1-1/2 inches in diameter and 7 inches
in length, extending about 1/2 inch above the top of the
outer wooden box. Food placed in the tube was fed by
gravity into a triangular metal food trough placed at an
angle in the lower comer of the box. The tilt of the food
trough helped to keep it continuously supplied and also
served as a light trap. The top of the trough was covered
with 1/2-inch wire mesh through which the animal had to
eat. This was added to prevent excessive spillage and
43
waste of food. In addition to the tilt of the food trough
and the presence of the food itself which also served as a
light barrier, additional precaution against the entry of
light through the food tube was insured through its being
fitted with a piston-like plunger which fit into the tube.
Attached to the plunger was a rod that served as a food-
level Indicator as well as a handle by which the piston
could be removed so that food could be added to the reser
voir. A sliding cap was mounted on the rod. It acted as
a cover for the orifice of the food tube and was, there
fore, still another means of preventing the entrance of
stray or uncontrolled light.
At the front or "working end" of the box was an
access door which consisted of a rectangular aperture
3 inches wide by 4 inches high. The inner surface of this
opening was lined with metal stripping to prevent gnawing.
On the outside of the box the aperture was surrounded by
strips of black sponge rubber. Above and below these
strips, tight hold-down springs were fastened. It was over
this aperture that an opaque masonite panel or the appro
priate stimulus panels were placed. When the opaque panel
was inserted between the sponge rubber and the hold-down
springs, a lightproof seal was affected. If a stimulus
panel were substituted for the masonite, it insured that
light could enter only through the stimulus pattern itself,
b. Provision for Visual Stimulation: The
44
stimulus panels were 5-inch by 7-inch photographic nega
tives. Three of these were made by photographing glossy
red geometric figures mounted on white paper. A fourth was
made by photographing an India ink stippled drawing on
white paper. These were made on photographic film which
gave only two values and, therefore, resulted in a negative
that contained only a completely opaque black or a clear
transparency. The stimuli were thus completely transparent
"cut-outs" on a completely opaque, black background. This
photographic process was selected to insure accuracy of
reproduction of the stimuli since many animals were to be
undergoing stimulation at the same time. For each stimulus
panel required, an appropriate photographic negative, along
with two sheets of paper from a 5-inch by 7-inch scratch
pad, was mounted between two sheets of clear glass of the
same size. The inclusion of the paper was necessary to
alter the stimulus "cut-outs" from a transparency to a
translucency. Had the stimulus remained transparent, it
could have served as a "window" through which the animal
might view other visual events in the laboratory. With the
addition of the two sheets of white paper, light was so
diffused that only faint shadows could be seen passing
close by, even when the object forming them was brightly
lit from behind. Nevertheless, the addition of the paper
did not eliminate the sharp definition of the contours of
the stimuli.
45
Film, paper and glass were bound together with
masking tape into a single unit. This constituted the
stimulus panel which was put over the front aperture of
the living cage and held in place by the hold-down springs.
The four types of stimulus patterns so photographed
and so mounted are those referred to as the Line, Angle,
Triangle, and Non-Pattemed Light stimuli. The Line, as
shown in Plate III, was a transparency 3 inches long and
1/8 inch wide centered on the opaque film in an orientation
60 degrees from an imaginary horizontal base. The Angle
panel, shown in Plate IV, consisted of two transparent
lines of 3 inches by 1/8 inch dimensions joined to form a
60-degree angle which was photographed with its apex at the
top-center of the panel. The third stimulus panel, illus
trated in Plate V, consisted of a transparent equilateral
triangle which, in effect, added only the horizontal base
to the angle. It thus consisted of three sides, each of
which was 3 inches by 1/8 inch in size and, of course, con
tained three 60-degree angles. The stimulus designated as
the Non-Pattemed Light stimulus, shown in Plate VI, was a
photographic negative of a stippled "nebula" with its
greatest density more or less centered and gradually becom
ing less dense toward its perimeter. When photographed,
this had a rather "astral" appearance with the light and
dark reversed so that the greatest amount of light entered
at the center and there was a gradation, without any
PLATE III
THE LINE STIMULUS
U6
PLATE IV
THE TWO-LINES-AND-INCLUDED
ANGLE STIMULUS
U8
PLATE V
THE DISCRIMINATION STIMULI
THE TRIANGLE IS THE SAME AS THAT USED FOR
THE TRIANGLE GROUP STIMULUS
50
PLATE VI
NON-PATTERNED LIGHT STIMULUS
52
54
definite boundaries, toward total blackness at the edges.
In all the panels, the opaque background surround
ing the figures insured that no inadvertent visual contour
would be furnished by the edges of the opening over which
the panel was placed.
Since the areas of the stimuli varied and the light
source for all of them was uniform, they admitted differ
ing amounts of light. A photographer's light meter gave
the following relative readings.The Line admitted 1.6
units of light, the Angle admitted 2.4 units. By a happy
coincidence, both the Triangle and the Non-Pattemed Light
panels admitted the same amount of light, 3.2 units.
c. Provision for Cleaning: From the descrip
tion of the living-cages thus far given, it may be seen
that they were equipped to supply adequate ventilation,
food, and water to the experimental animals while they were
kept in total darkness or in a highly restricted visual
environment. Another important consideration in designing
a lightproof living-cage was to devise a means whereby the
cages could be cleaned without the admission of light.
This was accomplished by having the inner floor of the cage
^Since these readings were made by the photographer
who made the photographic negatives, the exact conditions
under which the readings were made and the nature of the
units in which they are expressed is unknown. They were,
however, all made under the same conditions, i.e., at the
same distance from a uniform light source and with the
same meter.
55
made of 1/2-inch wire mesh through which waste food, feces,
and urine could drop into a drawer-like metal excrement
tray. A wooden drawer-pull extended one inch to either
side of the tray and was padded with black felt to prevent
the entrance of light. Before the excrement tray was
removed for cleaning, a light shield was inserted into a
narrow slit which was lined with black felt stripping and
was further light-shielded by an overhanging metal strip.
The light shield was always inserted before the excrement
tray was removed and was withdrawn only after the tray was
replaced in its proper position. The shield slipped under
the mesh flooring so that its insertion did not injure the
animal.
d. Additional Characteristics: The mesh floor
ing actually was only a part of the animal's actual living
quarters. Essentially the living quarters consisted of a
1/2-inch wire mesh cage suspended within the wooden light
proof outer box. The dimensions of this inner cage were
7 inches wide, 15 inches deep, and 6 inches high which, it
might be noted, were also the dimensions of the open mesh
cages in which the normal controls were kept. At any rate,
except for the front end of the box which contained the
water tube, the covered food trough and the metal-lined
stimulus aperture, there was a space of one inch between
the wire mesh walls of the inner cage and the wood walls
of the outer box. This space prevented the animal from
56
gnawing the wood.
As a final precaution against any uncontrolled entry
of light, all joints of the wooden box were calked and it
was coated inside and out with a non-toxic, black, matte
paint. As evidence that the box was truly light-tight when
closed with the masonite panel, a strip of film placed in
the box and exposed for one-half hour showed no fogging,
even though the box was in an illuminated room at the time.
Feeding Apparatus
Regulation of feeding to maintain a deprivation
schedule during discrimination training was not feasible
in the experimental living-cage. Therefore, when an animal
began training, the living-cage was cleaned and all food
removed. From that time until the time when training was
completed, all animals were fed one hour per day in a
special feeding apparatus as shown in Plate VII. This
consisted of a rectangular box 30 inches long, 16 inches
deep, and 9 inches high in its outer dimensions. It was
constructed throughout of 1/2-inch plywood.
Inside the box, raised 1-1/2 inches, was a 1/2-inch
wire mesh floor through which excess food and excrement
could drop into paper placed beneath it.
This large box was divided into twelve individual
feeding "stalls" or compartments, each of which had inner
dimensions of 4 inches in width, 7 inches in depth, and
PLATE VII
THE FEEDING APPARATUS
W A T E R BOTTLES
GUILLOTINE
DOOR
FLOOR J g
WIR E MESH
59
7 inches in height. Each of these stalls was accessible
through its own guillotine-type door.
Raised 3 inches from the top of the box, parallel
with it, and extending its length was a 6-inch wide panel
out of which twelve circles were cut. Each of these
circles was large enough to accommodate a water bottle.
These were placed in an inverted position in the rack thus
formed. Metal tubes, inserted in the black rubber stoppers
which capped the bottles entered each of the compartments
below through a small aperture bored through its "ceil
ing." The stopper rested over this aperture and made a
fairly effective light seal. Food was supplied to the
animals in the form of pellets which were simply placed on
the mesh floor of the stall.
No effort was made to keep the feeding apparatus
completely lightproof. However, it was kept effectively
darkened by the following means. The solid walls of the
box extended 1-1/2 inches below the mesh flooring. The
walls rather than the floor supported the box. Since the
box was placed on a level surface and kept in a dark room,
very little, if any, light entered through the bottom.
Furthermore, all surfaces inside and out were covered with
a black, matte paint to absorb light and to keep visibility
of contour at a minimum.
60
Discrimination Apparatus
The discrimination apparatus is illustrated in
Plates VIII and IX. It consisted of a box built of 3/4-
inch lumber with outside dimensions of 15 inches in length,
11-1/2 inches in width, and 9-1/2 inches in height.
Access to the box was by means of a simple lid which
lifted off. Except for the front edge, the sides of the
lid overlapped the box by one inch. The front edge of the
lid was covered with felt stripping and served as a
friction-hold on the shields which separated the rear com
partment from the stimulus panel assembly and the food
troughs.
There were two shields, each 10 inches by 9 inches
in size, which slipped into and were held in place by
grooves in the sides of the box. These shields could be
separately raised and lowered by means of finger grips at
their tops. One of the shields, that closest to the rear
of the box, was made of opaque metal. The other was
transparent lucite. When the felt stripping on the front
edge of the lid was abutted against either of the shields,
one or both could be kept raised until pushed down into
the closed position.
The front end of the box, separated from the rest
by the shields, consisted of two openings two inches apart.
Each of these was 3-1/2 inches wide and 4 inches high. An
animal extending its head through either of these openings
PLATE VIII
THE DISCRIMINATION APPARATUS
TOP: APPARATUS WITH OPAQUE SCREEN RAISED And
STIMULUS ASSEMBLY IN VIEWING POSITION
BOTTOM: LEFT— APPARATUS WITH STIMULUS ASSEMBLY
REMOVED TO SHOW FEEDING TROUGHS AND
APERTURES
RIGHT— RAT' S EYE-VIEW OF STIMULUS
ASSEMBLY WITH BOTH SCREENS RAISED
61
PLATE IX
CONSTRUCTION OF DISCRIMINATION APPARATUS
63
COVER
FELT
STRIPPING
OPAQUE SHIELD
TRANSPARENT SHIELD
ALIGNMENT PIN
TO POWER
TRANSFORMER
STIMULUS ASSEMBLY
FOOD TROUGHS
o. OUTER VIEW
COVEI
TRANSPARENT SHIELD
OPAQUE SHIELD
b. INNER VIEW
STIMULUS
ASSEMBLE
,TO POW ER
TRANSFORMER
65
could reach one of the food troughs in front of it. The
troughs were metal-lined 1/4-inch depressions extending
across the 4-inch width of the aperture and one inch
beyond them. They could accommodate wet mash and/or
pellets.
The complex part of the discrimination apparatus
consisted of its removable stimulus panel illustrated in
Plate X. This was composed of a unified group of three
identical swinging panels, each of them 6-1/2 inches tall
and 6>1/2 inches wide and in which there was a 4-inch by
4-inch opening covered with clear glass. Behind this
glass a visual stimulus could be mounted. Each of the
panels was effectively a box which had a one-inch depth as
well as the height and width stated. The back of each box
was covered with black, matte painted cardboard. A hole
lined with a rubber gasket was centered in the cardboard.
Through the hole a small light bulb was inserted which
illuminated the stimulus. The bulbs used were Tung-Sol
panel bulbs, 3.2V, .35 amp, with a miniature screw. Since
the stimulus figures mounted in these panels were trans
lucent "cut-outs" against an opaque ground, their centers
were completely opaque. The bulb, immediately behind the
central opaque area, could uniformly illuminate the
translucent pattern.
Power for the stimulus panel bulbs was supplied from
a 2.5V step-down transformer using current from a 110V wall
PLATE X
DETAILS OF STIMULUS ASSEMBLY
UPPER: OUTER VIEW
LOWER: INNER VIEW
66
POWER SOURCE FOR
STIMULUS LIGHTS
BATTERIES F O R
SIGNAL LIGHTS
M I C R O SWITCH
F O R S IG N A L
LIGHT
S IG N A L -
LIGHT
8TIMULUS LIGHTS
POW ER SOURCE FO R
STIMULUS LIGHTS
TRANSLUCENT STIMULUS 0
O P A Q U E B A C K G R O U N D , 3V;
STIMULUS CIRCLE - 3 *°
68
outlet, routed through a parallel circuit from a telephone
jack plug on the bar from which the panels were hung. This
bar fit across the top front of the box when it was prop
erly aligned. Two openings in the top of the bar permitted
two different alignment positions by placing one or the
other of them over a pin placed on the top of the center
post of the box. The bar was 2-3/4 Inches deep and 18
inches long and therefore extended both to the right and
left of the box. These extensions served as handles by
means of which the bar, and consequently the stimuli, were
moved from one alignment position to the other.
The three stimulus panels were hung from the under
side of this bar by means of piano hinges. Microswitches
were attached to the outer sides of the panels. These
operated small signal lights powered from two pencil flash
light batteries mounted on top of the bar by means of
clips. The microswitches were operated by pushing the
stimulus panels from the inside. Since the apparatus was
not a light-tight device, in order to minimize any extrane
ous visual stimulation the box was, as usual, painted with
a black matte finish. As an added precaution, discrimina
tion training was carried out in a darkened room and the
animals' responses were ascertained by the operation of the
signal lights. The microswitch on the center, or positive
stimulus panel, operated a red bulb. The two outer or
negative stimuli operated white bulbs. Since there were
69
only two openings in the discrimination apparatus, only
two of the three stimulus panels could be viewed by the
subjects at any one time. The use of three panels, with
the negative stimulus in the two outer panels and the
positive stimulus in the inner panel, made it possible, by
a simple manual shift of the stimulus panel assembly, to
vary the left-right presentation of the discriminanda.
The two outer panels, however, had only enough
"give" to operate the switches. They did not swing open
far enough to permit an animal to gain access to the food
trough. In both of these outer panels, identical "nega
tive" stimuli were mounted. The center panel, in which
the positive stimulus was inserted, however, did open out
sufficiently so that, providing both shields were raised,
an animal would be able to put its head out of the aperture
and eat from the food trough. However, even the positive
panel did not open far enough to enable the animal to get
out of the box entirely, nor to do much visual investiga
tion of the outside environment. In its most open posi
tion, it extended only to the outer edge of the food
trough. A prearranged, random sequence of left-right
presentations of the stimuli was followed simply by moving
the stimulus assembly from one alignment position to the
other.
With both the opaque and transparent shields in the
closed position, the stimulus panel assembly was lifted off
70
the box and repositioned. The closing of the shields pre
vented stray admission of light into the box, prevented
the subjects from looking out of the box into the labora
tory and also served as a visual cue demarcation of one
learning trial from the other.
Felt Transporting Sleeve
A final refinement which prevented the animals from
receiving any uncontrolled visual stimulation was the use
of a felt sleeve to transport them from one to another
device in which they had to spend some time. This felt
sleeve was a simple tube made by seaming a rectangular
piece of black felt cloth and drawing an elastic through
one end. The experimenter's arm was placed in the sleeve
with the elastic toward the shoulder. The sleeve was long
enough so that it completely enclosed the experimenter's
hand. In using it, the experimenter pulled the sleeve up
and placed the open end over the aperture of the apparatus.
He then reached in, picked up the animal in his hand, with
drew it through the door, into the sleeve, pulled the
sleeve down over animal and hand and reversed the process
when releasing the animal.
Further details of the operation of all the equip
ment should become evident from the description of the
procedures of the experiment in the following section.
Procedure
71
The experimental procedures are logically divided
into two parts. The first of these is concerned with the
treatment of the animals prior to training. This consti
tutes their experiential background. The second part
consists of the training procedures themselves. The ex
periential background of the animals includes such factors
as their breeding, weaning, handling, living conditions,
and visual stimulation. Training procedures are logically
divided into three parts: adaptation, pretraining, and
discrimination training.
Experiential Background
1. Breeding. Breeding stock was obtained from two
sources: the University of Southern California Biochemis
try Department and the California Caviary. Animals from
both sources were albino rats of the Wistar strain, 12
weeks to 6 months of age. Animals were bred by placing
two females in a single breeding cage with two males.
Matings were spaced at 3-week intervals so that no more
than two litters (or 12 experimental animals) would be
ready to begin training at the same time. Male and female
animals were allowed to remain together for a period of
18 days. Three days before the first expected day of
parturition the animals were separated and each female was
placed in a separate brood cage with fresh wood shavings.
72
Ad lib supply of Wayne Lab Blox and water were present at
all times. Animals were observed daily. Cages were
cleaned twice weekly unless parturition occurred before
scheduled cleaning time.
After the birth of the litter, mother and pups were
left undisturbed for three days. On the fourth day the
cages were cleaned, provided with fresh shavings, and the
litter culled, if necessary. A maximum of eight pups were
allowed to remain with the mother. After culling, mother
and pups were again left undisturbed until the pups were
nine days old. At that time they were removed to the
dark-type living-cage. Torn paper toweling was supplied
for bedding since shavings would have fallen through the
mesh flooring. Pups were removed first and allowed to
rest on the new nesting material for half an hour. The
mother was then placed in the cage with them.^
This procedure represents a modification of the
method originally proposed for breeding and rearing experi
mental animals. The original procedure called for placing
the pregnant females in a dark cage three days prior to
expected parturition to permit them to become accustomed
to their new surroundings. It was expected that the cages
would be checked daily until the birth of the litter.
When the pups were bom, the birthdate would be noted, the
litter culled, and the mother and pups allowed to remain
undisturbed for three weeks. At 2i days of age weaning was
to have occurred, the cage was to have been opened and the
pups transferred to their individual rearing-stimulus
cages. This weaning and transfer was to have been carried
out in complete darkness.
Ten pregnant females were treated in this manner.
In each instance the birth of a litter was noted. At wean
ing time, however, only one litter had survived. The rest
73
A clear glass panel was placed over the cage open
ing and allowed to remain for 24 hours. This gave the
mother the opportunity to explore the cage and familiarize
herself with its food and water mechanisms under conditions
of the established laboratory illumination. At the end of
the 24-hour period, when the pups were 10 days of age, the
clear glass was replaced by the lightproof masonite cover
so that mother and pups were in complete darkness. Wayne
Lab Blox Meal and water were available at all times.
had evidently been eaten by the mother since no traces of
the carcasses, beyond a few blood stains, remained. It is
likely that they were destroyed when still quite young.
The mothers appeared in good health and normal weight at
the end of the three-week period when the cages were
opened. It was impossible to ascertain the exact time of
the pups' demise because the cages had been left unopened
for the 21-day period which was to have been the interval
between birth and weaning.
No mention of this high incidence of cannibalism
could be found in the literature. Inquiry directed to
experienced "rat men," however, indicated that the total
darkness may have been a factor. The procedure was there
fore modified in the manner described. It represented a
compromise with the original plan in that the pups were
permitted to remain in a dimly lighted atmosphere for 10
days after birth, a period when tne eyes are still closed
and the retina is not fully mature. The assumption was
that, because of the immaturity of the visual mechanism
during this period, exposure to light could not affect
perception and, according to a communication from Dr. K.
Tansley of the Institute of Ophthalmology, this assumption
seems to be valid.
The modification of birth-to-weaning procedure cut
down the incidence of cannibalism to approximately two out
of ten litters as against the nine out of ten rate which
was obtained by the original method. It was thus possible
to obtain an experimental population.
74
Cleaning, feeding, and watering (as previously described)
required no admission of light into the cage. In this
manner, the pups, which constituted the experimental popu
lation, were kept in complete darkness from the age of ten
days up to the time of weaning.
2. Weaning. The pups were weaned when they were
21-24 days old. At the time of separation from the mother,
the laboratory was illuminated only by a photographer's red
darkroom bulb. The cage was opened and the mother removed
first. Pups were then removed, one at a time, and trans
ported in the felt sleeve to the individual living cages.
Members of the litter were randomly assigned to each of
the six stimulus conditions. Where more than six litter-
mates were available, the extra two were discarded at this
time.
3. Living Conditions. The rearing cages were
placed on stationary wooden racks, disposed in three tiers
along two sides of a four-foot wide aisle. Two inches of
air space was allowed on either side and six inches of
space was allowed above and below each cage. The tiers
were stepped back so that there was six inches of clearance
at the top-front of each cage to permit easy access to the
food-tube and clearance for the mounted water bottle.
Laboratory illumination was supplied by two
fluorescent lighting fixtures mounted three feet above the
75
highest tier in the center of the aisle and extending its
length. The stimulus panel of all cages faced this aisle.
Illumination was supplied to simulate a twelve-hour day-
night cycle.
While constant temperatures could not be maintained,
the laboratory was supplied with two electric heaters with
thermostatic controls. A thermometer was mounted on the
wall farthest from the locations of the two heaters. The
thermometer was checked three times a day, roughly corres
ponding to morning, afternoon, and night, although no time
schedule was maintained. By regulating the heaters at
these times, an effort was made to keep temperatures at an
optimum level. At no time did the thermometers read below
60 degrees or above 90 degrees Fahrenheit. For the most
part, temperatures were successfully kept within the 75-85-
degree range recommended by Farris and Griffiths (1949) and
Munn (1950).
Cages were cleaned twice a week without any admis
sion of additional light beyond that obtainable through
the stimulus panels. Animals were not handled at all from
the time of weaning up to the time of removal from the
cages for adaptation to the discrimination apparatus.
4. Feeding. All animals except the normal cage
controls were fed on a half and half mixture of Wayne
Krumbettes and Wayne Lab Blox Meal. Normals were fed
76
Wayne Lab Blox in the block form.
Wayne Krumbettes is a dry food which is formed into
small cylindrical shapes--approximately 3/8 to 1/2 inch in
length and 1/4 inch in diameter. No analysis of its in
gredients is available but it is supposed to supply a
complete diet for dogs.
Wayne Lab Meal is a diet designed especially for
laboratory rats and mice and is supposed to supply all
known nutritional requirements for these animals so that
no supplementary feeding is necessary. It is made by
grinding Wayne Lab Blox into a powder. The fineness of
the particles, however, gave the substance a tendency to
pack in the food tube. The addition of the coarser
Krumbettes eliminated this problem and insured an even
flow of food from the tube into the feeding dish. Many
investigators report the use of Purina Dog diow as an
unsupplemented laboratory diet, with no apparent ill
effects. It is likely that the two products are similar
in their ingredients and nutritional value.
Wayne Lab Blox is a compressed food shaped in the
form of tablets of uneven size but approximately 1-1/2
inches long, 3/4 inch wide, and 1/2 inch thick. These
dimensions made unfeasible its use in the experimental
living cages, since the blocks were too large to fall
freely through the food tube unless they were painstakingly
and evenly stacked.
77
Wayne Lab Blox is analyzed into the following com
ponents: crude protein, 24.0 per cent, crude fat, 4.0 per
cent, crude fiber, 4.5 per cent, and consists of the fol
lowing ingredients: animal liver meal, fish meal, dried
whole whey, com and wheat flakes, ground yellow com,
ground oats, de-hulled soybean oil meal, wheat germ meal,
wheat standard middlings, feeding cane molasses, dehydrated
alfalfa meal, soybean oil, brewers yeast, irradiated yeast,
dicalcium phosphate, salt, ferrous carbonate, copper sul
phate, potassium iodide, cobalt carbonate, manganese sul
phate, riboflavin, niacin, calcium pantothenate, choline
chloride, Vitamin A feeding oil, D-activated animal sterol,
and a-Tocopherol.
Animals gave the impression of making normal weight
gain and growth on this diet. However, it was impossible
to weigh them without exposing them to light. Therefore,
no objective data are available.
All animals had ad lib supplies of food and water
up to the time of the beginning of training. Food and
water were checked twice daily to insure a continually
adequate supply.
4. Visual Stimulation. At the time of weaning,
when animals were placed in their individual cages, the
assigned stimulus panel was placed at the cage opening.
The stimulus panel remained fixed up to and during the
78
training procedures. For 12 hours per day for approxi
mately 90 days, all animals were exposed to one of the
specified stimulus conditions. However, the longer an
animal required for learning to operate the apparatus and
to meet the criterion for the discrimination task, the
greater was the duration of his exposure to the stimulus
condition experienced in the living-cages, since the
stimulus panel was permitted to remain on the living-cages
during the period animals were undergoing training.
Training
The training of the animals was divided into three
parts. There was, first, a period of adaptation to
handling and to the discrimination apparatus. This was
followed by the pretraining period and then by the dis
crimination training which provided the measure of the
dependent variable in the study.
1. Adaptation Procedure. The adaptation of the
animals to handling and to the apparatus required a minimum
of five days and was begun when the animals were 86-94 days
old. Except for the first day, the animals were approxi
mately 23 hours food-deprived at the start of each training
session. Animals were transported from the living-cages to
the apparatus in the black felt sleeve. The laboratory was
kept dark whenever animals were being handled. The train
ing also was conducted in an unilluminated laboratory after
79
nightfall. In addition, a black felt cloth was draped
over the apparatus to keep it in complete darkness in order
to prevent the animals from experiencing any uncontrolled
visual stimulation.
On the first day of adaptation, the animals were
removed from their living cages. This was their first
handling since the time of weaning. They were placed in
the dark discrimination apparatus which had one open door
baited with wet mash but no visible stimuli. For 15 minutes
the left-hand door was propped open. For the next 15
minutes, the door on the right was opened and baited; the
door on the left was closed and the bait removed. After
30 minutes, the animals were returned to their living-
cages, supplied with ad lib water but no food.
On Day 2, the animals were 23 hours food-deprived.
They were placed in the discrimination apparatus and the
first day18 procedure repeated, except that the door was
first opened on the right and then on the left and baited
accordingly.
On Day 3, both doors were baited. The left door
was propped half open so that the animal would have to
push it slightly to get to the wet mash. The left door
remained half open for 15 minutes, was then closed and the
right door was propped half open for a 15-minute period.
Each time the position of the open door was shifted,
the opaque and transparent screens were lowered and the
80
stimulus panel lifted off the box. This was necessary, not
only to keep any stray light from entering the apparatus
but also in order to accustom the animals to the noises
that would normally occur in their subsequent training to
discriminate the visual stimuli.
On Day 4, both doors were closed. Both food troughs
were filled. In addition, the positive door was smeared
with wet mash. Animals had 15 minutes with the positive
door on the right and then 15 minutes with the positive
door to the left.
On Day 5, the animals were placed in the apparatus
with both doors closed and baited but not smeared with mash
or in any other way identified. They were given 15 minutes
with the unlatched door on the left and then 15 minutes
with the positive door on the right.
The situation all through the adaptation period al
lowed the animals to correct their errors. Thus, if they
tried to open the incorrect door and found it latched,
there was the opportunity to try the other door. During
the entire adaptation period, the apparatus was dark and
the doors unmarked so that there were no visual cues to
lead the animal to make the correct choice, nor did the
adaptation period give the animal any additional visual
experience to that which he obtained in his living-cage.
Olfactory cues were also minimized by the fifth day by
having both food bins baited with equal amounts of the
81
wet mash and no mash on the doors themselves.
For the greatest majority of the animals, these
five days of adaptation proved sufficient to enable them
to learn to operate the doors of the apparatus and to
obtain the wet mash placed in the feeding troughs. A few
of the animals, however, did not eat in the apparatus even
when the doors were left open, and a few of them would eat
only when the doors were propped open but would not push
the doors to get the food. For these animals the adapta
tion procedure was prolonged or modified. If an animal
did not eat in the apparatus for two days even when the
doors were opened, it was allowed to feed for an hour in
the regular feeding trough and then returned to its living-
cage. The following day it was placed in the adaptation
apparatus again. The appropriate door was left open for
half an hour and then the alternate door was opened for an
equal length of time. Both food troughs were baited. If
the animal still did not eat, the investigator held the
animal and guided it to the door. (This was done in a com
pletely darkened laboratory in order to prevent uncon
trolled visual stimulation.) Usually following this pro
cedure, the animal would approach the doors on the follow
ing day when it was again placed in the apparatus. If the
animal was still refractory, it was allowed to remain in
the apparatus for one houx; during which time alternate
doors were propped open every 15 minutes. Once the animal
82
ate through the opened door, its next day's training fol
lowed the usual procedure. If at any point in the adapta
tion procedure the animal again balked, it was allowed to
go back to the earlier stage and work its way through the
adaptation again. If an animal ate from the opened door
but spent two days in the apparatus without attempting to
open a closed door which was smeared with mash, the experi
menter would again hold the animal and guide it to the door
and push its nose against the door gently until the door
opened and the mash was obtainable, after which the animal
was allowed to remain in the apparatus for an hour while
alternate doors were propped half open at 15-minute inter
vals .
In no instance was an animal started on the pre
training procedure until it had become adapted to the
apparatus, opening the unsmeared doors promptly and eating
the mash from the feeding trough. Unfortunately, since no
difficulty was anticipated and since almost half of the
animals had already been adapted before difficulty was
encountered, no records were kept of the actual time spent
in adaptation or of the special assistance that had to be
given to the few animals that required it. At most, how
ever, only four or five of the 57 animals who were trained
required additional time in adaptation beyond the standard
procedure.
83
2. Pretraining Procedure. Once the animals had
been adapted to the apparatus, pretraining was begun. As
during adaptation, the entire pretraining procedure was
carried out in the dark laboratory, the apparatus was kept
dark, and no visual stimuli appeared on the doors. The
doors were unidentifiable by olfactory cues since both
feeding troughs were supplied with wet mash. The animals
were trained to an alternation pattern. A non-correction
method was employed. The criterion for ending pretraining,
however, was not the learning of an alternation habit but
the development of an average reaction time of five seconds
on ten consecutive trials. At all times during training,
the 23-hour food-deprivation schedule was maintained.
A pretraining trial proceeded in the following manner.
The laboratory was unilluminated. The animal was carried
from the living-cage to the apparatus in the felt sleeve.
It was placed in the apparatus which had both the trans
parent and opaque screens in the closed position. The
unilluminated stimulus panel was removed and mash placed
in the food troughs. The panel was then replaced. After
a five-second interval, the opaque screen was lifted.
After another five seconds, the transparent screen was
lifted. Although lifting the screens did not, in this
case, provide any change in visual stimulation, it provided
the auditory environment which would later accompany dis
crimination training.
84
As soon as Che animal pushed open a door, time was
recorded and five seconds allotted for the animal to obtain
the food reward. At the end of this interval, both the
opaque and transparent screens were lowered simultaneously.
The same operations were repeated for each trial, with the
"positive" door, i.e., the unlatched door, open alternately
on the left and right.
On the first day of pretraining, the animals were
given four to ten trials, depending upon the length of time
it took them to respond. A maximum of twenty minutes were
spent with each animal and a five-minute limit per trial
was adopted. If the animal did not respond in three
minutes, E opened the correct door and held it open for
two minutes, or less if the animal took the bait before
the maximum allotted time.
On the second day a twenty-one-minute maximum per
animal was allowed. On this day, however, a three-minute
criterion was adopted. If the animal did not push open
one of the closed doors within two minutes, E opened the
correct door and held it for one minute. Whether the
animal did or did not take the bait during this interval,
both screens were lowered, five seconds were allowed to
elapse, and a new trial begun.
On the third day the procedure of day two was
repeated. The three-minute criterion was maintained and
a maximum of twenty-one minutes were allotted to each
85
animal.
On the fourth day of pretraining a two-minute cri
terion per trial was adopted. This permitted ten trials
within a twenty-minute maximum per animal. If the animal
did not respond by pushing a closed door in one minute, E
held the correct door open for one minute. If the animal
did not respond during this interval, the screens were
lowered, the position of the correct door shifted, and,
after a five-second interval, a new trial begun.
On the fifth day and on all subsequent days required
for pretraining, the animals were permitted two minutes per
trial and E did not open doors. If the animal did not re
spond within the two-minute interval, the usual procedure
was followed for terminating one trial and beginning
another. A maximum of ten trials was given during each
day of pretraining, however, even if an animal required
less than the maximum time allotment. Training continued
at ten trials per day until the animal responded with an
average reaction time of five seconds per trial for a ten-
trial block.
At the end of each day's pretraining, the animals
were placed in the feeding apparatus. They were permitted
to remain in it for one-half hour. At the end of the half
hour, six to eight pellets of Wayne Lab Blox were placed
in each compartment. This amount of food was about two or
three times as much as an animal's normal daily intake.
86
The animals were then allowed to feed for one hour, after
which they were returned to their living-cages. Water was
available throughout the waiting and feeding period. All
transfers were made In a darkened laboratory and the
animals were transported Individually In the felt sleeve.
No food was available In the llvlng-cages, although a
supply of fresh water was available at all times.
In general, this procedure was adequate for pre
training . However, several animals, despite their previous
adaptation to the apparatus, did not open the doors or eat
at the beginning of the pretraining trials. When an animal
did not respond during the allotted time, It was placed In
an alternate but Identical training apparatus Instead of
the regular feeding box. In effect, the adaptation proce
dure was repeated. The feeding troughs were filled with
wet mash, and the doors closed. A total of one hour was
allowed for eating. This time was divided into fifteen-
minute intervals, during which alternate doors were left
unlatched so that the animal could push them open and
obtain food. If thirty minutes elapsed without the
animal's eating, the door was propped half open. If the
animal did not eat by the end of the hour, it was put in
its living-cage without food. If more than one day of
this procedure was required, the animal was placed in the
feeding box, pellets were supplied, and it was given an
hour to eat in addition to the hour spent in the apparatus.
87
In no instance was an animal deprived of food for more
than forty-eight hours.
Except for a few extremely refractory animals, a
day or two of this additional "adaptation" concurrent with
the pretraining was sufficient to bring the animal to
respond quickly enough to reach the pretraining criterion.
In a few instances, however, it was necessary to continue
the procedure and, occasionally, to modify it even further.
A few cases required repetition of the entire adaptation
procedure, beginning with the open door, the half opened
door, and the closed door smeared with wet mash before
they would venture to push a closed door open in the pre
training situation. Two animals did not respond even to
this treatment and had to be guided by having E hold them
and push them against the door and placing their noses in
the wet mash behind it. All the animals, however, eventu
ally learned to operate the apparatus and reached the
criterion for pretraining. No animal was discarded for
failure to learn this part of the procedure.
3. Discrimination Training. Discrimination train
ing was also given in blocks of ten trials per day. Pro
cedure and timing of the pretraining situation was dupli
cated in all respects except that during discrimination
training, the stimulus panel was illuminated, and the
triangle and circle between which the animal was required
88
to discriminate were shifted from the left to right posi
tion in a prearranged random sequence. The triangle was
the positive stimulus and the door on which it appeared
was always unlatched. The circle door was always latched
but the food troughs behind both doors were always baited
with wet mash to eliminate any identifying olfactory cues.
The day's trials were begun by removing the 23-hour
food-deprived animal from the living-cage and transporting
it to the apparatus through the darkened laboratory in the
felt sleeve. The animal was placed in the apparatus which
had both the transparent and the opaque screens in the
closed position. Food troughs were baited and the stimulus
panel placed in position. This took approximately five
seconds. At the end of this time the opaque screen was
lifted and the animal was given five seconds to view the
stimulus panel through the transparent screen which served
as a barrier to the doors. After the five-second viewing
time the transparent screen was raised and the animal per
mitted to respond. The animal's reaction time was noted
and five seconds were allotted for E to record time and
for the animal to obtain its reward if the correct door
had been selected. At the end of this time both the opaque
and transparent screens were simultaneously lowered. If
the animal made an incorrect choice, both screens were
lowered immediately. Regardless of whether the choice was
correct or incorrect, a five-second interval was given
89
before the start of the next trial. Between each trial
the stimulus panel was removed and replaced so that audi
tory cues were the same, regardless of whether the positive
stimulus remained at its previous setting or was shifted to
the alternate position.
The criterion adopted for discrimination learning
was eighteen correct responses out of twenty consecutive
trials. The criterion for non-learning was set at 300
trials without reaching the learning criterion.
Daily records of three measures of discrimination
learning were kept. The number of trials, as stated, was
set at ten per day, but records were also kept on the
number of errors and correct responses made in each block
of ten trials and the time required for each separate
response was recorded. A stop watch was used for measur
ing time on all intervals and responses. Measures of
time, therefore, must be considered only as crude approxi
mations since they were confounded with E's reaction time.
CHAPTER IV
RESULTS
The findings of the study may be grouped into two
broad categories. The first one is primarily descriptive.
It classifies and organizes. In it will be found informa-
tion about the samples of subjects and the scores they
earned in performing the task which served as the dependent
variable of the experiment. The second major category is
one which is primarily analytic. It presents the evidence
for support or rejection of the experimental hypotheses.
Each of these categories of findings is taken up in turn.
Descriptive Findings
The Experimental Subjects
1. Male versus Female Groups. The total sample of
fifty-four rats^ which was used in the study came from
nine litters bred in the laboratory. Since the distribu
tion of subjects involved splitting of litter-mates across
the six experimental stimulus conditions and had to be
*Ten litters were actually bred, but in one litter
three animals died before training was begun. The remain
ing three were trained but the data were not included since
they did not represent a complete litter.
90
91
carried out in the dark, it was impossible to ascertain the
sex of the animals before assigning them to their groups.
While it might have been preferable to have had equal
numbers of males and females in each of the stimulus
groups, it was assumed on the basis of a study by Forgus
(1958) that sex differences would not significantly affect
the discrimination learning scores. Further, it was
thought that if animals did differ on the basis of sex,
their random assignment to the treatment conditions would
effectively cancel out such differences.
To determine whether these assumptions were justi
fied by the data, sex was noted for all experimental
animals after they had completed discrimination training
and could be brought into the light. The data in Table 1
would seem to indicate that the distribution of animals
was a chance assignment of the sexes to the six stimulus
condition groups.
The Chi-squares show that while the obtained differ
ences between the number of males and the number of females
is a departure from the expected equal division of the
sexes, this kind of inequality might be expected to occur
by chance at least fifteen out of a hundred times. The
obtained difference consequently cannot be regarded as
statistically significant. Nor does the unequal division
of the sexes in each of the stimulus groups appear to be
cause for concern. Only one group, the Angle group, shows
92
TABLE 1
DISTRIBUTION OF SEXES IN THE TREATMENT GROUPS
Significance of
difference
Stimulus Number of subjects Approx.
conditions Males Females square P
Dark 4 5 0.09 0.75
Non-Pattemed 6 3 1.00 0.30
Line 3 6 1.00 0.30
Angle 1 8 5.29* 0.03*
Triangle 4 5 0.09 0.75
Normal Cage _3 _6 1.00 0.30
Total 21 33 2.24 0.15
♦Accepted as a significant departure from chance expectan
cies.
93
a distribution of males and females that is likely to
occur fewer than five times out of a hundred. However,
there is a likelihood that twenty times out of a hundred
one group out of six would have such a disparate division
of the sexes (Guilford, 1956).^
What was more important for the interpretation of
the results of the study, however, was to determine whether
males and females differed significantly in their perform
ance on the discrimination-learning task. To obtain some
evidence in this matter, the mean scores on the three
acquisition measures were estimated separately for the two
sexes, disregarding the stimulus conditions, on the grounds
that a random assignment of the sexes to these conditions
had been satisfactorily demonstrated. Casual inspection
of Table 2 will show that no significant differences were
found between the mean scores earned by males and females
on any of these measures.
It will be recalled that a criterion of eighteen
2
As determined by formula:
t - fl - f2
^ 1 + f2
which tests the departure of two frequencies from equality
when is the larger of the two frequencies and fi + f2 *
N. With one degree of freedom, t is equal to Chi, so the
results obtained from the above formula, when squared, give
the value of Chi-square. Probabilities of their occurrence
may then be found directly from the appropriate tables.
94
TABLE 2
DISTRIBUTION OF ACQUISITION SCORES
FOR MALES AND FEMALES
Acquisition
scores
Trials
Errors
Time
(in seconds)
Males
N-21
Mean Sigma
152.9 67.9
54.8 30.7
1089.2 425.3
Females
N-33
Mean Sigma
149.9 73.4
57.2 35.1
983.1 466.5
NOTE: No significant differences between means
of males and females on any of the
acquisition scores.
95
correct responses out of twenty consecutive trials was
adopted as an indication that discrimination*learning had
been accomplished. On this basis it may be concluded from
the data that males and females do not appreciably differ
from one another when total number of trials required,
total number of errors made, or total time of exposure to
the visual stimulation of the discrlminanda are used as
measures of discrimination*learning. Consequently, sex
distinctions were discarded in all further computations.
2. Litter versus Treatment Groups. An analysis of
variance in a two-way classification without replications
was run to describe the relative effects of litter and
treatment-group membership (Guilford, 1956). While this
type of analysis is ordinarily done to test an hypothesis,
in this instance it was used primarily as a descriptive
tool.
The split-litter design was adopted, not to enable
the determination of interaction effects nor to determine
the possible effects of litter-qiembership and its genetic
implications. The hypothesis under test simply required
the determination of the effects of varying the stimulus
conditions. The test of this hypothesis, however, re
quired some means of eliminating or controlling litter-
effects so that results could logically be attributed to
conditions rather than to litters. A rank-order type of
96
analysis of variance accomplishes this and was, therefore,
chosen as the statistical tool for hypothesis testing.
Nevertheless, it seemed a matter of interest to find
out whether litter-membership or treatment condition was
the source of greater variation in the performance of the
discrimination task. The fact that litters were split
across treatments made such an analysis possible.
Table 3 gives the results of this analysis for each
of the three measures of the dependent variable.
It is quite clear from these results that litter-
membership accounts for more variation in the solution of
this discrimination problem than the visual environment to
which the animal was subjected. It would appear, in fact,
to be the sole significant contributor to the differences
observed with all three measures. However, since this type
of analysis does not permit the computation of an interac
tion or error term, it was not possible, from these data,
to determine the variance of the treatment conditions if
the effects of litter variance were removed or to determine
the relative advantage or disadvantage contributed by the
visual environment to bright or dull rats.
The Acquisition Scores
1. Individual Treatment Groups. To recapitulate
briefly, three types of scores were derived from the experi
mental procedure. These were: (1) total number of trials
97
TABLE 3
ANALYSIS OF VARIANCE IN A TWO-WAY CLASSIFICATION WITHOUT
REPLICATIONS: LITTER VS. TREATMENT GROUPS
Source of
variation df
Sum
of squares Variance F P
TOTAL TRIALS
Litters 8 87,811.7 10,976.5 2.67 .025
Treatment 5 23,490.4 4,698.1 1.14 N.S.
Litter x
Treatment 40 164.405.6 4,110.1
- -
Total 53 275,701.7
- - -
TOTAL ERRORS
Litters 8 20,467.0 2,558.4 2.62 .025
Treatment 5 5,161.4 1,032.3 1.06 N.S.
Litter x
Treatment 40 39.063.4 976.6
- -
Total 53 64,691.9
- - -
TOTAL TIME (in aeconds)
Litters 8 4,559,393.4 569,924.2 3.50 .005
Treatment 5 486,865.8 97,373.2 .60 N.S.
Litter x
Treatment 40 6.516.636.4 162,915.9
- -
Total 53 11,562,895.5
- -
98
to reach the criterion; (2) total number of errors, or
incorrect choices, made during these trials; and (3) total
time of visual exposure of the discrlminanda during these
same trials.
Table 4 presents the arithmetic means, standard
deviations, medians, and semi-interquartile ranges for
these three acquisition scores for each of the six experi
mental treatment conditions.
From Table 4 it may be seen that the Total Trials
score gave the most symmetrical distribution of scores for
all groups. The means and medians were similar in numeric
value and both measures gave almost exactly the same rank
order assignments to each of the six groups. The Total
Error and Total Time scores showed some skewing. This was
most marked for the Non-Pattemed group, for which the
median was lower than the mean. The original data revealed
one exceptionally high score in this group. Its effect was
to elevate the mean to such an extent that the rank posi
tion was different for this group when ranking was done on
the basis of means compared to ranking based on medians.
Table 4 also shows that there was great variability
of the scores around the measures of central value. Inspec
tion seemed to indicate thaf there was less agreement in
the rank order assignments based on the Total Error and
Total Time scores than on the Total Trials scores for ranks
based on standard deviations compared to those based on the
99
TABLE 4
DISTRIBUTION OF THREE ACQUISITION SCORES
FOR INDIVIDUAL TREATMENT GROUPS
Stimulus condition
groups Mean Median Sigma
Q
TOTAL TRIALS
Dark 148.9 150 77.5 55.0
Non-Pattemed 141.1 130 61.5 52.5
Line 157.8 160 87.7 75.0
Angle 187.8 170 87.7 95.0
Triangle 153.1 150 43.8 25.0
Normal Cage 117.8 120 66.5 47.5
Total 151.1 145
TOTAL ERRORS
Dark 57.4 57.0 36.1 23.75
Non-Pattemed 50.1 36.0 32.1 26.75
Line 61.2 63.0 43.9 29.25
Angle 71.2 63.0 37.6 39.25
Triangle 58.0 56.0 16.8 9.25
Normal Cage 39.4 46.5 26.0 20.50
Total 56.2 53.0
TOTAL TIME
( in seconds )
Dark 954.9 893 479.8 318.7
Non-Pattemed 1018.1 886 431.6 402.5
Line 1043.1 1091 519.7 497.2
Angle 1191.4 1207 472.1 502.3
Triangle 1055.9 929 266.2 245.5
Normal Cage 883.0 866 500.4 489.7
Total 1024.4 921
Sigma denotes standard deviation.
Q denotes semi-interquartile range.
100
semi-interquartile range.
A quick check was made on these surmises drawn from
inspection of the descriptive statistics. A Spearman rank
order correlation coefficient was computed between the
means and medians of each of the six groups. The same
coefficient was computed for the two measures of variabil
ity, the standard deviation, and the semi-interquartile
range (Guilford, 1956).^ Table 5 shows the correlations
which were obtained.
A highly significant, almost perfect, correlation
was obtained between the ranks assigned to the groups by
the means and medians of the Total Trials scores. What
appeared to be a high correlation was obtained for the
ranks assigned to the six groups by the standard deviations
and semi-interquartile ranges, but it was not statistically
significant for so small a sample. Neither the Total Time
nor Total Error scores gave comparable agreement. On this
basis, for the individual groups, it would appear that the
Total Trials scores would be the measure of choice for the
dependent variable in testing the hypotheses of the study.
3
Calculations based on formula:
i i 6£ d2
rh° " 1 “ N ( & - _ 1}
When - sum of the squared differences between ranks and
N - number of pairs of measurements.
101
TABLE 5
RANK ORDER CORRELATIONS OF TWO MEASURES OF CENTRAL VALUE
AND TWO MEASURES OF DISPERSION
IN THREE ACQUISITION SCORES
Acquisition
scores
Spearman rank order correlation coefficients
Individual treatment
groups
N - 6
Composite treatment
groups
N - 3**
Mean vs.
median
Sigma vs.
Q
Mean vs.
median
Sigma vs,
Q
Total trials 0.99*
Total errors 0.89
Total time 0.89
0.92
0.89
0.54
1.00 1.00
0.5 0.5
0.5 1.00
*Significant at <.02 level on two-tail test.
**N of 3 too small to yield statistically significant
correlations.
102
2. Composite Treatment Groups. The alternate
hypotheses of the study called for comparisons of composite
groups which represented coarser scale divisions of the
independent variable than those represented by the six
individual treatment groups. Table 6 shows the distribu
tion of the three acquisition scores for these composite
groups.
Again it would appear that the most symmetrical
distributions were obtained from the Total Trials scores.
Referring back to Table 5 will show that complete agreement
in rank order assignments was obtained from this score both
between the two measures of central tendency and the two
measures of dispersion. Neither of the other two acquisi
tion scores gave this perfect correlation. It was impos
sible to determine significance levels for these correla
tions since they were computed on an N of 3. However, it
would again appear that the Total Trials scores would be
the measures of choice in making comparisons among the
composite groups just as they seemed to be the preferred
measures for determining the performance of each treatment
group individually. It would also seem likely that if the
Trials scores were used in calculations based on rank order
of the groups, it would be immaterial whether ranks were
determined from means or medians.
The differences in rank order correlation shown on
Table 5 are primarily due to differences in the symmetry of
103
TABLE 6
DISTRIBUTION OF THREE ACQUISITION SCORES
FOR COMPOSITE TREATMENT GROUPS
Stimulus condition
groups Mean Median Sigma
Q
i
rOTAL TRIA1LS
Dark and Non-Pattemed
Line and Angle
Triangle and Normal
145.0
172.8
135.5
140
165
135
99.0
124.2
79.6
55.0
87.5
27.5
Total 151.1 145
TOTAL ERRORS
Dark and Non-Pattemed
Line and Angle
Triangle and Normal
53.8
66.2
48.7
45.5
63.0
51.0
48.3
57.7
30.8
25.75
32.75
34.00
Total 56.2 53.0
TOTAL TIME
(in seconds)
Dark and Non-Pattemed
Line and Angle
Triangle and Normal
986.5
1117.3
969.4
892
1149
971
645.4
702.1
566.8
394.8
508.1
303.7
Total 1024.4 921
Sigma denotes standard deviation.
Q denotes semi-interquartile range.
104
the distributions of each of the three different acquisi
tion scores. They indicate only whether the means and
medians of a given score might be used interchangeably in
assigning ranks to the treatment groups. It has been shown
that only with the Trials scores would such an interchange
be possible without significantly affecting the ordinal
position of the various groups.
This type of correlation, however, did not provide
information as to whether the three acquisition scores
themselves were significantly interrelated. Simple logic
would indicate that they would be. When using eighteen
correct responses out of twenty consecutive trials as a
criterion of learning, it stands to reason that the more
trials an animal requires to learn, the more errors it is
likely to make and the more time it is likely to consume.
Nevertheless, this simple, logical conclusion is not
inevitable. While both the Error and Time scores must vary
with the Trials score, each has the possibility of varying
considerably from it. In a block of twenty trials, one
animal could make twenty errors. In a similar block of
trials another animal might make only three errors. Both
would fail to meet the learning criterion. It is obvious
that two animals could thus have the same Trials score but
differ widely on their Error scores, and vice versa. There
is even greater likelihood of concomitant variation of the
Trials and Time scores. The Time score incorporates the
105
Trials score because five seconds were uniformly allotted
as viewing time before the start of each trial. However,
since each animal was allowed to react to the stimuli with
out any prodding once discrimination-training was begun,
the Time score reflects the confounding of the trials to
criterion scores with the reaction time of the subjects as
well as with their error scores. A variety of possible re
sponses is thus introduced. One animal might react slowly
but make few errors. Another might be slow and inaccurate.
A third might respond both quickly and accurately or, yet
again, quickly and inaccurately. Comparably, two animals
might have required the same number of trials to reach
criterion but one may have reacted much more slowly than
the other on all of the trials.
It would merely belabor the obvious to indicate all
the possible variations that might be introduced. The
question raised by the logical possibility of these varia
tions required an empirical answer. To provide it, Pearson
correlation coefficients were computed from each individual
subject's responses on all three scores (Guilford, 1956).^
The following correlations were obtained: (1) between
^Using the formula:
nZxy - gx)gY)
where X and Y are original scores in variables X and Y, and
N - number of pairs of scores.
106
Trials and Errors scores, r - 0.93; (2) between Trials and
Time scores, r - 0.89; (3) between Errors and Time scores,
r * 0.83. A multiple correlation was then computed
(Guilford, 1956).Because N was less than 100, a correc
tion for bias due to small samples was made (Guilford,
1956).This resulted in an Rj.,23 “ 0.94. With correla
tions so high, it hardly seemed likely that any of the
acquisition scores were measuring significantly different
aspects of behavior. All the correlations obtained were
significant considerably beyond the .01 point. It would,
therefore, seem reasonable if the scores were to be used
interchangeably in further calculation.
However, since correlations were not perfect and
small differences could conceivably be magnified by group
ing, and particularly since assigning groups to ranks can
further effectively magnify small differences as well as
compress large ones, Spearman rank order coefficients of
correlation were computed for the ranks assigned by the
various scores. As may be seen from Table 7, for the
**From the formula
R1.23 “ f212 + f213 ~ 2r12r13r23 and R1#23 --\!r21.23
i *-2
1 - r 23
^From the formula CR^ ■ 1~(1 - R^) (^ ~ ™)
where N - number of cases in the sample, m - the number of
variables, and N - m - degrees of freedom.
107
TABLE 7
RANK ORDER CORRELATIONS OF THREE ACQUISITION SCORES
WITH EACH OTHER
Spearman rank order correlation coefficients
Acquisition
8cores
correlated
Individual treatment
groups
N - 6
Composite treatment
groups
N - 3**
with
each other
Ranked by
means
Ranked by
medians
Ranked by Ranked by
means medians
Trials x
errors 1.00* 0.92 1.00 0.5
Trials x
time 0.89 0.99* 1.00 0.5
Errors x
time 0.89 0.89 1.00 1.00
^Significant; at <.02 level on two-tail test.
**N of 3 too small to yield statistically significant
correlations.
108
Individual groups, only the means of the Trials and Errors
scores were perfectly correlated, i.e., only these two
scores assigned exactly the same rank order to each of the
six treatment groups. For the individual groups only one
other correlation was highly significant. That was the
correlation obtained between the ranks assigned to the
groups by the medians of the Trials and Time scores. For
the composite groups, on the other hand, the means gave
exactly the same rank order for all three scores.
On the basis of the findings summarized in Tables
5 and 7 and from the nature of inter-correlations of the
three acquisition scores, it was evident that the best
measure for use in computations requiring the ordinal
placement of the treatment groups would be the Total Trials
scores.
Analytic Findings; Hypothesis-testing
Statistics
The general hypothesis underlying the present study
was that an inverse relationship would be obtained between
the amount of learning required to perceive a total visual
form and the number of component elements of that form
which had been previously experienced. From the design of
the experiment, however, it was not possible to gather
information in a fashion which would permit the direct test
of this hypothesis. It was, therefore, broken down into
109
two subsidiary hypotheses, each of which could find em
pirical support and which could be subjected to appropriate
test. Each of these is considered in turn.
The First Experimental Hypothesis
Hypothesis I predicted that: Significant differ
ences would be obtained in the rank totals of the six
individual treatment groups. Stated in the form of a null
hypothesis, it would be predicted that none of the six
treatment conditions would have a differential effect on
the rank totals obtained. Both the experimental and the
null hypothesis should hold for any or all of the three
acquisition scores adopted as measures of discrimination-
learning.
Friedman's Two-Way Analysis of Variance by Ranks was
selected as the most appropriate and direct statistical
test of the null hypothesis for two main reasons (Siegel,
1956).
First, the manner of computation involved in this
test teased out the litter-effects which were shown to be
significant in the metric analysis of variance in a two-way
classification without replications. Because no replica
tions were possible in the experiment, however, the metric
analysis did not permit removing litter-effects so that
possible effects of the stimulus conditions on "equated"
litters could be determined. The Friedman test, on the
110
other hand, automatically "equates" all members from a
single litter because the ranking is done on each litter
across all the stimulus (or treatment) conditions. Thus,
if all had gone exactly as predicted from theory, the mem
bers of each litter would have been ranked from poorest to
best performance, in the following order: Dark, Non-
Pattemed Light, Line, Angle, Triangle, and Normal.
Second, while it might be argued that the measure
ment of trials, errors, and time is done in equal intervals
it would probably be incorrect to assume that equal inter
vals in these measures reflect equal behavioral or learning
intervals. Among other assumptions, the metric analysis of
variance requires measurement in an interval scale. The
Friedman deals with ordinal measures and probably provided
a more accurate interpretation of the data on this account.
Its use also obviated the necessity of making assumptions
of normal distribution which would have been particularly
questionable since the skewing of the Non-Pattemed Light
group's scores demonstrably altered its rank position when
ranking was done on the basis of means rather than medians.
Table 8 shows the rank totals of the three acquisi
tion scores for all six of the individual treatment groups.
This is followed by Table 9 which shows the results of
Friedman1s two-way analysis of variance of these rank
totals.
Inspection of Table 8 shows that the ranks of all
Ill
TABLE 8
RANK TOTALS OF THREE ACQUISITION SCORES
FOR INDIVIDUAL TREATMENT GROUPS
Rank
totals*
— t r * a t m e t i i k r o u p 8
Dark
Nfon-
Pattemed Line Angle Triangle Normal
Trials 30.5 30.5 33.5 41.0 33.5 20.0
Errors 32.0 30.0 32.0 41.0 33.5 20.5
Time 27.0 31.0 31.0 41.0 35.0 24.0
*Low ranks equal high performance or ease of learning.
112
but two of the groups were fairly homogeneous. The Angle
group, however, seems to give a consistently poor perform
ance on all three acquisition scores. The Normal group, on
the other hand, appears to have been consistently superior,
although this superiority is least marked on the Time
score.
These differences, however, were not sufficient to
give a significant X^r, the statistic derived from the
Friedman test, as may be seen by Table 9.
It is apparent from Table 9 that the groups did not
differ sufficiently from one another, when considered all
together, to justify rejection of the null hypothesis. It
must, therefore, be concluded that the experimental find
ings do not lend support to Hypothesis I.
Since no support for the primary hypothesis was
found, it was determined to test the alternate of this
hypothesis with the coarser groupings of the independent
variable. Alternate Hypothesis I predicted that: Signif
icant differences would be obtained in the rank totals of
the three composite treatment groups. Stated in the form
of a null hypothesis, it would be predicted that none of
the three composite treatment conditions would have a
differential effect on the rank totals obtained. Again,
both the experimental and the null hypotheses should hold
for any or all three of the acquisition scores. Table 10
gives the rank totals of the three acquisition scores for
113
TABLE 9
TWO-WAY ANALYSIS OF VARIANCE BY RANKS
FOR INDIVIDUAL TREATMENT GROUPS
Acquisition
score df
Sum of squared
rank totals x2r
Approx.
P
Trials 5 6186.0 7.34 0.20
Errors 5 6171.5 6.88 0.25
Time 5 6133.0 5.70 0.60
Probabilities derived from table of Chi-square.
114
TABLE 10
RANK TOTALS OF THREE ACQUISITION SCORES
FOR COMPOSITE TREATMENT GROUPS
T r e a t m e n t g r o u p s
Rank Dark and Line and Triangle and
totals non-patterned angle normal
Trials
Errors
Time
13.5
15.0
16.0
23.0
24.0
22.0
17.5
15.0
16.0
115
the composite groups.
It would appear from Table 10 that the effect of
combining the groups was to minimize the differences among
them. A general "flattening" of the distribution scores
seems to have occurred and this apparent effect is borne
out by the two-way analysis of variance by ranks summarized
in Table 11.
Since none of the X^r statistics for these composite
groups reached an acceptably significant confidence point,
it was concluded that the data lent no support for Alter
nate Hypothesis I. The null hypothesis that none of the
composite treatment conditions would have a differential
effect on the rank totals of the groups was accepted. How
ever a glance at Table 11 will show that, whatever variance
is indicated in the X^r is accounted for by the relatively
poor performance of the Line and Angle group. Only in the
Trials scoring do the rank totals for the other two groups
differ. In using Errors as the measure, where an almost
significant probability level is achieved, the variance is
entirely accounted for by the poor performance of this
group which, according to theory, should have scored in an
intermediate position between the other two.
The Second Experimental Hypothesis
From a statistical point of view it would appear
that no test of the second experimental hypothesis would be
116
TABLE 11
TWO-WAY ANALYSIS OF VARIANCE BY RANKS
FOR COMPOSITE TREATMENT GROUPS
Acquisition Sum of squared 2
score rank totals X R P
Trials 1017.5 5.00 .088
Errors 1026.0 6.00 .057
Time 996.0 2.67 .328
P derived from the ta^le of exact probabilities associ
ated with values of X ^ when N ■ 9 and k - 3.
117
justified, since it could yield no significantly supportive
results. Hypothesis II predicted that: 1£ significant
differences in rank totals occurred, they would be such
that further analysis would reveal that the groups would
be ranked on the basis of low to high scores on the
measures of discrimination-learning in an inverse order to
the amount of their prior visual experience. Inasmuch as
the Friedman test revealed no significant differences in
rank totals on any of the three acquisition scores, it
would seem that this hypothesis as well as its alternate
statement for the composite groups must be automatically
rejected and the null hypotheses accepted. In other words,
the sequence of ranks based on low to high scores of any or
all of the three measures of discrimination-learning were
not in an inverse order to the amount of prior visual expe
rience. That this negative statement is valid may be seen
as readily from casual inspection of Table 8 as from the
analysis of variance. It shows that on all three acquisi
tion scores the Angle group consistently gave the poorest
performance, whereas, in terms of its prior visual experi
ence, it should lie midway on the scale. Even if this
finding should not prove to be statistically significant,
the ordinal position of this group alone is sufficient to
reject the hypothesis of inverse relationship as stated.
The same is true of the position of the combined Line and
Angle group on the composite measures. Its position as
118
lowest group, revealed in Table 10, is contrary to its
predicted position which should have fallen midway between
the Dark and Non-Patterned groups and the combined Triangle
and Normal group for the alternate of Hypothesis II to have
been upheld.
From a psychological point of view, however, and
because it seemed to offer the possibility of generating
further experimental hypotheses, it seemed important to
carry the analysis of the experimental results further.
The over-all lack of significant differences in the rank
totals and the manifest departure of ordinal position from
that predicted made it necessary to reject both experi
mental hypotheses and their alternates. But it did not
obviate the possibility that some significant differences
could still be found in making comparisons of the various
treatment groups. It seemed conceivable, for example, that
the Normal groups might differ significantly from each of
the other treatment groups. Such a finding would have
important theoretical meaning even though it would not lend
support to the original hypotheses. Additional statistical
analysis of the data, therefore, seemed to belong more
properly in a portion of the thesis devoted to interpreta
tion and discussion of the theoretical implications of this
study and its indications of needed further research. They
are, therefore, dealt with in the chapter which follows.
CHAPTER V
DISCUSSION AND INTERPRETATION
-OF THE FINDINGS OF THE STUDY
The findings of the present study are clearly non-
supportive of the experimental hypotheses deduced from
Hebb's theory of perceptual development. If the inter
pretation of that theory has been correct and if the
assumptions and presuppositions underlying the experimental
design have been accepted, then Hebb1s theory that the per
ception of the identity of a total form is the result of a
gradual, summative integration of experience with its com
ponent parts must be regarded in error. Neither the pre
diction of significant differences in the rank totals of
the treatment groups nor the prediction of the rank order
in which these groups would fall was supported by the data.
Accordingly, it must be recognized that no inverse rela
tionship between prior part-experiences and later ability
to learn to perceive a total form has been demonstrated.
No hypothesis and certainly no theory, however, can
be categorically rejected merely on the basis of one set of
negative findings. Error may lie, not in the theoretical
statements, but in the experimental test. There is always
119
120
the possibility that a particular experiment may represent
an erroneous interpretation of the theory, just as there is
always the danger that other assumptions and presupposi
tions underlying a particular experiment may be unjusti
fied. Attention should, therefore, be turned first to a
critical evaluation of the experimental test before conclu
sions are drawn with respect to the theory. What are some
of the most probable sources of experimental error?
The most probable source of error would not seem to
lie in the execution of the details of procedure which were
carried out in the laboratory with scrupulous care. While
some error in the data is humanly unavoidable, any really
critical error is more likely to lie in the choice of the
procedures themselves. This study would be entirely inval
idated, for example, if it were shown that the operational
definitions of its independent and dependent variables do
not accurately reflect the variables which Hebb theoretic
ally places in an antecedent-consequent relationship.
Assessment of the Experiment and Its
Relationship to Hebb*s Hypothesis
Critical Evaluation of the
Independent Variable
On this basis, one of the criticisms that is likely
to be leveled against this study is that its operational
definition of part-experiences did not truly reflect Hebb's
121
meaning in his use of the term perceptual element. While
he did make the statement that "lines and angles . . . can
be treated as perceptual elements . . . and [are] likely to
be learned before more complex patterns are" (Hebb, 1950),*
he did not indicate anywhere in his text that these part-
experiences should be expected to have the same effect when
they are removed from the context of the total form. From
Hebb1s account of the development of a phase-sequence, it
is possible to conclude that while lines and angles are the
elements which must be learned, they are learned only as
segregated parts of a unified whole. In other words, be
fore an animal is able to learn to perceive a triangle as a
total form, it may have to move its eyes back and forth
along one of its sides, visually traverse the angles, and
repeat this process often enough so that, with the aid of
reverberatory neural circuits, eventually the triangle may
be seen as an entity. But it may be argued that this occuzs
normally only in the presence of the whole triangle and
that it need not necessarily follow that the same effect
would be achieved by a view limited to a segregated portion
of the triangle.
To this criticism, the performance of the Triangle
group of animals makes a powerful negative statement. Even
if it is conceded that the Non-Pattemed, Line, and Angle
^-Italics mine, FBL.
groups did not properly reflect Hebb's meaning of a percep
tual element, no such criticism can validly be directed
against the Triangle group. Whatever the elements that may
be postulated, they were all present and organized in this
stimulus pattern. Accordingly, the Triangle group's per
formance should have been significantly better than the
performance of the Dark-reared animals. It was not. Nor
was its performance significantly better than that of the
Non-Pattemed group. Since the visual experience of this
group did not Include exposure to any lines or angles in or
out of context, its performance, like the Dark group,
should have been inferior to that of the Triangle group
even if the most literal definition of Hebb's perceptual
elements were used. It must, therefore, be concluded that
even if the segregation of the part-experiences from the
total form represents an incorrect Interpretation of a
perceptual element, the findings of the study remain non-
supportive of Hebb's theory.
Parenthetically, it may be noted that this sort of
criticism leveled against the experiment would be a singu
larly poor defense of Hebb's theory. For, in his discus
sion of perceptual elements, he clearly indicates that they
are not merely constructs into which a total form may be
analyzed, but that they are the sufficient (and perhaps
necessary) antecedent experiences from which ability to
perceive a total form is synthesized. His whole
123
physiological model is designed Co account for the manner
in which perception is built up by the neural connection of
experiential elemental units. To claim the theory's
validity by assigning it only analytic properties would be
to win a Phyrric victory. The basic approach of the theory
would be destroyed in the process. For the theory to re
tain its explanatory power, the perceptual elements should
be capable of denotation isolated from the context in which
they appear when perception is fully developed.
Another criticism that may be leveled against the
independent variable is this. Despite the logical validity
in Hebbian terms of the segregation of the line and two-
lines-and-included-angles as part-experiences for later
perception of a triangle, they may not hold for the rat.
It is true that the aforementioned stimulus conditions
represent lines and angles in the same orientation as they
would appear in an upright equilateral triangle. This
characteristic should qualify them to meet Hebb's defini
tion of perceptual elements. However, it has been shown by
studies of pattem-equivalence that the rat tends to view
only the bottom portions of both stimuli when making a dis
crimination between a triangle and a square in a jumping
stand apparatus (Lashley, 1938). The same characteristic
is likely to hold for its discrimination of a triangle from
a circle in the apparatus used in the present study. If
this were the case, then the rat would not really be making
124
a distinction between an entire triangle and an entire
circle but between a straight line and a curved one.
Furthermore, since the straight line would be in a hori
zontal position, neither the line nor the two-lines-and-
ineluded-angle stimuli would have provided prior experience
with a line in this orientation.
While such an argument might disqualify the inde
pendent variable from fulfilling the requirements of a
scale of part-experiences, it still does not render the
results of the experiment supportive of Hebb's fundamental
thesis. The performance of the triangle group continues to
make a negative statement. If prior experience with either
parts or wholes has a facilitative effect on later percep
tion and if perception must be learned, then the Triangle
group should have performed significantly better than all
the other stimulus groups (barring, perhaps, the Normal
group). If only the bottom segment of the patterns was
used in the rat's discrimination of the triangle from the
circle, then none of the other groups had any relevant
visual experience prior to discrimination-training. The
Triangle group, however, should already have learned to
perceive the relevant portions of the triangular stimulus
and consequently should have been able to learn to dis
tinguish it from a circle more quickly than the other
groups. It did not.
This finding is non-supportive of the most basic of
125
the Hebbian tenets. For, if we strip his theory for the
moment of the question of parts and wholes or analysis and
synthesis, its fundamental position is that perception of
the identity of a total configuration must be learned.
Whether it is learned from prior experience with segregated
parts or parts analyzed during visual encounter with the
entire form, whether it results from a summative process or
not, whether or not the rat learns to perceive the entire
triangle and circle, are all of great theoretical signifi
cance, but they are not nearly so basic to Hebb's position
as it is to recognize that perception is in some way
learned. If it is learned, then some prior experience with
the to-be-perceived entity should facilitate its consequent
perception. If discrimination-learning is an adequate test
of perception, then the Triangle group, regardless of
whether it actually perceived a whole triangle or not,
should have turned in the superior performance on this task
if Hebb's basic position is valid.
Critical Evaluation of the
Dependent Variable
"If discrimination-learning is an adequate test of
perception"--ay, there's the rub! It has already been
pointed out that discrimination-learning confounds at least
three variables, among which the ability to perceive differ
ences is only one. This recognition cannot be lightly
dismissed and its implications will be further discussed in
a subsequent portion of this paper. However, considered as
a suitable dependent variable for a test of Hebb's theory,
it cannot be legitimately criticized. For even if we con
cede the weaknesses of discrimination-learning as a measure
of perceptual ability, Hebb does not do so. His perceptual
theory is, in point of fact, an attempt to explain certain
data derived from discrimination-learning experiments. In
his attempt to show that "quite simple diagrams are not
perceived directly as distinctive wholes . . . [and] the
perception of identity depends on a series of excitations
from the parts of the stimulating diagram," Hebb (1950)
acknowledges that the work of Senden (1932) and Riesen
(1947) is fundamental to his arguments. These two works to
which Hebb acknowledges his indebtedness both involve tests
of discrimination-learning. The first studies humans who
had congenital cataracts surgically removed and who thereby
achieved vision for the first time during adulthood
(Senden, 1932). Many of these patients could not use
purely visual cues to tell the difference between a tri
angle and a square when diagrams of them were placed side
by side. They could immediately distinguish them, however,
by touch or by counting corners. Riesen1s study (1947)
dealt with chimpanzees reared in darkness. These animals
were inferior to normally-reared controls, among other ways^
in their ability to learn to solve problems which were
dependent on making discriminations between forms. Hebb's
127
own later re-evaluation of his early study with rats, which
compared the performance of his dark-reared animals with
Lashley's normally-reared ones, dealt with their ability to
distinguish between horizontal and vertical striations in a
9
jumping stand apparatus (Hebb, 1949). All the studies
cited in the review of the literature which purport to stem
from Hebb's theory of perceptual development, and which
have been regarded as lending it support, use some form of
discrimination-learning as an index of ability to perceive.
Throughout these studies, as in Hebb's book, there is the
implicit assumption that differences in discrimination-
learning scores reflect differences in ability to perceive.
Clearly, then, any criticism of the use of the
dependent variable in this study as a measure of animal
perception is unjustified as long as that criticism pur
ports to remain within the framework of the Hebbian con
ception of perceptual ability. Once that framework is
discarded, there are cogent arguments against the use of
discrimination-learning as an index of perceptual ability.
But these arguments are more relevant to an evaluation of
^Hebb's actual laboratory investigation used 18
dark-reared animals for a variety of tests of perceptual
development. It was reported in the J. Genet. Psychol.,
1937, 51, 101-126. Comparison to Lashley's normal rats
must have been made subsequent to the publication. While
the comparison is reported in Organization of Behavior
which was published in 1949, no study other than the 1937
one is cited as a reference by Hebb.
128
the theory itself than to any experiment designed within
its structure.
If it is granted that the choice of the independent
and dependent variables in the study are valid empirical
interpretations of Hebb's theoretical antecedent-consequent
variables, then the negative findings of the study would
seem to justify a negatively critical appraisal of the
theory which failed to make an accurate prediction of the
experimental results.
Any good scientific theory "goes beyond the data."
One of its major functions is to fill in the gaps in em
pirical observation with plausible inference. These infer
ences in turn stimulate further observation and further
inferences and the spiral of scientific knowledge ascends.
The observations made in the present experiment
would seem to indicate that Hebb's inferences can no longer
be considered plausible. To contribute to the ascending
spiral of scientific knowledge, the next order of business
would call for a reappraisal of the data and the formula
tion of further plausible inferences to account for it.
Reappraisal of the Data and Formulation
of Alternative Hypotheses
When the present experiment was designed it was
recognized that only four stimulus groups would have been
logically sufficient for an appraisal of Hebb's theory of
129
perceptual development. The four required groups, it was
suggested, were the Non-Patterned, Line, Angle, and Tri
angle groups. Only in these groups was the visual stimulus
definable in Hebb's terms as containing the perceptual ele
ments entering into the final ability to perceive a total
form. The decision to include the Dark-reared and Nor
mally-reared groups was made to serve a threefold purpose.
First, it was considered important to replicate the
kind of design on which Hebb's theory had been founded.
His theory was formulated in an attempt to account for the
differences in the discrimination-learning ability of dark-
reared and normally-reared animals. The gap in the per
formance of these two groups was his primary empirical
datum. The importance attributed to lines and angles in
perceptual learning was his explanation of how this gap was
filled. Therefore, even though the four stimulus groups
would have been sufficient to provide support or non
support for his explanatory thesis, it was felt that,
should the study prove non-supportive, it would make a more
powerful negative statement if it could be demonstrated
clearly that the primary fact on which Hebb's explanation
had been built still held, and, by the same token, it would
lend more powerful support if predictions from the theory
were borne out.
Second, there were certain novel conditions intro
duced by the experimental design, and the inclusion of the
130
Dark and Normal groups, it was felt, would not only be a
replication of the work of others but would also provide a
check on the technique of the present experiment. Since
the difference in the performance of these two groups has
been demonstrated often enough to constitute a genuine,
reproducible behavioral phenomenon, it was felt that unless
the phenomenon could be shown to occur under the particular
conditions of the present experiment, the technical ade
quacy of the experimental design might be brought into
question.
Third, it was felt that the inclusion of these two
groups would provide a base from which alternative hypoth
eses might be generated, if the study should prove non-
supportive of the experimental hypotheses.
Since the study proved to be non-supportive, it
seemed important to see if the Dark and Normal groups did
actually differ under these experimental conditions. Also,
despite the obtained lack of difference in the over-all
rank totals of the groups which did not provide statistical
justification for it, psychological curiosity impelled
further analysis of the data to see whether significant
differences would be found, not only between the Dark and
Normal groups, but also in comparisons of all the other
131
3
groups with each other.
Reassessment of Data for Individual Groups
As a quick check to determine whether such differ
ences were present in the data, Wilcoxin's Matched-Pairs
Signed-Ranks Test was selected as an appropriate statis
tical tool (Siegel, 1956). Like Friedman's two-way
analysis of variance, it deals with ranks. But, because it
takes into account the magnitude of the differences between
pairs, it offered the possibility of showing up more
clearly than the Friedman test any differences that might
have existed between the members of a single litter as
signed to different stimulus groups. Only the Trials and
Error scores were used in this additional analysis because
the Time scores manifestly offered the poorest probability
of detecting significant differences.
The values of T, the statistic obtained from Wil
coxin's test of the significance of differences of ranked
signs obtained by the subjects when matched for litter
membership are given in Table 12 for both the Trials and
^The findings of the study were subjected to this
additional scrutiny with the full recognition that any re
appraisal of the data at this point would not have any
relevance to the test of the original hypotheses underlying
the study. The dangers involved in post-hoc evaluation of
data were also recognized and accounts for the fact that
these additional analyses are not included in the chapter
on results. Their limited, speculative value, it was felt,
placed them more properly in a chapter devoted to an inter
pretation and discussion of the findings.
132
TABLE 12
VALUES OF T FROM WILCOXIN'S MATCHED-PAIRS SIGNED-RANKS
TEST OF DIFFERENCES BETWEEN TREATMENT GROUPS
T r e a t m e n t gr ou p s
Treatment
groups
Non-
Fattemed Line Angle Triangle Normal
TRIALS
Dark 17.5 12.0d 11.0 22.5 8.0a
Non-Patterned
-
15.0C 10.0 22.0 11.0L
Line
- -
17.0 21.5 5.0b»c
Angle
- - -
16.5 7.5f
Triangle
— • •
3.5b»c
ERRORS
Dark 20.0 19.5 13.0 21.0 6.5a
Non-Patterned
-
22.0 11.5 14.0 14.5.
Line
- -
17.0 20.5 5.0b
Angle
- - -
16.0 7.Of
Triangle
“ *
4.0b»c
NOTE: All significance levels were computed for a
one-tall test since prediction of the direction of the
difference was made. The superscripts in the table have
the following meaning with respect to significance levels:
a - < .05; b - <.025.
N - 9 except in those Instances noted by the follow
ing superscripts: c ■ an N of 8; d - an N of 7. Where N's
are less than 9 they are due to zero difference scores in
one or more matched pairs.
133
Errors scores.
As may be seen from Table 12, the data uphold the
observations of significant differences in discrimination-
learning of Dark- and Normally-reared groups on which Hebb
placed so much importance. However, it is apparent that
the Normal group is not only superior to the Dark-reared
animals, it is also superior in its performance to all the
other groups with the exception of the Non-Pattemed group.
It seems worthwhile to comment on the fact that the Non-
Pattemed group's performance was not also significantly
inferior to the Normal group because of a single extreme
difference score. In one litter the Non-Pattemed animal
learned in eighty trials with only twenty-five errors,
while his normal litter-mate required 240 trials and made
ninety errors. The signed-rank assigned to this pair
accounts for the high numerical value of T. It seems
highly likely that a replication of the study would find
the Normal group's performance superior to all the other
groups.
If, as seems plausible, the Normal group is superior
to all the other groups, how can we account for this
difference?
The answer to this question is by no means apparent
from the data nor derivable from Hebb's theory as stated.
Certain factors that might account for the difference, how
ever, can be ruled out. The precautions exercised to
134
equate the dimensions of living quarters, for example,
would seem to rule out major differences in motor experi
ence. Handling of the animals was no different for the
Normal group so this may be ruled out as a factor. While
the metric analysis of variance indicated that litter-
membership is more significantly related to discrimination-
learning than prior visual experience, the use of the
split-litter technique undoubtedly minimized the differ
ences that might be attributed to this factor when stimulus
groups are compared only for rank positions within their
litter grouping. What plausible inferences remain? Out of
the infinity of possible explanations, several could be
superficially tested against existing data.
First, there was the possibility that the Normal
group had developed a different type of learning process
than the other groups. It was thought that one way in
which this difference might conceivably be displayed was
through a sudden as opposed to a gradual drop in a curve
of errors just prior to reaching criterion, an insight as
opposed to a trial-and-error type of curve. Accordingly,
a backward learning curve was constructed for each of the
six stimulus groups based on the mean error of the group
in its performance of each day's block of ten learning
trials. Table 13 is a tabular representation of these
curves and shows the mean daily error for each group for
seven days. This includes the five days of training just
135
TABLE 13
BACKWARD LEARNING SCORES FOR ALL GROUPS USING
MEAN DAILY ERROR PER BLOCK OF TEN TRIALS
Treatment
group
Mean d a i 1 y e r r o r
D* D-l D-2 D-3 D-4 D-5 D-6
Dark 0.9 0.9 3.9 2.9 3.6 4.1 3.6
Non-patterned 0.5 0.9 2.4 2.4 3.2 3.5 4.0
Line 1.0 1.3 3.8 3.8 3.0 3.7 4.7
Angle 1.1 0.9 4.0 3.1 3.9 3.7 4.8
Triangle 0.5 1.1 4.1 3.9 3.4 3.8 4.5
Normal 0.6 1.0 3.3 3.9 4.0 4.1 4.3
*D represents the last of the two days on which criterion
for learning was attained. Each of the other columns
represents the appropriately numbered day preceding the
criterion day. Both D and D-l were required to meet the
learning criterion for 18 correct responses out of 20
trials since only 10 trials were given on each day.
136
prior to reaching criterion as well as the two days needed
to meet criterion of eighteen correct responses (a maximum
of two errors) in twenty consecutive trials.
As is shown in Table 13, all the groups seem to
follow essentially the same learning pattern. There ap
pears to be a rather sudden sharp drop in the number of
errors on D-l, which was the first of the two criterion
days. At this point the animals seem to have "caught on"
to the solution to the problem. Prior to these days, the
errors fluctuate, sometimes showing a slight increase,
sometimes showing a slight decrease, but always remaining
well within the chance expectancies.
It may thus be seen that, although the Normal group
learned significantly faster than the other groups, all
groups, when they learn, appear to learn by the same gen
eral process. The gestaltists would call it Insightful
learning or learning through the restructuring of the
field. This is an interesting finding in itself and might
be worth checking into in a study designed specifically for
the purpose. But it does not contribute much to an under
standing of the superiority of the performance of the
Normal group. Differences in mode of learning would not
seem to be an appropriate explanatory hypothesis.
137
Reassessment of Data for Composite Groups
One finding of the study which seemed of interest
was the poor performance of the Line, Angle, and Triangle
groups. While these groups were not statistically dif
ferent than the other groups in over-all analyses of their
rank positions, they were consistently poorer on nearly all
measures, as reference to Table 4 will reveal. Since this
finding was completely contrary to what would be predicted
from Hebblan theory, it led to consideration of some
alternative explanation.
One possible explanation was related to the recently
published volume by Berlyne (1960) which discusses the
motivating effects of novelty. If novelty of a visual
stimulus can be defined in Berlyne's terms as being in
versely related to (1) how often relevantly similar pat
terns have been experienced before, (2) how recently they
have been experienced, and (3) how similar they have been
to the stimulus in question, then the Line, Angle, and
Triangle groups should certainly have found the triangle
in the discrimination apparatus less "novel" than did the
other three groups. If novelty is intrinsically interest
ing or in some way motivating, then the Dark, Non-Pattemed
and Normal groups might conceivably have been more highly
motivated than the other groups to learn to solve the
discrimination problem. While this cannot be offered as
an explanation of the superior performance of the Normal
138
group, It might help to explain the consistently poor per
formance of the three groups mentioned.
This possibility led to a further evaluation of
data to test a post-hoc hypothesis. It was thought that
one way of checking into the motivational impact of
novelty would be to investigate the first choice of stimu
lus by the subjects on their first exposure to the tri
angle and the circle in the discrimination apparatus. If
novelty were indeed the basis on which selection was made,
it seemed likely that the first choice would be one which
was most dissimilar to the stimulus conditions under which
the animals had been reared.
Accordingly, the subjects were divided into two
major groups. Group I included the Normal, Non-Pattemed,
and Dark groups. For all these groups the discrimination
situation was novel. For the Normal group it represented
a complete switch from their regular living conditions.
But for all of these groups it represented a situation in
which two stimuli were presented, neither of which had
been seen before. It was, therefore, postulated that the
members of this group could be expected to choose either
the triangle or circle as a first choice equally often
since both were equally novel. On the other hand, Group
II, consisting of the Line, Angle, and Triangle groups,
might be thought of as having seen, all their lives since
weaning, a stimulus similar in some degree to a triangle.
139
It would seem reasonable to predict that this group would
prefer the circle and therefore choose it more often than
the triangle if novelty were the primary determinant of
the first adient response.^
Table 14 shows the results of this post-hoc analy
sis of the re-grouped data.
Analysis of these findings show them to be exactly
the opposite of what was predicted on the basis of a
novelty hypothesis. Group I showed a significant prefer
ence for the triangle, while Group II showed no significant
preference for either stimulus. In Group I, twenty out of
the twenty-seven animals made their first response to the
triangle; in Group II, seventeen out of twenty-seven
responded first with a choice of the triangle.
It would seem unlikely that, on the basis of these
findings, the poor performance of the Line, Angle, and
Triangle groups in the original experiment could be attrib
uted to the lack of novelty in the discrimination task.
Instead of explaining the results of the original experi
ment, this finding only serves to point up a phenomenon
which in itself calls for explanation. What accounts for
^These hypotheses pose a different question than
the one to which Hebb addressed himself and imply an
opposite assumption. Since first choice is used as a
measure of novelty, there is a presupposition that the
first response in the discrimination apparatus is not only
toward a preferred stimulus, but also one which is per
ceived on first exposure and is, therefore, unlearned.
140
TABLE 14
CHOICE OF STIMULUS IN FIRST DISCRIMINATION-
LEARNING TRIAL
Group N
First
Triangle
response
Circle Chi-square P
I 27 20 7 6.26 •
o
N>
II 27
11
10 1.81 NS
Total 54 37 17 7.41 .01
Chi-square computed from formula t - f^ - f£
to test whether the frequencies in either of two groups
depart from equality, the expection under the null
hypothesis.
141
the rather overwhelming preference for the triangle on
first exposure to the discrimination stimuli when the
total number of subjects is considered as a whole? It
would seem that, regardless of prior visual experience or
the lack of it, a decided and significant number of first
responses were made toward the triangle.
A digression to "explain" one proposed "explana
tion" of the data is not really out of order at this
point. The finding that the triangle was responded to
more often than the circle at first exposure to both
stimuli in the discrimination apparatus could lead to seri
ous doubts about the experimental procedures employed.
One of the first suspicions aroused is likely to be that
the doors were in some way identifiable and provided a
positive attraction to the triangle door which had nothing
to do with the visual properties of the triangle at all.
The only answer to this possibility is that every precau
tion was exercised to eliminate preferences. Brightness
of the two stimuli was carefully equated; doors were
checked before each animal was placed in the apparatus to
make sure that no extraneous light entered when the stimu
lus panel was in viewing position; both doors were equally
baited with wet mash; animals were placed in the apparatus
as nearly equidistant from both doors as could be gauged;
and so forth.
However, despite all these precautions, some
142
unknown non-chance factor seems to have been operating in
the choice of the first stimulus. Some assurance was
needed that despite its presence it did not have a biasing
effect on the acquisition scores used in the test of the
study's major hypotheses. It was, therefore, determined
whether any significant differences could be obtained in
the mean number of trials to criterion needed by those
animals who first chose the triangle (the "correct" re
sponse) and those who first chose the circle. The mean
trials-to-criterion score for the thirty-seven subjects
who chose the triangle was 143.5. That for the seventeen
subjects who chose the circle was 169.3. The test of the
significance of the difference of these means yielded a t
of 1.20 which was not significant for the appropriate de
grees of freedom. Nor was either mean significantly
different from the grand mean of 151.6 obtained by the
entire group of fifty-four animals. Therefore, it would
not seem that the preference of one stimulus over the
other in the first choice had any prejudicial influence on
the acquisition scores for the solution of the discrimina
tion problem. Consequently they did not bias the data on
which the hypotheses of the study were tested.
The phenomenon of the first-choice preference for
the triangle over the circle for all subjects, however,
remains to be explained since it seems unlikely that the
explanation lies in faulty apparatus or in the intrinsic
143
novelty of the triangle as a stimulus.
One other aspect of the experimental procedure
might be held accountable for the phenomenon. Since the
stimuli were presented in a prearranged but random order
and it was not foreseen that first response might have any
special meaning, there was no apparent reason to vary the
presentation position of the stimuli on first trial,
especially since an alternation procedure had been used in
the pretraining of all subjects to the apparatus. It
therefore happens that all animals had the triangle pre
sented to their right on the first trial. Position
preferences are not unknown in rats. They may be thought
of as similar to handedness in humans. It may simply be
that most rats prefer to turn to the right at a choice
point, just as most humans prefer to reach for an object
with their right hands. Such a position preference might
inadvertently have influenced the first choice. If this
were the reason, then no ability to perceive differences
between the visual stimuli could be inferred. A decision
as to the meaning of the excessive first choice of the
triangle would have to await further investigation of the
position preference phenomenon.
Learning the alternation pattern prior to discrimi
nation training might have minimized position preferences
but it also might have introduced other complicating
factors. As Brogden (1951) has pointed out,
144
"Discrimination learning based on simultaneous presentation
of stimuli involves heterogeneous reinforcement of re
sponse." Any reward obtained by the animal can be regarded
as reinforcement of the animal's operating "hypothesis."
If an animal had adopted an alternation principle it could
have received partial reinforcement and this might compli
cate its rejection of the alternation pattern and the
adoption of the "correct" visual hypothesis for the solu
tion of the discrimination problem. If, as seems reason
able, animals raised in an impoverished visual environment
are less likely to adopt a visual hypothesis than animals
raised with normal visual stimulation, pretraining all of
the animals to the alternation pattern would run the risk
of introducing a bias which might have led to an unjusti
fied acceptance of the experimental hypothesis. Since
that hypothesis stated, in effect, that dark-reared animals
would require a greater number of trials to learn to make a
visual discrimination than normals and the nature of their
rearing predisposed them toward adopting positional cues,
training them to alternate might have reinforced this
tendency to the extent that many trials would have been
used before these animals rejected a position habit and
adopted a visual one, even though they might have been
quite capable of perceiving the visual cues all the time.
It would be an error of logic to conclude that the greater
number of trials needed in this type of instance was
145
attributable exclusively to visual perceptual disability
in the dark-reared rat.
The use of a time criterion for pretraining mini
mized the reinforcement of position habits but still
assured that all subjects had associated a motor habit with
the obtaining of a reward. At the start of discrimination-
training in the present experiment all animals were re
sponding quickly, opening the doors of the apparatus
easily, and receiving at least partial reinforcement for
their efforts.
The pretraining procedure called upon the experi
menter to hold open the correct door during the early
trials if the subject did not respond within specified
time limits. This makes it impossible to determine the
exact number of alternation errors made by the subjects
during pretraining trials. However, in a crude attempt to
determine whether alternation-training did occur in some
subjects to the possible prejudicial evaluation of experi
mental results, a lenient criterion for the learning of
alternation was adopted. The data were reviewed and the
last twenty pretraining trials examined. If only five
errors occurred during these last twenty trials, the
alternation was considered established. Only eleven out
of the fifty-four subjects met this criterion. These
eleven subjects were distributed across all six stimulus
groups. Three fell in the Normal group, one in the
146
Non-Pattemed Light group, and two each in the other three
groups. It would, therefore, appear unlikely that the
experimental conditions were meaningfully related to
learning the alternation habit or that group scores were
unduly affected by it.
Concluding Reappraisal of the Findings
However, it cannot be overlooked that another, and
perhaps very significant, aspect of behavior may be indi
cated by the preferred response to the triangle. It is
within the realm of possibility that the indicated prefer
ence is directly related to the animals' ability to per
ceive the difference between the triangle and circle on
first exposure to it. Hilgard (1951) mentions stimulus
preferences as one of the "natural propensities affecting
experimental design." He indicates that they should be
taken into account. But it is patently impossible to take
them into account in deprivation-type experiments where
the tacit assumption, indeed the basic hypothesis that the
animals must learn to perceive the forms before they can
react differentially to them, makes it illogical to expect
anything but random first responses. When it is demon
strated, however, that animals show a decided preference
for one stimulus over another when these stimuli represent
their very first visual encounter, then it is certainly one
plausible explanation to say that the animals were capable
147
of perceiving a difference between them without needing
to learn to perceive that difference.
Because of the importance of this kind of conclu
sion for the old nativist versus empiricist controversy,
further research to determine whether this preference is a
reproducible and consistent phenomenon would take on con
siderable theoretical significance. The present study,
however, cannot be regarded as offering evidence for a
nativist rather than an empiricist interpretation of the
development of perception.
It has already been stated that the "correctness"
of Hebb's physiological model for the underlying process
in the development of perceptual abilities is not crucial
for the empirical study of its behavioral implications.
The converse of this statement should be recognized as
similarly valid. It is still logically conceivable that
the physiological processes that Hebb postulates do actu
ally occur. While it is now difficult to defend the
thesis that these processes are correlated with the early
experience of lines and angles as perceptual units out of
which visual forms are constructed, it does not eliminate
the possibility that other elements may underlie percep
tion or that prior experience with these other elements is
necessary before a total visual form can be perceived.
Other lines of evidence, for example the work of Chow and
Nissen (1955) on interocular transfer and Forgus' study
148
(1958) on differential sensitivity to differences between
stimuli on successive presentation of training and testing
forms to dark- and normally-reared rats, both indicate
that some sort of learning is indeed a part of the devel
opment of perception and that this learning involves some
modification of central nervous processes. The findings
of the present study which show that the triangle group's
scores were more homogeneous than that of any other
group's may also be interpreted as indicating that some
learning occurred. While the triangle group, as a whole,
did not perform better than the other groups as would be
predicted by theory, the animals in this group showed con
siderably less variability than those in all other groups.
This might certainly be taken as evidence that some modi
fication of these animals had occurred from their prior
commerce with the visually presented triangle.
The problem, however, does not lie so much in
pointing out what the findings of the study do not indi
cate. The responsibility of the investigator calls for an
attempt to give a positive interpretation of experimental
results and to find some logical meaning in the data, even
if it does not support the experimental hypotheses. The
data derived from the present investigation seem to indi
cate that the Normal group had attained through learning
some superiority over all the other stimulus groups which
had been reared under less advantageous visual conditions.
149
It is submitted that this superiority is not neces
sarily due to a superiority in ability to perceive the
difference between a triangle and a circle. A more reason
able conclusion seems to be that the Normal group's prior
experience made it more prone to use available visual cues
in the solution of problems and it just so happened that
the experimenter "rigged" the apparatus so that only
through the adoption of a "visual hypothesis" could the
problem be solved to the experimenter's satisfaction. If
a position-hypothesis had been adopted, perhaps another
group would have demonstrated superiority. This inter
pretation seems to be the same as that offered by Beach
and Jaynes (1954) in their review of the literature on the
effects of infantile experience on adult behavior. They
stated that "... absence of the normal amount of visual
stimulation during the developmental period may result in
inability to respond adaptively to visual cues when such
cues first become available to the individual."
If this interpretation is correct, then the use of
discrimination-training would be a singularly poor means
of measuring perceptual ability, particularly in depriva
tion studies on infra-human species. In this type of
training, the ability to perceive becomes so confounded
with the animal's ability to associate visual stimuli with
the workings of the apparatus and with its possible prefer
ence for non-visual "hypotheses" due to early visual
150
deprivation, that the interpretation of a discrimination-
leaming score becomes difficult at best.
Where the nature of the prior experience is such
that one could reasonably expect differences in the nature
of the "hypotheses" adopted by the experimental subjects
to bias results in favor of the experimental hypothesis
under test, it becomes scientifically indefensible to
attribute differences in these scores to differences in
perceptual ability. As Nissen (1951) has said about more
complex learning problems,
. . . the effective cue cannot be discovered in a
single trial, but only on the basis of repeated
experience. Sometimes the animal is said to be
responding on the basis of a principle of alterna
tion, of oddity, of matching, and so on. In each
case a temporal integration of experience is
involved. It seems reasonable, therefore, that
this conceptual organizing may take up most of the
time used in "learning" a problem involving the
concept, and that its connection to a given overt
response may then occur very rapidly. (Nissen,
1951)
Similar suggestions for simpler learning situations have
also been made by Guthrie (1935), Lashley (1929), and
Krechevsky (1932).
It might be interesting to determine empirically
whether animals raised in a normal visual environment but
one which restricted locomotion would learn to make a
visual discrimination more rapidly than the ordinary cage-
reared rat. Restriction of movement might result in a
lowered tendency to adopt position cues and an even greater
151
reliance on vision than that shown by the usual laboratory
animal. Such a finding would, of course, bolster the
suggested interpretation of the data derived from the
present study.
Since such bolstering information is not presently
available, some comfort may be drawn from the fact that
earlier studies which were thought to support Hebb's
hypothesis can be interpreted from the present point of
view with equally acceptable logic. Forgus' experiment
(1954) on visual and proprioceptive deprivation did indeed
indicate that visual encounter with the three-dimensional
stimuli was the crucial factor in their later recognition
in the discrimination apparatus. His data, however, need
not be interpreted to mean that perceptual learning had
occurred during this early visual encounter. It would be
equally supportive of an explanation based on the hypothe
sis that rats deprived of visual experience are simply
less likely to utilize perceived visual cues in the solu
tion of problems. This same interpretation would fit the
data derived from Hebb's study (1937), from Hymovitch's
study (1952), the series of studies from Cornell, and all
the other studies derived from and considered supportive
of Hebbian theory, because they showed a significant
difference in the discrimination-learning scores of dark-
reared and normally-reared rats.
Moreover, this new interpretation also fits in well
152
with other lines of investigation which have not thus far
been related to studies on infantile visual deprivation.
For example, it fits in nicely with the findings of
studies designed to test "regression" hypotheses such as
those of Hunt (1941) and Wolf (1943). One of the most
recent of these studies, that by Gauron and Becker (1959),
is particularly relevant. They found that rats deprived
of vision in infancy lose more frequently in competition
for food with rats deprived of hearing when light is used
as a signal, whereas rats deprived of hearing lose to
visually-deprived rats when a buzzer is used as signal.
Apparently these animals are less likely to respond to
cues in the sense modality in which deprivation occurred
when they are under stress, even though the same responses
to the same cues had been made many times under less try
ing circumstances.
If the discrimination procedure is regarded as a
stress situation for the experimental animals, it is rea
sonable to draw an analogy between this study's findings
and that derived from the regression study of Gauron and
Becker.
Furthermore, the emphasis on the differential
adoption of "hypotheses" and their influence on discrimi
nation-learning scores would seem to fit in better with
the data in the present study which showed even greater
discrepancies between litters than between stimulus
153
conditions. They may reflect inherited differences in
intelligence and/or temperament that show up in greater
flexibility in problem-*solving. Conclusions with respect
to this possibility, however, must also await further re
search on animals bred for certain characteristics.
There seems to be little question that this re
interpretation of the meaning of the discrimination score
accounts for more of the data in the present study than
any of the other hypotheses which have been considered as
possible explanations for the findings.
If it should prove to be a valid interpretation of
the meaning of discrimination-learning, it could account
for the lack of predictive accuracy of Hebb's theory of
perceptual development. It would indicate that the theory
suffers not only from the likely possibility that its pro
posed perceptual elements are incorrect, nor even from the
more devastating possibility that perception of differ
ences between forms may be unlearned in the rat; but, even
more basic than either of these, the theory may suffer
from a fundamental error in regarding discrimination-
learning scores as measurements of perception. It may very
well be that any theory based on such an interpretation of
discriroination-leaming could not fail to fall into errors
of both prediction and explanation.
As Spence (1951) has stated, the task of the learn
ing theorist involves "the specification of the
154
experimental variables, environmental and organic, that
determine the observed behavioral changes that occur with
practice." However much we may regret the recognition
that discrimination-learning does not meet the require
ments for a simple, objective means of measuring perceptual
ability in infra-human species, it is incumbent upon us as
scientists to try to carry out our task as learning theo
rists and specify the other variables that may be involved
in a single discrimination score when evidence for their
existence comes to light.
Emphasis on the confounding variables in discrimi
nation -leaming could make a positive contribution to the
field of psychology if it leads to an attempt at more
careful specification of the experiences which influence
animal behavior in the discrimination situation. If this
study were to provide the impetus for such further inves
tigations, it would demonstrate the value of negative
findings to the progress of science. Just as early inves
tigators might have continued to attribute the cause of
malaria to cool, damp, night air had a negative instance
not been demonstrated, so the present investigator might
never have considered seriously the multiplicity of vari
ables in discrimination-learning had the outcome of this
study been in line with theoretically-based predictions
rather than proving non-supportive of them.
CHAPTER VI
SUMMARY, CONCLUSIONS, AND SUGGESTIONS
FOR FURTHER RESEARCH
Summary
The present study represents an attempt to test
empirically a deduction made from one aspect of Hebb's
theory of visual perceptual development. Its major hypoth
esis was that an inverse relationship would be obtained
between the amount of learning required to perceive a
total visual configuration and the number of its component
perceptual elements which had been previously experienced.
To test this hypothesis, an ordinal scale was con
structed in which the scale regions consisted of differing
amounts of the perceptual elements. These were elements
which were theoretically relevant to the later ability to
learn to perceive an equilateral triangle of specified
dimensions. These six regions, ranging from least to
greatest numbers of perceptual elements, were specified as
follows: dark, non-pattemed light, line, two-lines-and-
included-angle, triangle, and normal cage environment.
A treatment-by-litter experimental design was used
155
156
in which albino rats from a single litter were randomly
assigned to the stimulus conditions. Nine litters of six
animals each were distributed in this way, placing nine
animals in each one of the scale's six regions. This pro
cedure constituted the manipulation of the independent
variable since comparisons of the groups thus formed made
it possible to determine the relative effects of differing
amount of prior part-experiences on later ability to learn
to perceive a total form.
The dependent variable, the ability to perceive a
total form, was measured by three discrimination-learning
scores. These were: (1) number of trials, (2) number of
errors, and (3) total time of exposure to the visual
stimuli required for the subjects to learn to solve a
problem which involved the discrimination of an equilateral
triangle from a circle of equal total brightness.
From the basic hypothesis of the study, it was pre
dicted that: (1) significant differences would be obtained
in the rank totals of the six groups on one or more of the
three acquisition measures, and (2) if these differences
were found, rank order placement of the groups based on
low-to-high mean (or median) scores would be in inverse
order to the amount of the perceptual elements with which
they had had prior visual experience. It was, therefore,
predicted that the groups would be ranked in the following
order on the basis of low-to-high scores: normal cage,
157
triangle, angle, line, non-pattemed light, and dark.
However, it was thought that these stimulus groupings
might be too "finely divided" for differences among them
to be evident in the rather crude measure of discrimina
tion-learning. Therefore, if differences were not
apparent in the six stimulus groups, it was thought that
"coarser" stimulus groupings might support essentially the
same predictions. Since Hebb stresses the importance of
lines and angles as perceptual elements in the ability to
perceive total form, these coarser groupings were con
structed by combining the six stimulus groups into three
groups on the following basis. The dark and non-pattemed
light groups were combined since neither group had ever
had visual experience with lines or angles; the line and
two-lines-and-included-angle groups were combined, since
both had had some pattern vision and some experience with
perceptual elements but neither had seen a total triangle
(the complete form); the triangle and normal cage groups
were combined since both had had experience with the total
form, the triangle group most assuredly, and the normal
group by presumption. Making essentially the same predic
tion for these composite groups as had been made for the
individual stimulus groups, it was thought that signifi
cant differences in their rank totals would be found, and
their ranking, based on low-to-high scores, would be in
the following order: (1) normal-plus-triangle; (2) line-
158
plus-two-lines-and-included-angle; and (3) dark-plus-non-
pattemed light.
Analysis of the data, however, did not provide
support for any of these experimental hypotheses. Fried
man's analysis of variance by ranks showed no statistic
ally significant differences in the rank totals of either
the individual stimulus groups or the coarser composite
groups. Consequently there was no statistical justifica
tion to carry the analysis further. By implication, the
hypothesis of significant rank order based on low-to-high
scores would have been unsupported by an additional sta
tistical manipulation of scores. Even casual inspection
of the data, however, showed the futility of such further
analysis since the line and two-lines-and-included-angle
groups simply did not occupy the predicted intermediate
position in the rank ordering of either the individual or
the composite stimulus groups.
Conclusions
The main conclusion to be drawn from the experi
mental findings is very simple and clear. As long as the
measures used in the study are accepted as valid and the
experimental design is considered logically defensible, the
data must be regarded as non-supportive of Hebb's thesis
that an inverse relationship exists between the amount of
learning required to perceive a total visual form and the
159
number of its component perceptual elements to which there
has been prior exposure.
However, from post-hoc analyses of the data, some
suggestive results were found that provide a tentative
indication that discrimination-learning scores may not be
valid measures of perceptual ability at all. Some of the
data gathered in the study can be regarded as indications
that discrimination-learning scores may be measures of a
tendency to utilize discriminable visual cues in the solu
tion of a problem rather than a measure of the ability to
perceive the differences in the simultaneously presented
visual stimuli.
While these post-hoc findings in no way modify the
non-supportive character of the data with regard to the
Hebbian hypothesis, it is plain that if further research
were to validate this tentative interpretation of the
meaning of discrimination-learning scores, much more than
Hebb's hypothesis would be subject to re-evaluation. It
would require a reinterpretation of all the theories and
experimental findings that regard discrimination-learning
scores as an index of perceptual ability. Since such a
reinterpretation might lead to the search for many hitherto
unheeded factors in the discrimination-learning process and
in its development, the tentative interpretation of
discrimination-learning scores indicated by the present
study may well prove to be its major positive contribution.
160
Suggestions for Further Research
Throughout the preceding chapter, many suggestions
for further research were indicated. Some of these are
repeated here in summary form. As with almost any study
of behavior, the foremost suggestion for further research
is replication of the study. The findings of this study
have by no means been established as reproducible behav
ioral phenomena. Until they are so established, any
interpretations or conclusions based on them can only be
offered hesitantly and with many reservations because the
possibility of error is so great. It would seem especial
ly important, from a theoretical point of view, to vali
date through replication the following findings:
1. The superiority of the normal group over other
stimulus groups in discrimination-learning
ability.
2. The "insightful1 1 nature of the backward learn
ing curve for the solution of discrimination-
leaming-type problems.
3. The apparent preference of rats for the triangle
over the circle on first exposure to them.
If these phenomena can be established as regulari
ties in behavior, then a host of other research problems
suggest themselves. Among the most evident of these are:
1. Attempts to identify the factors in the normal
161
group's experience which accounts for its
superior performance over animals reared in
less visually advantageous environments.
2. Determination of the cues most often used by
animals raised under conditions which stimulate
or deprive them in one or more of the sensory
modalities.
3. Examination of the possible role of complexity
in a stimulus to subjects' preference for it.
For example, could the preference for the tri
angle over the circle be related to the greater
amount of "information" in its lines and
angles? If so, would a rectangle or hexagon
induce a greater number of adient responses
than a triangle from an animal who has seen
neither of these stimuli before?
The apparent influence of litter-membership on
discrimination-learning scores, which evidently outweighs
the influence of prior visual experience, would seem to
call for further investigation to find factors that might
conceivably be associated with these effects. For example:
1. What are the differences, if any, in the inter
action effects of "brightness" and/or "dullness"
with prior exposure to varying visual stimulus
conditions and later use of visual cues in the
162
solution of discrimination problems? Is a
bright rat raised in the dark as "smart" as a
dull rat raised in a normal environment?
2. What is the relevance to discrimination-learn
ing scores of such other litter-connected
factors as temperament? Is boldness or timidity
a factor in discrimination-learning as measured?
Is it more or less a factor than prior visual
experience?
Finally, an interesting series of divergent studies
is suggested by one of the troublesome incidental findings
connected with the execution of the present experiment.
The apparent increase in cannibalism in rat mothers forced
to rear their young in complete darkness may point to some
connection between stimulation by light and the neural
and/or humoral regulators of maternal behavior in the rat.
These are only a few of the most apparent sugges
tions for further research that emerge from the present
study. A lifetime could probably be spent pursuing all
the implications of even so small a laboratory study as
that reported here, a study based on one aspect of one
theory of perceptual development. Such are the ways of
science.
LIST OF REFERENCES
LIST OF REFERENCES
Allport, Floyd H. Theories of perception and the concept
of structure. New York: Wiley, 1955.
Bartley, S. Howard. Principles of perception. New York:
Harper, 1958.
Beach, F. A., and J. Haynes. Effects of early experience
upon the behavior of animals. Psychol. Bull., 1954,
51, 239-263.
Berlyne, D. E. Conflict, arousal and curiosity. New York:
McGraw-Hill, 1960.
Boring, Edwin G. A history of experimental psychology.
2d ed. New York: Appleton-Century-Crofts, 1950.
Brogden, W. J. Animal studies of learning. Chap. 16 in
S. S. Stevens (ed.)., Handbook of experimental psychol
ogy. New York: Wiley, 1951.
Chow, K. L., and H. W. Nissen. Interocular transfer of
learning in visually naive and experienced chimpanzees.
J. comp, physiol. Psychol., 1955, 48, 229-237.
Farris, Edmond J., and John Q. Griffith, Jr. (eds.). The
rat in laboratory investigation. 2d ed. Philadelphia:
Lippincott, 1949.
Forgays, D. G., and Janet W. Forgays. The nature of the
effect of free environment experience in the rat.
J. comp, physiol. Psychol.. 1952, 45, 322-328.
Forgus, R. H. The effect of early perceptual learning on
the behavioral organization of adult rats. J. comp.
phvsiol. Psychol.. 1954, 27, 331-336.
. Influence of early experience on maze learning
with and without visual cues. Canad. J. Psychol.,
1955, 9, 207-214.
_______. The effect of different kinds of form pre
exposure on form discrimination learning. J. comp.
physiol. Psychol., 1958, 51, 75-78.
164
165
Forgus, Ronald H. The interaction between form pre
exposure and test requirements in determining form
discrimination. J. comp, phvsiol. Psvchol.. 1958,
51, 588-591.
Gauron, E. F., and W. C. Becker. The effects of early
sensory deprivation on adult rat behavior under com
petition stress: an attempt at replication of a
study by Alexander Wolf. J. comp, physiol. Psvchol..
1959, 52, 689-693. --------
Gibson, E. J., and R. D. Walk. The effects of prolonged
exposure to visually presented patterns on learning to
discriminate them. J. comp, physiol. Psychol.. 1956,
49, 239-242.
Gibson, E. J., R. D. Walk, and T. J. Tighe. Enrichment
and deprivation of visual stimulation during rearing
as factors in visual discrimination learning. J. comp.
physio1. Psycho1.. 1959, 52, 74-81.
Griffiths, W. J., Jr., and W. F. Stringer. The effects of
intense stimulation experienced during infancy on
adult behavior in the rat. J. comp, physiol. Psychol.,
1952, 45, 301-306.
Guilford, J. P. Fundamental statistics in psychology and
education. 3d ed. New York: McGraw-Hill, 1956.
Guthrie, E. R. The psychology of learning. New York:
Harper, 1935.
Hebb, D. 0. The innate organization of visual activity:
I. Perception of figures by rats reared in total
darkness. J. Genet. Psychol.. 1937, 51, 101-126.
______. The effects of early experience in problem-
solving at maturity. Amer. Psychol., 1947, 2, 306-307.
The organization of behavior. New York: Wiley,
T959.
Helson, Harry (ed.). Theoretical foundations of psychol
ogy. New York: D. Van Nostrand, 1951.
Hilgard, Ernest R. Methods and procedures in the study of
learning. Chap. 15 in S. S. Stevens (ed.)., Handbook
of experimental psychology. New York: Wiley, 1951.
166
Hunt, J. McV. The effects of infant feeding frustration
upon adult hoarding in the albino rat. J. abnorm. soc.
Psychol.. 1941, 36, 338-360.
Hymovitch, B. The effects of experimental variations on
problem solving in the rat. J. comp, physiol.
Psvchol.. 1952, 45, 313-321.
Krechevsky, I. "Hypotheses" in rats. Psychol, rev..
1932, 39, 516-532.
Lashley, K. S. Brain mechanisms and intelligence.
Chicago: University of Chicago Press, 1929.
. The mechanism of vision XV. Preliminary studies
of the rat'8 capacity for detail vision. J. General
Psvchol.. 1938, 18, 123-193.
Michels, K. M., W. Bevan, and H. C. Strasel. Discrimina
tion learning and interdimensional transfer under
conditions of systematically controlled visual
experience. J. comp, physiol. Psychol., 1958. 51,
778-781.
Munn, N. L. Handbook of psychological research on the rat.
Boston: Houghton Mifflin, 1950.
Nissen, Henry W. Phylogenetic comparison. Chap. 11 in
S. S. Stevens (ed.)., Handbook of experimental psychol
ogy. New York: Wiley, 1951.
Riesen, A. H., The development of visual perception in man
and chimpanzee. Science. 1947, 106, 107-108.
. Arrested vision. Scient. Amer.. 1950. 183,
r?^19.
Senden, M. v. Raum-und Gestaltauffassung bei operierten
bllnd-geborenen vor und nach der operation. Leipzig:
Barth, 1932.
Siegel, Sidney. Nonparametric statistics for the behav
ioral sciences. New York: McGraw-Hill, 1956.
Solley, C. M., and Gardner Murphy. Development of the
perceptual world. New York: Basic Books, 1960.
Spence, Kenneth W. Theoretical interpretations of learn
ing. Chap. 18 in S. S. Stevens (ed.)., Handbook of
experimental psychology. New York: Wiley, 1951.
167
Stevens, S. S. Mathematics, measurement, and psycho
physics. Pp. 1-49 in S. S. Stevens (ed.)., Handbook
of experimental psychology. New York: Wiley, 1951.
Walk, R. D., E. J. Gibson, H. L. Pick, Jr., and T. J.
Tighe. Further experiments on prolonged exposure to
visual forms: the effect of single stimuli and prior
reinforcement. J. comp, physiol. Psychol.. 1958, 51,
483-487.
Wolf, A. The dynamics of the selective inhibition of
specific functions in neurosis. In S. S. Tomkins
(ed.)., Contemporary psychopathology. Cambridge:
Harvard University Press, 1943.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
The Effects Of Ether And Electroconvulsive Shock On One Trial Appetitive And Adversive Learning
PDF
The Modification Of Partial Reinforcement Effect As A Consequence Of Electrical Stimulation Of The Caudate Nucleus In Cats
PDF
The role of cues in the arousal of anxiety
PDF
A Study Of Affective Sets: The Effects Of Family And Non-Family Verbal Contexts On Word-Need Stimuli In A Word Association Experiment With Reference To Pleasant And Emotional Tones Of Associated...
PDF
Empathy And Social Perception
PDF
Dominance And Contiguity As Interactive Determinants Of Autonomic Conditioning
PDF
Behavioral Seriousness And Impulse-Control Balance In Delinquency
PDF
Validity Concomitants Of Various Scoring Procedures Which Attenuate The Effects Of Response Sets And Chance
PDF
An Effect Of Verbal Conditioning On Nonverbal Behavior
PDF
A Study Of The Effectiveness Of Teaching Methods Of Study To Selected High School Freshmen
PDF
A Factor Analytic Investigation Of The Effects Of Specific Visual Contentof Motion Pictures Used In Psychological Measurement
PDF
The Effect Of Discriminability On The Partial Reinforcement Effect In Human Gsr Conditioning
PDF
The Effect Of Subject Sophistication On Ratio And Discrimination Scales
PDF
Habituation Of Aggression In The Three-Spined Stickleback (Gasterosteus Aculeatus L.)
PDF
A Study Of The Relationship Of Temperament Variables To The Ability To Make Certain Judgments Of Emotional Behavior
PDF
A Study To Compare The Effectiveness Of Individual And Group Counseling Approaches With Able Underachievers When Counselor Time Is Held Constant
PDF
A Problem-Centered Approach To Introductory Psychology
PDF
Figural And Symbolic Divergent-Production Abilities In Adults And Adolescents
PDF
The Effects Of Group Experiences On The Aged
PDF
The effect of noise on intellectual performance as related to personality and social factors in upper division high school students
Asset Metadata
Creator
Libaw, Frieda Bornston
(author)
Core Title
The Effects Of Prior Part-Experiences On Visual Form Perception In The Albino Rat
Degree
Doctor of Philosophy
Degree Program
Psychology
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
OAI-PMH Harvest,psychology, experimental
Language
English
Contributor
Digitized by ProQuest
(provenance)
Advisor
Wyers, Everett J. (
committee chair
), Grings, William W. (
committee member
), Guilford, Joy P. (
committee member
), Jacobs, Alfred (
committee member
), Lefever, David Welty (
committee member
)
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c18-101955
Unique identifier
UC11358076
Identifier
6103801.pdf (filename),usctheses-c18-101955 (legacy record id)
Legacy Identifier
6103801.pdf
Dmrecord
101955
Document Type
Dissertation
Rights
Libaw, Frieda Bornston
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA
Tags
psychology, experimental