You are on page 1of 4

THE EFFECT OF PROLONGED EXPOSURE TO VISUAL PATTERNS ON LEARNING

TO DISCRIMINATE SIMILAR AND DIFFERENT PATTERNS1


ELEANOR J. GIBSON, RICHARD D. WALK, HERBERT L. PICK, JR., AND THOMAS J. TTGHE
Cornell University

The experiments to be reported are part of Triangles and circles were exposed on the cage
a program devoted to the investigation of the walls, and at 90 days discrimination learning
role of certain kinds of prolonged visual stim- of the same shapes began. Group d was con-
ulation on visual discrimination of patterns. trol group for EI. At no time during the experi-
The basic technique has been that of giving ment were there metal patterns on the cage
an experimental group an opportunity to view walls, but discrimination learning began at 90
patterned stimuli from birth until 90 days of days. Group £2 had triangles and circles ex-
age, and then requiring them to discriminate posed on the cage walls from birth on, like
these patterns. Performance on the discrimina- Group EI. But at 90 days they learned a dis-
tion is compared with that of a control group crimination between an ellipse and an isosceles
raised in the usual cage environment. In our triangle. These forms were chosen because
first experiment, a control group was raised in they are members of the families of perspective
ordinary mesh cages surrounded by card- transformations to which the circle and equi-
board walls. The experimental group was lateral triangle belong. The shapes were
raised in a similar environment, but a black approximately equal in area to the cage figures
metal cutout was hung on each of the four (circle was 2% in. in diameter, base of equi-
walls, circles on two sides and triangles on two lateral triangle was 3 in.). The ratio of the
sides. The results (1) showed a significant axes of the ellipse was 2^' to 1, and the base
difference between the two groups in speed of of the isosceles triangle was equal to the long
learning to discriminate the circle from the dimension of the ellipse. Group C-2 was a con-
triangle in the test series. Many questions trol for Group E2. No forms were exposed on
arose, at this point, as to the interpretation of the cage walls, but at 90 days, the animals
the difference. The one which interested us learned to discriminate the ellipse and the
most was whether the facilitation observed isosceles triangle.
was specific to the exposed figures (the only
ones tested), or whether it was a more general Method
effect, possibly dependent on the degree of Subjects. The 5s were albino rats born in the Cornell
similarity between the exposure and test laboratory. Litters were split when the animals were
1 to 3 clays old, so that control and experimental groups
patterns. It is this problem—the generality could be matched by distribution of littermates between
or specificity of the effects of prolonged ex- them. The N was 13 in Groups EI and Ci, 15 in Eo,
posure—with which the present experiments and 14 in Co.
are concerned. Rearing. The animals of all four groups were raised
in similar mesh cages, each cage containing a mother
EXPERIMENT I and a mixed litter averaging about seven young. The
mothers were removed at weaning, and sexes were sepa-
Is facilitation of discrimination by pro- rated. Each cage was surrounded by white cardboard
longed exposure a matter of identifying a walls, 6 in. from the cage, on four sides. Only the ceiling
particular pattern, or is some more general was visible above. The room was lighted at all times
(but the lighting fixtures were not visible to the rats).
process responsible? The experimental design Food and water were always available in the cages.
to answer this question included four groups of The two experimental groups were treated differently
animals. Group E\ was identical with the in one way only: black metal cutout triangles and
experimental group of the original experiment. circles were hung on the cage walls from birth on. There
was one cutout to each wall, and these were moved
1
This research was supported in part by a grant around the cage at intervals so that their spatial prox-
from the National Science Foundation. The first ex- imity to food and water would be random. The cutouts
periment was reported to the Experimental Psychology remained on the cage walls throughout, until discrimina-
Group in Oxford, England, April, 1956, and at the tion training was completed.
American Psychological Association in September, 1956. Apparatus for discrimination learning. A Grice-type
584
EXPOSURE TO VISUAL PATTERNS 585

discrimination box was used, like that described A comparison of errors for the two groups
by Gibson and Walk (1). Two discrimination again shows the experimental group superior
chambers adjoined each other so that S need
not be handled during a day's training. In a to the control group. The mean number of
given trial the animal was faced by two initial errors was 57 for Group C and 49 for
stimuli, a black triangle (equilateral for Groups Ei Group E. The difference is small, and, taken
and Ci and isosceles for K-2 and C?), and a black circle alone, not significant (p = .06).
(E! and Ci) or ellipse (E5 and C2). These were painted
on a white background, the same size as the cutouts, Variability among the animals was great.
and stood side by side. Food was obtained by pushing The control animals ranged from 75 to 222
open a small door in the center of the pattern painted trials to reach the criterion, and the cxperimen-
on the stimulus holder. The door was locked for the tals from 72 to 202. The importance of splitting
negative stimulus. Doors could be lowered over the two litters among the experimental conditions
stimulus patterns at the end of a trial, or if an error
were made. The animal then turned around, went seems obvious, as does the necessity of replica-
through the door separating the two choice compart- tion.
ments, and again faced the two patterns. No punish- Specificity of transfer. Table 1 presents
ment was employed, but if the animal made an error, means for the four grotips run in the present
both doors were instantly lowered over the stimulus
patterns, and the animal went to the other compartment experiment, for total errors and trials to
for a second choice. If the animal made three errors on criterion. Group £2, as well as Group E1( was
one trial, it was permitted to open the door in the superior to its control group in both errors
positive stimulus pattern without going to the other and trials. The t tests for EX vs. Cx and for Ea
compartment. A trial was thus denned as choice of the vs. Co were significant at the .05 level of con-
correct door (R or L) leading to reinforcement, and
might be preceded by as many as three incorrect choices. fidence for both trials and total errors, with a
Procedure. Discrimination training began at about one-tailed test. Since the ellipse-isosceles dis-
90 days. Pretraining the animals to open the small crimination was easier than the circle-equi-
doors and eat from the stimulus holders took 3 to 4 lateral discrimination, the two experimental
days and was done with plain black holders. Then
10 trials a day were given with the patterns in place. groups must be compared relative to their
The learning criterion was 18 correct out of 20 responses, control groups. Percentage facilitation was
with the last 10 in succession correct. The animals therefore calculated. As Table 1 shows, the
were run 6 days a week until they reached criterion, or facilitation was from 12% to 18%, and is no
until they had been run for 30 clays. greater for Group EI than for Group Eg. Similar
There were two Es, each of whom ran half the
animals. The circle or ellipse was positive for half the patterns are discriminated as easily as ones
animals, the triangle (equilateral or isosceles) for the identical with the exposed pattern after pro-
others. longed exposure.
Our results suggest, therefore, that what the
Res-lilts
animal learns from viewing the triangle and
Replication of original experiment. A com- circle on the cage walls is not a specific identi-
parison of Groups G! and Ei permits a check on fication of these two patterns. If pattern dis-
the results of our original experiment, Since crimination itself is a function of learning, as
the next experiment also duplicated these Hebb for one has suggested (4), the learning
groups, the two replications will be considered
here together. The mean number of trials to TABLE 1
reach criterion was 144 and 134, respectively, Mean Number of Trials to Learn, Mean Total Errors,
for the two experimental groups, and 163 for and Per cent Facilitation* for
both control groups. The original trend is thus All Groups, Experiment I
corroborated in both experiments. If animals %
from the two experiments are combined, there Facil-
Mean Total Mean Trials itation Facil-
A' Group Errors to Learn itation
is a difference of 23 trials between Groups EI (Total (Trials)
F.)
and Ci. The significance of this difference is
between the .01 and .02 level of confidence 13 Et 89.9 143.7 17 12
with a one-tailed test. It seems, therefore, 13 f~t
108.6 162.5
that the facilitating effect of exposure to 15 E, 59.2 97.6 15 18
visual patterns is confirmed, but the difference 14 C-> 69.6 120.1
is not as striking as it was in the original ex-
periment. * Per cent facilitation ~ -w°-£ x 100.
586 ELEAXOR GIBSOX, RICHARD WALK, HERBERT PICK, JR., AXD THOMAS TIGHK

process would have lo be one which accounts are present. We tried, therefore, to vary the
for some equivalence of stimuli. One hy- exposure patterns from the test patterns as
pothesis might be a learning theory of shape widely as possible. This turned out to be diffi-
constancy. As the animals walk and climb cult; if curves and corners are possibly the
about the cage, the patterns yield transforma- crucial features, what kind of pattern has
tions varying continuously in shape and size. neither? A double-barreled plan was finally
There will be constant common features with settled on, in an attempt to get around this
concomitant variation of other features. All problem. Horizontal and vertical stripes were
t h e dimensions of transformation might be chosen as different patterns for test stimuli,
categorized to yield something analogous to for groups which had circles and triangles as
a concept of triangle and circle. A second hy- exposure patterns. Also, other groups were
pothesis might assume that the common fea- run which had what were supposed to be
tures which stay constant through changes of "formless" objects on the cage walls, to pro-
viewing distance and orientation (here such vide something to fixate there but a something
attributes of shape as curves and corners) with no pronounced pattern characteristics.
become capable of operating as stimuli. The Rocks, painted black, were chosen for this.
animal might be learning to respond to de- The}' were smaller than the cutout shapes and
rivatives of points of stimulation, namely, were irregularly formed, with rough surfaces.
angles and curves. These are relational char- They hung inside the cage.
acteristics of stimulation; perhaps they must
be discovered by a learning process which re- Method
peated viewing permits (2).
Six groups of rats were run in this experiment. Their
But. the present results could be equally differences are summarized in Table 2. As usual, llie
well explained by a third hypothesis without
supposing that the animals' sensitivity to TABLE 2
patterns or features of them has been improved Experimental Design
at all. Perhaps the animals with patterns on
the cage walls develop auxiliary habits of Cage Discrimina- IDiscrimina-
Groups
looking which facilitate the visual discrimina- Stimuli ; tion 1 tion 2
tion (e.g., looking upward or "attending" to
the stimuli present for vision). Why such c, ; None T-C X'one
Co ; None H-V T-C
habits should be learned is not entirely clear, E, T-C T-C None
since looking upward, responding to visual E; T-C H-V T-C
stimuli instead of, say, olfactory stimuli, and CE, ; Rocks T-C XTone
so on are not reinforced. If looking itself is CE2 j Rocks H-V T-C
reinforcing, as Harlow suggests (3), we must Xote.—T, triangle; C, circle; II, horizontal stripes; V, vertical
assume here that the animal that has had stripes.
something to look at tends to look more.
TABLE 3
EXPERIMENT II Mean Number of Trials Required lo Reach Criterion
This experiment was designed in an effort and Mean Initial Errors for All Six Groups
to determine whether the generalized facilita- 1 Mean Trials Mean Errors
tion was due to something completely unspeci- V
lic, like viewing habits, or whether it was clue T-C H-V T-C H-V
I
to some process inherent in the development of
discrimination. Development of pattern dis- E, 10 ! 133.8 46.6
E,, 8 I 130 50.4 44.9 14.1
crimination might proceed along dimensions, C, 8 i 162.5 51.1 :
as has been suggested, and account for gen- C2 9 137.2 46 50.4 ; 12.6
eralization where similarities or common fea- CE, 11 ! 138.9 47
!
tures of stimuli are involved. In such a case, CE2 8 138.5 39.9 41.9 11.4
we should expect the generalization to break Xote.—T, triangle; C, circle; H, horizontal stripes; V, vertical
down when no common features of stimulation stripes.
EXPOSURE TO VISUAL PATTERNS 587

main experimental group had cutout triangles and of Group EI) does not permit a firm conclusion,
circles on the cage walls. Two such groups were run, since this group did not differ significantly
EI and E->. Group EI was tested on a triangle-circle
discrimination in the usual fashion at about 90 days. from Group CEi. The problem which concerned
Group E-i, however, at 90 days learned the horizontal- us, the exploration of the apparent generality
vertical stripe discrimination. Afterward, they also of the effect, is unfortunately not much clari-
discriminated the triangle and circle. Two control fied.
groups were run to match these; neither had cage pat-
terns. Group C\ learned the triangle-circle discrimination SUMMARY AND CONCLUSION'S
at 90 clays, as usual. Group Co learned the stripe dis-
crimination at 90 daj'S, followed by triangle-circle. 1. Two experiments were performed in an
Groups CEi and CE-> had rocks hung on each of the cage attempt to discover whether exposure of visual
walls. One group learned the triangle-circle discrimina-
tion at 90 days; the other first discriminated the stripes, patterns on cage walls has a specific facilitating
and afterward, the triangle and circle. effect on visual discrimination learning, or
Details of method were precisely as in the foregoing whether facilitation is of a generalized sort.
experiment, and need not be repeated. Litters were In one experiment the effect of pattern ex-
split at birth or two days of age and distributed among
three groups. Two -Es ran the animals, each running
posure was tested by discrimination of identical
half of each group. patterns and of similar patterns. In the other,
the exposure pattern and the pattern to be
Results discriminated were varied as much as possible.
The principal results are presented in Table 2. Exposure of visual patterns on the cage
3. The standard EI group was superior in learn- walls was found to facilitate learning a dis-
ing the discrimination to the standard control, crimination between the same patterns, con-
firming results of a previous experiment.
Ci, and the difference has already been dis-
cussed in the previous section. The other com- 3. Exposure of visual patterns on the cage
parisons merit only the briefest discussion, walls also facilitated learning a discrimination
since by /- tests none of the differences reached between patterns similar to, but not identical
with, the exposed ones.
a satisfactory significance. The three groups
which learned to discriminate the striations do 4. When cage pattern and test pattern were
not vary significantly from one another, al- made very different, no significant transfer
effects were found.
though two of them had some kind of cage
exposure. The striations, it may be noted, are
REFERENCES
a very easy discrimination. Furthermore,
learning the striations before learning the tri- 1. GIBSON, E. ]., & WALK, R. D. The effect of pro-
angle-circle discrimination lowers the mean longed exposure to visually presented patterns
on learning to discriminate them. J. coinp. pliysiol.
trials to criterion only in the case of the con- Psychol, 1956, 49, 239-242.
trol group, and even here the difference is not 2. GIBSON, J. J., & GIBSON, E. J. Perceptual learning:
significant. The main comparison which we Differentiation or enrichment? Psychol. Rev.,
wished to make was the ease of discriminating 1955, 62, 32-41.
3. HARI.OW, H. F. Motivation as a factor iu the acquisi-
the triangle and circle after viewing rocks, tion of new responses. In J. A. Brown et al.,
with the same discriminatioii after viewing Current llieory and research in motivation. Lincoln:
the triangle and circle. Group CEi, in this case, Univer. Nebraska Press, 1953.
does have lower mean scores than the control 4. HEBB, D. 0. The organization of behavior. New York:
group Ci, but the difference is not significant. Wiley, 1949.
The evidence in favor of specificity (superiority Received December 14, 1956.

You might also like