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
/
An Application Of A Two-Stage 'Attention' Model To Concept Formation In The Mentally Retarded
(USC Thesis Other)
An Application Of A Two-Stage 'Attention' Model To Concept Formation In The Mentally Retarded
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
I
This dissertation h as been
m icrofilm ed exactly as received 67— 1 7 ,7 0 2
SMITH, Ruth Ellen, 1930-
AN APPLICATION OF A TWO-STAGE "ATTENTION"
MODEL TO CONCEPT FORMATION IN THE MENTALLY
RETARDED.
U n iversity of Southern C alifornia, Ph.D ., 1967
P sych ology, clin ical
|
University Microfilms, Inc.. Ann Arbor, Michigan
AN APPLICATION OP A TWO-STAGE "ATTENTION" MODEL
TO CONCEPT FORMATION IN THE MENTALLY RETARDED
by
Ruth Ellen Smith
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)
September 1967
UNIVERSITY O F SO U TH ER N C A U FO R N IA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES, CALI FORNIA 9 0 0 0 7
This dissertation, written by
......... ilu.th...EHen..SinJ.£Jfci...........
under the direction of he.x:....Dissertation Com
mittee, and approved by all its members, has
been presented to and accepted by the Graduate
School, in partial fulfillment of requirements
for the degree of
D O C TO R OF P H IL O S O P H Y
Dean
DISSERTATION COMMITTEE
Chairman
THAT LOYAL GROUP OF FRIENDS
WITHOUT WHOSE SUPPORT AND ENCOURAGEMENT,
THIS DISSERTATION WOULD NEVER HAVE BEEN
BEGUN, MUCH LESS COMPLETED.
ACKNOWLEDGMENTS
The author wishes to express her thanks to
Dr. James T. Shelton, Superintendent and Medical Director
of Porterville State Hospital, and to the Administration
for their support during this study. The experimental
phase of the research was locally funded under project num
ber PVL 53. Gratitude is also due to the patients and to
the staff— both professional and ward level. Without their
cooperation and assistance this project could never have
been done.
Particular thanks go to my colleagues in the Psy
chology Division of Porterville State Hospital for making
time available to me to complete my research, for their
many helpful suggestions and for their boundless emotional
support.
Finally, I would like to express my thanks to my
Chairman, Professor Alfred Jacobs, and to the other members
of my committee for their guidance, help and patient
tolerance.
iii
TABLE OF CONTENTS
i
Page
DEDICATION............................................. ii
j
ACKNOWLEDGMENTS........................................ iii
I
LIST OF T A B L E S ........................................ vi
LIST OF FIGURES viii i
t
i
]
Chapter
I. THE PROBLEM........................ 1
The Background of the Problem
Introduction
Major theories of concept formation
Attention theory
Studies of concept formation in the
retarded |
Formulation of the Problem j
Hypothesis j
II. M E T H O D ........................................ 41
Experiment I
Subjects
Apparatus
Procedure
Experiment II
Subj ects
Apparatus
Procedure
III. RESULTS........................................ 66
Experiment I
Experiment II
iv
IChapter Page
IV. DISCUSSION..................................... 91
Implications for Further Research
V. SUMMARY......................................... 114
REFERENCES.............................................. 118
APPENDIX . ........................................... 132
i
i
v
LIST OF TABLES
Table Page
1. Descriptive Data by Treatment Groups for
Experiments I and I I ......................... 43
2. Means and Standard Deviations of Experiment I
(9 Card) Trials to Criterion ................ 67
3. Means and Standard Deviations of Experiment I
(9 Card) Error Scores ....................... 68
4. Analysis of Variance of Trials and Errors
Scores, Experiment I, Pretraining Groups x
M.A. Levels (PPVT) (Subjects Who Failed Were
Replaced).................................... 70
5. Analysis of Variance of Trials and Errors
Scores, Experiment I, Pretraining Groups x
M.A. Levels (PPVT) (Includes Subjects Who
Failed) ....................................... 71
6. Correlations for Experiment I ................ 73
7. Adjusted Cell N's and Mean Trials Based on
Three Levels of Raven Raw Score,
Experiment I .................................. 75
8. Number and Percentage of Subjects Able to Sort
All Three Generalization Cards Correctly,
Experiment I ................ 82
9. Means and Standard Deviations of Experiment II
(4 Card) Trials and Errors Scores ......... 84
10. Analysis of Variance of Trials and Errors
Scores, Experiment II, Pretraining Groups x
M.A. Levels (PPVT)........................... 85
11. Correlations for Experiment I I ................ 86
vi
Table Page
12. Number and Percentage of Subjects Able to
Sort All Three Generalization Cards Correct
ly, Experiment I I ........................... 89
13. Three Way Analysis of Variance of Trials
Scores by PPVT M.A. Levels, Pretraining
Groups and Number of Stimulus Cards (Four or
Nine) ......................................... 90
LIST OF FIGURES
Figure Page
1. Pretraining Material for Group I (Exposure)
and Group III (Form)........................ 47
2. Stimulus Cards for the Concept Sorting Task . 50
3. Auxiliary Cards Used for Labeling Concept
Categories (Set II) and for Testing Concept
Generalization (Set III)................... 51
4. Median Errors per Trial, by M.A. Levels, for
Each Treatment Group. Experiment I.
Failures Replaced............................ 77
5. Cumulative Percentage of Subjects Achieving
Errorless Performance on Each Trial until
All Are at Criterion. Group I, Exposure . . 78
6. Cumulative Percentage of Subjects Achieving
Errorless Performance on Each Trial until
All Are at Criterion. Group II, Dot
Matching..................................... 79
7. Cumulative Percentage of Subjects Achieving
Errorless Performance on Each Trial until
All Are at Criterion. Group III, Form
Matching..................................... 80
8. Cumulative Percentage of Subjects Achieving
Errorless Performance on Each Trial until
All Are at Criterion. Experiment II, All
Groups. Failures Are Excluded ............. 88
9. Frequency Distribution of Errors on Trial One,
Experiment I ................................ 100
viii
Figure Page
10. Frequency Distribution of Errors on Trial One,
Experiment II................................. 107
1
ix
CHAPTER I
THE PROBLEM
The Background of the Problem
Introduction
Psychologists have proposed many definitions of the
term concept. A sample of the definitions could include:
"common elements" (Hull, 1920), "relation between parts"
(Smoke, 1932), the "object of cognition" (Heidbreder, 1945)
"common cues resulting from the subject's own symbolic
processes" (Osgood, 1953), "an operation involving equilib
rium of intension and extension" (Inhelder and Piaget,
1964), and an "implicit encoding response" (Bower and
Trabasso, 1964).
Kendler (1964) indicates the basis for some of the
proliferation of definitions by pointing out that a concept
may serve one of three functions: association, cue, or
response. A concept may be the association between dissim
ilar stimuli and a common response, (S^-R^; S2-R1 ), as
when a subject learns a concept de novo. For example, a
2
1 child is told that a saucer, a ball, a stool, a merry-go-
round and the moon are all "round." After repeated experi
ence with different examples, the child learns the stimulus
characteristics which result in an object being called
round. In the situation in which a concept has already
been learned, it may serve as a cue and take on mediating
functions. A subject who knows the concept "vegetable,"
would be expected to learn the paired associates, "tomato-
GUZ, carrot-GUZ" more quickly than the pairs "tomato-GUZ,
elephant-GUZ" because tomato and carrot evoke a common
mediator. Third, a concept may serve as an observing or
attention response. Such a response determines to which
aspects of a stimulus the subject will respond. If a sub
ject makes an observing response to shape, he is more
likely to classify a ball on the basis of its roundness
than on the basis of its color (House and Zeaman, 1960a).
Major Theories of Concept Formation
The major theories regarding concept formation and
concept acquisition may be classified as single-stage S-R
and two-stage or mediated S-R. Three other theoretical
positions, which are concerned more with sorting behavior
than concept formation, will be discussed briefly to
|complete this section. Concept acquisition will refer to
I
the process of developing a concept when the rule for
classifying must be learned de novo. Concept identification
is the process of determining which of a number of rules
known to the subject is being used by the experimenter to
classify stimuli. Concept formation will be used to include
both concept acquisition and concept identification.
Single-stage S-R. Hull (1920) presented a "common
elements" theory of concept acquisition. From a list of 12
radicals, ^ he constructed 144 pictographs, each radical
being represented in 12 variations. He then made up 12
lists of pictographs, each list containing one variant of
each of the basic radicals. The subject was presented with
a list of pictographs and a list of nonsense syllables and
was required to learn which nonsense syllable was the cor
rect response for which pictograph. All 12 lists of picto
graphs were presented in turn, each with the same list of
nonsense syllables. Unknown to the subject, the correct
response for any pictograph depended on its radical. That
is, all pictographs constructed upon the same radical had
*A radical is a simple pattern of brush strokes
which forms the basic element in a Chinese character or
pictograph.
-the same nonsense syllable label. As the subject progressed
through the lists, different radicals began to "stand out,"
so that in time the subject was able to respond with the
correct nonsense syllable on the first presentation of a
pictograph. To Hull, the basic set of strokes, i.e., the
common form elements, which made up a radical defined a
concept and the nonsense syllable associated with that
radical was the name of the concept. By analogy, Hull
reasoned, the characteristics common to terriers, poodles,
and hounds define the concept named "dog."
Smoke (1932) disagreed with Hull that common ele
ments determine a concept such as "dog." A toy poodle and
a Great Dane have no identical parts, but both are dogs.
For Smoke, identifying common elements was discrimination
learning, not concept formation. He defined concept forma
tion as the recognition of relationships between elements.
For example, a concept labeled DAX might be defined as "one
figure inside another." The size or shape of the figures
would be irrelevant to the concept.
Gray (1931) would define a concept by naming all of
the responses and stimuli which are conventionally classi
fied under the concept word. For example, "table" might be
composed of the following responses: wood, square, four
|legs, flat surface, brown, used for writing, etc.
Bourne et al. (Bourne, 1957; Bourne, 1963; Bourne
and Bunderson, 1963; Bourne, Goldstein, and Link, 1964;
Bourne, Guy, Dodd, and Justesen, 1965; Bourne and Haygood,
1961; Bourne and Parker, 1964; Bourne and Pendleton, 1958;
Bourne and Restie, 1959), in an extensive research program,
have proposed a cue conditioning model of concept identifi
cation. Two processes are basic to the cue conditioning
model, the conditioning of relevant cues and the adaptation
of irrelevant cues. Any stimulus is considered to repre
sent a population of potential cues; size, shape, color,
position, etc. When a subject makes a response to a
stimulus object, he responds to a sample of cues. Whether
or not the sample of cues is conditioned to the response
depends on whether or not reinforcement follows the re
sponse. Over a series of trials, responses to the relevant
cues are reinforced more often than are responses to the
irrelevant cues and the probability of the subject respond
ing to the relevant cues approaches 1.00.
Two-stage models. Two-stage models postulate that
the subject learns an intervening variable in the process
of learning a concept. The intervening variable has been
identified by different theorists as perceptual, verbal or
"response produced cue."
Kounin (1941a, 1941b), a Lewinian, would not con
sider himself a mediationalist, however, it is possible to
reformulate his theory in terms of an intervening variable,
the "life space." Kounin assumed that mentally retarded
persons might be characterized as having more rigid bound
aries between the various regions of their life space.
Therefore, retarded subjects would be expected to find
greater difficulty than normal subjects in shifting from
one sorting task to another. He found that two thirds of a
group of older retardates and half of a group of younger
retarded subjects could not shift from sorting on the basis
of color to sorting on the basis of form although all of
the normal subjects could do so. Kounin also found that
retarded subjects would work longer at a dull task than
would normal subjects. Zigler (Zigler, 1961; Zigler, 1962;
Zigler, 1963; Zigler and deLabry, 1962) has argued that the
greater persistence of retardates than normals is attribut
able to differences in motivational factors rather than to
rigidity. Zigler asserts that the retarded are more
responsive to social reinforcers and are more likely to
comply with adult requests than are normal subjects.
7
|
Heidbreder (1945) advanced a perceptual theory of
concept formation. For her, human cognition ... may be
ordered with respect to two kinds of performances, the per
ception of concrete objects and the attainment of concepts"
(p. 1). She considered the perception of concrete objects
to be the dominant cognitive response. All other cognitive
responses, including concept learning, were approximations
or modifications of the perception of objects. Heidbreder
hypothesized that subjects would identify concepts based on
concrete objects more quickly than concepts based on numer-
osity since concrete objects are more tangible than are
numbers. (Heidbreder discusses the "perceived thingness"
of a stimulus, by which she seems to mean tangibleness or
some measure along a dimension of abstractness.) Heid
breder et al. (Heidbreder, 1946a; Heidbreder, 1946b;
Heidbreder, 1947; Heidbreder, 1948; Heidbreder, Bensley,
and Ivy, 1948; Heidbreder and Overstreet, 1948) conducted a
series of experiments using pictures paired with nonsense
syllables, and interpreted the results of the experiments
as supporting the hypothesis. The subjects learned the
nonsense syllable associated with "a picture of a face,"
more quickly than the nonsense syllable associated with "a
picture showing two lines intersected by a third." The
nonsense syllable associated with the latter concept was
learned more quickly than the nonsense syllable associated
with the concept, "a picture displaying three objects."
Osgood (1953) proposes that sorting or labeling
based on physical identity or common configuration among
the stimuli be defined as "labeling" rather than conceptual
behavior. Osgood defines concepts in terms of similarities
based on abstract or mediated properties rather than on
more concrete physical similarities. For example, an
orange and a banana are similar because both are fruit; an
orange and a baseball are similar because both are round.
Osgood would classify the former similarity as a concept,
the latter as labeling.
Kendler (1964) disagrees with Osgood's declaration
that a concept requires the subject to make an abstraction
to stimuli "having no property in common" since there are
no stimuli which do not have some property in common.
Kendler's work has been concerned with the mediation pro
cess in concept formation. "The subject is assumed to make
a hypothetical implicit response (r) which in some way
modifies the external source of stimulation to produce a
transformed stimulus (s) that elicits behavior" (Kendler
and Kendler, 1966, p. 282). The implicit response may be
verbal (Kendler and Karasik, 1958; Kendler and Kendler,
1962) or perceptual (Kendler, Glucksberg, and Keston, 1961).
Kendler and others have accumulated evidence supporting the
mediation model, however, there is conflicting evidence as
well.
A number of studies have compared reversal shift
with nonreversal shift in discrimination learning. In
reversal shift the negative cue becomes positive and the
positive cue negative. In nonreversal shift, cues from a
different stimulus dimension become relevant and the old
cues become irrelevant. Mediation would predict superior
ity for reversal shift since only the overt response needs
to be changed, e.g., the subject responds to the red object
instead of to the blue one; in nonreversal shift, the
mediated chain must be modified, e.g., the subject responds
to the round object instead of to the blue one. Kirk (1964)
points out that the superiority of reversal shift over non
reversal is predicted only if the rate of extinction of
overt responses is greater than the rate of acquisition of
mediators. Several studies have shown that reversal shift
is easier than nonreversal for college students, (Harrow
and Friedman, 1958; Isaacs and Duncan, 1962; P. J. Johnson,
1966; Kendler and D'Amato, 1955; Kirk, 1964), normal
children (Sanders, Ross, and Heal, 1965) and for fast
i
learners among kindergarten children (Kendler and Kendler,
1959). The superiority of reversal over nonreversal shift
was not maintained in a four-choice situation (Kendler and
Mayzner, 1956).
Goss (Goss, 1955; Goss, 1961a; Goss, 1961b; Goss
and Moylan, 1958) agrees with Kendler's S-r-s-R formulation,
however, Goss has been concerned only with verbal mediators.
If mediation is verbal, one might predict that subjects who
could sort stimuli correctly would also be able to give an
accurate verbal report of what they are doing. Numerous
studies show that subjects often cannot give such reports
(Bruner, Goodnow, and Austin, 1957; Green, 1955; Klugh and
Roehl, 1965; Osier and Weiss, 1962; Reichard, Schneider,
and Rapaport, 1944; Rommetveit, 1960; Smoke, 1932; Vygotsky,
1962). Goss explains that verbal reports may be inadequate
because (1) verbal mediation is occurring only during the
initial stages of learning, or (2) because the verbal
mediators may change, or (3) because the subject is describ
ing both mediating and terminating responses.
Wyckoff (1952) is concerned with yet a different
kind of mediating response. He formulated a two-stage
model consisting of an observing response and an effective
‘ response. The observing response was defined operationally j
las that response which "... results in exposure to the
|
pair of discriminative stimuli ..." (p. 431). For exam-
i
pie, the observing response for a pigeon was stepping on a
pedal, which changed a lighted key from white to one of the
to-be-discriminated colors. Such experiments have the
appearance of rigor, however they are not necessarily more
effective than the well designed discrimination apparatus
in assuring that the stimulus will impinge on the organism's
receptors.
Potpourri. Three theoretical positions remain which
cannot easily be fitted under the above rubrics.
Goldstein and Scheerer (1941) developed a number of
sorting tasks for studying concrete and abstract behavior.
They were chiefly interested in concreteness in sorting
behavior as a diagnostic sign of organic brain pathology,
however, the Weigl-Goldstein-Scheerer (W-G-S) color-form
sorting task has also been used to study sorting behavior
in various clinical groups (Silverstein and Mohan, 1962;
Silverstein and Mohan, 1964; Weiss, 1964). Goldstein and
Scheerer maintain that abstraction involves "conscious will"
(p. 23), by which they mean that the subject must be able
either to repeat the sorting under different conditions or
to verbalize the principle he used in sorting. For example,
if a subject can sort the color-form stimuli by shape when
all are white, but fails when the colored sides are turned
up, he is functioning at the concrete level, since he can
not sort correctly under new conditions.
Piaget (Inhelder and Piaget, 1964) has been con
cerned with the development of thinking in children. The
four stages of growth which he has formulated are well
known; his work on the development of classifying behavior
in children is less well known. After discarding language,
maturation, perception, or sensori-motor schema as inade
quate to explain the development of classification, he
concludes, ". . . the central problem of classification
must therefore be the differentiation and progressive co
ordination of extension and intension" (p. 16). By "inten
sion” Piaget means the set of properties which define a
class. By "extension" he means the set of members which
make up a class. But, intension determines what stimuli
will be included in a set, and extension determines which
properties define the set. It is only when the subject is
able thus to coordinate intension and extension that he can
understand the hierarchical relationships necessary for
true classification.
Computer simulation concludes this section. Hovland
(1952) presented an analysis of concept learning in terms
of an information processing model. He later amplified
this model into a simple simulation, suitable for conjunc
tive concepts (Hovland, 1960; Hovland and Hunt, 1960). He
presents a flow chart illustrating a relatively simple
o
TOTE process, by which the computer makes and progressively
refines lists of stimulus characteristics. The concept has
been defined when a final list of characteristics common to
all positive instances remains. The concept problem is
unsolved if no such list is found. The method used by the
computer is inefficient and requires considerable memory
storage, therefore, it is rarely used by human subjects
except when the concept is simple and the number of instan
ces few. More refined simulations have been developed but
they are beyond the scope of the present discussion.
Attention Theory
The notion of attention has been present in psycho
logical theorizing since the early nineteenth century.
Herbart attributed "intensity" to ideas. Wundt experimented
^Test, operate, test, exit.
on the range and fluctuation of attention, considering it
to encompass both simultaneous and successive events.
Titchener considered attention to be the attribute of
"clearness" in sensory processes, an attribute he later
called "attensity." And Cattell utilized tachistoscopic
presentation of various sized groups of objects in a classic
study on the range of attention.
The use of attention in a learning model is rela
tively new. House and Zeaman (House and Zeaman, 1962;
House and Zeaman, 1963; Zeaman and House, 1963) have devel
oped such a model for discrimination learning. Stage one
is the observing response (O); stage two the instrumental
response (R).
Suppose a subject, confronted with a pair of stimu
li; a 3 inch red pyramid to his left, a 2 inch blue sphere
to his right. The subject's attention may focus on the
dimension of shape, size, color, or location, assuming for
simplicity that he attends to only one dimension at a time.
(The authors have abandoned this "one-look" model, House and
Zeaman, 1963). Assume further that shape is the relevant
dimension and that sphere is the correct cue. (A cue is a
value of a dimension.) A subject making a relevant 0^ for
shape will notice the cues, sphere, s^ and pyramid, s^'.
If he makes an incorrect 0, he will be exposed to the
irrelevant cues of that dimension, e.g., red; blue (s2;
s2'). The subject makes an instrumental response and is
rewarded or he is not. If he has made the relevant 0, his
approach to will be rewarded 100% of the time, his
approach to s^' will be rewarded 0% of the time. If the
subject has made an irrelevant O, his approach to cues s2
and s2' will each be rewarded 50% of the time. That is,
assuming that the irrelevant cues are counterbalanced, a
subject who approaches "red," for example, will be respond
ing half of the time to the sphere and half of the time to
the pyramid. Stimulus dimensions which are not required
for sorting the stimuli into categories are called irrele
vant. Experiments have been done in which there were
redundant relevant dimensions, e.g., all triangles were
also blue, all circles were also red (Bourne and Haygood,
1961; Haygood and Bourne, 1964); constant irrelevant
dimensions, e.g., all stimuli are red (House and Zeaman,
1962; Pishkin, 1965); and variable irrelevant dimensions
(Campione, Hyman, and Zeaman, 1965; House and Zeaman, 1962;
Zeaman, Thaller, and House, 1964).
The assumptions of probability learning models lead
to the prediction that the subject will eventually respond
almost entirely to the cue which is rewarded 100% of the
'time. House and Zeaman expand their attention model to
include a mathematical formulation. Parameters are des
cribed which control individual differences: (1) in rate
of acquisition and extinction, (2) in initial probability
of paying attention to various dimensions. Formulae are
developed from these parameters to predict the probability
of a correct response, given direct or indirect reinforce
ment, nonreinforcement, or reversal or nonreversal shift
conditions. The mathematical description is immaterial to
this study and will not be detailed further.
Summarizing, House and Zeaman have described a two-
stage model composed of an observing response and an in
strumental response. The observing response is an inter
vening variable anchored at one end to the stimulus and at
the other to the subject's instrumental response to that
stimulus. 0 is elicited by the dimensional aspects of the
stimulus and results in attention to specific cue values of
that dimension. O and R differ in rate of acquisition and
rate of extinction.
House and Zeaman and their colleagues present evi
dence in support of their theory. Zeaman, House, and
Orlando (1958) suggested that retarded subjects were
'deficient in learning cue-producing responses to the posi-
i I
tive and negative stimuli in a discrimination task. They
had pretrained their subjects on junk^ discrimination to be
sure the subjects were familiar with the basic idea of a
discrimination task. House and Zeaman (1960b) found the
rate of acquisition of a discrimination task to be related
to IQ independently of M.A. They hypothesized that the
differences in acquisition rate were due to differences in
attention rather than to differences in learning the correct
instrumental response, i.e., duller subjects are slow to
learn 0 but having learned O, they learn R as quickly as
do brighter subjects. Zeaman and House develop the hypoth
esis further in their 1966 article. House and Zeaman
(1960a) tried to increase the attention value of the shape
dimension by using three dimensional stimuli instead of two
dimensional patterns. They found that subjects given pre
training on object discrimination were later able to solve
pattern discriminations, whether or not the same shapes were
used in both pretraining and test.
House and Zeaman (1962) trained retarded subjects
on a discrimination problem and then introduced: a reversal
^Multidimensional common objects, e.g., a pan lid,
a soap dish, a toy car.
!shift, a shift to new cues in the same dimension (intra- !
j j
dimensional shift), or a shift to the formerly irrelevant
dimension (extradimensional shift). The extradimensional
shift condition was found to be significantly more diffi
cult than either the reversal or the intradimensional shifts.
The reversal shift and intradimensional shift did not dif
fer significantly, which the authors interpret to mean that
there was positive transfer of the relevant 0, even when
the transfer task required reversal of the instrumental
responses. Campione, Hyman, and Zeaman (1965) did a follow-
up study, designed to demonstrate that the effectiveness of
overtraining was attributable to an increase in the sub
ject's attention to the relevant dimension rather than that
the subject's tendency to avoid the negative stimulus was
diminished, as had been suggested by critics. The study
was designed so that one group of subjects learned a dis
crimination task and then learned a reversal shift. A
second group learned the discrimination task, received 100
trials of overlearning, then learned (1) an intradimensional
shift, and finally a reversal shift of the second task or
(2) an extradimensional shift involving new stimuli and
then a reversal of this second task. Reversal after an
intradimensional shift was learned most quickly, next.
: ;
; i
'ordinary reversal, and finally, reversal after an extra-
i
dimensional shift. Since overlearning occurred on stimuli
\
different from those used in the reversal task, it is highly
unlikely that overlearning would have affected the subject's
response to the negative cue of the reversal task.
Using the "miniature experiment" described by Estes,
House and Zeaman (1963) investigated a phenomena they termed
"compound stimuli." In a task in which two dimensions are
relevant and redundant— for example, shape and color— the
subject may solve the task by responding to the correct
shape cue (e.g., square), or to the correct color cue
(e.g., red). But, theoretically, he may also respond to a
compound, e.g., "red square." The authors found that most
subjects solved the discrimination problem on the basis of
a single cue, however, some subjects, in fact, responded to
the compound stimulus. The ability of the subjects to
respond to compounds led the authors to reject their "one-
look" model in favor of a model which assumes that the
subject may attend to more than one dimension at a time.
Zeaman, Thaller, and House (1964) have followed up
some of the theoretical implications of compound stimuli.
They found that it is easier for subjects to solve discri
mination problems in which the irrelevant cue(s) remains
20
; i
constant, than problems in which the irrelevant cue(s) is !
varied. Bourne et al. (Bourne and Haygood, 1961; Bourne
and Pendleton, 1958; Bourne and Restle, 1959) found task
difficulty to vary directly with the number of both rele
vant and irrelevant dimensions. Bourne hypothesizes that
constant irrelevant cues do not control discriminative
behavior. He assumes that varying the number of irrelevant
cues reduces learning rate because the total number of cues
is increased. In contrast, Zeaman et al. speculate that
constant irrelevant cues, which permit stimulus compounds,
would make learning easier since the number of potentially
relevant cues would be increased. The authors present the
design of an experiment which would test contrasting predic
tions from their theory and that of Bourne et al.
Bower and Trabasso (Bower and Trabasso, 1963;
Bower and Trabasso, 1964; Trabasso and Bower, 1964a;
Trabasso and Bower, 1964b; Trabasso and Bower, 1966) have
applied attention theory to a cue-sampling model of concept
learning similar to Restle's (1955). They assume learning
to be an all-or-none occurrence and, therefore, use a
Markov chain to determine probabilities. Use of the Markov
chain assumes that the probability of a transition from the
initial state to the learned state on any trial, n, is
independent of the subject's performance on trial n-1 .
j ;
While Restle's theory is based on "strategy selection,"
Trabasso et al. postulate: (1) a stimulus selection pro
cess, during which the subject learns to attend to the val
ues of the relevant stimulus dimension; (2 ) a conditioning
process whereby cues are paired with their assigned re
sponses. Stage one is an encoding process, by which the
subject constructs some kind of internal representation of
a stimulus pattern. For human subjects this encoding may
take the form of questions he asks about the stimulus,
e.g., "What color is it?" The "stimulus-as-coded" (s-a-c)
might be the implicit response, "red." The s-a-c is the
unit which enters into response selection and is conditioned
to overt responses.
Bower and Trabasso*s hypothesis of all-or-none
learning has led to an intriguing series of experiments
using multiple reversals. The subject is given a task in
which cards are to be labeled "alpha" or "beta" on the
basis of, for example, size of figure. The subject is told
whether he is right or wrong. However, the subject's
second error is called "right," not "wrong" and the experi
menter shifts, without warning the subject, to reinforcing
the subject for labeling according to the shape of the
|figures. If the subject makes a second error under the new :
conditions, the experimenter responds "right" and shifts
!back to size as the relevant dimension. Since cue-sampling ;
theory assumes that the subject selects a new set of cues
only when told he has made an error, only those "errors"
which the experimenter called "wrong" were counted. Under
these conditions, the subjects learned a multiple reversal
task with as few errors as the subjects given a standard
situation.
In an earlier experiment, Trabasso (1963) sought to
vary attention by using different means of emphasizing the
stimulus dimensions of the Hovland floral designs (Hovland,
1953). The floral designs vary in type of flower, type of
leaf, angle of leaf to stem and position of leaves on the'
stem. Trabasso tried to vary the attention value of a
dimension by adding color to part of the design, by increas
ing the angle of leaf to stem, by making cues redundant or
by keeping the irrelevant dimensions constant. He found
that use of an emphasizer enabled the subjects to learn
faster, even if the emphasizer was a variable cue. For
example, if angle of leaf to stem was the relevant dimen
sion, adding color to the leaves facilitated learning, even
i
if the leaves were sometimes red and sometimes green. On
ja transfer task, where the emphasizer was removed, the
i '
i *
amount of transfer varied with the type of emphasizer. He
concluded that the emphasizer probably had its effect by
attracting attention to a dimension, e.g., type of leaf,
rather than by focusing attention on a specific cue, e.g.,
serrated leaf.
Finally, Mackintosh (1965) has developed an atten
tion theory from discrimination experiments with animals.
He advances the following experimental conclusions: (1)
pretraining an animal on one dimension facilitates learning
a discrimination involving that dimension, while pretraining
to ignore a dimension, that is, to attend to a different
dimension, hinders learning in which the ignored dimension
is now relevant; (2 ) overtraining facilitates reversal
learning of a visual discrimination task but not of a
spatial discrimination. He suggests that the facilitating
effects of overlearning may be due either to adaptation of
the irrelevant cue or to increased attention to the relevant
cue. The mechanism appears more likely to be one of in
creased attention because reversal learning was facilitated,
even when he introduced new irrelevant cues during over
training. Campione, Hyman and Zeaman (1965) using retarded
subjects, came to a similar conclusion. Mackintosh notes
several hypotheses regarding the differences between visual
and spatial discrimination tasks. Only two are of immedi
ate concern: (1 ) the number of irrelevant cues is impor
tant, and visual discrimination tasks contain more irrele
vant cues than do spatial tasks; (2 ) if the relevant cue
should happen, fortuitously, to be a preferred cue, (as is
"position" for rats), then overtraining would have little
effect on its dominance, relative to other cues. Mackintosh
also notes, as does Kirk (1964) that the superiority of
reversal over nonreversal shift can be predicted only if
the rate of extinction of the instrumental response is
faster than the rate of acquisition of an attention response,
otherwise reversal shift subjects would have no advantage
over nonreversal shift subjects.
In summary, some authors differentiate attention
models from orienting response or mediational response
models, although the distinction will not be maintained in
the present research. In the attention model, the subject
is assumed to make an observing response to a stimulus
dimension, which in turn exposes him to the cues of that
dimension. The observing response is an intervening vari
able and can only be measured by the subject's instrumental
response, i.e., his response to one of the stimuli. The
25
|subject is presumed to be responding to a cue (or a com- .
;pound) and it is his response to the cue which is rein
forced. Instrumental responses are assumed to extinguish
at a faster rate than are observing responses.
Studies of Concept Formation
in the Retarded
A number of studies have investigated concept for
mation in the retarded. These studies have been concerned
with: the effect of age and intelligence on concept learn
ing, the W-G-S Color-Form sorting test as an experimental
and a clinical tool, other concept formation tests as diag
nostic tools, the acquisition of semantic concepts, and the
role of mediation in concept learning.
Hull's (1920) comprehensive study of concept learn
ing involved a comparison of the performance of "constitu
tional inferiors" and other diagnostic groups, with that of
normals. He found that the retarded took considerably
longer to evolve concepts and were almost entirely lacking
in ability to define concepts. The degree of retardation
of Hull's subjects cannot accurately be determined.
Hermelin and O'Connor (1958) studied institutional
ized retardates (mean IQ = 40) with a type of discrimination
task. The subject was presented three pairs of pictures.
|For each pair, the correct picture was the one representing
i
a given concept, e.g., animal. Several such sets were
jgiven to each subject, however, for one set, the correct
stimulus was chosen arbitrarily by the experimenter, i.e.,
there was no concept the subject could use to help him
learn the discrimination. It was found that the subjects
could learn discriminations based on concepts more readily
than discriminations having no basis for choice.
Mishima and Tanaka (1966), dividing subjects by age
(6 , 10 and 14 years) and by WISC vocabulary score (high and
low) found both age and intelligence were significant in
determining the subject's score on an analogous progression
type of concept task.
A number of studies have investigated the perform
ance of retardates on the color-form sorting test of Weigl,
Goldstein, and Scheerer (1941). This test utilizes 12
stimuli which vary simultaneously in shape and color.
Shapes are round, square or triangle; colors are red, blue,
green, or yellow. The forms are colored on one side and
white on the reverse. The colored face is turned up and
the subject is asked to sort the pieces first one way, and
then "a different way" (called shifting). Halpin (1958)
studied 7-14 year old subjects (mean IQ = 60). She found
r" ................. 27 " i
i ;
the subject's preference for sorting by form or color was
not related to M.A. or to C.A., however, more subjects
sorted by color than by form. Ability to shift was related
to both M.A. and C.A. Halpin and Patterson (1954) found no
differences between brain injured and familial mentally
retarded children (mean C.A. = 121 months, mean IQ = 58) in:
the number of subjects able to shift, the number of subjects
who could be induced to shift, and the number of subjects
using form or color. The only significant difference found
was that the familially retarded subjects tended to make
patterns with the stimuli more often than did the brain
injured.
Korstvedt, Stacey, and Reynolds (1954) used a modi
fication of the W-G-S, adding another shape value (diamond)
and two more color values (black and white). With normal
(mean IQ = 109), borderline (mean IQ = 74) and mildly
retarded (mean IQ = 60) subjects, all aged 15 to 18, it was
found that normal subjects were more successful than the
retarded in sorting by both color and form, and more suc
cessful in verbalizing the sorting principle. The differ
ences between borderline and mildly retarded subjects was
not significant. Borderline retardates showed a preference
for form over color; the mildly retarded and normals showed
28
i :
I i
no such preference.
Silverstein and Mohan (1962, 1964) investigated the
relationship between performance on the W-G-S Color-Form
test and a number of organismic variables. With a popula
tion of hospitalized retardates (median IQ = 56, median
age = 35 years, median length of hospitalization = 15 years)
jthe investigators found no relationship between the vari
ables: age, sex, IQ, or length of hospitalization, and
variables of test performance: use of patterning, ability
to shift, ability to verbalize the sorting principle, or
percentage selecting color (form) in the initial sort.
Other studies have used different concept sorting
tasks as a diagnostic tool. McMurray (1954) used card
stimuli varying along three 3-valued dimensions: color,
form, and number. Subjects were either brain injured or
"endogenous" mentally retarded children. Subjects were
reinforced for sorting by color. When the subject reached
criterion, a shift was made, first to form and then to num
ber. Subjects were not told of the shift. Brain injured
subjects took more trials to shift than did the endogenous
mentally retarded.
Siegel (1957) developed the visual-verbal concept
formation test. A card containing four designs is presented
i 29
; i
|to the subject, who is asked to pick the three which "go
together," and then a different group of three. For exam
ple, if a card displayed two blue triangles, a blue circle
i
and a green triangle, the concepts would be, "three tri
angles" and "three blue objects." Failure to shift to a
second grouping was scored as a "single miss.” The subject
was also required to verbalize the sorting principles.
Failure to do so was scored a "double miss." The "single
miss" score differentiated M.R.'s from normal subjects and
from brain damaged subjects. The "double miss" score
differentiated within IQ levels, (correlation with IQ =
-.56), as well as between diagnostic groups: M.R.1s, brain
damaged, schizophrenics and normals.
Zaslow (1961) developed a test requiring the sub
ject to recognize a continuum. There were fourteen stimulus
figures, ranging from an equilateral triangle, through
figures with successively bowed sides to a perfect circle.
Subjects were asked to arrange the stimuli, with and without
the experimenter placing the triangle and the circle as
anchor points. Next, the experimenter arranged the designs
in a continuum and asked the subject to delimit the bound
aries of the concepts circularity and triangularity. Few
retarded adolescents (mean C.A. =14, mean IQ = 57) and few
I normal second and third graders were able to produce the
continuum. A high percentage of seventh and eighth graders,
high school students and college students produced continua
(81, 8 8 , and 89% respectively). A group of paretics (mean
C.A. =52, mean IQ = 80) and the second and third graders
tended to constrict the limits of the triangularity concept.
Mental retardates were either too constrictive or too
flexible, simply dividing the continuum in half. The con
cept of circularity did not differentiate groups. Zaslow
concluded that the concept of a continuum did not occur
until about age eleven. However, Piaget (Inhelder and
Piaget, 1964) found children able to produce seriations
based on length by ages seven to eight. Stacey and Cantor
(1953) repeated Zaslow's test with borderline (mean IQ = 73)
ancLmildly retarded subjects (mean IQ = 61). Their group
of mildly retarded subjects was comparable to Zaslow's
retarded group. They found that when the concept was pre
sented by the experimenter, more borderline than mildly
retarded subjects could grasp it.
Turning to work on semantic concepts, Stedman (1963)
studied semantic clustering in normal and retarded subjects
(C.A. = 15-34, IQ = 43-83, mean IQ = 64) by presenting the
subject with pairs of words to be read aloud several times,
and then recalled. Both normal and retarded subjects tend j
j l
! to cluster common types of words together In their recall. 1
The total number of word pairs recalled was smaller for the j
retarded subjects than for normal subjects. The retarded
recalled more coordinate and contrast clusters, whereas
normal subjects recalled more synonym, supraordinate, and
"action-of" clusters.
Griffith et al. (Griffith, 1960; Griffith, Spitz,
and Lipman, 1959) required subjects to abstract a verbal
concept which would relate three words. For example,
mountain, elephant and house may all be described as "big."
Subjects were asked to define the three words and were ques
tioned until the definition included the relevant concept.
For example, an elephant is: an animal, has a trunk, is
gray, and is big. Griffith found that normal seven year
olds, and retarded young adults (mean C.A. about 16) having
IQ's under 65 needed to have included the relevant abstrac
tion in their definition of two out of three, or four out
of six words for which a concept was requested before being
able to apply the concept to the whole group of words.
Brighter retardates (IQ above 65) and normal nine year olds
were able to report the abstraction if one of their defini
tions had included the concept. The authors concluded that
{subjects with higher M.A.'s are able to make more efficient
iuse of verbal mediators.
Milgram (1966) "matched" normal (mean C.A. = 6.1),
ieducable (mean M.A. = 6.1) and trainable (mean M.A. - 5.4)
subjects on the basis of mental age. The subject was pre
sented with a set of seven cards and asked to select the
"three that go together." If the subject could not do so,
the experimenter indicated the correct cards and presented
the next set. Eighteen such sets were given. The task was
difficult, even for the normal subjects and failed to dif
ferentiate between the groups. Subjects did differ in
their ability to verbalize the concept once the experimenter
had indicated the exemplars, "trainable" subjects giving
less adequate definitions than the two groups characterized
by the higher IQ's.
Johnson and O'Reilly (1964) varied the availability
of verbal mediators by presenting to the subjects either
pictures of birds, which varied in type of beak, color of
wings, and color of tail; or cards containing verbal phrases
describing the same dimensions varied in the pictures,
e.g., "red wing, hooked beak, orange tail." Normal twelve
year olds learned concepts based on the verbal phrases in
fewer trials than when the same concept was presented
pictorially. Rieber (1964) set up conflicting mediators by !
pretraining normal and retarded children (mean IQ = 70 for
M.R.'s) to associate line drawings to color names, (e.g., a ,
picture of a dog was paired with the word RED) and then to
associate colored buttons with slide pictures. For one
button, the color and picture associated together were
identical to the paired associates task (e.g., red button,
dog), for two other buttons the color and picture to be
learned were the reverse of pretraining, and for the fourth
button, a new color and new picture were used. Reversing
color and picture resulted in negative transfer for both
groups, with the retarded making more errors than the normal
subjects. Where the pairing was unchanged, both groups
showed positive transfer, indicating that pretraining was
effective in manipulating mediation. Semler and Iscoe
(1965) also found that conceptually dissimilar pairs of
words were learned more slowly than conceptually similar
pairs of words by retarded subjects, in a paired-associates
task.
Milgram and Furth (1963) studied the acquisition
rate of verbally mediated concepts, comparing educable
mentally retarded subjects (mean IQ = 70) with normal sub
jects. Subjects learned discrimination tasks based on
{sameness (two identical versus two different figures);
symmetry (symmetrical versus asymmetrical figures); and
opposition (biggest or smallest of two disks). Mental
retardates and normals performed equally well on "same";
M.R.'s were slightly better than normals on "symmetrical";
normals were significantly better than mental retardates on
"opposite." The authors concluded that normals were superi
or on opposition because it is verbally mediated. Mental
retardates do about as well as normal subjects when the
tasks can be mediated perceptually.
Ball (1964) investigated perceptual mediation in
concept formation. Subjects were mildly retarded patients
in a state hospital (mean IQ = 71). The "phi" technique
was used to test the subject's ability to discriminate
between a diamond and a triangle, and between a diamond and
a square. Four trials of a sorting task were then given
with "roundedness," "acute angles" and "four sidedness" as
the categories, and oval, triangular and quadrilateral
designs. The phi test was again administered. Subjects who
passed the sorting task showed a decreased threshold for
diamond-square discrimination and an increased threshold
for diamond-triangle discrimination. Subjects who failed
the sorting task did not show such a shift. Ball concluded
|that perceptual reorganization had occurred as a result of
the sorting task, with the subjects coming to regard diamond
as "four sided" rather than "angular."
In a theoretical article, Spreen (1965) discusses
the role of verbal mediators in the learning of the mentally
retarded. He suggests that M.R.' s may have a specific
deficiency in verbal mediation. However, he also points
out that age and degree of concept mastery interact. When
a concept is not well learned, an age deficiency in the
verbal mediation of a discrimination task is found. If
the concept is well learned, no such age differential occurs.
He suggests the necessity for studying the effects of such
variables as anxiety or incentive, as well as M.A. and type
of verbal task.
Findings of the mediation studies and the attention
models suggest that training the subject to respond to a
dimension should facilitate concept learning in which that
dimension was relevant. Bensberg (1958) investigated
"attention sets" in mentally retarded males (mean C.A. =
19.5, mean IQ = 47). Subjects first learned to associate
one of three buttons with figures on a card. Designs
varied in shape (circle, square, triangle) and color (red,
blue, yellow). One group learned to associate the buttons
36
with the shape cues, the other with the color cues. Half
of each group learned to criterion and half were given over
training. A control group learned to associate orally pre
sented animal names with the three buttons. All groups
were then given a transfer task in which the subject was to
learn which of three nonsense syllable names was associated
with each of nine irregular colored designs. Three differ
ent designs represented each shape value: rounded, jagged
and convex, and a nonsense syllable was paired with each
shape value. Color was irrelevant and had the values:
green, brown and gray, three shades of each. All subjects
received 1 2 presentations of the set of nine stimuli (108
trials). Total errors were recorded, the subject being
credited with an error if he responded incorrectly or if he
failed to respond within 4 seconds. As Bensberg predicted,
the subjects pretrained on color made more errors them the
control group. The control group, in turn, made more errors
than the group pretrained on form. Bensberg concluded that
attention set (or observing response) can be manipulated,
and that making a relevant observing response increases the
probability of a correct instrumental response.
Bensberg's results are as predicted, however, there
are certain methodological criticisms of the study. After
3 7 j
i i
Ipretraining, at the introduction of the transfer task, the
subject was reportedly "... shown one picture represent
ing each of the three basic forms and told the name of the
picture" (p. 140). It is not stated whether the three
examples presented were all of one color or were represent
ative of the three basic colors used in the experiment. A
subject's attention to color might be expected to be
different if exemplars of only one color were presented, in
contrast to having exemplars of all three colors. Second,
the effects of "attention set" on the error score are con
founded with errors caused by the difficulty retarded sub
jects have in learning nonsense syllables. Bryant (1965)
found that retardates who sorted the stimulus materials into
boxes learned a concept task based on shape more quickly
than did a peer group required to learn nonsense syllable
labels for the same concept. Third, no indication is given
how many subjects learned the transfer task; only error
scores are given. It would be predicted, and it was found,
that subjects trained to pay attention to color would make
more initial errors on a concept task based on form than
would subjects trained to attend to shape. However, do
subjects pretrained on color take significantly more trials
to learn a concept task based on form than do subjects
38
i - ;
pretrained on form, or perchance does the difference in
i 1
errors between the two groups diminish after a trial or two,
with no significant difference between trials to criterion
scores for the two groups? Bensberg's article does not
present data which would answer these questions.
Bryant (1965) taught moderately retarded subjects
to sort by shape, with size as the variable, irrelevant
dimension. Half of these subjects then learned an intra
dimensional shift task, with color a new, irrelevant
dimension. The other half of the subjects learned an extra
dimensional shift, with color relevant and size irrelevant.
Both groups of subjects learned the transfer task in fewer
trials than they had required for the original task, how
ever, they did not differ significantly from each other.
Control groups, which learned only the "shift" tasks, re
quired significantly more trials than the experimental
subjects. Since the two experimental groups were equally
facilitated in learning the transfer task, it may be
hypothesized: (1 ) that transfer was the result of non
specific learning, i.e., a general "warm up" effect; or
(2 ) that learning to ignore an irrelevant dimension (group
2 ) is just as important in a transfer task as learning to
respond to a relevant dimension (group 1). Although Bryant
39
I
called the second task "positive” or "negative" transfer,
i
there is no negative transfer in the sense of interference
from a previously relevant dimension. Therefore, the fail
ure to obtain a significant difference between the experi
mental groups does not necessarily conflict with attention
theory.
Formulation of the Problem
The work of Bensberg (1958) and of Bryant (1965)
may be reformulated in terms of an observing response model.
The effect of pretraining would then be explained as chang
ing the probability that the subject would make a relevant
observing response. The probability of a relevant observ
ing response might be either increased or decreased,
depending on the kind of pretraining. If the subject made
a relevant observing response, the probability would
increase that he would respond to the cue of the relevant
dimension, by definition of the observing response.
The present investigation is concerned with manipu
lating pretraining in mentally retarded subjects to deter
mine whether such training would affect the rate of
acquisition (trials to criterion) or number of errors in
a concept sorting task. Specifically, one group was given
40
I ,
jpretraining which was designed to increase the probability
i '
that the subject would make an observing response to the
relevant dimension— shape; the other two groups were control
groups.
Hypothesis
In a group of retarded patients, the rate of acqui
sition of a concept task can be increased through pretrain
ing in making a relevant observing response.
Prediction. In hospitalized retardates, trials to
criterion on a concept sorting task will be less in a group
which has had pretraining designed to increase the probabil
ity of a relevant observing response than in a group which
has not had such training.
I
I
CHAPTER II
METHOD
Experiment I
Sub jects
The subjects (referred to in this chapter as Ss),
24 females and 42 males, were all patients in a state hos
pital for the mentally retarded. Hospitalized retardates
were used because they were readily available to the exper
imenter. Only hospitalized Ss were used because Zigler's
results (1962, 1963) show that hospitalized retardates are
more responsive to social cues and to social reinforcers
than are non-hospitalized retardates. All but one1 S had
been hospitalized for at least ten months. The upper limit
for hospitalization at the present facility was thirteen
years, however, some Ss had been hospitalized previously.
Subjects were not equated for length of hospitalization,
since Silverstein and Mohan (1962, 1964) found no correla
tion between length of hospitalization and performance on
10ne S had been hospitalized only four months.
41
jthe W-G-S Color-Form test in a population of hospitalized
i
- retardates.
Subjects ranged in age from 13 years through 38
years and were classified in the mildly or moderately re
tarded IQ range— AAMD classification, "-2" or "-3" (Heber,
1959). (See Table 1 for a summary of descriptive data.)
I
All Ss were screened with the Peabody Picture Vocabulary
Test (PPVT), Form A whether or not they had currently valid
test results, in order to obtain comparable M.A. scores.
The PPVT was chosen as the selection device because it is
quick, is easily understood by retarded Ss, requires no
verbal response from S, and has norms based on a retarded
population. Only Ss earning M.A.'s in the range four years
to ten years were used in the study. Task difficulty dic
tated the lower M.A. limit, availability of Ss the upper.
Subjects were divided into three groups on the basis of
M.A.: 4 years, 0 months (4-0) to 5-11, 6-0 to 7-11, 8-0 to
9-11. Subjects from each M.A. level were randomly assigned
to one of the three treatment groups by the hospital's
Director of Research. Initially 10 Ss were assigned to
each treatment group, 4 each from M.A. groups 4-5 and 6-7,
and 2 from M.A. group 8-9. Experiment I was replicated
following Experiment II, however there were not sufficient
43
TABLE 1
DESCRIPTIVE DATA BY TREATMENT GROUPS
FOR EXPERIMENTS I AND II
Measure
Experimental Group
I
Exposure
II
Dot Matching
III
Form Matching
Ma SD Ma SD Ma SD
Experiment I
C.A. 24.6 6.06 2 1 . 6 5.33 23.1 6.25
M.A. (years) 6 . 8 1.47 6 . 8 1.43 6 . 8 1.23
IQ
51.3 7.14 51.8 8 . 1 2 51.3 6.64
Experiment II
Mb SD Mb SD Mb SD
C.A. 22.76 5.91 2 1 . 0 5.03 2 2 . 0 4.56
M.A. (years) 6.4 1.39 6 . 8 1 . 6 8 6 . 6 1.54
IQ
49.8 7.85 50.8 8.93 48.9 8.79
Note: M.A. and IQ obtained from the PPVT.
aN = 22
bN = 10
t
Ss 1:o permit a complete replication nor separate analyses |
of the data obtained from the replication. Therefore, data
from the replication were pooled with those from the ini-
i
tial experiment, yielding an N of 2 2 for each treatment
condition. (Six Ss were from the M.A. 4-5 level, 11 from
the M.A. 6-7 level, 5 from the M.A. 8-9 level.) An attempt
was made to balance sex distribution. There were 7 females
in treatment group I, 8 in group II and 9 in group III.
Subjects were equated for M.A. level because reviews
of the literature (Rosenberg, 1963; Zeaman and House, 1966)
indicate a relationship between performance on intelligence
tests and performance on learning or problem solving tasks.
A pilot study on the present population found a rank order
correlation (p) of -.45 between trials to criterion and
M.A., suggesting that experimental variance could be reduced
significantly by matching Ss on the M.A. variable.
Diagnostic category per se was not used as a selec
tion factor. The research literature is in disagreement
regarding the appearance or non-appearance of differences
between "brain-injured" or "exogenously" retarded children
and "familial" or "endogenously" retarded children. Gold
stein and Scheerer (1941); McMurray (1954); and Rose,
Smith, and Robles (1964) found significant differences;
; ......... 45 *
j .
iHalpin and Patterson (1954) and Meyers, Dingman, Attwell
i
jand Orpet (1961) found none. Since the diagnosis of brain
damage in a retarded patient is often based on unreliable
history or equivocal neurological findings, and the diag
nosis of familial is subject to equal limitations, there
seemed to be no useful purpose in pursuing diagnosis fur
ther. Blind, deaf, or severely spastic patients were not
included. A few non-ambulatory patients who were able to
use their hands were included. All Ss were able to verbal
ize, at least minimally, although a few refused to do so and
simply nodded or shrugged their responses during the
inquiry. If a S was uncooperative during initial screening
with the PPVT, he was not included in the experiment.
Apparatus
Apparatus consisted of three sets of pretraining
materials and cards for the concept sorting task.
For Group I (Exposure), the following stimuli were
used:
1. An 8* s x 11 inch sheet of paper on which had been
mimeographed four figures, one in each quadrant:
a cross, an equilateral triangle, a circle and
a hexagon. Figures were roughly two inches
high. (See Figure 1.) j
2. Four 2*s x 2*s inch white cards, each with one of
the same four geometric designs. Figures were
drawn in black India ink and were approximately
1*5 inches high.
3. Four two-inch square white cards having one,
two, or three small, circular spots in hori
zontal array in the center of the card. Two of
the cards displayed two dots. The dots were
drawn free hand in black India ink and were
approximately four millimeters in diameter.
4. A white cardboard box, 4*s x Ah x 3h inches,
with a 2*5 inch diagonal slit in the lid.
For Group II (Matching training) materials consisted
of the four dot cards referred to in (3) above and a 2 x 8
inch rectangle of white cardboard on which were drawn, in
black India ink, five groups of dots. Reading from left to
right, groups consisted of: two, one, three, two, and one
dots. Dots were the same size and in the same horizontal
array as on the two-inch cards.
Stimuli for Group III (Form) were: (1) the mimeo
graphed sheet described for Group I; (2) the four geometric
design cards, also described for Group I; (3) a rectangle
47
Figure I. P retrain in g material for Group I
(Exposure) and Group 111 (Form).
!(12 3/4 x 2% Inches) on which were drawn, reading from left j
i I
to right, a circle, an equilateral triangle, a parallelo
gram, a hexagon, a right angle triangle, and a cross.
Figures on the rectangle were drawn in black India ink,
were approximately I3 * inches high and were spaced approxi
mately inch apart.
Stimulus materials for the concept sorting task
consisted of fifteen cards and three wooden trays to hold
the cards. The stimulus cards were three-inch squares of
white cardboard, to the center of which (front and back)
were affixed designs cut from Colormatch (Craft Tint Manu-
o
facturing ), a commercial colored paper. The cards were
covered with 600 gage self-adhesive transparent plastic.
Each card represented one of three shape values (S^, S2, S^)
and one of three color values (C^, C2, C3>. Five different
designs were used for each shape value. For (four sided)
the designs were: a square, a rectangle, a parallelogram,
and two different trapezoids. For S2 (angular) there were
four modified Chinese characters and a modified Greek
letter "ir." The S3 (blob) designs were irregular, amorphous
shapes. The five designs used for each shape value had
218501 Euclid, Cleveland, Ohio, 44112.
been modified during -the pilot study until no shape was
1 systematically selected, assuring a kind of equivalence
ibetween the designs. Four shades, ranging from light to
dark, made up each of the color values, green, blue and
3
gray. (See Figures 2 and 3.) Five shades had originally
been chosen, however, in the pilot study, Ss were able to
discriminate only four. It was not found possible to
choose five discriminable shades without varying the hue as
well. Bensberg (1958) and Hermelin and O'Connor (1958)
found that retardates are able to learn concept tasks in
which nonidentical exemplars of the concept are used.
An attempt was made to match form and color cues
for ease of mediation. The author assumed that retarded Ss
would have little difficulty giving some kind of verbal
label to the S^ and S3 shapes but assumed S2 (angular)
designs would be somewhat more difficult for retardates to
label. Therefore, two color values were chosen which, it
was assumed, the retardates could name fairly easily (green
blue) and an achromatic color (gray). It was assumed
retardates would know the names of primary colors better
than the names of achromatic colors.
3Greens were #167, 171, 175, 176; blues #123, 130,
131, 132; grays # G-7, G-ll, G-13, 198.
50
GREEN, BLUE GRAY,
7
GREEN2 BLUE2
gray3
Figure 2. Stim ulus cords for the concept sorting
tosk. (Compore with Figure S.)
51
GREEN BLUE
G R A Y , GREEN
G RA Y ,
SET II
Figure 3. Auxiliory cord* used for labeling concept
categories ( SET II ) and for testing concept generalization
(SET III).
I The fifteen stimulus cards were divided into three
'unequal sets. Set I, consisting of nine cards representing
ithree shape and three color values, made up the manipulanda
to be sorted by S. (See Figure 2.) Set II consisted of
three cards which served as labels for the sorting trays.
The colors in Set II had duplicates in Set I, since there
were only four shades for each color value. However, care
was taken that the duplicate colors occurred in designs
having different shape values. For example, the shade of
gray used for S3 in Set II had been used for S^ in Set I.
(See Figure 3.) Set III, the cards to be used for testing
S's ability to generalize the concept, was composed of the
three remaining cards. The three shape values and the
three color values were represented by new shapes and new
shades differing from those in Sets I and II. (See Figure
3.)
The sorting trays, made of quarter inch plywood,
were wedge shaped, measuring 4*s x 4% x 1% inches, and pre
sented a slanting surface on which S could place the card.
To prevent the card from sliding off, a 2.5 x 2.5 x 55
millimeter strip of wood was glued one quarter inch above
the lower edge of the tray and midway between the two sides
A similar strip was glued three quarters of an inch below
jthe top edge. The latter strip served to support the label i
i !
card, as did a rectangle, 2% x 4% x % inches, tacked to the ;
back of the wedge. The whole tray was painted a flat white.;
Procedure
4
All Ss had seen E or the research assistant during
screening with the PPVT and had been told they might be
selected to help E with a "game" later. The subject was
brought to one of the offices at the front of his home
ward. A display of prizes: toys, candy, trinkets and coins
was shown to S and he was allowed to select the prize for
which he would work. The experimenter then left the room
and the research assistant administered the pretraining.
The experimenter did not know S's group membership until
the conclusion of the experiment. The decision to have
someone other than E do the pretraining was necessitated by
the finding, during pilot study, that Ss were responding
directly to E's enthusiasm (or lack of it) in reporting
"right" or "wrong." The experimenter's ignorance of S's
^The author gratefully acknowledges the cheerful,
patient, indefatigable help of Mr. Bert Itoga, summer
Student Professional Assistant, during the screening and
pretraining phases of this project.
5Verbatim instructions for the pretraining and con
cept sorting tasks are given in the Appendix.
54
; j
treatment group avoided direct contamination of the results.!
! j
Pretraining for Group I Ss was designed to control
for possible facilitating effects of "warm-up," rapport
building, or exposure to the experimental materials not
related to the pretraining in form matching. The subject
was given the mimeographed page and asked to fold it as
demonstrated. He was then given the cardboard box and asked
to put into it the small cards displaying the geometric
figures and the dots.
Pretraining for Group II Ss was designed to control
for the possibility that the pretraining was effective
because it taught S to match— to make comparisons, rather
than because it elicited the relevant observing response.
The rectangle displaying the sets of dots was placed before
S and he was handed, one at a time, the four small dot
cards, beginning with the card containing one dot, and asked
to, "Find a set that looks just like this." The set of
four cards was presented twice.
Pretraining for Group III Ss was designed to in
crease the probability that S would make a relevant observ
ing response. First S was given the mimeographed page and
a black waxed pencil and asked to trace the figure. Tracing
was used rather than copying because the cross and hexagon
Iwould be boo difficult for the lowest M.A. group to copy.
Next the rectangle displaying the six geometric figures was
placed before S. The small cards were handed to S, one at
a time, with the instructions, "Find the one that looks
just like this." The four cards were presented twice,
beginning with the circle. Smith and Means (1961) found
that transfer to a discrimination task was higher when S
traced the designs than when he matched them. Tracing and
matching were expected to reinforce each other in the pres
ent experiment. On the basis of the findings of a pilot
study, it was assumed that the tracing and form matching
would be effective in evoking an observing response for
form. A group of 6 subjects— 3 male and 3 female— had been
given the tracing and form matching described above. Six
other subjects received no such training. These 12 pilot
subjects were then given a matching task in which a stimulus
card was presented and S was asked to select the one of two
response cards "which goes best with it." One response
card matched the stimulus card in shape and one in color.
(It was assumed that form and color would be the predominant
responses in S's hierarchy.) The stimulus cards were
selected from Set II (labeling cards), and the response
cards from Set I (sorting cards). A graeco-latin square
1 56
i
I
jdesign permitted each label card to be paired with every
other appropriate response card. Each S received two such
matching tasks.
All the retardates were able to match the cards,
although normal adults find the task difficult because they
immediately see the two possibilities, form and color. A
few of the brighter retardates also seemed to be aware of
the conflicting cues. A few merely matched by color; seem
ing to be completely unaware of the form aspects.
A 2 x 2 contingency table, with pretraining or no
pretraining as classes and with "no form response" or "one
or more form responses" as scores, was analyzed using the
Fisher exact probability test (Siegel, 1956) . The proba
bility of the obtained distribution of frequencies under a
null hypothesis was .05, suggesting that pretraining did
establish the relevant observing response.
The subjects made virtually no errors in any of
the pretraining tasks, assuring that the groups had approx
imately equivalent reinforcement histories immediately
preceding the concept task.
At the conclusion of the pretraining, the research
assistant called E back into the room and then left. The
concept task proceeded as follows: The three trays were
iplaced before S and the three label cards (Set II) were j
I i
placed in their holders. Reading from left to right, as
;the trays faced S, the label cards were: a blue, four-sided
design, a green angular figure, a gray blob. Use of label
ing cards eliminated errors which might result simply
because S could not remember onto which tray a particular
shape value was to be placed. A standard set of instruc
tions was read to S and he was handed the stimulus cards to
sort.
A sorting task was chosen rather than a labeling
task (having S learn some kind of name for the stimulus)
because Bryant (1965) found that retarded Ss were able to
learn a concept task in fewer trials when the Ss manually
sorted the stimuli than when the Ss categorized the stimuli
by verbal labels. Second, retarded Ss are generally less
distractible when they can handle materials directly, than
when required merely to observe. Third, requiring a verbal
response would deleteriously affect the motivation of some
Ss, since retardates are often reluctant to use the verbal
skills available to them.
A simple concept task was chosen, with one relevant
dimension (shape) and one irrelevant dimension (color). A
conjunctive concept, e.g., color and form relevant, was
58
t
! |
presumed -too difficult for retarded Ss. A single irrelevant^
i i
dimension was used in order to limit task difficulty. Task
difficulty has been found to increase with the number of
irrelevant dimensions (Bourne, 1957; Bourne, 1963; Bourne
and Restle, 1959; Glanzer, Huttenlocher, and Clark, 1963;
iHovland, 1952). Designs were to be of equal height, how
ever, pilot study suggested that total mass appeared to be
more important than height, so the more solid designs were
reduced in height to equate for mass.
Form was arbitrarily chosen as the relevant dimen
sion. There is disagreement in the literature concerning
the relative ease of color and form. Bourne and Restle
(1959) found color more discriminable than form. Grant,
Jones, and Tallantis (1949) and Heidbreder, Bensley, and
Ivy (1948) found no significant differences between color
and form. Goldstein and Scheerer (1941); House and Zeaman
(1962); Korstvedt, Stacey, and Reynolds (1954); Ohlrich and
Ross (1966); Reichard, Schneider, and Rapaport (1944);
Sanders, Ross, and Heal (1965); and Zeaman, Thaller, and
House (1964) found a preference for form over color, with
various tasks and various subject populations.
After every response, E announced whether the choice
was right or wrong. Immediacy of feedback has been found
jto be an important variable in the rate of learning (Bourne,
;1957; Jacobs, 1950) and additive effects have been found
when both rights and wrongs are reported (Buss and Buss,
1956; Newman, 1966; Sullivan, 1964). An attempt had been
made to mechanize feedback of information using lights to
indicate correct and a buzzer to indicate a mistake. It
was found in the pilot study that some Ss had difficulty
learning or remembering the meaning of light and buzzer.
Frequently S would forget or interchange the meaning of the
signs. Because variability was being increased, rather
than decreased and since retardates are reportedly highly
responsive to social cues (Zigler, 1963), the light and
buzzer system was discontinued. The subject was required
to correct his errors. Because of the number of stimuli to
be sorted, it was considered advisable to use the correc
tion technique, even though House and Zeaman (1958a, 1958b)
found no differences attributable to the use of correction
or non-correction in a two choice discrimination task. An
error was scored each time6 S put a card in the wrong tray.
6To some of the Ss it was not transparently obvious
that if two trays were wrong in a three choice discrimina
tion, the third was correct. In order to avoid spuriously
inflating the error scores of such Ss and to avoid discard
ing them, no more than four errors per card per trial were
counted.
60
After the third error with the same card, S was told, "No,
;it doesn't 22 there. Try a different tray." If S failed
to correct an error, he was reminded, "That one is wrong.
Take it out and try it some place else." Occasionally a
S would remove the card after a correct response. Such Ss
were told, "That one is right. Leave it there and try the
next card."
When S had correctly placed all nine cards (one
trial), E picked up the cards in scrambled order, making
certain that no two cards representing the same shape value
were in consecutive order. Varying the order of presenta
tion eliminated serial position effects. The deck of nine
cards was presented to S until he achieved criterion— two
errorless trials in succession— or until he had received a
total of 20 trials in any one day. If the twentieth trial
was errorless but not criterion, S was given a twenty-first
trial. The subjects were seen on successive days until
they achieved criterion, or reached a total of 60 trials.
Subjects who did not learn within 60 trials were considered
to have failed.
^Initially a limit of 40 trials was set. Pilot
study suggested that 40 was a reasonable limit. A subse
quent loss of Ss led to modifying the limit to 60. As a
consequence, some Ss were reported to have failed at 40
trials, whereas others were successful at 50+ trials.
j If it was necessary to see S in a second or third \
j - i
isession, a modified form of pretraining was administered,
since it is unreasonable to assume that the observing
response for shape would remain constant from one day to
the next. For Group I, paper folding was eliminated. For
Group II, the set of dot cards was presented only once.
For Group III, tracing was eliminated and the form matching
cards were presented only once. It was not known exactly
what effects the modification in pretraining would have,
nor what learning effects might occur over the course of
several days. However, since the number of Ss requiring
two or three sessions was approximately equal for the three
experimental groups, any effects should be distributed
randomly over experimental conditions.
After S achieved criterion, the nine card deck was
removed and the Set III cards were handed to S, one at a
time, in standard order (S3 , S2» S^), with the question,
"Where should this one go?" Notation was made of S's place
ments but no feedback was given. The generalization cards
were presented as a test for S's understanding of the con
cept. It is possible that S had merely learned an arbitrary
sorting task, i.e., that this card goes in that box, but
not that four-sided designs go in that box. Goldstein and
Scheerer (1941), and Hunt (1962) do not consider a S to
ihave learned a concept unless he can apply the concept to
new stimuli or under altered conditions. Use of a non
verbal task provided a check for Ss who could not verbalize
the concept.
Finally, S was interviewed to try to determine
whether he had identified the intended concepts. A stand
ard interview schedule was not used because of the wide
range of verbal abilities found in the population used in
the present study. A modified schedule was used, with the
initial questions being quite general. If S could not
answer, the questions became more focused unt4. 1 * if neces-
• »
sary, E would suggest a name for the group of cards to see
if S would accept it. Although Hunt (1962) considers
verbalizing the classification rule to be essential to
mastery of a concept, there is considerable research which
shows that Ss, even normal' • adults, are able to use a con
cept correctly without being able to verbalize the rule
(Bruner, Goodnow, and Austin, 1957; Green, 1955; Klugh and
Roehl, 1965; Osier and Weiss, 1962; Reichard, Schneider,
and Rapaport, 1944; Rommetveit, 1960; Smoke, 1932; Vygotsky
1962).
63
At the conclusion of the experiment, S was given \
i
the prize he had selected, even if he had failed to learn
the concept task.
Experiment II
Initially it had been proposed to repeat Experiment
!l without the criterion cards labeling each category. It
had been hypothesized that such cards might serve not only
a memory function but might also cue S as to the possible
relevant dimensions. That is, given a single card, a £
might consider color, form, size, intensity, height, or
position on the card to be potentially relevant. Having
three labels present should suggest to S that color or form
were relevant, since three values of color and three values
of form were displayed, while no such values were presented
for the other dimensions. In practice, the difference would
probably be academic, since only a few of the brighter Ss
would be able to formulate so many hypotheses. Even grant
ing the formulation of a number of alternatives, the re
search findings which have been quoted would lead to the
expectation that color or shape would be dominant in the
hierarchy of retardates. Finally, few of these retarded Ss
would be expected to process so efficiently the information
j 64 !
I !
|available from the labeling cards. Therefore, faced with
what seemed to be an alarming rate of failure in Experi
ment I, an attempt was made to simplify the task by reducing
from nine to four the number of stimulus cards and from
three to two the number of sorting categories.
Subjects
Subjects were drawn from the same M.A. lists as in
Experiment I. Ten Ss were assigned to each treatment group,
4 from M.A. 4-5, 4 from M.A. 6-7, and 2 from M.A. 8-9.
There were 3 females in treatment Groups I and III, 4 in
Group II.
Apparatus.
Pretraining stimuli were the same used in Experi
ment 1. Four of the nine concept sorting cards from
Experiment I, Set I were used: the green parallelogram,
the green angular design, the blue square and the blue
angular design. The two relevant labeling cards (S1 and S£)
from Set II were employed. The green trapezoid and the *nr
design from Set III were used, however, the it figure was
reconstructed in blue (#123).
65
I
I Procedure
Pretraining was identical with that of Experiment I.
The concept sorting task varied only in that four stimulus |
cards and two sorting categories were used, instead of nine j
i
i
cards and three categories; and the four stimulus cards
j
were presented according to a predesigned table of random
i
orders instead of being picked up in scrambled order by E.
!
At the conclusion of Experiment II, there remained
too few Ss to do the originally proposed concept acquisition
task, using unlabeled trays. The remaining Ss were used in
i
an incomplete replication of Experiment I.
j
j
CHAPTER III
i )
RESULTS i
I
: i
Results are presented in terms of the two scores
t
obtained for each subject, total errors (errors) and trials ;
j
to criterion (trials). The trials to criterion score did
not include the two criterial trials, therefore, a subject
who made no errors on trial one or trial two received a ,
trials score of zero. Means and standard deviations for j
both trials and errors scores are presented in Tables 2 and ;
3. Analyses of variance were done using Lindquist's treat- j
I
ments x levels paradigm (1956). The three pretraining
groups (exposure, matching training, form) constituted the
i
"treatment" groups and the M.A. groups (4-5, 6-7, 8-9) the
"levels."
Experiment I
The 22 subjects in each treatment group were divi
ded among the M.A. groups as follows: 6 in M.A. level 4-5,
11 in level 6-7, 5 in level 8-9. Analyses of variance,
I
using trials or errors as the dependent variable, yielded j
66
67
TABLE 2
MEANS AND STANDARD DEVIATIONS OF EXPERIMENT I
(9 CARD) TRIALS TO CRITERION
Experimental Group
M .A. Group 1
Exposure
II
Dot Matching
III
Form Matching :
M SD M SD M SD
Includes Subjects Who Failed
i
4-5 29.67 24.01 11.50 7.71 23.67 16.43 I
l
6-7 3.64 3.50 21.27 22.19 14.27 17.53 !
i
8-9 14.00 8.51 2 . 2 0 2.48 8.80 1 1 . 1 2 ;
All M.A. Groups 13.09 16.81 14.27 17.67 15.59 16.27 |
1
Replaces Subjects Who Failed
4-5 7.33 6.80 a a 10.83 12.77
6-7 a a a a 10.82 15.59
8-9 a a a a a a
All M.A. Groups 7.00 6.95 10.36 13.38
aNo replacements. Measure same as above.
TABLE 3
MEANS AND STANDARD DEVIATIONS OF EXPERIMENT I
(9 CARD) ERROR SCORES
Experimental Group
M.A. Group
I
Exposure
II
Dot Matching
III
Form Matching :
M SD M SD M SD >
Includes Subjects Who Failed i
4-5 174.83 162.53 41.67 31.33 165.67 169.66
6-7 1 1 . 0 0 14.09 95.18 109.53 64.45 105.25
8-9 57.40 50.22 5.00 5.10 42.80 57.49
All M.A.
Groups
66.23 108.86 60.09 8 6 . 1 2 87.13 123.52;
i
Replaces Subjects Who Failed
i
4-5 26.50 33.81 a a 49.17 82.04
6-7 a a a a 35.45 62.93
8-9 a a a a a a
All M.A.
Groups
25.77 34.64 a a 40.86 64.45
i
I
aNo replacements. Measure same as above.
69
non-significant F ratios. (See Table 4.)
Statistical analyses had been done on data for
groups in which replacements were made for subjects who
failed. It seemed possible that eliminating the subjects
who failed, (three from Group I and three from Group III),
was attenuating potentially significant differences between
the experimental groups. Therefore, the replacements were
eliminated and the data of the original subjects restored.
The corrected data were reanalyzed, using the same treat
ments x levels paradigm. Variation attributable to M.A.
levels and that attributable to the treatments by levels
interaction were now found to be significant at the .05
level. (See Table 5.) The experimental hypothesis, that
the number of trials to criterion in a learning task would
be least for a group receiving pretraining designed to
increase the probability of a relevant observing response,
was not sustained. Subjects in the higher M.A. groups
tended to require fewer trials and made fewer errors than
did the duller subjects, however, there were exceptions.
The subjects in M.A. group 6-7 required fewer trials and
made fewer errors than M.A. group 8-9 subjects under experi
mental treatment I (exposure). For treatment Group II
(matching training), the M.A. 6-7 subjects required more
TABLE 4
ANALYSIS OF VARIANCE OF TRIALS AND ERRORS SCORES,
EXPERIMENT I PRETRAINING GROUPS X
M.A. LEVELS (PPVT)
(SUBJECTS WHO FAILED WERE REPLACED)
Source df MS F
Trials as Dependent Variable
M.A. level (M) 2 71.26 < 1
Pretraining group (G) 2 291.45 1.72
M x G 4 389.33 2.30
Error 57 168.98
Errors as Dependent Variable
M.A. level (M) 2 881.95 < 1
Pretraining group (G) 2 6508.92 1.59
M x G 4 9263.09 2.27
Error 57 4072.77
!
TABLE 5
ANALYSIS OF VARIANCE OF TRIALS AND ERRORS SCORES,
EXPERIMENT I PRETRAINING GROUPS X
M.A. LEVELS (PPVT)
(INCLUDES SUBJECTS WHO FAILED)
Source df MS F
Trials as Dependent Variable
M.A. level (M) 2 773.41 3.27*
Pretraining group (G) 2 34.41 < 1
M x G 4 760.99 3.22*
Error 57 236.15
Errors as Dependent Variable
M.A. level (M) 2 41590.88 4.39**
Pretraining group (G) 2 4423.10 < 1
M x G 4 26194.16 2.77*
Error 57 9470.77
*p <.05
**p <.025
72
l
i
j trials and made more errors than the duller M.A. 4-5 sub-
I
jects. The significant treatments x levels interaction is j
i
probably attributable to the shift in performance of the
! i
M.A. 6-7 subjects under different pretraining conditions. j
!
i
Product moment correlations were calculated between !
mental age score and trials or errors. (See Table 6 .) For !
i
trials, r = -.29; for errors, r = -.31 when failing sub
jects are included. Both correlations are significant at j
i
the .05 level, which is consistent with the results of the
i
analysis of variance. If failing subjects are replaced, i
the correlations drop to -.09 and -.07 respectively. The
decrease in correlation when failures are replaced ap- |
proaches significance (£ = .12). Changing the dependent
I
variable from trials to errors or from M.A. to IQ did not
i
appreciably alter the correlations.
Although the Peabody gave significant correlations
for the total group, correlations within experimental sub
groups were not significant. It was decided to administer
a test which might be more appropriate to a concept forma
tion task. The Raven Colored Progressive Matrices (here- j
i
inafter referred to as Raven) was selected, since it
involves simple reasoning. In the Raven, the subject is
required to select a stimulus, either to match a set of
TABLE 6
CORRELATIONS FOR EXPERIMENT I
73
Experimental Group
Total
I II III
Exposure Dot Matching Form Matching
Failures Replaced
Trials
--------
x Errors .96** .98** .95** .97**'
Peabody
M.A. (PMA) .24 -.26 -.07 -.09
IQ (PIQ) . 2 0 -.26 -.04 -.09
Raven
Raw Score
(RRS) -.18 -.48* - . 0 0 1 -.23
IQ (RIQ)
-.17 -.46* . 0 1 - . 2 1
Errors
,
x PMA .24 -.23 -.03 -.07
PIQ
.18 -.23 -.03 -.08
RRS - . 1 2 -.41* -.06 - . 2 0
RIQ
- . 1 1 -.38 -.04 -.18 i
PIQ x RIQ .05 .05
i
•
H
00
H
O
•
Failures Included
Trials
i
i
x Errors .98** .98** .90** .93**!
PMA -.27 -.26 -.35 -.29*
PIQ -.35 -.26 -.36 -.32**'
RRS -.43* -.48* -.24 -.39**1
RIQ -.43* -.46* -.25 - .38**1
Errors
i
x PMA -.33 -.23 -.36 -.31*
PIQ
-.42* -.23 -.39 -.35**
RRS -.44* -.41* -.30 -.38**
RIQ -.43* -.38 -.30 -.38**
PIQ x RIQ .15 .05
CO
o
.
o
H
•
*p < .05 **p < . 0 1
istimuli or to complete the Intersect of a 2 x 2 matrix.
.For example, if row 1 were: A B, and row 2: AA___, the
i
correct response would be BB. (The test Itself uses colored;
i
designs, not letters.) The Raven was administered to all j
i
subjects. The correlation of Raven IQ with trials or '
errors (-.38) was only slightly better than the correlation j
I
obtained with the Peabody, however somewhat more confidence ;
i
can be placed in the Raven correlations (£<.01). (See ,
j
Table 6 .) It will be noted that the correlations for ex-
i
perimental subgroups are now found to be significant.
The correlation between Peabody and Raven.IQ's is j
virtually nil for the present population. (See Table 6 .)
Kilburn, Sanderson, and Melton (1966) found a correlation
of .22 between scores on the Peabody and Raven for a similar
population. The differences in correlation between that
study and the present study are not statistically signifi
cant .
An analysis of variance was computed, using the
original experimental group as "treatments" but using Raven
scores to form "levels." Since the Raven does not yield an
M.A. score, a frequency distribution of raw scores was
made. Cutting points were established by inspection, to
divide the distribution into low, medium, and high scores.
75
Raw scores 7-12 included 23% of the subjects, scores 13-15
included 48%, scores 16-30 included 29%. It was not pos
sible to balance the distribution of subjects by raw score
level, since the Raven was administered post hoc. Table 7
gives the readjusted N's and the mean trials score for each i
cell. The analysis of variance yielded a significant F
!
ratio (F = 3.51, £ < .05) for variance based on Raven score '
level. The F ratios based on treatments, and the treatmentsj
i
x levels interaction were less than one.
TABLE 7
ADJUSTED CELL N'S AND MEAN TRIALS BASED ON
THREE LEVELS OF RAVEN RAW SCORE
EXPERIMENT I
Raven Score
Group
Experimental Group
Exposure
II
Dot Matching
III
Form Matching
Mean N Mean N Mean N
Low 28.20 5 27.33 3 16.00 8
Medium 8.62 8 16.64 1 1 17.00 12
High 8.67 9 6 . 1 2 8 5.50 2
j Learning curves were drawn to show graphically the
'differential effects of the experimental treatments.
Figure 4 indicates median errors, by trial, until all sub- j
i
. 1
jects reached criterion. Medians were used to avoid over- i
i
emphasizing a few extreme cases. In Figures 5, 6 and 7 the j
percentage of subjects achieving perfect (errorless)
performance is plotted cumulatively by trials. Groups II j
i
and III both contain individual subjects who required well |
above the median number of trials to achieve criterion.
Inspection of Figures 4 through 7 suggests that form train- !
ing had the effect of reducing random M.A. variability,
the three M.A. curves being virtually identical for Group
i
III. In order to compare the curves, the distributions j
were divided into blocks of three trials each, the last j
i
block including all trials over 21. For each block of
trials, the number of subjects achieving criterion was
i
counted and tabulated separately for each M.A. level. It |
i
was then possible to compute a 3 x 8 Chi square for each j
!
experimental group. Chi square for Group I was significant !
(£<.05); for Groups II and III it was not significant,
j
suggesting that the three M.A. groups learn at different
rates under the experimental condition, Exposure. Since
dividing the trials distribution into blocks of three
Median Errors Median Errors Median Errors
M A 4 ,5
MA 6 ,7
MA 6 ,9
\
—
Trials
GROUP I— EXPOSURE
M A 4 ,5
M A 6 ,7
MA 8 ,9
20
Trials
GROUP 11-DOT MATCHING
M A 4 ,5
MA 6 ,7
MA 8 ,9
20
Trials
GROUP III— FORM MATCHING
Figure 4. Median errors per trial, by M.A.
levels, for each treatment group. Experiment I.
Failures replaced.
PERCENTAGE OF SUBJECTS
s
100
90
80
70
60
M A 4/5
M A 6/7
M A 8/9
40
20 30 25 35 40 45
t
TRIALS
"i!
Figure 5. Cumulative percentage of subjects achieving errorless per- »j
formance on each trial until all are at criterion. Group I, Exposure.
PERCENTAGE OF SUBJECTS
100
90
80
70
60
50
40
30
20
10
h
r-
rihJ
MA 4,5
MA 6 , 7
MA 8£
1 L-/L J
65 10 15 20 25
TRIALS
30 35 40 60
Figure 6 . Cumulative percentage of subjects achieving errorless per
formance on each trial until all are at criterion. Group II, Dot matching.
PERCENTAGE OF SUBJECTS
100
90
80
70
r-
60
50
M A 4.5
M A 6.7
M A 8.9
4 0
30
20
35 30 50
i
TRIALS i
oo |
o
Figure 7. Cumulative percentage of subjects achieving errorless per
formance on each trial until all are at criterion. Group III, Form matching. j
yielded a number of expected frequencies of less than one, '
! !
Chi squares were recalculated using blocks of five trials.
;A11 Chi squares were now insignificant. Although the num-
ber of small expected frequencies was reduced by the latter
procedure, the grouping was too coarse to reflect differ
ences in learning rate.
Data obtained from the inquiry portion of the
experiment were tabulated by experimental groups. Table 8
lists the number of subjects in each group who were able to
sort the three new stimulus cards correctly. Neither men
tal age level nor pretraining were significant in determin
ing whether or not a subject would be able to generalize
the concept of shape to a new set of stimuli. Of the 42
subjects able to sort the new stimulus cards correctly,
33 were able to give acceptable verbal labels to the cate
gories, with minimal help from the experimenter. Of the 9
subjects who could not do so, 2 were in Group I, 3 in
Group II, and 4 in Group III. Ten subjects were able to
give adequate verbal reports, after failing to sort the
stimulus generalization cards correctly; 3 each in Groups I
and II, 4 in Group III. These data support the findings of
Luria (1961) and O'Connor and Hermelin (1959), who suggest
that the behavior of retardates is often not consonant
TABLE 8
NUMBER AND PERCENTAGE OF SUBJECTS ABLE TO SORT ALL THREE
GENERALIZATION CARDS CORRECTLY, EXPERIMENT I j
(REPLACES SUBJECTS WHO FAILED)
Experimental Group
M.A. Group
I
Exposure
II
Dot Matching
III
Form Matching
Total j
N %a N % N % N % !
I
4-5b 5 •
00
00
1 .17 5
on
00
•
1 1 .61
6-7° 7 .64 7 .64 6 .54 2 0 .61 j
8-9d 4 .80 5 1 . 0 0 2 .40 1 1 .73 j
j
All M.A.
Groups
16 .73 13 - .59 13 .59 42 .64 1
1
i
Note: Differences in frequencies in the column and j
row totals are non*-significant when tested by X^.
aDecimal percentages are given.
bCell N = 6
cCell N = 11
dCell N = 5
! 83 |
jwith their verbalizations.
i i
Experiment II
Trials and errors scores were computed for each
subject, as in Experiment I and the same treatments x levels
type of analyses of variance were done. PPVT M.A. scores
were used to form the "levels." There were 4 subjects per
cell for M.A. levels 4-5 and 6-7 and 2 subjects per cell
for M.A. level 8-9. No replacements were made for subjects
who failed. Table 9 presents means and standard deviations
for trials and errors data. Analyses of variance based on
trials or errors yielded no significant F ratios. (See
Table 10.) With the exception of the high errors and trials
scores earned by the 3 subjects who failed to learn the
task, data adhered to the expected inverse relationship
between M.A. score and trials (errors) score. Correlation
coefficients were comparable to those found in Experiment I.
(See Table 11.) Correlations with Raven score were lower
than in Experiment I, however, the difference was not sig
nificant (-.39 compared with -.11, £ = .09). Correlation
coefficients were obtained for one subgroup, Group III.
There were no significant differences between subgroup
correlation coefficients and r's for the total group, and
TABLE 9
MEANS AND STANDARD DEVIATIONS OF EXPERIMENT II
(4 CARD) TRIALS AND ERRORS SCORES
(INCLUDES SUBJECTS WHO FAILED)
Experimental Group
M.A. Group
I
Exposure
II
Dot Matching
III
Form Matching
M SD M SD M SD
Trials to Criterion
4-5 17.50a 28.48 4.75 6.95 32.50a 32.66
6-7 9.50 9.15 2.75 3.20 5.25 5.12
8-9 .50 .70 0 0 30.50a 41.72
All M.A. Groups 10.90 18.51 3.00 4.78 2 1 . 2 0 27.33
Total Errors
4-5 34.75a 62.90 4.25 6.55 48.25a 50.54
6-7 13.00 14.47 2 . 0 0 2.83 5.25 6.18
8-9 1 . 0 0 1.41 0 0 40.00a 55.15
All M.A. Groups 19.30 39.84 2.50 4.45 29.40 40.55
aIncludes one subject who failed.
TABLE 10
ANALYSIS OF VARIANCE OF TRIALS AND ERRORS SCORES,
EXPERIMENT II PRETRAINING GROUPS X
M.A. LEVELS (PPVT)
(INCLUDES SUBJECTS WHO FAILED)
Source df MS F
Trials as Dependent Variable
M.A. level (M) 2 469.52 1.25
Pretraining group (G) 2 832.90 2 . 2 2
M x G 4 297.80 < 1
Error 2 1 375.25
Errors as Dependent Variable
M.A. level (M) 2 1539.68 1.38
Pretraining group (G) 2 1846.43 1.65
M x G 4 677.16 < 1
Error 21 1117.67
86
I
TABLE 11 |
i
i
CORRELATIONS FOR EXPERIMENT II
Group III Total
Trials
x Errors .99** .97**
Peabody
M.A. (PMA) - . 2 2 -.23
IQ (PIQ)
- . 2 2 -.23
Raven
Raw Score (RRS) - . 2 1 - . 1 1
IQ (RIQ)
-.23 - . 1 2
Errors
x PMA -.28 -.25
PIQ -.29 -.26
RRS -.19 - . 1 2
RIQ
- . 2 2 -.13
PIQ x RIQ
in
•
.39*
*£ < .05 **£ < .01
Note: Correlations computed for Group HI as a
check. N of 10 is unreliable for correlations.
j
I
I........' ........ . ' .. 87 ' ~
since 1 0 is an unreliable N for correlations, no further
i I
subgroup correlations were computed.
i
i
Learning curves were drawn. Including the three j
subjects who failed made the curves very unstable, so they
were redrawn, eliminating failures. (See Figure 8 .) Small
cell N's make the curves unreliable so Chi square tests and |
j
medians were not computed.
Table 12 presents the number of subjects who were
able to sort the new stimulus generalization cards correct* •
i
ly. Total percentages for the three M.A. levels or the
three experimental groups do not differ significantly from
the comparable percentages for Experiment I (compare
i
Table 8 ). Chi square tests indicate that neither mental !
j
age nor experimental pretraining are significant in deter- !
mining whether or not the subjects will be able to general- j
ize correctly to new stimulus cards. Of the 17 subjects
able to generalize to new stimulus materials, 1 0 were able |
to give adequate verbal labels to the sorting categories— j
!
4 in Group I, 3 each in Groups II and III. Four subjects
were able to give names to one or both of the categories J
i
but had failed to sort the generalization cards correctly.
t
A triple analysis of variance was computed (three
M.A. levels, three experimental treatments, four or nine
88
100
Wl
75
M A 4,5
M A 6,7
M A 8,9
50
L .J
8 25
0)
25 20
Trials
GROUP I— EXPOSURE
100
OTl
75
M A 4,5
M A 6,7
MA 8,9
50
8 25
0)
25 20
Trials
GROUP II— DOT MATCHING
100
(0
col
m 75
o
0)
$ 50
• p
c
0)
g 25
0)
04
M A 4,5
MA 6,7
MA 8,9
Trials
GROUP III— FORM MATCHING
Figure 8 . Cumulative percentage of subjects
achieving errorless performance on each trial until
all are at criterion. Experiment II, all groups.
Failures are excluded.
89
TABLE 12
NUMBER AND PERCENTAGE OF SUBJECTS ABLE TO SORT
ALL THREE GENERALIZATION CARDS CORRECTLY
EXPERIMENT II
M.A. Group
Experimental Group
Total
I
Exposure
II
Dot Matching
III
Form Matching
N %a N % N % N %
4-5b 2
o
in
.
3 .75 2
o
in
.
7 .58
6-7b 2 .50 1 .25 2
o
in
.
5 .42
00
1
V O
0
2 1 . 0 0 2 1 . 0 0 1
o
in
*
5 .83
Total 6 .60 6 .60 5
o
in
.
17 .57
aDecimal percentages are given.
bCell N = 4
cCell N = 2
{stimulus cards) to determine whether the four card condi
tion was significantly easier than the nine card condition.
It was not. However, the high scores of the subjects who
failed may have attenuated significance. Mental age level
continued to contribute significantly to the variance, as
did the treatments by levels double interaction. That is,
experimental treatments seem to differ in effectiveness,
depending on the mental age of the subject. (See Table 13.)
TABLE 13
THREE WAY ANALYSIS OF VARIANCE OF TRIALS SCORES
BY PPVT M.A. LEVELS, PRETRAINING GROUPS AND
NUMBER OF STIMULUS CARDS (FOUR OR NINE)
(INCLUDES SUBJECTS WHO FAILED)
Source df MS F
M.A. level (A) 2 1034.65 3.78*
Experimental group (B) 2 376.53 1.38
Number of cards (C) 1 141.38 < 1
A x B 4 709.24 2.59*
A x C 2 208.37 < 1
B x C 2 490.78 1.79
A x B x C 4 349.56 1.28
Error 78 273.60
*£ < .05
CHAPTER IV
DISCUSSION
Mental age scores and learning ability are ordi
narily found to be correlated (Rosenberg, 1963; Stevenson,
1963; Zeaman and House, 1966), although correlations range
from .03 for acquisition of a conditioned response to .73
for learning set. A few studies suggest that IQ is also
correlated with learning, independently of mental age
(House and Zeaman, 1960b; O'Connor and Hermelin, 1959).
The few studies which have found no correlation between
learning and mental age tended to use very difficult tasks
or restricted ranges of intelligence (Stevenson, 1960).
The present study found correlations of approximately .30
between measures of intelligence and measures of learning.
(See Table 6 .) These correlations were large enough to be
considered statistically significant. The correlations
with IQ reported in Table 6 do not hold M.A. constant.
When mental age effects were held constant, a nonsignificant
partial correlation of -.14 was found for trials with
Peabody IQ. Thus the present study supports the thesis
91
92
that M.A. and learning are correlated but not that IQ and
learning are correlated.
It will be noted that the correlation between
Peabody and Raven IQ scores is quite low. Kilburn, Sander
son, and Melton (1966) found a similar low correlation
between Peabody and Raven scores, suggesting that the two
tests measure different mental abilities. It might, there
fore, be hypothesized that the moderate, but significant,
correlations between learning measures and Peabody IQ are
derived from a different subpopulation than are the comp
arable correlations involving Raven IQ. That is, for some
subjects, the Raven would best predict success on the con
cept task; for other subjects the Peabody would predict
best. This variability in the subject population will be
discussed further below.
It is of interest to notice the change in correla
tion when the subjects who failed were replaced by subjects
able to learn the task. Groups I and III each included
subjects who failed to learn the concept task; all Group II
subjects learned the task. Comparison of the subgroup
correlations shows that the Group II correlations are most
like the Group I and III correlations in which failures
were included. In fact, eliminating failures lowers the
correlation between measures of learning and measures of
intelligence, in one case (Group I, PPVT) shifting the
correlation from negative to positive. The latter correla
tion would imply that brighter subjects make more errors
and take longer to learn than duller subjects. Although
McIntyre (1964) and Weir and Stevenson (1959) have reported
such findings, the results, in the present research, appear
to be random. It would seem that the practice of eliminat
ing failing subjects, whether done wittingly or heedlessly,
could seriously bias experimental findings.
The significant treatment x levels interaction
(failures included) is difficult to explain. If the over
all trend for low trials scores to be associated with
higher IQ's is accepted, then Exposure training appears to
be easier for M.A. 6-7 and harder for M.A. 8-9 than would
be anticipated, and matching training appears to be easier
for M.A. 4-5 and harder for M.A. 6-7 than would be expected.
(See Tables 2 and 3). If mean number of trials is ranked
by M.A. level rather than by experimental group, it appears
that form matching is of intermediate difficulty for all
M.A. groups. Exposure training is the most difficult,
except for M.A. 6-7 subjects and matching training least
difficult, again excepting M.A. 6-7 subjects. For M.A.
6-7 subjects the rank order of matching training and
iexposure training is reversed. There is no theoretical
'position which easily explains these findings. Piaget
! (Inhelder and Piaget, 1958) indicates that a shift from the
"representational" stage to the "concrete operations" stage
takes place about six or seven years of age. Luria (1961)
points out that five to seven year olds begin to make use
of "internal speech" in problem solving. Zeiler (1964)
found that normal five year olds performed like seven year
olds in learning a two choice discrimination problem but
like three year olds in their response to a nonreversal
shift. He concluded that the different age groups were
using different mediation processes. None of these authors
approaches problem solving in terms of mental age, inde
pendent of chronological age, although Luria points out that
retardates often fail to form stable verbal connections and
poorly coordinate or synchronize their verbal and motor
responses. Assuming a shift to more verbal or abstract
thinking at about mental age six, Exposure training might
enhance a subject's performance on a concept task but the
same facilitation would be expected at M.A. 8-9. Further,
if use of verbal symbols enhances a child's ability to
generalize (Luria, 1961), then matching training should be
I
more effective with six and seven year olds than with four
and five year olds— the reverse of present experimental
findings. The better part of valor may be to assume, post
jhoc, that a £ <.05 is not significant and that the inter
action term reflects only random variability.
Non-confirmation of the experimental hypothesis was
disappointing. The various types of pretraining did not
seem to affect differentially the performance of hospital
ized retardates on a concept learning task. Since the pre
training was designed to increase (or leave unchanged) the
probability of occurrence of a relevant observing response,
the nonsignificant results do not lend support to the value
of a two-stage model in concept learning of retardates.
There is strong evidence in favor of the two-stage model,
however, in the concept learning of college students
(Bower and Trabasso, 1963; Bower and Trabasso, 1964) and
in the discrimination learning of the retarded (Zeaman and
House, 1963), which warrants careful consideration before
rejecting the two-stage model as inapplicable to concept
learning in the retarded.
The following hypotheses are advanced to explain
the non-confirmation of the experimental hypothesis;
1. The pretraining was not effective in eliciting
the relevant observing response.
2. The subject failed to generalize the observing
response elicited in pretraining to the concept
task.
3. The observing response for form was dominant
prior to pretraining, precluding establishment
of different levels of dominance by pretraining.
4. Learning the concept task interfered with the
effect of the observing response.
1. If the pretraining tasks were ineffective in
eliciting the relevant observing response, the following
would be plausible reasons:
a) The form training task failed to elicit an
observing response for form. Since the observing response
is by definition response to a dimension, and the matching
task required response to different geometric forms, then
if the subject matched by form, he must, perforce have made
a relevant observing response. The pilot study showed that
subjects who received form training matched by form signifi
cantly more often in a subsequent task than did subjects
who received no pretraining, which suggests that the Form
: ........................................... "" ' " 97....
i
|
|training was effective in eliciting an observing response
for form. It is quite possible that some subjects did not
match the geometric figures of the pretraining task by form,:
but rather on the basis of edge details, points, or by some *
other idiosyncratic method, in which case a relevant observ
ing response need not have occurred.
b) The pretraining tasks failed to affect
differentially the observing response for form. The two
control pretraining tasks, exposure and dot matching, were
not piloted, however, it seems unreasonable to believe that
simple exposure to line drawings or that matching numbers
of dots would increase the probability of an observing
response for form as much as would matching geometric
figures.
c) Some of the subjects did not have an ab
stract concept of form. That is, although a subject might
be able to respond concretely, "A ball is round," the
abstracted stimulus property, shape— particularly two-
dimensional shape— might be quite unknown. Since the ob
serving response is response to a dimension, not response
to a particular stimulus value, subjects who lacked an
abstract notion of shape could not have an observing
response for shape. The tracing and form matching tasks of
98
i
pretraining were designed to elicit an observing response
land were too brief to train a subject to acquire a concept
of shape de novo. In retrospect, it would have been advis- '
able either to test subjects' knowledge of the relevant
i
dimensions as had been done by Klugh and Janssen (1966) or
to train all subjects to a known level of proficiency on
a task involving the concept of form, and then attempt to
interfere with the observing response.
d) Fourth, there was simply too little practice
on the form training for the observing response to be
learned effectively. In particular, those subjects who
required more than one day to learn the concept task (5 in
Group 1, 3 in Group II, 7 in Group III) were exposed to
abbreviated pretraining on subsequent days. If the observ
ing response for form were poorly learned, the single pre
sentation of the matching figures might have been insuffi
cient to elicit full recall.
2. Subjects may have learned the relevant observing
response, yet failed to generalize the response to the con
cept sorting task. The use of a research assistant for the
pretraining, and failure to specify that the pretraining
tasks were relevant to the concept task may have exceeded
the ability of some of these subjects to generalize from one
Itask to another. However, if the pretraining were effec-
I
tive in changing the relative dominance of form over other
observing responses, it is unlikely that the nonspecific
activities which occurred during the approximately two-
minute interval between pretraining and the concept task
would serve to extinguish the observing response. "Failure
to generalize" seems a tenable but weak hypothesis.
3. A third hypothesis, that the observing response
for form was already dominant, is quite tenable. Goldstein
and Scheerer (1941), and Halpin (1958) found a certain per
centage of retarded subjects would sort colored forms by
shape, rather than color, when simply asked to sort the
stimuli. In the present investigation, four subjects in
Group I and two in Group II, groups given "irrelevant" pre
training, sorted the concept cards by form immediately,
making no errors whatsoever. In Group III, which received
pretraining designed to teach the relevant observing
response, two subjects sorted without errors. (See Figure
9.) Either the supposedly irrelevant pretraining did, in
fact, teach a relevant observing response, or else form was
already dominant in the response hierarchy of these sub
jects. It would be difficult to increase the probability
of occurrence of an observing response for form, relative
No. of Subjects No. of Subjects No. of Subjects
100
x = M.A. level 4,5
o = M.A. level 6,7 ;
* = M.A. level 8,9 '
# ** Failure
Each character rep
resents one subject
o
o
*
#
o
* *
o o
#
X o o o o X
*
o o X
*
#
0 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 13
Group I--Exposure
*
*
o
*
o
*
o o o o
*
X o X X o X X X o o o
0 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 13
Group II-— Dot Matching
*
o
o
*
# # #
*
o o o X
*
o X o o o X X o *
0 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 13
Number of Errors
Group III— Form Matching
Figure 9, Frequency distribution of errors on
trial one. Experiment I.
j 101 !
| ;
to other observing responses, for those subjects for whom
j
form was already dominant. Unfortunately subjects could
not be pretested to determine the dominance level for form
without thereby changing that dominance.
4. The verbal responses of the patients confirmed
that for some of them, learning the concept task was inter
fering with the effect of the observing response. Subjects
would, for example, sort by form saying, "The blue one goes
here. The green one goes here." Then, as if listening to
their own verbalizations, they would shift to sorting by
color, even though they would then make mistakes. If pre
training were effective, it should have been evident in the
subject's response on trial one, before he developed inter
ference from other responses. Figure 9 presents a frequency
distribution of errors on trial one by experimental groups.
Most of the Group III subjects seem to cluster at the lower
end of the frequency distribution and no subject made more
than eight errors. In contrast, Group I and II subjects
show a wider distribution and both groups contain subjects
who made as many as thirteen errors. Although Group III
subjects tend to make fewer errors, a Kruskal-Wallis one-way
analysis of variance by ranks indicates that these distri
butions do not differ significantly. Here is further
evidence of non-confirmation of the experimental hypothesis.
i
Re-examination of the studies of House and Zeaman
.and of Trabasso and Bower (House and Zeaman, 1962; Trabasso,
1963; Trabasso and Bower, 1964a; Trabasso and Bower, 1964b;
Trabasso and Bower, 1966) indicates a major difference from
the present investigation. All of these experimenters used
tasks having only binary dimensions in their investigations
of the two-stage model. Although the number of dimensions
,varied, all stimuli were classifiable as "A" or "B." How
ever, in 1964, Bower and Trabasso extended their study to a
four choice situation and indicated that when the number of
stimulus and response values was increased, the second phase
of the two-stage model became more time consuming. That
~~is, even though the subject has learned which dimension is
relevant, he requires more trials to learn the correct
responses when three or more stimuli are involved than when
only two are involved. It was found, in the present inves
tigation, that subjects would often learn the correct
placement for one of the shape values after a few trials
but seem to require considerably more to learn the remaining
two. It could be hypothesized that these "plateaus" con
tributed variability in the second stage of learning which
masked effects of pretraining on the initial observing
103 |
I
response phase. However, the trial one data reported above j
l
! would tend to vitiate the argument.
It could be argued that the subjects in the present
investigation failed to learn a concept task at all and,
therefore, would be uninfluenced by an observing response
for form. That is, a subject might learn quite arbitrarily
that this funny card goes in tray one and that odd card
goes in tray two, with no notion of grouping similar shapes
together. Hermelin and O'Connor (1958) found that retard
ates were able to learn such rote tasks, although rote
tasks were learned more slowly than concept tasks. Subjects
who failed to grasp the concept of shape would not be ex
pected to sort the new stimulus generalization cards cor
rectly, nor would they be expected to be able to give
appropriate verbal labels to the categories. Twenty-four
subjects failed to sort the generalization cards correctly.
A biserial correlation with trials to criterion as the con
tinuous variable and pass or fail on the generalization
task as the dichotomous variable was -.77 (standard error =
.25, £ = .001). The significant correlation lends support
to the hypothesis that some of the subjects learned only a
rote sorting task. Unfortunately the generalization task
was not administered to those subjects who failed the
104
concept task, so no comparison can be made. Experimental
'treatment did not significantly influence the number of sub-:
f
jects able to generalize.
Forty-two subjects could apply the sorting princi- ‘
pie to new stimuli and of these, 33 were able to give to
the three form categories some kind of name which indicated
recognition of the shape characteristic to that category.
For example, 19 subjects labeled the quadrilateral figures
"square," 8 labeled them "box"; 6 labeled the amorphous
figures "round," 5 labeled them "star" and 5 "cloud." The
assumption that the angular figures would be difficult to
name was not substantiated, 1 1 subjects naming them "cross"
and 10 "airplane." Nine of the subjects who sorted the
generalization cards correctly gave extremely inadequate
verbal labels or required considerable help in giving verbal
labels to the categories. Two of the subjects were in
Group I, 3 in Group II, four in Group III. Surprisingly
1 0 subjects who had failed the new card generalization task
were able to give adequate verbal labels. Six of the sub
jects required some help from the examiner, 4 gave verbal
labels without help. It may be hypothesized that these
subjects had failed to see that there was a principle by
which the stimuli could be sorted. Heidbreder (1947)
points out that her subjects' performance changed when they i
suddenly "discovered" that the nonsense syllables were
paired with concepts. And Reed (1946) and Shaffer (1961)
have shown that college students learn a task more quickly
if informed that there is an underlying concept which can
be utilized. In retrospect, the present study would have
been strengthened if the subjects had been informed directly
that there was a rule by which the stimuli were to be sorted
and that their task was to learn this rule.
The findings in Experiment II, the four-card exper
iment, are similar to those in Experiment I, however, none
of the F ratios are significant in the analyses of variance.
The triple analysis of variance indicated no significance
for variance attributable to the number of stimulus cards,
i.e., the four-card task was not significantly easier than
the nine-card task. However, the 3 subjects who failed
tended to inflate the scores a disproportionate amount
because of the small N, e.g., the mean trials score for
Group I, M.A. 4-5 cell, was 17.50 when the failing subject
was included, 3.33 when excluded. If the failing subjects
were excluded, a practice previously eschewed? a signifi
cantly lower trials (errors) score for the four-card condi
tion might have resulted, as would be predicted by the
findings of Bower and Trabasso (1964).
i
Correlations for Experiment II (Table 11) are not
I significantly different from their counterparts in Experi
ment I. (See Table 6 .) The correlation between trials and
Raven raw score (-.39 and -.11) yielded the only difference
which approached significance (£< .1 0 ).
The percentages of subjects able to sort the gener
alization cards, or to verbalize labels for the concept
groupings, or to do both did not differ from Experiment I.
(Compare Tables 8 and 12.) A biserial correlation of -.63
was found between trials and the ability to generalize.
This correlation is comparable to the -.77 found in Experi
ment I and lends further weight to the hypothesis that some
of the subjects learned the concept task by rote. Six
subjects, compared to 10 in Experiment I were able to give
verbal labels after failing to sort the stimulus generaliza
tion cards. The distribution of errors on trial one is
narrower than in Experiment I, as would be expected with
the reduction in response categories. There are no differ
ences in distribution between experimental groups. (See
Figure 10.)
It is difficult to assess the significance of the
various statistical results obtained in the present
-p
o
A *
•§ O *
" O O X
o X O X
25
( 0
+»
o
< u
w
« 4 - l
o
m
•P
o
a >
0 3
107
x = M.A. level 4,5
o = M.A. level 6,7
* = M.A. level 8,9
# = Failure
Each character rep
resents one subject
Group I— Exposure
*
*
o
o
0 X 0
X X X
g 0 1 2 3 4 5
Group II— Dot Matching
*
o #
o o #
X x Q
1 2 3 4
o Number of Errors
25
Group III— Form Matching
Figure 10. Frequency distribution
of errors on trial one, Experiment II.
j 108
|Investigation. Significance has been found for M.A. level
las a predictor of learning on the experimental task. On
ithe other hand, large standard deviations (Tables 2, 3, and
19), standard deviations which are often greater than their
means, even with failing subjects eliminated, indicates
that there is extreme intragroup variability. This intra
group variability often exceeds that between groups. The
reasons for the variability are several.
1. Subjects were chosen with no regard for their
emotional stability beyond the fact that they could cooper
ate in the task. Some of the subjects were "garden variety"
retardates with no particular emotional problems. For many
of these patients the experiment offered an interesting
break in routine. To some patients, the experiment was a
threat, for it presented them with a novel task which they
feared they might fail. A few subjects displayed a kind of
passive negativism, as if their attitude were, "I'll go
along with your silly game because I have to but don't
expect me to try very hard." Attempts had been made to
insure at least minimal motivation by offering a choice of
prizes but no doubt some of the subjects did not consider
the prizes to compensate for the effort which they had to
exert.
2. Mentally retarded persons are not a homogeneous j
j i
group. In fact, they are far more variable than the averagei
"college sophomore" or "white rat" population. Varieties
of physical limitations; of brain pathology; of past learn- ;
i
ing; of past experience with learning tasks, adults, and
the self as a coping organism; assorted reinforcement his
tories; and a variety of tranquilizer, energizer and
anticonvulsant medications result in a population of sub
jects differing widely in responses to people, situations
and things. Equating such subjects on the basis of mental
i
age score could be expected to reduce the total variability,
since mental age scores have some predictive validity for
learning (Rosenberg, 1963; Stevenson, 1963; Zeaman and
House, 1966) . However, at best an M'.A. score represents a
point measure of a variable range of performance. It has
already been noted that there is evidence of two subpopula
tions in the present study, one whose learning is best
predicted by the Peabody, another whose learning is best
predicted by the Raven.
3. A third source of variability is to be found in
the strategies used by the patients to solve the concept
task. Probability learning experiments by Weir (1964) have
shown that normal three to five year olds tend to repeat
110
j ;
the response which they just made about 80% of the time,
whether or not it was followed by reinforcement. Seven year
iolds do so only about 50% of the time and are beginning to
respond differentially, immediately repeating a response
more often when it is reinforced than when it is not. By
nine years, children are immediately repeating a nonrein
forced response only about 40% of the time. Subjects who
persist in a particular response, irrespective of the per
centage of reinforcement are termed "maximizers" by proba
bility theorists, in contrast to subjects who try to "fol
low" the probable ratio of reinforcements. Schenck and
Shepherd (1967), with a population of retardates (mean M.A.
- 8 years), found that 56% of their subjects were maximi
zers. Although a correction method was used in the present
investigation, making the task different from probability
learning experiments, the tendency of a subject to repeat
or to shift from a category which was just reinforced would
affect his error rate. That is, a subject who tended to
sort stimulus n + 1 into the same category as stimulus n
would have a high error rate, since the stimuli were so
arranged that stimuli representing the same shape value
were not in consecutive order.
Ill !
I
I
4. Finally, the apparent failure of some subjects
i
to understand that the concept task involved finding a rule j
for classifying the stimuli, rather than rote memory would
contribute to the variability in trials (errors) scores.
A brief investigation was carried out on these sub
jects who failed the concept task. There were a total of
1 2 subjects, 6 in the original group and 6 more among the
subjects selected to replace the failures. These 12 sub
jects do not appear to have any characteristics in common
which would differentiate them from the group of subjects
who learned. Five subjects who were classified as failures
under the 40 trial limit would probably have learned within
the 60 trial limit. All 5 of these subjects had made less
than 2 0 0 errors and all had learned the correct response
for several of the stimulus cards. There were a few sub
jects who seemed to have no concept whatsoever of the task.
After 40 trials, there was still no discernable pattern to
these subjects' errors, either of color matching or place
responding. It is possible that the latter group of
patients represents those who had never acquired an abstract
concept of shape.
i Implications for Further Research
! i
I
Aside from the shift to retarded subjects, a major
'difference between the present research on two-stage models ;
i
and that of previous investigators is the use of multi
valued dimensions. It is possible, of course, that the
two-stage model of learning is not applicable to concept
learning in the mentally retarded. It might be more parsi
monious first to try to adapt the bivalued tasks of Trabasso
and Bower to a retarded population. By limiting the number
of dimensions, it should be possible to bring the tasks
within the capabilities of the retarded.
Second, it would be useful to investigate the two-
stage model in the learning of normal children. It would
be particularly interesting to determine the relationship
between the observing response and the development of class
ification schema described by Inhelder and Piaget (1964).
Perhaps the transition from "graphic" to "non-graphic
collections" occurs only when the child has acquired an
observing response for the relevant dimension. Or perhaps
the transition from graphic to non-graphic collections is
the Anlage for developing the abstracted concepts required
by an observing response.
| Finally, it might be of interest to repeat the
present research, having first trained all subjects to a
given level of proficiency on a concept task to be certain
all were able to make the relevant observing response.
Part of the subjects would then be given a task designed to
interfere with the observing response, to determine whether
they then learned a new concept task more slowly than did
a group having no interfering task.
CHAPTER V j
1 !
; I
I
SUMMARY
i
!
I
Two-stage attention models have been used to explain,
concept learning by normal college students and discrimina- j
tion learning by the mentally retarded. The model postu
lates: (1 ) attention to a stimulus dimension— an interven
ing variable, (2 ) an overt response to a cue of that i
dimension.
i
I
The present research sought to assess the applica- |
I
bility of a two-stage model to concept learning in the j
mentally retarded by attempting to manipulate the attention ;
stage. It was hypothesized that pretraining designed to
i
increase the probability of occurrence of an observing
i
I
response for form would facilitate learning a concept task j
i
based on shape. |
!
The subjects were hospitalized retardates (C.A. 13- j
38 years) whose mental age scores ranged from 4 to 10 years.;
Three M.A. levels and three experimental treatments were j
used in a treatments by levels design.
Group I was exposed to the form matching materials j
114
115
{but received no matching training. Group II received
imatching training on materials in which form was irrele
vant. Group III was given both form tracing and form j
i
!
matching. All pretraining was administered by a research ;
i
i
assistant.
I
All subjects were then given a sorting task in which;
i
nine different, colored designs were to be sorted into three
categories on the basis of similar shape characteristics:
rounded, angular, or quadrilateral. Three shades of three
colors were used for the designs: green, blue and gray.
Each of the shape categories was labeled with a card repre- ,
senting the relevant shape value and one of the three color j
values, a different color for each label. The subject was
told immediately whether he was right or wrong and was
}
required to correct his mistakes, without help from the
experimenter. At the conclusion of a trial (nine cards)
the experimenter scrambled the cards and presented them
again. A subject sorted the stimulus cards until he sorted
the complete set correctly twice in succession or until he
reached a limit of sixty trials. In the latter case he was
considered to have failed. Subjects who reached criterion
were given three new cards to test their ability to general- 1
ize the shape concept to new stimuli. Finally, an inquiry
{was conducted to determine whether the subject could verbal-
I
ize a name for the concept categories. j
I I
The data for the statistical analyses were trials
to criterion and total errors. Two kinds of analyses were
1
done, one including subjects who failed and one excluding !
those who failed. The analysis which included failures
I
f
showed that mental age differences contributed significantly
to the variance, as did the interaction effects of the !
experimental treatments on the three mental age groups. 1
i
i
Experimental treatment effects were not significant. j
!
I
The experiment was repeated, reducing the number of ;
i
I
stimulus cards to four and the number of response categories!
to two, in order to test the hypothesis that the original
experiment was too difficult. Subjects tended to require
fewer trials; nevertheless, three subjects failed to learn i
the sorting task, raising the trials and errors scores for
the whole group. Neither treatments nor levels effects
were significant in the analysis of variance computed for
i
Experiment II. A three-way analysis of variance, combining
the data from Experiment I and Experiment II showed no
significant differences between the four card and the nine
card conditions.
Several hypotheses regarding the non-confirmation
i
of the main experimental hypothesis seemed reasonable.
First, the present experiment shifted from binary dimen
sions, which have generally been used, to three-valued
dimensions. Second, the observing response for form may
j
not have been dominant, relative to other observing re- j
!
sponses. Third, for a variety of reasons, discussed in the i
i
l
paper, there was an excessive amount of intragroup varia- i
!
bility. i
REFERENCES
118
i
REFERENCES !
Ball, T. S. Perceptual concomitants o£ conceptual reorgan
ization. Journal of Consulting Psychology, 1964, 28,
523-528.
Bensberg, G. J., Jr. Concept learning in mental defectives
as a function of appropriate and inappropriate "atten- j
tion sets." Journal of Educational Psychology, 1958,
49, 137-143.
j
Bourne, L. E. Effects of delay of information feedback and ;
task complexity on the identification of concepts.
Journal of Experimental Psychology, 1957, 54, 201-207. j
!
Bourne, L. E. Factors affecting strategies used in problems;
of concept formation. American Journal of Psychology, j
1963, 76, 229-238.
j
Bourne, L. E., and Bunderson, C. V. Effects of delay of |
informative feedback and length of postfeedback interval!
on concept identification. Journal of Experimental |
Psychology, 1963, 65, 1-5. J
Bourne, L. E., Goldstein, S., and Link, W. E. Concept j
learning as a function of availability of previously ;
presented information. Journal of Experimental
Psychology, 1964, 67, 439-448. ;
I
Bourne, L. E., Guy, D. E., Dodd, D. H., and Justesen, D. R.
Concept identification: the effects of varying length
and informational components of the intertrial interval.;
Journal of Experimental Psychology, 1965, 69, 624-629.
I
Bourne, L. E., and Haygood, R. C. Supplementary Report:
Effect of redundant relevant information upon the
identification of concepts. Journal of Experimental
Psychology, 1961, 61, 259-260.
119
Bourne, L. E., and Parker, B. K. Differences among modes
for portraying stimulus information in concept identi
fication. Psychonomic Science. 1964, 1, 209-210.
Bourne, L. E., and Pendleton, R. B. Concept identification
as a function of completeness and probability of in
formation feedback. Journal of Experimental Psychology
1958, 56, 413-420.
Bourne, L. E., and Restle, F. Mathematical theory of con
cept identification. Psychological Review, 1959, 6 6 ,
278-296.
Bower, G., and Trabasso, T. Reversals prior to solution in
concept identification. Journal of Experimental Psy
chology, 1963, 6 6 , 409- 418.
Bower, G. H., and Trabasso, T. R. Concept identification.
In R. C. Atkinson (ed.), Studies in mathematical psy
chology. Stanford: Stanford University Press, 1964.
Pp. 32-94.
Bruner, J. S., Goodnow, J., and Austin, G. A. A study of
of thinking. New York: Wiley, 1957.
Bryant, P. E. The transfer of sorting concepts by moder
ately retarded children. American Journal of Mental
Deficiency, 1965, 70, 291-300.
Buss, A. H., and Buss, E. H. The effect of verbal rein
forcement combinations on conceptual learning. Journal
of Experimental Psychology, 1956, 52, 283-287.
Campione, J., Hyman, L., and Zeaman, D. Dimensional shifts
and reversals in retardate discrimination learning.
Journal of Experimental Child Psychology, 1965, 2, 255-
263.
Glanzer, M., Huttenlocher, J., and Clark, W. H. System
atic operations in solving concept problems: a para
metric study of a class of problems. Psychological
Monographs, 1963, 77 (1, Whole No. 564).
jGoldstein, K.r and Scheerer, M. Abstract and concrete
behavior: an experimental study with special tests.
Psychological Monographs, 1941, 53 (2, Whole Mo. 239).
Goss, A. E. A stimulus-response analysis of the interac
tion of cue-producing and instrumental responses. j
Psychological Review, 1955, 62, 20-31. j
t '
Goss, A. E. Acquisition and use of conceptual schemes.
In C. N. Cofer (Ed.), Verbal learning and verbal
behavior. New York: McGraw-Hill Book Co., Inc., 1961.
Pp. 42-69. (a)
Goss, A. E. Verbal mediating responses and concept forma- j
tion. Psychological Review, 1961, 6 8 , 248-274. (b) !
Goss, A. E., and Moylan, M. C. Conceptual block-sorting as i
a function of type and degree of mastery of discrimina- ;
tive verbal responses. Journal of Genetic Psychology,
1958, 93, 191-198.
Grant, D. A., Jones, O. R., and Tallantis, B. The relative :
difficulty of the number, form and color concepts of a
Weigl-type problem. Journal of Experimental Psychology,
1949, 39, 552-557.
Gray, J. S. A behavioristic interpretation of concept for- j
mation. Psychological Review, 1931, 38, 65-72.
Green, E. J. Concept formation: a problem in human {
operant conditioning. Journal of Experimental Psychol- I
ogy, 1955, 49, 175-180. |
i
Griffith, B. C. The use of verbal mediators in concept
formation by retarded subjects at different intelligence!
levels. Child Development, 1960, 31, 633-641. j
Griffith, B. C., Spitz, H. H., and Lipman, R. S. Verbal
mediation and concept formation in retarded and normal j
subjects. Journal of Experimental Psychology, 1959, j
58, 247-251.
122
Halpin, V. 6 . The performance of mentally retarded children
on the Welg1-GoIdstein-Scheerer color form sorting test.
American Journal of Mental Deficiency. 1958, 62, 916- |
919. I
j
t
Halpin, V. 6 ., and Patterson, R. M. The performance of j
brain-injured children on the Goldstein-Scheerer tests. !
American Journal of Mental Deficiency, 1954, 59, 91-99. !
t
Harrow, M., and Friedman, G. B. Comparing reversal and
nonreversal shifts in concept formation with partial
reinforcement controlled. Journal of Experimental j
Psychology, 1958, 55, 592-598.
Haygood, R. C., and Bourne, L. E. Forms of relevant stim
ulus redundancy in concept identification. Journal of
Experimental Psychology, 1964, 67, 392-397. !
Heber, R. A manual on terminology and classification in
mental retardation. American Journal of Mental Defi
ciency, Monograph Supplement, 1959, 64, No. 2. ;
Heidbreder, E. Toward a dynamic psychology of cognition.
Psychological Review, 1945, 52, 1-22. j
Heidbreder, E. The attainment of concepts: I, Terminology i
and methodology. Journal of General Psychology, 1946,
35, 173-189. (a) j
Heidbreder, E. The attainment of concepts: II, The prob
lem. Journal of General Psychology, 1946, 35, 191-223.
(b) 1
i
Heidbreder, E. The attainment of concepts: III, The
process. Journal of Psychology, 1947, 24, 93-138.
Heidbreder, E. The attainment of concepts: VI, Exploratory
experiments on conceptualization at perceptual levels.
Journal of Psychology, 1948, 26, 193-216.
Heidbreder, E., Bensley, M. L., and Ivy, M. The attainment j
of concepts: IV, Regularities and levels. Journal of ;
Psychology, 1948, 25, 299-329.
Heidbreder, E., and Overstreet, P. The attainment of con
cepts: V, Critical features and contexts. Journal of
Psychology. 1948, 26, 45-69.
Hermelin, B., and O'Connor, N. The rote and concept learn- ;
ing of imbeciles. Journal of Mental Deficiency |
Research, 1958, 2, 21-27. j
I
j
House, B. J., and Zeaman, D. A comparison of discrimina
tion learning in normal and mentally defective children.!
Child Development, 1958, 29, 411-416. (a) j
i
House, B. J., and Zeaman, D. Visual discrimination learning!
in imbeciles. American Journal of Mental Deficiency, I
1958, 63, 447-452. (b) i
House, B. J., and Zeaman, D. Transfer of a discrimination
from objects to patterns. Journal of Experimental
Psychology, 1960, 59, 298-302. (a)
House, B. J., and Zeaman, D. Visual discrimination learn
ing and intelligence in defectives of low mental age.
American Journal of Mental Deficiency, 1960, 65, 51-58.
(b)
House, B. J., and Zeaman, D. Reversal and nonreversal |
shifts in discrimination learning in retardates. Jour- j
nal of Experimental Psychology, 1962, 63, 444-451. j
l
House, B. J., and Zeaman, D. Miniature experiments in the |
discrimination learning of retardates. In L. P. Lipsittj
and C. C. Spiker (Eds.), Advances in child development
and behavior. New York: Academic Press, 1963. Pp. 313-:
374. !
1
Hoviand, C. I. A "communication analysis" of concept !
learning. Psychological Review, 1952, 59, 461-472. j
Hovland, C. I. A set of flower designs for experiments in
concept formation. American Journal of Psychology,
1953, 6 6 , 140-142.
Hovland, C. I. Computer simulation of thinking. American
Psychologist, 1960, 15, 687-693.
124
;Hovland, C. 1., and Hunt, E. G. Computer stimulation of
concept attainment. Behavioral Science, 1960, 5, 265-
267.
Hull, C. L. Quantitative aspects of the evolution of con
cepts. Psychological Monographs, 1920, No. 28 (1,
Whole No. 123).
Hunt, E. G. Concept learning: an information processing
problem. New York: Wiley, 1962.
Inhelder, B., and Piaget, J. The growth of logical think- |
ing. New York: Basic Books, 1958.
i
Inhelder, B., and Piaget, J. The early growth of logic in
the child. New York: Harper, 1964. I
i
Isaacs, I. D., and Duncan, C. P. Reversal and nonreversal
shifts within and between dimensions in concept forma- |
tion. Journal of Experimental Psychology, 1962, 64,
580-585. j
Jacobs, A. Performance of children in a discrimination ;
problem as a function of symbolic guidance, delay of |
reward, and mental ability. Unpublished doctoral
dissertation, University of Iowa, 1950. i
I
Johnson, D. M., and O'Reilly, C. A. Concept attainment in !
children: classifying and defining. Journal of Edu- j
cational Psychology, 1964, 55, 71-74.
!
Johnson, P. J. Factors affecting transfer in concept- j
identification problems. Journal of Experimental Psy
chology , 1966, 72, 655-660.
Kendler, H. H. The concept of the concept. In A. W.
Melton (Ed.), Categories of human learning. New York:
Academic Press, 1964. Pp. 211-236.
Kendler, H. H., and D'Amato, M. F. A comparison of reversal!
shifts and nonreversal shifts in human concept formation;
behavior. Journal of Experimental Psychology, 1955, i
49, 165-174.
125
Kendler, H. H., Glucksberg, S., and Keston, R. Perception
and mediation in concept learning. Journal of Experi
mental Psychology, 1961, 61, 186-191. j
iKendler, H. H., and Karasik, A. D. Concept formation as a
function of competition between response produced cues. {
Journal of Experimental Psychology, 1958, 55, 278-283. j
Kendler, H. H., and Kendler, T. S. Vertical and horizontal ;
processes in problem solving. Psychological Review,
1962, 69, 1-16.
Kendler, H. H., and Kendler, T. S. Selective attention vs. ;
mediation: some comments on Mackintosh's analysis of
two stage models of discrimination learning. Psycho
logical Bulletin, 1966, 6 6 , 282-288.
Kendler, H. H., and Mayzner, M. S. Reversal and nonreversal;
shifts in card-sorting tests with two or four cate- j
gories. Journal of Experimental Psychology, 1956, 51,
244-248.
Kendler, T. S., and Kendler, H. H. Reversal and nonreversalj
shifts in kindergarten children. Journal of Experimen- J
tal Psychology, 1959, 58, 56-60. |
• > i
Kilburn, K. L., Sanderson, R. E., and Melton, K. Relation
of the Raven Coloured Progressive Matrices to two j
measures of verbal ability in a sample of mildly
retarded hospital patients. Psychological Reports,
1966, 19, 731-734.
Kirk, W. L., Jr. An analysis of the mediational approach
to reversal and nonreversal shifts in concept formation.;
Unpublished doctoral dissertation. University of Cali
fornia at Los Angeles, 1964.
Klugh, H. E., and Janssen, R. Discrimination learning by
retardates and normals: method of presentation and i
verbalization. American Journal of Mental Deficiency,
1966, 70, 903-906.
Klugh, H. E., and Roehl, K. Developmental level and con
cept learning: interaction of age and complexity.
Psychonomic Science, 1965, 2, 385-386.
126
jKorstvedt, A., Stacey, C. L., and Reynolds, W. F. Concept
formation of normal and subnormal adolescents on a
modification of the Weigl-Goldstein-Scheerer color form
sorting test. Journal of Clinical Psychology, 1954,
10, 88-90.
;Kounin, J. S. Experimental studies of rigidity: I, The
measurement of rigidity in normal and feeble-minded
persons. Character and Personality, 1941, 9, 251-272.
Kounin, J. S. Experimental studies of rigidity: II, The {
explanatory power of the concept of rigidity as applied j
to feeble-mindedness. Character and Personality, 1941,
9, 273-282. (b)
i
j
Lindquist, E. F. Design and analysis of experiments in j
psychology and education. Boston: Houghton Mifflin,
1956.
Luria, A. R. The role of speech in the regulation of normal
and abnormal behaviour. New York: Pergamon Press, 1961.
Mackintosh, J. J. Selective attention in animal discrimi- i
nation learning. Psychological Bulletin, 1965, 64, I
124-150. j
i
McIntyre, R. B. Effects of repetitious programing in the
acquisition of addition facts by educable retardates.
Unpublished doctoral dissertation, George Peabody
College, 1964.
McMurray, J. G. Rigidity in conceptual thinking in exo
genous and endogenous mentally retarded children.
Journal of Consulting Psychology, 1954, 18, 366-370.
i
Meyers, C. E., Dingman, H. F., Attwell, A. A., and Orpet,
R. E. Comparative abilities of normals and retardates
of M.A. 6 years on a factor-type test battery. Ameri
can Journal of Mental Deficiency, 1961, 6 6 , 250-258.
i
Milgram, N. A. Verbalization and conceptual classification |
in trainable mentally retarded children. American '
Journal of Mental Deficiency, 1966, 70, 763-765.
Milgram, N. A., and Furth, H. G. The influence of language
on concept attainment in educable retarded children.
American Journal of Mental Deficiency, 1963, 67, 733-
739.
Mishima, J., and Tanaka, M. The role of age and intelli
gence in concept formation of children. Japanese
Psychological Research, 1966, 8 , 30-37.
Newman, L. M. The effect of reinforcement variations on
concept formation in retarded boys. Dissertation
Abstracts, 1966, 27, 317-318.
O'Connor, N., and Hermelin, B. Discrimination and reversal
learning in imbeciles. Journal of Abnormal and Social
Psychology, 1959, 59, 409-413.
Ohlrich, E., and Ross, L. E. Reversal and nonreversal
shift learning in retardates as a function of over
training. Journal of Experimental Psychology, 1966,
72, 622-624.
Osgood, C. E. Method and theory in experimental psychology.
New York: Oxford University Press, 1953.
Osier, S. F., and Weiss, S. R. Studies in concept attain
ment: 111, Effect of instructions at two levels of
intelligence. Journal of Experimental Psychology, .
1962, 63, 528-533.
Pishkin, V. Dimension availability with antecedent success
or failure in concept identification. Psychonomic
Science, 1965, 2, 69-70.
Reed, H. B. Factors influencing the learning and retention
of concepts: I, The influence of set. Journal of
Experimental Psychology, 1946, 36, 71-87.
Reichard, S., Schneider, M. and Rapaport, D. The develop
ment of concept formation in children. American
Journal of Orthopsychiatry, 1944, 14, 156-161.
Restle, F. A theory of discrimination learning. Psychol
ogical Review, 1955, 62, 11-19.
128
Rieber, M. Verbal mediation In normal and retarded chil
dren. American Journal of Mental Deficiency# 1964, 6 8 ,
| 634-641.
Rommetveit, R. Stages in concept formation and levels of
cognitive functioning. Scandinavian Journal of Psy
chology, 1960, 1, 115-124.
Rose, D., Smith, R. E«, and Robles, A. Some problems in
perceptual handicap of mentally retarded children.
Journal of Genetic Psychology, 1964, 104, 123-133.
Rosenberg, S. Problem-solving and conceptual behavior. In
N. R. Ellis (Ed.), Handbook of mental deficiency. New
York: McGraw-Hill, 1963. Pp. 439-462.
Sanders, B., Ross, L. E., and Heal, L. W. Reversal and non
reversal shift learning in normal children and
retardates of comparable mental age. Journal of Exper
imental Psychology, 1965, 69, 84-88.
Schenck, H. u., and Shepherd, D. Learning strategies of
the retarded in a four-choice situation with probabil
istic reinforcement and varying stimulus conditions.
Paper read at Western Psychological Association, San
Francisco, May, 1967.
Semler, I. J., and Iscoe, I. Concept interference and
paired associates in retarded children. Journal of
Comparative and Physiological Psychology, 1965, 60,
465-466.
Shaffer, L. H. Concept formation in an ordering task.
British Journal of Psychology, 1961, 52, 361-369.
Siegel, S. Nonparametric statistics. New York: McGraw-
Hill, 1956.
Siegel, S. M. Discrimination among mental defective, nor
mal, schizophrenic and brain damaged subjects on the
visual-verbal concept formation test. American Journal
of Mental Deficiency, 1957, 62, 338-343.
129
Silverstein, A. B., and Mohan, P. J. Performance of men
tally retarded adults on the color form sorting test. i
American Journal of Mental Deficiency, 1962, 67, 458-
462.
Silverstein, A. B., and Mohan, P. J. A reanalysis of the
color form sorting test performance of mentally retarded
adults. American Journal of Mental Deficiency, 1964,
69, 402-404.
Smith, M. P., and Means, J. R. Effects of type of stimulus
pretraining on discrimination learning in mentally
retarded. American Journal of Mental Deficiency, 1961,
6 6 , 259-265.
Smoke, K. L. An objective study of concept formation.
Psychological Monographs. 1932, 42 (4, Whole No. 191).
Spreen, O. Language functions in mental retardation, a
review: II, Language in higher level performance.
American Journal of Mental Deficiency, 1965, 70, 351-
362.
Stacey, C. L., and Cantor, G. The use of Zaslow's test of
concept formation on a group of subnormals. Journal of
Clinical Psychology, 1953, 9, 51-53.
Stedman, D. J. Associative clustering of semantic cate
gories in normal and retarded subjects. American
Journal of Mental Deficiency, 1963, 67, 700-704.
Stevenson, H. W. Learning of complex problems by normal
and retarded subjects. American Journal of Mental
Deficiency, 1960, 64, 1021-1026.
Stevenson, H. W. Discrimination learning. In N. R. Ellis
(Ed.), Handbook of mental deficiency. New York:
McGraw-Hxll, 1963. Pp. 424-438.
Sullivan, P. W. The effects of verbal reward and verbal
punishment on concept elicitation in children. Dis
sertation Abstracts, 1964, 24, 4807-4808.
Trabasso, T. R. Stimulus emphasis and all-or-none learning
in concept identification. Journal of Experimental
Psychology, 1963, 65, 398-406.
Trabasso, T., and Bower, G. Memory in concept identifica
tion. Psychonomic Science. 1964, 1, 133-134. (a)
Trabasso, T., and Bower, G. Presolution reversal and
dimensional shifts in concept identification. Journal
of Experimental Psychology. 1964, 67, 398-399. (b)
Trabasso, T., and Bower, G. Presolution dimensional shifts
in concept identification: a test of the sampling with
replacement axiom in all-or-none models. Journal of
Mathematical Psychology, 1966, 3, 163-173.
Vygotsky, L. S. Thought and language. New York: M. I. T.
Press and Wiley, 1962.
Weir, M. W. Developmental changes in problem-solving
strategies. Psychological Review, 1964, 71, 473-490.
Weir, M. W., and Stevenson, H. W. The effect of verbaliza
tion in children's learning as a function of chronolog
ical age. Child Development, 1959, 30, 143-149.
Weiss, A. A. The Weigl-Goldstein-Scheerer color-form
sorting test: classification of performance. Journal
of Clinical Psychology, 1964, 20, 103-107.
Wyckoff, L. B. The role of observing responses in dis
crimination learning. Psychological Review, 1952,
59, 431-442.
Zaslow, R. W. A study of concept formation in normal, men
tal defectives, and brain-damaged adults. Genetic
Psychology Monographs, 1961, 63, 279-338.
Zeaman, D., and House, B. J. The role of attention in
retardate discrimination learning. In N. R. Ellis
(Ed.), Handbook of mental deficiency. New York: McGraw-
Hill, 1963. Pp. 159-223.
Zeaman, D., and House, B. J. The relation of I.Q. and
learning. In R. M. Gagnd (Ed.), Learning and individual
differences. Columbus, Ohio: Merrill, 1966. Pp. 192-
212.
Zeaman, D., House, B. J., and Orlando, R. Use of special
training conditions in visual discrimination learning
with imbeciles. American Journal of Mental Deficiency,
1958, 63, 453-459.
Zeaman, D., Thaller, C., and House, B. J. Variability of
irrelevant stimuli in discrimination learning of
retardates. Journal of Experimental Child Psychology,
1964, 1, 89-98.
Zeiler, M. D. Component and configurational learning in
children. Journal of Experimental Psychology, 1964,
6 8 , 292-296.
Zigler, E. Social deprivation and rigidity in the perform
ance of feebleminded children. Journal of Abnormal and
Social Psychology, 1961, 62, 413-421.
Zigler, E. Rigidity in the feebleminded. In E. P. Trapp
and P. Himelstein (Eds.), Readings on the exceptional
child. New York: Appleton-Century Crofts, 1962. Pp.
141-162.
Zigler, E. Rigidity and social reinforcement effects in
the performance of institutionalized and noninstitu
tionalized normal and retarded children. Journal of
Personality, 1963, 31, 258-269.
Zigler, E., and deLabry, J. Concept-switching in middle-
class, lower-class, and retarded children. Journal of
Abnormal and Social Psychology, 1962, 65, 267-273.
APPENDIX
132
APPENDIX
INSTRUCTIONS
Introduction to the Task
A display of toys was shown to the subject and he
was told:
"Today we are going to play a funny kind of card
game. If you learn it well, you can win a prize.
See all these things? Which one would you like to
win?"
The subject was allowed to choose a prize.
"O.K., we'll put the ___________ over here and when
you have learned the game, you can have it. First
Mr. Itoga will show you some things and then we
will play the card game."
The experimenter then left the room and the research assist
ant administered the pretraining.
Pretraining
Group I (Exposure). Group I subjects were presented
the mimeographed page and told:
"See this piece of paper. I want you to fold it,
just like I am doing."
A longitudinal fold was demonstrated with a second page.
If the subject's fold approximated the edges of the paper
he was told:
133
j "Good," or "That's O.K." "Now fold it again, like
! this."
A horizontal fold was demonstrated.
The square box was set before S; the four small
cards displaying the geometric designs were handed to S;
and he was instructed,
"O.K., now I want you to put these cards into this
box. Put them in here one at a time." (Indicate
slot.) "That's right."
Group II (Matching training). The rectangle dis
playing the sets of dots was placed before S and he was
handed, one at a time, the four small dot cards, beginning
with the card containing one dot. The subject was told,
"See these sets of dots. Find a set that looks
just like this. ______ Very good. Now try this
one."
If S is correct, or
"No, it doesn't go there. Try it again."
If S is incorrect.
Group III (Form). The subjects were presented the
mimeographed page and a black waxed pencil and told,
"See the pictures? I want to see if you can trace
them. Take this pencil and draw along the lines,
like this."
The experimenter demonstrated, tracing one side of a figure
"Draw all around it. _________ Good. Now do the
others the same way.”
Next the rectangle displaying the six geometric
figures was placed before S. The smaller cards were handed
to S, one at a time, with the instructions,
"Find the one that looks just like this.
Very good. Now try this one."
'if S were correct. Or,
"No, it doesn't go there. Try it again."
If S is in error.
Concept Sorting Task
The trays and label cards were placed before S.
"O.K., now we will play the card game and you can
try to win your prize. See all these cards?"
The nine cards of Set I were displayed.
"Some of them go here, and some of them go here,
and some of them go here,"
E indicates each tray in turn
"but I'm not going to tell you where they go. You
have to figure that out for yourself. Put the
cards in here, or here or here, one at a time. If
you are right, I will say, 'Right' and you can try
the next card. If you are wrong, I will say,
'Wrong.' Then you must take the card out and try
it some place else. When you can do all the cards
two times, with no mistakes, you can win the
O.K., try it."
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Transfer Of The Partial Reinforcement Extinction Effect Across Tasks In Normal And Retarded Boys
PDF
The Effects Of Sex, Assigned Therapist Or Peer Role, Topic Intimacy, And Expectations Of Partner Compatibility On Dyadic Communication Patterns
PDF
Effects Of Video Tape Feedback Versus Discussion Session Feedback On Group Interaction, Self Awareness And Behavioral Change Among Group Psychotherapy Participants
PDF
Verbal Reports Of Emotional States And Onsets And Offsets Of Conditioned Stimuli
PDF
The Effects Of A Self Shock Procedure On Hallucinatory Activity In Hospitalized Schizophrenics
PDF
The Effects Of Two Types Of Experimenter Intervention And Schedules Of Reinforcement On Verbal Operant Conditioning Of Affective Self-References
PDF
Prediction Of Therapeutic And Intellectual Potential In Mentally Retardedchildren
PDF
Defense Choice And Identification
PDF
Prognostic Expectancy Effects In The Desensitization Of Anxiety Over Invasion Of Body Buffer Zones
PDF
Placement Characteristics Of Educable Mentally Retarded And Educationallyhandicapped Children: A Multi-Statistical Approach
PDF
An Experimental Study Of Two Moral Responses Of Two Socioeconomic Groups Based On Four Paradigms
PDF
Confirmation Of Expectancy And Changes In Teachers' Evaluations Of Student Behaviors
PDF
The Enhancement Of Eeg - Alpha Production And Its Effects On Hypnotic Susceptibility
PDF
The Concept Of Sexual Identity In Normals And Transvestites: Its Relationship To The Body-Image, Self-Concept And Parental Identification
PDF
The effect of noise on intellectual performance as related to personality and social factors in upper division high school students
PDF
Imagery And Response Styles In Desensitization
PDF
Use Of College Students In A Social Therapy Program With Chronic Schizophrenics To Produce Changes In Mood And Social Responsiveness
PDF
The Effect Of Dissonance In Self-Esteem On Susceptibility To Social Influence
PDF
The Conditioning Of A Discriminative Stimulus Measured As An Orienting Reaction In Profoundly Retarded Blind Children
PDF
The Effects Of Justice, Balance, And Hostility On Mirth
Asset Metadata
Creator
Smith, Ruth Ellen
(author)
Core Title
An Application Of A Two-Stage 'Attention' Model To Concept Formation In The Mentally Retarded
Degree
Doctor of Philosophy
Degree Program
Psychology
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
OAI-PMH Harvest,Psychology, clinical
Language
English
Contributor
Digitized by ProQuest
(provenance)
Advisor
Jacobs, Alfred (
committee chair
), De Nike, L. Douglas (
committee member
), McIntrye, Robert B. (
committee member
)
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c18-580063
Unique identifier
UC11360177
Identifier
6717702.pdf (filename),usctheses-c18-580063 (legacy record id)
Legacy Identifier
6717702.pdf
Dmrecord
580063
Document Type
Dissertation
Rights
Smith, Ruth Ellen
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