THE EFFECTS OF OBSERVER AGE AND
TYPE OF TASK ON THE IMITATION
OF ADULT AND PEER MODELS
GERALD CLEMMER MUSSELMAN
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
Gerald Clemmer Musselman
To the primary sources of this work--my wife, Julie, and
my nephew, Kendall
The fact that this investigation deals with models provides the
author with the pleasant opportunity to thank four of his own who compose
his doctoral committee--Dr. William D. Wolking, chairman, Dr. Ben Barger,
Dr. Richard J. Anderson, and Dr. Audrey S. Schumacher. The author would
like to first of all express his appreciation for each of them as persons
and for the warm associations experienced with them. Secondly, he would
like to acknowledge his substantial debt to each for the various kinds
of guidance-knowledge-encouragement which made his professional training
as well as this research endeavor an enriching experience. Special
appreciation is offered to Dr. Wolking for his personal interest in the
author and his willingness to endure the author's impatience to assure a
This writer is also very grateful to his friend and colleague,
Dr. Philip Costanzo, for serving as the adult model in the experiment and
to Timothy Savoy, Jeffrey Grater, and Hugh Davis, Jr.,for serving as the
first, fifth, and seventh grade models, respectively.
TABLE OF CONTENTS
LIST OF TABLES
LIST OF FIGURES
RESULTS . .
SUMMARY . .
BIOGRAPHICAL SKETCH .
. . . . . . . . . . . iv
. . . v ii
LIST OF TABLES
1. Number of Subjects in Each Grade-Treatment Group . . . 18
2. Model Role Behaviors in the Two Experimental Groups . 20
3. Summary of Analysis of Variance for Task 1: Superstitious-
irrelevant Behavior . . . . . . . .... . 31
4. Summary of Analysis of Variance for Task 2: Guessing
Behavior . . . . . . . . . . . . . 31
5. Summary of Analysis of Variance for Task 3: Risk-taking
Behavior . . . . . . . . . . . . . 32
6. Summary of Analysis for Variance for Task 4: Spontaneous
Talking Behavior . . . . . . . .. .. . .. 32
7. Summary of Analysis of Variance for Task 5: Drawing Be-
havior . . . . . . . . . . . . . 33
8. Summary of Analysis of Variance of Task Behavior as a
Function of Age, Modeling Situation, and Task . . ... 33
9. Summary of Newman-Keuls' Tests on C Minus El Differences
for Tasks 1 Through 5 . . . . . . . . . 34
10. Summary of Newman-Keuls' Tests on C Minus E2 Differences
for Tasks 1 Through 5 . . . . . . . . . 34
11. Summary of Newman-Keul's Tests on El Minus E2 Model Potency
Differences for Tasks 1 Through 5 . . . . .... .35
12. Usage Across Age of the Finger Oscillation Device . .. .35
13. Number of Responses Identical to Responses of Models
Across Six Tasks . . . . . . . . ... . .. 36
14. Differences in Modeling Situation Potency . . . ... 36
15. Summary of Score-ranks in Treatment Groups Across the
Five Tasks . . . . . . . . ... .. . .. 37
LIST OF FIGURES
1. Scores on Task 1 According to Age and Modeling Group . 38
2. Scores on Task 2 According to Age and Modeling Group . 39
3. Scores on Task 3 According to Age and Modeling Group . . 40
4. Scores on Task 4 According to Age and Modeling Group . 41
5. Scores on Task 5 According to Age and Modeling Group . 42
6. Usage Across Age of the Finger Oscillation Device .... 43
7. Responses Identical to Responses of Models Across Six Tasks. 44
Imitation has been a concept in psychology dating back at least
to the turn of the century, when it was discussed by men such as Morgan
and McDougall. However, at that time, due to its association with
instinct, the concept was framed in terms not susceptible to empirical
verification. As a result, the concept of imitation remained suspect
and unscientific, but it experienced revivals of interest under the
guise of different terminology. The operant-reinforcement view of
imitation has been one of the more recent paradigms under which imitation
has been considered, especially in relation to early development and the
learning of language (Miller and Dollard, 1941; Bijou and Baer, 1965;
Skinner, 1953;1957). This point of view has led to important research
(Baer and Sherman, 1964) and clinical applications (Metz, 1965; Risley,
1966), especially with children.
However, the work of Bandura (1962;1965a;1965c), which will be the
basis of this investigation, represents another current, exciting renewal
of interest in the area of imitation under the labels of vicarious learn-
ing and modeling. As Bandura (1965a) points out, a primary source of new
behavior or new combinations of previously learned behaviors for humans
is the behavior of other humans. It is obvious that many complex and
important human behaviors are learned through one person observing the
behavior of another person, the model. It is also obvious that such
modeling procedures are a very efficient and, at times, indispensable way
of passing along behavior, even whole repertories of behavior. Modeling
procedures run the gamut of age ranges and become more explicit as age
increases, especially as verbal behavior increases. A combination of
verbal (symbolic) and demonstrative (live) aspects in modeling procedures
is assumed to be optimally effective. While all this is certainly no
new revelation, Bandura correctly stresses that such an emphasis is in
contrast to the emphasis that current learning theories place on operant
conditioning, learning by successive approximations, and reinforcement:
Research and theoretical interpretations of learning processes
have focused almost exclusively on a single mode of response
acquisition which is exemplified by the operant or instrumental
conditioning paradigm. In this procedure the organism is im-
pelled, in one way or another, to perform responses under
specific stimulus conditions and, through differential rein-
forcement of spontaneously emitted variations in behavior, new
patterns are developed or existing repertories are brought under
new discriminative stimulus control. It is generally assumed
that the principles governing the latter mode of response
acquisition account also for social-learning phenomena occurring
under naturalistic conditions.
The continued adherence to a relatively narrow range of learning
principles and procedures stems primarily from the fact that
certain critical conditions that obtain in real-life situations
are rarely, if ever, reproduced in laboratory studies of learning.
Thus, in laboratory investigations experimenters arrange com-
paratively benign environments in which errors will not produce
fatal consequences for the organism. By contrast, naturalistic
environs are loaded with potentially lethal consequences that
unmercifully befall those who happen to perform hazardous errors.
For this reason, it would be exceedingly injudicious to rely
primarily upon trial-and-error and successive approximations
methods in teaching children to swim, adolescents to drive
automobiles, or adults to master complex occupational and social
tasks. If rodents and pigeons toiling in Skinner boxes and various
mazes could likewise get electrocuted, dismembered, or extensively
bruised for errors that inevitably occur during early phases of
learning, it is a reasonably safe prediction that few of these
venturesome subjects would ever survive the shaping process.
Apart from questions of efficiency...and survival, it is doubtful
if many classes of responses would ever be acquired if social
training proceeded solely by the method of approximations through
differential reinforcement of emitted responses..."(Bandura,
1965c, p. 1).
It is also clear that while operant methods are very efficient
for strengthening and maintaining behavior already possessed by individ-
uals, such methods are very often inefficient for developing new be-
havior (Bandura, 1965a). The point is that work on imitative learning
and modeling is generating meaningful, much needed investigations into
the conditions of human learning in social situations. Social situations
represent the settings for a large percentage of human learning, and
there is much yet to be learned about the effects of such settings. The
study of modeling and social interaction effects also represents meaning-
ful work on human learning, especially developmental human learning,
which is more relevant to real life situations than is frequently the
case in other approaches to learning (Epstein, 1962).
Vicarious Learning and Identification
Given the importance of the area of study, the question becomes
one of how learning takes place through modeling--the question of vicar-
ious'learning. Bandura (1965c) has summarized various viewpoints on
vicarious learning and distinguished his position from the others (Bandura
and Walters, 1963). He notes that instead of imitation there are now
many terms used to refer to essentially the same phenomena-- "observational
learning," "copying," "social facilitation," "vicarious learning," "con-
tagion," "identification," and "role playing" (Bandura, 1965c, p. 2).
Because Bandura thinks that the same basic process of learning is involved
regardless of the models, stimulus conditions, content, and generality of
what is learned, he does not concern himself with the above variety of
terms. He uses the terms imitative, observational, and vicarious learn-
ing interchar,. :,bly to refer to the differences in behavior resulting
from modeling stimuli.
His definition and views of vicarious learning are as follows:
...a vicarious learn event is defined as one in which
new responses are acquired or the characteristics of
exist response re rtories are modified as a function
of observii the behavior of others and its reinforci
con Jences, without the modeled res ses beir overtly
performed by the viewer duri-- the exposure period. In
demonstrating vicarious learn phenomena, it is therefore
necessary to employ a nonresponse acquisition procedure in
which a subject simply observes a mode 's behavior, but
otherwise performs no overt instrumental res, -iss, nor is
administered any reinforcing stimuli during the riod of
acquisition. Any learnir- that occurs under these limiting
conditions is purely on an observational or covert basis.
This mode of res ise acquisition is accordir-ly designated
as no-trial learning, since the observer does not e in
any o)v 7 p ",:l; li although. . he may require multiple
&,j\,-; ..- in order to reproduce th-e mod led stimuli
accurately. moreover, the development of ed iat;ona res .-
es, in the form of i final and implicit verbal representations
of the perceived stimulus events, nay play a critical role
in the vicarious learning process (Bandura, 1565c, p. 2).
Unlike nost previous accounts of modelir- effects, which
tend to highlight the reinforcing stimulus control of
matchin- responses, the theory propounded '. the author em-
phasizes the function of representational processes in
observational learn Accordi o this forruat ion,
matching responses are acquired or the ais of stimulus
conti Jity and are mdiaed by cue- reducing s lic re-
sponses wch exercise discriminative stus control over
corres di overt rfor nces. Thu in eis d
re onse acquisition, i in and verb r rse nations
of od ling sti uli con iue the encuri ro uc of
observation-l i rinc ile the perc ta and i t i ve
aspects of vicarious learning are given ei sis, t is
recognized that motivational and reinforce ent variables may
influence indirectly the level of r';':" c :, l ; :'' :' : by
augmenting or reduci the occurrence of requisite observing
responses and facilitative covert rehearsal. There is
considerable research evidence, however, that the performance
of previously learned matching responses is primarily
governed by reinforcement-related variables... The theory
advanced ... suggests that vicarious learning may be analyzed
in the same manner as other associative learning processes
... The fact that vicarious learning experiments employ social
cues rather than nonsense syllables does not result in an
acquisition process that is fundamentally different from
traditional associative learning (Bandura, 1965c, pp. 41-42).
Bandura goes on to contrast his "contiguous sensory stimulation"
position with three other categories of learning positions and indicates
his position better accounts for the phenomena of vicarious learning.
According to Bandura (1965c, pp. 3-8),"associative and classical condition-
ing theories" (Humphrey, Allport, Holt, Piaget) do not account for novel
responses in the observer resulting from the model-observer interaction;
"instrumental conditioning theories" (Miller and Dollard, Skinner) do not
account for novel responses' being acquired through observation prior to
performance and reinforcement and often confuse acquisition and performance;
"sensory feedback theory" (Mowrer) involving classical conditioning cannot
adequately account with peripheral feedback and proprioceptive cues for
imitative responses without reinforcement of either model or observer.
Because the focus of the present work will be primarily on aspects of per-
formance rather than learning,these theoretical differences need not be
pursued further here.
Concerning the allied concept of identification, as compared with
imitation, the following stand is taken (Bandura, 1962, p. 215):
This type of learning is generally labeled "imitation" in
behavior theory, and "identification" in most theories of
personality. These concepts,however, are treated ... as
synonymous since both encompass the same behavioral phenom-
enon, i.e., the tendency for a person to match the behavior
or attitudes as exhibited by actual or symbolized models.
This appears to be the most fruitful position to take and coincides with
the "identification as behavior" viewpoint, one of the three main view-
points on the nature of identification found in a review of the literature
by Bronfenbrenner (1960). The "identification as behavior" position
appeared to be the most useful of the three positions in the opinion of
Bronfenbrenner. As he pointed out, the other two main views of identifi-
cation--"identification as a motive," and "identification as process"--
are less empirical in nature and do not lend themselves to empirical study.
While some investigators do not favor the "identification as behavior"
position (Kagan, 1958), the point is that even identification itself, as
a concept, especially as a concept connoting unconscious processes, seems
to hinder definitive work in the area it is concerned with delineating:
Identification, as nominally defined, tends to preclude
adequate and valid operational definition and empirical test.
Identification is a hypothetical construct, which seems
destined to remain a hypothetical construct. As such, it can
explain nothing. Nor does it seem to lead to a definition of
those variables and the process by which they interact which
would constitute an explanation, subject to experimental
verification (Martin, 1954, p. 214).
Review of Studies
The behavioral effects of exposure to modeling stimuli fall into
three main categories (Bandura, 1965a, pp. 320-321):
(1) ... the observer may acquire new responses that did not
exist in his behavioral repertoire. In demonstrating this
modeling effect experimentally, the model exhibits responses
that the observer has not yet learned to make, and he must
reproduce them in substantially identical form...
(2) Exposure to models may also strengthen or weaken inhibitory
responses in the observer. These inhibitory and disinhibitory
effects are evident when the frequency of imitative and non-
matching responses increases or decreases, often as a function
of rewarding or punishing response consequences to the model...
(3) ... the behavior of models may elicit previously learned
responses that match precisely or bear some resemblance to those
exhibited by the model. This response facilitation effect can
be distinguished from disinhibition when the behavior in question
is not likely to have incurred punishment and, therefore, any
increase in responsivity is not attributable to the reduction of
In more specific terms, some of the experimental results are as
follows. In order to assess the effectiveness of modeling and operant
conditioning, Bandura and McDonald (1963) attempted to change the moral
orientations (Piaget's objective and subjective moral orientations) of
5 to 11 year old children by the following three methods: (1) having
children observe adult models express moral judgements counter to the
group's orientation and be reinforced with approval, reinforcing the
children with approval for adopting the model's responses; (2) having
another group observe the models reinforced but not reinforcing the
children's matching behavior; (3) having a third group receive no exposure
to models but reinforcing the judgements of the children which were counter
to children's judgement tendencies. The children in the modeling groups
were not significantly different from each other but showed significantly
higher modifications in their moral judgements in the direction of the at-
tempted influence than the children in the group using only reinforcement,
who did not change significantly. Also, the groups maintained their altered
positions in a new situation in which the models and social reinforcement
were absent, suggesting modeling effects can be long-lasting.
Perhaps the classic studies were the ones in which the social trans-
mission of novel aggressive responses and film versus live models were
investigated (Bandura, Ross, and Ross, 1961; Bandura, Ross, and Ross, 1963b).
In the 1961 experiment, children from a nursery school, with an average age
of 52 months, were assigned to three conditions: (1) observation of aggres-
sive adult model; (2) observation of inhibited nonaggressive model; (3) no
exposure to a model. The experimental groups observed an adult model in-
teract physically and verbally with a large plastic doll and then were
frustrated and tested for imitative and nonimitative aggressive behavior.
Approximately the same paradigm was used in the 1963 experiment with same
age subjects from the same school. Three groups were used: (1) observation
of real-life aggressive model; (2) observation of the same aggressive model
on film; (3) observation of a model on film in the costume of a cartoon cat
showing the same aggressive behavior. Children who observed the aggressive
models showed roughly twice as much aggressive behavior as the children in
the nonaggressive and control groups. The nonaggressive group showed sig-
nificantly less aggressive behavior than the controls. Children who observed
the aggressive models showed many precisely matching physical and verbal
aggressive acts. This behavior occurred only rarely in the other groups.
The film-mediated aggressive models and the cartoon character were as ef-
fective as the real-life models in promoting aggressive behavior, although
the modeling effects were not as prominent in the cartoon character group.
In order to investigate whether reinforcement functions more as a
performance or more as a learning variable, Bandura (1965b) had nursery
school subjects observe an aggressive film-mediated model rewarded, punished,
or receive no consequences for his aggressive acts. The results indicated
that children in the model-punished group gave significantly fewer matching
responses than children in the other two groups. However, when the children
in each group were offered rewards if they could reproduce the model's
aggressive behavior,the previous differences among the three groups dis-
appeared, indicating an equal amount of learning in each group. The results
supported Bandura's contiguity theory--reinforcement of the model influenced
the observers' performances but not their acquisition of matching responses
The influence of adult and peer models has also been studied, in
relation to aggressive behavior and patterns of self-reinforcement. Bandura
and Kupers (1964) had 7 to 9 year old children in one group observe either
peer or adult models who adopted a high criterion for self-reinforcement
on a bowling game, while those in another group observed models employing
a low criterion for self-reinforcement. It was found that the subjects'
patterns of self-reinforcement were closely related to the models they had
observed, with the adult models having a more powerful influence than the
peer models. The influence of peer models and adult models has been studied
in another situation in an interesting investigation by Hicks (1965). Hicks
looked into the effects of adult and peer film-mediated aggressive models
on imitation and retention of modeled responses. Children from 3 to 6 years
of age viewed either male or female adult or peer aggressive models and then
were tested for imitative behavior two times. The first time was immediately
after the observation of the models and a mild frustration. The second time
was 6 months later, following the same mild frustration but without any re-
observation of the models. It was found that the peer male model group
showed the greatest immediate modeling influence. After 6 months, however,
the adult male model group showed the most matching behavior, although none
of the groups performed even close to the amount of imitative behavior that
they had displayed immediately after exposure to the models. Tests of re-
tention after 6 months also revealed a significantly greater number of
matching responses had been retained than had been performed.
In another investigation of modeling and self-reinforcement, Mischel
and Liebert (1966) found that fourth grade children tended to impose on
younger children, second grade children of the same sex as the subjects,
the identical self-reward criteria they imposed on themselves when playing
a game. The subjects previously played the game with an adult model who
imposed a level of self-reward on the subject which was equal to, higher
than, or lower than the level employed by the model. As predicted, in the
absence of the model the subjects on whom high criteria were imposed and
who also observed their model employ equally high criteria were the most
stringent in applying self-rewards to themselves and younger children. The
subjects who had observed high modeled criteria but had been permitted
self-reward for lower performances were the least stringent. When imposed
and observed criteria were discrepant the least stringent criteria tended
to be employed by the subject. There were no differences between the self-
reward criteria employed by the subject for himself and the criteria imposed
by the subject on younger children if the subject demonstrated the task to
the younger child immediately following his performance of the task alone.
However, if immediately after the modeling situation the subject demonstrated
the task to the younger child before performing it alone, there were some
differences between the self-reward used by the subject and imposed on
the younger child.
Studies by Mussen and Parker (1965) on 5 year old girls and by
Bandura and Huston (1961) on 3 to 5 year old children indicate that
nurturance or positive social interaction displayed by the model to the
subject leads to more imitation of task irrelevant (non-problem-solving)
behavior than to more imitation of task relevant behavior. However, the
opposite of this was found by Henker (1963) with 6 to 10 year old boys in
a study of the effects of rewarding, critical, and neutral models in various
situations. The play behavior of the subjects showed no difference in amount
of imitation, but task relevant behavior on a problem-solving task was re-
lated to the type of model observed and his previously displayed (experiment-
ally manipulated) competence. In this regard, Aronfreed (1964) noted an
absence of any consistent direct relationship between amount of parental
nurturance and extent of internalization.
There have been numerous other studies by Bandura and by other in-
vestigators related to the realm of modeling but not especially relevant or
more elucidating to the purposes of the present study. Surprisingly, how-
ever, there is a great paucity of developmental studies in the area of
modeling, as in many other areas. Most of the older studies in the area
dealt exclusively with a very narrow age group, usually preschool children,
and used criteria and situations such as verbal responses to questions con-
cerning imitation, simple manipulation of dolls, or verbal reports of what
they would do in hypothetical situations, rather than actual behavioral
samples in a "real-life" situation. Some examples are as follows. Hartup
(1964), in doll play situations with 3 to 5 year old children, found that
there was a moderate generalization across situations of same-sex imitation
but not of opposite-sex imitation. When nonimitation could occur as a
response choice it occurred more frequently than imitation. This nonimita-
tion tendency had been previously pointed out by McDavid (1959) in 3 to 5
year old children where imitative behavior seemed to be related to child-
rearing practices and sibling constellation but not to intelligence or
dependency. Brown (1956), studying sex-role preferences in 5 and 6 year
olds, found boys to have a stronger preference for the masculine role than
girls for the feminine role, no differences between upper and lower middle
class groups, and that boys having only sisters or both brothers and sisters
scored more feminine than boys having only brothers. Brown used a sex-role
preference scale called the It Scale for Children (ITSC). Hartup and Zook
(1960), also using the ITSC, studied the sex-role preferences of 3 and 4
year old children and found the 4 year old boys scored more masculine than
the 3 year old boys, and the 4 year old girls scored more feminine than the
3 year old girls. Lynn (1959) reviewed research on the development of
identification and suggested that with increasing age males become more
firmly identified with the masculine role, responding more to a cultural
stereotype of the masculine role. With increasing age females become less
firmly identified with the feminine role, responding more to aspects of
their own mother's role. Gelfand (1962) got results in work with fifth
grade children which indicated that failure on tasks in comparison to a
peer's success on the tasks led to matching of the peer's responses on a
picture preference task. Task performances inconsistent with the child's
self-evaluations led to more verbal conditioning in such children than in
children whose task performances were consistent with their self-attitudes
(performances experimentally manipulated). DeRath (1963) found prohibitive
verbal instructions paired with a film-mediated aggressive model to be
effective in inhibiting the performance of aggressive behavior in nursery
school boys. The inhibition group performed less aggressive responses than
a group which saw the same film but received no prohibitive instructions.
Implications for Future Study
A number of conclusions and suggested areas for further study emerge
from these investigations of imitative behavior. It has been conclusively
shown that modeling procedures can lead to changed behavior in the observer
and that live and filmed models can be equally effective. The primary
relevant variables seem to be the age of the model, the age of the observer,
the sex of the model and the observer, the type of task or the type of
behavior involved, and the performance situation.
The first and most obvious characteristic of the modeling studies
is that most of them have been done with preschool children or with very
limited age ranges. This characteristic has resulted in the inability to
generalize across age levels on significant variables such as the effects
of model age, type of task, or type of performance situation. Second, the
range of modeling tasks or behaviors has been very narrow,also. Aggressive
behavior has been the primary focus in most of the recent studies. Third,
the modeling paradigm has been oversimplified by having the subject observe
only one model. Outside the laboratory a person faced with a task is or
has usually been exposed to more than one model and more than one method
of doing the task, resulting in the necessity for dealing with the various
methods observed. He may combine aspects of different methods, employ
a few or no aspects of any method observed, or follow precisely one of the
observed methods. Such a situation is often made more complex by adult
models being in opposition to peer models and the changes in influence of
each according to the age of the observer. Fourth, the situation in which
the observer performs, following exposure to modeling stimuli, probably
affects what he will perform as opposed to what he learned. For example,
whether the observer performs in the presence of an adult, a peer, an older
person, a younger person, alone, or in a group is a factor relevant to what
These four variables have either not been studied at all or have
only been touched on in modeling experiments. This investigation attempts
to study the first three of these variables--age of observer, type of task
or behavior, number and age of models.
This investigation was an attempt to analyze the effects of observer
age and type of task on the imitation of adult and peer models. The term
imitation, as it is used herein, means "response facilitation effect," as
defined by Bandura. Thus, the emphasis is on performance rather than
Three different observer age levels and five types of tasks were
used. The situation studied was one in which a subject performed five tasks
following the observation of an adult and a peer performing the tasks
in contrasting ways.
The three different observer age levels were first, fifth, and
seventh graders. These three grade levels, equivalent to ages 6, 10, and
12, were hypothesized as different on a chronological basis and on the
basis of developmental theory. The three age-grade levels can be viewed
as approximating developmental phases III, IV, and V in the scheme result-
ing from Maier's (1965) integration of the developmental sequences of
Erikson, Piaget, and Sears. Apart from intra-personal characteristics of
these children in these three developmental phases there are inter-personal
interactions which seem to be different. At age 6 the child is still
making his entrance into the world of structured inter-personal relation-
ships with those his own age and with adults outside of his family. In
these early stages of schooling, adults, primarily in the form of teachers,
determine many things he does and require specific methods for doing various
tasks. Thus, the child is dependent on adult models at this stage for new
types of behavior and knowledge and must change other types of previously
learned behavior to conform to the school routine. In other words, the
adult is a very potent model for the 6 year old child. This potency of the
adult model, however, apparently declines with age but does not seem to
decline in a linear fashion. As the child grows older the adult model may
remain very potent for certain types of behavior in certain types of situa-
tions but may have very little influence, compared with a peer model, over
other types of behavior in other situations. This set of circumstances
would seem to be characteristic of the 10 year old child, in contrast to
the 6 or 12 year old child. The 12 year old could be seen as representing
the other end of the modeling spectrum, the generalized dominance of the
peer model instead of the adult model. The 12 year old has already ac-
quired much knowledge, many of the basic inter-personal skills, and a
system of social relationships in which conformity to peer cultures and
peer approval are highly reinforcing. It is suggested, therefore, that
these three age-grade ranges may be considered representative of a
model-potency continuum. The 6 year old group would be characterized by
generalized high adult model potency, the 10 year old group by lower, less
generalized adult model potency, and the 12 year old group by low adult
model potency and generalized high peer model potency. The use of these
three age-grade levels should enable one to test the assumption that there
are differences in model potency between adult and peer models as observer
1. Imitation will occur at all 3 age levels and on all 5 tasks,
though not necessarily at each age level on each task.
2. There will be differences in adult versus peer model potency
among the three age levels on each of the five tasks.
3. The adult model will be more potent than the peer model at the
lowest age level, and the peer rodel will be more potent than
the adult model at the highest age level on tasks 2 to 5.
4. There will be differences in modeling situation potency among
the three modeling situations on each of the five tasks.
5. The highest age level will show less superstitious-irrelevant
behavior on the superstitious-irrelevant task than the lowest
Male subjects were used exclusively. The total populations of male
students from the first, fifth, and seventh grades of the P.K. Yonge
Laboratory School, Gainesville, Florida were used.
The P.K. Yonge Laboratory School is a department in the College of
Education of the University of Florida and has an enrollment of approximately
930 students. Sixty pupils are enrolled in each grade from kindergarten
through the seventh grade, with an equal number of males and females in
each grade. Part of the school's enrollment formula is that 50% of the
pupils in each grade are children from the families of the academic faculty
of the University of Florida. The school devotes much attention to the
individual student and provides pre-internship participation and observation
experiences for undergraduate students in various fields. Practicum ex-
periences for graduate students are also provided. The school serves as a
research center for educational and behavioral science endeavors.
There was a total of 79 subjects used, 24 first graders, 28 fifth
graders, and 27 seventh graders. This represented all the males in each
grade except a few who were absent from school and missed the experimental
sessions. The subjects in each age-grade level were randomly assigned to
the three treatment groups--a control group and two experimental groups--
by using a table of random numbers and the procedure suggested by Lindquist
(1956). The number of subjects in each age and treatment group can be seen
in Table 1.
Number of Subjects in Each Grade-Treatment Group
Grade Level C El E2 Total
1 8 8 8 24
5 9 11 8 28
7 9 10 8 27
Total 26 29 24 79
The mean age of the first graders was 6.7 years with a range of
75-85 months. The mean age of the fifth graders was 10.7 years with a
range of 122-133 months. The mean age of the seventh graders was 12.6
years with a range of 144-158 months.
The three peer models were children of colleagues of the experiment-
er. None of the peer models was from P.K. Yonge Laboratory School. The
ages of the first, fifth, and seventh grade models were 6 years 9 months,
10 years 2 months, and 11 years 9 months, respectively. The adult model
was a 25 year old psychology graduate student.
Tasks and Apparatus
There were five tasks or types of behavior under study. These five
tasks were presented to the subjects by way of video tape. Each group at
each age-grade level was presented with a different video tape via closed
circuit television. The control group video tape consisted of the exper-
imenter presenting the task materials and instructions. The video tape
for the first experimental group consisted of the same presentation of
the task materials and instructions as contained in the control group
tape with the addition of a peer and an adult doing each of the tasks in
contrasting ways. The video tape for the second experimental group was
exactly like the first experimental group tape except that the model
roles were reversed. (See Appendix A.) Consequently, seven video tapes
were made. The same tape was used in all three control groups, but each
of the experimental groups had its own separate tape.
The five tasks were as follows:
1. Superstitious-irrelevant task: First, a finger oscillation de-
vice was presented. This device consisted of two 5-digit
counters, mounted on an 11 x 6 piece of wood, which automatic-
ally recorded the number of taps. The experimenter stated
that some people think tapping brings good luck and helps them
perform better on the tasks. He also stated that one could
tap as much as one wanted to, immediately, and at any time while
doing the tasks. One model said, "I think I'll tap." He then
tapped 25 times, initially, and 25 times between each task for
a total of 100 taps. The other model said, "I think I won't
tap." He did not tap at all, initially, nor at any other time
during the tasks.
2. Guessing task: Second, a glass jar containing 500 dried beans
was presented. The experimenter posed the question of how many
beans there were in the jar. One model guessed 300 beans, and
the other model guessed 700 beans.
3. Risk-taking task: Third, a dart board composed of 5 concentric
circles was presented. The experimenter explained that the
closer to the center of the board the darts hit, the higher the
score. The circles scored 10, 25,50, 75, and 100 points. The
experimenter further explained that points could also be scored
depending on how far from the board one stood to throw the darts.
Sixteen marks numbered 0 to 15, each one-half foot apart, were
pointed out on the floor. The closest one could stand was at
line 0, 4 feet from the board, and the farthest away one could
stand was line 15. The number of the line at which one stood
was added to the score of whatever each dart hit on the board.
It was emphasized that the closer one s:ccd to trc orct, t'e
easier it was to hit the hi ,int circ ut the f crt
away one stood, the more points one ot cdlc t ha; t csr
hit on the board. There were five dra s o no. A. n to
be thrown from the same line, whichvcr ie i .s c oscn.
The one model said, "I think .. c re t st
points by throwing from line 15,"and thry. froi ine i5. nc
other model said, "I think you can score the ost points
throwing from line O,"and threw from line 0.
4. "- wztazco,a taking tz I: Fourth, a micro e r.c tape reccrd-
er were presented in order to measure "tone of voice. irsrc-
tions were given to say whatever one wanted to se and s m
or as little as one wanted to say. The one rodl saic ne ;rd,
"hello," and the other model used 75 words, relying on o previous-
ly prepared script.
5. .rcwi-: task: Fifth, paper and 15 different cozor: p:cis
were presented. Instructions were given to draw tw- st cse
one could draw in 5 minutes, using as few or as cay coors as
one wished to use. One model said, "I think i'1 use a 1 5
colors in my drawing," and used all 15 colors. Tae o:her model
said, "I think I'll use just I color in my drawing," and used
just I color.
The role behaviors performed by the adult and peer models in the
two experimental groups are presented in Table 2.
Model Role Behaviors in the Two Experimental Groups
Task El E2
Adult Model Peer Model A, cul:'see Peer .Yde1
1. Number of taps 100 0 0 1
2. Number of beans 700 300 303 73-
3. Number of line 15 0 0
4. Number of words 75 1 1 75
5. Number of colors 15 1 1 15
The order of performance for the two models alternated from task to task
in the order stated above in the task descriptions.
The experiment took place at P. K. Yonge Laboratory School in
January 1967. The subjects were run in the following grade order,
seventh, first, and fifth. This order was determined by school space
restrictions and commitments. The experiment lasted 6 days, and each
grade was run in 2 days. The three groups per grade were run in the
following order. The C and El groups were run the first day in the
morning and afternoon, respectively. The E2 group was run the morning
of the second day.
The following experimental procedure was used for each group. The
group was called as a group and taken to the experimental room by the
experimenter. At this time no task apparatus was in sight, only the video
tape equipment. The experimenter then introduced himself and the situation,
told the subjects what the procedure was going to be, and gave the instruc-
tions. (See Appendix B.) The situation was presented as an attempt to find
out how well boys of different age and grade levels could do on the tasks.
Following an opportunity to ask questions, the group was shown the appro-
priate video tape and then sent back to their classrooms. The experimenter
then set up the task apparatus, called each of the subjects individually
into the room, had him perform each of the tasks, and sent him back to his
classroom. Each subject was asked the same questions and given the same
instructions during his individual performance. (See Appendix C.)
The length of the control group video tape was 7 minutes. The six
experimental group video tapes were approximately the same length,
ranging from 13 1/2 to 16 minutes. The mean individual subject-running
time was 9.6 minutes and ranged from 8.0 to 11.1 minutes.
Scoring and Analysis
The tasks were scored as follows:
behavior Number of taps
2. Guessing behavior Number of beans guessed
3. Risk-taking behavior Number of line
4. Spontaneous talking Number of words
5. Drawing behavior Number of colors
The experimenter recorded the subject's estimate on tasks 2 and 3.
The subject's behavior on tasks 1, 4, and 5 was recorded automatically
on the counter, tape, and paper, respectively, and thus required no active
scoring by the experimenter in the presence of the subject.
The scores for each task were analyzed with a 3 x 3 unweighted means
analysis of variance (3 age-grade levels x 3 modeling situations) and New-
man-Keuls' tests on row and column groupings of cell means. The hypotheses
of the investigation are mostly concerned with cell mean differences.
Therefore, row and column cell means were compared even when one of the
two F's was not significant or when both were not significant but the
interaction F was significant (Winer, 1962).
Hypothesis 1, concerning the presence of imitation, was tested by
comparing the three treatment means at each age-grade level. A significant
difference between either of the experimental groups and the control group
was defined as establishing the presence of imitation at that age-grade
level. If such a difference occurred at least once in each of the five
tasks, and if each age-grade level showed a significant difference at
least once, hypothesis 1 would be confirmed.
Hypothesis 2, concerning differences in adult versus peer model
potency at each age-grade level, was tested by the analyses used to test
hypothesis 1. However, for hypothesis 2, attention was focused on signif-
icant differences between the two experimental groups at each age-grade
level. A significant difference between the two experimental groups at
an age-grade level was defined as establishing the presence of a difference
in model potency at that age-grade level. If such a difference occurred
at least once in each of the five tasks, and if each age-grade level showed
a significant difference at least once, hypothesis 2 would be confirmed.
Hypothesis 3, concerning directional differences in model potency
related to observer age, was tested by employing the directional results
of hypothesis 2. A significant difference in the predicted direction at
both the lowest and highest age-grade levels on a task was defined as es-
tablishing the presence of a difference in model potency between the
highest and lowest age-grade levels on that task. If such differences
occurred on tasks 2 through 5, hypothesis 3 would be confirmed.
Hypothesis 4, concerning differences in modeling situation potency
among the 3 age-grade levels, was tested by comparing the 3 age-grade
levels under each of the three modeling situations. A significant differ-
ence between any two of the three age-grade levels under a modeling situa-
tion was defined as establishing the presence of a difference in potency
for that modeling situation. If such a difference occurred at least once
in each of the five tasks, hypothesis 4 would be confirmed.
Hypothesis 5, concerning differences on the superstitious-
irrelevant task related to observer age, was tested by comparing the
main effects for the highest and lowest age-grade levels in the analysis
of variance for task 1. A significantly lower total for the highest age-
grade level than for the lowest age-grade level was defined as establish-
ing the presence of less superstitious-irrelevant behavior in the highest
age-grade level than in the lowest age-grade level. Such a difference
would also confirm hypothesis 5.
For an over-all view, the scores for all five tasks were trans-
formed into standard scores and analyzed in a 3x3x5 analysis of variance
(age x modeling situation x task). Also, a procedural statistical control
in the form of a Kendall coefficient of concordance was run on the five
task scores for each subject in each treatment condition at each age-
grade level to determine if task performance was affected by the amount
of time elapsed between introduction to the tasks and the opportunity to
perform the tasks.
The .05 level of significance was employed in all of the analyses.
A Kendall coefficient of concordance was run on each of the nine
age-treatment group conditions to determine if task performance was re-
lated to the amount of time between introduction to the tasks and the
opportunity to perform. That is, with a mean subject-running time of
approximately 10 minutes, over an hour elapsed between the first and
last subject's chance to perform. The Kendall coefficient of concordance
(W) was significant on only one of the nine groups, the seventh grade
control group. However, even this group did not show a linear increase
or decrease in scores as a function of time. These findings indicate
that when the time period between the modeling stimuli and the opportunity
to perform is short--l to 2 hours--performance is not strongly influenced
by temporal closeness or distance between modeling stimuli and opportunity
The main results of the experiment are seen in the six analyses of
variance summaries (Tables 3 to 8) and the four summaries of Newman-Keuls
comparisons of cell means (Tables 9,10,11,14). The cell means for the first
five analyses of variance summaries have been plotted (Figures 1 to 5) to
present a better picture of the location of each group on the various
tasks. The scores for tasks 1, 2, and 4 were transformed to achieve homogen-
eity of variance, following significance at the .05 level on the Cochran
test for homogeneity of variance, by the function: x' = log0 (x + 1)
None of the five hypotheses were confirmed in the form originally
stated. However, some support was found for hypothesis 1, and much
support was found for hypotheses 4 and 5.
Hypothesis 1 predicted imitation, defined as a significant differ-
ence between El and/or E2 and C, would occur at all three age levels and
on all five tasks, though not necessarily at each age level on each task.
The above stated differences were found to be significant at only one age
level, first graders, and on only one task, task 2. (See Tables 9 and 10.)
Here both El and E2 differed significantly from C. However, during the
course of the experiment a behavior was noted on task 1 which obviously
differentiated the experimental groups from the control groups. This
behavior was the number of times the finger oscillation device was used,
as opposed to the number of taps which was scored on task 1. That is, the
experimental video tapes showed the models either using the device multiple
times or not at all. The control tape showed the experimenter explaining
the device and using it one time for demonstration purposes. (See Appen-
dix A.) These differences can be seen in Table 12 and Figure 6. The
differences indicate the presence of imitation on task 1 across the three
age-grade levels. In addition, when imitation in terms of responses ident-
ical to model responses (e.g., 100 or 0 responses on task 1, a guess of
700 or 300 on task 2, etc.) is considered across age and task, one finds
differences between control and experimental groups. (See Table 13 and
Figure 7). This analysis included as a task the behavior discussed im-
mediately above. Consequently, when viewed in terms of number of responses
identical to responses of models and when viewed across age and task in-
stead of separately by age and task, one can see that in an over-all manner
some imitation did occur. Also, imitation occurred more strongly in E2
than in El.
Hypothesis 2 predicted differences in adult versus peer model poten-
cy, defined as significant differences between El and E2, among the three
age levels on each of the five tasks. From Table 11 it can be seen that
such a difference was found at only one age level, first graders, and on
only one task, task 2.
Hypothesis 3 predicted that the adult model would be more potent
than the peer model at the lowest age level and vice versa for the highest
age level. Potency was defined in terms of the direction of the E1-E2
difference. If the E1-E2 difference was positive, this indicated adult
model potency because the adult model displayed high task behaviors in the
El situation and low task behaviors in the E2 situation, while the peer
model displayed low task behaviors in the El situation and high task be-
haviors in the E2 situation. Conversely, a negative E1-E2 difference
indicated peer model potency. Again from Table 11 it can be seen that,
as noted under hypothesis 2, there was only one significant difference in
model potency. This significant difference, however, was in the direction
opposite that predicted in hypothesis 3. At the first grade level the
direction of the difference in model potency was in the predicted direction
on three of the five tasks but not significant. At the seventh grade level
the difference in model potency was in the predicted direction on only one
of the five tasks. It was only on task 4 that both the first and seventh
age-grade levels had differences in model potency in the predicted direct-
Hypotheses 4 and 5 were not confirmed. However, the data presented
below strongly indicate that these two hypotheses deserve more study.
Hypothesis 4 predicted differences in modeling situation potency on
each of the five tasks. This was defined as differences on a task in the
three age-grade levels among the three treatment groups. Significant
differences were found on four of the five tasks, tasks 2,3,4, and 5, as
can be seen in the three treatment groups under each task in Table 14.
Differences also occurred on task 1, but were significant only at the .10
level. In terms of scoring highest, lowest, and in between the highest
and lowest, Table 15 summarizes the score-ranks of Table 14 for each level
according to treatment group. The summation across treatment groups re-
flects the score placements in the control group. The seventh graders
scored highest most often, the first graders scored lowest most often, and
the fifth graders scored in between most often.
Hypothesis 5 predicted that the highest age-grade level would show
less superstitious-irrelevant behavior than the lowest age-grade level.
This was indeed the case as can be seen from Table 3 and Figure 1. The
seventh graders scored lower in all three treatment groups on task 1 than
did the first graders. However, the seventh and first grade row means in
the task 1 analysis of variance differed less strongly than predicted.
The difference was significant at the .10 level but not at the .05 level.
Finally, from an over-all standpoint, Table 8 represents the
results when the scores for all five tasks were transformed into stand-
ard scores and analyzed in a 3 x 3 x 5 analysis of variance (age x model-
ing situation x task). No significant main effects of age, modeling
situations, or task were found. However, the significant age x task and
age x modeling situation x task interactions indicate that the main effects
were masked. That is, task behavior was dependent on the combination of
age and task as well as the combination of age, modeling situation, and
task presented to the subject, not just on one of the three factors. The
highly significant age x task interaction also emphasizes the importance
of the task in determining the nature of the results at the different age
Consequently, in terms of differences significant at or beyond the
.05 level, the results of this experiment provide evidence that imitation
occurs across the three age-grade levels considered but in a less power-
ful and less age-related manner than hypothesized. The results also in-
dicate that there is practically no significant differential imitation of
adult and peer models as a function of observer age or type of task. Most
of the differences in task behavior found when the tasks were considered
individually were a function of main effects, primarily age, instead of
the predicted age-treatment group interactions. However, the score-ranks
in Tables 14 and 15 indicate that these age differences were task-related
and affected by treatment group. These interactions are demonstrated in
Table 8 where age, modeling situation, and all five tasks were analyzed in
a 3 x 3 x 5 analysis of variance. The high age-task interaction emphasizes
the importance of the task used to study modeling and the need to employ
more than one task in such studies.- This finding confirms the criticism
made earlier concerning the tendency in previous research to use one task
and the lack of valid generalization that results.
Summary of Analysis of Variance for Task 1: Superstitious-
Source SS ds MS
A (age) 2.67 2 1.34 2.53b
B (models) 2.58 2 1.29 2.43b
AB 4.81 4 1.20 2.26b
Within cell 37.37 70 .53
a Scores transformed by the function: x' = 1og x + 1)
Summary of Analysis of Variance for Task 2: Guessing Behavior
Source SS df MS F
A (age) .27 2 .14 NS
B (models) 1.34 2 .67 6.70**
AB .18 4 .04 NS
Within cell 7.41 70 .10
a Scores transformed
; p < .01
by the function: x' x log10(x + 1)
Summary of Analysis of Variance for Task 3: Risk-taking
Source SS df MS F
A (age) 423.85 2 211.92 10.38**'
B (models) 47.04 2 23.52 NS
AB 73.24 4 18.31 NS
Within cells 1428.87 70 20.41
p < .001
Summary of Analysis of Variance for Task 4: Spontaneous
Source SS df MS F
A (age) 4.90 2 2.45 12.25
B (models) .18 2 .09 NS
AB .89 4 .22 NS
Within cells 13.90 70 .20
a Scores transformed
*** p < .001
by the function: x' = log (x + 1)
Summary of Analysis of Variance for Task 5: Drawing Behavior
Source SS df MS F
A (age) 51.41 2 25.70 2.41
B (models) 4.54 2 2.27 NS
AB 113.96 4 28.49 2.67'
Within cells 746.25 70 10.66
p < .10
p < .05
Summary of Analysis of Variance of Task Behavior as a Function
of Age, Modeling Situation, and Taska
Source SS df MS F
A (age) 348.92 2 174.46
situation) 225.16 2 112.58
AB 1737.81 4 434.45 1.48 NS
groups 20535.14 70 293.36
C (task) 51.41 4 12.85 -
AC 5130.38 8 641.30 9.32""
BC 1043.63 8 130.45 l.D IS
A[0 2751.85 16 171.f" 2.
groups 19259.24 280 68.78
Scores transfo-red into standard scores:
SD = t
p < .01
Summary of Newman-Keuls' Tests on C Minus El Differences
for Tasks 1 Through 5
Grade Task 1 Task 2 Task 3 Task 4 Task 5
5 + +
7 + -
p < .05
Summary of Newman-Keuls' Tests on C Minus E2 Differences
for Tasks 1 Through 5
Grade Task 1 Task 2 Task 3 Task 4 Task 5
1 + +
5 + + -
7 + + +
p < .01
Summary of Newman-Keuls' Tests on El Minus E2 M'odel
Differences for Tasks 1 Thr.ouTh 5
Grade Task 1 Task 2 Task 3 Task 4 Task 5
5 + -
7 0 + +
p < .01
Usage Across Age of the Finger Oscillation Device
Times Used C El X2 E2 X2
1 24 14 ,, 10
+1 2 15 14
p < .01
p < .001
Responses Identical to Responses of
Models Across Six Tasks
Responses C El X2 E2 X2
Identical 24 41 40
Dissimilar 132 133 104
Total Responses 156 174 144
Sp < .10
p < .02
Differences in Modeling Situation Potency
Anova cell mean Task 1 Task 2 Task 3 Task 4 Task 5
Highest 5 l7: |7 7 7
Middle 1 55 5** 5
Lowest 7 1 1
Highest 1 5 17* 7 1""
Middle 7 7 5 5
Lowest 5 1 1 5
Highest 15 1 5 7 1
Middle 1 5 7 15 5
Lowest 7 7 1 1 7
p < .05
Scores joined by a common line do not dif
and vice versa.
Differences tested by Newman-Keuls' test.
'fer significantly from each other
Summary of Score-ranks in Treatment Groups Across the
Score-rank of C El E2 Total
Anova Cell mean 1 5 7 1 5 7 1 5 7 1 5 7
Highest 0 1 4 2 1 2 2 2 1 4 4 7
Middle 1 4 0 0 2 3 1 3 1 2 9 4
Lowest 4 0 1 3 2 0 2 0 3 9 2 4
Figure 1. Scores on Task 1 According to Age and Modeling Group.
C El E2
Figure 2. Scores on Task 2 According to Age and Modeling Group.
Figure 3. Scores on Task 3 According to Age and Modeling Group.
-. 1 .30
z 1.20 -
S 1 .10 -
0 1 .00
S .90 1i
S .70 -
C El E2
S Grade 1
D Grade 5
Figure 4. Scores on Task 4 According to Age and Modeling Group.
[ Grade 1
D] Grade 5
E Grade 7
Figure 5. Scores on Task 5 According to Age and Modeling Group.
C El E2
Times Used: 1 +1
Figure 6. Usage Across Age of the Finger Oscillation Device.
Type of Response:
Figure Responses Identical to Responses of Models Across Six Tasks.'
Given the indications of imitation when age levels are combined,
the main question arising from these results is why there is a lack of
imitation per se according to age level and of differential adult and
peer imitation according to age level. The following considerations seem
to contain the most relevant issues.
The MuZti-model Choice Situation
A two-model peer-adult situation of the type employed in this study
is unique in the modeling literature, to the knowledge of the experimenter.
Such a situation presents a more complex stimulus and choice situation to
the subject than the one model situation. As Bandura (1965a) points out,
children observing multiple adult models will not necessarily perform as
smaller replicas of a model but may combine components of both models'
performances in their response to produce a "new" response. In other
words, imitation can occur without a subject's response being just like
the response of one of the models. Therefore, the "new" response given
by the subject in this experiment (a response between the response of the
two models) may have represented a constructive attempt to incorporate
aspects of both models' performances and have been truely imitative, but
not of one model only. The assessment of this is discussed later in rela-
tion to dependent variable measures. The concern of this experiment with
the effects of multiple peer-adult models reflects Bandura's statement
(1965a, p. 338): "...learning principles based on a two-person model
[one model, one subject] may be subject to strict limitations, since the
introduction of additional social variables can produce significant changes
in the functional relationships between relevant variables." The lack of
strong imitation and differential imitation according to age (imitation
of one model exclusively) may be one of the effects of observing two models
of different ages perform in contrasting ways.
Effects Related to Reinforcement.
On the other hand, most of the previous experiments have been con-
cerned with the effects on imitation of variables other than age of model.
alone--reinforcing consequences to the model (Bandura, Ross, and Ross,
1963c), response consequences to the model (Bandura, 1965b), model's eval-
uation of responses or standard setting (Bandura and Kupers, 1964; Mischel
and Liebert, 1966; Bee and Colle, 1967), and model's rewarding power and
control of future resources (Bandura, Ross, and Ross, 1963z; Mischel and
Grusec, 1966). In the present experiment there were no model response
evaluations, consequences, rewarding powers, controls of future resources,
or reinforcements either by the models themselves or by anyone else. With
no reinforcements, evaluations, etc.,of either model presented to the
subject, model age would seem to be the most outstanding difference between
the two models in the present study. The results indicate that the age
of the model may not be a very potent factor apart from reinforcement.
Effects of Perceived Similarity
However, an alternative answer to the influence of reinforcement
was suggested in a recent study by Rosekrans (1967) on a group of 11 to
14 year old boys. She found evidence that perceived similarity of the
subject to the model (in terms of interests, skills, background, and group
membership) was a factor affecting both frequency of imitation and size of
imitative behavior repertoire. Both the frequency of imitation and the
size of the imitative behavior repertoire were greater when the observer
perceived himself as similar to the model than when the observer perceived
himself as dissimilar. Positive and negative response consequences to the
model had only minimal differential effects on imitation. Evidence was
also found which indicated that the learning as well as the performance of
imitative responses was increased by high perceived similarity. The results
of Rosekrans' study indicate that perceived positive similarity may be a
more influential variable than response consequences in producing imitation.
Special Characteristics of Subject Population
In terms of methodological considerations, it would be important to
replicate this study on another population of first, fifth, and seventh
graders and compare the results with the present findings on the P.K. Yonge
Laboratory School population. This is important because the P. K. Yonge
Laboratory School population is a somewhat atypical student population.
As mentioned previously in the description of the school, one-half of the
students are from families of academic faculty at the University of Florida.
The student population could be described over-all as relatively high in
intelligence and socio-economic status. The students are given individual
attention and encouraged in individual pursuits. Also, the students are
continually participating in experiments and are sophisticated concerning
experimental procedures. In other words, the P.K. Yonge Laboratory School
population could be described as having a nonimitative bias. However, the
fact that even on such a population imitation was found, albeit not as
powerfully as predicted, suggests both that imitation is a factor involved
in determining what behavior a child displays, and that much stronger
imitation might occur in children without the special characteristics of
the children in this population. The effects of intellectual level and
socio-economic status are still to be determined. A study now in progress
(O'Connor, 1967), investigating the relationships of imitation to intel-
lectual level, socio-economic status, and achievement in white and Negro
first grade children, should make the importance of thesevariables clearer.
The Task in Imitation Research
In terms of dependent variables, this experiment shows that the kind
of task one uses has an influence on the results one finds. The difference
between the oldest and youngest subjects on the superstitious-irrelevant
task offers evidence that imitation becomes less obvious as age increases
as a function of increasing ability to discriminate between behaviors rel-
evant and irrelevant to the situation or task. That is, imitation becomes
more focused and "reasoned." This in turn leads to a decreasing need to
respond in exactly the same manner as the model. It follows, therefore,
that the more unfamiliar the task,the more difficult it is to make the
discrimination between the relevant and irrelevant dimensions. Consequently,
it is more necessary to employ such unfamiliar tasks as age increases in
order to reveal the imitation that may be occurring. The tasks employed
in this experiment probably erred on the side of being too familiar and/or
unstimulating--i.e., subjects had previously established response tenden-
cies on the tasks. It could be said that the primary imitation data in the
experiment were found on the task which, in the opinion of the experimenter,
was the most novel and stimulating. This was task 1, the superstitious-
irrelevant task. From the observations of subject behavior, this opinion
seemed to be confirmed. As mentioned above, it is hypothesized that the
more familiar the task the more likely one is to know the relevant and
irrelevant dimensions and feel comfortable in relying only on oneself in
handling the task. This is the case because of having had past experience
with the task and having the consequent response tendencies or repertoire
associated with the task. If this is indeed true, task novelty is a
critical variable in eliciting imitative behavior and assuming, as Bandura
(1965a) points out when speaking of response facilitation effect, that the
behavior in question is not likely to have been previously punished or to
possess some kind of reinforcement history.
Measurement and the Dependent Variables
Concerning the measurement of the dependent variables, it appears
that a reduction in the possible ways of responding on a task would result
in a more precise measure of imitation. In the present study an attempt
was made to provide many alternative responses to a task and to allow imita-
tion to vary in intensity. As a consequence, response continue were provided
for each task with the models' responses representing both ends of the con-
tinuum. This procedure, while seeming to provide a task situation more
comparable with daily encountered task situations, made it more difficult
to know whether or not an individual response was imitative. By employing
task situations with more discrete but previously shown equally attractive
response alternatives, one could achieve a clearer response classification.
For example, one could allow only four discrete responses to a task--
the adult model response, the peer model response, a response incorporating
some aspects of both the adult and peer models' responses, and a response
displaying none of the aspects of the above three responses. This situation
would make a non-imitative attempt by the subject clearer. This situation
would also prevent extremely deviant scores, which were possible and found
on the open ended tasks, tasks 1,2, and 4, where it was possible to respond
higher than the model displaying the high task behavior. The resulting
deviant scores on these tasks necessitated the score transformations as
previously indicated on tasks 1,2, and 4.
Foci for Future Research
It would be important to know whether the finding that filmed models
can be as effective as live models, established with nursery school child-
ren (Bandura, Ross, and Ross, 1963b), holds across age. This could be
determined by using live and video-tape models at several age levels. The
characteristics of the observation and performance situations should also
be investigated. In the present study the subject observed the modeling
stimuli in a group of peers. This could be compared with situations in
which the subject observes the modeling stimuli alone or with adults. In
the present study the subject performed in the presence of an adult, the
experimenter. This situation could be compared with one in which the sub-
ject performed alone, in the presence of a peer, or in the presence of a
qutin of se differences which was not addressed in th-is i research, a
question of sex differences which was not addressed in this research, and
the effects resulting from the various combinations of this variable
with the variables previously discussed.
From the modeling literature and the present study, the following
factors and directions seem to be the most salient for future research in
1. Comparison of multiple versus single model situations;
2. Discrete response categories on tasks in multiple model
3. Response consequences versus no response consequences to models
in multiple model situations;
4. Task novelty;
5. Effects of perceived similarity in multiple model situations;
6. Filmed versus live models on older children;
7. Manipulations of the performance situation such as individual
versus group observation of stimuli, observation group
8. Manipulations of sex differences in models and observers.
This investigation attempted to assess the effects of multiple
observer ages and types of tasks on the imitation of adult and peer
models when both adult and peer models are observed by the subject.
First, fifth, and seventh grade males were randomly divided in each
grade into two experimental groups and one control group. Seventy-nine
subjects were used. The three groups received different modeling
stimuli as a group via video-tape. The control group video-tape con-
sisted of the experimenter presenting the materials and instructions for
the five tasks. The video-tape for the first experimental group consisted
of the same presentation of the task materials and instructions as con-
tained in the control group tape with the addition of a peer and an adult
doing each of the tasks in contrasting ways. The video-tape for the
second experimental group was exactly like the first experimental group
tape, except that the model roles were reversed. The five tasks involved
the following behaviors: tapping a finger oscillation device, guessing
the number of beans in a jar, throwing darts, talking spontaneously, and
drawing. Following the observation of the tape, the subjects were indi-
vidually given the opportunity to perform each of the tasks to see if
imitation per se occurred and whether the peer or adult was imitated.
It was found that with a time period of one to two hours between
the modeling stimuli and the opportunity to perform, performance was not
substantially influenced by temporal closeness or distance from the obser-
vation of the modeling stimuli. Imitation was found to occur across the
three age-grade levels considered but in a less powerful and less age-
related manner than hypothesized. The results also indicated that there
was practically no significant differential imitation of adult and peer
models as a function of observer age or type of task. Most of the differ-
ences in task behavior found when the tasks were considered individually
were a function of main effects, primarily age, instead of the predicted
age-treatment group interactions. When all five tasks were considered
together, significant interactions between age and task and among age,
modeling situation, and task were found. Thus, task behavior was de-
pendent on the combination of age and task as well as on the combination
of age, modeling situation, and task presented to the subject, not just
on one of the three factors. The high age-task interaction emphasized the
importance of the task used in modeling research and the need to employ
more than one task to get a good view of the nature of modeling at various
The results were discussed in terms of the uniqueness of the two-
model peer-adult situation in this study, the assessment of imitation in
a two-versus a one-model situation, lack of response consequences to the
models, the importance of subject variables, importance of task character-
istics, and discreteness of response. Various experimental manipulations
remaining to be studied were noted as well as factors and directions salient
for future research.
Approximate Transcript of Experimental
(Camera opens on experimenter, adult model, and peer model)
"Hello. What is your name?"
"My name is John Clemens." (Name assigned to all peer models.)
"How old are you, John?'
"I'm (appropriate number) years old."
"What grade are you in?"
"I'm in (appropriate)grade."
"And what is your name?"
"My name is Mr. Costanzo."
"How old are you, Mr. Costanzo?"
"I'm 25 years old."
"All right. Let's begin." (Camera on experimenter only.)
First of all, I want to show you this finger oscillation
device. All it is, is this little lever which you can push
and a counter that counts the number of times you tap the
lever. Like this. (Experimenter taps 5 times and lays
the device down.) Some people think tapping brings good
* The control group tape is the experimental tape with all model-porticns
deleted. In the El situation Model 1 represents the adult model, and
Model 2 represents the peer model. In the E2 situation Model 1 represents
the peer model, and Model 2 represents the adult model.
luck and helps them perform better on the tasks; so I'm
going to give you the chance to tap now, if you want to,
as many times as you want to. Also, you tap at any
time while you are doing the other tasks. Do you w.ant
(Camera on Model 1.) "I think I'll tap. (Taps 25 tires
immediately and 25 times between each task for a total
of 100 taps.)
(Camera on Model 2.) "I think I won't tap." ("ver t .)
(Camera on experimenter and container only.) '"kay. 'ext
I have this jar filled with beans. I know the exact
number of beans in the jar, but you do not. So I want
you to try to guess the best that you can how many beans
there are in the jar. How many beans do -~ think are in
(Camera on Model 2 and jar.) "I'11 guess 300 beans."
(Camera on Model 1 and the jar.) "I'll guess 700 beans."
(Camera on Model 1 as he goes to tp-ing device and t
(Camera on experimenter and dartboard.) '~"Oy. o here
you see a dartboard and darts. You will notice ha he
closer to the center of the board your darts hit,te hi '.r
your score is. The rings score, going from the outside to
the center of the board, 10, 25,50,75, and 100 points.
However, you can also score points according to how far
away from the board you stand. On the floor (Camera
switches to marks on floor) there are 16 marks numbered
zero to 15, each one-half foot apart. (Camera on exper-
imenter and marks.) The closest to the board you may
stand is here at the first mark, and the farthest away
you may stand is here at number 15. You get added to
the score of whatever your dart hits on the board, the
number of the line you stand at to throw. For example,
if you stand at the first line, line 0, and throw a dart
that hits the 25 point circle, you get 25 points. If
you stand at line 15 and throw a dart that hits the 25
point circle, you get 40 points--25 for hitting the 25
point circle plus 15 points for standing at line 15. You
should remember, however, that the closer to the board
you stand the easier it is to hit the high point circles,
but the farther away you stand the more points you get
added to the score of what your dart hits on the board.
You get 5 darts to throw. You must throw all 5 darts
from the same line, whichever line you choose. It is up
to you to choose the spot that you think you can score
the most points from. Where do you want to throw from?"
Model 1: (Camera on Model 1 showing what line he is standing at
and stays on him while he throws. The dartboard and
where his darts hit are not shown.)
"I think you can score the most points by throwing from
Model 1 :
line 15." (Throws from line 15.)
(Camera on Model 2 shower what line he is standing on
and stays on him while he throws. The dartboard and
where his darts hit are not shown.)
"I think you can score the most points by throwing from
line 0." (Throws from line 0.)
(Camera switches to Model 1 and tapping device where Model 1
taps 25 times.)
(Camera on experimenter and tape recorder.) .. t I ant
to get a measure of the tone of your voice by having you
talk into the microphone of this tape recorder. I want
you to say as much or as little as you want to say. What
do you want to say?"
(Camera on Model 2.) "Hello."
(Camera on Model 1.) "Mly name is (appropria t) and I live
here in Gainesville, Floriua. I had a nice Christmas and
New Year vacation. I watched football on TV and
went to visit some of relatives. I have hobbies, like
to go fishing, read, p hbal i to the beach, swim, and
I collect different things and 1ie to go places. i ,;sh
I had more time to do som of t things I like to do."
(Carera on Mode as he goes to apping device and t is
25 times.) I 've t, ed 1- ti s in all !"
(Camera on experimenter and table on which is paper and
two sets of 15 different colored pencils.) 'All right.
Here is some paper and 15 different colored pencils.
I want you to draw the best house that you can draw.
You may use as many or as few of the colors as you wish.
You have 5 minutes to work on your drawing. You may go
ahead now and begin to draw.
(Camera on Model I and Model 2 sitting side by side at
"I think I'll use all 15 colors in my drawing."
"I think I'll use just I color in my drawing."
(Camera stays on both models for 2 minutes during which
Model 1 changes colors frequently and Model 2 continues
with the same color. At no time are either the drawings
or the kinds of colors visible.)
(Experimenter walks into picture behind the models after
2 minutes, following a brief fade-out and fade-in.) "Five
minutes is up. Stop drawing. That's all the tasks there
are. Thank you very much."
(End of tape.)
Hello. I'm Mr. Musselman from the University of Florida. I have
asked you to be here today to help me find the answer to the question I
am studying. How well can boys and men of different ages do on the tasks
I am going to show you? I have been giving these tasks to a few boys
from other schools and to a few men, but I want to give them to many boys
of the same age so I can tell how well boys of a certain age and grade
can do on them. Because all of the boys in your grade will do the tasks,
please do not tell them anything about the tasks or instructions so that
the difficulty of the tasks is the same for them as it is for you.
Watch and listen to the instructions of how to do the tasks. I
want you to do the best job you can possibly do on the tasks. Do each
task in the way you think is the best possible way to do it.
This is what you are going to do. I will give you the instructions
for the tasks and show you what the tasks are. Then you will go back to
your classroom. You will be called one at a time to come and do the tasks.
When you finish you will go back to your classroom and another one of you
will come in and do the tasks. Let me again emphasize that you not talk
about the tasks with each other or anyone else. I want you to do the tasks
the way you think you can do them the best.
Okay. Now I want to show you what the tasks are and give you the
instructions. Before we begin do you have any questions?
Control Group: I made a film of the instructions and tasks which I will
show to you now.
Experimental Groups: I thought it would be a good idea to show you some
people doing the tasks in addition to giving the instructions and showing
you the equipment so that everyone would know how to do the tasks. So a
few days ago while I was giving the tasks to some people,! asked if I could
film two of them doing the tasks. They agreed and I filmed them. I will
introduce them to you and then begin.
Individual Performance Procedure for Each
(Subject enters room.) "If you have any questions now or
anytime, please ask them."
"Do you want to tap?"
"How many beans do you think are in the jar?"
"Where do you want to throw from?"
"What do you want to say?"
"I want you to draw the best house that you can draw. You
have 5 minutes to work on your drawing. You may go ahead
now and begin to draw."
(After five minutes.) "Five minutes is up. That's all.
Thank you very much."
Summary of Row Scores
Aronfreed, J. The origin of self-criticism. Psychological Review, 1964,
Baer, D., and Sherman, J. Reinforcement control of generalized imitation
in young children. Journal of Experimental Child Psychology,
1964, 1, 37-49.
Bandura, A. Behavioral modification through modeling procedures. In L.
Krasner and L. Ullmann (Eds.), Research in behavior modification.
New York: Hold, Rinehart and Winston, 1965a.
Bandura, A. Influence of model's reinforcement contingencies of the
acquisition of imitative responses. Journal of Personality and
Social Psychology, 1965b, 1, 589-595.
Bandura, A. Social learning through imitation. In M. Jones (Ed.),
Nebraska symposium on motivation. Lincoln, Nebraska: University
of Nebraska Press, 1962.
Bandura, A. Vicarious processes: A case of no-trial learning. In L.
Berkowitz (ed.), Advances in experimental social psychology.
Vol. II. New York: Academic Press, 1965c.
Bandura, A. and Huston, Aletha. Identification as a process of incidental
learning. Journal of Abnormal and Social Psychology, 1961, 63,
Bandura, A., and Kupers, Carol. Transmission of patterns of self-
reinforcement through modeling. Journal of Abnormal and Social
Psychology, 1964, 69, 1-9.
Bandura, A., and McDonald, F. Influence of social reinforcement and the
behavior of models in shaping children's moral judgement. Journal
of Abnormal and Social Psychology, 1963, 67, 274-281.
Bandura, A., Ross, Dorothea, and Ross, Sheila. A comparative test of the
status envy, social power, and secondary reinforcement theories of
identificatory learning. Journal of Abnormal and Social Psychology,
1963a, 67, 527-534.
Bandura, A., Ross, Dorothea, and Ross Sheila. Imitation of film-mediated
aggressive models. Journal of Abnormal and Social Psychology,
1963b, 66, 3-11.
Bandura, A., Ross, Dorothea, and Ross, Sheila. Transmission of aggression
through imitation of aggressive models. Journal of Abnormal and
Social Psychology, 1961, 63, 575-582.
Bandura, A., Ross, Dorothea, and Ross, Sheila. Vicarious reinforcement
and imitative learning.' Journal of Abnormal and Social Psychology,
1963c, 67, 601-607.
Bandura, A., and Walters, R. Social learning and personality development.
New York: Holt, Reinhart and Winston, 1963.
Bee, Helen, and Colle, H. The origins of standards of excellence:
Modeling vs. direct reinforcement. Paper presented at the Society
for Research in Child Development meeting, New York, 1967.
Bijou, S., and Baer, D. Child Development. Vol. II. Universal stage of
Infancy. New York: Appleton-Century-Crofts, 1965.
Bronfenbrenner, U. Freudian theories of identification and their
derivatives. Child Development, 1960, 31, 15-40.
Brown, D. Sex-role preferences in young children. Psychological Mono-
graphs, 1956, 70, No. 14 (Whole No. 421).
DeRath, G. The effects of verbal instructions on imitative aggression.
Michigan State University, 1963, Unpublished Doctoral Dissertation.
Epstein, S. Comments on Dr. Bandura's paper. In M. Jones (Ed.), Nebraska
symposium on motivation. Lincoln, Nebraska: University of Nebraska
Gelfand, Donna. The influence of self-esteem on rate of verbal conditioning
and social matching behavior. Journal of Abnormal and Social Psy-
chology, 1962, 65, 259-265.
Hartup, W. Patterns of imitative behavior in young children. Child De-
velopment, 1964, 35, 183-191.
Hartup, W., and Zook, E. Sex-role preferences in three and four year old
children. Journal of Consulting Psychology, 1960, 24, 420-426.
Henker, Barbara Anne. The effect of adult model relationships on children's
play and task imitation. Ohio State University, 1963, Unpublished
Hicks, D. Imitation and retention of film-mediated aggressive peer and
adult models. Journals of Personality and Social Psychology, 1965,
Kagan, J. The concept of identification. Psychological Review, 1958, 65,
Lindquist, E. Design and anaTc'is of experimr-nts in pal .7' and
education. Boston: Houghton Mifflin Company, 1953.
Lynn, D. A note on sex differences in the development of masculine and
feminine identification. Psychological Review, 1959, 66, 126-135.
Maier, H. Three theories of child development. New York: Harper and
Martin, W. Learning theory and identification: IlI. The development of
values in children. Journal of Genetic Psychology, 1954, 84,
McDavid, J. Imitative behavior in preschool children. Psychological
Monographs, 1959, 73, No. 16 (Whole No. 486).
Metz, J. Conditioning generalized imitative in autistic children. Journal
of Experimental Child Psychol7..:- 1965, 2, 389-399.
Miller, N., and Dollard, J. Social learning and imitation. New Haven:
Yale University Press, 1941.
Mischel, W., and Grusec, Joan. Determinants of the rehearsal and trans-
mission of neutral and aversive behaviors. Journal of P rsonalit,
and Social Psychology, 1966, 3, 197-205.
Mischel, W., and Liebert, R. Effects of discrepancies between observed and
imposed reward criteria on their acquisition and transmission.
Journal of Personality and Social Psychology, 1966, 3, 45-53.
Mussen, P., and Parker, Ann. Mother nurturance and girls' incidental
imitative learning. Journal of Personality and Social Ps;c'o. ',
1965, 2, 94-97.
O'Connor, A. h." relationship of vicarious lear arr intiZncc
academic achievement among ,ro and white fi*r p.'ps in
integrated classes. In process, University of Florida, 1967.
Risley, T. The establishment of verbal behavior in deviant children.
Unpublished doctoral dissertation, University of Washington, 1966.
Rosekrans, Mary. Imitation in children as a function of perceived similar-
ity to a social model and vicarious reinforcement. Paper presented
at the Society for Research in Child Development meeting, New York,
Siegel, S. Nonparametric statistics for the behavioral sciences. New York:
Skinner, B. Science and human behavior. New York: Macmillan, 1953.
Skinner, B. Verbal behavior. New York: Appleton-Century-Crofts, 1957.
Winer, B. Statistical principles in experimental design. New York:
McGraw Hill, 1962.
Gerald Clemmer Musselman was born July 15, 1940, in Souderton,
Pennsylvania. In June, 1958, he was graduated from Christopher Dock
High School, Lansdale, Pennsylvania. In June, 1963, he received the
degree of Bachelor of Science from Ursinus College, Collegeville, Penn-
sylvania. In 1963, he enrolled in the Graduate School of the University
of Florida. He worked as a graduate assistant for the Departments of
Psychology and Psychiatry. In August, 1965, he received the degree of
Master of Arts. From September, 1965, to the present he has been pur-
suing work toward the degree of Doctor of Philosophy. In 1964 he was
awarded a United States Public Health Service Fellowship which he held
for two years. In 1966 he was awarded a Vocational Rehabilitation Ad-
ministration Traineeship, which he still holds. At the present time he
is completing his year of clinical internship at the J. Hillis Miller
Health Center, University of Florida.
Gerald Clemmer Musselman is married to the former Julie Groff
Landis. He is a member of Pi Gamma Mu and Psi Chi.
This dissertation was prepared under the direction of the
chairman of the candidate's supervisory committee and has been
approved by all members of that committee. It was submitted to the
Dean of the College of Arts and Sciences and to the Graduate Council,
and was approved as partial fulfillment of the requirements for the
degree of Doctor of Philosophy.
Dean, College of Arts and Sciences
Dean, Graduate School