Journal oi Personality and Social Psychology
1965, Vol. 1, No. 6, 589-595
INFLUENCE OF MODELS' REINFORCEMENT CONTINGENCIES
ON THE ACQUISITION OF IMITATIVE RESPONSES '
ALBKRT BANDURA 2
Stanford University
In order to test the hypothesis that reinforcements administered to a model
influence the performance but not the acquisition of matching responses, groups
of children observed an aggressive film-mediated model either rewarded,
punished, or left without consequences. A postexposure test revealed that
response consequences to the model had produced differential amounts of
imitative behavior. Children in the model-punished condition performed sig-
nificantly fewer matching responses than children in both the model-rewarded
and the no-consequences groups. Children in all 3 treatment conditions were
then offered attractive reinforcers contingent on their reproducing the model's
aggressive responses. The introduction of positive incentives completely wiped
out the previously observed performance differences, revealing an equivalent
amount of learning among children in the model-rewarded, model-punished,
and the no-consequences conditions.
It is widely assumed that the occurrence of sponse and simultaneously rewards the ob-
imitative or observational learning is con- server. If the modeled responses are thus
tingent on the administration of reinforcing paired repeatedly with positive reinforcement
stimuli either to the model or to the observer. they gradually acquire secondary reward
According to the theory propounded by Miller value. The observer can then administer posi-
and Bollard (1941), for example, the neces- tively conditioned reinforcers to himself sim-
sary conditions for learning through imitation ply by reproducing as closely as possible the
include a motivated subject who is positively model's positively valenced behavior. In the
reinforced for matching the rewarded behavior second, or empathetic form of imitative learn-
of a model during a series of initially random, ing, the model not only exhibits the responses
trial-and-error responses. Since this concep- but also experiences the reinforcing conse-
tualization of observational learning requires quences. It is assumed that the observer, in
the subject to perform the imitative response turn, experiences empathetically both the re-
before he can learn it, this theory evidently sponse-correlated stimuli and the response
accounts more adequately for the emission of consequences of the model's behavior. As a re-
previously learned matching responses, than sult of this higher-order vicarious condition-
for their acquisition. ing, the observer will be inclined to reproduce
Mowrer's (1960) proprioceptive feedback the matching responses.
theory similarly highlights the role of rein- There is some recent evidence that imita-
forcement but, unlike Miller and Dollard who tive behavior can be enhanced by noncontin-
reduce imitation to a special case of instru- gent social reinforcement from a model (Ban-
mental learning, Mowrer focuses on the clas- dura & Huston, 1961), by response-contingent
sical conditioning of positive and negative reinforcers administered to the model (Ban-
emotions to matching response-correlated dura, Ross, & Ross, 1963b; Walters, Leat, &
stimuli. Mowrer distinguishes two forms of Mezei, 1963), and by increasing the reinforc-
imitative learning in terms of whether the ob- ing value of matching responses per se through
server is reinforced directly or vicariously. In direct reinforcement of the participant ob-
the former case, the model performs a re- server (Baer & Sherman, 1964). Nevertheless,
1 reinforcement theories of imitation fail to ex-
This investigation was supported by Research plain the learning of matching responses when
Grant M-5162 from the National Institutes of Health,
United States Public Health Service. the observer does not perform the model's re-
2
The author is indebted to Carole Revellc who sponses during the process of acquisition, and
assisted in collecting the data. for which reinforcers are not delivered either
589
, 590 ALBERT BANDURA
to (he model or to the observers (Bandura et quences to the model would result in signifi-
al., .1961, 1%3a). cant differences in the performance of imita-
The acquisition of imitative responses under tive behavior with the model-rewarded group
the latter conditions appears to be accounted displaying the highest number of different
for more adequately by a contiguity theory classes of matching responses, followed by
of observational learning. According to the the no-consequences and the model-punished
latter conceptualization (Bandura, in press; groups, respectively. In accordance with previ-
Sheffield, 1961), when an observer wit- ous findings (Bandura et al., 1961, 1963a) it
nesses a model exhibit a sequence of responses was also expected that boys would perform
the observer acquires, through contiguous as- significantly more imitative aggression than
sociation of sensory events, perceptual and girls. It was predicted, however, that the in-
symbolic responses possessing cue properties troduction of positive incentives would wipe
that are capable of eliciting, at some time out both reinforcement-produced and sex-
after a demonstration, overt responses corre- linked performance differences, revealing an
sponding to those that had been modeled. equivalent amount of learning among children
Some suggestive evidence that the acquisi- in the three treatment conditions.
tion of matching responses may take place
through contiguity, whereas reinforcements METHOD
administered to a model exert their major in- Subjects
fluence on the performance of imitatively The subjects were 33 boys and 33 girls enrolled in
learned responses, is provided in a study in the Stanford University Nursery School. They ranged
in age from 42 to 71 months, with a mean age of
which models were rewarded or punished for SI months. The children were assigned randomly to
exhibiting aggressive behavior (Bandura et one of three treatment conditions of 11 boys and 11
al., 1963b). Although children who had ob- girls each.
served aggressive responses rewarded subse- Two adult males served in the role of models, and
one female experimenter conducted the study for all
quently reproduced the model's behavior while 66 children.
children in the model-punished condition
failed to do so, a number of the subjects in Exposure Procedure,
the latter group described in postexperimental The children were brought individually to a semi-
interviews the model's repertoire of aggressive darkened room. The experimenter informed the child
responses with considerable accuracy. Evi- that she had some business lo attend to before they
dently, ihey had learned the cognitive equiva- could proceed to the "surprise playroom," but that
during the waiting period the child might walch a
lents of the model's responses but they were televised program. After the child was scaled, the
not translated into their motoric forms. These experimenter walked over lo the television console,
findings highlighted both the importance of ostensibly tuned in a program and then departed. A
distinguishing between learning and perform- film of approximately S minutes duration depicting
the modeled responses was shown on a glass lenscrcen
ance and the need for a systematic study of in the television console by means of a rear projec-
whether reinforcement is primarily a learning- tion arrangement, screened from the child's view by
related or a performance-related variable. large panels. The televised form of presentation was
1 n the present experiment children observed utilized primarily because attending responses to
televised stimuli arc strongly conditioned in children
a film-mediated model who exhibited novel and this procedure would therefore serve to enhance
physical and verbal aggressive responses. In observation which is a necessary condition for the
one treatment condition the model was se- occurrence of imitative learning.
verely punished; in a second, the model was The film began with a scene in which the model
generously rewarded; while the third condi- walked up to an adult-size plastic Bobo doll and
ordered him to clear the way. After glaring for a
tion presented no response consequences to the moment at the noncompliant antagonist the model
model. Following a postexposure test of imi- exhibited four novel aggressive responses each ac-
tative behavior, children in all three groups companied by a distinctive verbalization.
were offered attractive incentives contingent First, the model laid the Bobo doll on its side, sat
on it, and punched it in the nose while remarking,
on their reproducing the models' responses so "Pow, right in the nose, boom, boom." The model
as to provide a more accurate index of learn- then raised the doll and pommeled it on the head
ing. It was predicted that reinforcing conse- with a mallet. Each response was accompanied by
1965, Vol. 1, No. 6, 589-595
INFLUENCE OF MODELS' REINFORCEMENT CONTINGENCIES
ON THE ACQUISITION OF IMITATIVE RESPONSES '
ALBKRT BANDURA 2
Stanford University
In order to test the hypothesis that reinforcements administered to a model
influence the performance but not the acquisition of matching responses, groups
of children observed an aggressive film-mediated model either rewarded,
punished, or left without consequences. A postexposure test revealed that
response consequences to the model had produced differential amounts of
imitative behavior. Children in the model-punished condition performed sig-
nificantly fewer matching responses than children in both the model-rewarded
and the no-consequences groups. Children in all 3 treatment conditions were
then offered attractive reinforcers contingent on their reproducing the model's
aggressive responses. The introduction of positive incentives completely wiped
out the previously observed performance differences, revealing an equivalent
amount of learning among children in the model-rewarded, model-punished,
and the no-consequences conditions.
It is widely assumed that the occurrence of sponse and simultaneously rewards the ob-
imitative or observational learning is con- server. If the modeled responses are thus
tingent on the administration of reinforcing paired repeatedly with positive reinforcement
stimuli either to the model or to the observer. they gradually acquire secondary reward
According to the theory propounded by Miller value. The observer can then administer posi-
and Bollard (1941), for example, the neces- tively conditioned reinforcers to himself sim-
sary conditions for learning through imitation ply by reproducing as closely as possible the
include a motivated subject who is positively model's positively valenced behavior. In the
reinforced for matching the rewarded behavior second, or empathetic form of imitative learn-
of a model during a series of initially random, ing, the model not only exhibits the responses
trial-and-error responses. Since this concep- but also experiences the reinforcing conse-
tualization of observational learning requires quences. It is assumed that the observer, in
the subject to perform the imitative response turn, experiences empathetically both the re-
before he can learn it, this theory evidently sponse-correlated stimuli and the response
accounts more adequately for the emission of consequences of the model's behavior. As a re-
previously learned matching responses, than sult of this higher-order vicarious condition-
for their acquisition. ing, the observer will be inclined to reproduce
Mowrer's (1960) proprioceptive feedback the matching responses.
theory similarly highlights the role of rein- There is some recent evidence that imita-
forcement but, unlike Miller and Dollard who tive behavior can be enhanced by noncontin-
reduce imitation to a special case of instru- gent social reinforcement from a model (Ban-
mental learning, Mowrer focuses on the clas- dura & Huston, 1961), by response-contingent
sical conditioning of positive and negative reinforcers administered to the model (Ban-
emotions to matching response-correlated dura, Ross, & Ross, 1963b; Walters, Leat, &
stimuli. Mowrer distinguishes two forms of Mezei, 1963), and by increasing the reinforc-
imitative learning in terms of whether the ob- ing value of matching responses per se through
server is reinforced directly or vicariously. In direct reinforcement of the participant ob-
the former case, the model performs a re- server (Baer & Sherman, 1964). Nevertheless,
1 reinforcement theories of imitation fail to ex-
This investigation was supported by Research plain the learning of matching responses when
Grant M-5162 from the National Institutes of Health,
United States Public Health Service. the observer does not perform the model's re-
2
The author is indebted to Carole Revellc who sponses during the process of acquisition, and
assisted in collecting the data. for which reinforcers are not delivered either
589
, 590 ALBERT BANDURA
to (he model or to the observers (Bandura et quences to the model would result in signifi-
al., .1961, 1%3a). cant differences in the performance of imita-
The acquisition of imitative responses under tive behavior with the model-rewarded group
the latter conditions appears to be accounted displaying the highest number of different
for more adequately by a contiguity theory classes of matching responses, followed by
of observational learning. According to the the no-consequences and the model-punished
latter conceptualization (Bandura, in press; groups, respectively. In accordance with previ-
Sheffield, 1961), when an observer wit- ous findings (Bandura et al., 1961, 1963a) it
nesses a model exhibit a sequence of responses was also expected that boys would perform
the observer acquires, through contiguous as- significantly more imitative aggression than
sociation of sensory events, perceptual and girls. It was predicted, however, that the in-
symbolic responses possessing cue properties troduction of positive incentives would wipe
that are capable of eliciting, at some time out both reinforcement-produced and sex-
after a demonstration, overt responses corre- linked performance differences, revealing an
sponding to those that had been modeled. equivalent amount of learning among children
Some suggestive evidence that the acquisi- in the three treatment conditions.
tion of matching responses may take place
through contiguity, whereas reinforcements METHOD
administered to a model exert their major in- Subjects
fluence on the performance of imitatively The subjects were 33 boys and 33 girls enrolled in
learned responses, is provided in a study in the Stanford University Nursery School. They ranged
in age from 42 to 71 months, with a mean age of
which models were rewarded or punished for SI months. The children were assigned randomly to
exhibiting aggressive behavior (Bandura et one of three treatment conditions of 11 boys and 11
al., 1963b). Although children who had ob- girls each.
served aggressive responses rewarded subse- Two adult males served in the role of models, and
one female experimenter conducted the study for all
quently reproduced the model's behavior while 66 children.
children in the model-punished condition
failed to do so, a number of the subjects in Exposure Procedure,
the latter group described in postexperimental The children were brought individually to a semi-
interviews the model's repertoire of aggressive darkened room. The experimenter informed the child
responses with considerable accuracy. Evi- that she had some business lo attend to before they
dently, ihey had learned the cognitive equiva- could proceed to the "surprise playroom," but that
during the waiting period the child might walch a
lents of the model's responses but they were televised program. After the child was scaled, the
not translated into their motoric forms. These experimenter walked over lo the television console,
findings highlighted both the importance of ostensibly tuned in a program and then departed. A
distinguishing between learning and perform- film of approximately S minutes duration depicting
the modeled responses was shown on a glass lenscrcen
ance and the need for a systematic study of in the television console by means of a rear projec-
whether reinforcement is primarily a learning- tion arrangement, screened from the child's view by
related or a performance-related variable. large panels. The televised form of presentation was
1 n the present experiment children observed utilized primarily because attending responses to
televised stimuli arc strongly conditioned in children
a film-mediated model who exhibited novel and this procedure would therefore serve to enhance
physical and verbal aggressive responses. In observation which is a necessary condition for the
one treatment condition the model was se- occurrence of imitative learning.
verely punished; in a second, the model was The film began with a scene in which the model
generously rewarded; while the third condi- walked up to an adult-size plastic Bobo doll and
ordered him to clear the way. After glaring for a
tion presented no response consequences to the moment at the noncompliant antagonist the model
model. Following a postexposure test of imi- exhibited four novel aggressive responses each ac-
tative behavior, children in all three groups companied by a distinctive verbalization.
were offered attractive incentives contingent First, the model laid the Bobo doll on its side, sat
on it, and punched it in the nose while remarking,
on their reproducing the models' responses so "Pow, right in the nose, boom, boom." The model
as to provide a more accurate index of learn- then raised the doll and pommeled it on the head
ing. It was predicted that reinforcing conse- with a mallet. Each response was accompanied by
