Journal of Abnormal and Social Psychology1963, Vol. 66, No. 1, 3-11

IMITATION OF FILM-MEDIATED AGGRESSIVE MODELS r

ALBERT BANDURA, DOROTHEA ROSS,a AND SHEILA A. ROSS

Stanford University

In a test of the hypothesis that exposure of children to film-mediated aggressivemodels would increase the probability of Ss' aggression to subsequent frustra-tion, 1 group of experimental Ss observed real-life aggressive models, a 2ndobserved these same models portraying aggression on film, while a 3rd groupviewed a film depicting an aggressive cartoon character. Following the exposuretreatment, Ss were mildly frustrated and tested for the amount of imitativeand nonimitative aggression in a different experimental setting. The overallresults provide evidence for both the facilitating and the modeling influenceof film-mediated aggressive stimulation. In addition, the findings reveal thatthe effects of such exposure are to some extent a function of the sex of themodel, sex of the child, and the reality cues of the model.

Most of the research on the possible effectsof film-mediated stimulation upon subsequentaggressive behavior has focused primarily onthe drive reducing function of fantasy. Whilethe experimental evidence for the catharsisor drive reduction theory is equivocal (Albert,1957; Berkowitz, 1962; Emery, 1959; Fesh-bach, 1955, 1958; Kenny, 1952; Lovaas,1961; Siegel, 1956), the modeling influenceof pictorial stimuli has received little researchattention.

A recent incident (San Francisco Chronicle,1961) in which a boy was seriously knifedduring a re-enactment of a switchblade knifefight the boys had seen the previous eveningon a televised rerun of the James Dean movie,Rebel Without a Cause, is a dramatic illustra-tion of the possible imitative influence offilm stimulation. Indeed, anecdotal data sug-gest that portrayal of aggression throughpictorial media may be more influential inshaping the form aggression will take when aperson is instigated on later occasions, than inaltering the level of instigation to aggression.

In an earlier experiment (Bandura &1 This investigation was supported in part by Re-

search Grants M-4398 and M-S162 from the NationalInstitute of Health, United States Public HealthService, and the Lewis S. Haas Child DevelopmentResearch Fund, Stanford University.

The authors are indebted to David J. Hicks forhis generous assistance with the photography and toJohn Steinbruner who assisted with various phasesof this study.

2 This research was carried out while the juniorauthor was the recipient of an American Associationof University Women International Fellowship forpostdoctoral research.

Huston, 1961), it was shown that childrenreadily imitated aggressive behavior exhibitedby a model in the presence of the model. Asucceeding investigation (Bandura, Ross, &Ross, 1961), demonstrated that children ex-posed to aggressive models generalized aggres-sive responses to a new setting in which themodel was absent. The present study soughtto determine the extent to which film-mediatedaggressive models may serve as an importantsource of imitative behavior.

Aggressive models can be ordered on a reality-fictional stimulus dimension with real-lifemodels located at the reality end of the con-tinuum, nonhuman cartoon characters at thefictional end, and films portraying humanmodels occupying an intermediate position. Itwas predicted, on the basis of saliency andsimilarity of cues, that the more remote themodel was from reality, the weaker would bethe tendency for subjects to imitate the be-havior of the model.

Of the various interpretations of imitativelearning, the sensory feedback theory ofimitation recently proposed by Mowrer (1960)is elaborated in greatest detail. According tothis theory, if certain responses have beenrepeatedly positively reinforced, proprioceptivestimuli associated with these responses acquiresecondary reinforcing properties and thus theindividual is predisposed to perform the be-havior for the positive feedback. Similarly,if responses have been negatively reinforced,response correlated stimuli acquire the capac-ity to arouse anxiety which, in turn, inhibitthe occurrence of the negatively valenced

A. BANDUEA, D. Ross, AND S. A. Ross

behavior. On the basis of these considerations,it was predicted subjects who manifest highaggression anxiety would perform significantlyless imitative and nonimitative aggression thansubjects who display little anxiety over ag-gression. Since aggression is generally con-sidered female inappropriate behavior, andtherefore likely to be negatively reinforced ingirls (Sears, Maccoby, & Levin, 1957), it wasalso predicted that male subjects would bemore imitative of aggression than females.

To the extent that observation of adultsdisplaying aggression conveys a certain degreeof permissiveness for aggressive behavior, itmay be assumed that such exposure not onlyfacilitates the learning of new aggressive re-sponses but also weakens competing inhibitoryresponses in subjects and thereby increases theprobability of occurrence of previously learnedpatterns of aggression. It was predicted, there-fore, that subjects who observed aggressivemodels would display significantly more ag-gression when subsequently frustrated thansubjects who were equally frustrated but whohad no prior exposure to models exhibitingaggression.

METHOD

SubjectsThe subjects were 48 boys and 48 girls enrolled in

the Stanford University Nursery School. They rangedin age from 35 to 69 months, with a mean age of52 months.

Two adults, a male and a female, served in therole of models both in the real-life and the humanfilm-aggression condition, and one female experi-menter conducted the study for all 96 children.

General ProcedureSubjects were divided into three experimental

groups and one control group of 24 subjects each.One group of experimental subjects observed real-life aggressive models, a second group observed thesesame models portraying aggression on film, while athird group viewed a film depicting an aggressivecartoon character. The experimental groups were fur-ther subdivided into male and female subjects sothat half the subjects in the two conditions involv-ing human models were exposed to same-sex models,while the remaining subjects viewed models of theopposite sex.

Following the exposure experience, subjects weretested for the amount of imitative and nonimita-tive aggression in a different experimental settingin the absence of the models.

The control group subjects had no exposure to

the aggressive models and were tested only in thegeneralization situation.

Subjects in the experimental and control groupswere matched individually on the basis of ratingsof their aggressive behavior in social interactions inthe nursery school. The experimenter and a nurseryschool teacher rated the subjects on four five-pointrating scales which measured the extent to whichsubjects displayed physical aggression, verbal ag-gression, aggression toward inanimate objects, andaggression inhibition. The latter scale, which dealt withthe subjects' tendency to inhibit aggressive reactionsin the face of high instigation, provided the measureof aggression anxiety. Seventy-one percent of thesubjects were rated independently by both judgesso as to permit an assessment of interrater agreement.The reliability of the composite aggression score,estimated by means of the Pearson product-momentcorrelation, was .80.

Data for subjects in the real-life aggression con-dition and in the control group were collected aspart of a previous experiment (Bandura et al., 1961).Since the procedure is described in detail in theearlier report, only a brief description of it will bepresented here.

Experimental Conditions

Subjects in the Real-Life Aggressive condition werebrought individually by the experimenter to theexperimental room and the model, who was in thehallway outside the room, was invited by the ex-perimenter to come and join in the game. The subjectwas then escorted to one corner of the room andseated at a small table which contained potatoprints, multicolor picture stickers, and colored paper.After demonstrating how the subject could designpictures with the materials provided, the experi-menter escorted the model to the opposite cornerof the room which contained a small table andchair, a tinker toy set, a mallet, and a 5-footinflated Bobo doll. The experimenter explained thatthis was the model's play area and after the modelwas seated, the experimenter left the experimentalroom.

The model began the session by assembling thetinker toys but after approximately a minute hadelapsed, the model turned to the Bobo doll andspent the remainder of the period aggressing towardit with highly novel responses which are unlikelyto be performed by children independently of theobservation of the model's behavior. Thus, inaddition to punching the Bobo doll, the modelexhibited the following distinctive aggressive actswhich were to be scored as imitative responses:

The model sat on the Bobo doll and punched itrepeatedly in the nose.

The model then raised the Bobo doll and pommeledit on the head with a mallet.

Following the mallet aggression, the model tossedthe doll up in the air aggressively and kicked itabout the room. This sequence of physically aggres-sive acts was repeated approximately three times,

IMITATION OF FILM MODELS

interspersed with verbally aggressive responses suchas, "Sock him in the nose . . . ," "Hit him down ...,""Throw him in the air ... ," "Kick him ... ," and"Pow."

Subjects in the Human Film-Aggression conditionwere brought by the experimenter to the semi-darkened experimental room, introduced to thepicture materials, and informed that while the sub-jects worked on potato prints, a movie would beshown on a screen, positioned approximately 6feet from the subject's table. The movie projectorwas located in a distant corner of the room andwas screened from the subject's view by large woodenpanels.

The color movie and a tape recording of thesound track was begun by a male projectionist assoon as the experimenter left the experimental roomand was shown for a duration of 10 minutes. Themodels in the film presentations were the same adultmales and females who participated in the Real-Lifecondition of the experiment. Similarly, the aggressivebehavior they portrayed in the film was identicalwith their real-life performances.

For subjects in the Cartoon Film-Aggression con-dition, after seating the subject at the table withthe picture construction material, the experimenterwalked over to a television console approximately3 feet in front of the subject's table, remarked,"I guess I'll turn on the color TV," and ostensiblytuned in a cartoon program. The experimenter thenleft the experimental room. The cartoon was shownon a glass lens screen in the television set by meansof a rear projection arrangement screened from thesubject's view by large panels.

The sequence of aggressive acts in the cartoon wasperformed by the female model costumed as a blackcat similar to the many cartoon cats. In order toheighten the level of irreality of the cartoon, thefloor area was covered with artificial grass and thewalls forming the backdrop were adorned withbrightly colored trees, birds, and butterflies creatinga fantasyland setting. The cartoon began with aclose-up of a stage on which the curtains wereslowly drawn revealing a picture of a cartoon catalong with the title, Herman the Cat. The remainderof the film showed the cat pommeling the Bobo dollon the head with a mallet, sitting on the doll andpunching it in the nose, tossing the doll in the air,and kicking it about the room in a manner identicalwith the performance in the other experimental con-ditions except that the cat's movements were char-acteristically feline. To induce further a cartoon set,the program was introduced and concluded withappropriate cartoon music, and the cat's verbal ag-gression was repeated in a high-pitched, animatedvoice.

In both film conditions, at the conclusion of themovie the experimenter entered the room and thenescorted the subject to the test room.

Aggression Instigation

In order to differentiate clearly the exposure andtest situations subjects were tested for the amount

of imitative learning in a different experimentalroom which was set off from the main nurseryschool building.

The degree to which a child has learned aggres-sive patterns of behavior through imitation be-comes most evident when the child is instigated toaggression on later occasions. Thus, for example, theeffects of viewing the movie, Rebel Without a Cause,were not evident until the boys were instigated toaggression the following day, at which tune theyre-enacted the televised switchblade knife fight inconsiderable detail. For this reason, the children inthe experiment, both those in the control group,and those who were exposed to the aggressive models,were mildly frustrated before they were broughtto the test room.

Following the exposure experience, the experimenterbrought the subject to an anteroom which containeda varied array of highly attractive toys. The ex-perimenter explained that the toys were for the sub-ject to play with, but, as soon as the subject becamesufficiently involved with the play material, theexperimenter remarked that these were her very besttoys, that she did not let just anyone play withthem, and that she had decided to reserve these toysfor some other children. However, the subject couldplay with any of the toys in the next room. Theexperimenter and the subject then entered the ad-joining experimental room.

It was necessary for the experimenter to remainin the room during the experimental session; other-wise, a number of the children would either refuseto remain alone or would leave before the termina-tion of the session. In order to minimize any in-fluence her presence might have on the subject's be-havior, the experimenter remained as inconspicuousas possible by busying herself with paper work at adesk in the far corner of the room and avoidingany interaction with the child.

Test for Delayed Imitation

The experimental room contained a variety oftoys, some of which could be used in imitativeor nonimitative aggression, and others which tendedto elicit predominantly nonaggressive forms of be-havior. The aggressive toys included a 3-foot Bobodoll, a mallet and peg board, two dart guns, anda tether ball with a face painted on it which hungfrom the ceiling. The nonaggressive toys, on theother hand, included a tea set, crayons and coloringpaper, a ball, two dolls, three bears, cars and trucks,and plastic farm animals.

In order to eliminate any variation in behaviordue to mere placement of the toys in the room, theplay material was arranged in a fixed order for eachof the sessions.

The subject spent 20 minutes in the experimentalroom during which time his behavior was rated interms of predetermined response categories by judgeswho observed the session through a one-way mirrorin an adjoining observation room. The 20-minutesession was divided in S-second intervals by means

A. BANDURA, D. Ross, AND S. A. Ross

TABLE 1

MEAN AGGRESSION SCORES TOR SUBGROUPS OP EXPERIMENTAL AND CONTROL SUBJECTS

Response category

Total aggressionGirlsBoys

Imitative aggressionGirlsBoys

Mallet aggressionGirlsBoys

Sits on Bobo doll"GirlsBoys

Nonimitative aggressionGirlsBoys

Aggressive gun playGirlsBoys

Experimental groups

Real-life aggressive

KModel

65.876.8

19.218.4

17.215.5

10.41.3

27.635.5

1.87.3

MModel

57.3131.8

9.238.4

18.728.8

5.60.7

24.948.6

4.515.9

Human fi lm-aggressive

FModel

87.0114.5

10.034.3

49.220.5

10.37.7

24.046.8

3.812.8

MModel

79.585.0

8.013.3

19.516.3

4.50.0

34.331.8

17.623.7

Cartoon film-aggressive

80.9117.2

7.816.2

36.812.5

15.35.6

27.571.8

8.816.6

Controlgroup

36.472.2

1.83.9

13.113.5

3.30.6

17.840.4

3.714.3

a This response category was not included in the total aggression score.

of an electric interval timer, thus yielding a totalnumber of 240 response units for each subject.

The male model scored the experimental sessionsfor all subjects. In order to provide an estimateof interjudge agreement, the performances of 40%of the subjects were scored independently by asecond observer. The responses scored involved highlyspecific concrete classes of behavior, and yielded highinterscorer reliabilities, the product-moment coeffi-cients being in the .90s.

Response Measures

The following response measures were obtained:Imitative aggression. This category included acts of

striking the Bobo doll with the mallet, sitting onthe doll and punching it in the nose, kicking thedoll, tossing it in the air, and the verbally aggressiveresponses, "Sock him," "Hit him down," "Kick him,""Throw him in the air," and "Pow."

Partially imitative responses. A number of subjectsimitated the essential components of the model'sbehavior but did not perform the complete act, orthey directed the imitative aggressive response tosome object other than the Bobo doll. Two responsesof this type were scored and were interpreted aspartially imitative behavior:

Mallet aggression. The subject strikes objectsother than the Bobo doll aggressively with the mallet.

Sits on Bobo doll. The subject lays the Bobo dollon its side and sits on it, but does not aggresstoward it.

Nonimitative aggression. This category includedacts of punching, slapping, or pushing the doll,

physically aggressive acts directed toward objectsother than the Bobo doll, and any hostile remarksexcept for those in the verbal imitation category;for example, "Shoot the Bobo," "Cut him," "Stupidball," "Knock over people," "Horses fighting, biting."

Aggressive gun play. The subject shoots darts oraims the guns and fires imaginary shots at objects inthe room.

Ratings were also made of the number of be-havior units in which subjects played nonaggressivelyor sat quietly and did not play with any of thematerial at all.

RESULTS

The mean imitative and nonimitative ag-gression scores for subjects in the variousexperimental and control groups are presentedin Table 1.

Since the distributions of scores departedfrom normality and the assumption of homo-geneity of variance could not be made formost of the measures, the Freidman two-wayanalysis of variance by ranks was employedfor testing the significance of the obtaineddifferences.

Total Aggression

The mean total aggression scores for sub-jects in the real-life, human film, cartoon film,and the control groups are 83, 92, 99, and 54,

IMITATION OF FILM MODELS

respectively. The results of the analysis ofvariance performed on these scores reveal thatthe main effect of treatment conditions issignificant (xr2 = 9.06, p < .OS), confirmingthe prediction that exposure of subjects toaggressive models increases the probabilitythat subjects will respond aggressively wheninstigated on later occasions. Further analysesof pairs of scores by means of the Wilcoxonmatched-pairs signed-ranks test show thatsubjects who viewed the real-life models andthe film-mediated models do not differ fromeach other in total aggressiveness but all threeexperimental groups expressed significantlymore aggressive behavior than the controlsubjects (Table 2).

Imitative Aggressive Responses

The Freidman analysis reveals that ex-posure of subjects to aggressive models is alsoa highly effective method for shaping subjects'aggressive responses (xr2 — 23.88, p < .001).Comparisons of treatment conditions by theWilcoxon test reveal that subjects who ob-served the real-life models and the film-mediated models, relative to subjects in thecontrol group, performed considerably moreimitative physical and verbal aggression(Table 2).

Illustrations of the extent to which someof the subjects became virtually "carboncopies" of their models in aggressive behaviorare presented in Figure 1. The top frameshows the female model performing the fournovel aggressive responses; the lower framesdepict a male and a female subject reproduc-

ing the behavior of the female model theyhad observed earlier on film.

The prediction that imitation is positivelyrelated to the reality cues of the model wasonly partially supported. While subjects whoobserved the real-life aggressive models ex-hibited significantly more imitative aggressionthan subjects who viewed the cartoon model,no significant differences were found betweenthe live and film, and the film and cartoonconditions, nor did the three experimentalgroups differ significantly in total aggressionor in the performances of partially imitativebehavior (Table 2). Indeed, the availabledata suggest that, of the three experimentalconditions, exposure to humans on film por-traying aggression was the most influential ineliciting and shaping aggressive behavior.Subjects in this condition, in relation to thecontrol subjects, exhibited more total aggres-sion, more imitative aggression, more partiallyimitative behavior, such as sitting on the Bobodoll and mallet aggression, and they engagedin significantly more aggressive gun play. Inaddition, they performed significantly moreaggressive gun play than did subjects whowere exposed to the real-life aggressive models(Table 2).

Influence of Sex of Model and Sex of Child

In order to determine the influence of sexof model and sex of child on the expression ofimitative and nonimitative aggression, thedata from the experimental groups were com-bined and the significance of the differences^between groups was assessed by t tests for

TABLE 2SIGNIFICANCE OF THE DIFFERENCES BETWEEN EXPERIMENTAL AND CONTROL GROUPS

IN THE EXPRESSION OF AGGRESSION

Response category

Total aggressionImitative aggressionPartial imitation

Mallet aggressionSits on Bobo doll

Nonimitative aggressionAggressive gun play

X,'

9.0623.88

7.368.057.288.06

t

<.OS<.001

.10>/>>.05<.05

.10>/>>.05<.OS

Comparison of treatment conditions"

Live vs.Film

P

nsns

ns

<.01b

Live vs.Cartoon

P

ns<.OS

ns

ns

Film vs.Cartoon

P

nsns

ns

ns

Live vs.Control

P

<.01<.001

ns

ns

Film vs.Control

P

<.01<.001

<.os<.05

Cartoon vs.Control

P

<.oos<.oos

<.005

ns

» The probability values are based on the Wilcoxon teat.b This probability value is based on a two-tailed test of signlfic

A. BANDUKA, D. Ross, AND S. A. Ross

FIG. I. Photographs from the film, Social Learning of Aggression through Imitationof Aggressive Models.

uncorrelated means. In statistical comparisonsinvolving relatively skewed distributions ofscores the Mann-Whitney U test was em-ployed.

Sex of subjects had a highly significanteffect on both the learning and the perform-ance of aggression. Boys, in relation to girls,exhibited significantly more total aggression(t = 2.69, p < ,01), more imitative aggression(t = 2.82, p<.005), more aggressive gunplay (z = 3.38, p < .001), and more nonimita-tive aggressive behavior (t = 2.98, p < .005).Girls, on the other hand, were more inclinedthan boys to sit on the Bobo doll but refrainedfrom punching it (z = 3.47, p < .001).

The analyses also disclosed some influencesof the sex of the model. Subjects exposed tothe male model, as compared to the femalemodel, expressed significantly more aggressivegun play (z = 2.83, p < .005). The mostmarked differences in aggressive gun play((7 = 9.5, ^ < . 0 0 1 ) , however, were foundbetween girls exposed to the female model(M = 2.9) and males who observed the malemodel (M=19.8) . Although the overallmodel difference in partially imitative be-

havior, Sits on Bobo, was not significant,Sex X Model subgroup comparisons yieldedsome interesting results. Boys who observedthe aggressive female model, for example,were more likely to sit on the Bobo dollwithout punching it than boys who viewedthe male model ( f /=33 , p<.Q5). Girlsreproduced the nonaggressive component ofthe male model's aggressive pattern of be-havior (i.e., sat on the doll without punchingit) with considerably higher frequency thandid boys who observed the same model (U =21.5, p<.Q2). The highest incidence ofpartially imitative responses was yielded bythe group of girls who viewed the aggressivefemale model (M=10.4), and the lowestvalues by the boys who were exposed to themale model (If = 0.3). This difference wassignificant beyond the .05 significance level.These findings, along with the sex of child andsex of model differences reported in thepreceding sections, provide further support forthe view that the influence of models inpromoting social learning is determined, inpart, by the sex appropriateness of the model'sbehavior (Bandura et al., 1961).

IMITATION OF FILM MODELS

Aggressive Predisposition and Imitation

Since the correlations between ratings ofaggression and the measures of imitative andtotal aggressive behavior, calculated separatelyfor boys and girls in each of the experimentalconditions, did not differ significantly, thedata were combined. The correlational anal-yses performed on these pooled data failed toyield any significant relationships betweenratings of aggression anxiety, frequency ofaggressive behavior, and the experimentalaggression measures. In fact, the array meanssuggested nonlinear regressions although thedepartures from linearity were not of sufficientmagnitude to be statistically significant.

DISCUSSION

The results of the present study providestrong evidence that exposure to filmed ag-gression heightens aggressive reactions inchildren. Subjects who viewed the aggressivehuman and cartoon models on film exhibitednearly twice as much aggression than didsubjects in the control group who were notexposed to the aggressive film content.

In the experimental design typically em-ployed for testing the possible cathartic func-tion of vicarious aggression, subjects are firstfrustrated, then provided with an opportunityto view an aggressive film following whichtheir overt or fantasy aggression is measured.While this procedure yields some informationon the immediate influence of film-mediatedaggression, the full effects of such exposuremay not be revealed until subjects are in-stigated to aggression on a later occasion.Thus, the present study, and one recentlyreported by Lovaas (1961), both utilizing adesign in which subjects first observed filmedaggression and then were frustrated, clearlyreveal that observation of models portrayingaggression on film substantially increasesrather than decreases the probability of ag-gressive reactions to subsequent frustrations.

Filmed aggression, not only facilitated theexpression of aggression, but also effectivelyshaped the form of the subjects' aggressivebehavior. The finding that children modeledtheir behavior to some extent after the filmcharacters suggests that pictorial mass media,particularly television, may serve as an im-

portant source of social behavior. In fact, apossible 'generalization of responses originallylearned in the television situation to the ex-perimental film may account for the signifi-cantly greater amount of aggressive gun playdisplayed by subjects in the film condition ascompared to subjects in the real-life and con-trol groups. It is unfortunate that the qualita-tive features of the gun behavior were notscored since subjects in the film condition,unlike those in the other two groups, devel-oped interesting elaborations in gun play (forexample, stalking the imaginary opponent,quick drawing, and rapid firing), character-istic of the Western gun fighter.

The view that the social learning of ag-gression through exposure to aggressive filmcontent is confined to deviant children(Schramm, Lyle, & Parker, 1961), finds littlesupport in our data. The children who par-ticipated in the experiment are by no meansa deviant sample, nevertheless, 88% of thesubjects in the Real-Life and in the HumanFilm condition, and 79% of the subjects inthe Cartoon Film condition, exhibited varyingdegrees of imitative aggression. In assessingthe possible influence of televised stimulationon viewers' behavior, however, it is importantto distinguish between learning and overtperformance. Although the results of thepresent experiment demonstrate that the vastmajority of children learn patterns of socialbehavior through pictorial stimulation, never-theless, informal observation suggests thatchildren do not, as a rule, perform indis-criminately the behavior of televised char-acters, even those they regard as highly at-tractive models. The replies of parents whosechildren participated in the present study toan open-end questionnaire item concerningtheir handling of imitative behavior suggestthat this may be in part a function of negativereinforcement, as most parents were quick todiscourage their children's overt imitation oftelevision characters by prohibiting certainprograms or by labeling the imitative behaviorin a disapproving manner. From our knowl-edge of the effects of punishment on behavior,the responses in question would be expectedto retain their original strength and couldreappear on later occasions in the presence ofappropriate eliciting stimuli, particularly if

10 A. BANDUEA, D. Ross, AND S. A. Ross

instigation is high, the instruments for ag-gression are available, and the threat of nox-ious consequences is reduced.

The absence of any relationships betweenratings of the children's predisposition to ag-gression and their aggressive behavior in theexperimental setting may simply reflect theinadequacy of the predictor measures. It maybe pointed out, however, that the reliabilityof the ratings was relatively high. While thisdoes not assure validity of the measures, itdoes at least indicate there was consistency inthe raters' estimates of the children's aggres-sive tendencies.

A second, and perhaps more probable, ex-planation is that proprioceptive feedbackalone is not sufficient to account for responseinhibition or facilitation. For example, theproprioceptive cues arising from hitting re-sponses directed toward parents and towardpeers may differ little, if any; nevertheless,tendencies to aggress toward parents are aptto be strongly inhibited while peer aggressionmay be readily expressed (Bandura, 1960;Bandura & Walters, 1959). In most socialinteraction sequences, proprioceptive cuesmake up only a small part of the total stimulus•complex and, therefore, it is necessary to takeinto consideration additional stimulus com-ponents, for the most part external, whichprobably serve as important discriminativecues for the expression of aggression. Con-sequently, prediction of the occurrence or in-hibition of specific classes of responses wouldbe expected to depend upon the presence of acertain pattern of proprioceptive or introcep-tive stimulation together with relevant dis-criminative external stimuli.

According to this line of reasoning, failureto obtain the expected positive relationshipsbetween the measures of aggression may bedue primarily to the fact that permissivenessfor aggression, conveyed by situational cuesin the form of aggressive film content andplay material, was sufficient to override theinfluence of internal stimuli generated by thecommission of aggressive responses. If, in fact,the behavior of young children, as comparedto that of adults, is less likely to be underinternal stimulus control, one might expectenvironmental cues to play a relatively im-

portant role in eliciting or inhibiting aggres-sive behavior.

A question may be raised as to whether theaggressive acts studied in the present experi-ment constitute "genuine" aggressive re-sponses. Aggression is typically denned asbehavior, the goal or intent of which is injuryto a person, or destruction of an object(Bandura & Walters, 1959; Bollard, Doob,Miller, Mowrer, & Sears, 1939; Sears, Mac-coby, & Levin, 1957). Since intentionality isnot a property of behavior but primarily aninference concerning antecedent events, thecategorization of an .act as "aggressive" in-volves a consideration of both stimulus andmediating or terminal response events.

According to a social learning theory ofaggression recently proposed by Bandura andWalters (in press), most of the responsesutilized to hurt or to injure others (for ex-ample, striking, kicking, and other responsesof high magnitude), are probably learned forprosocial purposes under nonfrustration con-ditions. Since frustration generally elicitsresponses of high magnitude, the latter classesof responses, once acquired, may be called outin social interactions for the purpose of injur-ing others. On the basis of this theory it wouldbe predicted that the aggressive responsesacquired imitatively, while not necessarilymediating aggressive goals in the experimentalsituation, would be utilized to serve suchpurposes in other social settings with higherfrequency by children in the experimentalconditions than by children in the controlgroup.

The present study involved primarily vicar-ious or empathic learning (Mowrer, 1960)in that subjects acquired a relatively complexrepertoire of aggressive responses by the meresight of a model's behavior. It has been gen-erally assumed that the necessary conditionsfor the occurrence of such learning is that themodel perform certain responses followed bypositive reinforcement to the model (Hill,1960; Mowrer, 1960). According to thistheory, to the extent that the observer experi-ences the model's reinforcement vicariously,the observer will be prone to reproduce themodel's behavior. While there is some evidencefrom experiments involving both human(Lewis & Duncan, 1958; McBrearty, Marston,

IMITATION OF FILM MODELS 11

& Kanfer, 1961; Sechrest, 1961) and animalsubjects (Darby & Riopelle, 1959; Warden,Fjeld, & Koch, 1940), that vicarious rein-forcement may in fact increase the probabilityof the behavior in question, it is apparentfrom the results of the experiment reportedin this paper that a good deal of human imita-tive learning can occur without any reinforcersdelivered either to the model or to the ob-server. In order to test systematically theinfluence of vicarious reinforcement on imita-tion, however, a study is planned in which thedegree of imitative learning will be comparedin situations in which the model's behavior ispaired with reinforcement with those in whichthe model's responses go unrewarded.

REFERENCES

ALBERT, R. S. The role of mass media and the effectof aggressive film content upon children's aggres-sive responses and identification choices. Genet,psychol. Monogr., 19S7, 55, 221-285.

BANDURA, A. Relationship of family patterns tochild behavior disorders. Progress Report, 1960,Stanford University, Project No. M-1734, UnitedStates Public Health Service.

BANDURA, A., & HUSTON, ALETHA C. Identificationas a process of incidental learning. J. abnorm. soc.Psychol., 1961, 63, 311-318.

BANDURA, A., Ross, DOROTHEA, & Ross, SHEILA A.Transmission of aggression through imitation ofaggressive models. 7. abnorm. soc. Psychol., 1961,63, 575-582.

BANDURA, A., & WALTERS, R. H. Adolescent aggression.New York: Ronald, 1959.

BANDURA, A., & WALTERS, R. H. The social learning,of deviant behavior: A behavioristic approach tosocialization. New York: Holt, Rinehart, & Wins-ton, in press.

BERKOWITZ, L. Aggression: A social psychologicalanalysis. New York: McGraw-Hill, 1962.

DARBY, C. L., & RIOPELLE, A. J. Observational learn-ing in the Rhesus monkey. J. comp. physiol.Psychol, 1QS9, 52, 94-98.

DOLLARD, J., DOOB, L. W., MlLLER, N. E., MOWRER,O. H., & SEARS, R. R. Frustration and aggression.New Haven: Yale Univer. Press, 1939.

EMERY, F. E. Psychological effects of the Westernfilm: A study in television viewing: II. The ex-perimental study. Hum. Relat., 1959, 12, 21S-232.

FESHBACH, S. The drive-reducing function of fantasybehavior. J. abnorm. soc. Psychol., 19S5, 50, 3-11.

FESHBACH, S. The stimulating versus cathartic ef-fects of a vicarious aggressive activity. Paper readat the Eastern Psychological Association, 19S8.

HILL, W. F. Learning theory and the acquisition ofvalues. Psychol. Rev., 1960, 67, 317-331.

KENNY, D. T. An experimental test of the catharsistheory of aggression. Unpublished doctoral disserta-tion, University of Washington, 1952.

LEWIS, D. J., & DUNCAN, C. P. Vicarious experienceand partial reinforcement. J. abnorm. soc. Psychol.,1958, 57, 321-326.

LOVAAS, 0. J. Effect of exposure to symbolic ag-gression on aggressive behavior. Child Develpm.,1961, 32, 37-44.

McBREARiy, J. F., MARSTON, A. R., & KANFER, F. H.Conditioning a verbal operant in a group setting:Direct vs. vicarious reinforcement. Atner. Psy-chologist, 1961, 16, 425. (Abstract)

MOWRER, O. H. Learning theory and the symbolicprocesses. New York: Wiley, 1960.

SAN FRANCISCO CHRONICLE. "James Dean" knifingin South City. San Francisco Chron., March 1,1961, 6.

SCHRAMM, W., LYLE, J., & PARKER, E. B. Televisionin the lives of our children. Stanford: StanfordUniver. Press, 1961.

SEARS, R. R., MACCOBY, ELEANOR E., & LEVIN, H.Patterns of child rearing. Evanston: Row, Peterson,1957.

SECHREST, L. Vicarious reinforcement of responses.Amer. Psychologist, 1961, 16, 356. (Abstract)

SIEGEL, ALBERTA E. Film-mediated fantasy aggressionand strength of aggressive drive. Child Develpm.,1956, 27, 365-378.

WARDEN, C. J., FJELD, H. A., & KOCH, A. M.Imitative behavior in cebus and Rhesus monkeys./. genet. Psychol, 1940, 56, 311-322.

(Received September 21, 1961)