television violence and aggression

8/6/2019 Television Violence and Aggression

1/8

... --_-.ilBulletin1986, Vol.100, No.3,364-371 Copyright 1986 by the American Psychological Association, Inc.0033-2909/86/$00.75

TelevisionViolenceandAggression:TheDebateContinuesLynetteFriedrich-Cofer

University of North Carolina at Chapel HillAletha C. Huston

DepartmentofHuman DevelopmentUniversity ofKansas

Many social scientistsbelieve there is a causal relationbetween viewing television violenceand ag-gression. That majority consensus has recently been challenged on the grounds that the data areinconsistent, that different methods of study may be subject to systematic biases, and that the find-ings have limitedgeneralizability to real-world violence. In this review we reply to thesechallenges.Weconclude that thedatasupportabidirectionalcausal relation between viewing television violenceandaggression, that the potential threats to the internal and externalvalidity ofstudiesusing differentmethods are not likely to produce a positive bias, and that the findings can be generalized. Theavailable research is placed in a theoretical context encompassing multiple psychological processesand developmental change,and social policy implications are discussed.

Sincethe advent ofmoviesand particularly television, therehasbeen social concern about thepotentialeffects ofmedia vio-lenceon the attitudes, values,and aggressive behaviorofyoungviewers. Alarge bodyofpsychological research on these ques-tionswassummarizedand evaluated in the 1972Surgeon Gen-eral's Report (Surgeon General's Scientific Advisory Commit-tee on Television and Social Behavior, 1972) and the 1982 re-port from the National Institute of Mental Health (Pearl,Bouthilet, & Lazar, 1982).Their conclusion that television vio-lence has acausal effect onaggressive behaviorforchildrenandadolescents was recently reaffirmed by the American Psycho-logicalAssociation(1985).

Themajorityconsensusof the social science community, re-flected in these reports, hasrecently been challenged. Freedman(1984)concluded that the available empirical evidence does notsupport a causal relation between television violence and ag-gression and doesnot justify efforts to influence public policy.Heargued that laboratory research is irrelevant to the issue be-cause it lacks external validity, and that field experiments andlongitudinal studies have produced weak and inconsistent re-sults.

Cook, Kendziersky, andThomas(1983) concluded that thereis asmallbut reliableeffect of television violenceon aggression,but raised conceptual and methodological questions about theinterpretations in the National Instituteof Mental Health re-port. They argued that convergenceof findings from differentmethods (e.g., laboratory experiments and field studies) couldbedue to common biases across methods, and they questionedwhether the aggression measured inmost studiescan begener-

The authors made equal contributions to the article; therefore, theyare listed in alphabetical order.

This work wassupported in part by National Institute of MentalHealth Grant MH 39596 to Aletha C. Huston.

We are grateful to John C. Wright, Charles Cofer, and Eli Rubinsteinfor their critical reading of the manuscript and helpful suggestions.

Correspondence concerning this article should be sent to LynetteFriedrich-Cofei; Department of Psychology, University of North Caro-lina, Chapel Hill, North Carolina 27514.

alized to socially important forms of violence such as delin-quent andcriminalbehavior.

In thisreviewwedefend the majorityconsensusand reply tosome of the issues raised byFreedman (1984) and byCook etal. (1983). The internal and external validity of the availablestudies is discussed. Particular attention is given to potentialsources of bias inherent in the stimuli, settings, and measuresused, and to thegeneralizabilityof the findings to situations ofsocial concern. Finally, both theoretical and social policy im-plications are discussed.

Experimental Studies in Laboratory andFieldSettingsLaboratory Experiments

Most reviewers agree that laboratory experimentsareconsis-tent in showing that viewing violent television leads toaggres-sivebehavior forboth childrenandadolescents (Andison, 1977;Freedman, 1984; Stein&Friedrich, 1975).Laboratory experi-mentsgenerallyhave high internal validity because randomas-signment ofsubjects to treatments and manipulation of the in-dependent variable permit causal inference and ensure that thedifferences betweengroupsare not a function ofother unmea-sured variables.

StimuliFreedman (1984) raised threemajor questions about the ex-

ternal validity of laboratory studies. One challenge concernedthe degree towhichthe stimuli used represent real-world televi-sion. In the early studies ofyoung children, filmswere experi-mentally constructed (Bandura &Walters, 1963);however, inmany studiesof children and adults, real television programsor films were shown(Berkowitz, 1984; Collins&Getz, 1976;Stein &Friedrich, 1975). In almost all of these investigations,nonviolent control programs were included in an effort to iso-late the violent content of the experimental treatments as theimportant variable.

The stimuli used in laboratory experiments werenot atypi-callyviolent or unlike normal programming. Content analyses

364


2/8

TELEVISION AND AGGRESSION 365since 1968have demonstrated that thereare 5 or 6incidentsofviolence per hour in prime time televisionand from 15 to 25incidents per hour in cartoons (Signorielli, Gross, & Morgan,1982). Indeed, ethical considerations often led experimentersto useprograms thatdid not reflect maximum levelsofviolenceavailable in real-world television.

Selling. A second criticism of laboratory studies is thattheir artificial settings focus attention on the television contentandlead toexperimenter demand forimitation. Although suchdemand mayoccur, thereis noevidencethat itaccountsfor theeffects of aggressive television. On the contrary, for children,aggressive televisionismorelikely toproduce aggressive behav-iorwhen theexperimenter leaves the child alone than when theadult remains duringthe test ofaggression (Stein&Friedrich,1975). Similarly, in studies of adolescents and adults, the evi-dence contradicts the hypothesis that experimenter demandac-counts for aggressive imitation (Berkowitz & Donnerstein,1982).

Measures of aggression. The third major criticism of labo-ratory studies concernsthe external validityofmeasuresof ag-gression. Some studies of children have measured "playful"punching ofBobodollsor other aggressivetoy play, but inter-personal aggression in play groups or in classrooms was ob-servedinother studies(Ellis & Sekyra, 1972;Ross, 1972; Steuer,Applefield, &Smith, 1971). Analog measures, suchasadminis-tering shockorpressing buttons to interfere with another's ac-tivity, constitute another indexof aggression (Collins &Getz,1976; Liebert &Baron, 1972).Thevalidityofanalog measuresissupportedby the finding hat children's responseson a"hurt"machine were correlated with naturally occurring interpersonalaggression (Johnston, DeLuca.Murtaugh, &Diener, 1977).Fi-nally, children'sselections ofaggressive,avoidant, orprosocialsolutionsto interpersonalconflict formed anindexininvestiga-tionsacross awideagerange (Collins, 1973;Leifer &Roberts,1972).Thepreponderance of resultsfor allthese measureswaspositive (Stein&Friedrich, 1975).

Summary. Laboratory studies makean important contri-bution becauseof their high internal validity. Challenges o ex-ternal validity suggesting that the television programs and be-haviors measured are unrepresentative of the "real world" orthat the results are due to experimenter demand do not receivestrong support from acareful examination oftheliterature. Thepotential biases in the laboratory method are both positive andnegative.The effects of television violence could be magnifiedbecause the impact of other variables isminimized, but theymightalso be underestimated because the stimuli used are briefand often less violent than the television typically availableathome.

Field ExperimentsThefield experiment appears at first glance to be the method

ofchoice forevaluating causal hypothesesinreal-world settings.It has the advantages of random assignment of people to view-ing conditions and experimental manipulation of the stimuli.External validity is enhanced by the use of natural settings,relatively long time periods, and measurement of everydaybehavior.

Field experiments have been conducted with preschool chil-

dren in nursery school settings, withyoung adolescent boys inresidential settings,and with male adolescent delinquentswhowere incarcerated. The results have been mixed, but weassessthemdifferently thanFreedman (1984)did in several respects.Someaspectsofthese studieswere not presented by Freedman;therefore, theyaresummarizedbriefly here.

Inone fieldexperiment (Friedrich &Stein, 1973), childreninanursery schoolwere assignedtoviewviolentcartoons,neutralfilms, orprosocial televisionfor 12 days across4weeks.Aggres-sivebehaviorwas recorded during free play. The main effect ofthe television program was not significant, but as predicted,therewas a significant Baseline LevelXTreatment interaction.For high-initial-aggression children, the change scores for theviolent television group were significantlydifferent from theneutral group (Friedrich&Stein, 1973,p. 38); theviolent tele-vision group remained high, whereas the neutral group de-creasedconsiderably. Becausethe mean baseline scores for thetwo groups were similar, regression to the mean should havebeen equallylikely forboth groups. Childrenwhowatched vio-lent television also declined in self-controlthat is, tolerancefor minor delays, spontaneous obedience of school rules, andtaskpersistence.

Freedman (1984) devoted considerable attention to an earlyfield study of adolescent boys in residential settings whowereassigned towatch violent ornonviolent televisiondietsforsev-eral weeks (Feshbach &Singer, 1971). In three schools, boyswatchingnonviolent programs were significantly moreaggres-sive than those watching violent programs, but the study's manymethodological problems raised serious questions about its in-ternalvalidity (Liebert,Sobol, &Davidson, 1972). Wells(1973)replicated Feshbach & Singer's basic design with much im-provedmethods, including baselinedataand objective behaviorratings. Boys who viewed a violent diet were slightly,but notsignificantly, higher in physical aggression than those whoviewed nonviolent television.The difference was significant forboys whowere above averagein initial aggression and forboyswho liked their television diets.

In a series ofexperimentsin Belgium and theUnited Statesthat was heavily criticized by Freedman (1984), violent andnonviolent movies were shown to groups of institutionalizeddelinquentandneglectedboys for 1 week (Leyens,Parke, Cam-ino, &Berkowitz, 1975;Parke, Berkowitz, Leyens,West, & Se-bastian, 1977). In the Belgian study, physical aggression in-creased significantly after viewing inboth cottages assigned toviolentfilms, but did not increase in theneutral filmcottages.Totalaggression, including both physical and verbal aggression,increased primarily in theviolent filmcottagethatwasinitiallymore aggressive. In two U.S. studies total aggression wassig-nificantly higherin twocottages viewing television violencefor5 days thanincottages viewing neutralfilms.

Theexternal validityof fieldexperimentsisenhancedby theuseofreal television programs overaperiod from one toseveralweeks, placement in real-life settings, and measurement ofnat-urally occurring aggression. Nevertheless, there are seriousmethodological problems in some cases.

Stimuli. In existing field experiments, the violent stimuliwere typical of materials viewed frequently by the age groupstudied. In several instances, however, there were major prob-lemsin findingcontrol stimuli that wereasattractiveas the vio-


3/8

366 LYNETTE FRIEDRICH-COFER AND ALETHA C. HUSTONlentprograms (Feshbach&Singer, 1971;Parkeetal., 1977,firststudy; Wells, 1973).Onlytwostudies included nonviolent pro-grams that elicited asmuch attention (Friedrich&Stein, 1973)orwere rated asattractive (Parkeetal., 1977, finalstudy)as theviolent programs. The frustration resulting from being re-quired towatch disliked programs couldhavegeneratedaggres-sionthat counteracted the effects ofviewingviolence in the fieldexperiments that failed to find an effect of violent television.When program attractiveness was controlled by comparingboys in both treatments who liked their assigned televisiondiets, the violent diet group had significantly higher levels ofphysical aggression than the nonviolent diet group (Wells,1973).

Settings. Although fieldexperiments occur in real-life set-tings, the locationsdo not represent themost common viewingenvironmentthehome. Adolescents in residential institutionsmay not represent the larger population well.One fieldexperi-ment conducted on adults at home isfraught with threats to itsinternal validity (Loye,Gorney, &Steele, 1977).

Potential invalidityisalsointroduced by the reactionsofsub-jects and staff to thedisruption of their routines and to the im-position of restrictions on their freedom ofchoiceintelevisionviewing.These problems were probably minimizedin the pre-school setting by the brevity of the sessions and the fact thattelevisionviewing sessionsfit anaccustomed pattern ofleavingthe classroom for special activities (Friedrich &Stein, 1973).Similarly, themanipulationbyParkeet al. (1977)wasprobablyrelatively unintrusivebecauseit lasted only 1 weekand replacedundesirable activities. In the other investigations (Feshbach&Singer, 1971;Wells, 1973), the experiment required a substan-tial amount of time, and students were sometimes required togiveupdesired activities.

Measures of aggression. The aggressive behaviors codedrepresent theoretically central characteristics of children andadolescents, includingphysical attacks onother people (hitting,banging, pushing, kicking, squeezing, choking, holdingdown,andthrowing objects), threats and verbal abuse, and destructionofproperty. These serious forms of interpersonaland object ag-gression are conceptually distinct from playful and fantasy ag-gression, and the two typesofaggressionare not correlated inthebehaviorofyoungchildren (Friedrich &Stein, 1973).

Externalvalidityisachieved,however, withsome lossof inter-nal validity. Freedman (1984) criticized Parke et al.'s studiesbecause whole cottages received the same treatment and wereobserved interactingwithoneanother,but were treated asinde-pendent cases in statistical analyses. Interdependence of sub-jects' behavior could reduce error variance in all existing fieldstudies.1 However, the likely bias depends on whether wholegroups experienced the same treatment (Parke et al., 1977) orwereassignedtodifferent treatments (Feshbach&Singer, 1971;Friedrich &Stein, 1973; Wells, 1973). In the former case, themean difference might be inflated; in the latter it mightbe re-duced, causing a negative bias.

Ona more substantive level, external validity is enhanced byadministering the same television treatment to whole groupsbecause real-world media effects occur in ongoing family andpeer groups that share viewing patterns and whose aggression isinterdependent. Wholefamilieshavesimilar levelsofaggression(Patterson, 1976), as do adolescent male friendship groups

(Cairns &Cairns, inpress).Ifviolent televisionhasdirect effectson a few members of such groups, it may well reverberatethroughout the family orpeer group system.

Freedman's discussion of field experiments implies that theresultsare spotty and inconsistent because significant findingsoccurred for only a few of the dependent variables measured.The variableson which significant effects occurred were mostoften composites ofsubcategories, not achance selection frommultiple measures. For example, significant findings in theParke et al. (1977) studies occurred for general aggression, acomposite of all categories, or on physical aggression, a com-positeof four. Significant effects inFriedrich and Stein's(1973)studyoccurred forinterpersonal aggression, acompositeofsev-eral subcategories representing the most theoretically centralform ofaggression.In fact, the failure to find effects on playful,fantasy aggressionis ofsubstantive interest.

Summary. The best designed field experiments form amoderately consistent patternsuggestingan effect of televisionviolenceonaggressionand self-control, particularly forsubjectswith relatively high baseline levelsof aggression. Negative re-sults occur most often in studies with seriousthreats to internalvaliditycreated by unappealing, nonviolent control treatmentsanddisruptionof ongoingsocial settings that mayincrease sub-jects' reactivity to the manipulation.

Theoverall biasof fieldexperiments appears to be in the di-rectionofunderestimating theeffects of televisionviolence. Theexperimental manipulation is weak because the programsshownare a tiny fraction of the televisionthat the subjects havewatched in their lives or even during the experiment in somecases, but the televisiontreatmentmust besufficiently robust toproduceeffects onaggressioninmethodologically noisy naturalcontexts inwhich many extraneous variablescontributeto thevariance and the usual laboratory controls are not operating.

The rate of aggression in natural settings is probablyunderes-timated by observations collected by adults in settings in whichteachers or counselors are present because aggression is nega-tively sanctioned in schools and institutions. The inhibitingeffects of adults should affect both experimental and controlgroups equally,but the reduction in levelmaymilitate againstfindingdifferences between treatments.

Correlational StudiesA large number ofcorrelational studies involving thousands

of subjectsacross widely differing levelsof age, socioeconomicstatus, and ethnic background have yielded consistently mod-est, but positive correlations (rangingfrom .10 to .35) betweenviewing naturally occurring violence and aggression (Freed-man, 1984; Stein &Friedrich, 1975). Interpretation of thesecorrelations hinges on two questions: causal direction and pos-siblecontributions ofthird variables to the association betweenviewingandaggression.Longitudinal Studies

Longitudinal studies provide the opportunity to assess tem-poral relationsamong variables and to test causal hypotheses.

1 The only way to achieve statistical independence would be to ob-serve only oneperson in each group setting, a procedure that is unrealis-tic in fieldstudies.


4/8

TELEVISION AND AGGRESSION 367

The available studies support the hypothesis that the relationsbetween viewing violence and aggression persist over time(Eron, Huesmann, Lefkowitz, & Walder, 1972; Huesmann,1982; Huesmann, Lagerspetz, &Eron, 1984; Milavsky, Stipp,Kessler,&Rubens, 1982; Singer&Singer, 1981; Singer, Singer,&Rapaczynski, 1984). All show positive correlations betweenviewingat onetimeand aggressionat another.

Cumulative and enduring effects. Freedman (1984) pro-posed that if television violence has a causal effect, it should becumulative and the correlations should increase with age.Bythat line of reasoning, correlations also ought to increase withage if aggressive personality attributes caused violent viewing.Hence, the fact that correlations do not increase consistentlywithage has little bearing on the issue of causal direction.

Causal analyses. The longitudinal method is emphasizedby Freedman (1984) and Cook etal.(1983) as an excellent toolfor testing causal relations. The basic logic of causal analysesof longitudinal data is to determine whether a variable (e.g.,television viewing) measured at one point predicts the othervariable (e.g., aggression) measured at alater point. If naturallyoccurring variations in one variable predict later variations inthe other variable, then the hypothesis thatA causes B is sup-ported.Such analyses may reflect delayed effects of televisionviewingor critical periods duringwhich television has particu-lar effects, as Freedman (1984) suggests, but they can be inter-preted more parsimoniously as assessments of temporal re-lations.

Longitudinal findings support a bidirectional model of cau-salitytelevision violenceinfluences aggression,andaggressivepredispositions influence thepreference for television violence.In the first longitudinal investigation (Eronet al., 1972), cross-lag correlations demonstrated, for male subjectsonly, that vio-lence viewing at age 8predicted aggressionat 18, but that ag-gression at age 8 did not predict violenceviewing at 18. How-ever,anycausalconclusion was weakened bystatistical artifactsin the cross-lag procedure (Cook et al., 1983). Nevertheless, ofa large number ofparent, family, and socioeconomic variablesmeasured at age 8, television was the single best predictor ofaggression in 18-year-olds (Eronetal., 1972).

Singerand Singer(1981)presented correlations betweenpre-schoolchildren'sviewing assessed by parental diary records andobserved aggression in preschoolers for four wavesofdatacol-lected in 1year. Children's viewing of "action" programs waspositively related to later aggression for both sexes in all sixpairsofwavecomparisons. The reverse comparison, predictinglater waves of viewing from earlier aggression, also yielded con-sistently positive correlations. In elementary school,viewingof"realisticaction TV" predicted later aggressionand children'sbelief in a"scary world"(Singer etal., 1984).

Amore sophisticated analysis performed by Huesmann et al.(1984) consisted of regressions designed to determine whetherearlier viewing predicted later aggression after initial levels ofaggression and age were controlled. Their data consisted ofthree annual wavesof data collected in the United States andFinland for children whowere in the first or third grade at theinitial wave.The reverse causal direction was tested by regres-sions predicting later violenceviewing from earlier aggression,controlling for grade leveland initialviewing.

The findingssupportedabidirectional model: Early violence

viewing predicted later aggression, and initial aggression pre-dicted later violence viewing.Thecoefficients were consistentlypositive forbothgenders in both countries; some reached statis-tical significance, others did not. Amongboys, violenceviewingpredicted later aggression primarily for those who identifiedwith television characters.

Milavskyet al. (1982)performed regressionsforelementaryschool children assessed six times during 3 years and adolescentboys assessed five times during 3 years in a panel design. Vio-lence viewing in the earlier wave was tested as a predictor ofaggression in the later wave, controlling for earlier aggression.They estimated LISREL models as well. In general, the coeffi-cients were positive; some weresignificant, but many were not.Theauthors concluded that television viewingmadeanegligiblecontribution to the variance accounted for.

Although Freedman (1984) accepted their negative conclu-sion, Cook et al. (1983) noted that the contribution of viewingtelevision violence was positive in most analyses, and that thecoefficients were largerthelongerthe timelagbetweenthemea-sures ofviewingand aggression. They criticized the analysis forusing low-power statistical procedures and failing to probemodels that might demonstrate the cumulative impact oftelevi-sion or interactions with subject variables such as gender andsocioeconomic status (Cooketal., 1983).

Two other problems are equally serious. The causal analysiswaslimited because theydid not test the reverse hypothesis thatearlier aggression predicted later violenceviewing. Second, nat-urally occurring aggression scoresare typicallyhighlyskewed,but no correction for skewing was reported. Unlike analysis ofvariance, causal modeling analysesare readily affected bynon-normal distributions, making any modeldifficult to replicate.

Summary. Longitudinal investigations support a small, butconsistent effect of viewing violenceon aggression. Earlyview-ingwas positively related to later aggression in all studies, evenwhen earlier aggressionwaspartialed out. Some relations werestatisticallysignificant and some were not, but there were virtu-allyno instancesofnegative relations between viewingand ag-gression. If the real relation were zero, one would expect an ap-proximately equal number of positive and negative corre-lations.

Causal analysesof longitudinaldatacould bebiased towardunderestimation of the association of variables because the ini-tial correlation betweenviewingandaggressionispartialed out.Asa result, the factors that contributed to theassociation, in-cluding earlier television viewing, are excluded. What theseanalyses evaluate is whether variations in viewingat one timecontribute tochangesinaggressive behaviorat alater time.

Third VariablesA second major question about correlational studies is the

possibility of selection bias or the contribution of third vari-ables to the associationofviewingand aggression. Cook et al.(1983) suggested that relevant background variables have notbeen adequately controlled incross-sectionalsurveys. However,inspection of the data showscareful attention to awide rangeof theoretically relevant background andpersonality variables.

Several early studies of elementary and adolescent childrenexamined social class, IQ, school achievement, age, parental ag-


5/8

368 LYNETTE FRIEDRICH-COFER AND ALETHA C. HUSTON

gression, parental warmth, restrictiveness, punitiveness, aspira-tions for the child, parental viewing of violence, control of tele-vision viewing, and stylesof family communication (Chaffee &McLeod, 1972;Lefkowitz, Eron, Walder, &Huesmann, 1972;Mclntyre&Teevan, 1972).Similarly, measures ofperceived re-alism of television, aggressivefantasy, sex role identification, ag-gressivepredispositions, parental viewing, parental aggression,social class, and the child's achievement levelwere entered inregressions by Huesmann etal. (1984). None of these variablesaccounted for the relation between viewing and aggression.

Belson (1978)conducted astill more extensive test ofpoten-tial selection variables. A stratified sample of 1,565 boys aged12-17living in Londonparticipatedin two extensive interviewsin which data about television viewing history, aggressive be-havior, and numerous other variables wereobtained.Measuresof viewing and aggression were derived. In addition, a pool of227 possible predictors of violenceviewingand aggression wasassembled. They included age, demographic information,neighborhood qualities, localityand home conditions, physicalstrength, physical maturation, attitudes, typeof school, schoolperformance and attitudes, child rearing practices, reportsofchildhood behavior and temperament, and nonaggressive delin-quentbehaviors (e.g., minor theft).

High and low viewers of violence were statistically equatedfor a subset of variables selected from the pool to maximizetheir correlation with aggression and with viewing violence.2This procedure partialed out thecontribution ofalarge numberof possible third variables that might account for the relationbetween viewing and aggression. The aggression scores of thehigh- and low-violence viewers remained significantly differenton all indices of aggression.

Although it is logically impossible to ensure that all possiblevariables contributingtoselection biashavebeen accounted for,the extensive analyses by Huesmann et al. (1984) and Belson(1978)provide strong evidence for the conclusion that the rela-tion of viewing violence to aggression is not accounted for byassociated background andpersonality variables.

Potential Sources of BiasCorrelationalstudieshavehigh external validity because they

are nonintrusive assessments of naturally occurring viewingand behavior. Internal validity is more difficult to evaluate.Viewing isusually assessed byself-reportsof unknownvalidity.There appears to be no strong reason to assume that such re-portsoverestimate orunderestimate violence viewing systemat-ically.

Freedman (1984) argued that measures of viewing violencecould represent televisionviewing in general. For example, inBelson's (1978)study, serious acts of aggression were correlatedwith total television viewed and the amount of nonviolent tele-vision viewed as well as with violent television viewing. How-ever, violence viewing washighly correlated with totalviewing(r =.87) and with nonviolent program viewing(r = .72).Whenthe amount of violent television viewed was controlled, therewasno relation of total viewingto aggressive behavior.

Measures of aggression. The measures of aggression in-clude self-reports in interviews, questionnaires, peer nomina-tions, parent reports, and observations.Extensive evidenceex-

ists for the validity of peer nominations (Huesmann, Eron,Lef-kowitz, &Walder, 1984).Self-reports from interviews, such asthose conducted by Belson(1978), alsohave good evidence forvalidity, but self-reports onquestionnaires are subject to morethreats to validity (Elliot &Ageton, 1980;Olweus, 1979).

Themeasures incorrelational studies appear better suited toassess long-term learned patterns of behavior,attitudes,andval-ues than tomeasuring change over time.Peer nominations, forexample, are based on children's long-term experience witheach other and on reputations that may not alter easily even ifbehavior changes. In a quantitative review of the outcomes ofpsychotherapywith children, peer ratings ofbehavior changedconsiderably less than indices based on direct observation, ther-apist and parent ratings, or subject performance (Casey & Ber-man, 1985). Self-reports include behaviors exhibited over sev-eral months or even years. Most of the measures incorrelationalstudies, including longitudinal studies designed to measurechange, appear to assess trait rather than stateand may not besensitive tochangesinbehavior over relatively short periods.

ConclusionConvergence

The weight of theevidence from different methods of investi-gation supports the hypothesis that television violence affectsaggression. Virtually all reviewers agree that laboratory studiesof children and adults demonstrate positive findings and thatfieldsurveys produce modest but consistently positive corre-lations. Our review of the field experiments and longitudinalcausal analyses disputes Freedman'snegative conclusions. Thefieldexperiments produced mixed results, but onbalance, theyare positive. Thelongitudinal causal analyses indicate small butconsistent relations of television viewing to later aggression.The convergence argument is valid only if shared biases arenot operating in the same direction across laboratory experi-ments, fieldexperiments, and correlational studies.Our reviewsuggests that both positive and negative potential biases existwithin as well as between methods. Laboratory experimentshave high internal validity. The assertions that they are posi-tivelybiased by the selection of unusually violent materials, ex-perimenter demand, and artificial or unusual measures of ag-gression are not supported by acareful examination of the liter-ature. One can make an equally good case that they arenegativelybiased by selection of benign and brief stimuli, exper-imenter-produced inhibition of aggression, and measures thatfail to sample existing aggressive tendencies.

Infieldexperiments,thebias appears more likelyto benega-tive than positive because the treatment is relatively weak incomparison tonaturally occurring television viewing,but mustinfluence naturally occurring behavior. There is considerabledanger of negative reactions to control treatments and to theintrusion of anexperimental manipulation intoan ongoingso-cialsetting,and aggression may be underestimated by adult ob-servation.

2 The analysisprocedure was the stable correlates technique, a statis-tical procedure that is analogous to multiple regression with fewer as-sumptions(Belson, 1978).


6/8

TELEVISION AND AGGRESSION 369In correlational studies, selection bias is the major potential

threat to validity.Severalcareful efforts havebeen made to testthe contribution of awide rangeofdemographic, intellectual,personality, and social attributes to the correlation ofviolenceviewing with aggression. In all cases, the correlation remainsrobust. It is reasonable to conclude that selection bias does notexplain the correlation between natural viewing of violence andaggressive behavior.

Finally, longitudinal causal analysesmayalso underestimateeffects because themethodsand measuresof aggression arebet-ter suited to detecting stability of behavior than to detectingchange.

Generalization to Socially ImportantForms of Aggression

The forms of aggression measured in studiesof televisiongowell beyond "boisterousness" and "incivility" (Cook et al.,1983, p. 193); theyhaveclear relevanceto serious forms of ag-gression directly.Forexample, the categoryofaggression mostclearly related to television violence in Belson's (1978) studywas"seriousacts of aggression"; it included firing a revolver atsomeone, attackswith a knife, setting fire to a building, hittingsomeone in the face with a broken bottle, and knocking some-oneoffabike. Several other studiesofadolescents include delin-quent and criminal actions. Most correlational and observa-tional studies focus on physical aggressionhitting, kicking,biting, threatening, andother forms ofdirectphysical attack.

In addition, physicaland verbal childhood aggression of thetypemeasured bypeer nominations predicts adult physicalag-gression, criminal behavior, and serious crimes (Farrington,1979; Huesmann, Eron, Lefkowitz, &Walder, 1984; Olweus,1979;Parke& Slaby, 1983).If aconstant diet oftelevision vio-lence is causally related to childhood aggression, then it mayindeed contribute to the violence that concerns the society atlarge.

Moreover, criminal behavior is not the only socially impor-tant form ofviolence.Familyviolenceis amajor social problemaffecting many more people than criminal activity.In onefol-low-up,childhood aggression (measured by peer nominations),predicted the amount of physical punishment that men andwomen used with their children and the amount of spouseabuse indulged in by men (Huesmann, Eron, Lefkowitz, &Walder, 1984).

Theoretical ImplicationsThe social issues surrounding television violence sometimes

obscure careful theoretical analysis. The theories guiding manyearlyempirical investigations were social learningand arousaltheories, both ofwhichled to the prediction that viewing vio-lence would increase aggression.Theexplanatorypower ofboththeorieswaslimitedbysimple unidirectional modelsofcausal-ity. Freedman's (1984) critique of the literaturewasbased onthe assumption of a unidirectional model. After rejecting thehypothesis that television violence causes aggression, he ac-ceptedbydefault the reverse hypothesis, that aggressive person-alityattributes causeapreference for viewing violence (Freed-man, 1984, p.244).

Recent social learning theories articulate the reciprocaleffects ofenvironmental variables and qualities of the individ-ual (e.g., Bandura, 1978; Mischel, 1979). The theory and re-search supporting a bidirectional model of the relations be-tweentelevision violence and aggression is consistent with thismore sophisticated conceptualization (Stein & Friedrich,1975). Independent assessment of each direction of causalitysupports the prediction that both are important.

More complete models might specify interacting variables(e.g., Berkowitz, 1984). It is reasonable to expect that the rela-tion between viewing and aggression varies with individualviewerattributes (e.g., age,gender,cognitions about aggression,and aggressive predisposition), program attributes (e.g., pro-duction techniques, plot context), and environmental variables(e.g., cues for aggression, amount of adult-imposed structure,and existing patterns of social interchanges among peers).

A complete understanding of television effects is not ad-vanced by interpreting such variations asevidence of inconsis-tent results. For example, Freedman(1984)saw no "convincinga priorireason"whyviewingviolence shouldpredictaggressionfor male subjects but not female subjects in the research byEron et al. (1972) and concluded that their result was "weak-ened considerably by the lack of effect for girls" (p. 241).Yet,gender differences in aggression from early childhood on arewelldocumented (Parke& Slaby, 1983), and theyare magnifiedin adolescence. The incidence of violent offenses increasesabruptly forboysafter puberty;at age 18, thereisa 10:1 differ-encein themalerfemalearrest rateformurder (Cairns&Cairns,1985).

Process. The processes relating television violence to ag-gression can be divided into immediate, short-term effects onbehavior (e.g., activation, arousal, disinhibition, and behaviorcontagion) and long-termeffects involving learning ofhabitualbehaviorpatterns, attitudes, andvaluesabout aggression. Labo-ratory and field experimentsshow the short-term effects moreclearly than they do the long-term learnedpatterns. Of course,onewould probably not predict long-termeffects from asinglelaboratory exposure or a brief field experiment, given subjects'widetelevision experiencebeforeand after anexperiment.

Even ifmuchof the effect of television violence were limitedto short-term activation or arousal (and we do not believethatit is), such an effect isimportant. Aggressiveactionscommittedimmediatelyafter viewingcan bejust asharmful as those insti-gated byhabitual patterns ofbehavior.Given the frequency ofviewingviolent televisionand the millionsofchildrenand ado-lescentswhowatchiteveryday,evenshort-termeffects cancon-stitute a major social problem.Correlational data, including multivariate investigations,demonstrate long-term patterns of learned behavior,attitudes,and values.Iftelevision violence produces repeated instigationor disinhibition of aggression, then children exposed to it fre-quentlywillprobablyengageinaggressionoften. The daily rep-etition of simple, direct, violent problem solutions on televisionmay function tomaintainhighlevelsof aggression and tocoun-teract societal demands for more mature, self-controlled be-havior.

SocialPolicy ImplicationsThe debate about television effects cannot bedivorced from

social and political contexts. For more than 30 years the princi-


7/8

370 LYNETTE FRIEDRICH-COFER AND ALETHA C. HUSTON

pal playerspoliticians, broadcast industry executives, refor-mers, andacademic researchershavebeen lockedin alternat-ing positions ofdisputeand accommodation (Rowland, 1983).Social science research was elevated to a central role in argu-ments directed toward establishing a relation between violenttelevision and aggressive behavior. The networks questioned so-cial scientists' conclusions and conducted their own research.Themajor thrust was to press forproofof unidirectional, directcausality.

Responsible psychologists and other social scientists havebeen obliged to acknowledge the limits of their theoretical andresearch paradigms. Social science research alonewill not settlethe policy questions about television fare because current con-ceptualizations and methods cannot provide irrefutable causalproof. We can at best argue probabilities.Theweight of theoryand convergent evidence supports the likelihoodthat televisioncontributes toaggressionformany youngpeople.

The policy questions are becoming more urgent and morecomplicated with the increased availability of cable moviechannels and videotape rentals that contain much more explicitviolence than is shown onbroadcast television. Perhaps socialscientists need to acknowledgethatwhiletheir contributions tothe policy debate are important, the dialogue must be widenedto include consideration of societal values. The Federal Com-munications Act states that programs on the public airwavesare to serve the "public interest, convenience, and necessity."The question is not merelyout of fallible empirical methods,but what uses and consequent effects of the media are in thepublic interest.

ReferencesAmerican Psychological Association. (1985). Violence on TV. A social

issue release rom the Board of Social and EthicalResponsibility forPsychology. Washington, DC: Author.

Andison, F. S.(1977). TV violence and viewer aggression: A cumulationof study results: 1956-1976.Public Opinion Quarterly, 41, 314-331.

Bandura, A. (1978). The self system in reciprocal determinism. Ameri-can Psychologist. 33, 344-358.

Bandura, A., & Walters, R. (1963). Social learning andpersonality de-velopment. NewYork: Holt, Rinehart &Winston.

Belson, W.(1978). Television violence and the adolescent boy. Hamp-shire, England: Saxon House.

Berkowitz, L. (1984). Some effects of thoughts on anti- and prosocialinfluences of media events: A cognitive neoassociation analysis. Psy-chologicalBulletin, 95,410-427.

Berkowitz, L.,&Donnerstein, E.(1982). External validity is more thanskin deep: Some answers to criticisms of laboratory experiments.American Psychologist, 37, 245-257.

Cairns, R., &Cairns, B. (1985).The developmental-interactional viewofsocial behavior: Four issues ofadolescentaggression. In D. Olweus,J. Block, & M. Radke-Yarrow(Eds.), Development of antisocial andprosocial behavior: Theories, researchand issues (pp. 315-342).New^brk: Academic Press.

Cairns, R. B., & Cairns, B. D. (in press). On social values and socialdevelopment: Gender and aggression. In L. K. Friedrich-Cofer (Ed.)Human natureandpublic policy: Scientific viewsof women, children,and families. New\brk:Praeger.

Casey, R. J. & Herman, J. S. (1985). The outcomes of psychotherapywith children. Psychological Bulletin, 98,388-400.

Chaflee, S. H., & McLeod, J. M. (1972). Adolescent television use inthe family context. In G. A. Comstock & E. A. Rubinstein (Eds.),

Television and social behavior: Vol.3. Television and adolescent ag-gressiveness (pp. 149-172). Washington, DC: U.S. GovernmentPrinting Office.

Collins, W. A. (1973). Effect of temporal separation between motiva-tion, aggression, and consequence: Adevelopmental study. Develop-mental Psychology, 8, 215-221.

Collins, W.A., &Getz, S. K. (1976). Children's social responses follow-ing modeled reactions to pronocation: Prosocialeffects of a televisiondrama. Journal of Personality. 44,488-500.

Cook, T. D., Kendziersky, D. A., & Thomas, S. V. (1983). The implicitassumptions of television: An analysis of the 1982 NIMH Report onTelevision andBehavior. PublicOpinion Quarterly, 47, 161-201.

Elliot, D. S., & Ageton, S. S. (1980). Reconciling race and class differ-ences in self-reported and official estimates of delinquency.AmericanSociological Review, 45,95-110.

Ellis, G. T.,& Sekyra, F. (1972). The effect ofaggressive cartoons on thebehavior offirst grade children.Journal of Psychology, 81, 37-43.

Eron, L. D., Huesmann, L. R., Lefkowitz, M. M., & Walder, L. O.(1972). Does television violence cause aggression?American Psychol-ogist, 27, 253-263.

Farrington, D. P. (1979). Longitudinal research on crime and delin-quency. In N. Morris & M. Tondry (Eds.), Crimeandjustice:An an-nual review of research (Vol. 1, pp. 289-348). Chicago: University ofChicago Press.

Feshbach, S., &Singer, R. (1971). Television and aggression. San Fran-cisco: Jossey-Bass.

Freedman, J. L. (1984). Effect of television violence on aggressiveness.Psychological Bulletin,96, 227-246.

Friedrich, L.K., &Stein, A. H. (1973). Aggressive and prosocialtelevi-sion programs and the naturalbehavior ofpreschool children.Mono-graphs of the Society for Research in Child Development, 38(4, SerialNo. 151).

Huesmann, L. R. (1982). Television violence and aggressive behavior.In D. Pearl, L. Bouthilet, & J. Lazar (Eds.), Television and behavior:Ten years of scientific progress and implications for the eighties:Vol.2. Technical reviews (pp. 220-256).Washington, DC: NationalInsti-tute ofMental Health.

Huesmann, L. R., Eron, L. D., Lefkowitz, M. M., & Walder, L. O.(1984). Stability of aggression over time and generations. Develop-mental Psychology, 20, 1120-1134.

Huesmann, L. R., Lagerspetz, K., & Eron, L. D. (1984). Interveningvariables in the TV violence-aggression relation: Evidence from twocountries. Developmental Psychology, 20, 746-775.

Johnston, A., DeLuca, D., Murtaugh, K., &Diener, E. (1977). Valida-tionof a laboratoryplay measure ofchildaggression. Child Develop-ment, 48, 324-327.

Lefkowitz, M. M., Eron, L. D., Walder, L. Q, & Huesmann, L. R.(1972). Television violence and child aggression: A follow-up study.In G. A. Comstock &E. A. Rubinstein (Eds.), Television and socialbehavior: Vol.3. Television and adolescent aggressiveness (pp. 35-135). Washington, DC: U.S. Government Printing Office.

Leifer, A.D.,& Roberts, D. F.(1972). Children's responses to televisionviolence. In J. P.Murray, E. A. Rubinstein, &G. A. Comstock (Eds.),Television and socialbehavior: Vol. 2. Television and sociallearning(pp. 43-180).Washington,DC:U.S. Government Printing Office.

Leyens, J. P., Parke, R. D., Camino, L., & Berkowitz, L. (1975). Effectsof movie violence on aggression in a field setting as a function ofgroup dominance and cohesion. Journal of Personality and SocialPsychology, 32,346-360.

Liebert,R. M., &Baron, R. A. (1972).Some immediateeffects of tele-vised violence on children's behavior. Developmental Psychology, 6,469-475.

Liebert, R. M., Sobol, M. P., & Davidson, E. S. (1972). Catharsis ofaggression among institutionalized boys: Fact or artifact? In G. A.


8/8

TELEVISION AND AGGRESSION 371Comstock. E. A. Rubinstein, & J. P. Murray (Eds.), Television andsocial behavior: Vol. 5. Television effects, further explorations (pp.351-358).Washington,DC: National Institute ofMental Health.

Loye, D., Gorney, R., & Steele, G. (1977). An experimental field study.Journal of Communication, 27, 206-216.

Mclntyre, J. J., &Teevan, J. J. (1972). Television violence and deviantbehavior. In G. A. Comstock & E. A. Rubinstein (Eds.), Televisionand socialbehavior: Vol.3. Television and adolescent aggressiveness(pp. 383-435).Washington,DC: U.S. Government Printing Office.

Milavsky, J. R., Stipp, H. H., Kessler, R. C, & Rubens, W. S.(1982).Television andaggression:Apanelstudy. New York: Academic Press.

Mischel, W. (1979). On the interface of cognition and personality: Be-yond the person-situation debate.AmericanPsychologist, 34, 740-754.

Olweus, D. (1979).The stability of aggressive reaction patterns in hu-man males: A review.Psychological Bulletin,86,852-875.

Parke, R. D., Berkowitz, L., Leyens, J. P., West, S. G., & Sebastian,R. J. (1977). Some effects of violent and nonviolent movies on thebehavior of juvenile delinquents. In L. Berkowitz(Ed.), Advances inexperimental social psychology (Vol. 10, pp. 135-172). New York:Academic Press.

Parke, R. D., & Slaby, R. G. (1983). The development of aggression.In P. H. Mussen & E. M. Hetherington (Eds.), Handbook of childpsychology: Vol. 4.Socialization, personality, and social development(pp. 547-641).New York:Wiley.

Patterson,G. R. (1976). The aggressive child: Victim and architect of acoercive system. In L. A. Hamerlynck. E. J. Mash, & L. C. Handy(Eds.), Behavior modification and families: Vol. 1. Theory and re-search (pp. 1-44).NewYork: Brunner &Mazell.

Pearl, D.,Bouthilet, L.,& Lazar, J. (Eds.). (1982). Television andbehav-ior: Ten years of scientific progress and implications for the eighties.(Vols. 1 and 2). Washington, DC: U.S. Government Printing Office.

Ross, L. B. (1972). The effect of aggressive cartoonson the group playof children. Unpublished doctoral dissertation, Miami University,Oxford, OH.

Rowland, W. D., Jr. (1983). The politicsof TV violence: Policy uses ofcommunication research. Beverly Hills, CA: Sage.

Signorielli, N., Gross, L., &Morgan, M. (1982). Violence in televisionprograms: Ten years later. In D. Pearl, L. Bouthilet, & J. Lazar (Eds.),Television and behavior: Ten years of scientific progress and implica-tions or the eighties: Vol.2. Technical reviews (pp. 158-173). Wash-ington, DC: U.S. Government Printing Office.

Singer,J. L., & Singer, D. G. (1981). Television, imaginationand aggres-sion: A study of preschoolers.Hillsdale, NJ: Erlbaum.

Singer, J. L., Singer, D. G., & Rapaczynski, W. S. (1984). Family pat-terns and television viewingaspredictorsofchildren's beliefs and ag-gression. Journal of Communication, 34, 73-89.

Stein, A. H.,&Friedrich, L. K. (1975). Impact of television on childrenandyouth. In E. M. Hetherington (Ed.), Review of child developmentresearch(Vol. 5, pp. 183-256).Chicago: University of Chicago Press.

Steuer, F.B., Applefield, J. M., & Smith, R. (1971). Televised aggressionand the interpersonal aggression of preschool children. Journal of Ex-perimental ChildPsychology, 11, 442-447.

Surgeon General'sScientificAdvisory CommitteeonTelevision and So-cial Behavior. (1972). Television and growing up: The impact of tele-vised violence. Washington, DC: U.S. Government Printing Office.

Wells, W. D. (1973).Television and aggression: Replication of an exper-imental field study. Unpublished manuscript, University of ChicagoGraduate School ofBusiness.

ReceivedAugust 1,1985Revision received December 20,1985

Change in Distributionof APAConvention "CallforPrograms"In an effort to facilitatedistributionof the APA"Call forPrograms" for theannualconvention,the "Call" for the 1987 convention will appear in the December issue of the APA Monitorinsteadof being aseparatemailingto APAmembers.The 1987 conventionwillbe in New\brkfrom August 28 toSeptember 1.Deadlinefor submissionofprogramand presentation propos-als is January 20, 1987. Additional copies of the"Call"will be available from the APA Conven-tionOffice inDecember.

television violence and aggression

Documents