Intervening Variables in the TV Violence-Aggression ... · PDF fileIntervening Variables in the TV Violence-Aggression Relation: ... posure to TV violence increases a child's ag-

Developmental Psychology1984, Vd 20, No 5. 746-775

Copyright 1984 by theAmerican Psychological Association, lnc

Intervening Variables in the TV Violence-Aggression Relation:Evidence From Two Countries

L. Rowell HuesmannUniversity of Illinois at Chicago

Kirsti LagerspetzAbo Akademi, Turku, Finland

Leonard D. EronUniversity of Illinois at Chicago

Samples of 758 children in the United States and 220 children in Finland wereinterviewed and tested in each of 3 years in an overlapping longitudinal designcovering Grades 1 to 5. For girls in the United States and boys in both countries,TV violence viewing was significantly related to concurrent aggression and signif-icantly predicted future changes in aggression. The strength of the relation dependedas much on the frequency with which violence was viewed as on the extent of theviolence. For boys the effect was exacerbated by the degree to which the boyidentified with TV characters. Path analyses suggested a bidirectional causal effectin which violence viewing engenders aggression, and aggression engenders violenceviewing. No evidence was found that those children predisposed to aggression orthose with aggressive parents are affected more by TV violence. However, a numberof other variables were found to be correlates of aggression and violence viewing.The most plausible model to explain these findings seems to be a multiprocessmodel in which violence viewing and aggression affect each other and, in turn, arestimulated by related variables. Observational learning undoubtedly plays a role,but its role may be no more important than the attitude changes that TV violenceproduces, the justification for aggressive behavior that TV violence provides, orthe cues for aggressive problem solving that it furnishes.

Antisocial aggressive behavior, like manyother behaviors, appears to be overdetermined.Many factors—hereditary, paranatal, hor-monal, environmental, familial, and cognitiveones—probably influence how aggressively achild behaves. Severe antisocial, aggressive be-havior seems to occur most often when there

This research was supported in part by Grants MH-28280 and MH-3I886 from the National Institute ofMental Health to L. Rowell Huesmann, and by grantsfrom the Council of Social Sciences, Academy of Finland(Valtion Yhteiskuntatieteelhnen Toimikunta, SuomenAkatemia) to Kirsti Lagerspetz.

The authors wish to acknowledge the assistance of PatBrice, Paulette Fischer, Rosemary Klein, and Vappu Vie-mero, who contributed greatly to the operation of theproject and to the analyses of data. Also assisting in thecollection of the data were Birgitta Hagg, Raija-LeenaHolmberg, Patricia Jones, Esther Kaplan-Shain, LiisaKarkkainen, Bo Lillkall, Rebecca Mermelstein, SusanMoloney, Sharon Morikawa, Vita Musonis, Erica Rosen-feld, James Stewart, Ann Washington, and Linda White.

Requests for reprints should be sent to L. Rowell Hues-mann, Department of Psychology, University of Illinoisat Chicago, Box 4348, Chicago, Illinois 60680.

is a convergence of a number of these factorsduring a child's development (Eron, 1982), butno single factor by itself seems capable of ex-plaining more than a small portion of the in-dividual variation in aggression. For example,in a 10-year study of 427 children in NewYork State (Eron, Huesmann, Lefkowitz, &Walder, 1972; Lefkowitz, Eron, Walder, &Huesmann, 1977), no single variable corre-lated more highly than about .40 with aggres-sion (16% of the variance). Yet aggressive be-havior is reasonably stable over time and canbe reliably measured (Olweus, 1979). Thetechnique used to measure aggression in thecurrent study and the Eron et al. 10-year studyyielded an internal consistency of .97, a 1-month stability of .91, and a 10-year stabilityof .38. In Finland, similar findings over a 10-year period have been reported by Pitkanen-Pulkkinen (1981).

One environmental factor to which partic-ular attention has been directed is televisionviolence. In most of the developed countries,the mass media, especially television, have be-

746

TV AND AGGRESSION 747

come socializes that compete in importancewith home, school, and the neighborhood. Be-cause violence appears on television to a muchlarger extent than in real life, the TV provideseven more opportunities to experience vio-lence than does real life. Evidence from bothlaboratory and field studies has led most re-viewers to conclude that aggressiveness andviewing violence are interdependent to somedegree (Andison, 1977; Chaffee, 1972; Corn-stock, 1980; Eysenck & Nias, 1978; Hearold,1979; Huesmann, 1982a; & Lefkowitz &Huesmann, 1980). More aggressive childrenwatch more violent television. The relation isnot strong by the standards used in the mea-surement of intellectual abilities, but the re-lation is highly significant statistically and issubstantial by the usual standards of person-ality measurement with children. More im-portant, the relation is highly replicable.

Although the relation between violenceviewing and aggression is clear, the reason forthe relation is not—nor has it been establishedthat the relation goes in only one direction.Too often researchers have viewed the causalrelation between violence viewing and aggres-sion as an either/or proposition when, in fact,the causal effect is probably bidirectional.

Does TV Violence EngenderAggressive Behavior?

There can be little doubt that in specificlaboratory settings exposing children to violentbehavior on film or TV increases the likelihoodthat they will behave aggressively immediatelyafterward. Large numbers of laboratory studieshave demonstrated this effect both before andafter the 1972 Surgeon General's report ap-peared (Comstock, 1980). This effect has usu-ally been attributed to observational learning(Bandura, 1977; Bandura, Ross, & Ross, 1961,1963a, 1963b), in which children imitate thebehaviors of the models they observe. Just asthey learn cognitive and social skills fromwatching parents, siblings, and peers, they canlearn to behave aggressively from watching vi-olent actors. Similarly, children can learn tobe less aggressive by watching prosocial mod-els. For example, in Finland, Pitkanen-Pulk-kinen (1979) showed that the aggressive be-havior of 8-year-old boys could be reduced bywatching films depicting constructive solutions

to conflicts that appear frequently in children'severyday life. The question is whether the pos-itive correlations between violence viewing andreal aggressive behavior found in the field existbecause children are imitating the violent be-haviors they see on TV, or because of one ofthe other processes previously mentioned orfor some other reason.

The Lefkowitz, Eron, Walder, & Huesmannstudy (1977; Eron et al., 1972) provided thefirst substantial evidence from a field settingthat implicated television violence as a causeof aggressive behavior. Without rehashing tiredarguments, the results suggested that excessiveviolence viewing increases the likelihood thata child will behave aggressively. Althoughmany researchers have appropriate reserva-tions about the analyses used to extract causalinferences from these longitudinal observa-tional data (Comstock, 1978; Kenny, 1972),the critiques advocating a complete rejectionof the results (e.g., Armour, 1975; Kaplan,1972) contained such serious errors of rea-soning that they have not had a major impact(Huesmann, Eron, Lefkowitz, & Walder, 1973,1979).

Since the Lefkowitz et al. (1977) study, anumber of other observational studies and fieldexperiments have suggested that violenceviewing is indeed a precursor of aggression(Belson, 1978;Gransberg&Steinbring, 1980;Hennigan et al., 1982; Leyens, Parke, Camino,& Berkowitz, 1975; Loye, Gorney, & Steele,1977; Parke, Berkowitz, Leyens, West, & Se-bastian, 1977; Singer & Singer, 1981; Stein &Friedrich, 1972; Williams, 1978). The one re-cent longitudinal study (Milavsky, Kessler,Stipp, & Rubens, 1982) that led its authors toa conclusion of no causal effect has been in-terpreted by others as suggesting a weak pos-itive effect similar to that found in the abovestudies and in the current study (Cook, Kend-zierski, & Thomas, in press). Even if one con-siders Milavsky et al.'s results to be negative,it still seems reasonable to conclude that ex-posure to TV violence increases a child's ag-gressiveness under many conditions. The ma-jor problem with such a conclusion is that theboundary conditions under which the effectoccurs, as well as the essential precursor andintervening variables, and the processesthrough which they affect aggression are notwell specified.

748 L. HUESMANN, K. LAGERSPETZ, AND L. ERON

Culture, Age, and Sex asIntervening Variables

One can hardly consider the boundary con-ditions under which media violence might in-fluence a child's behavior without consideringthe child's culture, age, and sex. Although sub-stantial differences in sociocultural environ-ments exist among the children of any country,many cultural factors remain relatively con-stant. Therefore, to obtain reasonable variationof these sociocultural factors and to test thegeneralizability of any model, one would likea sample that includes subjects from severalcountries. Although a number of researchershave reported results from other countriescomparable to those from the United States(e.g., Belson, 1978; Granzberg & Steinbring,1980; Krebs & Groebel, 1977; Murray & Kip-pax, 1977; Williams, 1978), only a few havestudied the effects of television violence withcomparable methodologies in more than onecountry (e.g., Parke et al., 1977).

Similarly, to determine more precisely howthe relation between violence viewing andaggression develops with age, one must ex-amine multiple cohorts of children longitu-dinally during the periods when the relationis likely to be emerging. A number of re-searchers have attempted to determine the ageat which children are most susceptible to tele-vision violence. Eron et al. (1972) argued thatonce an individual has reached adolescence,behavioral predispositions and inhibitorycontrols have become crystalized to the extentthat a child's aggressive habits would be dif-ficult to change. Collins (1973; 1982; Collins,Berndt, & Hess, 1974; Newcomb & Collins,1979) consistently found that young childrenare less able to draw the relation between mo-tives and aggression and, therefore, may bemore prone to imitate inappropriate aggressivebehaviors. Hearold's (1979) review of the areagenerally supported these views, but suggestedthat the relation between violence viewing andaggression may increase again among adoles-cent boys. Perhaps the more important ques-tion, however, is at what age do children beginto be affected by behaviors viewed on TV.McCall, Parke, and Kavanaugh's (1977) ex-periments indicated that children as young as2 years were facile at imitating televised be-haviors, and some imitation was observed in

even younger children. Singer and Singer's(1981) study, mentioned earlier, provided ev-idence of adverse effects for television violenceon children who were only 3 or 4 years old.

An important finding in early field studiesof aggression and television violence was thatfemales were less affected by violence viewingthan were males (Bailyn, 1959; Eron, 1963).In our 10-year longitudinal study, conductedbetween 1960 and 1970, we found no corre-lation between a girl's violence viewing andher later aggressiveness (Eron et al., 1972). Theexplanations offered for the difference in re-sults for boys and girls have been speculativebut lead to some testable propositions. Onehypothesis would be that observational learn-ing was occurring for boys but not for girlsbecause there were no aggressive females por-trayed on TV. Because Bandura, Ross, andRoss (1963a, 1963b) and Hicks (1965) re-ported that both boys and girls more readilyimitated male TV actors than female actorsin laboratory studies, this explanation seemsunlikely. However, the hypothesis can be testedby evaluating observed programs for male andfemale violence because both types of aggres-sors now appear on TV. A second hypothesisfor sex differences is that females have beensocialized so strongly to be nonaggressive thatTV violence has little effect on them. Underthat theory one might expect girls from cul-tures in which female aggressiveness is moreacceptable to be more influenced by televisionviolence. Thus, an effect might be found nowin samples of American girls even though onewas not found in the 1960s. Third, sex dif-ferences might be due to differences in thecognitive processes of boys and girls. For ex-ample, under several of the explanatory modelstelevision violence would be more likely toelicit aggressive behavior if the violence de-picted were perceived as a realistic response.Because there is evidence that girls think tele-vision violence is less realistic than boys do(Lefkowitz et al., 1977), one would expectthem to be less affected than boys are underthis theory.

Design of the Field Study

In light of these considerations, a longitu-dinal, cross-cultural field study was undertakento determine the boundary conditions under


which the television violence/aggression re-lation obtains, to determine the relevant in-tervening variables, and to shed light on theprocess through which television violenceviewing relates to aggression. Samples fromsix countries—the United States (N = 758),Australia (N = 289), Finland (N = 220), Israel(Ar = 189), the Netherlands (N = 469), andPoland (N = 237) were included, but completedata presently are available only from theUnited States and Finland. Therefore, theconclusions reported in this article are basedon the data from these two countries only.

In Finland and in the United States twocohorts of children (first graders and thirdgraders) were interviewed and tested threetimes at 1-year intervals. Thus, the span ofGrades 1 to 5 was covered with an overlappinglongitudinal design. Most children's parentswere interviewed during the first wave, andsome were interviewed again during the finalwave. Because both cohorts of children wereinterviewed in Grade 3, developmental effectscan be separated from cohort effects.

Intervening Variables

The intervening variables selected for at-tention in this article were chosen for theirtheoretical relevance to the relation betweentelevision violence viewing and aggression.Three have already been mentioned: culture,age, and sex. However, a number of other cog-nitive, familial, and social variables were alsostudied.

Frequency of Viewing

One important mediating variable ob-viously would be the frequency with which achild watches television in general and the fre-quency with which he or she watches violentprograms in particular. A violent program thatis viewed only once in a while would not beexpected to have as much effect as a violentprogram viewed regularly. Two studies thatwere done in areas where television was re-cently introduced (Granzberg & Steinbring,1980; Williams, 1978), suggested that fre-quency of viewing was a crucial variable. Inthese and in a study by McCarthy, Langner,Gersten, Eisenberg, and Orzeck (1975),amount of television viewed appeared to be a

critical potentiating variable in elucidating therelation between violent television and ag-gressive behavior.

Sex-Typed Behaviors

The role of culture in socializing the sexesin regard to the appropriateness of aggressionhas already been mentioned. However, sex isnot always a complete measure of sex-typedbehavior. Girls who are aggressive might, forexample, be more comfortable with the ste-reotypical male role than with the female role.

Identification With TV Characters

Although the weight of evidence from lab-oratory studies (Bandura, 1963a; 1963b;Huesmann, 1982a) seemed to indicate that allviewers are most likely to imitate a heroic,white, male actor, individual differences shouldnot be ignored. It may be that some childrenidentify much more with some actors, and thisidentification mediates the relation betweenviolence viewing and aggressiveness. Such anidentification would be important not just inan observational learning model but also in amodel that emphasizes norms or standards ofbehavior. The more the child identifies withthe actors who are aggressors or victims, themore likely is the child to be influenced bythe scene, believing that the behaviors are ap-propriate and to be expected.

Fantasy

Fantasizing about aggression has seldombeen investigated as a mediator of a positiverelation between TV violence and aggression.Some theorists have argued that a child whoreacts to television violence by fantasizingabout aggressive acts might actually becomeless aggressive (Feshbach, 1964). However, noresearcher has ever reported finding such anegative correlation in a field study. In fact,a more compelling argument exists that fan-tasizing about aggressive acts should lead togreater aggression by the child. From an in-formation-processing perspective the rehearsalof specific aggressive acts observed on TVthrough daydreaming or imaginative playshould increase the probability that aggressiveacts will be performed.


Reality

Another potential mediating variable wouldbe a child's ability to discriminate betweenfantasy and reality as portrayed on television(Smythe, 1954). Violent scenes perceived asunrealistic by the child should be less likelyto affect the child's behavior according to sev-eral models, and some evidence for such anoutcome has been provided by Feshbach(1976).

Intelligence

Intelligence has often been postulated as athird variable that might account for the as-sociation between television-violence viewingand aggression; however, the evidence usuallyhas not supported such a thesis. In many ob-servational studies, researchers have measuredand partialed out the effect of intelligence butstill detected a significant relation between vi-olence viewing and aggression (Belson, 1978;Eron et al., 1972; Singer & Singer, 1981).

Popularity

Individual differences in popularity amongone's peers may also play a role in the aggres-sion/violence viewing equation. Previousstudies (Lefkowitz et al., 1977; Schramm, Lyle,& Parker, 1961) have shown that youngsterswho are unpopular or have poor social rela-tions spend more time watching television.Furthermore, in the Lefkowitz et al. 10-yearstudy, evidence was presented implicating lowpopularity as a cause of television viewing.

Parental Roles

Because the parents' social class, education,and aggressiveness might be correlated withthe child's aggressiveness, the role of parentsas mediating variables must also be examined.More interesting than any of these general so-cial variables, however, would be an exami-nation of the parent's television viewing habitsand attitudes and knowledge about the child'sviewing. According to a number of theories,the parent's beliefs about the realism of tele-vision violence might exacerbate or mitigatethe effects television has on the child. Fur-thermore, whether the parent watches whatthe child watches might be indicative of the

parent's concern and attitudes about televi-sion's effect. A number of experimental studieshave suggested that intervention by an adultmight mitigate the effect of violence on thechild (e.g., Hicks, 1968; Huesmann, Eron,Klein, Brice, & Fischer, 1983); however, noevidence of such an effect has been reportedin a field study. Similarly, one would also liketo test whether television-viewing habits aretransmitted from parent to child. There issome evidence that they are (Soneson, 1979;von Feilitzen, 1976), but the question remainswhether parents who watch more violence alsohave children who watch more violence.

Initial Aggression

Finally, one might want to consider initialaggression as a mediating variable as well. Ithas often been written that the effect of tele-vision violence is more notable among childrenpredisposed to aggression (Rubinstein, Coin-stock, & Murray, 1972). The theoreticalmeaning of such a statement is cloudy. If itmeans that severe aggression is an overdeter-mined behavior resulting from the convergenceof many factors, it is as true a statement abouttelevision violence as it would be about anyother cause of aggression. If it means that tele-vision violence only has a noticeable effect onhighly aggressive children, it requires empiricalvalidation. Some evidence is available fromlaboratory experiments both in favor of thepredisposition hypothesis, and contrary to it(Dorr & Kovaric, 1980; Kniveton & Stephen-son, 1973; Surgeon General's Advisory Com-mittee on Television and Social Behavior,1972). In a study by Lagerspetz and Engblom(1979), the aggressive play of aggressive chil-dren showed no increase after the viewing ofviolent film, whereas the aggression of sub-missive and cooperative children did show anincrease. The only available evidence from fieldstudies suggests that the positive relation be-tween violence viewing and aggression occursamong children at all levels of aggression(Huesmann et al., 1973). However, confir-mation of this finding in children from culturesdiffering in attitudes toward aggression wouldbe valuable.

In the next two sections we present datafrom the U.S. sample and contrast the resultswith the data from the Finnish sample. In the


final section we integrate the results into amodel of the relation between television vio-lence viewing and aggression.

United States Study

Subjects

The original group of subjects in the UnitedStates was composed of 672 children in thepublic schools of Oak Park, Illinois, an eco-nomically and socially heterogenous suburbof Chicago, and 86 children from two inner-city parochial schools in Chicago. The poolfrom which these subjects were drawn con-sisted of all the children in the first and thirdgrades of these schools in 1977. The parochialschools were added to increase the ethnic andsocioeconomic heterogensity of the sample,although Oak Park is by no means uniformlymiddle class. In 1977 it ranked 110th in me-dian family income ($19,820) among Chica-go's 201 largest suburbs. Minority enrollmentin the public schools ranged from 5% to 18%,with an average of 14%. Of the two parochialschools, one was composed of a predominantlylower-middle-class Hispanic population andthe other a predominantly lower-middle-classintegrated population. One of the Oak Parkschools was dropped from the study beforeany data were collected because the principalfelt the study would be too disruptive.

Having selected the schools, we compiledclass lists of all first- and third-grade childrenand solicited their parents' permission for themto participate in the study. Through repeatedwritten and personal contacts we raised thefinal permission rate to 76%.' Of the remainingchildren, 14.8% declined to participate,whereas 9.2% never responded. The responserates were similar in all of the schools. Thisprocedure yielded a final pool of 841 childrenfrom which samples could be selected. Eighty-three children were used for pilot testing, leav-ing a sample of 758 for the study.

Of the 758 subjects, who thus constitutedthe United States sample at the start of thestudy in 1977, 384 were girls (207 first gradersand 177 third graders) and 374 were boys (194first graders and 180 third graders). In 1978,for the second wave, 607 children remainedin the sample, and in 1979 for the third wave,there were 505 children. Almost all of the sub-

ject attrition resulted from children leavingthe school systems.

During the first and second waves of datacollection in the United States, parents of 591children were interviewed (177 both motherand father, 378 mother only, and 36 fatheronly for a total of 768 interviews). After thethird wave, one of the parents was reinter-viewed for 300 children (285 mother only, 15father only). An attempt was made to interviewat least one parent of" every child, and thosewho were not interviewed represent the morereluctant parents. Thus, the United States par-ent sample is somewhat biased and excludesless cooperative parents. The effect of this biasis examined in the Results section.

Procedure

The children were initially interviewed andtested in the United States during the springof 1977. They were retested during the springof 1978 and 1979. Thus, we obtained data onthe original first grade children in the first,second, and third grades, and on the originalthird grade children in the third, fourth, andfifth grades. This overlapping longitudinal de-sign has the advantage of permitting the testof causal and developmental theories morereadily. Parents were interviewed once duringthe first or second wave and, if possible, againduring the third wave.

Data were collected from four sources: thechild, the child's peers, the school, and theparents. From each child we collected data onhis or her use of fantasy, preferred sex-typedbehaviors, self-rating of aggression, favoritetelevision programs, frequency of viewing fa-vorite programs, identifications with TV char-acters, and judgment of how realistic violentTV programs were. From each child's class-mates we obtained a peer rating of popularity

1 The response rate to our original letter sent in themail was 55% Telephone follow-ups raised the rate onlyto about 65% Then, a few days before testing was to begin,we gave a prize to each child for whom a letter had beenreturned, regardless of the decision on the letter. The otherchildren were given another copy of the permission letterto take home and were told that they would receive thesame prize if they returned the letter the next day, nomatter how it was signed. This procedure raised our finalresponse rate to about 91% and our permission rate to76%.


and aggressiveness by the peer-nominationprocedure used in previous studies (Eron,Walder, & Lefkowitz, 1971). In the parent in-terviews we collected data on demographiccharacteristics of the family; child-rearingpractices; and parents' personalities, attitudes,and TV preferences. Most of these parentscales had been used in our previous studies.The variables measured included the parents'education and occupations, the parents' ag-gressiveness as measured by the sum of scales4 and 9 on the Minnesota Multiphasic Per-sonality Inventory (MMPI), hours of TVviewing and four favorite programs (giving aTV violence score), the parents' own viewingof the child's favorite programs, judgment ofhow realistic TV violence is, and the parents'estimates of the child's viewing. The finalsource of data was school records. Data werecollected on the older children's standardizedachievement test performance during the firstwave. The only major change in procedureover the three waves of child interviews oc-curred between first and second grade. Eachfirst-grade child received the peer-nominationprocedure and the television viewing questionsin an individual 30-min session. The othermeasures were administered in group sessions.For the second and higher grades, however,both sessions were group sessions. The indi-vidual sessions for first graders were necessi-tated by their low reading level.

Each group session was conducted by atleast two experimenters and contained nomore than 25 children. The teacher and chil-dren who did not participate in the study werenot present except during peer nomination insplit-grade classes. In these classes childrenfrom the other grades were also used as ratersto obtain enough nominators for valid scores.Each page of the interview booklet was a dif-ferent color, so the monitors could be surechildren were on the correct page. Each answerblank was denoted by a different picture, sothe children could readily keep their place.Each group session required about 40 min.

The parent interviews were conducted in-dividually by research assistants. Some wereconducted in a field office and some in theparents' home. In either case, only the parentand interviewer were present in the room whilethe questions were being asked. For the mostpart, the interviewer read the questions and

recorded the parent's answer. However, theMMPI and a few other sensitive test questionswere placed at the end of the booklet. Theparent read those questions and marked theanswers while the interviewer waited. Parentswho traveled to the field office for interviewswere paid their travel expenses.

Child Measures

Peer-nominated aggression and popularitywere measured by a slightly modified versionof the Peer-Rating Index of Aggression (Walder,Abelson, Eron, Banta, & Laulicht, 1961): Eachsubject in the sample names all of the studentsin his class who have displayed 10 specific ag-gressive behaviors during the school year. Re-liability as well as concurrent, predictive, andconstruct validity of this measure have beenamply demonstrated (Eron, 1980; Lefkowitzet al., 1977). A slight deviation from the pre-vious use of this technique was the instructionthat subjects were only to consider behaviorsthat had occurred during the current schoolyear. A more major modification was that inthe first-grade classes the procedure as notedabove was conducted individually. These chil-dren nominated their peers by naming orpointing to the appropriate child's picture ona class photograph containing individual pic-tures of each child in the class. Coefficientalpha in the current study calculated only onthe United States sample was .97 and test-retest reliability over one month was .91.

The child's self ratings of aggression werebased on four items in which the child ratedhis or her similarity to fictional children de-scribed as engaging in specific aggressive be-haviors (e.g., "Steven often gets angry andpunches other kids." "Are you just like Steven,a little bit like Steven, or not at all likeSteven?"). Coefficient alpha was .54. Self-rat-ings correlated .38 with peer ratings of aggres-sion in the United States and .39 in Finland.

Sex-typed behavior was indicated by pref-erences for sex-typed toys and games. Thismeasure of preference for sex-typed activities,based on Nadelman's (1974) procedure, madeup a booklet of four pages, each of which con-tained six pictures of children's activities. Twopictures of each set had been previously ratedas masculine, two as feminine, and two asneutral by 67 U.S. college students who had


been asked to designate the activities as popularfor boys and/or for girls. The task for the chil-dren was to select the two activities they likedbest on each page; they received a score forthe number of masculine, feminine, and neu-tral pictures they chose. The reason for in-cluding a neutral category was that, eventhough less masculine boys may not like tra-ditionally feminine activities, they might preferneutral over traditionally masculine ones.Similarly, for girls, we anticipated that thosewho did not prefer traditionally feminine ac-tivities might also eschew masculine activitiesbut would subscribe to neutral ones. Coeffi-cient alpha is not an appropriate measure ofreliability for this scale. However, one-monthtest-retest reliabilities ranged between .55 and.60. Only the masculine and neutral sex-typedpreferences are included in the present analysesinasmuch as preference for feminine activitiesdid not relate to aggression, either positivelyor negatively.

Aggressive fantasy and active-heroic fantasywere measured by two scales devised by Ro-senfeld, Huesmann, Eron, and Torney-Purta(1982). Each scale has six items, for example,"Do you sometimes have daydreams abouthitting or hurting somebody you don't like?"or "When you are daydreaming, do you thinkabout being the winner in a game you like toplay?". Coefficient alphas for these scales were.64 and .61, respectively, and one-month test-retest reliabilities, .44 and .62.

The children's regularity of TV viewing andTV violence viewing scores were derived fromtheir responses to several lists of programs. Oneach list they were asked to select the one theywatched most. For the program selected, theythen marked whether they watched it once ina while (I), a lot but not always (2), or everytime it's on (3). A child's regularity of TVviewing was computed as the sum of theseresponses for the selected programs. Thus, theregularity score is a measure of the child'sfrequency of viewing his or her favorite showsrather than of hours of viewing.

Violence in favorite television programs wasdetermined from the same program selections.Two psychology graduate students, who them-selves had small children, rated all programsfor the amount of visually portrayed physicalaggression on a 5-point scale from not violentto very violent. Each rater received a paragraph

giving detailed descriptions of what constitutedviolent behavior (e.g., graphically portrayedphysical assaults). Interrater reliability was .75.Previous analyses have shown (Lefkowitz etal., 1977) that this method yields ratings com-parable to those obtained from more objectivecontent analyses (e.g., SignoreUi, Gross, &Morgan, 1982). A child's TV violence scorewas the sum of the violence ratings of showsthat the child had indicated he or she watchedmost, weighted by the regularity with whichthe child reported watching the program.2

In the United States, eight lists of 10 pro-grams each were used to obtain TV frequencyand violence scores. The 80 shows used werethe most popular ones for children ages 6-11during the current year (1977, 1978, or 1979)according to Nielsen ratings. The programswere arranged into lists; each list had an equalnumber of violent programs. The violent andnonviolent programs on a list were equatedfor popularity. An attempt was made to equateviolent and nonviolent programs for time andday of week shown and for sex of the centralcharacter. The one-month test-retest reliabilitywas .75 for frequency and .76 for violenceviewing.

Realism of television programs was ratedby the children. They were given a list of vi-olent shows, including cartoons, and wereasked, "How true do you think these programsare in telling what life is really like—just likeit is in real life, a little like it is in real life, ornot at all like it is in real life?" The subject'stotal realism score was the sum of the ratingson the items. One-month test-retest reliabilitywas .74; coefficient alpha was .72.

Identification with TV characters was de-rived from children's ratings that indicated

2 Several different weightings for frequency were triedm the test-retest subsample A multiple regression analysispredicting aggression from the product of the frequencyand the violence rating indicated that the strongest relationcould be obtained if violence ratings, scaled from mostviolent (4) to nonviolent (0), were multiplied by frequency,scaled from watching every time its on (10) to watchingonly once in a while (0). This result was cross-validatedwith the enure first wave of U.S data separately for boys& girls. Therefore, for all waves these scalings were usedto compute the violence viewing scores in both countries.In other words, a program that is only viewed once in awhile does not contribute to a child's violence viewingscore, no matter how violent the program is.


how much they believed they acted like certaintelevision characters. These characters in-cluded two aggressive males, two aggressivefemales, two unaggressive males, and two un-aggressive females. For each character, thechildren were asked "How much do you actlike or do things like the character?" The sumof their responses constituted a reliable "iden-tification with aggressive character" score.Coefficient alpha for this measure was .71 andtest-retest reliability over one month was .60.

Parent Measures

The parents reported their own and theirspouses' education and occupation. The fa-ther's occupation was coded to obtain a social-class score for the family (Warner, Meeker, &Eels, 1960). Each parent's aggression wascomputed as the sum of Scales 4 and 9 of theMMPI, which has been shown to be a reliableand valid measure of antisocial behavior(Huesmann, Lefkowitz, & Eron, 1978). Theparent's TV violence viewing was calculatedas the sum of violence scores for the four pro-grams listed as favorites and weighted by thefrequency with which the parent reportedwatching them. The parent's judgment of TVrealism was derived in the same way as thechild's. A measure of how frequently the parentwatched the child's programs was obtained byasking the parents how often they watched thespecific programs their children had selectedas their favorites (without telling the parentthis). Finally, the parents estimated both theirown and their children's weekly hours of TVviewing.

Results

Figure 1 shows how peer-nominated aggres-sion changed for boys and for girls over thecourse of the 3 years. One can see that in theUnited States the occurrence of peer-nomi-nated aggressive behavior increased from firstto fifth grades as has been reported previously(Eron, Huesmann, Brice, Fischer, & Mermel-stein, 1983). To give this graph more substan-tive meaning, one should remember that anaggression score of .20 indicates that a childwas nominated on aggression questions 20%of all possible times.

Within the United States, the amount ofviolence shown on TV did not change to an

appreciable extent over the course of the 3years (Signorelli et al., 1982). Furthermore,during each year approximately the samenumber of violent shows (25%) was includedin the 80 programs from which the childrenmade their selections; thus, it is appropriateto compare TV violence viewing from year toyear. Figure 2 indicates (as reported previouslyby Eron et al., 1983) that TV violence viewingin the United States sample peaked in the thirdgrade. This suggests that the third-grade timeperiod would be particularly important in thetelevision violence-aggression association.

Violence-viewing scores are more difficultto interpret than are aggression scores. Becauseeach score represents the product of theamount of violence occurring on the selectedshows with the self-reported viewing regularity(i.e., frequency of viewing the shows), the samescore can have somewhat different meanings.For example, a score of 80 (the mean score

40

35

Q

ua.oo<oa)

Z

oza.aiaJa.

<LU

30

25

20

15

10

05 —

USA Boys

USA Girls

Jk FINN Boys

.-•FINN Girls

1 2 3 4 5GRADE

Figure 1. Cohort, grade, and sex differences in aggression.


for third-grade boys) could be obtained if thechild reported watching two of the most violentprograms every time they were on, or if thechild reported watching four of the moderatelyviolent shows every time they were on, or ifthe child reported watching eight of the slightlyviolent shows every time they were on. Nev-ertheless, no child in the United States scoredin the upper quartile on violence viewing un-less he or she watched at least four highly vi-olent shows regularly.

Because of the overlapping longitudinal de-sign, we could separate out the effects of sex,grade, and cohort on aggression and televisionviolence. A Sex X Cohort X Wave repeatedmeasures analysis of variance (ANOVA) con-firmed that boys were consistently more ag-gressive than were girls, f\l, 486) = 24.3, p <.001, and watched more television violencethan did girls, F{\, 420) = 26.1, p < .001. Inaddition, aggression increased developmentally

140

100

o2 60

FINLAND

v > • FINN Boys

___#- - - • FINN Girls

zHI

<LU

2 0

80

60

40

201

USA

1 2 3 4 5GRADE

Figure 2. Cohort, grade, and sex differences in TV violenceviewing.

with age, as indicated by the significant maineffects for both wave, F{2, 972) = 34.6, p <.001, and cohort, F(l, 486) = 22.83,p < .001.The main effect for cohort, however, cannotbe taken as evidence of a cohort effect, becauseit is confounded with age. Rather, one mustexamine the contrast between the two cohorts'aggression in the overlapping grade (i.e., thethird grade). This difference was not signifi-cant.

For both aggression and TV violence view-ing, significant Wave X Cohort interactionswere obtained, F(2, 972) = 20.43, p < .001,and F(2, 840) = 21.69, p < .001, respectively.The interactive effect on aggression of waveand cohort is at least partially a result of thechange in peer-nomination procedure betweenfirst and second grades (from pointing to pic-tures to crossing out names). However, the in-teractive effect of wave and cohort on TV vi-olence viewing reflects the obtained peak inthe third grade, as shown in Figure 2.

Because of these sex and grade differencesin aggression and TV viewing, most of theremaining analyses are conducted separatelyfor each subset of subjects or else grade andsex are used as covariates. In this way, spuriousrelations due to common age or sex effects ontwo variables are avoided. Overall correlationsthat are reported are generally averages of thesubsample correlations. This procedure usu-ally yields lower correlations than does usingthe entire sample together. For example, thecorrelation between TV violence and aggres-sion among all subjects in the first wave was.221, whereas the average of the correlationsfor the four grade-sex subsets in the first wavewas .185. The overall correlation is heightenedbecause both boys and older children act moreaggressively and watch more violence than dogirls and younger children.

Television Violence and Aggression

As Table 1 reveals, positive correlations werefound between overall TV violence viewingand peer-nominated aggression in all gradesin the United States. A child's overall TV vi-olence score represents how regularly the childis exposed to TV violence. Unlike our previousresults, which were significant only for boys(Eron et al., 1972), positive correlations wereobtained for both boys and girls. In fact, on


Table 1Correlations Between TV Violence Viewing and Peer-Nominated Aggression

Cohort

First grade

Third grade

Grade

123

345

U.S.

.16*

.20*

.15

.24**

.18"

.20*

N

188144114

16713199

Males

Finn

.03

.27*

.09

.04

.38*»

.28"

N

555450

494443

U.S

.22"

.25*

.28***

• 13t.26*'*29««

Females

N

201158124

168138107

Finn

.14

.02

.33**

.05

.16

.04

N

644844

514641

t p < .10. * p < .05. •• p < .01. ••* p < .001. All probabilities are two-tailed.

the average the correlations were somewhathigher for girls. Averaging data over the threewaves of measurements, one obtains an overallcorrelation of .29 (p < .001) for girls and .25(p < .001) for boys. Although it might betempting to attribute the correlations betweenself-reports of viewing violent programs andaggression to methodological or socioculturalartifacts, such explanations find little supportin the data. Response bias due to impulsivity,for example, was measured (Rosenfeld et al.,1982) and does not explain the relation. Fur-thermore, the validity of the TV-violence mea-sure and these results is supported by the find-ing that these children's scores, based on theirown choices from the program lists, correlatedsignificantly with an alternative violence scorebased on the parents' reports of the child'sfour favorite programs (r = .29, N = 572,p < .001).

The correlations between self-reported reg-ularity of viewing favorite shows (regardlessof violence) and peer-nominated aggressionwere also positive and significant for both boysand girls in the United States (average r =.226, 11 of 12 significant). In fact, a pure self-reported regularity score correlated as highlywith aggression as a violence-viewing score forsome grade-sex combinations in the UnitedStates. Of course, regularity in this contextmeans the frequency with which the childwatches his or her favorite programs, not thetotal numbers of hours watched per week. Asa matter of fart, the child's report of the fre-quency with which he or she watched his orher favorite program correlated only .20 (N =540, p < .001) in the United States with themother's report of the number of hours her

child watched TV each week, which averaged15 hours per week. The mother's report offrequency did not correlate significantly witheither boys' or girls' aggression. Thus, the re-lation between violence viewing and aggressionseems to depend strongly on the regularity ofexposure. Regular exposure to mild violenceis associated with higher aggression much morethan is infrequent exposure to severe violence.

Rather than use cross-lagged correlations,which have several drawbacks (Huesmann,1982b; Rogosa, 1980), we use multiple regres-sions to estimate causal effects. To interpretregression coefficients as causal coefficients,several conditions must be met. First, the cri-terion (endogenous) variable must not causeany of the predictor (exogenous) variables inthe proposed structural model. In the currentcase, this condition is satisfied by using laggedobservations on the exogenous variables, soone knows they cannot have been caused bythe endogenous variable. Note that this doesnot rule out the possibility that the endogenousvariable could be causing the exogenous vari-able's future values. Second, none of the un-measured causes of the endogenous (criterion)variable can be causes of an exogenous (pre-dictor) variable. This condition certainly is notsatisfied in these analyses, though later the ef-fect of adding various third variables into thecausal system is discussed. Until each suchthird variable is introduced, however, a plau-sible alternative hypothesis to a causal effectis a spurious relation due to the third variable.The third condition required is that the vari-ables must be corrected for their measurementerror (i.e., their unreliability, if inferencesabout causal relations among true scores are


to be made). We present both the correctedand the uncorrected regression coefficients, al-though most variables are so highly reliablethat the corrections make little difference.

Table 2 contains a multiple regression anal-ysis indicating the effect of television violenceon later aggression while controlling for earlieraggression. The analysis reveals a significantpositive effect of TV violence viewing on lateraggression for girls in the United States butonly a marginal effect for boys in the UnitedStates. The means for girls in Table 3 illustratethese results more clearly. No matter what theinitial level of girls' aggression, those whowatch more TV violence are likely to becomemore aggressive than are those who watch less.However, the data do not support a conclusionof unidirectional causality. Regressions pre-dicting TV violence viewing from aggressionrevealed that those girls who are more ag-gressive also are more likely to become heavierviolence viewers regardless of their initial levelof violence viewing. The bidirectionality of

the effect suggests that a simple observationallearning model is not sufficient to explain thecorrelation between violence viewing andaggression for girls.

These regressions also reveal that aggressionis a reasonably stable behavior for both boysand girls in the age range studied. The cor-responding 3-year longitudinal correlations foraggression were .72 for boys and .57 for girls.The differences in these and in the multiplecorrelations for boys and girls indicate thataggression is more predictable and stable forboys than for girls. Such a result is consistentwith other recent research (Huesmann, Eron,Lefkowitz, & Walder, in press; Olweus, 1979).On the other hand, TV violence viewing wasmuch less stable with longitudinal correlationsof .19 for boys and .31 for girls.

Sex of Viewer and of TV-Character Observed

As we have maintained, one possible ex-planation for differential effects of violence on

Table 2Multiple Regressions Relating Aggression to TV Violence Viewing, Controlling for Initial Levelof the Dependent Variable

Predictor

GradeFirst wave aggressionTV violence viewing

in first and secondwaves

GradeFirst wave TV

violence viewingAggression in first

and second waves

Note The numbers int p < . 1 0 . * p < . 0 5 . •

U.S. (N -•

.251"*

.526W

.135**

R2 = .407,FIX 217) =

Standardized regression coefficients

Girls

= 221)

Dependent

(.253)(-526)

(152)

= 49.7,

Finn (A' = 85) U.S. (A' =

variable: Third-wave aggression

-.097.649"*

.012

R2 = .390,7=13,81) =p< .001

(-.106)(.671)

(.012)

17.3,

132".695***

.079

R2 = .547,HX 187) =p< .001

Boys

-- 191)

(.132)(.714)

(.086)

•• 7 5 . 3 ,

Dependent variable: Third-wave TV violence viewing

- .216"*

.293*"

.257***

R2 = .191,FIX 217) =p< .001

(-.249)

(.298)

(.294)

= 17.1,

-.085

.411 —

.148

R2 = .190,F{3, 81) =/x .OOl

(-.146)

(.566)

(.158)

6.35,

-.364*'*

.190"*

.165'

R2 = .170,F\X 187) =p<.001

parentheses are the regression coefficients corrected for the i• p < .01. •*• p < .001.

(-.429)

(257)

(.178)

= 12.7,

Finn (N

-.172*.768*"

.110R2 = .575,F\X 76) =p< .001

-.067

.248*

.123

R2 = .069,F\X 76) =p < .15

= 80)

(-.187)(.792)

(.126)

34.2.

(-.117)

(.346)

(.147)

1.86,

unreliability of the measures.


Table 3U.S Girls Mean Peer-Nominated Aggression inthe Third Wave as a Function of Their EarlierTV Viewing and Aggression

Initial peer-nominated aggression

High

Medium

Low

Total

TV violencefirst two

Low

.284(A'=21)

.131(N = 63)

.096(A' = 37)

.147(A' = 121)

viewing forwaves

High

.334(N = 42)

.167(N = 56)

.115(N = 24)

.214(N = 122)

Note Standard deviations ranged from . 10 to. 18 and werecorrelated with means.

boys and on girls is that they respond differ-ently when observing male and female ag-gressors. To test the hypothesis that girls areaffected more by violent programs depictingfemale aggressors, we scored TV programsseparately for the amount of violence perpe-trated by males and by females. However, wefound that regardless of the sex of the viewerthere were higher correlations between thechild's aggressiveness and the child's viewingof a male character's violence than betweenthe child's aggressiveness and the child's view-ing of a female character's violence. In theUnited States, the average correlation of a girl'saggression with female aggressor violence was.168 (overall r = .23, p< .001), compared to.236 (overall r = .30, p < .001) with male ag-gressor violence. Thus, it does not seem rea-sonable to attribute the recent emergence ofa relation between violence viewing andaggression in girls in the United States to theappearance of female models on TV.

One of the problems with using gender asa measure of identification with a TV modelis that aggression is highly correlated with sex-role orientation (Eron & Huesmann, 1980;Lefkowitz et al., 1977). Girls who are aggres-sive may in fact identify more with male actorsthan with most female actors. More importantthan the sex of the model may be the behaviorsthe model is performing and, if masculine ac-tivities are intrinsically more appealing to somesubjects of either sex, then these subjects could

be more likely to attend to male charactersand be influenced most by their behaviors.

In the current study, we measured sex roleorientation by having children select the ac-tivities they preferred most from pictures rep-resenting stereotypic male, female, and neutralactivities. The results of ANOVAS with thesevariables did not support the hypothesis thata child's preference for stereotypic male, fe-male, or neutral activities interacted with TVviolence viewing to affect aggression. Nev-ertheless, when we divided girls into those whoscored high and those who scored low on pref-erence for stereotypic male activities, we founda greater correlation between violence viewingand aggression among the "high male" girls(r = .30, N = 110, p< .001) than among the"low male" girls (r = .15, N = 130, p < .09).Thus, the emergence of a positive correlationbetween violence viewing and aggression ingirls at this time as opposed to previous times(Eron, 1963) might be explained by a changein sex role norms occurring over this period.However, the most consistent significant find-ing about sex role and aggression was that apreference for neutral activities (non sex-typed)was associated with lower aggression in bothboys (average r = -.147, 3 of 6 significant;overall r = -.24, p < .001) and girls (averager = -.158, 3 of 6 significant; overall r = -.25,p< .001).

Identification With TV Characters

Gender and sex-typed activity preferencestap only one dimension of similarity betweena child and a TV character. More specific in-dividual differences were measured by askingthe child how much he or she acts like specificTV characters. As Table 4 shows, the extentto which a child thought he or she acted likeTV characters was significantly positively cor-related with the child's aggressiveness in mostage-gender groups, although the effect wasmore pronounced for boys in the United States(overall r = .22, p < .01) than for girls in theUnited States (overall r. ns). It is interestingthat the correlations are just about the samefor "identification with aggressive characters"and for "identification with all characters" (thetwo variables correlated .87 with each other).Children seem to see themselves either as act-ing like most TV characters or as being like


Table 4Correlations Between Identification with Aggressive TV Characters and Peer-Nominated Aggression

Males Females

Cohort

First grade

Third grade

Grade

!23345

U.S.

.15*

.18*

.08

.12

. 3 3 " '

.35***

N

18714011216413197

Finn

.06

.36"

.22

.25t

.28f

.18

N

545443484341

U.S.

.17*

.26**

.18*

.10

.09

.11

N

196157122165134105

Finn

.06

.16

.59***

.26t

.04

.21

N

504837484537

f p < .10. • p < .05. *• p < .01. ••• p < .001. All probabilities are two-tailed

no character. Children who identify most arealso those who watch the most TV violence(overall r = .37, p < .001 for boys; overall r =.32, p< .001 for girls).

Longitudinally, the correlation betweenidentification with TV characters during thefirst 2 years and aggression in the 3rd year ishighly significant for boys (r = .23, N = 191,

p < .001). Moreover, multiple regressions forboys revealed a significant longitudinal pre-diction to later aggression from earlier iden-tification. However, the most significant causalcoefficients were derived when the mulipli-cative product of violence viewing, and iden-tification was used to predict later aggression.This critical result is shown in Table 5.

Table 5lauic j

Multiple Regressions Relating Boys Aggression to a Multiplicative Composite of TV Violence Viewingand Identification with Aggressive TV Characters, Controlling for Initial Levelof the Dependent Variable

Standardized regression coefficients

Predictor U.S. Boys (N= 191) Finn Boys (N = 80)

Dependent variable. Third-wave aggression

GradeFirst wave aggressionProduct of TV violence viewing

and identification withaggressive characters in firstand second waves

.151*"

.679*"

.152*

(160)(.677)

(.199)

.174*7 4 7 * "

.205*

(-.202)(754)

(.271)

R2 = .563, FIX 187) = 80.3.p < . 0 0 1

R2 = .604, F\X 76) = 38.6,/ x .OOl

Dependent variable: Product of TV

GradeProduct of TV violence viewing

and identification withaggressive characters in firstwave

Aggression in first and secondwaves

violence viewing and identification with

-.236**»

.384*"

.102R2 = .232, /T[3, 187)

/ x .OOl

(-.267)

(.612)

(.059)= 18 9,

aggressive TV characters in third wave

-.080

.319*'

.178R2 = .1

(-.186)

(.528)

(.213)129, FIX 76) = 3.74,

p < .02

Note Numbers in parentheses are the itgnwon coefficients corrected for the unreliability of the measures.t/> < .10. *p< .05. "p< .01. ••• P < .001.

760 L. HUESMANN, K. LAGERSPETZ, AND L ERON

Table 6US and Finnish Boys Mean Peer-Nominated Aggression in the Third Wave as a Function ofTheir Earlier TV Viewing and Aggression

Initial peer-nominated aggression

High

Medium

Low

Total

Product

Low

.419(N = 25)

.195(N = 38)

.098(W = 46)

.206(N = 109)

of TV violence viewing and identificationcharacters for first two waves

U.S. boys

High

427(A7 = 49)

.214(N = 35)

.148(N = 27)

.292(A'= 111)

Low

.158(A1 = 16)

.119(N = 10)

.050(A-= 18)

.105(N= 44)

with aggressive TV

Finn boys

High

.318(A'= 15)

.128(A1 = 19)

.071(JV = 11)

.177(N = 45)

Nole Standard deviations ranged from .08 to .28 and were correlated with means

Table 5 contains the multiple regressionspredicting a boy's later aggression from theproduct of earlier violence viewing and iden-tification with aggressive characters after initialaggression was partialed out.3 On the basis ofthe significant regression coefficient for theproduct, one concludes that those boys whowatched violent television and identified withaggressive TV characters became more ag-gressive 2 years later, regardless of their initiallevel of aggressiveness. Thus, identificationwith aggressive characters seems to act like acatalyst, increasing the effect of TV violencein boys. Identification with aggressive TVcharacters by itself is a good predictor ofaggression, but not as significant a predictoras its product with TV violence viewing.

The relation is illustrated by the means forboys in Table 6. Regardless of his initial levelof aggressiveness, the boy who watches moreviolence and identifies with the violent char-acter is more likely to be higher in aggressionafter 2 years. Furthermore, as Table 5 illus-trates, this effect is only weakly bidirectional.Aggression is not as good a predictor of thesetwo TV variables as the TV variables are ofaggression. Note that this effect only obtainsfor boys. For girls there are positive correla-tions, but the identification variable does notadd to the power of the violence-viewing vari-able in predicting a girl's aggression.

It is particularly interesting that this effectis equally as strong when one uses the variable"identification with all characters" instead of

"identification with aggressive characters." Ofcourse, the two variables are highly correlated.Still, the results show that even the boy whobelieves he acts like nonviolent TV charactersis more likely to increase in aggressiveness afterregular violence viewing.

Judgment of TV Realism

The children's perception of television vi-olence as realism declined significantly withage (Eron et al., 1983) but remained substantialin every grade. For example, 45% of first grad-ers thought the program "Charlie's Angels"showed life just like it really is, as did 29% ofthird graders and 7% of fifth graders.

In the earlier 10-year longitudinal study(Lefkowitz et al., 1977) it was found that girlsthought television was significantly less realisticthan did boys. It was hypothesized that thismight be one of the reasons for the lack, atthat time, of a significant longitudinal relationbetween girls' violence viewing and theiraggression. In accordance with this hypothesis,girls and boys in the United States, who arenow more equally affected by violence, nowalso perceive television violence to be equallyrealistic. Furthermore, the relation between achild's aggression and belief that television vi-

3 Such a product captures the linear effects of TV vi-olence and identification with TV character, as well astheir interaction. By itself the interactive effect was notsignificant when the main effects were partialed out.


35 —

oCOCOUJ

,GG

R

<tQUJ

.25

.20

J 10

USA Boys

USA GirlsPFINN Boys

*FINN Girls

LO MED HI LO MED HI LO MED HI

IDENTIFICATIONTV VIOLENCE TV REALISM WITH

AGGRESSIVETV CHARACTERS

Figure 3 Relation between aggression in the final wave and earlier TV viewing habits and attitudes inFinland and in the United States.

olence represents life as it really is was nowsignificant for both sexes (average r = .12,4of 6 significant for girls (overall r, ns); averager = . 18, 4 of 6 significant for boys overall r =.22, p < .01). Similarly, believing that TV vi-olence is realistic was correlated with violenceviewing (average r = .29, 6 of 6 significant forboys; average r = .30, 5 of 6 significant forgirls). Although no interactive effect of realismand violence viewing on aggression could bedetected with regression analyses, these cor-relations reveal the interlocking relation of thevariables.

In Figure 3, the main effects for each of theTV variables on aggression are summarized.The high, medium, and low groups representapproximately equal portions of the samplein each case. Although the interaction betweenviolence viewing and identification is not il-lustrated in this diagram, the direct relationbetween each of the variables and aggression,regardless of gender, is clear.

Aggressive Fantasy

Another potential mediating variable is theextent to which children fantasize about

aggression. Although some catharsis theoristsmight expect such fantasizing to be negativelyrelated to peer-nominated aggression, we hadhypothesized a positive relation. In fact, fan-tasizing about aggressive acts correlated pos-itively with peer-nominated aggression, espe-cially in the older cohort. In the United States,the average correlation in the older cohort was.182 for boys (3 of 3 significant, overall r =.23, p < .001) and .261 for girls (3 of 3 sig-nificant, overall r = .20, p < .001). However,we could find no evidence that high fantasizerswere more affected by television violence thanwere low fantasizers; that is, there was no in-teractive effect of violence viewing and extentof aggressive fantasy on peer-nominatedaggression. Use of aggressive fantasy did notcorrelate significantly with TV violence, butuse of active-heroic fantasy did correlate sig-nificantly with violence viewing for girls in theUnited States (average r - .154, 3 of 6 sig-nificant).

Aggressive Predispositions

An oft-repeated claim about the relationbetween TV violence and aggression is that


"only those predisposed to be aggressive areaffected." In fact, the evidence from this fieldstudy directly contradicts such a statement ifby predisposed one means already aggressive.In analyzing each variable, we controlled forprevious level of aggression and still obtainedsignificant results. Tables 3 and 6 show thattelevision violence can affect any child re-gardless of his or her initial level of aggression.None of the theories that have been proposedto explain the aggression-television-violencerelation have provided for a mechanism bywhich previous level of aggression would in-teract with TV violence viewing, and no ev-idence of such an interaction was found inthis study. Thus, the claim only makes senseat the more general level—that is, for aggres-sion to become a serious problem, it is nec-essary that a number of factors be present toincrease a child's aggression. No one factorwould probably be sufficient.

Parental Influences

The analysis of the role of the parentsin the aggression-television-violence relationmust be preceded by a discussion of certainbiases introduced by incomplete sampling ofparents. In the United States, there were 167children (22%) for whom there was no parentinterview. An attempt was made to interviewall parents; so those who were not interviewedwere those who were more likely to be absentfrom the home, less cooperative, and/or lessinterested in their children. Therefore, it wasnot surprising to discover that the children ofthese parents differed in a number of waysfrom the children whose parents were inter-viewed. The children of the parents who werenot interviewed were significantly more ag-gressive, F\\, 592) = 9.26, p < .003, scoredsignificantly higher on TV violence viewing,F{\, 558) = 9.95, p < .002, and frequency ofTV viewing, F\\, 558) = 5.1, p < .025, andwere marginally less popular, F( 1, 592) = 2.91,p < .09. They thought TV violence was sig-nificantly more like real life than did the otherchildren, F( 1, 556) = 32.7, p < .001, and theyidentified much more with aggressive char-acters, F(l, 541) = 14.6, p < .001. In otherwords, due to self-selection, the parent sampleis biased in favor of parents of the better ad-justed children. This introduces two problems.

First, one could expect observed relations be-tween parent variables and child variables tohave been attenuated from the true relationsbecause of the restrictions in the range thatthe sampling produced. Second, one must beconcerned whether any relation discovered isrepresentative of all children or only of thebetter adjusted children. A comparison of thewithin-child correlations in the two samplesprovides some assurance at least on this latterconcern. For example, in the United States,the overall first wave correlation between vi-olence viewing and aggression among childrenwhose parents were interviewed was .255,compared to .237 for children whose parentswere not interviewed. Of course, it is impos-sible to know whether the same consistencywould hold for correlations with parent vari-ables.

One might also be concerned that the chil-dren who dropped out during the course of 3years were significantly different from the chil-dren who remained. However, there was nosignificant difference in TV violence viewingor aggression between those children whocompleted the study and those who droppedout; although the drop outs did score signifi-cantly higher in identification with TV char-acters and judgment of TV realism. It is pos-sible that parents and children who cooperatedthroughout the study represent a higher so-cioeconomic level (SES); thus, the relation be-tween SES and other variables may be atten-uated.

Parents' Aggression

Recall that our measure of parent aggressionwas the sum of Scales 4 and 9 on the MMPI(Huesmann et al., 1978). For boys in theUnited States, we found a significant rela-tion between peer-nominated aggression andmothers' MMPI scores (r = .245, N = 248,p < .001). For girls, the correlations betweenmothers' MMPI scores and aggression werenot significant although they were positive.Similarly, no significant relation with fathers'MMPI scores was found for either sex, al-though the correlations were consistently pos-itive.

The positive relation between the mother'sscore on Scales 4 and 9 of the MMPI and herson's aggression is particularly notable because


the mother's aggression is based on self-reportof attitudes, opinions, and behavior, whereasthe child's aggression score is based on peerobservations. It is unlikely that the mother'sscores on the MMPI are affected much by thechild's behaviors. A causal regression analysisdid not yield a significant causal coefficientfor mother's MMPI on child aggression, orvice versa. It may be that the effect of mother'spersonality on the child's aggression exerts it-self when the child is younger. Therefore, al-though the overall level of aggression of a 6-to 11-year-old child would correlate withmother's aggressiveness, changes in aggressionduring these ages would not.

Although mother's and son's aggressionwere correlated, we found no evidence thatthe parents' aggression either exacerbated ormitigated the effect of TV violence on theirchildren. Neither parent's aggressiveness wassignificantly correlated with the child's vio-lence viewing, and no significant interactiveeffect was found for parent aggression and childTV violence viewing on child aggression. Therewas no evidence to support the hypothesis thatchildren of more aggressive parents are affectedmore by TV violence.

Parents' TV Viewing

As described earlier, a parent's report ofchild's violence viewing correlated with his orher child's report, lending validity to the mea-sures. However, the parent's own self-reportedviolence viewing was uncorrelated with thechild's. This was somewhat surprising inas-much as one might expect the entire familyto be watching the same programs. The par-ents' reported hours of TV viewing did cor-relate positively with the child's hours as es-timated by the parents (r = .38, N = 582, p <.001) in the United States, and less significantlywith the child's frequency as estimated by thechild. Also, in the United States, the morehours of TV a child watched and the morefrequently a child watched his or her favoriteprograms, the more the parents reportedwatching the child's programs themselves ( r=.30, iV = 5 56,/7<.001 a n d r = .17, « = 556,p < .001). Given these correlations, the lackof a relation between parent and child TVviolence scores suggests that at least somefamilies treat violent programs differently from

other programs. Further evidence in supportof this hypothesis is developed below in thesection on social class. The only other signif-icant correlation between parent's and child'sviewing habits was between the parent's andthe child's perceptions that violent shows tellabout life just like it really is. The more aparent believed this, the more the child be-lieved it, though the effect was not large (r =. 14, N = 564, p < .001). Despite these modestrelations between parent and child viewing,no correlations between parent viewing andchild aggression were discovered.

Social Class, Parents' Education,and Child Achievement

Correlations of the parents' education andsocial class with child aggression and TVviewing are displayed in Table 7. One can seethat there are consistent, if weak, effects ofboth social class and parent education on thechild's TV viewing and aggression. Socioeco-nomic status of the family was defined for thestudy by the father's occupation. In the UnitedStates, the lower the parents' education andSES, the more aggressive was the child, themore the child watched TV and TV violence,the more the child believed the violent showstell about life like it is, and the more the childidentified with TV characters. Similarly, thelower the parents' education and social status,the more aggressive they were, the more theywatched TV, and the more they believed TVwas realistic. Unexpectedly, however, the morepoorly educated and lower-status parentswatched less TV violence and more of theirchildren's programs than did the better edu-cated and higher status parents. Higher statusparents watched less TV, but more of whatthey watched was violent.

Although the child's intelligence was notmeasured directly, academic achievement wasmeasured by the California Achievement Test(Tiegs & Clark, 1970) during the first wave ofthe study for the older cohort in the UnitedStates. The correlations of achievement withaggression and TV violence viewing are shownin Table 7. One can see that academic achieve-ment was significantly inversely related to thechild's aggression and violence viewing. Thelower the child's academic performance, thegreater was the child's aggression, the more


violence the child watched, and the more thechild thought that TV violence was like reallife. The correlation between achievement andaggression in the United States was one of thehighest correlations obtained for aggression.Low achievement may be both a frustrator,instigating aggressive responses, and a markerof a reduced capacity to deal intelligently withdifficult social situations. Aggression, in turn,may interfere with intellectual development(Huesmann, Eron, & Yarmel, 1983). The low-achieving child may turn to TV violence toobtain vicariously the successes that he or shecannot obtain in school. In turn, increasedTV viewing and increased aggression maycontribute to a further decline in academic

achievement. In considering such speculations,however, one must remember that lowachievement is correlated with lower SES (r =.25, N = 268, p < .001) and lower parentaleducation (r = .35, N = 258, p < .001) in theUnited States. Thus, the relation betweenachievement and aggression is confoundedwith the relation between social class andaggression.

As we mentioned, the social-class variablesunexpectedly correlated positively with pa-rental violence viewing. One can see from Ta-ble 7 that the same is true of the child's ac-ademic achievement. Although the child'sachievement correlates negatively with thechild's violence viewing and negatively with

Table 7Relation of the Parent's Social Class* and Education and the Child's Academic Achievement* to theChild's Aggression and TV Viewing Habits

Social classParent's educationChild's

achievement

ChildAggressionTV violenceTV regularityJudged TV

realismIdentification

with TVcharacters

ParentsMother's

aggressionFather's

aggressionTV violenceTV frequencyJudged TV

realismWatch child's

TV programsReport of

child's TVfrequency

Socialclass

1.0.59***

.24"*

- . 1 7 * "- .11*- . 1 4 "

- . 2 1 * "

- . 2 1 * "

- . 1 2 "

-.16*

- . 2 7 " *

- . 1 1 "

- . 2 5 * "

- . 2 4 " *

United States

Parent'seducation

1.0

.34*"

-.09*- . 1 5 * "- . 1 6 * "

- . 2 8 * "

- . 2 3 * "

-.09*

—.09*

- . 3 1 * "

- . 1 3 "

- . 2 7 * "

- . 2 7 * "

Child'sachievement

1.0

- . 4 0 " *- . 2 0 " *- . 2 3 * "

- . 3 1 * "

—

—

—.16"

-•12t

-.12t

-.13*

Socialclass

1.0.48*"

—

—

-.16*

—

- . 2 0 "

-.18*

• 13t

Finland

Parent'seducation

1.0

—

—

-.15f

- . 1 8 "

—

-.14f

Child'sachievement

1.0

- . 2 3 "

- . 2 4 "

-.16*

-.14t

* Social class was recoded from father's occupation; therefore, a high score means higher socioeconomic status.b Achievement was measured by scores on the California Achievement Test in the United States and by reading gradesin Finland.t p < .10. * p < .05. " p < .01. " * p < .001. All probabilities are two-tailed.


all other parent and child viewing measures,the child's achievement correlates positivelywith the parent's violence viewing. These re-sults are consistent with the fact that parentalviolence viewing is uncorrelated with child vi-olence viewing in our population. Together,these results suggest that violent shows aretreated differently than other shows by manyfamilies. It may be that, on the average, violentshows appeal more to sophisticated adults.These more highly educated adults eschew thetype of shows their children watch. At the sametime, these adults, concerned about the vio-lence, discourage their children from viewingthe programs. Although such a hypothesiscannot be tested with the current data, it isconsistent with the correlations.

To test whether education, SES, and intel-ligence might completely account for the re-lation between violence viewing and aggres-

sion, these variables were added to the multipleregression equations predicting later aggressionfrom earlier violence viewing (and for boy's,earlier identification with TV characters). InTable 8, both the raw regressions and theregressions corrected for the unreliability ofthe measures are shown. Unfortunately, be-cause of the incomplete survey of parents andschool achievement records, the samples weregreatly reduced in size for these analyses.However, the causal coefficients for the reducedsample were approximately the same as forthe total sample before the covariates wereintroduced, and the introduction of achieve-ment, education, and social class did notchange them much.

Popularity

The final intervening variable to be ex-amined is the child's popularity. In the United

Table 8Multiple Regressions Showing Effects of Soaoeconomic Variables and Academic Achievement on theRelation Between TV Violence Viewing and Aggression

Predictor of third-wave aggression

Grade'First wave aggressionTV violence viewing

for girls in firstand second waves

Product of TVviolence viewingand identificationwith aggressiveTV characters forboys in first andsecond waves

Parents'socioeconomicstatus

Parents' educationalstatus

Child's academicachievement

U S

.341***

.169*

.059

.080

-.340*"*R2 = .342,W, 83) =p<.001

Standardized regression

Girls

(N = 89)

(.318)

(.193)

(.165)

(.175)

(-.375)

8.63,

Finn

-.094.650"*

.049

—

.154

-.022

.059R2 = .432.F{6, 69) =p< .001

(N = 76)

(-.105)(.667)

(.057)

—

(.210)

(.021)

(.081)

8.75,

U.S

.731'

—

.188'

.034

-.031

045

R2 = .

coefficients

1

(N = 84)

'•* (.782)

—

• (.292)

(.036)

(.012)

(.121)593,) = 22.7,

p < .001

Joys

Finn

- 140.673»"

—

.210"

-.166

-.087

-.129R2 = .564,

W, 74) =p < .001

(N = 81)

(-.138)(.665)

—

(.303)

(-.282)

(-.145)

(-.152)

16.0,

Note Numbers in parentheses are the regression coefficients corrected for the unreliability of the measures. Becauseachievement data were only available for third-grade subjects in the United States, grade was not a predictor for them.*P< .05. *• p < .01. •*• p < .001.


States, popularity was negatively correlatedwith aggression in every grade among bothboys (average r = —.347, 6 of 6 significant)and girls (average r = —.295,5 of 6 significant).

In the United States, regressions similar tothose in Table 2 demonstrated that earlieraggression was predictive of later unpopularityeven after initial popularity was partialed out(B = - .19, p < .001 for girls; B = - .14, p <.02 for boys). At the same time, earlier un-popularity was found to be predictive of lateraggression (B = —.10, p < .08 for girls; B =—. 14, p < .01 for boys). Thus, aggression seemsto lead to unpopularity, and unpopularity toaggression in the United States.

Although unpopularity did not interact withviolence viewing to effect aggression, unpopu-larity among girls in the United States wascorrelated with higher levels of violence view-ing (average r = . 161) and increased frequencyof viewing (average r = .120). Furthermore, alongitudinal regression analysis revealed thatfor both sexes unpopularity led to an increasein TV violence viewing over the 2 years, butTV violence viewing did not lead to a decreasein popularity. It may be that unpopular youthsin the United States turn to a heavier diet ofTV violence to experience vicariously the so-cial rewards that TV heros obtain.

Finland Study

There are major differences between thepopulations of the United States and Finlandthat may be very important for this study. Tobegin with, Finland has only 4.7 million in-habitants, and the presence of a nonCaucasianamong their inhabitants is extremely rare. So-cioeconomic differences in Finland, as cal-culated by income, have a smaller range ofvariation than in the United States. At thesame time, although Finland is a friendlyneighbor of the Soviet Union, it is a capitalisticcountry with much Western influence, re-flected in its mass media. In fact, Finland hasone of the highest standards of living in theworld. The Finnish language belongs to theFenno-Ugrian group, which is not related tothe Indo-European languages. The children inthe current study had not yet learned otherlanguages well and thus were not able to un-derstand the dialogue of the programs im-

ported from other countries, even the otherScandinavian ones, because imported pro-grams always have subtitles and are neverdubbed.

Subjects

The original group of subjects in Finlandwas composed of 220 children from two publicschools in the Turku area. Turku is the thirdlargest city in Finland with about 160,000 in-habitants. As Finland's "gate to the West,"Turku is a passenger and freight port withshipbuilding, food, and textile industries. Itwas the capital of Finland until 1812 and ispresently an educational center for two uni-versities. Founded in 1229, Turku has the old-est cultural traditions in Finland.

One of the schools investigated was situatedin the center of the city and the other was ina suburb. Both were public schools, as arealmost all schools in Finland, and there wereno obvious social class differences between theschool populations.

In Finland, children enter the first grade inthe year of their seventh birthday. Thus, thesubjects were on the average about 9 monthsolder than were their U.S. counterparts. How-ever, the procedure used to select the subjectswas very similar to that used in the UnitedStates. In Finland, the final permission ratewas 83% of the children. Most of the othersdeclined to participate; only a few never re-sponded.

Of the 220 subjects who constituted theFinnish sample during the first wave in 1978,115 were girls (64 first graders and 51 thirdgraders) and 105 were boys (56 first gradersand 49 third graders). In 1979, for the secondwave, 192 children remained in the sample,and in 1980, for the third wave, there were178. Again, subject attrition was primarily dueto children changing schools. The mortalityrate in Finland (19%) was substantially lowerthan in the United States (33%), reflecting thegreater mobility of the U.S. population. InFinland, parents of 193 children were inter-viewed during the first or second waves (144both mother and father, 47 mother only, and2 father only, for a total of 337 interviews).Of these children, 134 had at least one parentreinterviewed after the third wave (61 both


mother and father, 47 mother only, 16 fatheronly).

Procedure

The procedure in Finland was essentiallyidentical to that used in the United States. Thechildren were seen for the first time in 1978and subsequently in 1979 and 1980. A few ofthe child measures were changed slightly toreflect cultural differences and requirements.For measuring TV viewing preferences onlyfive lists of seven programs each were usedbecause of the lower number of programsavailable in Finland. Similarly, only 7 violentprograms were evaluated for TV realism com-pared to 10 in the United States. A few of thepictures used in measuring preferences for sex-typed activites were altered on the basis ofratings of stereotypicality made in Finland.Nevertheless, the computed reliability of everyvariable used in Finland was comparable tothat in the United States.

The comparability of the TV violence mea-sures used in Finland with that used in theUnited States cannot be established with cer-tainty because of programming differences.However, interrater reliability for evaluatingthe violence of TV shows was comparable to

; that found in the United States (.75). The Fin-nish raters received the same training and in-structions as the raters in the United States;that is, only graphically portrayed physicalaggression was seen as violence. In addition,

* most of the violence on Finnish TV occurred- in imported programs. Indeed, on the average,- one fourth of the shows used in the Finnish? lists appeared on the United States lists.

Results

As in the United States, peer-nominatedaggression increased over the course of Grades

, 1 through 5 in Finland. The developmentalincrease shown in Figure 1 is indicated bysignificant main effects for both wave, F\2,348) = 8.40, p < .001, and cohort, F\ 1, 174) =4.45, p < .04. Also, boys were again consis-tently more aggressive than were girls, F(l,174) = 8.70, p < .004, and watched signifi-cantly more violence than did girls, F{ 1,174) =43.5, p < .001. The difference between the

aggression score of the two cohorts in Finlandin Grade 3, F(l, 190) = 6.31, p < .02, seemsto be due to both retesting and subject attrition.For example, when one compares only thosesubjects present for all three waves, the dif-ference is still substantial (p < .06), indicatingthat even after attrition is removed as an ex-planation, the effect remains.

The means for violence viewing in Finlandare diagrammed in Figure 2. Because of pro-gramming differences in the two countries,direct numerical comparisons with the UnitedStates are inappropriate. In Finland, there areonly two channels and they operate primarilyin the evening hours. In the Chicago area, thereare five VHF channels in addition to severalUHF channels and cable TV, each of whichis available to the subjects at least 16 hoursper day. This difference is reflected in themothers' report of how frequently their chil-dren watch TV. For example, in the older co-hort, children in Finland were reported towatch an average of 10.12 hours per week,whereas in the United States they watched anaverage of 14.99 hours per week. Nevertheless,the data in Figure 2 do suggest substantialviolence viewing among Finnish children. Forexample, to obtain a score of 120 (the meanfor third-grade boys in 1978), a child wouldhave had to watch at least three highly violentprograms every time they were shown.

An ANOVA of these data revealed that boyswatched significantly more violence than didgirls in Finland, F(l, 174) = 43.5, p < .001.Furthermore, in contrast to the United States,where violence viewing peaked in the thirdgrade, average viewing increased from the firstto fifth grades in Finland, as indicated by sig-nificant main effects for wave, F\2, 348) =4.42, p < .013, and cohort, F(l, 174) = 5.91,p< .016.

The overall level of aggression was signifi-cantly lower in Finland than in the UnitedStates for all grades and both sexes, F{ 1,664) =44.8, p < .001; however, this difference is dif-ficult to interpret. The substantially higherlevels of peer-nominated aggression in theUnited States might represent real differencesin aggression or could represent differences inthe children's propensities to nominate peers.Because scores on other variables based onpeer nominations (popularity) did not differ


between countries, the latter is an unlikely ex-planation, but it cannot be eliminated.

TV Violence and Aggression

The correlations between violence viewingand aggression in Finland are shown in Table1. Although less consistent than in the UnitedStates, the correlations are mostly positive andreplicate the U.S. results, especially for boys.However, only one of the correlations was sig-nificant for girls. The greater variation is per-haps a function of the smaller sample size andthe fewer numbers of TV programs used inmeasuring violence viewing in Finland. Theoverall correlations were .22 for boys (p < .04)and .16 for girls (ns).

Although children's self-reported regularityof viewing their favorite programs correlatedwith aggression in the United States, this wasnot the case in Finland. Frequency-weightedviolence viewing was a much better predictorof aggression than was regularity of viewingalone (.22 vs .14 for boys).

The regressions testing the antecedent-con-sequent relation of violence viewing andaggression in Finnish children are included inTable 2. The effect of a boy's earlier violenceviewing on his later aggression is positive butweak, as in the United States. However, theeffect for girls is completely insignificant. Al-though there are significant bidirectional ef-fects for girls in the United States, the onlyeffect in Finland is a weak trend for earlieraggression to engender later violence viewing.At the same time, as in the United States, thepredictability of aggression is lower for girls(R2 = .39) than for boys (R2 = .58), althoughaggression is again reasonably stable for both.

These results suggest, as do U.S. data, thatmultiple processes are operating and inter-vening variables should be examined.

Sex of Viewer and TV Character Observed

Just as in the United States, the averagecorrelation of girls' aggression with their view-ing of aggressive female characters was nohigher than was the correlation with aggressivemale characters. Thus, the lack of a relationbetween violence viewing and aggression ingirls cannot be attributed to paucity of femalecharacters on Finnish television.

In Finland, a child's preference for sex-typedactivities related to aggressiveness in much the

same way as in the United States. Most notably,a preference for neutral activities was asso-ciated with lower aggression in both boys (av-erage r = —.193, 2 of 6 significant, overall r =- .33 , p < .001) and girls (average r = - .158,3 of 6 significant, overall r = - .20, p < .06).However, preference for sex-typed activityagain did not interact with violence viewingto affect aggression.

Identification with TV Characters

Perhaps the most interesting result in theUnited States data was that a boy's identifi-cation with TV characters (how much hethinks he acts like TV characters) interactedsignificantly with violence viewing to affectaggression. As the correlations in Table 4,regressions in Table 5, and means in Table 6indicate, almost identical effects were foundin Finnish boys. Longitudinally, the correlationbetween a Finnish boy's identification withaggressive TV characters during the first 2years of the study and his aggression in the3rd year was .38 (N = 189, p < .001). However,the interactive product of violence viewing andidentification was again the best predictor oflater aggressiveness. Also, as in the UnitedStates the effects were as strong or stronger foridentification with all characters as for iden-tification with aggressive characters (overall r =.43 between aggression and identification withall characters). Identification with TV char-acters is the one variable that seems to ex-acerbate the effect of violence viewing amongboys in both countries.

Judgment of TV Realism

As reported above, in the United States,boys and girls perceived TV to be equally re-alistic. In contrast, Finnish girls perceived TVviolence to be less realistic than did Finnishboys, 7=1(1,151)= 10.53, p<. 002. The childrenin Finland also seemed to perceive violenttelevision as being more realistic than did theirpeers in the United States, F\\, 580) = 135,p < .0001, and did not decrease their judgmentwith age as did the children in the UnitedStates. However, these latter two differenceswere probably methodological artifacts.4

4 Because several of the same programs appeared onthe realism lists in both countries, these differences couldnot be attnbuted solely to program differences. Also, the


We found in Finland, as in the UnitedStates, weak but significant positive relationsbetween the child's aggression and how muchthe child believed that violent shows tell aboutlife like it is (average r = .20, 4 of 6 significantfor Finnish girls; and average r = .08, 1 of 6significant for Finnish boys). Similarly thechild's realism score again correlated with hisor her violence viewing (average r = .22, 1 of6 significant for boys; average r = .19, 2 of 6significant for girls). Thus, in both countriesviolence viewing, belief in the realism of TVviolence, and aggression mutually covary.

Aggressive Fantasy

Fantasizing about aggressive acts, in Finlandas in the United States, correlated positivelywith peer-nominated aggression among sub-jects in the older cohort (average r = .22, 1 of3 significant for boys, average r = .16, 1 of 3significant for girls). However, we could againfind no evidence that aggressive fantasizerswere more affected by TV violence.

Parental Influences

In Finland, we obtained parent interviewsfor 88% of the children and could find nosignificant differences between those children

differences were very likely not due to differential cali-bration of the questions because the children in the twocountries scored the same when asked the question, "Howtrue do you think newspapers are in telling what life isreally like?" However, most of the violent programs onthe lists presented to the Finnish children were importedprograms coming from other countries (e.g., the UnitedStates, Australia, France, and England). Thus, the languagespoken by the characters was unintelligible to the Finnishchildren. Furthermore, the imported films show remoteplaces and cultures that the children probably have neverseen. In responding to the realism questions, the Finnishyoungsters might have interpreted the referent for thequestions to be the country of origin of the programs andnot Finland, and they might very well believe that thefilms depict life just as it is in these foreign countries.

To test this theory, a subsample of 190 Finnish childrenwere retested in the fall of 1981. They were asked twoquestions about each of three violent programs: "Howaccurately does the program depict life in Finland?" and"How accurately does the program depict life in the countryof origin (England, Germany or the United States)?" Themean for the first question was significantly lower thanthe mean for the second question. Furthermore, it was notsignificantly different than the children's mean realismjudgment for programs in the United States. Thus, childrenin both countries have about the same perception of howrealistically TV violence portrays life in their country.

whose parents were interviewed and those whowere not. The correlation between mother'saggressiveness and child aggression was positivebut not significant, unlike the case in theUnited States, where it was significant. In Fin-land as in the United States, neither parent'saggressiveness was significantly correlated withthe child's violence viewing,5 and no evidencewas obtained to support the hypothesis thatchildren of more aggressive parents are moreaffected by television violence.

The parent's own self-reported televisionviolence viewing did not correlate with thechild's (as in the United States), even thoughthe parents' total hours of viewing did correlatewith their estimate of the child's total viewing(r = .40, N = 146, p < .001). Again, no evi-dence could be found that parent TV viewingexacerbated or mitigated the relation betweenthe child's violence viewing and aggressive be-havior.

Social Class, Parent's Education, andChild's Achievement

One can immediately see from Table 7 thatsocial class and parent's education in Finlandhad little effect on a child's aggression and TVhabits. In fact, they also had little effect onthe parent's own aggression and TV habits.Nevertheless, most of the correlations are inthe same direction as in the United States, andthe lack of significant relations may reflect thegreater homogeneity of Finnish society.

On the other hand, academic achievementwas just as strongly related to a child's aggres-sion and TV habits as in the United States.Particularly notable is the negative relationbetween both parents' and child's frequencyof TV viewing and academic achievement. Al-though low achievement was related to lowSES (r = .26, p < .01) and to low parentaleducation (r = .22, p < .05), this confoundingcannot explain the relation between aggressionand achievement because the social class vari-ables did not correlate with aggression. More-

5 In Finland, a self-report measure of parental aggressionthat included statements about feelings (e.g., "I feel likeswearing") and statements about behaviors (e.g., "I losemy temper at people") did correlate significantly positivelywith the child's violence viewing (r = .24, N = 134, p <.01). However, it did not correlate significantly with thechild's aggression or the parents' own violence viewing.


over, the regressions in Table 8 illustrate thatthe relation between violence viewing andaggression cannot be explained by either thesocial-class variables or the child's academicachievement. In Finland and in the UnitedStates TV variables retain their significant re-lation to later aggression even when these thirdvariables are partialed out.

Discussion

Positive relations were obtained between TVviolence viewing and aggression among boysin both Finland and the United States andamong girls in the United States. Although anumber of other environmental, TV-viewing,cognitive, and parental variables correlatedwith aggression, only a boy's identification withTV characters seemed to exacerbate the effectof the violence viewing on aggression. Theseresults for boys were remarkably similar inboth countries, and path analyses suggestedthat the effect was strongest from violenceviewing to aggression; that is, the combinationof a boy's violence viewing and identificationwith TV characters is a better predictor oflater aggression than aggression was of laterviolence viewing and identification. For girls,however, such an interactive effect was not ob-tained and the relation between violence view-ing and aggression was clearly bidirectional.The finding that identification with TV char-acters is an important mediator for boys isconsistent with the results of an interventiondesigned to mitigate the effect of TV violenceon aggression (Huesmann et al., 1983). In thisintervention, those children whom the inter-vention affected most were those children whoidentified least with TV characters.

The longitudinal relation between violenceviewing and aggression was found to be mostlyindependent of the child's perception of howrealistic the violence is, the child's use of ag-gressive fantasy, the sex of the aggressive char-acter, and the parent's TV viewing habits. Themore a child watched a preferred violent show,however, the more effect it had. In the UnitedStates, the regularity with which the childwatched TV shows (i.e., the self-reported reg-ularity) seemed to be as important a predictorof aggression as was the level of violence inthe shows. In Finland, the violence contentwas clearly more important. This difference

may reflect the greater variety of TV pro-gramming and pervasiveness of violence in TVprogramming in the United States.

The hypothesis that only those children whoare already aggressive are affected by TV vi-olence received little support. The strength ofthe longitudinal relation varied little as a func-tion of the child's preexisting level of aggressionor as a function of the child's parents' aggres-sion.

There was also little support for the hy-pothesis that the significant relation now ap-pearing between a girl's violence viewing andher aggression is due to the recent increase inthe number of female aggressors on TV. Itseems more plausible that the recent greateremphasis on the need and desirability for fe-males to be assertive has caused girls to havefewer inhibitions about performing the ag-gressive behaviors they observe on TV. In thisstudy, those girls in the United States whopreferred more masculine activities exhibiteda stronger correlation between violence viewingand aggression than did girls with more fem-inine or neutral preferences. In other words,a significant relation has emerged for girls,probably because girls in the United States arebehaving more like boys behave than they havein the past.

Although for boys, the TV variables werebetter longitudinal predictors of aggressionthan aggression was of the TV variables, theprocess actually appears to be circular for bothsexes. Television violence viewing for girls andviolence viewing catalyzed by identificationwith TV characters for boys leads to height-ened aggressiveness, which in turn leads tomore TV violence viewing. A number of me-diating variables appear to play a role in thiscycle. In both countries, those children whowere less successful in school watched moreTV, perhaps because they found it more sat-isfying than schoolwork. They preferred themore violent shows, they identified more withTV characters, and they believed that the vi-olence they observed on TV reflects real life.They are exposed to more violence and havemore opportunity to learn aggressive acts. Be-cause their intellectual capacities are morelimited, the easy aggressive solutions they ob-serve may be incorporated more readily intotheir behavioral repertoire. In any case, theheavy viewing isolates them from their peers


and gives them less time to work toward ac-ademic success. Similarly, at least in the UnitedStates, children who behave aggressively areless popular. Perhaps because their relationswith their peers tend to be unsatisfying, lesspopular children (at least less popular girls)watch more television and view more violence.The violence they see on television may re-assure them that their own behavior is appro-priate or may teach them new coercive tech-niques, which they then attempt to use in theirinteractions with others. Thus, they behavemore aggressively, which in turn makes themeven less popular and drives them back to tele-vision. The cycle continues with aggression,academic failure, social failure, and violenceviewing feeding on each other.

A number of the other variables may havesimilar circular relations to TV violence view-ing and aggression. Although none of theseinteracted significantly with violence viewingto affect aggression, they correlated with bothvariables. For example, it was found that themore a parent believes TV violence is realistic,the more the child believes it is realistic; andthe more a child believes this, the more thechild views violence and acts aggressively. Theviolence the child views and the aggression thechild commits may combine to increase thechild's belief in the reality of violence, whichin turn would increase the likelihood of thechild behaving aggressively. Similarly, a child'sfrequency of aggressive or active-heroic fantasywas correlated with the child's aggression inboth countries and with violence viewingamong girls in the United States. Thus, thehypothesis that aggressive fantasizing serves asa rehearsal process for aggressive acts remainsplausible. Finally, children of either sex whopreferred less strongly sex-typed activities wereless aggressive and watched less violence.

An interesting and unexpected finding wasthat parental violence viewing did not correlatewith child violence viewing in either countrydespite many other parent-child viewing sim-ilarities. Furthermore, those parents who weremore educated, had higher status jobs, andhad children who were achievers watchedsomewhat more violence. Only speculativereasons can be offered for these parents' pref-erences for violent programs; however, the ob-tained data demonstrated that their childrenusually were not watching the same violent

programs, either because of parental controlsor other factors. This result coupled with thefailure of the sociocultural variables to explainthe TV violence/aggression relation reinforcesour conviction that television violence can af-fect children from all strata of society.

Cultural Differences

In interpreting the results of the study, somereaders may be concerned that the meaningof high violence viewing is quite different inFinland than in the United States. This is cer-tainly true in the sense that no ratio scalecomparison is possible between the countries.The violence viewing situation for the Finnishchild is quite different because most violentshows are foreign without a dubbed sound-track. However, these differences should notbe of great concern for they only highlight theimpressive communalities found between thetwo countries. The results for boys were verysimilar despite the differences in the viewingenvironment. Among other things, this sug-gests that the visual presentations of aggressiveacts, independent of plot and dialogue, maybe sufficient to engender aggression.

Although the longitudinal effects for boysare very similar in the United States and Fin-land, the results for girls are not. Furthermore,the correlations of behavior with the TV vari-ables are much less consistent in Finland. Al-though it is impossible to state with certaintywhich cultural differences cause these incon-sistencies, differences in television program-ming and attitudes toward aggression inwomen are two possibilities.

Regular TV broadcasting did not start inFinland until 1955, but by 1981 the TV net-work covered 99.9% of the country andreached 98.8% of the population. The numberof TV sets per capita in Finland in 1976 wasthe ninth highest in the world: 300 sets per1,000 persons. The United States had thehighest frequency with 571 sets per 1,000 per-sons. In Finland, radio and TV broadcastingis carried out by the National BroadcastingCompany (Oy Suomen Yleisradio Ab). Thegovernment owns 99.9% of the shares of thiscompany and 0.1% are owned by 60 differentorganizations. The National BroadcastingCompany rents broadcasting time to theCommercial Broadcasting Company (Mainos-


TV), which broadcasts about 20% of the pro-grams, typically including films, series, andentertainment. The frequency of violence isdeliberately limited. However, some highly vi-olent shows are broadcast. The violence hasfor the most part been included in foreignseries and in films. In 1976, 37% of all pro-grams were imported and these included vir-tually all of the violent programs shown onFinnish television. Most of the imported filmsand series came from the United Kingdom,the United States, France, and Sweden. Theimported programs show remote places andpeople that the Finnish children have perhapsnever seen in real life. As a result, the childrenmay have difficulty judging the reality of theseshows. Furthermore, foreign language pro-grams are never dubbed (they have subtitles),so the characters in them speak a language thechildren do not understand unless they areable to read the Finnish subtitles.

The viewers in Finland have fewer oppor-tunities to view violence on TV than do viewersin the United States. Because there are onlytwo channels, a person cannot choose to ob-serve violence every time he or she wants to.The maximum exposure to violence on TVthus may not be as high for a child in Turkuas it is for a child living in the Chicago area.For example, a systematic observation of thetwo channels in Turku during one week inMay, 1981, revealed that only 47 cases of vi-olent aggressive acts were depicted during theweek's 77 hours of programming (an averageof 0.61 acts per programming hour or 6.7 perday). This compares to a much greater fre-quency of violent acts on TV in the UnitedStates. Gerbner and Gross (1981) found thatviolent acts occur on TV in the United Statesat a rate of 5 acts per prime-time hour and18 acts per weekend daytime hour. These sta-tistics have not changed much in the last 10years. Furthermore, because there are a greatervariety of shows available in the United States,children can watch violence almost always ifthey wish.

The substantially lower level of violence onTV may account for the greater inconsistenciesin the Finnish data. When there are only afew violent programs, a child's violence scorebecomes more susceptible to distortion by theother specific stimulus properties associatedwith the programs. Thus, we suspect that the

true relation between violence viewing andaggression is probably just as strong for boysin Finland as for boys in the United States.However, the differences for girls cannot be soeasily explained. One possibility is that Finnishsociety might be placing less emphasis ontraining girls to be aggressive than is Americansociety. Such aggressiveness might be unnec-essary because women in Finland obtained amore equalitarian social status many years ago.For example, Finland was the second countryin the world to give women the right to vote(in 1906). Women's and men's roles have longbeen less differentiated in Finland than in theUnited States. A large proportion of profes-sionals in Finland are women, and the women'sliberation movement has not been a strongforce in Finland. At the same time, in Finlandas in most Western countries, femininity isconsidered a desirable trait for women. As aresult, Finnish girls may be more encouragedto inhibit whatever aggressive behaviors arestimulated by TV violence than are Finnishboys. This hypothesis is supported by the find-ing that Finnish girls perceived TV violenceas less realistic than did Finnish boys, whereasthey were no sex differences in the UnitedStates. This pattern of results for Finnish girlsresembles that found in the United States be-fore the women's liberation movementemerged as a major force (Eron & Huesmann,1980; Lefkowitz et al., 1977).

Conclusion

Antisocial aggressive behavior in childrenis most often the result of a number of pre-cipitating factors. These factors provide thelearning conditions under which aggressive re-sponding becomes habitual. Although no sin-gle factor can make a child aggressive, eachcontributes. In the current study we have ex-amined the conditions under which exposureto media violence is most likely to lead toheightened aggressiveness.

Samples of 758 children in the United Statesand 220 children in Finland were interviewedand tested in each of 3 years in an overlappinglongitudinal design covering Grades 1 to 5.For girls in the United States and boys in bothcountries, TV violence viewing was signifi-cantly related to concurrent aggression andsignificantly predicted future changes in


aggression. The strength of the relation de-pended as much on the regularity with whichviolence was viewed as on the quality of theviolence. For boys, the effect was exacerbatedby the degree to which the boy identified withTV characters. Path analyses suggested a bi-directional causal effect in which violenceviewing engenders aggression and aggressionengenders violence viewing. No evidence wasfound that only those children already ag-gressive or those with aggressive parents areaffected by TV violence. However, a numberof other variables were found to be correlatesof aggression and violence viewing. In bothcountries, the child most likely to be aggressivewould be one who (a) watches violent pro-grams most of the time they are on, (b) believesthese shows portray life just like it is, (c) iden-tifies strongly with the aggressive characters inthe shows, (d) frequently has aggressive fan-tasies, and (e) if a girl, prefers boys' activities.In addition, such a child is likely to (a) havea more aggressive mother, (b) have parentswith lower education and social status, (c) beperforming poorly in school, and (d) be un-popular with his or her peers.

The most plausible model to explain thesefindings seems to be a multiprocess model inwhich violence viewing and aggression mu-tually encourage each other and are enhancedby these related variables. Observationallearning undoubtedly plays a role, but its rolemay be no more important than are the at-titude changes produced by TV violence, thejustification for aggressive behavior providedby TV violence, or the cues for aggressiveproblem solving furnished by TV violence.

References

Andison, F. S. (1977). TV violence and viewer aggressionA cumulation of study results 1956-1976. Public Opin-ion Quarterly, 41, 314-331.

Armour, D. J. (1975, April). Measuring the effects of tele-vision on aggressive behavior (RAND Report #R-1759-MF) Santa Monica, CA: Rand Corporation.

Bailyn, L. (1959). Mass media and children: A study ofexposure habits and cognitive effects. PsychologicalMonographs, 73, 1-48.

Bandura, A. (1977). Social learning theory, EnglewoodCliffs NJ: Prentice Hall.

Bandura, A., Ross, D., & Ross, S. A. (1961). Transmissionof aggression through intimation of aggressive models.Journal of Abnormal and Social Psychology, 63, 575-582.

Bandura, A., Ross, D., & Ross, S. A. (1963a). Imitation

of film-mediated aggressive models. Journal of Abnormaland Social Psychology, 66. 3 -11 .

Bandura, A., Ross, D., & Ross, S. A. (1963b). Vicariousreinforcement and imitative learning. Journal of Ab-normal and Social Psychology, 67, 601-607.

Belson, W (1978). Television violence and the adolescentboy Hempshire, England- Saxon House.

Chaffee, S H (1972). Television and adolescent aggres-siveness (overview). In G A. Comstock & E. A. Rubin-stein (Eds.), Television and social behavior Vol 3 Tele-vision and adolescent aggressiveness (pp. 1-34). Wash-ington, DC: U.S. Government Printing Office.

Collins, W. A (1973). Effect of temporal separation be-tween motivation, aggression, and consequences: A de-velopmental study Developmental Psychology, 8, 215—221

Collins, W. A. (1982). Cognitive processing aspects of tele-vision viewing. In D. Pearl, L. Bouthilet, & J. Lazar(Eds.), Television and behavior- Ten years of scientificprogress and implications for the eighties Vol 2 Tech-nical reviews (pp. 9-23). Washington, DC: U.S. Gov-ernment Printing Office.

Collins, W. A., Berndt, T. J., & Hess, V. L. (1974). Ob-servational learning of motives and consequences fortelevision aggression. A developmental study. Child De-velopment, 45, 799-802.

Comstock, G. A (1978). Contribution beyond controversy(Review of Growing up to be violent A longitudinalstudy of the development of aggression by M. M. Lef-kowitz, L. D Eron, L. O. Walder, & L. R. Huesmann).Contemporary Psychology, 23, 807-809.

Comstock, G. A. (1980) New emphases in research onthe effects of television and film violence. In E. L. Palmer& A. Dorr (Eds.), Children and the faces of televisionTeaching, violence, selling (pp 129-148). New Vbrk:Academic Press.

Cook, T. D., Kendzierski, D. A., & Thomas, S. J. (inpress). Television research for science and policy: Analien perspective in the NIMH Report on Televisionand Behavior. Public Opinion Quarterly

Dorr, A., & Kovaric, P. (1980). Some of the people someof the time—but which people' Television violence andits effects. In E Palmer & A. Dorr (Eds.), Children andthe faces of television Teaching, violence, selling (pp.183-200) New York- Academic Press

Eron, L D. (1963) Relationship of TV viewing habitsand aggressive behavior in children. Journal of Abnormaland Social Psychology. 67. 193-196.

Eron, L D. (1980). Prescription for reduction of aggression.American Psychologist, 35. 244-252

Eron, L D. (1982) Parent child interaction, televisionviolence, and aggression of children. American Psy-chologist, 37. 197-211.

Eron, L. D., & Huesmann, L. R. (1980). Adolescentaggression and television. Annals of New York Academyof Science 347. 319-331.

Eron, L. D , Huesmann, L. R., Bnce, P., Fischer, P., &Mermelstem, R. (1983). Age trends in the developmentof aggression, sex typing, and related television habits.Developmental Psychology, 19, 71-77.

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

Eron, L D., Walder, L. O, & Lefkowitz, M. M. (1971).


Learning of aggression in children, Boston, MA: Little,Brown.

Eysenck, H. J., & Nias, D. K. (1978). Sex, violence andthe media. London- Maurice Temple Smith.

Feshbach, S. (1964). The function of aggression and theregulation of aggressive drive. Psychological Review, 71,252-272.

Feshbach, S. (1976). The role of fantasy in the responseto television. Journal of Social Issues. 32, 71-85.

Gerbner, G., & Gross, L. (1981) The violent face of tele-vision and its lessons. In E. Palmer & M. Dorr (Eds.),Children and the faces of television Teaching, violence,selling New York: Academic Press, 149-162.

Granzberg, G., & Steinbring, J. (1980). Television and theCanadian Indian (Tech Rep.). Winnipeg, Manitoba,Canada: University of Winnipeg, Department of An-thropology.

Hearold, S. L. (1979). Meta-analysis of the effects of tele-vision on social behavior Unpublished doctoral disser-tation, University of Colorado.

Hennigan, K. M., Del Rosario, M L., Heath, L., Cook,T. D., Wharton, J. D., & Calder, B. J (1982). The impactof the introductions of television on crime in the UnitedStates. Journal of Personality and Social Psychology, 42,461-477.

Hicks, D. J. (1965). Imitation and retention of film-me-diated aggressive peer and adult models. Journal of Per-sonality and Social Psychology, 2, 97-100.

Hicks, D. J. (1968). Effects of co-observer's sanctions andadult presence in imitative aggression. Child Develop-ment, 39, 303-309.

Huesmann, L. R. (1982a). Television violence and ag-gressive behavior. In D. Pearl, L. Bouthilet, & J. Lazar(Eds.), Television and behavior Ten years of scientificprogress and implications for the eighties Vol 2 Tech-nical reviews (pp. 126-137). Washington, DC: U.S.Government Printing Office.

Huesmann, L R. (1982b). Process models of social be-havior. In N. Hirschberg (Ed.), Mullivanate methods inthe social sciences. Applications Hillsdale, NJ: Erlbaum.

Huesmann, L. R., Eron, L. D., Klein, R., Brice, P., &Fischer, P. (1983) Mitigating the imitation of aggressivebehaviors by changing children's attitudes about mediaviolence. Journal of Personality and Social Psychology,44. 899-910

Huesmann, L. R., Eron, L. D., Lefkowitz, M. M., & WaldetL O. (1973). Television violence and aggression- Thecausal effect remains. American Psychologist, 28, 617-620.

Huesmann, L. R., Eron, L. D., Leflcowitz, M. M., & Walder,L. O. (1979). Causal analyses of longitudinal data. Anapplication to the study of television violence andaggression (Tech. Rep.). Chicago: University of Illinoisat Chicago. Department of Psychology.

Huesmann, L. R., Eron, L. D., Leflcowitz, M. M., & Walder,L O. (in press). The stability of aggression over timeand generations. Developmental Psychology

Huesmann, L. R., Eron, L D., & Yarmel, P. (1983, August).Intellectual functioning and aggression Paper presentedat the meeting of the American Psychological Associ-ation, Anaheim, CA.

Huesmann, L. R., Leflcowitz, M. M., & Eron, L. D. (1978).Sum of the MMPI Scales F, 4, and 9 as a measure ofaggression. Journal of Consulting and Clinical Psy-chology, 46. 1071-1078.

Kaplan, R. M. (1972). On television as a cause of aggres-sion. American Psychologist, 27, 968-969.

Kenny, D. A. (1972). Threats to the internal validity ofcross-lagged panel inference, as related to "Televisionviolence and child aggression: A follow-up study." InG. A. Comstock & E. A. Rubinstein (Eds.), Televisionand social behavior Vol 3 Television and adolescentaggressiveness (pp. 136-140). Washington, DC: U.S.Government Printing Office.

Kniveton, B H., & Stephenson, G. M. (1973). An ex-amination of individual susceptibility to the influenceof aggressive film models. British Journal of Psychiatry,122, 53-56.

Krebs, D., & Groebel, J. (1977). The effects of televisionviolence on children (Tech Rep) Aachen, West Germany:Rheimsh-Westfalischen Technischen Hochschule.

Lagerspetz, K. M J., & Engblom, P. M. (1979). Immediatereactions to TV violence by Finnish pre-school childrenof different personality types, Scandinavian Journal ofPsychology, 20, 43-53.

Lefkowitz, M. M., Eron, L D., Walder, L. Q, & Huesmann,L. R. (1977). Growing up to be violent A longitudinalstudy of the development of aggression. New York: Per-gamon.

Lefkowitz, M. M , & Huesmann, L. R. (1980). Concom-itants of television violence viewing in children. InE L. Palmer & A. Dorr (Eds.), Children and the facesof television Teaching, violence, selling (pp. 165-182).New York: Academic Press.

Leyens, J. P., Parke, R. D., Camino, L., & Berkowitz, L.(1975). Effects of movie violence on aggression in a fieldsetting as a function of group dominance and cohesion.Journal of Personality and Social Psychology, 32, 346-360.

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

McCall, R B., Parke, R. D., & Kavanaugh, R D. (1977).Imitation of live and televised models by children oneto three years of age. Monographs of the Society forResearch in Child Development, 42, 95 pp.

McCarthy, E. D., Langner, T. S., Gersten, J C , Eisenberg,J. G., & Orzeck, L. (1975). Violence and behavior dis-orders. Journal of Communication, 25, 71-85

Milavsky, J. R., Kessler, R., Stipp, H., & Rubens, W. S.(1982). Television and aggression: Results of a panelstudy. In D. Pearl, L. Bouthilet, & J. Lazar (Eds.), Tele-vision and behavior Ten years of scientific progress andimplications for the eighties Vol 2' Technical reviews(pp. 138-157). Washington, DC- U.S. GovernmentPrinting Office.

Murray, J. P., & Kippax, S. (1977). Television diffusionand social behavior in three communities: A field ex-periment. Australian Journal of Psychology, 39, 31-43.

Nadelman, L. (1974) Sex identity in American childrenMemory, knowledge, and preference texts. Develop-mental Psychology, 10, 413-417.

Newcomb, A. F., & Collins, W. A. (1979). Children's com-prehension of family role portrayals in televised dramas'Effects of socioeconomic status, ethnicity, and age. De-velopmental Psychology, 15. 417-423.

Olweus, D. (1979). The stability of aggressive reactionpatterns in human males: A review. Psychological Bul-letin. 85, 852-875.

Parke, R. D., Berkowitz, L., Leyens, J. P., West, S., &Sebastian, R. J. (1977). Some effects of violent and non-


violent movies on the behavior of juvenile delinquents.In L. Berkowitz (Ed.), Advances in experimental socialpsychology (Vol. 10, pp. 135-172). New York- AcademicPress.

Pitkanen-Pulkkinen, L. (1979). Self control as a prereq-uisite for constructive behavior In S. Feshbach & A.Fraczek (Eds.), Aggression and behavior change (pp.250-270). New York: Praeger.

Pitkanen-Pulkkinen, L. (1981). Concurrent and predictivevalidity of self reported aggressiveness. Aggressive Be-havior, 7. 97-110.

Rogosa, D. (1980). A critique of cross-lagged correlationPsychological Bulletin, 88. 245-258.

Rosenfeld, E., Huesmann, L. R., Eron, L. D., & Tomey-Purta, J. V. (1982). Measuring patterns of fantasy be-havior in children. Journal of Personality and SocialPsychology, 42, 347-366.

Rubinstein, E. G., Comstock, G. A., & Murray, J. P. (1972).Television and social behavior Vol 4 Television in day-to-day life Patterns of use Washington, DC: U.S. Gov-ernment Printing Office.

Schramm, W., Lyle, J., & Parker, E. B. (1961). Televisionin the lives of our children Stanford, CA: Stanford Uni-versity.

Signorelh, N., Gross, L., & Morgan, M. (1982). Violencein television programs 10 years later. In D. Pearl, L.Bouthilet, & J. Lazar (Eds.), Television and behaviorTen years of scientific progress and complications forthe eighties Vol 2 Technical reviews (pp. 158-174)Washington, DC- United States Government PrintingOffice.

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

Smythe, D. W. (1954). Reality as presented by television.Public Opinion Quarterly, 18, 143-156.

Sonesson, I. (1979). Forskolebarn och TV [Elementaryschoolchildren and TV]. Malmb, Sweden. Esselte Stu-dium.

Stem, A H , & Fredrich, L. K. (1972). Television contentand young children's behavior. In J P. Murray, E. A.Rubinstein, & G. A. Comstock (Eds), Television andsocial behavior Vol 2 Television and social learningWashington, DC: U.S. Government Printing Office.

Surgeon General's Advisory Committee on Television andSocial Behavior. (1972). Television and growing up Theimpact of televised violence (Report to the SurgeonGeneral, U.S Public Health Service). Washington, DC:Government Printing Office.

Tiegs, E. W, & Clark, W. W. (1970). California Achieve-ment Test. New York- McGraw-Hill.

von Feihtizen, C (1976). Bam och TV i socialisenng-sprocessen [TV m the socialization of children]. Stock-holm, Sweden: Svenges Radio (PUB, nr. 28-75/76).

Walder, L. Q, Abelson, R., Eron, L. D., Banta, T. J., &Laulicht, J. H. (1961). Development of a peer-ratingmeasure of aggression. Psychological Reports (Mono-graph suppl 4-19), 9, 497-556

Warner, W L., Meeker, M., & Eels, K. (1960). Social classin America New York: Harcourt.

Williams, T M. (1978, September) Differential impact ofTV on children A nalural experiment in communitieswith and without TV Paper presented at the meetingof the International Society for Research on Aggression,Washington, DC.

Received March 21, 1982Revision received May 19, 1983 •

Documents

Intervening Variables in the TV Violence-Aggression ... · PDF fileIntervening Variables in the TV Violence-Aggression Relation: ... posure to TV violence increases a child's ag-