Does viewing violent media really cause criminal violence? A methodological review (original) (raw)

Does viewing violent media really cause criminal violence? A methodological review

Joanne Savage*
Department of Justice, Law and Society, American University, 4400 Massachusetts Avenue, NW, Washington, DC 20016-8043, USA

Received 9 September 2002; received in revised form 26 October 2003; accepted 30 October 2003

Abstract

The topic of media violence has been the subject of heated debate in recent decades. There is a vast empirical literature on the effects of television on aggression but no published comprehensive review has ever focused on those studies that use criminal aggression as their outcome. The present paper represents an attempt to fill this void and provide a resource for those who do not wish to delve into four decades of original research in order to assess this line of investigation. Studies are evaluated based on contemporary standards of research in the field of criminology. Although the possibility that television and film violence has an impact on violent criminality remains, it is concluded here that, despite persistent published reviews that state the contrary, the body of published, empirical evidence on this topic does not establish that viewing violent portrayals causes crime.

© 2003 Elsevier Ltd. All rights reserved.

Keywords: Television violence; Media violence; Violent crime

1. Introduction

A vast empirical literature on the impact of television exposure on aggression has accrued over the last four decades, mostly in the fields of psychology and broadcasting. Reviews of this literature typically conclude that there is evidence that viewing violent media is as-

[1]

1. • Tel.: +1-202-885-2974.
     E-mail address: jsavage@american.edu (J. Savage). ↩︎

sociated with aggression (e.g., Donnerstein & Linz, 1995; Huesmann & Miller, 1994; Paik & Comstock, 1994; Wood, Wong, & Chachere, 1991). It is not this particular conclusion that is going to be called into question in the present paper. Instead, the present paper focuses on the extension of that conclusion to apply to violent or criminal behavior. For example, Donnerstein and Linz (1995) conclude that studies “strongly suggest that exposure to media violence is a causal factor in criminal behavior,” (p. 250). Palermo (1995) writes “There is no doubt …\ldots. that excessive and extended exposure to television violence may promote violence in some children …” (p. 19). Huesmann and Miller (1994) summarize, “… the existing empirical studies do provide support for the conjecture that the current level of interpersonal violence in our societies has been boosted by the long-term effects of many persons’ childhood exposure to a steady diet of dramatic media violence” (p. 155), and Sege (1998) concludes, “Although this phenomenon is complex and multifactorial, with deep historical roots, one of the best documented causes of the modern upsurge in violence appears to be childhood exposure to television violence” (p. 129).

This idea has been reified through the popular press and other published works. Although many publications have tamed their rhetoric on this topic in recent years (Anderson and Bushman, 2002), one still may find many highly overstated comments on the TV violenceviolent criminal behavior relationship. For example, in the abstract for a recent piece on the effects of television violence we find: “More than 1000 studies have proposed a link between teen violence and violent TV programs” (Mudore, 2000, p. 24). Bergenfield (1994) writes, “Thirty years of research have proved that exposure to TV violence is hazardous to children’s health and welfare” (p. 40). The medical community has been very active in recent years on this topic and literature searches on media violence result in many hits related to statements by the American Medical Association and the American Academy of Pediatrics. A report on the policy statement for the American Academy of Pediatrics quotes it as saying that “Exposure to and the influence of media violence directly correlates to violent behavior” (Chatfield, 2002; p. 735). Articles with titles such as “Teaching Kids to Kill” (Grossman, 2000) and “Television’s Bloody Hands” (McCain, 1998) are commonplace.

It is likely that it is this kind of rhetoric from the popular press (vs. the more carefully worded scholarly summaries) that has inspired calls for action. For example, the American Medical Association has expressed its “vigorous opposition” to television violence and its support for minimally restrictive measures “to protect children from the harmful effects of such programming” (American Medical Association, 1993). A report released by Senator Orrin Hatch as Senate Judiciary Committee Chairman is said to blame television for 10%10 \% of youth violence, to identify media violence as one of the principal causes of youth violence, and to recommend 14 actions federal government should take to counter effects of media violence on youth (as reported by Albiniak & McConnell, 1999). Representative Henry Hyde forwarded the “Hyde Amendment,” also known as the Children’s Defense Act in 1999, which would have made it a federal felony, punishable by 5 years in jail, to expose children under 17 to materials with sexual or violent content (reported by Reid, 1999).

Due to the immense attention media violence effects have received from policymakers, it is paramount for criminologists to evaluate the literature for its implications for crime. Of course, there are many reasons why one might expect that viewing violence would affect

violent behavior. Basic principles of learning would suggest that small children will imitate behavior (Bandura, 1977). If the scenes viewed are exciting, this may also provide physical stimulation (excitation) that might have an immediate effect on behavior and, perhaps, a rewarding sensation that would then be associated with violence (classical conditioning). There is also the likelihood that vicarious reinforcement would affect the viewer-watching the perpetrator of violence receiving rewards for violent behavior might provide important information to the viewer about the consequences of behaving violently.

The most sophisticated and widely accepted perspective on the relationship between viewing violent TV and aggression has been forwarded by Huesmann (1986a). Huesmann and Miller (1994) argue that " . . the most plausible hypothesis is that habitual exposure to violent television programs teaches children aggressive habits which are maintained well into adulthood" (p. 165). Huesmann’s theory is based on the proposition that social behavior is guided by cognitive scripts that are stored in a person’s memory. Aggressive people are those who regularly retrieve and employ scripts that emphasize aggressive responding. Children can learn aggressive scripts from many sources-including watching television. The process is reciprocal-troubled and aggressive children often watch more television and identify with television characters to a greater extent than other children, and watching more television reinforces these violent scripts. The theory also addresses other complexities such as the importance of cues in the environment necessary for the retrieval of scripts.

However, although the theory makes sense, and there are a very large number of studies that examine the effects of media exposure on aggression, there are comparatively few studies that actually measure the effects of media violence on criminal aggression. The majority of the published studies alluded to in the popular press use, as an outcome measure, a measure of aggression that is not violent nor criminal. Most typically, they employ a shock box where subjects are told to administer shocks to another individual in a learning scenario. The outcome measure is often the maximum level of shock that each subject chooses to administer, and evidence of an effect is demonstrated if the group who had viewed a violent television show in an earlier experimental session set shock levels at a higher level, on average, than a group who viewed a control program.

Kaplan (1984) questioned the validity of using laboratory measures of aggression, such as the shock machine, for understanding real-life violence and aggression and concluded that naturalistic observation in field settings held the greater promise for understanding this phenomenon. Although most scholarly works provide very carefully worded conclusions based on such studies, the belief prevails that watching violent television or movies causes violent behavior. There are some obvious reasons why one might question the generalizability of findings using this type of outcome to behaviors that are violent and criminal. For example, there is a psychological difference between pressing a button and hitting or shooting someone directly. There is also the matter of “demand characteristics” of the experimental situation, where subjects are told to shock the other person-there is no rule or law against it-and, indeed, they are being encouraged to do so (see Felson, 1996, for further discussion). We can imagine that many ordinary people who would not be willing to violate the law or harm others normally might be willing to administer pain or harm to another person in a setting where it is legal and expected (e.g. as part of service in the

armed forces, as a police officer, or as a nurse who must give a shot to a patient) and where retaliation is unlikely to occur. There is no empirical basis for the assumption that there is an uninterrupted linear relation between legal, laboratory aggression and crimes such as aggravated assault and robbery. Given that the law, social disapproval, socialization, and formal punishment stand between this type of aggression and criminal aggression, the assumption is dubious at best.

The purpose of the present paper is to provide a review of all the published studies, in English, that examine the effects of viewing television or film violence on criminal behavior. Because previous reviews related to this topic are plentifully available, the present paper will emphasize the methodological problems that cause one to question the conventional wisdom. This is not the only review that has expressed skepticism regarding the aggression-violence relationship (see discussions by Felson, 1996; Freedman, 1984, 1986, 1988, 1992; Jensen, 2001; McGuire, 1986; Zimring & Hawkins, 1997). However, the present review is the only comprehensive review of the literature limited to studies that examine the effect of media violence on behavior that is criminally violent; that is, the overt expression of physical force against another individual and that might constitute criminal behavior, broadly defined. This review highlights the methodological issues that are relevant for understanding whether viewing media violence causes violent criminal behavior and, as such, it is hoped that it will provide a resource for those who want to understand the effects of media violence on crime but who do not wish to delve into the original literature themselves. Because the majority of reviews of this topic have suggested that viewing media violence causes violent behavior, great care is taken here to point out the plentiful evidence that contradicts this conclusion.

Empirical articles and reviews of the literature were identified, first, by using an electronic search of books and articles through a major university consortium database. Criminal Justice Abstracts was also consulted for recent years. Then, the bibliographies of all the items obtained were reviewed and empirical articles referenced therein were evaluated. This process was repeated until all empirical articles were obtained that fit the established criteria. Inclusion of studies was based on outcome measure; only studies where criminally violent behavior or “analogous” behavior are used are reviewed. Violent and analogous behavior includes behavior that is both physically hurtful or coercive and violates a rule, law, or norm. For example, in the present analysis, although the use of electric shock in a teacher-learner experiment is not included (because while it may be physical and harmful it is not a violation of a rule or law), children pushing, hitting, or kicking one another is included because it is physical and possibly harmful and violates rules that are probably well known to the children. This is seen as analogous to criminal violence, whereas following instructions in an experimental paradigm is not. Because of the emphasis on certain longitudinal studies in recent years, they are included although their outcome measure, peer-nominated aggression, represents overt violence only peripherally.

The review is organized by study methodology starting with aggregate-level studies (cross-sectional and longitudinal), continuing with individual-level studies (experimental and quasi-experimental) and ending with correlational studies and prospective longitudinal

studies. The review emphasizes methodology, using, as its criteria, methodological standards in the field of criminology. This methodological approach is important because, as we shall see, a careful evaluation of numerous studies calls into question the conclusions reported by their authors and later reviewers.

2. Methodological criteria

Table 1 summarizes the criteria for evaluating studies organized by study type. The criteria are based on contemporary standards of research in criminology and criminal justice and knowledge gained from reviewing several decades of television and media research. Each study was evaluated based on the set of appropriate criteria, although comments in the text of this paper are restricted to those of greatest relevance for determining the contribution of that study to our understanding of the effects of media exposure on criminal behavior.

In each section, more detail will be given on the specific methodological criteria sought for that particular study type. The criteria that span all of the studies shall be discussed here. First, for a study to be taken seriously, the authors should report the measures used and analyses conducted in a systematic fashion so that others could replicate the study if desired. This should go without saying, but several widely cited studies in this area do not meet even this basic criterion. Second, sample sizes should be large enough to make statistical inferences possible. In the interest of saving space, in this paper comments on sample size are made only when samples are thought to be too small to allow significance testing.

Measures should be reliable and valid. Reliability is not of great concern in the present review because many authors took care to demonstrate reliability. With respect to validity, outcome measures closer to criminal violence should be weighed most heavily. A few studies using a marginal measure of criminal aggression, peer-nominated aggression, are used and, all else being equal, studies using this measure should be given less weight. The best measures of the independent variable, television violence exposure, include a reliable estimate of television viewing that is weighted by an independent assessment of violent content of the shows watched. Weaker measures include those where television viewing overall is used (instead of television violence viewing) or where a rating of preferences for programs is used rather than an estimate of actual viewing time.

Of interest throughout this review is the establishment of temporal order and the control of spurious factors. Evidence suggests that aggressive people like to watch violent material so the actual causal progression may be opposite that being tested (aggression may cause exposure to TV violence) or the relationship may be spurious (aggressive predisposition causes violent behavior and a taste for violent programs). The control factors of greatest interest here are violence proneness (usually referred to as trait aggressiveness) and parental neglect or abuse because of their likely influence on television-viewing habits and violent behavior. Studies also control for factors such as income and parental education fairly regularly, which is important.

Table 1
Methodological criteria for studying the effects of media violence on criminally violent behavior
Common criteria across all types of studies

1. Systematic reporting adequate for replication
2. Adequate sample size for statistical inferences
3. Reliable and valid measures: measures of actual exposure to violent television, outcome measures of violent criminality analogous behaviors (e.g., hitting, pushing, etc.)
4. Establish or address temporal order
5. Control for potentially spurious factors
6. Test for interactions to demonstrate whether effects are limited to persons with certain characteristics (i.e., trait aggressiveness/violence proneness)

Criteria for aggregate level studies
Cross-sectional studies
Hypothesis: Jurisdictions where residents are exposed to more violent television are expected to have higher violent crime rates.
Attention to:
-Temporal order
-Control for factors likely to be associated with exposure to TV and crime such as SES, education levels, routine activities, and demographics (e.g., racial composition, age distribution of the population).
Longitudinal studies
Hypothesis: Jurisdictions where there has been an increase in exposure violent television are expected to experience increases in crime or violent crime.
Attention to:
-Assurance that other historical factors were not the cause of the change in crime
-A sensible time period including a prespecification of any expected time lags
Criteria for individual level studies
Experiments and quasi-experiments (group comparisons)
Hypothesis: Subjects exposed to violent television will subsequently respond more aggressively than subjects exposed to a control condition.
Attention to:
∙\bullet Equivalent groups (ideally, random assignment)
∙\bullet Realistic viewing experience (complete program, program from normal television)
∙\bullet Appropriate control condition (if violent treatment is exciting, control must also be exciting)
Correlational studies
Hypothesis: Subjects who watch more violent television will behave more violently, controlling for related factors
Attention to:
-Measures of violent television exposure and criminal behavior that are developmentally appropriate
∙\bullet Valid measures of exposure (measures of actual frequency of exposure to programs weighted by violence ratings are best)
-Temporal order-exposure to television must precede aggression or increase in aggression
-Control for important factors that may be associated with both television exposure and antisociality-most importantly

Trait aggressiveness
Socioeconomic status (SES)
Intelligence/ability to judge reality
Popularity (when using peer-nominated aggression measures)
Experience of neglect or abuse

Table 1 (continued)
Correlational studies
Overall TV viewing (if interest is in violent viewing per se)
Parenting factors (use of television, aggression, disciplinary practices, education, etc.)

• Simultaneous entry of control factors better than one-by-one control method

Prospective longitudinal studies
Hypothesis: Children who are exposed to a greater amount of violent television over time will behave in a more criminally aggressive manner later in childhood and into adult life.
Attention to:
-A sensible time period
-Control of early wave criminal aggression (early control should match as closely as possible later measure of violent aggression)
-Sample size-must be adequate to obtain variance in criminally violent behavior
-Measurement of criminal violence at outcome
-Control for parental factors such as abuse, neglect, education, etc.

Finally, the analysis of interaction effects has become the standard in later studies because evidence suggests that the effects of television violence may be limited to persons with a predisposition for aggression, and may be opposite for others.

3. Aggregate-level studies

3.1. Cross-sectional studies

It could be argued that the most desirable studies for those interested in public health or violent crime generally would be those that examine aggregate crime levels across jurisdictions or over time to see if television viewing or violence viewing affect crime rates. Interest in rates is due, in part, to their relevance for policy. Although we may find that television violence affects some individuals in the laboratory, if there is no measurable effect on crime rates it would not be necessary to target it as a priority for reform. Unfortunately, there have been only two cross-sectional studies and four longitudinal studies that have attempted to evaluate this matter at the aggregate level.

The most careful of the aggregate cross-sectional studies (and one of the most important studies overall) was done by Messner (1986), who examined variations in the violent crime rate within the United States. Messner constructed a measure of violent television exposure for standard metropolitan statistical areas (SMSAs) within the United States based on Nielsen audience size estimates of violent shows and found that such exposure was, contrary to expectations, significantly negatively related to violent crime rates, controlling for about 10 factors that could potentially confound the relationship. The findings were robust enough to warrant explanation, and Messner hazarded a routine activities interpretation-perhaps people who watch more violent television are less likely to associate with deviant peers, to socialize with deviant subcultures, and may avoid trouble by staying at home.

Lester (1989) compared 20 countries rated for film violence and found no correlation between film violence and homicide rates. It should be pointed out that the sample size is very small, the measure of “film violence” is not described and no control variables are used. Lester’s finding is suggestive that countries high in “film violence” do not seem to have higher homicide rates-but the effect could have been suppressed by a variety of factors such as wealth that might be associated with greater film violence and less homicide.

3.2. Longitudinal studies

Four aggregate-level longitudinal studies on this topic have been published to date (Berkowitz & Macaulay, 1971; Centerwall, 1989; Hennigan et al., 1982; Phillips, 1983). These studies attempt to test whether television exposure overall or certain types of exposure affect crime trends.

Berkowitz and Macaulay (1971) tested a “contagion” effect whereby violent crimes were expected to increase after the assassination of President Kennedy in November 1963, and after the mass murder of eight nurses in Chicago by Richard Speck in July 1966. Graphs of violent crime trends suggest that unusual increments occurred in overall violent crimes, aggravated assaults, and robberies following those events. However, the data are not as convincing as the authors take them to be. First of all, the “normal” trends for these crimes vary a great deal from month to month and there was a general increase overall during this period so that the two “unusual” peaks are not especially obvious to anyone not looking for them. Second, the effect was not evident for homicide, and if the effect were really due to “contagion” or “imitation,” we would expect the effect to be more evident for homicide than for other violent crimes.

Hennigan et al. (1982) conducted a time series analysis of UCR data and concluded that the introduction of television in the United States was consistently associated with increases in larceny, but not violent crime, burglary, nor auto theft. Those findings held for both cityand state-level analyses. The authors compared crime trends for jurisdictions that received television broadcasts before the Federal Communications Commission (FCC) freeze on broadcasting licenses between late 1949 and mid-1952 to those for jurisdictions that only received television afterward. It is important to note that Hennigan et al. were looking at the introduction of television, overall, and their independent variable was not a focused measure of violence exposure. The illustrations provided in this paper do not provide the “visual analyst” a very convincing case, but larceny trends vary at a statistically significant level, in the hypothesized direction between the two types of jurisdictions. It is important to emphasize that the finding did not apply to any violent crime and the authors themselves tentatively attributed the association to the viewing of high levels of material consumption rather than increased exposure to violent portrayals (supporting a “relative deprivation” hypothesis rather than social learning).

Phillips (1983) examined daily homicide rates following heavyweight prizefights and found convincing evidence of an effect on homicide 3 and 4 days after the eventsparticularly pronounced after highly publicized fights. The most important criticism of the analysis is that the 3 - to 4 -day lag in effect was not predicted a priori-one would expect that

if enough time lags were tested, some would turn out to be significant. Furthermore, the effect was small. Phillips estimated 11.6 extra homicides in the United States per fight and did not report significance levels for monthly or annual homicide rates. During this period the number of U.S. homicides was approximately 20,000 per year, which is approximately 1667 per month, and it is probable that a change of 11 or 12 extra homicides per fight would not be detectable at the monthly or annual level.

Furthermore, one could plausibly speculate that the “extra” homicides would have occurred eventually anyway. Perhaps a few violent persons murdered someone due to imitation, excitement, frustration, or altercations associated with the fight or related gambling losses, but those persons may have been bound to kill someone at some point soon anyway (it is easy to imagine that a person who would kill in response to a prizefight might do the same if caught in a traffic jam on a hot day, fired from a job, or faced with a colicky baby at 3 a.m.). Phillips argues that this is unlikely because there are no corresponding drops in homicide below the normal level within the 3 -week periods he analyzed. However, there is no reason to believe that all the “extra” homicides were destined to occur within that period so it is highly unlikely that such a drop would be evident. Because significance of “extra” homicides is not established for monthly or yearly homicides, it is not possible to conclude there were more homicides in the long run due to prizefights than there would have been otherwise.

A widely cited study by Centerwall (1989) presented a comparison of homicide trends in the United States, Canada, and South Africa. Centerwall observed that 10-15 years following the introduction of television, the annual incidence of white homicide deaths in the United States and overall homicides in Canada increased by over 90%90 \%, while no such increase occurred in South Africa, where the government delayed television broadcasting until 1975. Like Hennigan et al. (1982), Centerwall’s independent variable-the introduction of tele-vision-is not a powerful measure of exposure to media violence. Centerwall takes this comparison as very strong evidence that television causes violent crime and contends that television is responsible for approximately 10,000 homicides annually (computing a “relative risk” based on the fact that reported homicides doubled in the 10−1510-15 years after the introduction of television and assuming, therefore, that half the nation’s homicides must be due to television). Centerwall has republished these observations repeatedly and his articles have been widely cited and accepted as important evidence for a link between television and crime.

Many strong objections could be raised from this type of analysis. First, it is important to highlight the fact that Centerwall’s (1989) measurement and analysis are not documented in any systematic way and as such, the study is not replicable and does not meet the scientific standard normally accepted in the social sciences. Second, the comparison of individual countries is plagued by a plethora of idiosyncratic, unmeasurable factors that might affect any one nation’s crime rates. In this study, no convincing case is made that the United States, Canada, and South Africa are roughly identical on the myriad social factors associated with crime that might have changed in the 1950s and 1960s. Centerwall claims that he examined a wide array of possible confounding factors (including changes in age distribution, urbanization, etc.), but he does not present the analysis or the measures used. Most troubling is the oft-cited estimate that 10,000 homicides a year in the United States are due to television,

which relies on the tenuous assumption that nothing else changed in the United States and Canada (but not South Africa) during the 1960s that could have caused homicide to increase.

In the same paper, Centerwall (1989) follows up with the observation that increases in homicide rates in nine U.S. regions corresponded to the timing of their acquisition of television. Again, no systematic description of data or analysis is presented; the correlation between the timing of the acquisition of television and the timing of the region’s subsequent increase in homicide is reported to be .82(p=.003).82(p=.003). It is unclear why regions are used instead of cities or states, and the sample size is very small. Without further documentation, Centerwall’s findings are merely suggestive and should not be weighed very heavily. Unfortunately, these findings have received a great deal of attention, have been published and cited in prominent publications, and have been interpreted as support for the conclusion that television violence causes violent crime.

Although it is contended here that Centerwall’s report has received an exaggerated amount of attention, it is the case that the United States did experience marked, astonishing increases in reported crime, particularly violent crime and homicide, starting in the early 1960s, and high rates of violent crime became the norm by the mid-1970s. This phenomenon does bear explanation and Centerwall’s suggestion that television may be responsible is an important hypothesis-not only because Americans would have been exposed to more violence, but also for a variety of reasons that range from a greater emphasis on material gain (Merton, 1938) to a decline in “small-town” informal social control. The extent to which television influenced these crime trends is yet unclear.

4. Individual-level studies

4.1. Experiments, quasi-experiments (group comparisons)

A dozen studies comparing groups in an experimental or quasi-experimental setting have been reported. Approximately four to five of these suggest that watching violent television or films is associated with “violent” or analogous behavior-but there are some serious qualifications. Five of the studies found no effect of violence exposure, and four findings suggest a negative effect-children who watched the control television programs were more violent than those who watched the violent programs. The methodological rigor among these studies is comparatively better than among those that reported the hypothesized effect.

An early study by Steuer, Applefield, and Smith (1971) matched 10 preschool children based on their usual exposure to television and exposed half of them to daily “treatments” of violent programming (mostly cartoons) and half to nonaggressive programming. Separate graphs for each of five matched pairs suggest that the children were very similar in aggressive behavior during baseline (aggressive behavior included hitting, kicking, choking, etc.), and that three of the children in the experimental condition diverged very visibly from their matched control in aggressive behavior after the introduction of the TV diet. These findings are some of the more visually convincing, but the very tiny sample size, and matching procedure (which cannot eliminate many rival hypotheses) prevents statistical inferences.

Feshbach and Singer (1971) studied boys living in three private schools and four homes for boys in California and New York. One of the earliest field experiments on this topic, the findings do not support a relationship between viewing violent television and criminal aggression. Within each institution, boys were randomly assigned to watch a diet of violent or nonviolent television over a 6-week period. Overall, there were over twice as many instances of “fistfighting, hitting, and kicking” among controls as there were among the treatment group. When the institutions were analyzed individually, the statistical significance of this finding held for all four boys’ homes (where, presumably, the boys were more troubled) but not the private schools.

Although fears have been raised by reviewers that the boys in the control condition were angry about being deprived of their favorite shows, the authors contend that they made every effort to reduce possible sources of frustration, permitting boys to drop out and even permitting a few boys to watch Batman, which was not on the list of nonviolent programs. Although one could criticize some aspects of the methodology, the very consistent negative findings across separate sites argue that a diet of violent television does not increase and may actually reduce aggression of a type that is analogous to criminality among troubled boys. This conclusion is tempered by the authors’ report that some of the boys suspected that the study was about aggression. The effect as reported suggests that either violent television reduces aggression or that aggression-inducing effects of television are very easily overcome by minimal social constraints and influences.

Hapkiewicz and Roden (1971) showed children either a violent cartoon, a nonviolent cartoon, or no cartoon and then watched pairs of them try to watch a “peep show” with only one eye hole. There were no significant differences between groups on “aggressive behaviors” ( 85%85 \% of which were “pushing”) measured during this period, although a table suggests that aggression scores were much higher for the “no-cartoon group” (the mean score of 25.8 in the no-cartoon group was approximately double that in the aggressive and nonaggressive cartoon groups). In a later study, Hapkiewicz and Stone (1974) used a larger sample and found some evidence of an effect only when “realistic” violence was shown. There was no effect of treatment on aggression for girls. The authors concluded that cartoon violence may reduce aggression.

Friedrich and Stein (1973) observed the play behavior of preschoolers in a naturalistic setting. Although they collected a large number of measures of various constructs, of particular interest here is the comparison of the impact of diets of aggressive television (Batman and Superman cartoons), prosocial television (Mr. Rogers’ Neighborhood) and neutral films on physical aggression (the measure of which, unfortunately, included some types of nonverbal aggression that are not analogous to criminality). The description of the neutral and prosocial conditions suggests that there was no effort to make them as “exciting” as the aggressive TV shows-a problem since the effects of a violent program could be due to the fact that it is “exciting.” The authors do not report a full analysis of the dependent variable of greatest interest here, but the analysis of an overall measure of aggression (which includes mostly noncriminal behavior) revealed a statistically significant interaction suggesting that the aggressive TV condition affected children who were initially high in aggression and not children who were initially low in aggression. This is

an early demonstration of an interaction effect that obtains more frequently in the later, more analytically rigorous studies.

Leyens, Parke, Camino, and Berkowitz (1975) compared four cottages of institutionalized boys in a private secondary school in Belgium. The boys lived at the school because they lacked adequate care at home or had problems with the court, their school, or their parents. Prior to the experimental treatment, boys in two of the cottages had low base rates of aggression and boys in the other two had high base rates. During a “movie week,” children in one of each of the high- and low-aggressive cottages were shown a violent film each night while an (almost certainly less exciting) neutral film was shown in the other two cottages (another important difference in the films is that the “violent” films were almost all American and the neutral films were almost all French-language films; the authors refer to them as “comedies”). Many types of aggression were studied; for present purposes the most important was “physical aggression,” defined as physical contact of sufficient intensity to potentially inflict pain on the victim (hitting, slapping, choking, kicking). Unfortunately, the observers did not distinguish between play behavior and “real” physical aggression.

Results suggested that the violent films induced aggression in the high-aggressive cottage. The “rate” of physically aggressive behavior increased from .015 to .049 from the baseline to the movie week. Since the measure of aggression includes play fighting, it is not possible to ascertain the size of the effect on “real” aggression. One can certainly imagine one or two boys wanting to imitate the movies in play, and that the violent movies were more imitable (e.g., one of the violent movies was Zorro [en garde!]). The effect only occurred in the measures taken immediately after the movie screenings and did not persist through the noontime observations made on each of the following days (which supports the concern that the effect was due to arousal or play fighting).

Huston-Stein, Fox, Greer, Watkins, and Whitaker (1981) assigned 66 preschool boys and girls to four TV conditions: high action-high violence, high action-low violence, low action-low violence, and no TV. This approach is important because by using an exciting comparison treatment, the researchers can disentangle the effect of watching something exciting versus something violent per se. Subsequent observations of free play suggested that while “serious preschool aggression” decreased from baseline in the no-television and lowaction groups, it stayed approximately the same for the high-action groups. There was no effect of violent content.

Although preschool-type aggression is not closely analogous to adult criminal behavior, one would expect it to be easier to induce using violent media stimuli. This was not the case in the Huston-Stein et al. (1981) study. The findings suggest that the excitatory nature of the material may be related to aggression (though remember, aggression in the exciting TV group did not increase-it just did not decrease), but that violent content had no effect on aggressive behavior. The decline in aggression among the “low low” and “no TV” groups suggests that the baseline measures may not have represented “typical” behavior.

In a natural experiment, Joy, Kimball, and Zabrack (1986) compared three Canadian towns (Notel, Unitel, Multitel) with varying exposure to television over the period of time during which television was introduced into the “Notel” community. Researchers observed greater increases in schoolyard aggression in Notel than in the other jurisdictions. Like Centerwall’s

(1989) work discussed earlier, this study has been cited repeatedly as evidence that exposure to television causes aggression.

There are several methodological limitations that restrict the generalizability of this study. First, the communities were not randomly assigned or “equivalent” in the beginning. The authors state that Notel (no television) was typical and similar to the comparison community Unitel (one television station) and suggest that census data and resident reports corroborate this statement (detail is not provided). A convincing case is made that Notel was not an unusual town; the reason Notel had no television was simply because it was in a valley, a “geographic blind spot,” and the transmitter for the area could not reach it. What is puzzling is that the youngsters in the three towns had similar levels of aggression during baseline. If television causes aggression, Multitel (multiple television stations) should have had much higher levels of baseline aggression, followed by Unitel. Clearly, something else was different about the Notel children-or, television does not cause aggression.

Also, the dependent measure included a variety of aggressive schoolyard activities, some of which were analogous to criminality but many of which were not. One has to wonder to what extent imitative play fighting played a role in Notel’s increase in aggression. The findings of this study are suggestive that the introduction of television was associated with increased obnoxious playground behavior. It would be more instructive if actual crime counts in the three towns could be compared.

Violent media exposure is just one of numerous predictors used by Kruttschnitt, Heath, and Ward (1986) to distinguish between violent inmates and a matched sample of nonconvicted men. There were significant differences between groups on a retrospective measure of exposure to violent TV. The authors themselves do not put much confidence in the findings because the size of the coefficient is smaller than those for a variety of other variables. However, the finding does favor an effect of violent media exposure. The matched design, however, has many drawbacks, the most important of which is that predispositions for aggression were not controlled so the alternative hypothesis that aggressive kids like to watch violent shows has not been eliminated. Heath, Kruttschnitt, and Ward (1986) expanded on this analysis and found an interaction effect-that “high exposure to television during childhood years was related to the commission of a violent crime during young adulthood if violence was also present in the home” (p. 186).

Josephson (1987) marked a turning point in this line of research-her methods are a marked improvement over prior studies of this type. Josephson randomly assigned boys to watch a violent or nonviolent show and further manipulated frustration and violent cues. Her primary interest was in the action of aggressive cues and frustration but she does report that, controlling for initial aggression, frustration, and a variety of interactions, the main effect of TV violence on floor hockey aggression was negative; the boys who watched violence were less aggressive as a group than those who watched a control program. She tested interaction effects and found that violent TV content was associated with higher aggressive behavior among groups with relatively high average scores on initial characteristic aggressiveness. She further points out that “In most cases, the very opposite effects occurred among groups of boys with low mean levels of characteristic aggressiveness” (p. 888). Again, the interaction between trait aggressiveness and viewing violence appears to be important.

Sprafkin, Gadow, and Grayson (1987,1988)(1987,1988) used learning-disabled (LD) and emotionally disturbed (ED) children as subjects. The authors made some effort to use exciting control cartoons (Lassie’s Rescue Rangers) and they found no main effect of condition (violent cartoon vs. control cartoons) on short-term physical aggression in LD kids. Among the emotionally disturbed children there was significantly more physical aggression during recess following control cartoons. They also found an interaction such that low-IQ kids behaved more aggressively after the control cartoon.

An unusual aspect of the design was that all the children saw both violent and nonviolent cartoons in randomized order from one day to the next so that any effects the researchers detected would have to be short term. If witnessing violent cartoons has an effect that persists for days, for example, this type of design would miss it.

In summary, the group comparisons reviewed here provide very little evidence that viewing violent media is positively associated with spontaneous, physically aggressive behavior in natural settings. Of the four studies reporting a positive effect, only one did not have serious doubts cast on the findings by the authors or a critical review. Numerous studies of greater methodological soundness have found null effects and even negative effects of viewing violent media on violent behavior.

It is important to point out that this set of studies were short term in nature and do not provide evidence related to long-term effects of a diet of violent television or film exposure. It is not necessary for short-term effects to be evident for there to be a long-term effect on behavior.

4.2. Correlational research

Eron’s (1963) early study was the starting point for a long series of longitudinal research conducted by Eron, Huesmann, and Walder (1972) (discussed in the next section). The original study, not originally designed to test the effects of television, found a correlation between peer-rated aggression among third graders and the violence rating of the subjects’ three favorite television shows. There was also a significant negative relationship between the number of hours of television viewed (estimated by the mother but not corroborated by father’s estimates) and aggression. Only simple correlations were reported and no control variables were used. This finding was in many ways a springboard for several decades of research on this association. By today’s standards, however, the findings leave us wanting; like many other early correlational studies temporal order has not been established (it could be that aggressive children simply like violent programs - not that violent programs caused their aggression), and because no controls were used, we must wonder if the relationship is spurious.

The other purely correlational studies can be condensed and summarized because they were very similar to one another in design, and their findings are not ambiguous. Unlike the other types of studies reviewed here, most of the correlational studies find a significant positive correlation between exposure to or preferences for violent television and measures of delinquency or physical aggression (Belson, 1978; McIntyre & Teevan, 1972; McLeod, Atkin, & Chaffee, 1972a, 1972b; Robinson & Bachman, 1972; Thornton & Voigt, 1984).

Like some studies reported earlier (Friedrich & Stein, 1973; Josephson, 1987), Robinson and Bachman (1972) found that the relationship between preference for violent television and aggression was only evident among the group highest in prior aggression.

Some of the correlational results do not support a significant relationship between viewing violence and violent behavior. A reanalysis of the McIntyre and Teevan (1972) data found that overall television viewing was associated with violent behavior for girls but not boys (Hartnagel, Teevan, & McIntyre, 1975). The authors conclude, “We are forced, then, to conclude that the TV violence predictors, both objective and perceived, do not matter significantly in explaining violent behavior” (p. 347). McCarthy, Langner, Gersten, Eisenberg, and Orzeck (1975) found that “longer viewing hours” was associated with higher scores on fighting and delinquency but not violence of preferred programs.

Of the four types of studies used as categories in the present paper (aggregate, group comparisons, correlational and prospective longitudinal) the pattern of findings among correlational studies appears to provide the greatest amount of support for the hypothesis that exposure to violent television causes criminally aggressive behavior but it is very weak support. The finding of a correlation between viewing violence or a preference for it and aggressive behavior is consistent with the hypothesis that exposure to television violence causes aggression, but it is also consistent with the hypothesis that aggressive children choose to watch violent programs. Without a proper control for “trait aggressiveness” in any of these studies, the causal order cannot be determined. These studies should not be weighed heavily in an assessment of the evidence even though there are so many of them.

4.3. Prospective longitudinal studies

The prospective longitudinal design is the most relevant to the most prominent explanation for media effects-that a diet of violent television and movies over time contributes to the aggressive socialization of a child who will, over a long period of time, develop aggressive habits (cf. Huesmann, 1986a; Huesmann & Miller, 1994; Huesmann, Moise, & Podolski, 1997). Several prospective longitudinal studies have been cited widely to support the hypothesis that viewing violent television causes criminality.

Although prospective longitudinal studies probably are among the highest quality of those evaluating media violence, most of them use, as their dependent measure, an unsatisfactory measure of aggression and, therefore, their findings should not be overemphasized. The measure used is “peer-nominated aggression” obtained by asking children in a classroom to nominate which of their classmates behave antisocially (see Table 2). Each subject’s score is based on the number of other children who nominated him or her for the items.

The only items that are relevant for our purposes are “Who starts a fight over nothing?” (which could be physical) and “Who pushes or shoves other children?” Unfortunately, these items are embedded in a scale dominated by far less relevant items (see Table 2). Of most concern is that the measure is unlikely to identify the most violent children, because there are no items that reflect severe, but infrequent, violence. It may also miss children who are highly aggressive at home, with siblings, for example, but who are not aggressive at school. The child who bullies his little sister may appear to be low in aggressiveness while a little girl who

Table 2
Items from a peer-nominated aggression measure

1. Who does not obey the teacher?
2. Who often says, “Give me that?”
3. Who gives dirty looks or sticks out their tongue at other children?
4. Who makes up stories and lies to get other children into trouble?
5. Who does things that bother others?
6. Who starts a fight over nothing?
7. Who pushes or shoves other children?
8. Who is always getting into trouble?
9. Who says mean things?
10. Who takes other children’s things without asking?

From Huesmann and Eron (1986, p. 32).
is verbally aggressive may score high on this measure. In spite of this problem, as a group, the set of longitudinal studies are the most methodologically rigorous and have some potential for elucidating this phenomenon, therefore we include the three main sets of longitudinal studies and a more recently reported study here.

4.4. The Eron/Lefkowitz/Huesmann et al. studies

The first set of studies was conducted by Eron and his colleagues starting with third-grade subjects and was initially reported in 1963 (this study was discussed earlier among the correlational findings (Eron, 1963)). Those subjects were followed up at age 13 (reported in Lefkowitz, Eron, Walder, & Huesmann, 1972), age 19 (Lefkowitz, Eron, Walder, & Huesmann, 1977) and finally, when they were 30 years old (Huesmann, 1986a).

Controlling for initial levels of aggression (an important milestone in the present line of research), Lefkowitz et al. (1972) reported a partial correlation of .25 (significance not reported) between age 8 preference for violent programs and age 13 peer-nominated aggression. This approach, used in other studies as well, predicts change in aggression over a period of time using information related to violent television exposure. The authors went on to report a multivariate regression, but the procedure was run “stepwise” and thus is of little use since "television violence preference at age 8 " was entered into the model first, not effectively controlling for the other factors. Although the study does not contradict the authors’ hypothesis, its support is weak since alternative hypotheses were not effectively eliminated.

Lefkowitz et al. (1977) followed up the same subjects at age 19. The peer-nominated aggression measure is used again and is even less satisfying because the subjects were no longer in school together and may have been basing their responses on impressions from years past. Furthermore, the peer-nominated aggression items seem less relevant to a young adult sample than they were for a child sample (e.g., “Who does not obey the teacher?” “Who makes up stories and lies to get other children into trouble?”). The authors found a significant relationship between preference for violent television at age 8 and aggression at age 19 , controlling for a variety of factors one by one. Most important among these factors

was peer-rated aggression at age 8 . This study was among the earliest to attempt to isolate the effects of television violence on change in aggression and to establish the temporal order of the effect. The authors also controlled for other measures from age 8 (one by one): father’s occupation, father’s aggressiveness, child’s IQ, mother’s aggressiveness, parental punishment of child, parents’ mobility orientation, hours of TV watched, and a series of factors measured contemporaneously (age 19) including father’s occupation, subjects’ aspirations, and hours of watching television.

The estimated correlation between early preference for violent television and later aggression is .31 and is almost certainly larger than it would be if all the control variables were entered simultaneously. For present purposes, the finding is highly suggestive, but the outcome is not as close to “criminal” conduct as one would like. Based on such a finding, it would still be speculative to say that television violence would have a significant net effect on criminal aggression. This is the most important finding reported to date but not terribly satisfying-nevertheless, reviewers have interpreted it as convincing evidence that exposure to television violence causes violent crime.

Huesmann (1986a) briefly reports a 22-year follow-up of this study 1{ }^{1} in a theoretical article that presents a model of media effects on behavior. Figures are presented briefly at the end of the article and they appear to demonstrate correlations between early preferences for television violence, frequency of TV viewing, and the number and seriousness of convictions at age 30 . This would qualify as a very important finding, but the details of measurement and analysis are not described and no freestanding empirical article reporting this finding has been published to date. 2{ }^{2} Nevertheless, the finding has been emphasized dramatically by reviewers.

In summary, there are several problems with overstating the correlations discovered by the Eron/Lefkowitz/Huesmann group in this line of research-first, the initial measure of exposure to violent television is limited to preferences for violent shows-not exposure to violent television. The most important finding is that early preferences for violent television are associated with peer-rated aggression at age 19, controlling for initial levels of aggression. Unfortunately, the measure of aggression includes mainly obnoxious, irritating behaviors and few unlawful ones so generalization to criminal violence is tenuous. The most anticipated finding was one where early-wave aggression was related to seriousness of conviction record at age 30, controlling for earlier peer-nominated aggression. The casual reporting of the finding invites skepticism. Even if accepted as stated, there are some alternative explanations that prevent a great deal of confidence in that finding. In particular, it is quite likely that using peer-nominated aggression in third grade as a control is not appropriate when predicting later

[1]

1. 1{ }^{1} Although studies by Huesmann, Eron, Lefkowitz, and Walder (1984) and Huesmann, Lagerspetz, and Eron (1984) have both been cited as the 22-year follow-up of this study, said to report a relationship between preferences for violent TV at age 8 and seriousness of criminal convictions at age 30 , neither study presents any such comparison (various published misattributions exist, suggesting that reviewers are citing material they did not read).
   2{ }^{2} It has been reported in the popular press (Rhodes, 2000) that the correlation between age-8 TV violence viewing and adult violent crime was due to three subjects who were the only subjects to have committed violent crimes; all three had scored high on age- 8 TV violence viewing. ↩︎

criminality instead of later peer-nominated aggression. It is likely that such a measure “misses” children with serious antisocial tendencies who would not have been identified by the early measure. Still, this finding is of keen interest if replicated with a larger sample and valid measures of violent behavior would provide part of the basis for a convincing argument that TV violence is an important cause of violent crime.

4.5. The Milavsky et al. study

Though reviews of the literature repeatedly allude to the authors as researchers for a major television network presumably to discredit their work, Milavsky et al. also report a rigorous test of the impact of television violence on aggressive behavior (Milavsky, Kessler, Stipp, & Rubens, 1982a, refers to their portion of the NIMH report; Milavsky, Stipp, Kessler, & Rubens, 1982b, refers to their book).

Elementary school boys and girls and teenaged boys were used to test the hypothesis that television violence viewing measured earlier in the study would predict aggression and delinquency measured later in the study, controlling for prior aggression and other factors. The authors present multivariate models controlling for SES and other variables including “stability” of aggression (Wave 1 aggression) that do not indicate a significant relationship between early, violent TV viewing and later aggression for boys. They report a small but significant effect for girls. The authors explored the data in great detail for curvilinearity, for effects among subgroups (such as those based on income, occupation, city, family size, fatherhood involvement, strictness, etc.) and found no evidence, above what we might expect by chance in such a large number of analyses, for a relationship. The outcome measure was a modified version of the peer-nominated aggression measure Huesmann and his colleagues had previously used-Milavsky et al. used factor analysis to create the final index, which included four items (Who tries to hurt others by saying mean things to them? Who makes up stories or lies to get someone else into trouble? Who tries to hurt others by pushing or shoving? Who hits or punches other people to hurt them?); this index is somewhat more satisfactory than those used by the Eron/Lefkowitz/Huesmann group (Lefkowitz et al., 1977) because violent behaviors such as pushing, shoving, and punching are weighted more heavily.

One important difference between the Milavsky et al. study and the other longitudinal studies is that the authors chose to control for early-wave violent TV exposure as well. By partialling out variability in initial TV habits, the authors are essentially testing whether the exposure specific to the study period affects the outcome variables. This is a much more conservative test of the hypothesis than is reported in other studies and would make it more difficult to detect an effect-in particular if the appropriate age group is not used.

In a more relevant set of analyses (because the outcome is criminal in nature), Milavsky et al. (1982b) examined onset of delinquency for the teenage sample who were asked during each wave of data collection if they had committed any of a series of delinquent acts (badly beaten someone up, gotten arrested, been in a gang fight, stolen a car, knifed someone). Because of an unfortunate anomaly with the incidence rates, the authors chose to see if they could predict onset of delinquency using TV violence exposure rather than using another measure of delinquency. Although a global significance test of a large number of comparisons

suggested that exposure to television violence was not related to onset of delinquency, a visual inspection of the pattern of findings reported certainly appears to provide some evidence for a positive effect of exposure to violent television on onset of delinquency. The probabilities of onset of delinquency increased from .09 for teenagers in the 5th percentile of TV violence exposure to .18 for those in the 95 th percentile. There were other similar patterns also reported, but Milavsky et al. choose to rely on significance tests and conclude that their data do not support an effect.

A problem with the design in this particular comparison, though, may have resulted in an exaggeration of the effect-Milavsky et al. did not control for “trait aggressiveness” or some other indicator of initial variability in the propensity for delinquency in this comparison. They argue that a control was not necessary because the boys chosen for the “onset” analysis had not committed a delinquent act as of the first wave and therefore did not “vary.” It could easily be argued, however, the subjects did vary in their propensity to become delinquent or in trait aggressiveness and the authors should have at least used the peer-nominated aggression measure from the early wave as a control for personality differences (though this is not completely satisfactory as argued earlier, it was the best measure they had for the purpose). Based on the other findings where such a control nullified some significant simple correlations, one would expect it to reduce the visually evident (but not statistically significant) effect here too.

4.6. The Huesmann, Eron, et al. cross-national studies

Because of some dissatisfaction with the earlier series of studies, Huesmann and colleagues organized a replication of the Eron/Lefkowitz/Huesmann study, designed especially to test whether or not exposure to violent television had independent effects on later aggression. A well-designed, very ambitious project, Huesmann and Eron ran the study in the United States while it was being replicated simultaneously in four other countries: Australia (Sheehan, 1986), Finland (Lagerspetz & Viemero, 1986), Israel (Bachrach, 1986), and Poland (Fraczek, 1986). The studies are all reported in one volume.

Like Milavsky et al. (1982a, 1982b), Huesmann and Eron, 1986 used a multiple-wave design. In each country two cohorts of children-one in first grade and one in third gradewere followed up for 3 years. The findings of most interest for Americans are those reported by Huesmann and Eron (1986) based on the sample of children in the United States. They found the expected positive simple correlations between television violence viewing and aggression for boys and girls in almost all comparisons. More importantly, though, controlling for initial aggression, TV violence viewing in the first and second waves had a significant positive correlation with third-wave aggression for girls but not boys. This is the effect of primary importance and it did not hold for boys. One could take this as strong evidence, given the high quality of the study, that exposure to TV violence is not an important predictor of aggression for boys, who are of most interest because they are disproportionately involved in violence.

The authors followed up by creating a second version of the independent variable-a multiplicative composite of TV violence viewing and a measure of "identification with

aggressive characters" (which had demonstrated a very high correlation with aggression). They found that the composite was significantly, positively related to Wave 3 aggression, controlling for Wave 1 aggression for boys. It is unclear why this is a surprise, because Huesmann has elsewhere convincingly reported that aggression is a stable trait and “identification with aggressive characters” is correlated with it. It is also unclear why the composite measure is seen as a valid substitute for the purer TV violence exposure variable. The modified independent variable is a mixture of exposure to violent television and a personality-related measure and, thus, we cannot know the extent to which aggression in the later wave is due to TV or personality (in fact, we must suspect it is due to personality since the correlation for identification was high and the partial correlation for TV violence was null). The follow-up finding has been widely cited as evidence for the effect of violent television on aggressive behavior, in spite of the fact that the original analysis suggested that violent television did not cause aggression in the preadolescent boys in the sample.

Furthermore, in data from the four other countries where the study was conducted, in no case was the model of most interest based on the study’s original design (Wave 3 peernominated aggression == Wave 1 aggression + Wave 1 and 2 TV violence viewing + covariates) reported as statistically significant. In fact, most of the authors did not report this model at all. Huesmann and Eron’s (1986) finding of a significant association between TV violence exposure and later aggression for girls was unique among all the findings reported. Yet the international findings have also been taken to support the hypothesis that viewing violence causes aggression because of findings reported using “modified” variables. In Finland, Lagerspetz and Viemero (1986) used the product of television violence viewing and identification with aggressive characters as their independent variable and found a significant relationship between this composite variable and Wave 3 aggression. Fraczek (1986), reporting the Polish study, uses “violence of preferred shows” in Waves 1 and 2 as the independent variable and finds that this is not significantly related to third-wave aggression (though the authors and subsequent reviewers characterize the finding as “marginally significant,” P<.10P<.10 ).

The Israeli study is reported by Bachrach (1986), who uses the proper independent variable-television violence viewing in Waves 1 and 2, but instead of using Wave 3 aggression as the dependent variable, he reports findings for “the ratio of aggression to avoidance of aggression” for reasons unexplained. Controlling for cohort, initial aggression, initial peer aggression avoidance, the multiple regression reveals a positive and statistically significant relationship between early television violence viewing and this ratio. Finally, Sheehan (1986) does compute the model anticipated by the study’s design and reports that television violence exposure did not have significant effects on aggression among the Australian children. This is the only study that limits itself to reporting the expected multiple regression model for both sexes using the pure peer-nominated aggression and TV violence viewing variables. It is also the only study that reports completely null effects.

Huesmann’s (1986b) own summary assessment of the evidence from these studies suggests there is a consistent, convincing pattern that supports the TV violence viewing-aggression relationship. A more critical evaluation would conclude that there is very little evidence from this set of studies that exposure to television violence causes aggressive behavior. There was

no relationship between early-wave TV violence viewing (measured in its most appropriate form) and later wave peer-nominated aggression for any subject grouping except the American girls. 3{ }^{3}

4.7. Johnson et al. (2002)

A very recently released study by Johnson, Cohen, Smailes, Kasen, and Brook (2002) was published in Science and reported on the front page of the Washington Post. This study collected self-report and maternal report data beginning in 1975 in a study of children in the community. The independent variable was hours of television viewing per day, divided into three groupings ( <1 h,1−3 h,>3 h<1 \mathrm{~h}, 1-3 \mathrm{~h},>3 \mathrm{~h} ). This is a measure of overall television viewing, not violence viewing per se, so the findings should be interpreted accordingly. The authors found that net of controls, there was a higher prevalence of assaults, robberies, and aggressive acts more generally among individuals who watched more than 1 hour of television in adolescence (average age 14) 2 to 8 years later when they were young adults (measurements were made when their average age was 16 and again at 22) than among individuals who watched less than 1 hour of television. Though not statistically significant, it was clear that prevalence rates were also higher among those watching more than 3 h than 1−3 h1-3 \mathrm{~h}. Furthermore, the prevalence of violent behaviors at average age 30 (ages ranged from approximately 25 to 35 ) among those viewing more than 3 h per day as a young adult (average age 22) were higher than among those who watched less than 3 h of television a day. Some, not all, of the findings held when the authors controlled for previous and subsequent television viewing.

The analysis is difficult to evaluate thoroughly because the article is short and does not contain a great deal of detail. Furthermore, it was difficult to find the detail in other works cited. First, the subject ages seem somewhat inappropriate for testing the theory that has received the most attention in recent years-that a diet of television violence leads to violent behavior in children. Many studies suggest that aggressive tendencies are established rather early in life and the earliest measures in this study were taken when subjects were an average age of 14 years old (ages ranged from about 9 to 19). Even Huesmann would probably argue that this is too late to test his theory. The controls used in the first analysis are listed as childhood neglect, growing up in an unsafe neighborhood, low family income, low parental education, and psychiatric disorder. It is likely that these are dummy coded and, thus, lose some of their power. For example, the measure of childhood neglect is based on information from a central registry and retrospective self-reports and may not be adequate for controlling for the subtleties of parental monitoring and supervision. Another concern is that parental harshness or abuse is not controlled. The addition of a control for neglect is a welcome one, but it is not clear that this measure has adequate validity for the purpose.

[1]

1. 3{ }^{3} Huesmann et al. are preparing a report of the longitudinal follow-up of these children. The published report is currently limited to abstracts from a symposium on “The effects of childhood aggression and exposure to media violence on adult behaviors, attitudes, and mood” (Huesmann, Moise, Podolski, & Eron, 1999). ↩︎

Previous research has been very consistent in suggesting that aggressive children prefer violent television so it is very important in any study of this type to control for early tendencies for violent aggression if the outcome to be examined later is violence. The only control used here is “psychiatric disorder” based on the Diagnostic Interview Schedule for Children (DISC-I). Although the authors do not explain the measure in full, it appears that a dummy code was used for presence or absence of psychiatric symptoms (including aggressive symptoms). This very broad measure is really not adequate for controlling for early violent behavior.

The second set of comparisons that were reported examining the effects of television viewing in young adulthood (average age 22) are a little more ambiguous. The authors report that control factors were used but do not specify which ones. The control factors used in the prior analysis seem inappropriate (it does not make sense to control for neglect among subjects aged 17-27), and because violent behaviors were measured in young adulthood, this measure should have been used as a control. There are additional problems with this comparison - at this age, of course, watching a great deal of television could be associated with unemployment, depression, intelligence, whether the subject has children or not, whether the subject is a high school dropout or not, marital problems, and a variety of factors that could be related to violence and are unmeasured and uncontrolled here.

Although the Johnson et al. (2002) findings suggest increased violence among persons who watch more television, and add to the literature by suggesting this correlation lasts into adulthood, they do not add significant evidence in favor of the thesis that viewing violent television causes violent behavior.

In summary, although findings from the prospective longitudinal studies have been taken as strong evidence that viewing violence causes violent aggression, a careful reading suggests that evidence for an effect on criminal behavior is practically nonexistent and the evidence for an effect on aggression is very weak at best. If peer-nominated aggression is proximate for violent behavior, as is presumed by many, the evidence suggests, instead, that viewing violent television does not affect it significantly.

5. Summary and conclusions

Table 3 displays a tightly summarized version of what has been discussed throughout this paper. Although there are numerous positive effects evident in the table, they are concentrated among studies of least methodological relevance for studying the effect of television violence on criminally violent behavior (see Table 3). These include the correlational studies (which, for the most part, do not establish temporal order and have inadequate controls for spurious factors) and the prospective longitudinal studies that have, for the most part, relied on peernominated aggression as the outcome and that have not reported consistent significant effects based on the statistical model implied by the original design of those studies.

Although a tally of the findings summarized in Table 3 is not a perfect way to compare effects, it is a quick way to get a sense of the findings reported. If we limit ourselves to studies of “highest” and “medium” relevance for testing this hypothesis, we find 23 markings.

Table 3
Summary of study findings by methodological relevance for testing hypothesis that exposure to television violence causes criminally violent behavior

Table 3 (continued)

O Overall null effect.

• Overall positive effect of exposure to TV violence indicator on aggressive behavior measure.
• Overall negative effect of exposure to TV violence measure on aggressive behavior measure.

⊕\oplus Interaction effect-exposure associated with increased aggression for subjects high in trait aggressiveness.
a{ }^{a} Based on statistical test that best tests the research question of interest here. These are intended to summarize findings for study and do not necessarily reflect the conclusions drawn by the researchers themselves.

Of these, seven summary findings report a positive effect but three of those are for girls only. Four summary findings report a negative effect (more media violence, less violent behavior). Nine findings are null and three reflect an interaction such that viewing violence had a positive effect on those already high in trait aggression. On the balance, for boys, there appears to be no more evidence for a positive effect than there is for a negative effect of media violence on violent behavior. Although it could be the case that most of the studies missed the effect due to methodological limitations, it is not appropriate nor is it common practice to conclude that the effect must have been missed in those studies. What is common practice is to evaluate the methodology of studies that report significant findings, see if there are rival hypotheses, and temper our conclusions to the extent that there are. Of the “high” and “medium” relevance studies reporting positive findings we find a time lag that was not prespecified, a matched design with potential for spuriousness, and a lack of control for prior aggressiveness-all very significant problems that without further study mitigate against our confidence in these findings. Of course this conclusion would be different if we accept the interpretations of the prospective cross-national studies provided by their authors, which have not been accepted at face value in this review.

Schramm, Lyle, and Parker (1961), in the very first lines of one of the first full-length studies of television and North American children, wrote that “No informed person can say simply that television is bad or that it is good for children. For some children, under some conditions, some television is harmful. For other children, under the same conditions, or for the same children under other conditions, it may be beneficial. For most children, under most conditions, most television is probably neither particularly harmful nor particularly beneficial” (p. 1). This early conclusion probably holds today.

Unfortunately for the serious scholar, most published reviews and discussions of this topic frequently cite conclusions of authors without addressing the inadequacies of the research that produced them. At present, it is safe to say that the research has established several correlations. Children with aggressive tendencies tend to prefer violent programming more than children without them. Some studies find a correlation between exposure to violent programming and aggressive behavior. However, that correlation may be due to a variety of spurious factors that have not been adequately eliminated. In the laboratory, subjects who view violent material often behave more aggressively, on average, than those who did notbut mild aggression is typically measured as an outcome, and demand characteristics probably exaggerate this effect beyond applicability to criminal behavior.

However, saying there is a lack of evidence supporting a hypothesis does not mean there is convincing evidence that the null hypothesis is true and correct. As an early reviewer of this paper pointed out, it is important to acknowledge the methodological challenge of measuring the long-range impact of exposure to violent television. Criminality is rare and it would take very large samples of the general population to obtain enough variance to reliably predict it. Effects are probably interactive or indirect (e.g., exposure could change attitudes or beliefs, making one more prone to violence at a later date), and these effects are hard to measure. Exposure could change the way one raises one’s own children, could influence the choice of whether or not to buy a gun, or keep it in the house. If the effect operates in these complex ways, it is not surprising that it is not obvious in the current body of literature. Furthermore,

considering that most of the research discussed here used violent portrayals that are not nearly as violent as the material available today, and the research paradigms do not test a steady diet of these programs (more likely since the advent of video recorders and cable television), the effect sizes may be weaker than what we might get today if we conducted the same studies again.

Huesmann, in numerous papers has proposed a very compelling theory about why and how TV violence may influence aggressive behavior (e.g., Huesmann, 1986a; Huesmann & Miller, 1994; Huesmann et al., 1997). Given the logic of his and other social learning theories, it would be hard to deny that television could have an effect on behavior. Any parent has witnessed the obvious and direct effects of “The Power Rangers” or “The Three Stooges” on the play behavior of his or her children. But the question addressed here is not whether or not the effect is plausible, but whether the effect has been demonstrated convincingly in the scientific literature-and the answer is “not so far.” Furthermore, it is also quite plausible that the role that a child’s biological status, parents, neighborhood, schools, and salient features of the child’s environment overwhelm any effects that viewing television is likely to have on all but a few very neglected children. As Jenkins (1999) put it, “Real life trumps TV every time.” And if the effect is limited to a few trait-aggressive or neglected individuals (and evidence of an interaction of this type is mounting) it may not be appropriate to focus policy on television violence, but on the causes of early childhood aggression and parental neglect.

Future studies of this matter could, first, make use of old data and reanalyze some of the early experimental studies on aggression to see if there are interaction effects. There is emerging evidence that the effect may exist for highly aggressive children and this bears further exploration. Second, a well-designed longitudinal study, using a criminal outcome and adequate controls would be of keen interest. An important component would be a good baseline measure of violent behavior (not simply aggressive tendencies) as a control. Such a study would require clever sampling to ensure variance in the outcome (criminal violence), for example, oversampling of high-risk children. If Huesmann (1986a) were correct about the dynamics of this process, starting with a younger sample would also be very useful, because “trait” aggression is evident in young children. It is of paramount importance to distinguish between violent and property offending because this has important implications for the theories being tested and, of course, it is essential that the measure of exposure be very specific (e.g., an estimate of television viewing weighted by violence rating of preferred programs as has become the norm). Control factors should include those used by previous researchers with an emphasis on validity of the control for early violence and parental abuse, neglect, education, and supervision.

At the aggregate level, which is interesting because of its implications for policy, studies replicating that of Messner (1986) would be useful. This type of study requires a good measure of violent conduct, a good measure of violence viewing, and controls for factors related to both, such as population demographics, unemployment, poverty, inequality, routine activities, education, and the like. Admittedly, because the role of individual-level factors in the etiology of crime rates are probably limited due to many characteristics of a social environment that affect crime rates at aggregate levels, the prospects for discovering media violence effects on aggregate crime are not strong.

Because legislators and other policymakers make frequent calls to reduce media violence, this line of research, spanning over 40 years now, is still relevant and topical and bears further scrutiny. At this point it must be said, however, that there is little evidence in favor of focusing on media violence as a means of remedying our violent crime problem.

Acknowledgements

This paper was originally inspired by an invitation to present a talk to the Free Expression Network at the Freedom Forum in Rosslyn, Virginia, about the criminologist’s perspective on the causes of violence. I wish to thank Chris Finan and Amy Isbell of the Freedom of Expression Network and Paul McMasters of the Freedom Forum for their invitations and L. Rowell Huesmann for correspondence regarding his work.

References

Albiniak, P., & McConnell, B. (1999). TV to blame, Hatch report says. Broadcasting and Cable, 129(34), 19. American Medical Association (1993). Television violence. JAMA, 270(23), 2870.
Anderson, C. A., & Bushman, B. J. (2002). The effects of media violence on society. Science, 295, 2377-2379, March 29.
Bachrach, R. S. (1986). The differential effect of observation of violence on Kibbutz and city children in Israel. In L. R. Huesmann, & L. D. Eron (Eds.), Television and the aggressive child: A cross-national comparison (pp. 201-238). Hillsdale, NJ: Erlbaum.
Bandura, A. (1977). Social learning theory. Englewood Cliffs, NJ: Prentice-Hall.
Bassett, H. T., Cowden, J. E., & Cohen, M. F. (1968). The audio-visual viewing habits of selected subgroups of delinquents. The Journal of Genetic Psychology, 112, 37-41.
Belson, M. (1978). Television violence and the adolescent boy. Hampshire, UK: Saxon House.
Bergenfield, M. (1994). Media violence. Parents, 69(10), 40-45.
Berkowitz, L., & Macaulay, J. (1971). The contagion of criminal violence. Sociometry, 34, 238-260.
Centerwall, B. S. (1989). Exposure to television as a cause of violence. In G. Comstock (Ed.), Public communication and behavior (pp. 1-58). New York: Academic Press.
Chatfield, J. E. (2002). Influence of media violence on children. American Family Physician, 65(4), 735-736.
Donnerstein, E., & Linz, D. (1995). The media. In J. Q. Wilson, & J. Petersilia (Eds.), Crime (pp. 237-266). San Francisco, CA: Institute for Contemporary Studies Press.
Eron, L. D. (1963). Relationship of TV viewing habits and aggressive behavior in children. Journal of Abnormal and Social Psychology, 67, 193-196.
Eron, L. D., Huesmann, L. R., Lefkowitz, M. M., & Walder, L. O. (1972). Does television violence cause aggression? American Psychologist, 27, 253-263.
Felson, R. B. (1996). Mass media effects on violent behavior. Annual Review of Sociology, 22, 103-128.
Feshbach, S., & Singer, R. D. (1971). Television and aggression: An experimental field study. San Francisco: Jossey-Bass.
Fraczek, A. (1986). Socio-cultural environment, television viewing, and the development of aggression among children in Poland. In L. R. Huesmann, & L. D. Eron (Eds.), Television and the aggressive child: A crossnational comparison (pp. 119-160). Hillsdale, NJ: Erlbaum.
Freedman, J. (1986). Television violence and aggression: A rejoinder. Psychological Bulletin, 100, 372-378.
Freedman, J. (1988). Television violence and aggression: What the evidence shows. In S. Oskamp (Ed.), Applied social psychology annual: Television as a social issue, vol. 8 (pp. 144-162). Newbury Park, CA: Sage.

Freedman, J. (1992). Television violence and aggression: What psychologists should tell the public. In P. Suedfeld, & P. E. Tetlock (Eds.), Psychology and social policy (pp. 179-189). New York: Hemisphere.
Freedman, J. L. (1984). Effect of television violence on aggressiveness. Psychological Bulletin, 96(2), 227−246227-246.
Friedrich, L. K., & Stein, A. H. (1973). Aggressive and prosocial television programs and the natural behavior of preschool children. Child Development Monograph, 38(4).
Grossman, D. (2000). Teaching kids to kill. National Forum, 80(4), 10-14.
Hapkiewicz, W. G., & Roden, A. H. (1971). The effect of aggressive cartoons on children’s interpersonal play. Child Development, 42, 1583-1585.
Hapkiewicz, W. G., & Stone, R. D. (1974). The effect of realistic versus imaginary aggressive models on children’s interpersonal play. Child Study Journal, 4, 47-58.
Hartnagel, T. F., Teevan Jr., J. J., & McIntyre, J. J. (1975). Television violence and violent behavior. Social Forces, 54,341−35154,341-351.
Heath, L., Kruttschnitt, C., & Ward, D. (1986). Television and violent criminal behavior: Beyond the Bobo doll. Victims and Violence, 13, 177-190.
Hennigan, K. M., Del Rosario, M. L., Heath, L., Cook, T. D., Wharton, J. D., & Calder, B. J. (1982). Impact of the introduction of television on crime in the United States: Empirical findings and theoretical implications. Journal of Personality and Social Psychology, 42, 461-477.
Huesmann, L. R. (1986a). Psychological processes promoting the relation between exposure to media violence and aggressive behavior by the viewer. Journal of Social Issues, 423, 125-139.
Huesmann, L. R. (1986b). Cross-national communalities in the learning of aggression from media violence. In L. R. Huesmann, & L. D. Eron (Eds.), Television and the aggressive child: A cross-national comparison (pp. 239-258). Hillsdale, NJ: Erlbaum.
Huesmann, L. R., & Eron, L. D. (1986). The development of aggression in American children as a consequence of television viewing. In L. R. Huesmann, & L. D. Eron (Eds.), Television and the aggressive child: A crossnational comparison (pp. 45-80). Hillsdale, NJ: Erlbaum.
Huesmann, L. R., Eron, L. D., Lefkowitz, M. M., & Walder, L. O. (1984). Stability of aggression over time and generations. Developmental Psychology, 206, 1120-1134.
Huesmann, L. R., Lagerspetz, K., & Eron, L. D. (1984). Intervening variables in the TV violence-aggression relation: Evidence from two countries. Developmental Psychology, 205, 746-775.
Huesmann, L. R., & Miller, L. S. (1994). Long-term effects of repeated exposure to media violence in childhood. In L. R. Huesmann (Ed.), Aggressive behavior: Current perspectives (pp. 153-186). New York: Plenum Press.
Huesmann, L. R., Moise, J. F., & Podolski, C. -L., (1997). The effects of media violence on the development of antisocial behavior. In D. M. Stoff, J. Breiling, & J. D. Maser (Eds.), Handbook of antisocial behavior (pp. 181-193). New York: Wiley.
Huesmann, L. R., Moise, J. F., Podolski, C. -L., & Eron, L. D. (1999). The prediction of young adult aggression in the 1990s from childhood exposure to media violence in the 1970s: Gender differences and moderating factors for USA youth. Aggressive Behavior, 25, 19.
Huston-Stein, A., Fox, S., Greer, D., Watkins, B. A., & Whitaker, J. (1981). The effects of TV action and violence on children’s social behavior. Journal of Genetic Psychology, 138, 183-191.
Jenkins, H. (1999). Response during question-and-answer session after his discussion on media violence for the Free Expression Network, Freedom Forum, Rosslyn, Virginia, September 28.
Jensen, G. F. (2001). The invention of television as a cause of homicide: The reification of a spurious relationship. Homicide Studies, 5(2), 114-130.
Johnson, J. G., Cohen, P., Smailes, E. M., Kasen, S., & Brook, J. S. (2002, March 29). Television viewing and aggressive behavior during adolescence and adulthood. Science, 295, 2468-2472.
Josephson, W. L. (1987). Television violence and children’s aggression: Testing the priming, social script, and disinhibition predictions. Journal of Personality and Social Psychology, 535, 882-890.
Joy, A., Leslie, K. M., & Zabrack, M. (1986). Television and children’s aggressive behavior. In T. Williams (Ed.), The impact of television (pp. 303-360). Orlando, FL: Academic Press.

Kaplan, R. M. (1984). The measurement of human aggression. In R. M. Kaplan, V. J. Konečni, & R. W. Novaco (Eds.), Aggression in children and youth (pp. 44-72). Boston, MA: Martinus Nijhoff.
Kruttschnitt, C., Heath, L., & Ward, D. (1986). Family violence, television viewing habits, and other adolescent experiences related to violent criminal behavior. Criminology, 243, 235-267.
Lagerspetz, K., & Viemero, V. (1986). Television and aggressive behavior among Finnish children. In L. R. Huesmann, & L. D. Eron (Eds.), Television and the aggressive child: A cross-national comparison (pp. 81-117). Hillsdale, NJ: Erlbaum.
Lefkowitz, M. M., Eron, L. D., Walder, L. O., & Huesmann, L. R. (1972). Television violence and child aggression: A follow-up study. In G. A. Comstock, & E. A. Rubinstein (Eds.), Television and social behavior. Television and adolescent aggressiveness, vol. 3 (pp. 35-135). Washington, DC: U.S. Government Printing Office.
Lefkowitz, M. M., Eron, L. D., Walder, L. O., & Huesmann, L. R. (1977). Growing up to be violent: A longitudinal study of the development of aggression. New York: Pergamon.
Lester, D. (1989). Media violence and suicide and homicide rates. Perceptual and Motor Skills, 69, 894.
Leyens, J. P., Parke, R. D., Camino, K., & Berkowitz, L. (1975). Effects of movie violence on aggression in a field setting as a function of group dominance and cohesion. Journal of Personality and Social Psychology, 32, 346−360346-360.
McCain, R. S. (1998). Television’s bloody hands. Insight on the News, 14(47), 37-38.
McCarthy, E. D., Langner, T. S., Gersten, J. C., Eisenberg, J. G., & Orzeck, L. (1975). Violence and behavior disorders. Journal of Communication, 254, 71-85.
McGuire, W. J. (1986). The myth of massive media impact: Savagings and salvagings. In G. Comstock (Ed.), Public communication and behavior, vol. 1 (pp. 173-257). Orlando, FL: Academic Press.
McIntyre, J. J., Teevan, Jr., J. J., & assistance of Hartnagel, T. (1972). Television violence and deviant behavior. In G. A. Comstock & E. A. Rubinstein (Eds.), Television and Social Behavior: Vol. 3, Television and Adolescent Aggressiveness (pp. 383-435). Washington, DC: U.S. Government Printing Office.
McLeod, J. M., Atkin, C. K., & Chaffee, S. H. (1972a). Adolescents, parents, and television use: Self-report and other measures from the Wisconsin sample. In G. A. Comstock, & E. A. Rubinstein (Eds.), Television and social behavior. Television and adolescent aggressiveness, vol. 3 (pp. 239-313). Washington, DC: U.S. Government Printing Office.
McLeod, J. M., Atkin, C. K., & Chaffee, S. H. (1972b). Adolescents, parents, and television use: Adolescent selfreport measures from Maryland and Wisconsin samples. In S. H. Comstock, & E. A. Rubinstein (Eds.), Television and social behavior. Television and adolescent aggressiveness, vol. 3 (pp. 173-238). Washington, DC: U.S. Government Printing Office.
Merton, R. (1938). Social structure and anomie. American Sociological Review, 3, 672-682.
Messner, S. (1986). Television violence and violent crime: An aggregate analysis. Social Problems, 33, 218−235218-235.
Milavsky, J. R., Kessler, R., Stipp, H., & Rubens, W. S. (1982a). Television and aggression: Results of a panel study. In D. Pearl, L. Bouthilet, & J. Lazar (Eds.), Television and behavior: Ten years of scientific progress and implications for the 80s. Technical reviews, vol. 2 (pp. 138-157). Washington, DC: U.S. Government Printing Office.
Milavsky, J. R., Kessler, R., Stipp, H. H., & Rubens, W. S. (1982b). Television and aggression: A panel study. New York: Academic Press.
Mudore, C. F. (2000). Does TV violence kill? Current Health, 26(6), 24-26.
Paik, H., & Comstock, G. (1994). The effects of television violence on antisocial behavior: A meta-analysis. Communications Research, 21(4), 516-546.
Palermo, G. B. (1995). Adolescent criminal behavior-Is TV violence one of the culprits? International Journal of Offender Therapy and Comparative Criminology, 39(1), 11-22.
Phillips, D. P. (1983). The impact of mass media violence on U.S. homicides. American Sociological Review, 48, 560−568560-568.
Reid, C. (1999). AAP, ABA hail Hill’s rejection of media violence bill. Publishers Weekly, 246(25), 11.

Rhodes, R. (2000, November 23). The media-violence myth. Rolling Stone, 55-58.
Robinson, J. P., & Bachman, J. G. (1972). Television viewing habits and aggression. In G. A. Comstock, & E. A. Rubinstein (Eds.), Television and social behavior. Television and adolescent aggressiveness, vol. 3 (pp. 372-382). Washington, DC: U.S. Government Printing Office.
Schramm, W., Lyle, J., & Parker, E. B. (1961). Television in the lives of our children. Stanford, CA: Stanford University Press.
Sege, R. D. (1998). Life imitating art: Adolescents and television violence. In T. P. Gullotta, G. R. Adams, & R. Montemayor (Eds.), Delinquent violent youth: Theory and interventions (pp. 129-143). Thousand Oaks, CA: Sage.
Sheehan, P. W. (1986). Television viewing and its relation to aggression among children in Australia. In L. R. Huesmann, & L. D. Eron (Eds.), Television and the aggressive child: A cross-national comparison (pp. 161-200). Hillsdale, NJ: Erlbaum.
Singer, J. L., & Singer, D. G. (1981). Television, imagination, and aggression: A study of preschoolers’ play. Hillsdale, NJ: Erlbaum.
Sprafkin, J., Gadow, K. D., & Grayson, P. (1987). Effects of viewing aggressive cartoons on the behavior of learning disabled children. Journal of Child Psychiatry, 28, 387-398.
Sprafkin, J., Gadow, K. D., & Grayson, P. (1988). Effects of cartoons on emotionally disturbed children’s social behavior in school settings. Journal of Child Psychology, 29, 91-99.
Steuer, F. B., Applefield, J. M., & Smith, R. (1971). Televised aggression and the interpersonal aggression of preschool children. Journal of Experimental Child Psychology, 11, 442-447.
Thornton, W., & Voigt, L. (1984). Television and delinquency: A neglected dimension of social control. Youth and Society, 154, 445-468.
Wood, W., Wong, F., & Chachere, J. G. (1991). Effects of media violence on viewers’ aggression in unconstrained social interaction. Psychological Bulletin, 109, 371-383.
Zimring, F. E., & Hawkins, G. (1997). Crime is not the problem: Lethal violence in America. New York: Oxford University Press.