 How to cite: Anderson, C. A., Sakamoto, A., Gentile, D. A., Ihori, N., & Shibuya, A., Yukawa, S., Naito, M., & Kobayashi, K. (2008). Longitudinal effects of violent video games on aggression in Japan and the United States. Pediatrics, 122, e1067-1072.
Click here for a pdf of this article
INTRODUCTION
In the late 1980s American children played video games about four hours a week (1). They now average 13 hours overall, with boys averaging 16 to 18 hours per week (2). Furthermore, 90% of American children between the ages of 8 and 16 play video games at home (3). Children's favorite games often are violent (4). Currently, of all games classified by the industry's ratings group as appropriate for everyone aged 10 and older (E10+), over 90% contain violence (5). Over 75% of teen gamers under 17 report playing Mature-rated video games (the most graphically violent type) despite industry-wide restrictions (3). In a recent "secret shopper" study, over 80% of attempts by underage children to purchase an M-rated video game from rental stores were successful (3). If playing violent video games has harmful effects on some portion of players, then the vast majority of American youth are highly exposed to an unnecessary risk factor.
The general public typically define "violent media" as only those television shows, films and video games that include graphic images of blood and gore, but media violence researchers also include products without such images. Violent media are those that depict characters intentionally harming other characters who presumably wish to avoid being harmed. Thus, even children's video games that lack depictions of blood and gore can, and frequently do, include violence. "Aggression" also is defined differently by behavioral scientists than by the general public. Social and developmental psychologists typically define "aggression" as behavior that is intended to harm another person who is motivated to avoid that harm. In other words, aggression is an act carried out by one person with the intent of hurting another person; it is not an emotion, thought, or intention. For most social and developmental scientists, "violence" is the most extreme form of physical aggression, specifically physical aggression that is likely to cause serious physical injury.
Past research on violent video games has discovered consistent links to increased levels of aggression (6, 7). Existing experimental studies demonstrate that playing a violent video game causes an immediate increase in aggressive behavior, aggressive thoughts, and aggressive emotions (6, 8, 9). Existing cross-sectional studies (i.e., correlational studies that measure independent and dependent variables at one point in time) clearly link violent video game play to high levels of aggression and violence in real world contexts. They also rule out a number of non-causal explanations (10, 11).
However, establishing long-term causal effects of violent video games also requires longitudinal studies. Only one published longitudinal study with children has specifically examined longer term effects of exposure to violent video games (10), and no studies have investigated longitudinal effects in low violence cultures.
Longitudinal studies have investigated television and media violence in general among children and adolescents (12-14), and have demonstrated their causal longitudinal impact. Furthermore, these studies suggest that the long-term impact of television violence is larger for children than for adolescents. Nonetheless, the interactive nature of video games—their capacity to reward and punish the player for various actions, their immersive qualities, the fact that the user is an enactor as well as an observer of aggression—means that research specifically focusing on longitudinal violent video game effects is badly needed.
METHODS
Participants
In the present research, three samples of male and female school children were assessed at two points in time. Two samples are from Japan (15, 16); one is from the U.S. (17). Table 1 displays sample size and age ranges of the three samples. Although there are important developmental differences between middle childhood and adolescence (18), the psychological mechanisms postulated as underlying media violence effects are the same for each age such as priming processes; the learning of aggression-related scripts, attitudes, hostile attribution bias, and normative beliefs; and emotional desensitization (see 10, 19, for detailed descriptions of the short-term and long-term mechanisms).
Procedure
For all participants we assessed how much they habitually played violent video games, and how physically aggressive they had behaved in recent months. Table 1 describes the three samples. The samples varied in grade level (from 3rd to 12th graders), time lag between the two assessments (3 to 6 months), measure of habitual video game violence exposure (HVGV), and measure of recent physical aggressiveness.
Habitual Video Game Violence Exposure
Two of the studies—the younger Japan sample and the U.S. child sample—assessed habitual video game violence exposure (HVGV) in ways that fairly directly take into account violent content of favorite games and amount of time playing those violent games. The U.S. sample listed their three favorite video games and then rated each on amount of violent content and on how frequently they played each of the three games. HVGV for this sample was computed by multiplying the violent content rating by the frequency of play for each listed game, then averaging the three scores. This has been the standard procedure for several years (20). The younger Japan sample indicated how frequently they had played each of eight types of video games (fighting action, action, action role playing game, shooting, adventure, simulation, sports, puzzle). Based on prior content analyses of popular video games among Japanese children, HVGV was computed by averaging the frequency of play for the five types of games that are predominantly violent (fighting action, action, action role playing game, shooting and adventure).
The third study assessed HVGV in a somewhat less direct way. Participants listed their most favorite game genre and three additional favorite genres, and reported how many hours per week they spent playing any type of video game. For each participant, we assigned a favorite genres violence score which could range from 0 to 5. If their "most favorite" genre was a violent type, they received 2 points; if it was a nonviolent type they received a zero. For the remaining three favorites, they received an additional point for each that was a violent type of genre. We then multiplied the favorite genres violent score by the total number of hours per week spent playing video games.
Aggressive Behavior
For both Japan samples, the measure of aggressive behavior was self-reported trait physical aggression. For the younger sample, a 6-item Japanese version of the Buss and Perry physical aggression scale was used (21). This self-report measure asks about frequency of physically aggressive behaviors. This scale has been validated in a wide range of studies, including prior media violence studies. For the older Japanese sample, a 1-item self-report measure of frequency of physical aggression (involving punching or kicking someone) in the last month. For the U.S. sample, the measure of aggressive behavior was an index of teacher, peer, and self-reports of physical aggression, such as hitting, kicking, and getting into fights in the last year.
RESULTS
Despite the differences between samples in measures of HVGV, physical aggression, country, and age, each sample yielded statistically reliable positive correlations between Time 1 HVGV and Time 2 physical aggression of a magnitude that falls in the medium to large range for longitudinal predictors of physical aggression and violence (see Table 1). The weighted average longitudinal correlation across the three samples was: r+ = .28, Z = 11.65, p < .0001, 95% confidence interval = +.26 to +.31. The corresponding Odds Ratio is 2.10. Interestingly, the largest of these lagged correlations was for the sample that: (a) used the most direct measure of HVGV; (b) used multiple reports of aggressive behavior; (c) had the longest lag between the two measurement time periods; and (d) had the youngest participants—the U.S. sample (r = .40). The smallest correlation was from the sample with the least direct measure of HVGV, the shortest lag, a single item measure of physical aggression, and the oldest participants (r = .23). These two correlations are significantly different from each other, Z = 2.79, p < .01.
For our main analyses, we used the maximum likelihood structural equation procedures of the LISREL 8.5 statistical package to conduct a path analysis on the three correlation matrices (22). The first (baseline) model forced the estimated path weights linking the measured variables to be equal across the three samples. This model should yield the optimum fit if the variations in methods, ages, and sampled country have relatively little impact on the structure of the relations among the conceptual variables. However, if the longitudinal effect of HVGV on later aggression differs across samples, for example if it is stronger in the U.S. than the Japan samples, this baseline model should not fit the data very well. This model also provides a baseline for comparative testing of additional hypotheses about the relative magnitude of the longitudinal effect of HVGV on later physical aggression. Table 2 displays the results.
The baseline model fit the data quite well, as indicated by the measures of fit in Table 2 (23). Table 2 presents three additional and slightly different models, each of which was tested to see whether it fit the data better than the baseline model. One model allowed an independent estimate of the HVGV longitudinal path weight (B) for each sample. These path weights were similar for the two samples with younger participants and better measures of HVGV and physical aggressiveness, larger than the path weight the third sample, as shown in the bottom row of Table 1. However, this model did not yield a significantly better fit than the baseline model, as shown by the nonsignificant Chi-square difference test in Table 2.
A second comparative model specifically tested the hypothesis that the longitudinal effect is larger for the U.S. sample than for the two Japanese samples. The Chi-square fit test in Table 2 revealed that this model also was not significantly better than the baseline model.
A third model (labeled the "Age Model" in Table 2) tested the hypothesis that the older sample with the shorter lag and the weakest measures of HVGV and physical aggressiveness would yield a smaller longitudinal effect than the two younger samples (Japan 7th-9th graders, U.S. 3rd-5th graders). This model yielded a fit that was somewhat better than the baseline model. The estimated HVGV longitudinal path for the two younger samples (B = .152) was larger than the corresponding path for the older sample, B = .075.
Figure 1 displays the results of this path analysis. As expected, sex of participant strongly predicted HVGV and physical aggressiveness. Boys played more violent video games and were more physically aggressive than girls. Furthermore, physical aggressiveness at Time 1 was an extremely good predictor of physical aggressiveness at Time 2, consistent with much prior research which shows that the best predictor of future aggression is history of past aggression. Of primary importance, though, is the finding that across two very different cultures HVGV predicts physical aggression 3 to 6 months later, even after controlling for prior aggressiveness and sex. This result strongly supports the view that playing violent video games is a causal risk factor for relative increases in later physical aggressiveness. The main alternative explanation of prior cross-sectional correlation studies—that the association between amount of violent video game play and physical aggressiveness is merely an artifact of "naturally" aggressive children preferring violent video games—is ruled out by the longitudinal design and analysis. By controlling for participants' aggressiveness at Time 1, these longitudinal results also control for the innate aggressiveness of the participants as well as other factors that influence trait aggressiveness.
DISCUSSION
This study adds two critical pieces of evidence on the issue of the potential aggression-enhancing effects of violent video games. First, it confirms that habitually playing violent video games leads to increases in physical aggression some months later in children and adolescents, relative to those who do not play violent video games. Second, it demonstrates that such longitudinal effects occur in highly individualistic cultures with high societal levels of physical aggression and violence (the U.S.), and in more collectivistic cultures with low levels of physical aggression and violence. That both cultures yielded significant longitudinal effects of approximately the same magnitude illustrates the power of violent video games to affect children's developmental trajectories in a harmful way. These findings also contradict another popular alternative hypothesis: that only highly aggressive children (either by nature, culture or other socialization factors) will become more aggressive if repeatedly exposed to violent video games.
A third finding of importance was the trend of the longitudinal effect of video game violence to be larger in the younger samples. Of course, the younger two samples also had somewhat longer time lags and somewhat better measures of habitual exposure to video game violence and physical aggression, so it is not clear which of these sample differences contributed to this trend. Additional studies are needed in which the same measures are used with varying ages and longer time lags.
Of course, a short lag should theoretically make finding an effect of Time 1 video game violence exposure on Time 2 aggression (controlling for Time 1 aggression) less likely, because aggressive behavior is generally fairly stable across time, especially across shorter time lags than longer ones (24, 25). Therefore, it may be that the long-term effects of violent media exposure on later aggression and violence will be larger with longer time lags than were used in the present samples.
The study also is limited by the fact that the measures were not identical across samples. On the other hand, this fact also demonstrates the robustness of the violent video game effect across different measures of the same conceptual variables. In this way, the use of somewhat different measures of video game habits and of physical aggression in our three samples provides conceptual replication within this one study.
Additional research also is needed to further examine underlying psychological mechanisms of longitudinal change. Although prior research suggests that exposure to violent models, in either the real world or in entertainment media, teaches a host of aggression-enhancing behavioral scripts, attitudes, and beliefs, these effects have been tested most directly in short-term studies (10, 19, 26). Similarly, future research should further investigate the characteristics of violent games that may make some less harmful than others. For example, there is some evidence from television research that a focus on the pain and suffering of the victims of violence may reduce its harmful impact, whereas glamorizing the violent actions of attractive perpetrators may increase the harmful impact (19). Measures of video game exposure that more clearly distinguish among different types of violent video games may allow tests of these important theoretical and practical questions.
Youth violence is a public health issue in the U.S., because it accounts for so may deaths (25). Only accidental injury consistently leads homicide as the cause of death of one to twenty-four year olds. For those ten to twenty-four years old, homicide is the leading cause of death for African Americans, and the second leading cause for Hispanics. Finally, it is worth noting that in 2005, twelve to twenty year olds committed 28 percent of the single-offender and 41 percent of the multiple-offender violent crimes in the U.S., despite comprising only thirteen percent of the population (27).
Even so, such extreme violence is relatively rare in the age groups we studied (relative to milder forms of physical aggression). Thus, longitudinal studies of extreme violence will require much larger sample sizes (e.g., 25,000) and much longer time periods (e.g., 20-30 years) (12). But because physical aggressiveness in youth is one of the largest risk factors for later violence, an understanding of factors that increase (or decrease) youth aggression is vitally important if we are to understand and reduce violence in modern society. Prior experimental studies have clearly shown causal mechanisms by which violent video games can lead to long term changes in aggressive personality. Cross-sectional studies have repeatedly linked habitual video game violence to mild and severe forms of physical aggression, while ruling out plausible alternative explanations. The present study fills an import gap in the literature by confirming—with longitudinal data—prior empirical and theoretical work suggesting that frequent playing of violent video games is an important causal risk factor for youth aggression.
References
1. Harris MB, Williams R. Video games and school performance. Education 1985;105:306-309.
2. Martin S, Oppenheim K. Video gaming: General and pathological use. Trends & Tudes 2007; 6:1-7.
3. Walsh D, Gentile DA. Minneapolis, MN: National Institute on Media and the Family; c 2007 [updated 2007 Dec 4; cited 2007 Dec 6]. The MediaWise® 12th Annual Video Game Report Card; [about 3 pages]. Available from: http://www.mediafamily.org/research/report_vgrc_2007.shtml
4. Buchman DD, Funk JB. Video and computer games in the ’90s: Children’s time commitment and game preference. Children Today 1996;24:12-16.
5. Gentile DA. The rating systems for media products. In: Calvert SL, Wilson BJ, editors. Blackwell Handbook of Child Development and the Media. Malden, MA: Blackwell Publishing; in press.
6. Anderson CA, Carnagey NL, Flanagan M, Benjamin AJ, Eubanks J, Valentine J C. Violent video games: Specific effects of violent content on aggressive thoughts and behavior. Advances in Experimental Social Psychology 2004;36:199-249.
7. Dill KE, Dill JC. Video game violence: A review of the empirical literature. Aggression and Violent Behavior: A Review Journal 1998;3:407-428.
8. Bartholow BD, Sestir MA, Davis MD. Correlates and consequences of exposure to videogame violence: Hostile personality, empathy, and aggressive behavior. Personality and Social Psychology Bulletin 2005;31: 1573-1586.
9. Irwin AR, Gross AM. Cognitive tempo, violent video games, and aggressive behavior in young boys. Journal of Family Violence 1995;10:337-350.
10. Anderson CA, Gentile DA, Buckley KE. Violent Video Game Effects on Children and Adolescents: Theory, Research, and Public Policy. New York: Oxford University Press; 2007.
11. Krahé B, Möller, I. Playing violent electronic games, hostile attributional style, and aggression-related norms in German adolescents. Journal of Adolescence 2004;27:53-69.
12. Huesmann LR, Moise-Titus J, Podolski CL, Eron, L. Longitudinal relations between children's exposure to TV violence and their aggressive and violent behavior in young adulthood: 1977-1992. Developmental Psychology 2003;39:201-221.
13. Johnson JG, Cohen P, Smailes EM, Kasen S, Brook JS. Television viewing and aggressive behavior during adolescence and adulthood. Science 2002;295:2468-2471.
14. Slater MD, Henry KL, Swaim RC, Anderson LL. Violent media content and aggressiveness in adolescents. Communication Research 2003;30:713-736.
15. Reanalysis of data originally reported in Naito M, Kobayashi K, Sakamoto A. Terebigemu no shiyou to kougekisei no ingakankei no kentou: chugakusei ni taisuru paneru kenkyu [Testing a causal relationship between video game use and aggression: A panel study on junior high school students]. Proceeding of the 40th convention of the Japanese Society of Social Psychology 1999;40:288-289.
16. Reanalysis of data originally reported in Yukawa S, Sakamoto A. Terebi oyobi terebigemu niokeru bouryoku ga seishounen no kougekisei ni oyobosu eikyo: chugakusei oyobi koukousei wo taisho toshita juudan deta no bunseki [The effects of television and videogame violence on aggression of youth: A longitudinal study of junior high school and high school students]. Proceeding of the 42nd convention of the Japanese Society of Social Psychology 2001;42:502-503.
17. A similar longitudinal analysis with several additional variables and fewer participants (because of missing data on these additional variables) was reported in Anderson et al., 2007. The sample used in the present study consisted of the 364 participants who were assessed with a 5 to 6 month interval and had complete data on the variables used in this study.
18. Gentile DA, Sesma A. Developmental approaches to understanding media effects on individuals. In: Gentile, DA, editor. Media violence and children. Westport, CT: Praeger; 2003. p. 18-37.
19. Anderson CA, Berkowitz L, Donnerstein E, et al. The influence of media violence on youth. Psychological Science in the Public Interest 2003;4:81-110.
20. Anderson CA, Dill KE. Video games and aggressive thoughts, feelings, and behavior in the laboratory and in life. Journal of Personality and Social Psychology 2000;78:772-790.
21. Buss AH, Perry M. The aggression questionnaire. Journal of Personality and Social Psychology 1992;63:452 459.
22. Joreskog K, Sorbom D. LISREL 8: User's Reference Guide. In. Chicago: Scientific Software International, Inc.; 1996.
23. Generally, a structural equation model of this type is considered to fit the data well if the fit indexes are greater than .95 (maximum possible is 1.00), if the overall Chi-square fit test yields a nonsignificant p-value (> .05), or if the root mean square error of approximation is less than .05 (minimum possible is .00).
24. Eron LD, Huesmann LR. The stability of aggressive behavior--even unto the third generation. In: Lewis M, Miller SM, editors. Handbook of developmental psychopathology: Perspectives in developmental psychology. New York, NY: Plenum Press; 1990. p. 147-156
25. Services USDoHaH. Youth violence: A report of the Surgeon General. In: Services USDoHaH, Centers for Disease Control and Prevention, NCfIPaC, Substance Abuse and Mental Health Services Administration, Center for Mental Health Services, National Institutes of Health, National Institute of Mental Health, editors. Rockville, MD; 2001. Available from: http://www.surgeongeneral.gov/library/youthviolence/order.htm
26. Huesmann LR. Psychological processes promoting the relation between exposure to media violence and aggressive behavior by the viewer. Journal of Social Issues 1986;42:125–139.
Acknowledgments
Author Contributions: Dr Anderson had full access to the U.S. sample data and to the correlation matrices for the two Japanese samples and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Anderson, Sakamoto, Gentile, Ihori, Shibuya
Acquisition of data: Sakamoto, Gentile, Ihori, Shibuya
Analysis and interpretation of data: Anderson, Sakamoto, Gentile, Ihori, Shibuya
Drafting of the manuscript: Anderson, Gentile
Critical revision of the manuscript for important intellectual content:
Statistical analysis: Anderson
Obtained funding: Gentile, Anderson, Sakamoto
Administrative, technical, or material support: None
Study supervision: Anderson, Sakamoto, Gentile
Financial Disclosures: None
Funding/Support: National Institute for Media and the Family; Centers for Disease Control; The Broadcasting Policy Division in the Information and Communications Policy Bureau at the Ministry of Public Management, Home Affairs, Posts, and Telecommunications in Japan
Role of the Sponsor: None of the sponsors had any role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.
Disclaimer: The opinions expressed are those of the authors only and do not represent the official position or policies of the National Institute for Media and the Family; the Centers for Disease Control; or The Broadcasting Policy Division in the Information and Communications Policy Bureau at the Ministry of Public Management, Home Affairs, Posts, and Telecommunications in Japan.
Acknowledgment: We thank Frederick O. Lorenz for his statistical assistance.
Author Affiliations: Center for the Study of Violence, Department of Psychology, Iowa State University (Anderson, Gentile); Graduate School of the Humanities and Sciences, Ochanomizu University (Sakamoto, Ihori, Kobayashi); National Institute on Media and the Family (Gentile); Institute for Media and Communications Research, Keio University (Shibuya); Graduate School of Comprehensive Human Sciences, University of Tsukuba (Yukawa); Faculty of Economics, Takasaki City University of Economics (Naito).
Note: This version may differ in several small ways from the final published version
| 
