a review of the literature - dr. douglas · violent video game effects on children and adolescents...

Violent video game effects on children and adolescents

A review of the literature

D. A. GENTILE 1, 2, 3, W. STONE 3rd 1, 2

1Department of PsychologyIowa State University, Ames, IA

2Institute of Science and SocietyIowa State University

3National Institute on Mediaand the Family, Ames, IA

Studies of violent video games on children andadolescents were reviewed to: 1) determine themultiple effects; 2) to offer critical observa-tions about common strengths and weakness-es in the literature; 3) to provide a broader per-spective to understand the research on the ef-fects of video games. The review includes gen-eral theoretical and methodological consider-ations of media violence, and description ofthe general aggression model (GAM). The lit-erature was evaluated in relation to the GAM.Published literature, including meta-analyses,are reviewed, as well as relevant unpublishedmaterial, such as conference papers and dis-sertations. Overall, the evidence supports hy-potheses that violent video game play is relat-ed to aggressive affect, physiological arousal, ag-gressive cognitions, and aggressive behaviours.The effects of video game play on school per-formance are also evaluated, and the reviewconcludes with a dimensional approach tovideo game effects. The dimensional approachevaluates video game effects in terms ofamount, content, form, and mechanics, andappears to have many advantages for under-standing and predicting the multiple types of ef-fects demonstrated in the literature.Key words: Video games - Child - Violence.

“Lead has been recognized as a poi-son…and has been the focus of public

health regulation in much of the developedworld for the better part of the past century.

The nature of regulation has evolved in re-sponse to increasing information providedby vigorous scientific investigation of lead’seffects. In recognition of the particular sen-sitivity of the developing brain to lead’s per-nicious effects, much of this legislation hasbeen addressed to the prevention of child-hood lead poisoning”.

The quote above is from a review articletitled “Lead neurotoxicity in children: basicmechanisms and clinical correlates”.1 Therehave been more studies on the effects ofmedia violence than on lead exposure. Whatif we replaced the word “lead” above with“media violence”? In contrast, although me-dia violence has been recognized as a riskfactor for healthy child development since atleast 1972, there does not seem to have beena corresponding serious focus on the publicpolicy implications in most countries.Perhaps a direct comparison between leadexposure and media violence is unfair, butconsider the conclusions of meta-analyseson each.

Needleman et al.2 conducted a meta-analy-

Address reprint requests to: Dr. D. A. Gentile, Iowa StateUniversity, Department of Psychology, W112 LagomarcinoHall, Ames, IA 50011-3180. E-mail: [email protected]

Vol. 57, N. 6 MINERVA PEDIATRICA 337

REVIEWSMINERVA PEDIATR 2005;57:337-58

GENTILE VIOLENT VIDEO GAME EFFECTS ON CHILDREN

sis of research comparing lead exposure andthe intelligence quotients (IQs) of children.They found 24 studies that used proper con-trols which they included in the meta-analy-sis. Partial correlations ranged from -0.27 to-0.003. These numbers may seem small, butthey are large enough that few doubt the ef-fect of lead on child IQ and they are signifi-cant when considering the number of chil-dren. As a result of such research docu-menting the health risks of lead exposure,legislation strongly enforces limits on theamount of lead in the environment.

Anderson et al.3 conducted a meta-analy-sis of research comparing media violence ex-posure and aggressive behavior. They found284 studies with 51 597 participants usingproper measures. Correlations between me-dia violence and subsequent aggression frommeta-analytic reviews ranged from 0.11 to0.31. These results are stronger than thoseon lead exposure, and the number of childrenwho are exposed to media violence is greater.Yet, public policy debates continue to ques-tion whether any actions should be taken.Indeed, publicly the issue is so contentiousthat although almost no one would considerthe validity of the introductory quote withrespect to the effects of lead, many wouldfind it laughable to use such language whendiscussing the effects of violent media.

The Anderson et al.3 meta-analysis focusedprimarily on studies of television and film vi-olence. However, in the past 15 years, inter-active video games have become increas-ingly violent. At the same time, greater num-bers of children have begun playing videogames for increasing amounts of time.Although several excellent narrative andmeta-analytic reviews of violent video gamestudies exist, this review will primarily focuson studies including children and adoles-cents. This paper has 3 goals: 1) to summa-rize the literature on the multiple effects vi-olent video games have on younger players;2) to offer critical observations about commonstrengths and weaknesses in the literature;3) to provide a broader perspective withinwhich to understand the research on the ef-fects of video games.

Theoretical and methodologicalconsiderations

Several criticisms of the media violenceand violent video game literatures exist.4, 5

Some of these criticisms are legitimate, andmany are illegitimate. Due to space con-straints, only some will be discussed here.6, 7

One of the most serious of the illegitimatecriticisms is that critics often consider studies“in a theoretical vacuum”.7 No single study isperfect, and each type of study design hasstrengths and weaknesses. Critics often at-tempt to document the relative flaws in stud-ies without referring to the broader theoret-ical framework within which the studies wereconducted. This is like pointing out that aparticular car model made by Ford has a flaw,and therefore all internal combustion enginesmust not work. Just because one car has aflaw we cannot ignore that the theory of en-gine mechanics is sound, and many othercars with different flaws also work. Therefore,it is important to understand the broader the-ories that guide research on media violenceeffects.

Many theoretical models have been ad-vanced to describe, explain, and predict ag-gressive behavior. Most of these models havebeen specific to one domain or to a particu-lar level of analysis. For example, some lookat the level of neural networks (e.g.,Berkowitz’s cognitive-neoassociation theo-ry), some focus on arousal processes (e.g.,Zillman’s excitation transfer theory), and somefocus on cognitive processes (e.g.,Huesmann’s script theory). Each of these do-main-specific theories has received empiricalsupport.8 Research has shown that media vi-olence exposure does lead to priming of ag-gressive thoughts, increased physiologicalarousal, and the development of aggressivescripts and beliefs. Recently Anderson et al.created the general aggression model (GAM)to help unify the many more specialized the-ories into one broad model.9

The GAM is designed to describe and pre-dict likelihood of aggressive behavior in boththe short term (or episode) and long-term.Aggressive behavior is usually defined by re-searchers as behavior that is intended to harm

338 MINERVA PEDIATRICA Dicembre 2005

VIOLENT VIDEO GAME EFFECTS ON CHILDREN GENTILE

another person. Repeated interactions with vi-olent video games are hypothesized to haveseveral effects over the long term. Figure 1displays how single interactions can lead tolong term changes. Any given interaction canbe social (e.g., an argument with another per-son) or individual (e.g., playing a violentvideo game by oneself), and is consideredan “episode.” An episode is influenced bothby person variables and by situation vari-ables. The person variables include a per-son’s personality characteristics, their existingattitudes, beliefs, scripts, biological predis-positions, etc. The situation variables includespecific proximal characteristics such aswhether the individual was just insulted,whether there are onlookers, whether the in-dividual just played a violent video game,etc. The person and situation variables com-bine to affect the individual’s present internalstate. There are at least 3 routes by which aperson’s internal state could be influenced, in-cluding arousal, affect, and cognition (al-though these are not typically wholly inde-pendent of each other). The individual thenhas an opportunity to decide on how to re-spond and takes some action, which yieldssome result. The results of this episode feeddirectly into the next episode, affecting boththe situation variables as well as the internalstate of the individual (please note that this isan oversimplification of the episodic and de-cision processes in the model; for details seeAnderson et al.9, 10 and Anderson, Gentile,Buckley, in press).

If the results of an episode are rewardingor punishing, it is certainly likely that the in-dividual may learn something, leading tosome change. For example, if you insultsomeone else and get hit in response, hope-fully you will learn to be more careful withyour insults. This leads to some change inthe person, although in any single episode itis likely to be very little change – most per-sonality traits are highly stable over the shortterm. However, it is possible that people maylearn even when there is no decision-action-result process. Consider playing a violentvideo game in which one is rewarded forsuccessful violent actions. The violent gameis likely to increase physiological arousal,

prime aggressive thoughts, and lead tochanges in emotional states. Because these allhappen together as a result of the game, theycan become associated with each other.


Time

Short-term effects

Epi

sode

1E

piso

de 2

Epi

sode

N

Personvariables

e.g. aggressive beliefs, hostilityempathy, desensitization, etc.

Situationvariables

Personvariables

Decision

Action

Result

Present internal stateCognition

Affect Arousal

Personvariables

Decision(impulsive orthoughtful)

Action

Result

Present internal stateCognition

Affect Arousal

Person variables(e.g. hostility,

beliefs, attitudes,scripts)

Stuation variables(e.g. provocation,media violence)

Long-term

effects

Stab

ility

Cha

nge

Cha

nge

Figure 1.—The general aggression model: episodic process-es leading to long-term change.


Because game playing is generally consid-ered to be fun, relaxing, and rewarding, theselinks between aggressive thoughts, feelings,and arousal are rewarded, and may, therefore,also lead to some change in the person (al-though again, the short-term change is like-ly to be small). However, repeated exposureover time provides multiple learning trials,leading to changes that are likely to be larg-er and measurable. Figure 2 displays some ofthe personality variables believed to be in-fluenced by an individual’s ongoing bio-so-cial interactions. What is learned across timedepends on the experiences an individualhas, including interactions with biologicalfactors (such as biological predispositions tolearn certain types of associations throughrewards and punishments). Violent videogames may affect individuals by having in-

fluences on their aggressive beliefs, schema-ta, scripts, and by desensitizing them to ag-gression.

Understanding the theory helps to orga-nize the many types of studies that are con-ducted. It helps researchers and others un-derstand the types of effects they should ex-pect, as well as which effects they shouldnot expect. This allows researchers to an-swer the often-heard criticism, “I’ve played vi-olent video games for years and never shotanyone”. As Figures 1 and 2 show, no re-searcher would expect to see a simple andextreme effect such as playing a violent gameand then picking up a gun and shootingsomeone. We would expect that years of vi-olent video game exposure would be relatedto increased personality trait hostility; how-ever, we would not expect that a single


Repeated violent video game playLearning, rehearsal and reinforcement ofaggression-related knowledge structures

Aggressivebeliefs andattitudes

Aggressiveperceptualschemata

Aggressiveexpectationschemata

Aggressivebehaviorscripts

Aggressiondesensitization

Increase inaggressivepersonality

Cognition

Present internal state

Arousal

Affect

Person variablese.g., hostile personality, hostile attribution bias

Situational variablese.g., peer group, social situations

Rep

eate

d le

arni

ng tr

ials

acr

oss

time

(dis

tal i

nflu

ence

s)

Indi

vidu

al s

ocia

l epi

sode

(pro

xim

al in

flue

nces

)

Figure 2.—The general aggression model: developmental processes.


episode of playing a violent video gamewould change trait hostility – personality traitsby definition are stable over the short term.But we would expect that playing a violentvideo game might lead to increased arousal,aggressive affect, or aggressive cognitions inthe short-term, and maybe they would leadto aggressive behaviors immediately post-play (but note that behavior is farther re-moved from the situational event of playingthe video game than arousal, affect, and cog-nition, so it would have to be a pretty strongeffect to influence behavior).

Several legitimate criticisms of video gameresearch have been raised, 4 of which will bedescribed here.6

1) Some studies claiming to measure ag-gressive behavior use dependent variablesthat are not true aggressive behavior. A sur-prising number of studies claim to measurethe short-term effects of playing a violentgame on aggressive behavior, but use traitor personality aggression scales as measuresof aggressive behavior. As mentioned above,there is no reason to expect that playing avideo game for 20 min would change a trait.Some studies have also used hitting inani-mate objects as a measure of aggressive be-havior. Most modern aggression researchersdefine aggression as behavior intended toharm another person 10, 11, and would there-fore not consider hitting objects to be trueaggressive behavior.

2) Many studies use sample sizes that aretoo small to detect an effect. Meta-analyses 12

have suggested that the average effect size isabout r=0.20. Therefore, the number of studyparticipants (N) should be at least 200 forpower of 0.80 (power is the likelihood of be-ing able to find a legitimate difference be-tween groups). Studies that use sample sizesunder 200 are unlikely to find effects of vio-lent video games – not because there is no ef-fect, but because there is not enough statis-tical power to find it.

3) Many studies use “violent” and “nonvi-olent” games that are not particularly differ-ent from each other. This was especially aconcern for studies conducted in the 1980sand early 1990s. Several studies used games

such as Missile Command as a violent game,which included no aggression at all towardpeople. Some studies have used games thatincluded violence in the alleged non-violentcondition. If the games are not sufficientlydistinct from each other in terms of violence(or if the “violent” game is not particularlyviolent), then we should not expect to findmuch of a difference in their effects.

4) Many studies do not match the violentand nonviolent games on potentially con-founding dimensions such as frustration, dif-ficulty, or excitement. An experimenter mayhave done everything else correctly—select-ed violent and nonviolent games that areclearly distinct, had a sample size over 200,and measured appropriate dependent vari-ables—but still find no result or even a non-intuitive result if the games were not matched.For example, if the nonviolent game wasmore boring or frustrating than the violentgame, then participants may be more ag-gressive after playing the nonviolent gamethan the violent game! Therefore, the highestquality studies tend to match violent and non-violent games on several dimensions in orderto ensure that any differences are due to vi-olent content and not to other unrelated di-mensions.

In general, there are 3 dominant types ofstudy designs, and each has distinct strengthsand weaknesses. They are experimental, cor-relational (or cross-sectional), and longitudi-nal. The are several taxonomies under whichone could classify the different type of re-search designs. The 3 types discussed hereare not exhaustive, but are the most typicaland comprise the majority of media effectsstudies. The major strength of experimentalstudies is that causality can be determined.This is done by randomly assigning partici-pants to different groups, and treating eachgroup identically except for one variable (the“independent variable”). If the groups areshown to differ in some measurable way (the“dependent variable”) at the end of the study,the most probable explanation is that the in-dependent variable caused the changes inthe dependent variable. Experimental studiesare usually used to measure short-term ef-fects, although experimental studies could



be designed to test long-term effects. Highquality experimental studies share at least 4characteristics: they have sample sizes of atleast 200, they use games that are equated onpotentially confounding dimensions, theyuse violent and nonviolent games that aretruly violent and nonviolent, and have validmeasures of aggression or aggression-relatedvariables (Anderson, Gentile, Buckley, inpress). Over 30 experimental tests of the ef-fects of violent video games have been con-ducted with children and adults, but veryfew share all 4 of these characteristics.

The major weakness of experimental stud-ies is that researchers cannot measure “reallife” aggression. For example, it is unethicalto allow participants to hit each other as ameasure of aggressive behavior. Thus, re-searchers use other means such as the noise-blast paradigm.13-15 It is then incumbent up-on researchers to demonstrate that their lab-oratory measures have real-world validity.With the noise-blast paradigm, specifically,several studies have shown that people whogive higher blasts in the laboratory are alsomore likely to physically aggress in the realworld.16-20 Therefore, the noise-blast para-digm is a valid measure of aggressive be-havior.

Correlational (or cross-sectional) studiesallow researchers to overcome the primaryweakness of experimental studies becausethey can use a wide range of “real world”measures of aggression. Thus, researcherscan study the relations between violent gameexposure and several types and severities ofaggression. Correlational studies are usuallyused to measure long-term effects (althoughshort-term correlational studies can be de-signed). Correlational studies can also incor-porate multiple informants (e.g., self, peers,parents, teachers), thereby gaining addition-al levels of ecological validity. High qualitycorrelational studies also share several char-acteristics: adequate sample size (at least200); a reliable measure of exposure to vio-lent video games; and a reliable measure ofaggression or of an aggression-related vari-able (Anderson et al., in press).

The major weakness of correlational stud-ies is that causality cannot be determined.

However, correlational studies are importantin that they can support, refine, or refutecausal theories. More importantly, correla-tional studies are strong where experimentalstudies are weak—that is, they can find rela-tions between media violence and real-worldaggression. Conversely, experimental stud-ies are strong where correlational studies areweak—that is, they can document causal re-lations.

Longitudinal studies allow researchers toovercome some of the weaknesses of bothexperimental and correlational studies. In alongitudinal study, researchers study the samepeople over a period of time. They can thusobserve patterns of behavior and documentboth short-term and long-term effects.Longitudinal studies can be either experi-mental or correlational in design, but evencorrelational designs can provide some causalinformation, as it is possible to determinewhat precedes what over time. High qualitylongitudinal studies should share the samecharacteristics of high quality experimentaland correlational studies.

Violent video game effects studies

Researchers using the GAM would predict4 main types of short-term effects of violentvideo games: increased aggressive affect, in-creased physiological arousal, increase ag-gressive cognitions, and increased aggres-sive behaviors. Although few of the studiesreviewed here have relied on the GAM astheir theoretical basis, most fit with it verywell.

Aggressive affect

The GAM predicts that violent video gamesmight cause both short-term and long-termincreases in aggression-related feelings, suchas state anger, anxiety, or trait hostility. At thetime of this writing, there have been 6 ex-perimental (Hind,21 Funk et al.,22 Fleming etal.,23 Ballard et al.,20 Cohn and Brooks un-published), 2 correlational (Durkin et al.,24

Anderson et al. in press), and 2 longitudi-nal studies 20, 25 of video games and aggres-



sion-related affect conducted with childrenand adolescents (Table I). None of thesestudies meets all of the criteria for high qual-ity studies described above. Several havesmall sample sizes,20, 22, 23 which may be re-lated to many failing to find statistically sig-nificant effects. It should be noted, howev-er, that all studies including those with non-significant results showed trends in the pre-dicted direction (e.g., higher hostile feelingsafter playing a violent game). This is per-haps surprising given the wide range of the-oretically relevant dependent measures.Some studies measured hostility, frustration,anger, or anxious feelings—all clearly ag-gression—related emotions. Others, how-ever, measured depressed mood,24 prefer-ences for playing violent or non-violent

games,21 and emotional desensitization toplaying violent games.25

The 2 studies with large sample sizes bothsuffer from the problem that neither mea-sured violent game play, but only total amountof game play.24, 25 Amount of game play perse is not theoretically predicted to increaseaggressive emotions, only amount of violenceexposure. However, there is a strong corre-lation between amount of play and violenceexposure;26 children who spend more timeplaying video games also tend to play moreviolent games. This probably accounts for therelation found in these 2 studies – amount iseffectively a proxy for violence exposure.

None of the studies on affect with childrencan be considered to be definitive. In additionto the problems already described, none of


TABLE I.—Studies of violent video games on affect.

Study N Age Major findings

Experimental studies on affect

Cohn (1995 - unpublished) 124 6th-8th graders Higher, but non-significant, hostile and anxiousfeelings after V game compared to NV game

Hind (1995) 21 102 15-18 years Incarcerated adolescents preferred playing ga-mes with aggressive actions more than NV ga-mes and more than a non-incarcerated controlgroup

Funk et al. (1999) 22 35 3rd-5th graders Higher, but non-significant, frustration levelsand parents after V game

Brooks (1999 - unpublished) 120 6th-7th graders Higher frustration and negative affect levels afterV game compared to NV game, watching V andNV games, or watching V or NV television pro-grams

Fleming et al. (2001) 23 71 8-12 years V game increased state anger marginally signifi-cantly (P = 0.052), while simultaneously impro-ving overall mood compared to a paper-and-pen-cil task

Ballard et al. (2001) 20 41 12-18 years V games increased frustration (but were harder toplay than NV game)

Correlational studies on affectDurkin et al. (2002) 24 1 304 10th graders High computer game use correlated with greater

depressed mood compared to low computer ga-me use

Anderson et al. (in press) 189 14-19 years V game play correlated with trait hostility andtrait anger

Longitudinal studies on affectBallard et al. (2001) 20 41 12-18 years Repeated play of V games (3 times over 3 weeks)

resulted in lower facial displays of disgust andsmiling

Ihori et al. (2003) 25 807 5th and 6th grade Increased amount of video game play (all play,not specifically V games) related to later hostility

V = violent, NV = non-violent; VR = virtual reality.


the experimental studies used matched vio-lent and non-violent video games. However,regardless of the various methodologicalproblems and various approaches to mea-suring aggression-related affect, they con-verge on the finding that violent game play isrelated to aggressive affect.

Although the focus of this review is onchildren and adolescents, it is important tonote that these results also converge with thepreponderance of results from studies of col-lege students and adults. In a recent meta-analysis of studies of violent video gamesand aggressive affect, no differences in ef-fect sizes were found between child and adultsamples.27 Combining all samples, the meta-analysis 27 revealed a significant average ef-fect size of approximately r=0.17.

Physiological arousal

The GAM predicts that violent video gamesmight cause short-term increases in physio-

logical arousal, such as heart rate, blood pres-sure, or the so-called “stress” hormones. Atthe time of this writing, there have been 9 ex-perimental studies of video games and phys-iological arousal conducted with childrenand adolescents (Table II 20, 23, 28-31 and Cohn,Brooks and Matsuda et al., unpublished). Asa group, the results from these studies ap-pear to be relatively weak, but there is evi-dence that both violent and non-violentgames can increase physiological arousal 20,

23, 29 (Brooks MC, unpublished data, 1999).This fact increases the importance of match-ing the violent and non-violent games on di-mensions other than violence; unfortunately,none of the studies matched the violent andnon-violent games.

Winkel et al.28 created 3 games on an AppleIIe computer varying in terms of “aggres-siveness.” They tested appropriately whetherthe 3 versions were different from each oth-er (which they were) and from a no-game


TABLE II.—Studies of violent video games on physiological arousal.


Experimental studies on affectWinkel et al. (1987) 28 56 8th graders No effect of V or NV games on HRSegal et al. (1991)29 32 16-25 years Increased HR, BP, and oxygen consumption when

playing Ms. Pac Man compared to standing butinactive

Irwin et al. (1995)30 60 2nd graders No effect of V or NV games on HRCohn (1995 - unpublished) 124 6th-8th graders Higher, but non-significant, HR after V game com-

pared to NV gameLynch (1999)31 40 9th-12th grade V game play increased stress hormones (epi-

nephrine and nor-epinephrine) among hostileadolescents

Brooks (1999 - unpublished) 120 6th-7th graders Higher HR after V game compared to NV game,watching V and NV games, or watching V or NVtelevision programs

Fleming et al. (2001)23 71 8-12 years V game increased HR and self-reported arousalcompared to NV game or pencil-and-paper game

Ballard et al. (2001)20 41 Adolescents “Horror” genre VG increased arousal (HR andBP?) more than other sports or fighting genres;across games, HR initially decreased and then in-creased throughout game play; SBP decreasedin anticipation of playing, remained low duringplay, but increased immediate after playing

Matsuda et al. (2003-unpublished) ? Unknown Playing a V game (and some NV games) decrea-ses prefrontal cortex activation (indicating atten-tion and less self monitoring) compared to otherNV games and to just watching the games

V = violent, NV = non-violent; VR = virtual reality; HR = heart rate, BP = blood pressure, SBP = systolic blood pressure, DBP = dia-stolic blood pressure.


condition. However, it is unclear whetherany of the versions would be considered tru-ly aggressive (meeting the definition of in-tentional harm to another person, rather thanshooting at shapes). Their design included8 conditions, but with 56 participants total,each cell only included 7 participants. Thissample size leaves little power for findingdifferences in highly variable data such asheart rate.

One of the more interesting findings is re-ported by Lynch.31 In this study, stress hor-mones (epinephrine, nor-epinephrine, testos-terone, and cortisol) were measured in urinefollowing 45 min of violent game play com-pared to 45 min of quiet resting. Overall thereappeared to be no effect of playing violentgames. However, there was an interactionwith personality trait hostility. High hostileadolescents showed significant increases inepinephrine and nor-epinephrine (and a mar-ginally significant increase in testosterone)compared to low hostile adolescents. Lynchrecruited his participants from science andmath classes in high school. It is likely that ifasked, all participants would have reportedthat it was more relaxing to play video gamesthan to stay in class. However, hostile ado-lescents were physically unable to relax inthe presence of the aggressive stimulus.Although this study did not include a non-vi-olent control game, it suggests that the ef-fects of violent games on physiologicalarousal may not be as straightforward as re-searchers might hope, but may interact inimportant ways with personality or other vari-ables (such as amount of previous violentgame play).

In contrast to these studies, the literaturewith adult participants seems somewhatstronger. Combining child and adult samples,a recent meta-analysis 27 again reveals a sig-nificant average effect size of approximatelyr=0.17.

Aggressive cognition

The GAM predicts that violent video gamesare likely to cause both short-term and long-term increases in aggressive cognitions, suchas aggressive priming, hostile attribution bias,

and pro-violence attitudes. At the time of thiswriting, there have been 5 experimental stud-ies,32-36 8 correlational studies (Dominick,37

Colwell et al.,38 Funk et al.,35 Funk et al.,36

Funk et al.,39 Krahé et al.,40 Gentile et al.,26

Anderson et al. in press), and 1 longitudinalstudy of video games and aggressive cogni-tion (Anderson et al. submitted) conductedwith children and adolescents (Table III).Some of the correlational studies fit our cri-teria for high quality studies,26, 40 but most ofthe experimental studies have seriousmethodological problems. For example,Graybill et al.32 found that playing a violentgame increased assertive fantasies more thanplaying a non-violent game, but they also re-port that the violent game was more frus-trating to play than the non-violent game.Without the games being matched on con-founding dimensions such as this, it is im-possible to know whether it was the violentcontent that accounts for any observed dif-ferences. Funk et al.36 use both a correla-tional and an experimental approach to testwhether violent video games affect empathyin children. Because empathy is typically con-sidered to be a trait rather than a state, it is un-likely that measures would be sensitiveenough to find changes in empathy afterplaying a video game for 15 min. The au-thors found a significant long-term relation-ship in the correlational study (habitual vio-lent video game playing was correlated withlower empathy scores), but no significantshort-term relationship in the experimentalstudy. The experimental studies also use smallsample sizes – indeed, Funk et al.36 note thattheir observed statistical power is only 0.05.This is not a complete list of methodologicalflaws. For example, Graybill et al.33 title theirarticle “Effects of playing versus observingviolent versus nonviolent video games onchildren’s aggression,” but they then com-bine observers and players for data analysis.There is less theoretical reason to believethat observers would be affected in the sameway or to the same extent as players.

As a group, the correlational studies onaggressive cognitions are methodologicallystronger than the experimental studies. Krahéet al.40 gathered survey data from 231 German



8th graders. They found that children whoplay violent games more frequently and whowere likely to recommend violent games tofriends were more likely to endorse norma-tive beliefs about the acceptability of ag-gressive behavior. These beliefs in turn me-diated a significant relationship with hostileattribution bias. Hostile attribution bias is aperceptual/cognitive bias related to aggressivebehavior that has been validated in manystudies.41-43 It refers to how one views the

world when ambiguous negative events oc-cur. For example, if a child is bumped in thehallway, does he assume it was accidentalor was intentionally hostile. If the child has abias to attributing hostility to others’ actions,that child is very likely to respond aggres-sively to others. In an experimental study,Kirsh 34 found that playing a violent videogame can cause increased hostile attributionbias in the short term.

In a study of 607 adolescents, violent game


TABLE III.—Studies of violent video games on aggressive cognition.


Experimental studies on aggressive cognitionsGraybill et al. (1985) 32 116 2nd, 4th, and V game increased assertive fantasies and de-

6th grade creased defensive fantasies compared to NV ga-me

Graybill et al. (1987) 33 79 7-11 years V games increased aggressive thought accessibi-lity and liking compared to NV games

Kirsh (1998) 34 52 3rd-4th grade V game increased hostile attribution bias com-pared to NV game

Funk et al. (2000) 35 35 8-12 years V game increased aggressive thoughts and de-creased empathic thoughts, but not significantly

Funk et al. (2003) 36 31 5-7 years and No effect of V game on aggressive story com-35 8-12 years pletions or empathy

Correlational studies on aggressive cognitionsDominick (1984) 37 250 10th-11th grade Arcade game playing (but not home game

playing) correlated with aggressive cognitionsColwell et al. (2000) 38 204 12-14 years Game exposure correlated with aggressive co-

gnitionsFunk et al. (2000) 35 35 8-12 years Naming a V game as favorite correlated with ag-

gressive thoughtsFunk et al. (2003) 36 31 5-7 years and History of V game play correlated with lower

35 8-12 years empathyFunk et al. (2004) 39 150 4th and 5th grade Greater V game play positively correlated with

proviolence attitudes and negatively correlatedwith empathy

Krahé et al. (2004) 40 231 8th grade Greater V game play positively correlated with ac-ceptance of physical aggression and indirectlywith hostile attribution bias

Gentile et al. (2004) 26 607 8th and 9th grade Greater V game play positively correlated withtrait hostility

Anderson et al. (in press) 189 14-19 years V game play correlated with positive attitudes toviolence in war, attitudes toward intimate partnerviolence, and normative aggression beliefs

Longitudinal studies on aggressive cognitionsAnderson et al. (in press) 430 3rd-5th grade V VG play early in a school year related to later

increases in hostile attribution bias after control-ling for sex, race, total screen time, parental in-volvement, and earlier hostile attribution bias

V = violent, NV = non-violent; VR = virtual reality.


play, personality trait hostility (aggressivecognitions/beliefs/attitudes), and antisocialand aggressive behaviors were measured.26

Violent game exposure predicted antisocialand aggressive behaviors, but these rela-tionships were also mediated by increasedhostility. Therefore, as predicted by theory,long-term exposure to violent video gamesmay affect aggressive behavior by first af-fecting beliefs and attitudes about aggres-sion, which later change one’s likelihood ofbehaving aggressively. However, these stud-ies are correlational and it is unclear whethervideo games are the causes of increased ag-gressive cognitions.

In the only longitudinal study conducted todate, 430 3rd-5th graders were surveyed at 2points during the school year (Anderson,Gentile, Buckley, in press). Measures of vio-lent video game exposure, hostile attribution

bias, and aggressive behaviors were taken, in-cluding gathering information about behav-iors from multiple informants (self-report,peer-nominations, and teacher reports).Violent video game play early in the schoolyear was significantly related to changes inhostile attribution bias later in the schoolyear, even after controlling for several po-tentially confounding variables (sex, race, to-tal screen time, and parental involvement inchildren’s media habits). As predicted, chan-ges in hostile attribution bias mediated chil-dren’s later aggressive behaviors (Figure 3).

These results are consistent with the studiesof violent video games and aggressive cogni-tion with adults. Because of the larger numberof studies using children and adolescents,Anderson 44 was able to conduct a meta-analy-sis of studies with just child participants. Acrossstudies, the average effect size was significant


Total screentime

Video gameviolence

Parentalinvolvement

Sex: 1=male1=female

Time 1 Time 1 and 2 Time 2 Time 2

Hostileattribution bas

Schoolgrades

Peerrejection

Verbalaggression

Physicalaggression

Relationalaggression

Prosocialbehavior

0.18**

0.12*

0.15*0.22***0.16**

0.16**

0.10+

0.12*0.27***0.25***0.18**

0.10+

0.15**

0.26***

0.19***

0.20***

0.37***

0.29***

0.09+

Figure 3.—Longitudinal path analysis of 430 elementary school children documenting different effects of amount andcontent on school performance and aggressive/prosocial behaviors. +) P=0.10. *) P=0.05. **) P=0.01. ***) P=0.001.



TABLE IV.—Studies of violent video games on aggressive behavior.


Experimental studies on aggressive behaviorCooper et al. (1986) 45 84 5th graders V game increased aggressive toy choice and play

for girls, but not aggressive or prosocial beha-vior (punishment and reward time) toward otherchildren

Graybill et al. (1987) 33 79 7-11 years V games did not decrease helpful and increasehurtful responses compared to NV games

Chambers et al. (1987) 46 160 3rd-4th and V games decreased prosocial (money dona-7th-8th grades tions) behaviors more more than NV game

Silvern et al. (1987) 47 28 4-6 years Increased aggressive behavior and decreased pro-social behavior in free play after V game play orviolent cartoon viewing (no NV control)

Winkel et al. (1987) 28 56 8th graders No effect of V game on aggressive behavior(money deduction)

Brusa (1987 - unp) 32 6-year-olds Physical and verbal aggression toward same-sexpeer higher, but not significantly, after playingCentipede (V game) than after playing computerpinball.

Shutte et al. (1988) 48 31 5-7 years V game increased physically aggressive beha-viors to children and toys in free play

Irwin et al. (1995) 30 60 2nd graders V game increased verbal and physical aggressionto a peer in a frustrating situation and to toys ina free-play situation

Cohn (1995 - unp) 124 6th-8th graders No significant differences in aggressive behaviorafter V game compared to NV game (noise blast)

Anderson et al. (in press) 161 9-12 years Playing V E- and T-rated games increased ag-gressive behavior (noise blast)

Correlational studies on aggressive behaviorDominick (1984) 37 250 10th-11th grade Arcade game playing (but not home use) corre-

lated with self-reported physical aggression andaggressive delinquency

McClure et al. (1986) 49 290 9th-12th grade High-rate VG playing not correlated with psy-chopathic deviance or neuroticism

Lin et al. (1987) 50 189 4th–6th grade Arcade use of games (but not home use) corre-lated with teacher ratings of impulsiveness and ag-gressiveness

Fling et al. (1992) 51 153 6th-12th grade Amount of VG play correlated with self-reportedand teacher ratings of aggressive behavior

Cohn (1995 - unp) 124 6th-8th graders Higher aggressive behavior shown in lab by hi-gh experience Mortal Kombat players comparedto low experience players (noise blast)

Funk et al. (1996) 52 357 7th and 8th grade For girls, but not boys, V game play associatedwith lower self-esteem, poorer behavioral con-duct, and other self-perceptions

van Schie et al. (1997) 53 346 7th and 8th grade Greater amount of video game play negativelycorrelated with prosocial behavior but not signi-ficantly correlated with aggressive behavior

Wiegman et al. (1998) 54 278 7th and 8th grade Greater amount of video game play and prefe-rence for violent games negatively correlated withprosocial behavior; preference, but not amount,positively correlated with aggressive behavior

Janey (1999 - unp) 201 5th and 6th grade Amount of VG play associated with self-reportedaggressive boys behaviors and attitudes after con-trolling for masculine ideology and father availa-bility, but not with teacher-reported aggression

unp = unpublished; V = violent, NV = non-violent; VR = virtual reality; HR = heart rate, BP = blood pressure, SBP = systolic bloodpressure, DBP = diastolic blood pressure

Continued Table IV


at r=0.24. In his more recent meta-analysis,across studies with either children or adults, theaverage effect size was r=0.23.

Aggressive behavior

The GAM predicts that if violent videogames can cause increases in arousal, ag-gressive affect, and aggressive cognitions,then these in turn may cause increases in ag-gressive behavior. At the time of this writing,there have been 10 experimental studies(Cooper et al.,45 Graybill et al.,33 Chambers etal.,46 Silvern et al.,47 Winkel et al.,28 Shutte etal.,48 Irwin et al.,30 Brusa and Cohn unpub-

lished, Anderson et al. submitted), 18 corre-lational studies (Dominick,37 McClure et al.,49

Lin et al.,50 Fling et al.,51 Funk et al.,52 vanSchie et al.,53 Wiegman et al.,54 Colwell etal.,38 Funk et al.,55 Durkin et al.,24 Colwell etal.,56 Gentile et al.,26 Vandewater et al.,57 Cohnand Janey unpublished, Anderson et al. andAnderson et al. in press), and 3 longitudinalstudies (Ihori et al.,25 Slater et al.,58 Andersonet al. submitted) of video games and aggres-sive behavior conducted with children andadolescents (Table IV). Again, only a minor-ity of the studies meet our criteria for highquality studies. Most of the experimental stud-ies do not match violent and non-violent


Table IV continued.


Correlational studies on aggressive behaviorColwell et al. (2000) 38 204 12-14 years Game exposure correlated with aggressive beha-

viorsFunk et al. (2000) 55 364 4th and 5th grade Preference for violent games correlated with poo-

rer behavioral conduct (self-perception)Walsh (2000) 64 137 8th and 12th grade Greater game exposure correlated with physical

figthsDurkin et al. (2002) 24 1 304 3 10th graders High computer game use correlated with greater

aggressive behavior compared to low computergame use

Colwell et al. (2003) 56 509 12-14 years Amount of play and V game exposure predictedaggression in samples of UK and Japanese ado-lescents

Gentile et al. (2004) 26 607 8th and 9th grade Greater V game play positively correlated withphysical fights after controlling for hostility

Vandewater et al. (2005) 57 225 7-12 years Family conflict correlated with increased V ga-me play for girls, but not for boys

Anderson et al. (in press) 615 9-12 years and V game play correlated with violent behavior17-29 years

Anderson et al. (in press) 189 14-19 years V game play correlated with verbal and physicalaggression and violent behavior

Longitudinal studies on aggressive behaviorIhori et al. (2003) 25 807 5th and 6th grade Increased amount of video game play related to

later indirect aggression (boys and girls) and la-ter physical aggression (boys)

Slater et al. (2003) 58 2 550 3 6th and 7th grade Media violence exposure (including video ga-mes) increases aggressive cognitions and beha-viors, which in turn increase media violence ex-posure, which increases aggressive cognitionsand behaviors etc. over a two-year period (una-ble to separate out VGs)

Anderson et al. (in press) 430 3rd-5th grade V VG play early in a school year related to laterincreases in verbal and physical aggression (as ra-ted by self, peers, and teachers) after controllingfor sex, race, total screen time, hostile attributionbias, parental involvement, and earlier verbal andphysical aggression

unp = unpublished; V = violent, NV = non-violent; VR = virtual reality; HR = heart rate, BP = blood pressure, SBP = systolic bloodpressure, DBP = diastolic blood pressure.


games on other confounding dimensions.Many of the experimental studies were con-ducted in the 1980s, before games includedmuch violent content against other gamecharacters; therefore there cannot have beenparticularly strong differences between theviolent and non-violent games (for example,Brusa, unpublished data, 1987, usedCentipede as a violent game, which wouldnot be considered to be a violent game by to-day’s standards). Several include measuresof aggressive behavior that are probably nottruly aggressive behavior. For example, someused the choice of an aggressive or non-ag-gressive toy as measures of aggression 45 andothers used money donations.31, 37 There isnothing inherently wrong with using thesetypes of proxies for aggressive behavior, butit is important that these measures be vali-dated to show that children who donate lessmoney are also the types of children whowould be likely to hit other children outsideof the laboratory. This type of validity studyhas been conducted with noise blasts, sug-gesting that noise blasts are a valid approachto studying aggressive behavior in the labo-ratory (at least for adults). Unfortunately, veryfew experimental studies have been con-ducted with children since the mid-1990s (al-though there have been several excellentstudies with undergraduates and adults). Inthe most recent, 161 nine- to twelve-year-olds were randomly assigned to play either aviolent or non-violent video game rated asage-appropriate by the Entertainment RatingsSoftware Board. The games were matchedon frustration levels, and although enter-tainment levels differed, they were not relat-ed to aggressive responding. After playingthe video game, children played a competi-tive reaction time game against a same-sexpartner, who was in reality the computer.When they were slower than their partner,they received a blast of white noise throughheadphones. When they were faster, they setthe level of noise blast their partner received.Children who played the violent video gamegave significantly more high-intensity noiseblasts than children who played the non-vi-olent game.

Many of the correlational and one of the

longitudinal studies share the problem ofconfounding amount of video game play withthe content of games played 24, 25, 37, 38, 49-53

(Janey BA, unpublished data, 1999). As not-ed earlier, amount of play is not a good in-dicator of violence exposure, although theyare correlated, at least with recent data. Forany studies conducted prior to about 1992,the correlation between amount of play andviolence exposure is probably low becausethere were far fewer truly violent games. Forthis reason, it is surprising that amount ofplay still predicts aggressive behaviors inmost of these studies. In comparison, all ofthe correlational studies where amount of vi-olent game exposure is measured show acorrelation between violence exposure andaggressive behavior (for at least some of thesample). Correlational studies can not demon-strate causality – it could be argued that play-ing violent games increases aggression, orthat naturally hostile children prefer violentgames. However, although correlational stud-ies can not prove causality, they could dis-prove causality if no relation is found be-tween variables and they can provide supportfor causal theories. For example, in a sampleof 607 8th and 9th graders, violent game ex-posure was correlated with physical fights.26

Because personality trait hostility was mea-sured in this study, it allowed a test of the hy-pothesis whether the correlation between vi-olent game exposure and fights was spuri-ously caused by the third variable of hostilepersonality. If hostility were the necessaryfactor, then only hostile children would get in-to physical fights, and children with the low-est hostility scores would not get into fightsregardless of their video game habits. Figure4 displays the percentages of students who re-port being involved in physical fights withinthe previous year. Children with the lowesthostility scores are almost 10 times more like-ly to have been involved in physical fights ifthey play a lot of violent video games than ifthey do not play violent games (38% com-pared to 4%). Children with the highest hos-tility scores are over 2 times more likely to beinvolved in fights if they play a lot of violentgames than if they do not (63% compared to28%). In fact, the least hostile children who



play a lot of violent video games are morelikely to be involved in fights than the mosthostile children if those children do not playviolent video games.

Slater et al.58 addressed the developmentalcourse of media violence exposure and ag-gressive cognitions and behavior in a longi-tudinal study of 2 550 6th and 7th graders.Students were surveyed 4 times across 2 yearsabout their media violence exposure (movies,video games, and Internet). In short, theirdata provide evidence of a downward spiral,where media violence exposure increasesaggressive cognitions and behaviors, whichin turn increase media violence exposure,which further increases aggressive cognitionsand behaviors. Unfortunately, the authorscombined all media violence, so it is unclearhow much of this effect may be due to vio-lent video games.

In the longitudinal study of 3rd-5th graders,violent video game exposure was measuredseparately from other media violence andseveral subtypes of aggressive behavior weremeasured. Figure 3 displays the results ofpath analyses, in which variables at Time 1are shown at the left and variables at Time 2are shown at the right. Video game violenceexposure and total screen time both increasedhostile attribution bias, which in turn was re-lated to increased verbally, physically, andrelationally aggressive behavior, as well asdecreased prosocial behavior. Verbal ag-gression and prosocial behavior were in turnrelated to Time 2 peer acceptance. Videogame violence exposure was also directly re-lated (over and above the mediated path viahostile attribution bias) to increased verbal ag-gression, increased physical aggression, anddecreased prosocial behavior. Having par-ents involved in children’s media habits re-sulted in children showing less relational ag-gression and more prosocial behavior.

The results of studies with children andadolescents are again consistent with thestudies with adults. Meta-analyses 44 of thestudies with children show a significant over-all relation between violent video game ex-posure and aggressive behavior of r=0.18.Combining studies with children and adults,27

the overall relation is r=0.19.

Summary

In general, combining the results of ex-perimental, correlational, and longitudinalstudies, there is a preponderance of evidencethat violent video game play is related to ag-gressive affect, physiological arousal, ag-gressive cognitions, and aggressive behav-iors. We feel that it is fair to consider violentvideo games (and media violence in gener-al) as one risk factor for aggressive behavior.They are not the only risk factor for aggres-sion, nor are they the largest risk factor.However, they appear to be a significant riskfactor when one considers the large numberof children exposed to them. Furthermore,among the dozens of documented risk factorsfor aggressive behavior,59 media violence isunique in that it is the risk factor that is mosteasily controlled.

In this review, we have attempted to notemethodological problems that are common inthe literature. Because so many of the stud-ies have at least one methodological problem,it could be argued that the relation betweenviolent games and aggression might be arti-ficially high because of the flaws. In his meta-analysis, Anderson 27 coded whether eachstudy had any of 9 types of methodological


Perc

enta

ge o

f s

tude

nts

invo

lved

in p

hysi

cal f

ight

s

70

40

30

20

0Low hostility High hostility

60

50

10

High violent game playLow violent game play

63

38

28

4

Figure 4.—Percentages of 8th and 9th graders involved inphysical fights, split by hostility and violent video game ex-posure. The low and high hostility bars represent childreniin the bottom and top quartiles on a measure of trait hos-tility, respectively. The low and high violent video gameplay bars represent children in the bottom and top quar-tiles on a measure of exposure to violent video games.Data are extracted from Gentile, Lynch, Linder & Walsh,2004. Total N=607.27


flaws and then compared the meta-analyticresults between studies using best practices(i.e., having none of 9 flaws) with the stud-ies with flaws. In all cases (for arousal, ag-gressive affect, cognition, and behavior), thebest practice studies showed higher effectsizes than the studies with flaws. Therefore,the consistent relation between violent videogames and aggression is not due to method-ological flaws – indeed, when studies aremore carefully controlled, the effect seemsto be more easily found.

School performance

Although studies of violent video gamesget the most attention, many other effectshave been studied. One that is of particularconcern for children and adolescents is the ef-fects on academic achievement. Successfulor poor academic performance can have life-long consequences. Many studies have doc-umented negative correlations between videogame use and school performance for chil-dren, adolescents, and college students 13, 26,

38, 60-64 (Anderson et al., in press). Overall,the preponderance of studies demonstrates aconsistent negative correlation between recre-ational video game play and school perfor-mance. However, there is also a widespreadbelief that some types of computer or soft-ware use could have positive effects onschool performance.

Durkin et al.24 have recently attempted toargue that there is no negative relation be-tween video game use and school grades,aggression, or many other variables. Specifi-cally, they reported that students who neveruse computers have lower grades (meangrade point averages, GPA=2.53) than lowuse (2.79) and high use (2.61) students (us-age measured with a single seven-point Likertscale from “never” to “daily”). Yet, they basedthese conclusions on survey data gatheredin 1988, before games became particularlyviolent, and when there was a strong rela-tionship between computer use and socio-economic status. That is, children from poor-er families were far less likely to own com-puters in 1988. Children from poorer families

also typically have poorer school perfor-mance. Therefore, this is not a fair compar-ison. The fair test is whether the low useand high use groups differ, since both groupscome from families that own computers. TheDurkin et al. data actually show that highuse students have significantly lower GPAsand exhibit significantly more aggressive be-havior than low use students. This com-pletely changes the interpretation of theirdata, as it can explain why students whonever use computers also have higher de-pressed mood, lower self-esteem, higher dis-obedience, substance use, and truancy, andlower GPAs than computer users. All of theserelations would be predicted by lower so-cioeconomic status.

Other studies have been more careful intheir analyses. In one early study, childrenwho used computers to play games per-formed more poorly in school, whereas thosewho used computers for schoolwork per-formed better in school.62 Furthermore, manyvideo games have been designed specifical-ly to teach academic skills and concepts. Thequality of research in this domain varies great-ly, but a recent meta-analysis of high-qualitystudies shows that educational software hasa significant positive effect on academic skillsand student achievement.65 The average cor-relation is 0.35 for educational games teach-ing reading skills, and 0.45 for games teach-ing math skills.

Given these apparently conflicting results,how are we to understand the effects of videogames on school performance? The majorityof studies on school performance have beenatheoretical, and provide little predictive pow-er about when we would expect video gamesto have positive or negative effects. However,2 recent studies shed some light suggestingthat there may be several dimensions alongwhich video games may have effects.

In a correlational study of adolescents,path analyses revealed that the amount oftime adolescents spent playing video gamesdirectly predicted poorer grades, but wasnot directly related to aggressive behaviors.However, playing violent games directlypredicted aggressive behaviors, but did not



predict poorer grades.26 This pattern wasreplicated in a longitudinal study of ele-mentary school children (Figure 4, Andersonet al., in press). Children who spent a lot oftime playing video games had poorer gradeslater in the school year, but amount of playdid not directly predict aggressive or proso-cial behaviors. In contrast, playing violentgames directly predicted increased aggres-sive behaviors and decreased prosocial be-haviors, but did not directly predict grades.These studies and others suggest that thereare at least 4 theoretically independent di-mensions along which video games canhave effects: amount, content, form, andmechanics.

The dimensional approach to video game ef-fects

AMOUNT

As shown in Figure 4, the total amount oftime that children and adolescents spendplaying video games appears to be respon-sible for the negative correlation betweenvideo games and school performance. Thisprovides some support for the displacementhypothesis, which states that one way elec-tronic media have effects on children is bydisplacing other activities.66 In essence, everyhour that children play entertainment videogames is an hour that they are not doinghomework, reading, exploring, creating, orany of several other activities that may havemore educational benefits. In addition, sev-eral other scientifically-documented effectsare likely to be related to amount of play.Some research documents that the amount oftime spent with video games can be linked tolower activity levels, higher weight, and high-er risk of obesity.67-69 In addition, there are anumber of muscular and skeletal disorders as-sociated with heavy computer or video gameuse, such as tendonitis and nerve compres-sion. There is even a form of tendonitisnamed “Nintendinitis”, caused by repeated-ly pressing game-controller buttons with one’sthumb.70

However, total amount may not be the on-ly aspect of amount that matters – how chil-

dren distribute their time with video gamesmay also enhance or reduce the effective-ness of learning from them. Educational psy-chologists have shown that learning andtransfer are most likely if the learner prac-tices some each day (distributed practice)rather than “cramming” – trying to learn itall at once in one long session (massed prac-tice). From an educational standpoint, videogames encourage a close-to-optimal combi-nation of massed and distributed practice.Initial attempts at the game provide immedi-ate feedback and most people will keep play-ing until they begin to show some progress.Such massed practice eventually begins toproduce diminishing returns (when a plateauis reached or fatigue sets in). However, therepetition will have begun to develop bothphysical and mental skills on parts of thetask. Each subsequent encounter with thegame provides memory benefits – namely,relearning anything that was forgotten, pro-viding new cues for memory, interpretingnew information or examples with what is al-ready in memory, and reorganizing the mem-ory accordingly. This combination of massedpractice to build sufficient initial mastery toplay the game, followed by distributed prac-tice over days or weeks is optimal for learn-ing.71-73 Although there is a great deal of re-search on this aspect of repetition andamount in traditional education, there is al-most none with video games.

There is also some evidence that videogame “addiction” is a problem that the med-ical community is beginning to face 74, 75

(Gentile, Tapscott, submitted). In a recent setof analyses with younger adolescent (8th/9th

grade) and older adolescent (undergraduate)samples, the prevalence of video game ad-diction ranged between 6% and 13% ofgamers. Among younger adolescents, ad-dicted gamers averaged 21 h per week play-ing video games (compared to 8 h/weekamong non-addicted gamers). Although thereis not yet a medically agreed-upon diagnosisfor video game addiction, it appears to havehigh construct validity (Gentile, Tapscott, inpress). Furthermore, addicted gamers displaypatterns of comorbidity similar to those dis-played by other types of addicts, and ad-



dicted gamers display higher emotional re-activity to games than non-gamers and non-addicted gamers (Gentile, Tapscott, submit-ted). We do not mean to imply, however,that addiction can be reduced to amount ofplay. To be considered addicted, one mustplay to the extent that it damages one’s socialand/or professional life. However, addictedgamers are also more likely to suffer any ef-fects that may be due to amount of play.

CONTENT

Most of the research on video games hasdocumented what are likely to be effects ofthe content of the games. For example, the re-search reviewed here showing effects of vi-olent games represents the effects of violentcontent. As described above, there is evi-dence that violent content has effects thatare specific to aggressive and prosocialthoughts, feelings, and behaviors, and thatthese effects are theoretically and empirical-ly independent of the effects of amount (al-though there is a significant correlation be-tween amount and violence exposure – peo-ple who spend a lot of time playing videogames also tend to be exposed to more vio-lent content). There are also studies showingthat specially designed video games can teachchildren healthy skills for the self-care ofasthma and diabetes.76, 77 These games havesucceeded on multiple levels, modifying theattitudes, skills, and behaviors that they weredesigned to teach. Recent advances in virtu-al reality (VR) technology have also made itpossible for physicians and therapists to in-corporate them into therapeutic programs.For example, studies have shown that VRprograms can be effective in helping patientsmanage the pain of burns and wound care,as well as desensitizing phobic patients totheir fears, such as fear of heights, flying, andspiders.78, 79 Each of these studies on violentvideo games, educational video games, sys-tematic desensitization via virtual reality, andhealth promotion video games documentseffects of game content. In the case of edu-cational games, these are intentional contenteffects; in the case of violent games, theseare unintentional content effects.

FORM

There is a large body of literature with tele-vision and film on what are called “formalfeatures.” In general, this literature examineshow people understand the formal conven-tions of the medium and what they signify.For example, most people extract differentmeanings from editing cuts and dissolves. Ifa couple enters a bedroom and the scenegradually dissolves to the next morning, mostpeople take a very different meaning awayfrom it than if the scene abruptly cuts to thenew scene the next morning. Similarly, sev-eral of the studies on video games suggestpatterns of effects that are not due to thecontent per se, but to the form in which it ispresented. There appear to be many featuresthat are capable of differential effects, only 3of which will be discussed here.

First, some games require the use of 2Drepresentations to provide 3D informationand navigation.80 If players play these typesof games, then we should be able to docu-ment improvements in their ability to use 2Dinformation for 3D navigation. Although da-ta are limited at this point, 2 studies suggestthat these skills may be learnable and showtransfer. In the first of these, Greenfield etal.81 showed that skill in a game requiring3D navigation was related to 3D mental vi-sualization skill. Second, Rosser et al.82

showed that demonstrated skill on videogames and past experience with video gameswere the best predictors of surgeons’ ad-vanced laparoscopic surgical skills.Surprisingly, video game experience and skillwere better predictors of advanced surgicaltechniques than either amount of medicaltraining or number of laparoscopic surgeriesperformed. These findings are difficult to ex-plain using only amount or content types ofexplanations, but seem to fit a form argu-ment - that more experience with games leadsto increased skill in using 2D representationsfor 3D navigation.

Second, some games require constant scan-ning of the screen for information – for ex-ample, in action games it is critical to be con-stantly scanning all parts of the screen be-cause an enemy could jump out from any-



where and a quick reaction is necessary. Ifplayers play these types of games, then weshould be able to document improvements invisual attention skills to computer screens.This pattern has indeed been found, com-paring different types of games that requiredifferent types of attention.83, 84

Third, some games portray their subjectswith more realism than others. If players playgames with similar content, but with variedrealism, then we should be able to documentbetter learning and transfer. There are fewdata to support this at the present time, butit makes intuitive sense when consideringsimulation games. For example, if one want-ed to learn to fly an airplane, more realisticsimulators should result in better learningthat would transfer to outside situations. Someresults appear to suggest this may be cor-rect, such as the finding that playing MortalKombat with the blood setting turned on (sothat injuries are accompanied by depictionsof spurts of blood from the body) led togreater increases in hostility and arousal thanwith it turned off.85 In this study, the gamecontent is equally violent in both cases, butthe form is different. Furthermore, effect sizeshave been increasing as games have beenbecoming more realistic.6 However, theremay be many other reasons for this result,so this interpretation should be viewed withcaution. Certainly gamers cite realism as oneof the primary facets that are important tothem.86 Future studies will need to be con-ducted to determine the amount to whichrealism (and what dimensions of realism)makes a difference.

MECHANICS

The types of mechanical input/output de-vices used to play the games could also showeffects. If players play games with identicalcontent, but with varied mechanical inter-faces, where some have increased similarityto reality, then researchers should be able todocument better learning and transfer. Forexample, playing a driving simulation gamewith a wheel and pedals should improve dri-ving skill more than playing the same gamewith a mouse and a keyboard. Similarly, if

surgical simulators are designed with inputdevices that are similar to actual surgical in-struments, then learning and transfer shouldbe improved compared to input devices withless verisimilitude. To date, no studies havebeen conducted to test this hypothesis. Thisissue is further complicated because me-chanics are not entirely independent fromform. Game movements are guided both byvisual information gathered from the screenand use of the input/output devices, andthese inform each other (for a complete de-scription of how visual information, move-ment, and proprioception are linked seeGibson 87). For example, in a first-personshooter game such as Halo, gamers can oftenshoot at an enemy in a standard game viewor through a magnifying scope on theweapon their character is using. The opticsare very different under these 2 situations,as a small movement with the mouse changeswhat is seen on the screen a small amount inthe standard view, but a large amount whenviewed through the scope. Research is need-ed to determine how mechanics affect videogame effects.

Conclusions

When attempting to understand all of thevaried empirically-identified effects of videogames, conceptualizing the effects along the4 dimensions of amount, content, form, andmechanics is useful. One clear benefit of thisapproach is that it allows researchers andthe public to move beyond the dichotomousthinking that has too often characterized thedebate around video games, in which thediscussion is reduced to a question ofwhether video games are good or bad. As thepreceding analysis shows, this is too sim-plistic a question. The same video gamecould have both positive and negative ef-fects, depending on what dimension oneconsiders. For example, playing the GrandTheft Auto series of games (in which oneplays a criminal sociopath) a lot each daycould hamper school performance (amounteffect), increase aggressive thoughts and be-haviors (content effect), improve visual at-



tention skills (form effect), and improve dri-ving skills if one plays with a driver’s wheeland pedals or shooting skills if one plays us-ing a gun input/output device (mechanicseffects).

Riassunto

Effetti sui bambini e sugli adolescenti dei video gameviolenti: una revisione della letteratura

Per 1) determinare i multipli effetti, 2) offrire os-servazioni critiche circa le affermazioni presenti inletteratura e 3) fornire una prospettiva più ampia percomprendere la ricerca sugli effetti dei video game so-no stati rivisti gli studi eseguiti su bambini e adole-scenti che giocano con i video game violenti. Questarevisione comprende le considerazioni generali teo-riche e metodologiche sulla violenza dei mezzi diinformazione e la descrizione del general aggressionmodel (GAM). Tutti i lavori pubblicati sono stati va-lutati tenendo conto del GAM. Sono stati rivisti tuttii lavori pubblicati, comprese le meta-analisi, così co-me dati importanti non ancora pubblicati, quali co-municazioni a congressi. In generale, l’evidenza sup-porta le ipotesi che giocare con video game violentiabbia un effetto sull’aggressività, sulla fisiologia, sul-la percezione dell’aggressività e sui comportamenti ag-gressivi. Sono anche stati valutati gli effetti dei videogame violenti sul rendimento scolastico, e questa re-visione concorda con l’approccio dimensionale aglieffetti dei video game. L’approccio dimensionale va-luta gli effetti dei video game in termini di quantità,contenuto, forma e meccanicità e sembra presentaremolti vantaggi per la comprensione e la predizione deidiversi tipi di effetti dimostrati in letteratura.

Parole chiave: Video game - Età pediatrica - Violenza.

References

1. Lidsky TI, Schneider JS. Lead neurotoxicity in children:basic mechanisms and clinical correlates. Brain2003;126:5-19.

2. Needleman HL, Gatsonis C. Low level lead exposureand the IQ of children. JAMA 1990;263:673-8.

3. Anderson CA, Bushman BJ. The effects of media vio-lence on society. Science 2002;295:2377-8.

4. Freedman JL. Media violence and its effect on aggres-sion. Toronto, Canada: University of Toronto PressInc.; 2002.

5. Interactive Digital Software Association. Video Games& Youth Violence: Examining the Facts. Washington,DC: Interactive Digital Software Association; 2001.

6. Gentile DA, Anderson CA. Violent video games: thenewest media violence hazard. In: Gentile DA editor.Media violence and children. Westport, CT: PraegerPublishing; 2003. p.131-52.

7. Huesmann LR, Taylor LD. The case against the caseagainst media violence. In: Gentile DA editor. Media

Violence and Children. Westport, CT: Praeger Publi-shing; 2003. p.107-30.

8. Carnagey NL, Anderson CA. Theory in the study ofmedia violence: the general aggression model. In:Gentile DA editor. Media Violence and Children.Westport, CT: Praeger Publishing; 2003. p. 87-105.

9. Anderson CA, Carnagey NL. Violent evil and the gen-eral aggression model. In: Miller A editor. The SocialPsychology of Good and Evil. New York: GuilfordPublications; 2004.p.168-92.

10. Anderson CA, Huesmann LR. Human aggression: a so-cial-cognitive view. In: Hogg MA, Cooper J editors.Handbook of Social Psychology. London: SagePublications; 2003.p.296-323.

11. Geen RG. Human Aggression. Philadelphia, PA: OpenUniversity Press; 2001.

12. Anderson CA, Bushman BJ. Effects of violent videogames on aggressive behavior, aggressive cognition, ag-gressive affect, physiological arousal, and prosocialbehavior: a meta-analytic review of the scientific liter-ature. Psychol Sci 2001;12:353-9.

13. Anderson CA, Dill K. Video games and aggressivethoughts, feelings, and behavior in the laboratory andin life. J Pers Soc Psychol 2000;78:772-90.

14. Anderson CA, Carnagey NL, Flanagan M, Benjamin AJ,Eubanks J, Valentine JC. Violent video games: specif-ic effects of violent content on aggressive thoughtsand behavior. Adv Exp Soc Psychol 2004;36:199-249.

15. Persky S, Blascovich J. The Price of Technology:Immersive Virtual Video Games and Aggression.16thAnnual Convention of the American PsychologicalSociety; 2004, May 27-30, Chicago, IL.

16. Anderson CA, Bushman BJ. External validity of “trivial”experiments: the case of laboratory aggression. RevGen Psychol 1997;1:19-41.

17. Bernstein S, Richardson D, Hammock G. Convergentand discriminant validity of the Taylor and Buss measuresof physical aggression. Aggress Behav 1987;13:15-24.

18. Carlson M, Marcus-Newhall A, Miller N. Evidence fora general construct of aggression. Pers Soc PsycholBull 1989;15:377-89.

19. Giancola PR, Zeichner A. Construct validity of a com-petitive reaction-time aggression paradigm. AggressBehav 1995;21:199-204.

20. Ballard ME, Panee CD, Engold ED, Hamby RH. Brieflongitudinal effects of violent video game play. Societyfor Research in Child Development Conference, 2001,19 April, Minneapolis, MN.

21. Hind PA. A study of reported satisfaction with differ-entially aggressive computer games amongst incar-cerated young offenders. Criminol Legal Psychol1995;22:28-36.

22. Funk J, Hagan J, Schimming J. Children and electron-ic games: a comparison of parents’ and children’s per-ceptions of children’s habits and preferences in a UnitedStates sample. Psychol Rep 1999;85:883-8.

23. Fleming MJ, Rickwood DJ. Effects of violent versusnonviolent video games on children’s arousal, aggres-sive mood, and positive mood. J Appl Soc Psychol2001;31:2047-71.

24. Durkin K, Barber B. Not so doomed: computer play andpositive adolescent development. J Appl Dev Psychol2002;23:373-92.

25. Ihori N, Sakamoto A, Kobayashi K, Kimura F. Doesvideo game use grow children’s aggressiveness? Resultsfrom a panel study. In: Arai K editor. Social contribu-tions and responsibilities of simulation and gaming.Proceedings of the 34th annual conference of the in-ternational simulation and gaming association. Tokyo,Japan: Japan Association of Simulation and Gaming;2003.p.221-30.



26. Gentile DA, Lynch PJ, Linder JR, Walsh DA. The ef-fects of violent video game habits on adolescent hos-tility, aggressive behaviors, and school performance. JAdolesc 2004;27:5-22.

27. Anderson CA. An update on the effects of playing vi-olent video games. J Adolesc 2004;27:113-22.

28. Winkel M, Novak DM, Hopson H. Personality factors,subject gender, and the effects of aggressive videogames on aggression in adolescents. J Res Pers 1987;21:211-23.

29. Segal KR, Dietz WH. Physiologic responses to playinga video game. Am J Dis Child 1991;145:1034-6.

30. Irwin RA, Gross AM. Cognitive tempo, violent videogames, and aggressive behavior in young boys. J FamViolence 1995;10:337-50.

31. Lynch PJ. Hostility, type A behavior, and stress hor-mones at rest and after playing violent video games inteenagers. Psychosom Med 1999;61:113.

32. Graybill D, Kirsch JR, Esselman ED. Effects of playingviolent versus nonviolent video games on the aggres-sive ideation of aggressive and nonaggressive chil-dren. Child Study J 1985;15:199-205.

33. Graybill D, Strawniak M, Hunter T, O’Leary M. Effectsof playing versus observing violent versus nonviolentvideo games on children’s aggression. Psychology: AQuarter’s J Hum Behav 1987;24:1-8.

34. Kirsh SJ. Seeing the world through Mortal Kombat-col-ored glasses: violent video games and the develop-ment of a short-term hostile attribution bias. Childhood1998;5:177-84.

35. Funk JB, Buchman DD, Myers M, Jenks J. Asking theRight Questions in Research on Violent ElectronicGames. Annual Meeting of the American PsychologicalAssociation; 2000, 7 August, Washington, DC.

36. Funk JB, Buchman DD, Jenks J, Bechtoldt H. Playingviolent video games, desensitization, and moral eval-uation in children. Appl Dev Psychol 2003;24:413-36.

37. Dominick JR. Videogames, television violence, and ag-gression in teenagers. J Commun 1984;34:136-47.

38. Colwell J, Payne J. Negative correlates of computergame play in adolescents. Br J Psychol 2000;91:295-310.

39. Funk JB, Baldacci HB, Pasold T, Baumgardner J.Violence exposure in real-life, video games, television,movies, and the internet: is there desensitization? JAdolesc 2004;27:23-39.

40. Krahé B, Möller I. Playing violent electronic games,hostile attributional style, and aggression-related normsin German adolescents. J Adolesc 2004;27:53-69.

41. Crick NR. Relational aggression: the role of intent at-tributions, feelings of distress, and provocation type.Dev Psychopathol 1995;7:313-22.

42. Crick NR, Dodge K. A review and reformulation of so-cial information-processing mechanisms in children’ssocial adjustment. Psychol Bull 1994;115:74-101.

43. Crick NR, Grotpeter JK. Relational aggression, genderand social-psychological adjustment. Child Develop-ment 1995;66:710-22.

44. Anderson CA. Violent video games and aggressivethoughts, feelings, and behaviors. In: Calvert SL, JordanAB, Cocking RR editors. Children in the Digital Age.Westport, CT: Praeger Publishers; 2002.p.101-19.

45. Cooper J, Mackie D. Video games and aggression inchildren. J Appl Soc Psychol 1986;16:726-44.

46. Chambers JH, Ascione FR. The effects of prosocial andaggressive videogames on children’s donating andhelping. J Genet Psychol 1987;148:499-505.

47. Silvern SB, Williamson PA. The effects of video gameplay on young children’s aggression, fantasy, and proso-cial behavior. J Appl Dev Psychol 1987;8:453-62.

48. Shutte NS, Malouff JM, Post-Gorden JC, Rodasta AL.Effects of playing videogames on children’s aggres-

sive and other behaviors. J Appl Soc Psychol 1988;18:454-60.

49. McClure RF, Mears FG. Videogame playing and psy-chopathology. Psychol Rep 1986;59:59-62.

50. Lin S, Lepper MR. Correlates of children’s usage ofvideogames and computers. J Appl Soc Psychol 1987;17:72-93.

51. Fling S, Smith L, Rodriguez D, Thornton D, Atkins E,Nixon K. Videogames, aggression, and self-esteem: asurvey. Soc Behav Pers 1992;20:39-46.

52. Funk JB, Buchman DD. Playing violent video and com-puter games and adolescent self-concept. J Commun1996;46:19-32.

53. van Schie EGM, Wiegman O. Children and videogames:leisure activities, aggression, social integration, and schoolperformance. J Appl Soc Psychol 1997;27:1175-94.

54. Wiegman O, van Schie EGM. Video game playing andits relations with aggressive and prosocial behaviour.Br J Soc Psychol 1998;37:367-78.

55. Funk JB, Buchman DD, Germann JN. Preference for vi-olent electronic games, self-concept, and gender dif-ferences in young children. Am J Orthopsychiatry2000;70:233-41.

56. Colwell J, Kato M. A comparison of negative corre-lates of computer game play in UK and Japanese ado-lescents. In: Arai K editor. Social Contributions andResponsibilities of Simulation and Gaming. Tokyo,Japan: Japan Association of Simulation and Gaming;2003.p.693-702.

57. Vandewater EA, Lee JH, Shim M. Family conflict and vi-olent electronic media use in school-aged children.Media Psychol 2005;7:73-86.

58. Slater MD, Henry KL, Swaim RC, Anderson LL. Violentmedia content and aggressiveness in adolescents.Commun Res 2003;30:713-36.

59. U.S. Department of Health and Human Services. YouthViolence: A Report of the Surgeon General. Rockville,MD: U.S. Government Printing Office; 2001.

60. Creasey GL, Myers BJ. Video games and children: ef-fects on leisure activities, schoolwork, and peer in-volvement. Merrill-Palmer Q 1986;32:251-62.

61. Harris MB, Williams R. Video games and school per-formance. Education 1985;105:306-9.

62. Lieberman DA, Chaffee SH, Roberts DF. Computers,mass media, and schooling: functional equivalence inuses of new media. Soc Sci Comput Rev 1988;6:224-41.

63. Roberts DF, Foehr UG, Rideout VJ, Brodie M. Kids &media @ the new millennium. Menlo Park, CA: KaiserFamily Foundation; 1999.

64. Walsh D. Testimony submitted to the United StatesSenate Committee on Commerce, Science, andTransportation. Hearing on the Impact of InteractiveViolence on Children. March 21, 2000. http://com-merce.senate.gov/hearings/0321wal1.pdf

65. Murphy R, Penuel W, Means B, Korbak C, Whaley A.E-DESK: A Review of Recent Evidence on theEffectiveness of Discrete Educational Software. MenloPark, CA: SRI International; 2001. http://ctl.sri.com/pub-lications/downloads/Task3_FinalReport3.pdf, AccessedMay 19, 2004.

66. Huston AC, Donnerstein E, Fairchild H, Feshbach ND,Katz PA, Murray JP et al. Big world, small screen: therole of television in American society. Lincoln,Nebraska: University of Nebraska Press; 1992.

67. Berkey CS, Rockett HR, Field AE, Gillman MW, FrazierAL, Camargo CA Jr et al. Activity, dietary intake, andweight changes in a longitudinal study of preadolescentand adolescent boys and girls. Pediatrics 2000;105:E56.

68. Subrahmanyam K, Kraut RE, Greenfield PM, Gross EF.The impact of home computer use on children’s activ-ities and development. Future Child 2000;10:123-44.



69. Vandewater EA, Shim M, Caplovitz AG. Linking obesityand activity level with children’s television and videogame use. J Adolesc 2004;27:71-85.

70. Brasington R. Nintendinitis. N Engl J Med 1990;322:1473-4.

71. Ellis HC, Hunt RR. Fundamentals of Cognitive Psycho-logy. 5th ed. Madison, WI: Brown & Benchmark; 1993.

72. Anderson JR. The Architecture of Cognition. Cambridge,MA: Harvard University Press; 1983.

73. Glaser R. Educational thinking: the role of knowledge.Am Psychol 1984;39:93-104.

74. Fisher S. Identifying video game addiction in childrenand adolescents. Add Behav 1994;19:545-53.

75. Griffiths MD, Hunt N. Dependence on computer gamesby adolescents. Psychol Rep 1998;82:475-80.

76. Lieberman DA. Interactive video games for health pro-motion: effects on knowledge, self-efficacy, social sup-port, and health. In: Street RL, Gold WR, Manning T ed-itors. Health promotion and interactive technology:theoretical applications and future directions. Mahwah,NJ: Lawrence Erlbaum Associates; 1997.p.103-20.

77. Lieberman DA. Using interactive media in communi-cation campaigns for children and adolescents. In: RiceRE, Atkin CK editors. Public Communication Cam-paigns. 3rd ed. Thousand Oaks, CA: Sage Publications;2001.p.373-88.

78. Hoffman HG. Virtual reality therapy. Sci Am 2004;291:58-65.

79. Wiederhold BK, Wiederhold MD. Virtual Reality

Therapy for Anxiety Disorders: Advances in Evaluationand Treatment. Washington, DC: American Psycholo-gical Association; 2005.

80. O’Keefe BJ, Zehnder S. Understanding media devel-opment: a framework and case study. J Appl DevPsychol 2004;25:729-40.

81. Greenfield PM, Camaioni L, Ercolani P, Weiss L, LauberBA, Perucchini P. Cognitive socialization by computergames in two cultures: inductive discovery or masteryof an iconic code? J Appl Dev Psychol 1994;15:59-85.

82. Rosser JC Jr, Lynch PJ, Haskamp LA, Yalif A, Gentile DA,Giammaria L. Are Video Game Players Better atLaparoscopic Surgery? Medicine Meets Virtual RealityConference; 2004, January 15-17, Newport Beach, CA.

83. Green CS, Bavelier D. Video games modify visual at-tention. Nature 2003;423:534-7.

84. Greenfield PM, DeWinstanley P, Kilpatrick H, Kaye D.Action video games and informal education: effectson strategies for dividing visual attention. J Appl DevPsychol 1994;15:105-23.

85. Ballard ME, Wiest JR. Mortal Kombat: the effects of vi-olent videogame play on males’ hostility and cardio-vascular responding. J Appl Soc Psychol 1996;26:717-30.

86. Wood RTA, Griffiths MD, Chappell D, Davies MNO. Thestructural characteristics of video games: a psycho-struc-tural analysis. CyberPsychol Behav 2004;7:1-10.

87. Gibson JJ. The Ecological Approach to VisualPerception. Boston, MA: Houghton Mifflin; 1979.


a review of the literature - dr. douglas · violent video game effects on children and adolescents...

Documents