Professional Documents
Culture Documents
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Commentaries
Twenty-Five Years of Research
on Violence in Digital Games
and Aggression Revisited
A Reply to Elson and Ferguson (2013)
Brad J. Bushman1,2 and L. Rowell Huesmann3,4
1
Abstract. In this commentary, we first analyze Elson and Fergusons (2013) attempt to
offer a theory that would explain why exposure to family, community, school, and
media violence could be related to increased aggression, but not cause such aggression.
We conclude that the new theory they offer is not very new. It differs from
dominant social learning theories only in its claim that the relation between exposure
to violence and aggression is almost entirely due to people who are genetically or
biologically predisposed to be aggressive also exposing themselves to more violence.
We show this assertion is strongly contradicted by existing experimental and
longitudinal data. We also show that Elson and Fergusons so-called exhaustive
review of empirical data on the topic is seriously flawed; that their claim that effect
sizes are trivial is not supported by the math; and that their claim that scholars who
believe that violent video games cause aggression are an extreme group in a divided
field is contradicted by surveys that show the vast majority of researchers believe
violent video games increase aggression. We point out that their claim that scholars
who believe in media violence effects are having a moral panic has no theoretical or
empirical support, whereas the contrasting argument that researchers who produce
violent media themselves, or use it extensively, are biased by the force of cognitive
consistency and experience a reactance of regulatory panic does have support from
psychological theory.
Keywords: violent video games, aggression, violence
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
48
Commentaries
explain how observing violence in the home, school, community, or culture would increase the risk of violence but
observing it in the mass media or in video games would
not increase the risk?
Most writers who have denied that media violence has
any effects have avoided this issue and not discussed any theory. To their credit, Elson and Ferguson (2013) finally try to
address this issue by presenting their Catalyst theory,
which proposes that an aggression-prone personality develops mostly through biological and genetic dispositions that
are moderated by environmental aspects. The catalysts
they list are stressors like financial or relationship difficulties,
but also factors like having delinquent peers. In other words,
almost any environmental factor can moderate predispositions. Thus, there is nothing new in their theory, and it cannot
explain the obtained results of most media violence studies.
In fact, Elson and Ferguson have simply summarized the
dominant conclusions of a century of social, developmental,
and personality psychology with one different emphasis
biological and genetic dispositions. Even the term catalyst
is inferior to the term calibration that evolutionary psychologists have used to explain now environmental forces
modify genetic predispositions. Their summary is not a theory or a model, and it is certainly not falsifiable, except for
its predictions about the effects of exposure to violence that
already have been falsified. Nor are most of their assertions
different from the underlying assertions of the theories they
are trying to replace the General Aggression Model (GAM;
e.g., Anderson & Bushman, 2002), social learning theory
(e.g., Bandura, 1977), cognitive associationistic theory
(e.g., Berkowitz, 1990), social-information-processing theory (e.g., Dodge, 1986), script theory (e.g., Huesmann,
1988), evolutionary theory (e.g., Buss & Shackelford,
1997) except in the degree the emphasis is put on genetic
predispositions. Elson and Ferguson write that the aggression-prone personality develops mostly due to biological
and genetic dispositions. We know of no current aggression
researchers who do not believe there are important biological
predisposing factors (including genetically caused ones) that
make aggression more or less likely. The difference is that
most scholars, looking at decades of research, have concluded that these predisposing factors are not the lone or even
the main cause of aggression, but they strongly interact with
learning (both conditioning and observational learning) to
determine encoded social-cognitive structures that influence
behavior. Situational factors then precipitate aggressive
behaviors by stimulating aggressive emotions and thoughts
(Bandura, 1977; Berkowitz, 1962; Caspi, Taylor, Moffitt,
& Plomin, 2000; Dodge, Bates, Pettit, 1990; Eron, Walder,
& Lefkowitz, 1971; Huesmann & Kirwil, 2007).
Finally, let us consider the two key characteristics
that Elson and Ferguson claim for their theory. First, they
say that it considers individuals (to be) active modelers of their own behavior (p. 4). This is certainly true,
but not new, as every scholar who has studied self-perception theory or observational learning from Bandura
on has concluded. Children observe their own behavior
(as well as others) and draw conclusions from their
Commentaries
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
49
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
50
Commentaries
Empirical Evidence
In the empirical evidence section of their article, Elson and
Ferguson (2013) use an informal vote-counting approach to
evaluate the evidence, especially noting the number of
video game studies that found nonsignificant results. It is
well known that vote-counting procedures are the very last
resort for reviewing research evidence (see Bushman &
Wang, 2009). It is also well known that the simplest explanations for a null result are that the study was poorly conducted and therefore had a large error variance, had low
statistical power, or both. Null results are especially problematic when they are not predicted by theory. For example, Elson and Ferguson note that a few studies found
null effects for aggressive emotions, including a study by
Ballard, Hamby, Panee, and Nivens (2006). In this study
14 adolescents played a nonviolent NBA basketball game,
14 played a violent Resident Evil game, and 13 played a
violent Mortal Kombat game, in three separate sessions.
Although none of the effects were statistically significant,
they were far from trivial in size. The effect-size estimates
ranged from d = 0.48 to d = 0.72. The average effect was
d+ = 0.61 (95% confidence interval = 0.49 to 0.73,
p < .0001), which exceeds the conventional value for a
medium-sized effect (i.e., d = 0.50; see Cohen, 1988).
Yet, the statistical power was less than .50 (a coin toss)
for each comparison due to the small sample sizes.
The more important review that Elson and Ferguson
(2013) did not conduct was a meta-analysis to test their
new theory that is, testing whether early predispositions (early aggressive behavior or, even better, one of
the supposed genetic precursors of early aggressive behavior) cause media violence use independently of prior media
violence use. They should also test whether the effect of
early aggression on later media violence use is bigger than
the effect of early media violence use on later aggression.
Of course, finding such effects would not negate the argument that media violence causes aggression, but absent
such effects one cannot make a plausible argument that
genetic predispositions cause the relation between media
violence and aggression.
Commentaries
51
Number of effects
140
15
25
N
68,313
Unknowna
2,722
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Note. N = number of research participants. aFerguson and Kilburns (2009) meta-analysis included 15 video game studies, 7 television
studies, and 5 studies from movies alone or mixed media. The total sample for all studies is 12,436, but the number of participants in
video game studies is unknown. The Anderson et al. (2010) meta-analysis (136 research reports, 381 effects, 130,295 participants)
also examined the effects of violent video games on aggressive thoughts, angry feelings, physiological arousal, helping behavior, and
feelings of empathy, but only the aggressive behavior results are reported in this table.
of the laboratory experiment is that it allows causal inferences to be made; the primary weakness is that the setting
and measures are less realistic than in the real world. For
example, in the real world people aggress against others by
hitting, stabbing, or shooting them. Obviously, researchers
cannot allow participants to engage in such behaviors in
their laboratories. Thus, researchers use aggression measures such the amount of hot sauce given to target (usually
a confederate) who hates spicy food, or the intensity and
duration of unpleasant noise given to a target through headphones. Elson and Ferguson are critical of these laboratory
aggression measures: Naturally, children (and adults)
wishing to be aggressive do not chase after their targets
with jars of hot sauce or headphones with which to administer bursts of white noise (p. 6). Although this is generally
true, there are exceptions. For example, one woman was
found guilty of child abuse for punishing her child with
hot sauce (Hopper, 2011). Loud noise has also been used
to torture others, such as at Guantnamo Bay (Smith, 2008).
There are two types of realism: experimental and mundane (Aronson & Carlsmith, 1968). Experimental realism
refers to whether participants get so caught up in the procedures that they forget they are in an experiment. Mundane
realism refers to whether the setting physically resembles
the real world. Laboratory experiments are generally low
in mundane realism, but they can be high in experimental
realism. Elson and Ferguson (2013) focus on mundane realism. However, experimental realism is more important than
mundane realism in determining whether the results of a
study will generalize to the real world (Berkowitz &
Donnerstein, 1982).
Elson and Ferguson (2013) also criticize laboratory
measures for being unstandardized, especially the competitive reaction time task measure. However, effect-size estimates do not differ for studies that use a variation of the
competitive reaction time task compared to studies that
use other aggression measures (Bushman, Rothstein, &
Anderson, 2010).
Two different research teams using meta-analytic techniques have supported the validity of laboratory aggression
measures. One meta-analysis found impressive levels of
convergence across a wide range of laboratory aggression
measures (Carlson, Marcus-Newhall, & Miller, 1989).
Another meta-analysis found that real and laboratory
measures of aggression are influenced in similar ways by
situational variables (e.g., alcohol, provocation, anonymity)
2013 Hogrefe Publishing
and by individual difference variables (e.g., trait aggressiveness, participant sex, Type A personality) (Anderson &
Bushman, 1997).
Furthermore, several media violence field experiments
have shown similar effects, and they have more external
validity than laboratory experiments (for a summary of
some of them see Anderson et al. 2003). Additionally, randomized field experiments have shown that reducing exposure to media violence or teaching children that they should
not model it reduces aggression (e.g., Moller, Krahe,
Busching, & Krause, 2012).
Meta-Analytic Reviews
Elson and Ferguson (2013) describe the results from three
meta-analytic reviews (i.e., Anderson et al., 2010; Ferguson
& Kilburn, 2009; Sherry, 2001). Although they dismiss the
Anderson et al. (2010) meta-analysis on several grounds
(for a response to these criticisms see Bushman, Rothstein,
& Anderson, 2010 and Huesmann, 2010), it is by far the
most comprehensive meta-analysis on the effects of violent
video games on aggressive behavior (see Table 1).
Elson and Ferguson (2013) also claim that the metaanalytic effect-size estimates are inflated due to publication
bias. However, as Hannah Rothstein, an expert on publication bias in meta-analysis (e.g., Rothstein, Sutton, &
Borenstein, 2005), has repeatedly pointed out Ferguson
misused publication bias procedures in his meta-analytic
reviews (Bushman et al., 2010; Rothstein & Bushman,
2012). Furthermore, the Anderson et al. (2010) metaanalysis included extensive testing for possible publication
bias, and found none. We would argue if anything, the
meta-analytic effects are deflated rather than inflated. In
the experimental studies, there are at least two reasons
why. First, it is unethical for researchers to expose participants to age-inappropriate video games. For example, IRB
committees will not allow researchers to have 12-year-old
children play T games (for teens 13 and older), even
though many 12-year-olds play T- and even M-rated (for
players 17 and older) games outside the laboratory (e.g.,
Kutner & Olson, 2008). Theoretically, the effects should
be larger if children had played more graphic games.
Second, participants in laboratory experiments typically
play games for only 15-30 minutes, even though the
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
52
Commentaries
average player plays video games about 11 h a week (FactCheck.org, 2013). In many ways it is quite impressive that
playing a violent video game for just 1530 min, on a
single occasion can have significant and measurable effects
on aggressive behavior. For the cross-sectional and longitudinal studies, many of the effect-size estimates are likely
deflated because they are not corrected for attenuation
due to unreliable measures, and because of overcorrection
by partialling out effects of correlated variables (Prot &
Anderson, 2013).
Elson and Ferguson (2013) misrepresent the Anderson
et al. (2010) meta-analysis in other ways too. For example,
they state, Further, the authors included many of their own
unpublished studies and those of close colleagues but they
did not solicit unpublished studies from authors whose work differed in results from their own, thus setting
up selection bias problems. These claims are untrue, as
a careful reading of the methods section reveals. Anderson
et al. (2010) stated, we contacted the authors of reports
that appeared to have additional unpublished data that
could be used to compute effect-size estimates that met
the best practice criteria (emphasis added). In other words,
for any study that met the inclusion criteria but that was
missing statistical information needed to compute an effect
size for the meta-analysis, they contacted the author to get
the additional information. For example, if an article
reported only multiple regression results from a study that
measured both TV and video game violence exposure, they
contacted the author and asked for additional information
that would allow them to compute the relevant effect size
of violent video game exposure (e.g., raw correlations).
There were only a handful of such cases, and the decision
of whether to ask for the additional information was based
on whether or not the original report met the inclusion criteria, but did not include sufficient information to compute
the effect size. It was not based on whether the authors were
critics or not of the link between exposure to violent games
and aggression.
Furthermore, they did not ask anyone for unpublished
studies, did not receive any unsolicited unpublished studies,
and did not include any of their own unpublished studies, as
a review of the articles supplemental materials clearly
shows. Indeed, of the 88 included papers that reported
aggressive behavior, only two were unpublished one from
Japan, and a dissertation. Other unpublished doctoral dissertations were also uncovered in the original search, but
none of them met the inclusion criteria, or had already been
published.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Commentaries
Conclusion
We conclude that the new theory of Elson and Ferguson
(2013) is not new, and contains many truths that the theories they criticize also contain. It differs from these other
theories mainly in their claim that the relation between
exposure to violence and aggression is entirely due to people who are genetically or biologically predisposed to be
aggressive exposing themselves to more violence and people who are genetically or biologically predisposed to be
nonaggressive exposing themselves to less violence. This
assertion is strongly contradicted by existing experimental
and longitudinal data.
Elson and Ferguson (2013) aimed to give an exhaustive review of results obtained in laboratories and the field
(p. 4). Unfortunately, they used a vote-counting type
approach in which they counted studies that found positive
results, mixed results, and null results. This crude approach
gives the impression that the findings are inconsistent, even
though they are not. It focuses on the statistical significance
of effects rather than on the magnitude of effects. In metaanalytic reviews, no confidence interval even comes close
to including the value zero. For example, in meta-analysis
by Anderson et al. (2010) the average correlation between
exposure to violent video games and aggression was .19,
with a 95% confidence interval ranging from .18 to .20.
2013 Hogrefe Publishing
53
References
Abelson, R. P., Aronson, E., McGuire, W. J., Newcomb, T. M.,
Rosenberg, M. J., & Tannenbaum, R. H. (Eds.). (1968).
Theories of Cognitive Consistency: A sourcebook, Chicago,
IL: Rand-McNally.
Anderson, C. A., Berkowitz, L., Donnerstein, E., Huesmann,
R. L., Johnson, J., Linz, D., . . . Wartella, E. (2003). The
influence of media violence on youth. Psychological Science
in the Public Interest, 4, 81110.
Anderson, C. A., & Bushman, B. J. (1997). External validity of
trivial experiments: The case of laboratory aggression.
Review of General Psychology, 1, 1941.
Anderson, A. A., & Bushman, B. J. (2002). Human Aggression.
Annual Review of Psychology, 53, 2751. doi: 10.1146/
annurev.psych.53.100901.135231
Anderson, C. A., Sakamoto, A., Gentile, D. A., Ihori, N.,
Shibuya, A., Yukawa, S., . . . Kobayaski, K. (2008).
Longitudinal effects of violent video games on aggression
in Japan and the United States. Pediatrics, 122, 10671072.
Anderson, C. A., Shibuya, A., Ihori, N., Swing, E. L., Bushman,
B. J., Sakamoto, A., . . . Barlett, C. P. (2010). Violent video
game effects on aggression, empathy, and prosocial behavior
in Eastern and Western countries: A meta-analytic review.
Psychological Bulletin, 136, 151173.
Aronson, E., & Carlsmith, J. M. (1968). Experimentation in
social psychology. In G. Lindzey & E. Aronson (Eds.),
Handbook of social psychology (2nd ed., Vol. 2, pp. 179).
Reading, MA: Addison-Wesley.
Ballard, M. E., Hamby, R. H., Panee, C. D., & Nivens, E. E.
(2006). Repeated exposure to video game play results in
decreased blood pressure responding. Media Psychology, 8,
323341.
Bandura, A. (1977). Social learning theory. Englewood Cliffs,
NJ: Prentice Hall.
Berkowitz, L. (1962). Aggression: A social psychological
analysis. New York, NY: McGraw-Hill.
Berkowitz, L. (1990). On the formation and regulation of anger
and aggression: A cognitive-neo-associationistic analysis.
American Psychologist, 45, 494503.
Berkowitz, L., & Donnerstein, E. (1982). External validity
is more than skin deep. American Psychologist, 37,
245257.
Boxer, L. R., Huesmann, L. R., Dubow, E. F., Landau, S.,
Gvirsman, S., Shikaki, K., & Ginges, J. (2013). Exposure to
violence across the social ecosystem and the development of
aggression: A test of ecological theory in the IsraeliPalestinian conflict. Child Development, 84, 163177.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
54
Commentaries
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Commentaries
55
Brad J. Bushman
School of Communication
Ohio State University
3127 Derby Hall
154 North Oval Mall
Columbus, OH 43210-1339
USA
Tel. +1 614 688-8779
Fax +1 614 292-2055
E-mail bushman.20@osu.edu