Chronic Violent Video Game Exposure And Desensitization To ...
Journal of Experimental Social Psychology 42 (2006) 532–539
www.elsevier.com/locate/jesp
Chronic violent video game exposure and desensitization to violence:
Behavioral and event-related brain potential data
Bruce D. Bartholow a,¤, Brad J. Bushman b,c, Marc A. Sestir d
a Department of Psychological Sciences, University of Missouri-Columbia, USA
b Institute for Social Research, University of Michigan, USA
c Vrije Universiteit, Amsterdam, The Netherlands
d Department of Psychology, University of North Carolina at Chapel Hill, USA
Received 21 April 2005; revised 26 July 2005
Available online 7 October 2005
Abstract
Numerous studies have shown that exposure to media violence increases aggression, though the mechanisms of this eVect have
remained elusive. One theory posits that repeated exposure to media violence desensitizes viewers to real world violence, increasing
aggression by blunting aversive reactions to violence and removing normal inhibitions against aggression. Theoretically, violence desensi-
tization should be reXected in the amplitude of the P300 component of the event-related brain potential (ERP), which has been associated
with activation of the aversive motivational system. In the current study, violent images elicited reduced P300 amplitudes among violent,
as compared to nonviolent video game players. Additionally, this reduced brain response predicted increased aggressive behavior in a
later task. Moreover, these eVects held after controlling for individual diVerences in trait aggressiveness. These data are the Wrst to link
media violence exposure and aggressive behavior to brain processes hypothetically associated with desensitization.
© 2005 Elsevier Inc. All rights reserved.
Keywords: Aggression; Desensitization; Violent media; Event-related potential ERP; Habituation; P300; Brain; Video games
Introduction
ple, however, becoming desensitized to blood and gore can
have deleterious social consequences, such as reducing inhi-
Most people naturally have aversive reactions to the
bitions against behaving aggressively.
sight of blood and gore. Some people (e.g., soldiers, sur-
Hundreds of studies have shown that exposure to media
geons) must overcome these reactions in order to eVectively
violence increases aggression (see Anderson & Bushman,
perform their duties. This example illustrates the process of
2001; Anderson et al., 2003). Media violence is believed to
desensitization, deWned as diminished psychological or
increase aggression, at least in part, by desensitizing viewers
emotional responsiveness to a stimulus after repeated expo-
to the eVects of real violence (e.g., GriYths & Shuckford,
sure to it (see Wolpe, 1982). Desensitization can be adaptive
1989; Smith & Donnerstein, 1998). Media violence initially
because it enables people to ignore irrelevant information
produces fear, disgust, and other avoidance-related motiva-
and focus instead on relevant information. For most peo-
tional states (Cantor, 1998). Repeated exposure to
media violence, however, reduces its psychological impact
and eventually produces aggressive approach-related
We thank Bill Gehring and Johan Hoorn for their helpful comments
motivational states (Cline, Croft, & Courrier, 1973; Linz,
on an earlier version of this article. Portions of this work were presented at
Donnerstein, & Adams, 1989), theoretically leading to sta-
the 2005 annual meeting of the Society for Personality and Social
ble increases in aggression.
Psychology.
*
Extant research on media violence desensitization has
Corresponding author. Fax: +1 573 882 7710.
E-mail address: BartholowB@missouri.edu (B.D. Bartholow).
been limited in a number of respects. For example, although
0022-1031/$ - see front matter © 2005 Elsevier Inc. All rights reserevd.
doi:10.1016/j.jesp.2005.08.006
B.D. Bartholow et al. / Journal of Experimental Social Psychology 42 (2006) 532–539
533
desensitization is believed to have both cognitive and emo-
variable such as aggressive personality (Freedman, 2002).
tional components (see Funk, Bechtoldt-Baldacci, Pasold, &
In other words, aggressive individuals are simply drawn to
Baumgartner, 2004), most research has focused only on the
violent media. If this hypothesis is correct, controlling for
emotional component (see Smith & Donnerstein, 1998).
individual diVerences in aggressive disposition should elim-
Additionally, very few studies to date have examined how
inate or signiWcantly reduce the eVects of exposure to media
repeated exposure to media violence aVects brain processes
violence on relevant outcomes. Critics also argue that labo-
(but see Kronenberger et al., 2005), and no studies have
ratory experiments showing increased aggression following
examined potential links between physiological indices of
violent media exposure simply reXect a priming eVect, likely
desensitization and aggressive behavior. Strong support for
lasting only a few minutes, which does not carry over into
the desensitization account of media violence eVects will be
the “real world” (e.g., Fowles, 1999; Freedman, 1984). The
indicated only if processes associated with desensitization can
desensitization account is at odds with both of these alter-
be linked to increases in aggressive behavior.
native interpretations, predicting that media violence expo-
Theoretically, desensitization should be reXected in the
sure leads people to aggress (not that aggressive people seek
amplitude of the P300 component of the event-related
out violent media) and that repeated exposure has lasting
brain potential (ERP). ERPs are voltage deXections in the
deleterious consequences. The current study tests these
electroencephalogram (EEG) that reXect the engagement of
competing interpretations.
various information processing activities in the brain (see
In this study, violent and nonviolent video game players
Fabiani, Gratton, & Coles, 2000). The amplitude of the
completed a visual oddball task in which neutral, violent,
P300 component of the ERP, often associated with working
and negative nonviolent target images were presented in a
memory updating (Donchin & Coles, 1988), also has been
neutral image context while ERPs were recorded. Later,
shown to reXect the extent of evaluative categorization dur-
participants engaged in a competitive task during which
ing processing of aVective or emotionally relevant stimuli
they could blast another “participant” with loud noise; this
(e.g., Cacioppo, Crites, Berntson, & Coles, 1993; Ito,
task was used to measure aggression. We predicted that
Larsen, Smith, & Cacioppo, 1998b). Large P300 amplitudes
violent video game players would show deWcits in P300
are elicited over midline parietal scalp regions by stimuli
amplitude to violent images, but not to negative nonviolent
that are evaluatively inconsistent with a preceding context
images, relative to nonviolent game players. We further pre-
(Bartholow, Fabiani, Gratton, & Bettencourt, 2001;
dicted that violent game players would behave more
Cacioppo et al., 1993; Ito et al., 1998b). For example, infre-
aggressively in the competitive task, and that P300 ampli-
quent negative target images presented in a context of
tudes to violent (but not nonviolent) images would be
frequently presented neutral images elicit large P300s (Ito
inversely related to aggressive behavior. Finally, we pre-
et al., 1998b). It follows, then, that violent images presented
dicted that these eVects would remain when individual
in a context of neutral images should also elicit large P300s.
diVerences in aggressiveness were statistically controlled.
To the extent that an individual is desensitized to violence,
however, the P300 elicited by violent images should be
Method
reduced. Moreover, to the extent that a P300 reduction
reXects motivational processes associated with desensitiza-
Participants
tion to violence, the P300 reduction should be restricted to
evaluative categorization of violent images and not to nega-
Thirty-nine healthy, male undergraduates (mean
tive images more generally.
age D 19.5), all right-handed with normal or corrected-to-
Furthermore, if desensitization helps to explain the link
normal vision, volunteered in exchange for course credit.
between media violence and aggression, and if electrocorti-
Participants were recruited using an internet-based experi-
cal responses to violent stimuli reXect desensitization, then
ment sign-up procedure advertising a study on “the eVects
decreases in P300 amplitude to violent images should be
of picture viewing on reaction times.”
associated with increases in aggression. A number of stud-
ies have shown that the P300 elicited by negative informa-
Measures
tion reXects activation of the aversive/withdrawal
motivational system (e.g., Cacioppo, Crites, Gardner, &
Video game violence exposure
Berntson, 1994; Ito et al., 1998b). Given that aggression is
As in previous research (Anderson & Dill, 2000;
incompatible with withdrawal motivation (see Harmon-
Bartholow, Sestir, & Davis, in press), participants com-
Jones, 2003), and that desensitization theoretically weakens
pleted a questionnaire in which they listed their Wve favorite
the aversive motivation system pertaining to violence
video games and then rated each game, on scales anchored
(Cantor, 1998; Funk et al., 2004), there should be an inverse
at 1 and 7, in terms of how often they play the game and the
relationship between P300 amplitudes elicited by violent
violence of the game’s content and graphics. For each
stimuli and indices of aggressive behavior.
game, we multiplied the sum of the violent content and vio-
Critics of the link between media violence and increased
lent graphics ratings by the “how-often” rating. These Wve
aggression often claim that media violence exposure eVects
scores were averaged to form an overall index of video
are spurious, masking the eVect of some unmeasured third
game violence exposure ( D .82). For descriptive purposes,
534
B.D. Bartholow et al. / Journal of Experimental Social Psychology 42 (2006) 532–539
we refer to individuals with relatively high scores on this
Bernstein, Richardson, & Hammock, 1987; Giancola &
measure as violent video game players, and those who
Zeichner, 1995), and we have used it successfully in many of
received relatively low scores as nonviolent video game
our previous studies (e.g., Bartholow & Anderson, 2002;
players. However, participants were not dichotomized into
Bartholow et al., in press; Bushman, 1995).
these categories for analyses.
Electrophysiological recording
Trait aggressiveness
Individual diVerences in aggressiveness were assessed
The EEG was recorded from 28 electrodes Wxed in an
using the Irritability Scale (30 items; D .87; Caprara et al.,
electrode cap (Electro-cap International, Eaton, OH) at
1985) and the Aggression Questionnaire (29 items; D .88;
standard scalp locations. All EEG electrodes were refer-
Buss & Perry, 1992). The Irritability Scale contains items
enced online to the right mastoid (an average mastoid refer-
such as, “I easily Xy oV the handle with those who don’t lis-
ence was derived oZine). EEG was continuously recorded
ten or understand.” Responses were made on scales
and stimulus-locked ERP epochs of 1400 ms were derived
anchored at 1 (This doesn’t characterize me at all) and 5
oZine (referenced to 200 ms pre-stimulus baseline). EEG
(This characterizes me very well). The Aggression Question-
was ampliWed with a Neuroscan Synamps ampliWer and
naire (AQ) contains 4 subscales, labeled Physical Aggres-
Wltered on-line at 0.05–30 Hz at a sampling rate of 250 Hz.
sion (9 items; D .78; e.g., “If somebody hits me, I hit
Impedance was kept below 5 k . Ocular artifacts (blinks)
back”); Verbal Aggression (5 items; D .85; e.g., “I can’t
were removed from the EEG using a regression-based pro-
help getting into arguments when people disagree with
cedure (Semlitsch, Anderer, Schuster, & Presslich, 1986).
me”); Anger (7 items; D .84; e.g., “Some of my friends
Trials containing voltage deXections of §75 microvolts
think I’m a hot-head”); and Hostility (8 items; D .77; e.g.,
( V) after ocular artifact removal were rejected prior to
“At times I feel I have gotten a raw deal out of life”).
averaging. OV-line averages were derived according to par-
Responses were made on scales anchored at 1 (Extremely
ticipant, electrode, and stimulus conditions, and low-pass
uncharacteristic of me) and 6 (Extremely characteristic of
Wltered at 12 Hz (12 dB roll-oV).1 Initial inspection of the
me). High scores on these measures are considered reliable
waveforms conWrmed that the P300 was largest at the mid-
and valid self-report indices of trait aggressiveness (e.g.,
line parietal (Pz) electrode site. Therefore, for each partici-
Bushman & Wells, 1998; Caprara et al., 1985; Harris, 1996).
pant, the P300 was identiWed by selecting the largest
positive peak between 300 and 800 ms post-stimulus at Pz.2
Aggressive behavior
P300 amplitude was computed by averaging over the
Participants were led to believe that they were competing
300 ms around that peak (i.e., 150 ms before and after) in
against another participant (the ostensible partner) in a
each condition.
reaction time task to see who could press a button faster
following an auditory tone. They were further told that the
Picture viewing task
slower person on a given trial would receive a blast of noise
through a pair of headphones, the intensity and duration of
All images used in this study were taken from the Inter-
which were set by the other person prior to that trial. At the
national AVective Picture System (IAPS; Lang, Bradley, &
end of each trial, the participant saw a message stating
Cuthbert, 2001). Lang et al. (2001) had participants rate the
either “YOU WON!” or “YOU LOST!” and received a
valence of each image using scales ranging from 1 (com-
noise blast on losing trials. Noise intensities ranged from 60
pletely unhappy) to 9 (completely happy) and their arousal
decibels (level 1) to 105 decibels (level 10). A nonaggressive
to each image using scales ranging from 1 (completely calm)
no-noise setting (level 0) was also oVered. In addition to set-
to 9 (completely aroused). Valence and arousal ratings for
ting the intensity, the winner also determined the duration
the images used in this study are presented in Table 1.
of the loser’s suVering using a noise duration setting rang-
Examples of the neutral images included pictures of a man
ing from 0.25 s (level 1) to 2.5 s (level 10). In eVect, each par-
on a bicycle; a man opening a backpack on a street corner;
ticipant controlled a weapon that could be used to blast the
a towel laying on a table; and a mushroom. Example vio-
other person with loud noise on winning trials. Aggression
lent images included a man holding a gun to another man’s
was deWned as the average intensity and duration of noise
head on a subway; a man holding a gun in another man’s
(standardized and summed) that the participant set for the
ostensible partner over the 25 trials of the task.
Actually, there was no partner—the computer controlled
1 Applying a low-pass Wlter to the relatively low frequency activity em-
trial wins and losses as well as noise intensities and dura-
bodied by the P300 is important in order to reduce the inXuence of higher
tions ostensibly set by the partner. The participant lost the
frequencies, particularly when peak amplitude measures are used (e.g.,
Wrst trial, and half of the remaining 24 trials in a random
Fabiani, Gratton, Karis, & Donchin, 1987).
2
pattern, with intensity levels varying randomly between lev-
EEG data were recorded from the additional scalp locations for ex-
ploratory analyses not relevant to the psychological hypotheses being test-
els 2 and 9. Duration levels varied randomly between 0.5
ed in this study. Therefore, we do not present those analyses here. It is
and 2.0 s. Previous research has established the validity and
worth noting, however, that the eVects of video game violence exposure
reliability of this widely used aggression measure (e.g.,
and image type were similar at all midline electrodes.
B.D. Bartholow et al. / Journal of Experimental Social Psychology 42 (2006) 532–539
535
Table 1
position 3, 4, or 5; thus, participants viewed a total of 480
Average valence and arousal ratings for the IAPS images used in this
images. Images were displayed for 1 s each, separated by a
study
1 s inter-stimulus interval. Participants were instructed to
Image type
Valence ratings
Arousal ratings
think about their reactions to each image. Trials were sepa-
Neutral
5.13 (0.66)
2.70 (0.66)
rated by a 2.5 s inter-trial interval, during which the word
Violent
2.39 (0.17)
6.75 (0.22)
“ready” was displayed on the monitor. A short break
Negative nonviolent
2.26 (0.43)
6.09 (0.39)
(approximately 2 min) was inserted between blocks.
Note. IAPS, International AVective Picture System. Numbers in parenthe-
ses are standard deviations. The identiWcation numbers (from the IAPS
Procedure
manual; Lang et al., 2001) for the neutral images were 5875, 7493, 2749,
5410, 2840, 2850, 2870, 2880, 8465, 9210, 5500, 7000, 7002, 7009, 7010,
7025, 7030, 7035, 7040, 7080, 7090, 7140, 7217, 7224, and 7050. The violent
After obtaining informed consent, the researcher applied
images were numbered 3500, 3530, 6313, 6350, and 6540, and the nonvio-
all electrodes and explained the experimental tasks. Partici-
lent negative images were numbered 3170, 6415, 9570, 9800, and 9910.
pants were told that the purpose of the study was to exam-
ine how viewing diVerent kinds of images would aVect
mouth on a bus; and a man holding a knife to a woman’s
response speed in an unrelated, competitive task. The
throat. Example negative nonviolent images included a
researcher then left the room for several minutes, allegedly
baby with a large tumor on her face; a neo-Nazi skinhead
to explain the tasks to the other participant. Upon his
in front of a swastika Xag; and a decaying dog corpse.
return, the participant started the picture viewing task.
As can be seen in Table 1, valence ratings for violent and
After completing the picture viewing task, the participant
negative nonviolent images were very similar (d D 0.43).
waited 3 min while the experimenter allegedly set up the
However, the Lang et al. sample rated the violent images as
second task on the other participant’s computer. The exper-
more arousing than the negative nonviolent images
imenter then returned again and read the instructions for
(d D 2.16). Although the statistical signiWcance of the mean
the competitive task, after which the participant completed
diVerences presented in Table 1 cannot be calculated with-
the task. Finally, participants were interviewed for suspi-
out Lang et al.’s data set, power analyses based on the eVect
cion, debriefed, and dismissed.
sizes just noted showed that a sample size of approximately
200 would be required to detect a signiWcant diVerence
Results
between the valence ratings 80 percent of the time with
D .05 (two-tailed). In contrast, a signiWcant diVerence
Data from 5 participants were discarded (2 had a high
between the arousal ratings would be detected 80% of the
proportion of EEG artifacts and 3 were suspicious that
time using a sample size of only 10 (see Cohen, 1988).
they were not competing against anyone during the com-
Images were presented in 2 blocks of 48 trials each. In
petitive task). Thus, all analyses were based on data from 34
each block, target images were either neutral and negative
participants.
or neutral and violent (half of the target images in each
Simple bivariate associations among the main study
block were neutral). Block order was varied randomly
variables are given in Table 2. Consistent with previous
across participants. Each trial consisted of 4 context images
research (Anderson & Dill, 2000; Bartholow et al., in press),
(always neutral), and 1 target image presented randomly in
video game violence exposure was strongly associated with
Table 2
Bivariate associations among the main study variables
1
2
3
4
5
6
7
8
9
1. VVE
—
2. Aggression
.57¤¤¤
—
3. Violent P300
¡.64¤¤¤
¡.48¤¤
—
4. Negative P300
.00
¡.20
.34¤
—
5. Neutral P300
.17
.04
.05
.43¤¤
—
6. Irritability
¡.14
.269
¡.34¤
¡.17
¡.07
—
7. AQ-H
.05
.24
¡.31¤
¡.15
¡.18
.56¤¤¤
—
8. AQ-A
¡.23
¡.17
¡.14
¡.24
¡.09
.70¤¤¤
.58¤¤¤
—
9. AQ-P
.04
.12
¡.38¤
¡.22
.09
.65¤¤¤
.58¤¤¤
.52¤¤¤
—
10. AQ-V
.21
.10
¡.11
¡.18
.03
.32¤
.23
.33¤
.309
Note. VVE, video game violence exposure; Aggression, composite aggression score from competitive task; Irritability, Irritability Scale scores; AQ-H,
Aggression Questionnaire hostility subscale; AQ-A, Aggression Questionnaire anger subscale; AQ-P, Aggression Questionnaire physical aggression sub-
scale; AQ-V, Aggression Questionnaire verbal aggression subscale.
¤ p < .05.
¤¤ p < .01.
¤¤¤ p < .001.
9 p < .10.
536
B.D. Bartholow et al. / Journal of Experimental Social Psychology 42 (2006) 532–539
increased aggression during the competitive reaction time
A
task. More pertinent to our current hypotheses, violent
Neutral
Violent
Nonviolent game players
video game exposure was inversely associated with the
Violent game players
amplitude of the P300 elicited by violent images during the
-10.0
picture viewing task, which was inversely associated with
-5.0
aggressive behavior. However, our hypotheses are more
)
V
appropriately tested by the analyses presented next in
µ
0.0
(
e
which the eVects of violent video game exposure on all
d
5.0
image types is compared and in which individual diVerences
litu
p
m 10.0
in aggressiveness are controlled.
A
Our Wrst major prediction was that P300 amplitudes to
15.0
violent images would be smaller for violent game players
20.0
than for nonviolent game players, and that this eVect would
-200
50
300
550
800
1050
not generalize to other negative images. A general linear
Time (ms)
Picture
model showed that the predicted interaction between the 3-
onset
level image type variable (neutral, negative, violent) and
B
video game violence exposure scores (continuous variable)
Neutral Negative
was signiWcant, F (2, 64)
Nonviolent game players
D 12.03, p < .0001 (Greenhouse-
Geisser adjusted; D .99). Separate regression analyses
Violent game players
within each image type showed that, as expected, P300
-10.0
-
amplitudes to violent images decreased as a function of
-5.0
-
increased violent video game exposure, t (32) D ¡4.66,
V)
0.0
p < .0001, b D ¡.20 (see Fig. 1A). However, P300 amplitudes
µ
e (
to neutral images were not aVected by violent video game
d
u
5.0
exposure, t (32) D 1.00, p D .33, b D .03, nor were P300 ampli-
lit
p
m 10.0
tudes to negative nonviolent images, t (32) D 0.02, p D .98,
A
b D .00 (see Fig. 1B). This latter Wnding rules out the possi-
15.0
bility that evaluative categorization of all negative stimuli is
20.0
blunted in violent game players, relative to nonviolent
-2
- 00
50
300
550
800
1050
game players, supporting the idea that violent game players
Time (ms)
Picture
are speciWcally desensitized to violence.
onset
The analysis just presented examined brain responses to
Fig. 1. Event-related potential waveforms measured at the Pz (midline
violent, negative, and neutral images in an absolute sense.
parietal) electrode site as a function of picture type and levels of prior
Another, perhaps more focused way of conceptualizing
exposure to video game violence. For purposes of presentation, partici-
desensitization is in terms of the degree to which responses
pants were categorized as either violent (gray lines) or nonviolent (black
to violent images are attenuated relative to responses to
lines) video game players using a median split (though the continuous
measure was used in all analyses). (A) The P300 elicited by neutral images
equally negative nonviolent images. To test this possibility,
(solid lines) and violent images (dashed lines). (B) The P300 elicited by
we computed a new variable for each participant represent-
neutral images (solid lines) and negative nonviolent images (dashed lines).
ing the diVerence between their P300 responses to negative
images and violent images and examined the association
between this diVerence score and video game violence expo-
the fact that Irritability scores also accounted for signiW-
sure. This association was positive, t (32) D 3.73, p < .001,
cant variance in P300 amplitudes to violent images,
b D .20, D .55, indicating that increased exposure to video
t (27) D ¡2.03, p < .05, b D ¡2.00, D ¡.37. None of the AQ
game violence was associated with a larger diVerence
subscales accounted for signiWcant variance in P300 ampli-
between the P300 response to negative images and violent
tude to violent stimuli ( s ranged from ¡.25 to .27; ps > .05).
images, which can be seen by comparing panels A and B of
Inspection of Fig. 1A suggests that in addition to the
Fig. 1.
amplitude diVerence reported previously, the peak latency
It is possible that the eVect of video game violence expo-
of the P300 elicited by violent images also appears to be
sure on P300 amplitudes to violent stimuli is due to individ-
aVected by video game violence exposure. A general linear
ual diVerences in aggressiveness. To examine this
model conWrmed that the P300 to violent images peaked
alternative explanation of our eVects, we tested the associa-
later among violent video game players than among nonvi-
tion between P300 amplitudes elicited by violent images
olent video game players, t (32) D 2.19, p < .05, b D 8.18,
and violent video game exposure while simultaneously
D .36. This eVect was not weakened by the inclusion of
covarying Irritability scores and scores on each AQ sub-
Irritability Scale scores and AQ subscale scores,
scale. The eVect of video game violence exposure on P300
t (27) D 2.68, p < .05, b D 9.75, D .45; the AQ-anger subscale
amplitudes to violent images remained signiWcant in this
was the only other signiWcant predictor in this model,
analysis, t (27) D ¡5.55, p < .0001, b D ¡.22, D ¡.71, despite
t (27) D 2.66, p < .05, b D 7.20,
D .56. A separate general
B.D. Bartholow et al. / Journal of Experimental Social Psychology 42 (2006) 532–539
537
linear model indicated that, in contrast, P300 latency elic-
mentioned previously and aggressive behavior during the
ited by nonviolent negative images was not aVected by vio-
competitive task. This association was positive, t (32) D 2.12,
lent video game exposure, t (32) D ¡.03, p > .90, b D ¡.02,
p < .05, b D .29, D .35, indicating that a larger diVerence in
D ¡.01 (see Fig. 1B).
the P300 elicited by negative images relative to violent
Our second major prediction was that the P300 elicited
images was associated with increased aggression.
by violent images would signiWcantly predict aggressive
Our contention is that exposure to violent video games
behavior in the competitive reaction time task. A regression
leads to desensitization, as evidenced by reduced cortical
analysis showed the predicted inverse relationship: smaller
responses to violent images and increased aggressive behav-
P300 amplitudes to violent images (seen primarily among
ior. However, because the video game violence exposure
violent video game players) during the picture viewing task
measure used here combines the frequency of game play
were signiWcantly associated with higher levels of aggres-
with the violence of game content, it is possible that the
sion, t (32) D ¡3.11, p < .01, b D ¡.46, r D ¡.48 (see Fig. 2A).
eVects just reported are due to a more general phenomenon
In contrast, P300 amplitudes elicited by negative nonvio-
associated with frequently playing any video games, not
lent images were not signiWcantly associated with aggres-
violent games per se. To examine this possibility, we broke
sion, t (32) D ¡1.16, p D .25, b D ¡.19, r D ¡.20. Finally, a
the video game violence exposure score into constituent
separate regression analysis showed that the more time par-
average game content and gaming frequency scores for
ticipants spent playing violent video games, the more they
each participant, and examined their respective inXuences
aggressed against their ostensible partner, t (32) D 3.88,
on the main dependent variables. Game content and overall
p < .001, b D .17, r D .57 (see Fig. 2B). This relationship also
frequency of game play were only modestly correlated,
held after controlling for individual diVerences in aggres-
r D .30, p D .06, suggesting considerable independence in
siveness, t (32) D 3.68, p < .001, b D .16, D .57.
these dimensions. A regression equation in which both the
We also examined the association between the P300
content and frequency scores were used to simultaneously
diVerence score variable (negative images-violent images)
predict P300 amplitudes elicited by violent images showed
that violent content was a strong predictor of reduced P300
amplitude, t (31) D ¡3.90, p < .001, b D ¡1.73, D ¡.54, but
gaming frequency was much less strongly associated,
4
A
r = -0.48
t (31) D ¡1.90, p D .06, b D ¡0.93, D ¡.27. A similar regres-
3
)
d
sion equation showed that whereas violent content was a
e
z 2
unique predictor of aggressive behavior during the compet-
rdi
1
itive task, t (31) D 2.66, p < .05, b D 1.27, D .42, frequency of
nda
t
a
play was not, t (31) D 1.67, p > .10, b D 0.86, D .25. Thus,
0
although there is some evidence that the frequency of video
i
on (s -1
s
game play has an independent e
s
Vect on the cognitive pro-
r
e -2
cessing of violent images, these analyses largely support the
g
g
A
conclusion that exposure to violent games speci
-3
Wcally (and
not just any games) is responsible for our reported eVects.
-4
0
5
10
1
15
20
Discussion
P3
P 00 to
to vi
v olent image
s (
s μV)
V
B
Previous research has shown that playing violent video
5
r = 0.57
= 0.5
games increases aggressive behavior and decreases helping
4
d)
ze
behavior (see Anderson, 2004; Anderson & Bushman,
3
di
2001). One possible explanation for these e
ar 2
Vects is that peo-
nd
ple become desensitized to violence after prolonged expo-
1
sure to it, leading to reduction of normal inhibitions against
0
aggression and making individuals less responsive to the
on (sta -1
si
pain and su
-2
Vering experienced by victims of violence
es
(Carnagey, Bushman, & Anderson, 2005; Funk et al., 2004).
gr -3
Ag
The present research advances this desensitization account
-4
by showing that repeated exposure to violent video games
-5
0
10
1
20
2
30
3
40
4
50
60
6
is reXected in the brain as blunted evaluative categorization
Video g
o am
a e v
e i
v o
i lenc
l
e ex
e
posure
of violent stimuli. Compared to nonviolent video game
players, violent video game players showed reduced P300
Fig. 2. Scatterplots depicting the magnitude of association between
amplitude and increased P300 latency to violent images but
aggressive behavior (i.e., intensity and duration of noise blasts given to the
not to other, equally negative nonviolent images. The
ostensible partner during the competitive reaction time task) and the
amplitude of the P300 elicited by violent images during the picture view-
latency of the P300 component generally is associated
ing task (A), and scores on the video game violence exposure measure (B).
with stimulus evaluation or categorization time (see
538
B.D. Bartholow et al. / Journal of Experimental Social Psychology 42 (2006) 532–539
Fabiani et al., 2000). Thus, the increased latency among vio-
ences in self-reported aggressiveness, it would be of interest
lent video game players indicates that it took these individ-
in future research to examine potential links between P300
uals longer to categorize the violent images. The fact that
responses to violent images, aggressive behavior, and anti-
the ERP Wndings held even after controlling for individual
social personality disorder.
diVerences in aggressiveness is inconsistent with some alter-
As noted previously, the P300 is known to be associated
native accounts suggesting that media violence eVects are
with working memory updating (Donchin & Coles, 1988),
spurious (e.g., Freedman, 2002).
considered by some to be a key component of so-called
Moreover, P300 amplitudes elicited by violent images
executive cognitive function (Miyake, Friedman, Emerson,
during the picture-viewing task were inversely associated
Witzki, & Howerter, 2000). Recent brain imaging data indi-
with aggression during the competitive reaction time task.
cate that adolescents high in media violence exposure show
This Wnding is consistent with work showing that aggres-
abnormal frontal lobe function during the performance of
sive individuals tend to show deWcits in P300 amplitude
executive tasks (Mathews et al., 2005), a Wnding consistent
during simple stimulus discrimination tasks (Harmon-
with other work showing executive dysfunction among vio-
Jones, Barratt, & Wigg, 1997; Mathias & Stanford, 1999).
lent video game players (Kronenberger et al., 2005). As in
However, this Wnding is inconsistent with the notion,
the current study, these eVects were not attributable to
advanced by some, that the eVects of video game violence
diVerences in trait aggression in either of these other
exposure on aggression are short-lived and have no long-
reports, although Mathews et al. (2005) reported that the
term consequences for gamers (e.g., Freedman, 1984). More
pattern of brain activation seen in their violent media
generally, these Wndings are the Wrst to link violent video
exposed participants resembled that of a separate group of
game exposure to a reduction in brain activity known to
participants diagnosed with disruptive behavioral disorder.
reXect activation of the aversive motivational system (see
The current research extends this recent work by examining
Ito et al., 1998b), and to link this brain activity to aggressive
a diVerent neural index of executive dysfunction associated
behavior. Nevertheless, it will be important to replicate
with processing violence, and by showing links between this
these Wndings using a longitudinal design in which partici-
brain activity and aggressive behavior measured in the lab.
pants can be randomly assigned to violent and nonviolent
One important limitation of the current Wndings
media exposure conditions.
deserves mention. Although our hypotheses (and indeed,
As discussed previously, although the violent and nega-
our Wndings) suggest a potential mediational role for P300
tive images used here were rated as equally unpleasant by
amplitude in the link between video game violence expo-
participants in the norming sample, the violent images were
sure and aggressive behavior, speciWc tests for mediation
more arousing on average than the negative nonviolent
did not support this idea. Thus, although these three vari-
images (see Table 1). Other researchers have argued that
ables are clearly signiWcantly associated, the brain response
self-reported arousal ratings of IAPS images reXect the
to violent images did not account for the eVect of violence
degree of activation of the aversive motivational system
exposure on aggression in this study. It will be important in
(Ito, Cacioppo, & Lang, 1998a). The fact that violent video
future research to identify potential mediators of this eVect.
game players showed reduced cortical responses to violent
In summary, this study is the Wrst to link video game vio-
images, but not to other negative images, is consistent with
lence exposure and aggressive behavior to brain processes
prior research and theory indicating that desensitization
hypothetically reXecting desensitization in the aversive
limits emotional arousal to violent stimuli (see Smith &
motivational system. These Wndings, along with other
Donnerstein, 1998), and suggests that this decreased
recent research (Kronenberger et al., 2005; Mathews et al.,
arousal signiWes a relaxation of avoidant motivational pro-
2005), suggest that chronic exposure to violent video games
cesses speciWcally associated with violence.
speciWcally—and not just frequent playing of any video
These Wndings pertaining to arousal could have impor-
games—has lasting deleterious eVects on brain function
tant implications for linking violence desensitization to
and behavior.
aggressive behavioral disorders. Antisocial personality dis-
order is typiWed by low arousal in the aversive motivational
system (e.g., Fowles, 1988) and is a signiWcant risk factor for
References
aggressive and violent behavior (e.g., Langbehn, Cadoret,
Yates, Troughton, & Stewart, 1998). Recently, Herpertz
Anderson, C. A. (2004). An update on the eVects of violent video games.
et al. (2005) reported that adolescent boys with conduct dis-
Journal of Adolescence, 27, 113–122.
Anderson, C. A., & Bushman, B. J. (2001). EVects of violent video games
order show abnormally low autonomic responses to valen-
on aggressive behavior, aggressive cognition, aggressive aVect, physio-
ced images like those used in the current study (although
logical arousal, and prosocial behavior: A meta-analytic review of the
those researchers combined responses to violent and other
scientiWc literature. Psychological Science, 12, 353–359.
negative images). Some evidence suggests that symptoms
Anderson, C. A., Berkowitz, L., Donnerstein, E., Huesmann, R. L., John-
associated with antisocial personality disorder (e.g., delin-
son, J., Linz, D., et al. (2003). The inXuence of media violence on youth.
Psychological Science in the Public Interest, 4, 81–110.
quency) are positively correlated with exposure to video
Anderson, C. A., & Dill, K. E. (2000). Video games and aggressive
game violence (Anderson & Dill, 2000). Thus, although the
thoughts, feelings, and behavior in the laboratory and in life. Journal of
current Wndings were not dependent upon individual diVer-
Personality and Social Psychology, 78, 772–790.
B.D. Bartholow et al. / Journal of Experimental Social Psychology 42 (2006) 532–539
539
Bartholow, B. D., & Anderson, C. A. (2002). EVects of violent video games
Funk, J. B., Bechtoldt-Baldacci, H., Pasold, T., & Baumgartner, J. (2004).
on aggressive behavior: Potential sex diVerences. Journal of Experimen-
Violence exposure in real-life, video games, television, movies, and the
tal Social Psychology, 38, 283–290.
internet: Is there desensitization? Journal of Adolescence, 27, 23–39.
Bartholow, B. D., Fabiani, M., Gratton, G., & Bettencourt, B. A. (2001). A
Giancola, P. R., & Zeichner, A. (1995). Construct validity of a competitive
psychophysiological analysis of cognitive processing of and aVective
reaction-time aggression paradigm. Aggressive Behavior, 21, 199–204.
responses to social expectancy violations. Psychological Science, 12,
GriYths, M. D., & Shuckford, G. L. J. (1989). Desensitization to television
197–204.
violence: A new model. New Ideas in Psychology, 7, 85–89.
Bartholow, B. D., Sestir, M. A., & Davis, E. B. (in press). Correlates and
Herpertz, S. C., Mueller, B., Mutaz, Q., Lichterfeld, C., Konrad, K., & Her-
consequences of exposure to video game violence: Hostile personality,
pertz-Dahlmann, B. (2005). Response to emotional stimuli in boys with
empathy, and aggressive behavior. Personality and Social Psychology
conduct disorder. American Journal of Psychiatry, 162, 1100–1107.
Bulletin.
Harmon-Jones, E. (2003). Clarifying the emotive functions of asymmetri-
Bernstein, S., Richardson, D., & Hammock, G. (1987). Convergent and dis-
cal frontal cortical activity. Psychophysiology, 40, 838–848.
criminant validity of the Taylor and Buss measures of physical aggres-
Harmon-Jones, E., Barratt, E. S., & Wigg, C. (1997). Impulsiveness, aggres-
sion. Aggressive Behavior, 13, 15–24.
sion, reading, and the P300 of the event-related potential. Personality
Bushman, B. J. (1995). Moderating role of trait aggressiveness in the eVects
and Individual DiVerences, 22, 439–445.
of violent media on aggression. Journal of Personality and Social Psy-
Harris, M. B. (1996). Aggressive experiences and aggressiveness: Relation-
chology, 69, 950–960.
ship to ethnicity, gender, and age. Journal of Applied Social Psychology,
Bushman, B. J., & Wells, G. L. (1998). Trait aggressiveness and hockey
26, 843–870.
penalties: Predicting hot tempers on the ice. Journal of Applied Psychol-
Ito, T. A., Cacioppo, J. T., & Lang, P. J. (1998a). Eliciting aVect using the
ogy, 83, 969–974.
International AVective Picture System: Bivariate evaluation and
Buss, A. H., & Perry, M. P. (1992). The aggression questionnaire. Journal of
ambivalence. Personality and Social Psychology Bulletin, 24, 855–879.
Personality and Social Psychology, 63, 452–459.
Ito, T. A., Larsen, J. T., Smith, N. K., & Cacioppo, J. T. (1998b). Negative
Cacioppo, J. T., Crites, S. L., Jr., Berntson, G. G., & Coles, M. G. H. (1993).
information weighs more heavily on the brain: The negativity bias in
If attitudes aVect how stimuli are processed, should they not aVect the
evaluative categorizations. Journal of Personality and Social Psychol-
event-related brain potential? Psychological Science, 4, 108–112.
ogy, 75, 887–900.
Cacioppo, J. T., Crites, S. L., Jr., Gardner, W. L., & Berntson, G. G. (1994).
Kronenberger, W. G., Mathews, V. P., Dunn, D. W., Wang, Y., Wood, E.
Bioelectrical echoes from evaluative categorizations: I.A late positive
A., Giauque, A. L., et al. (2005). Media violence exposure and executive
brain potential that varies as a function of trait negativity and extrem-
functioning in aggressive and control adolescents. Journal of Clinical
ity. Journal of Personality and Social Psychology, 67, 115–125.
Psychology, 61, 725–737.
Cantor, J. (1998). “Mommy, I’m scared: How TV and movies frighten chil-
Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (2001). International aVective
dren and what we can do to protect them. San Diego, CA: Harvest/Har-
picture system (IAPS): Instruction manual and aVective ratings. Tech-
court.
nical Report A-5, The Center for Research in Psychophysiology, Uni-
Caprara, G. V., Cinanni, V., D’Imperio, G., Passerini, S., Renzi, P., & Trav-
versity of Florida.
aglia, G. (1985). Indicators of impulsive aggression: Present status of
Langbehn, D. R., Cadoret, R. J., Yates, W. R., Troughton, E. P., & Stewart,
research on irritability and emotional susceptibility scales. Personality
M. A. (1998). Distinct contributions of conduct and oppositional deW-
and Individual DiVerences, 6, 665–674.
ant symptoms to adult antisocial behavior. Archives of General Psychi-
Carnagey, N. L., Bushman, B. J., & Anderson, C. A. (2005). The eVect of
atry, 55, 821–829.
video game violence on physiological desensitization and helping. Manu-
Linz, D., Donnerstein, E., & Adams, S. M. (1989). Physiological desensiti-
script submitted for publication.
zation and judgments about female victims of violence. Human Com-
Cline, V. B., Croft, R. G., & Courrier, S. (1973). Desensitization of children
munication Research, 15, 509–522.
to television violence. Journal of Personality and Social Psychology, 27,
Mathews, V. P., Kronenberger, W. G., Wang, Y., Lurito, J. T., Lowe, M. J.,
360–365.
& Dunn, D. W. (2005). Media violence exposure and frontal lobe acti-
Cohen, J. (1988). Statistical power analyses for the behavioral sciences (2nd
vation measured by functional magnetic resonance imaging in aggres-
ed.). Hillsdale, NJ: Erlbaum.
sive and nonaggressive adolescents. Journal of Computer Assisted
Donchin, E., & Coles, M. G. H. (1988). Is the P300 component a manifesta-
Tomography, 29, 287–292.
tion of context updating? Behavioral and Brain Sciences, 11, 357–427.
Mathias, C. W., & Stanford, M. S. (1999). P300 under standard and sur-
Fabiani, M., Gratton, G., & Coles, M. G. H. (2000). Event-related brain
prise conditions in self-reported impulsive aggression. Progress in
potentials. In J. T. Cacioppo, L. Tassinary, & G. G. Berntson (Eds.),
Neuro-Psychopharmacology and Biological Psychiatry, 23, 1037–1051.
Handbook of psychophysiology (pp. 53–84). Cambridge, UK: Cam-
Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., & Howerter,
bridge University Press.
A. (2000). The unity and diversity of executive functions and their con-
Fabiani, M., Gratton, G., Karis, D., & Donchin, E. (1987). The deWnition,
tributions to complex “frontal lobe” tasks: A latent variable analysis.
identiWcation, and reliability of measurement of the P300 component
Cognitive Psychology, 41, 49–100.
of the event-related brain potential. In P. K. Ackles, J. R. Jennings, &
Semlitsch, H. V., Anderer, P., Schuster, P., & Presslich, O. (1986). A solu-
M. G. H. Coles (Eds.), Advances in psychophysiology (Vol. 1, pp. 1–78).
tion for reliable and valid reduction of ocular artifacts, applied to the
Greenwich, CT: JAI.
P300 ERP. Psychophysiology, 23, 695–703.
Fowles, D. C. (1988). Psychophysiology and psychopathology: A motiva-
Smith, S. L., & Donnerstein, E. (1998). Harmful eVects of exposure to
tional approach. Psychophysiology, 25, 373–391.
media violence: Learning of aggression, emotional desensitization, and
Fowles, J. (1999). The case for television violence. Thousand Oaks, CA: Sage.
fear. In R. G. Geen & E. Donnerstein (Eds.), Human aggression: Theo-
Freedman, J. L. (1984). EVects of television violence on aggressiveness.
ries, research, and implications for social policy (pp. 167–202). San
Psychological Bulletin, 96, 227–246.
Diego, CA: Academic Press.
Freedman, J. L. (2002). Media violence and its eVects on aggression: Assess-
Wolpe, J. (1982). The practice of behavior therapy (3rd ed.). New York:
ing the scientiWc evidence. Toronto: University of Toronto Press.
Pergamon Press.
Document Outline
