Violent Video Games and Physical Aggression

Violent Video Games and Physical Aggression: Evidence for a Selection Effect Among Adolescents

Johannes Breuer University of Münster

Jens Vogelgesang University of Erfurt

Thorsten Quandt University of Münster

Ruth Festl University of Münster and University of

Hohenheim

Longitudinal studies investigating the relationship of aggression and violent video games are still scarce. Most of the previous studies focused on children or younger adolescents and relied on convenience samples. This paper presents data from a 1-year longitudinal study of N � 276 video game players aged 14 to 21 drawn from a representative sample of German gamers. We tested both whether the use of violent games predicts physical aggression (i.e., the socialization hypothesis) and whether physical aggression predicts the subsequent use of violent games (i.e., the selection hypothesis). The results support the selection hypotheses for the group of adolescents aged 14 to 17. For the group of young adults (18–21), we found no evidence for both the socialization and the selection hypothesis. Our findings suggest that the use of violent video games is not a substantial predictor of physical aggression, at least in the later phases of adolescence and early adulthood. The differences we found between the age groups show that age plays an important role in the relationship of aggression and violent video games and that research in this area can benefit from a more individu- alistic perspective that takes into account both intraindividual developmental change and interindividual differences between players.

Keywords: video games, violence, aggression, adolescents, young adults

From the earliest investigations into the rela- tionship of video game1 use and aggression in the 1980s (Cooper & Mackie, 1986; Dominick, 1984; Silvern & Williamson, 1987; Winkel, Novak, & Hopson, 1987) until today, hundreds

of experimental and correlational studies have been conducted. Despite the large number of studies, the debate about the link between video games and aggression is ongoing, not only in politics and the mass media, but also within academia (Bushman & Huesmann, 2014; Elson & Ferguson, 2014a, 2014b; Krahé, 2014; War- burton, 2014). While all of the available meta- analyses (Anderson et al., 2010; Ferguson, 2007; Ferguson & Kilburn, 2009; Sherry, 2001, 2007) found a relationship between aggression and the use of (violent) video games, the size and interpretation of this connection differ largely between these studies; as do the defini-

1 We use the term video games as an umbrella term that includes all types of digital games, whether they are played on a PC, home consoles, handhelds, or mobile devices. We decided to use “video game” because it is the most common term in the literature and it is easier to read than the composite “computer and video games” or the more aca- demic denomination “digital games.”

This article was published Online First February 16, 2015.

Johannes Breuer, Department of Communication, Uni- versity of Münster; Jens Vogelgesang, Department of Com- munication, University of Erfurt; Thorsten Quandt, Depart- ment of Communication, University of Münster; Ruth Festl, Department of Communication, University of Münster, and Department of Communication, University of Hohenheim.

The research leading to these results has received funding from the European Union’s Seventh Framework Pro- gramme (FP7/2007–2013) under grant agreement number 240864 (SOFOGA).

Correspondence concerning this article should be ad- dressed to Johannes Breuer, Department of Communication, University of Münster, Bispinghof 9-14, 48143 Münster, Germany. E-mail: johannes.breuer@uni-muenster.de

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

Psychology of Popular Media Culture © 2015 American Psychological Association 2015, Vol. 4, No. 4, 305–328 2160-4134/15/$12.00 http://dx.doi.org/10.1037/ppm0000035

305

 

 

tions and measurement of violent content and aggression in the studies that were included in these meta-analyses. In addition, some meta- analyses only found a relationship for aggres- sive thoughts or feelings, but not for aggressive behavior. There is also a controversy about what exactly causes this link and, most impor- tantly, about the direction of the (potential) ef- fects.

Experimental research on video games and aggression has been criticized for a lack of ecological validity and the unstandardized use of measures of aggression that have not been properly validated (Ferguson & Rueda, 2009; Ferguson, Smith, Miller-Stratton, Fritz, & Hei- nrich, 2008; Ritter & Eslea, 2005; Tedeschi & Quigley, 1996). The issue of the real-world implications of findings from laboratory studies is further complicated by the fact that they can only investigate short-term effects that often only last for a few minutes (Barlett, Branch, Rodeheffer, & Harris, 2009). Cross-sectional correlational research, on the other hand, typi- cally has larger samples, but is unsuitable for making any claims about the direction of the effect. Longitudinal studies combine the advan- tages of cross-sectional and experimental stud- ies, as they use larger samples than most exper- imental studies and allow to sort out the temporal precedence between the variables of interest. Although it is still possible that addi- tional variables are responsible for the temporal order, given a sound control of potentially rel- evant third variables, panel studies allow to make claims about long-term effects that both cross-sectional and experimental research do not allow. Nonetheless, while panel data can help to determine direction and strengths of effects by testing for covariation and controlling for temporal order, only controlled experiments provide the means to actually prove causality (Finkel, 1995). Compared with the abundance of cross-sectional survey studies and experi- mental research, panel studies on video games and aggression are still scarce. The meta- analysis by Anderson et al. (2010), for example, included 34 effect sizes from longitudinal stud- ies2 and Ferguson and Kilburn (2009) used data from five longitudinal studies. While several longitudinal studies use a composite score for media violence that includes video games (e.g., Ferguson, Ivory, & Beaver, 2013; Gentile, Coyne, & Walsh, 2011; Krahé, Busching, &

Möller, 2012; Krahé & Möller, 2010; Ostrov, Gentile, & Crick, 2006), there are relatively few that look specifically at the effects of video games. Among those studies that explicitly in- vestigate video games, some only look at rela- tively brief periods of several months, and al- most all studies rely on convenience samples and focus on children or adolescents.

In longitudinal research on media violence and aggression, there are two seemingly com- peting hypotheses. The socialization hypothesis states that the repeated use of violent media leads to an increase of aggression over time, whereas the selection hypothesis is based on the idea of selective exposure (Zillmann & Bryant, 1985) and posits that individuals who are more aggressive will tend to choose (more) violent media content. The downward spiral model (Slater, Henry, Swaim, & Anderson, 2003) combines these hypotheses by proposing that individuals higher in trait aggression will choose more violent media content, which, in turn, increases their level of aggression. As with the experimental and cross-sectional studies, evidence from longitudinal studies on the rela- tionship between (violent) video games and ag- gression is mixed at best. Some studies found a media effect (Anderson et al., 2008; Hopf, Hu- ber, & Wei�, 2008; Möller & Krahé, 2009), while others report selection effects (von Salisch, Vogelgesang, Kristen, & Oppl, 2011), provide evidence for both (Slater et al., 2003), or found no effects (Ferguson, 2011; Ferguson, Garza, Jerabeck, Ramos, & Galindo, 2013; Fer- guson, San Miguel, Garza, & Jerabeck, 2012; Wallenius & Punamäki, 2008; Williams & Skoric, 2005).

A limitation of the previous longitudinal studies is that almost all of them rely on con- venience samples that are mostly composed of students from elementary schools, high schools, or colleges located in the areas where the re- spective researchers are based. Most studies also focus on specific grades, thereby reducing the age range of participants. In addition, even longitudinal studies often only test one direction of effects; mostly the socialization hypothesis.

2 Anderson et al. (2010) do not report the number of longitudinal studies in their paper. This number should be substantially lower than the number of effect sizes, as most longitudinal studies include cross-sectional and longitudinal effects (often also for different dependent variables).

306 BREUER, VOGELGESANG, QUANDT, AND FESTL

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

 

 

The goal of the current study was to address some of these issues by testing both the social- ization and the selection hypothesis and com- paring these relationships for adolescents and young adults, as these groups differ with regard to their developmental stage as well as their access to (violent) video games.

Theories Explaining Long-Term Effects of Video Games on Aggression

In the field of media violence research, there are three comprehensive theoretical models that aim at explaining the relationship between vio- lent video games and aggression. The most pop- ular is the General Aggression Model (GAM; Anderson & Bushman, 2002). The GAM com- bines the assumptions of social learning (Ban- dura, 1977), excitation transfer (Zillmann, 1983), and cognitive neoassociation (Berkow- itz, 1990). Long-term effects of video game violence are explained mainly by mechanisms of social learning and cognitive neoassociation. Put briefly, the GAM posits that the repeated use of violent media causes a learning, re- hearsal, and reinforcement of aggressive be- liefs, attitudes, perceptual and expectation sche- mata, and behavioral scripts, as well as an emotional desensitization to violence. In their combination, all of these processes can lead to an increase in aggressive personality and, ulti- mately, affect the likelihood to (re-) act aggres- sively in social encounters in the real world.

Although the GAM allows to formulate spe- cific hypotheses about the effects of violent video games and has been widely used in pre- vious research, it has been criticized for its overreliance on social learning, the neglect of biological factors, the conceptualization of me- dia use(r)s as passive, and the insufficient dis- tinction between real and fictional violence (Ferguson & Dyck, 2012). An alternative theory that focuses more on genetic factors and attri- butes of the social environment is the Catalyst Model (Ferguson et al., 2008). In essence, the Catalyst Model suggests that the roots of (vio- lent) criminal and aggressive behavior are ge- netic and proximal social factors, such as family and peer influences, and their interaction, whereas distal social influences, such as media violence, only have a negligible effect (Fergu- son, Ivory, et al., 2013). In this model, violent media are considered as stylistic catalysts in-

stead of sources of aggression. This means that individuals with an increased tendency for ag- gressive behavior may model violent acts they have seen in the media, whereas the actual inclination to (re-) act aggressively is not influ- enced or caused by violent media. The main limitation of the Catalyst Model is that it is difficult to test, as the measurement of genetic and proximal social risk factors poses substan- tial challenges to the methods of social science research. To date only three studies have sys- tematically tested the Catalyst Model and found support for its main assumptions (Ferguson, Ivory, et al., 2013; Ferguson et al., 2008; Surette, 2013).

The Downward Spiral Model by Slater et al. (2003) is a theory that accounts for both the socialization and the selection hypothesis. The Downward Spiral Model has also been called the negative feedback loop model by its authors (Slater, 2003) and describes a reciprocal rein- forcement of aggressive personality and prefer- ence for violent media content. Basically, the model assumes a circular relationship between current and future aggressive tendencies and use of violent media. While the inclusion of both socialization and selection effects is a strength of this model, it does not make any detailed statements about the role of other variables, such as personal experiences with violence, that could potentially moderate the relationship be- tween media use and aggression. As the down- ward spiral can only be studied in longitudinal designs that ideally also include more than two waves, there have been few studies that actually tested this model (Ferguson, 2011; Möller & Krahé, 2009; Slater et al., 2003; von Salisch et al., 2011; Willoughby, Adachi, & Good, 2012) and only one of these studies provided some empirical support for it (Slater et al., 2003).

Longitudinal Studies on Video Games and Aggression

As mentioned before, the number of longitu- dinal studies on video games and aggression is still relatively small. As the present study was concerned with video games, the overview in this section will focus on studies that were pub- lished in peer-reviewed journals and explicitly looked at the relationship between aggression and video games, and not violent media content in general. One of the earliest studies focusing

307VIOLENT VIDEO GAMES AND PHYSICAL AGGRESSION

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

 

 

on video games was the short-term longitudinal field study by Williams and Skoric (2005). This study investigated the effect of one particular massively multiplayer online role-playing game (MMORPG) on aggressive cognitions and be- haviors. Later studies looked at longer periods (typically between 1 and 2 years) and video games in general or at least specific genres or types of games (mostly “violent” games; with varying definitions of what “violent” means). In our review of the literature, we found 11 journal publications that present longitudinal data from studies dealing specifically with the relationship of aggression and video games. Table 1 sums up their methods (sample, design, and measures) and main findings. Overall, there are vast dif- ferences between these studies with regard to both the direction (socialization vs. selection) and the size of the effects they found. A big part of the inconsistencies in the results can be at- tributed to major methodological discrepancies between the individual studies. The longitudinal studies differ from one another in various re- spects, including size, origin, and composition of the sample; measures of aggression and ex- posure to violent video games; control vari- ables; and number of and time lag between waves (Table 1). While the differences in some crucial categories, such as the measures for ag- gression and exposure to violent content, are quite substantial, other features are much more homogeneous across studies. Although sample sizes vary between N � 143 (Ferguson, Garza, et al., 2013; Möller & Krahé, 2009) and N � 1,492 (Willoughby et al., 2012), almost all of them are convenience samples and the large majority include only children and/or adoles- cents (Table 1).

Summing up the comparisons in Table 1, it can be noted that both the methods and results of longitudinal studies on the link between ag- gression and video game use are very heteroge- neous. This heterogeneity of findings and mea- sures is somewhat contrasted by a relative homogeneity in the age and recruitment of the samples.

Aggression, Violent Video Games, and Age

Our review of previous longitudinal studies on video games and aggression revealed that the majority of them worked with convenience samples of children and teenagers, with the ex-

ception of the study by Williams and Skoric (2005) that used a self-selected online sample that also included adult players. Accordingly, the age range of the samples typically only spans a few years (M � 4.8 years for the nine studies in Table 1 that report the age range of their sample). Due to the limited age range of most studies, few of them have investigated the role of participant age in detail. While control- ling for participant sex is done in most studies, only a few control for age (Wallenius & Pu- namäki, 2008; Williams & Skoric, 2005) or specifically look at potential differences in the size and direction of effects between age groups (Anderson et al., 2008; Ferguson, Garza, et al., 2013; Willoughby et al., 2012).3 Most of the studies that did compare between age groups also found differences in terms of effect size. In the study by Willoughby et al. (2012), there were only small socialization effects from grades 9 to 10 (� � .06) and 11 to 12 (� � .08), but not from grades 10 to 11, when controlling for all of the measured third variables. Ander- son et al. (2008) found stronger socialization effects for the younger samples (� � .15) than for the older sample (� � .08). However, the study by Ferguson, Garza, et al. (2013) that found no effect of exposure to video game vi- olence on aggression, bullying, and delinquency also found no differences between the groups of late childhood (ages 10–11), preadolescence (12–13), and adolescence (14–17). With regard to age differences, von Salisch et al. (2011) suggest that the selection effect they found in their study with third and fourth graders may be replaced by socialization effects once media preferences have become more stable at an older age.

In a review of the literature on violent video games and aggression, Kirsh (2003) laments the absence of a developmental perspective. For the case of video game violence and aggression, this is especially problematic, as research has shown that video game preferences differ be- tween age groups and also change over time (Greenberg, Sherry, Lachlan, Luca, & Holm- strom, 2010). Genres that typically include

3 While Willoughby et al. (2012) compared the effect sizes across three waves for the same sample, Anderson et al. (2010) calculated a combined model that distinguished between younger and older participants with the data from the two Japanese studies and the one American study.

308 BREUER, VOGELGESANG, QUANDT, AND FESTL

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

 

 

T ab

le 1

O ve

rv ie

w of

L on

gi tu

di na

l St

ud ie

s on

V id

eo G

am es

an d

A gg

re ss

io n

St ud

y C

ou nt

ry Sa

m pl

e Sa

m pl

e ty

pe

M ea

su re

of vi

de o

ga m

e vi

ol en

ce M

ea su

re (s

) of

ag gr

es si

on C

on tr

ol va

ri ab

le s

W av

es

T im

e la

g be

tw ee

n w

av es

So ci

al iz

at io

n ef

fe ct

Se le

ct io

n ef

fe ct

W ill

ia m

s &

Sk or

ic (2

00 5)

U SA

N �

21 3;

A ge

14 –6

8 (M

� 27

.7 )

Se lf

-s el

ec te

d Se

lf -r

ep or

te d

ho ur

s of

pl ay

fo r

M M

O R

PG A

sh er

on ’s

C al

l 2

N or

m at

iv e

B el

ie fs

in A

gg re

ss io

n ge

ne ra

l sc

al e

(H ue

sm an

n &

G ue

rr a,

19 97

); tw

o be

ha vi

or al

qu es

tio ns

on ag

gr es

si ve

so ci

al in

te ra

ct io

ns

Se x,

ag e

2 1

m on

th N

or m

at iv

e be

lie fs

ab ou

t ag

gr es

si on

: B

� .2

5, n.

s. ;

A rg

um en

ts w

ith fr

ie nd

s: B

� �

1. 63

, n.

s. ;

A rg

um en

ts w

ith pa

rt ne

r: B

� �

.0 4,

n. s.

N ot

te st

ed (fi

el d

st ud

y)

W al

le ni

us &

Pu na

m äk

i (2

00 8)

Fi nl

an d

N �

31 6;

A ge

(T 2)

12 –1

5 (M

� 13

.8 )

C on

ve ni

en ce

Se lf

-r ep

or te

d ra

tin gs

of vi

ol en

ce in

ga m

es pl

ay ed

an d

fr eq

ue nc

y of

pl ay

in g

ac tio

n, fig

ht in

g, an

d sh

oo tin

g ga

m es

T en

ite m

s fr

om th

e D

ir ec

t &

In di

re ct

A gg

re ss

io n

Sc al

e (B

jö rk

qv is

t, L

ag er

sp et

z, &

Ö st

er m

an ,

19 98

)

Se x,

ag e-

gr ou

p, pa

re nt

–c hi

ld co

m m

un ic

at io

n

2 2

ye ar

s �

� .0

1, n.

s. N

ot te

st ed

(t ab

le co

nt in

ue s)

309VIOLENT VIDEO GAMES AND PHYSICAL AGGRESSION

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

 

 

T ab

le 1

(c on

ti nu

ed )

St ud

y C

ou nt

ry Sa

m pl

e Sa

m pl

e ty

pe

M ea

su re

of vi

de o

ga m

e vi

ol en

ce M

ea su

re (s

) of

ag gr

es si

on C

on tr

ol va

ri ab

le s

W av

es

T im

e la

g be

tw ee

n w

av es

So ci

al iz

at io

n ef

fe ct

Se le

ct io

n ef

fe ct

Sh ib

uy a

et al

. (2

00 8)

Ja pa

n N

� 59

1; A

ge (T

1) 10

–1 1

(m ea

n ag

e no

t re

po rt

ed )

U nc

le ar

(m os

t lik

el y

co nv

en ie

nc e)

D ic

ho to

m ou

s va

ri ab

le (v

io le

nt ye

s/ no

) fo

r th

re e

fa vo

ri te

vi de

o ga

m es

an d

21 co nt

ex tu

al va

ri ab

le s

ab ou

t th

e ty

pe of

vi ol

en t

co nt

en t

in ea

ch ga

m e

A gg

re ss

io n

Sc al

e fo

r C

hi ld

re n

(B us

s &

Pe rr

y, 19

92 ;

Sa ka

i et

al .,

20 00

); an

ti- vi

ol en

ce no

rm s;

re ce

nt ag

gr es

si ve

be ha

vi or

Se x,

ar ea

of liv

in g,

w ee

kl y

vi de

o ga

m e

us e

2 1

ye ar

H os

til ity

: �

� .1

3, p

� .0

5 fo

r bo

ys an

d n.

s. fo

r gi

rl s;

Ph ys

ic al

an d

ve rb

al ag

gr es

si on

, an

tiv io

le nc

e no

rm s,

an d

ag gr

es si

ve be

ha vi

or :

al l

n. s.

fo r

bo th

gi rl

s an

d bo

ys (e

ff ec

t si

ze s

no t

re po

rt ed

)

N ot

te st

ed

H op

f et

al .

(2 00

8) G

er m

an y

N �

31 4;

M ea

n ag

e (T

1) �

12

C on

ve ni

en ce

Fr eq

ue nc

y of

pl ay

in g

vi ol

en t

ga m

es fr

om a

lis t

of 19

po pu

la r

tit le

s

It em

s on

vi ol

en ce

be lie

fs ,

de lin

qu en

cy ,

ve rb

al ag

gr es

si on

, ph

ys ic

al ag

gr es

si on

, an

d de

vi an

ce in

sc ho

ol (T

ill m

an n,

H ol

le r-

N ov

itz ki

, H

ol ta

pp el

s, M

ei er

, &

Po pp

, 19

99 )

Sc ho

ol an

d cl

as sr

oo m

cl im

at e,

w el

l- be

in g

in sc

ho ol

, se

lf –

re gu

la tio

n, se

lf –

ef fic

ac y

in sc

ho ol

, em

ot io

na l

re ac

tio ns

to vi

ol en

ce ,

m at

er ia

lis tic

va lu

e or

ie nt

at io

ns ,

m ed

ia ed

uc at

io n

by pa

re nt

s, po

ve rt

y, ag

gr es

si ve

ne ss

, pe

ac ea

bl en

es s

2 2

ye ar

s A

gg re

ss iv

e be

ha vi

or in

sc ho

ol :

� �

.1 8

(s ig

ni fic

an ce

le ve

l no

t re

po rt

ed );

D el

in qu

en cy

: �

� .2

9 (s

ig ni

fic an

ce le

ve l

no t

re po

rt ed

)

N ot

te st

ed

310 BREUER, VOGELGESANG, QUANDT, AND FESTL

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

 

 

T ab

le 1

(c on

ti nu

ed )

St ud

y C

ou nt

ry Sa

m pl

e Sa

m pl

e ty

pe

M ea

su re

of vi

de o

ga m

e vi

ol en

ce M

ea su

re (s

) of

ag gr

es si

on C

on tr

ol va

ri ab

le s

W av

es

T im

e la

g be

tw ee

n w

av es

So ci

al iz

at io

n ef

fe ct

Se le

ct io

n ef

fe ct

A nd

er so

n et

al .

(2 00

8) U

SA N

� 36

4; A

ge 9–

12 (m

ea n

ag e

no t

re po

rt ed

)

U nc

le ar

(m os

t lik

el y

co nv

en ie

nc e)

Se lf

-r ep

or te

d am

"Is this question part of your assignment? We can help"

ORDER NOW