Burkhardt Lenhard 2021

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 29

AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 1

This article has been accepted for publication in Media Psychology, published by Taylor
& Francis:

Burkhardt, J. & Lenhard, W. (2021) A Meta-Analysis on the Longitudinal, Age-Dependent


Effects of Violent Video Games on Aggression. Media Psychology. DOI:
10.1080/15213269.2021.1980729

A Meta-Analysis on the Longitudinal, Age-Dependent

Effects of Violent Video Games on Aggression

Johanna Burkhardt & Wolfgang Lenhard

Author Note

Johanna Burkhardt and Wolfgang Lenhard, Institute of Psychology, Julius-

Maximilians-University of Würzburg, Germany.

There are no conflicts of interest, which might have influenced the conduction,

analysis and interpretation in this study. The data and scripts are available via

https://osf.io/9d7jf/. We would like to thank Marlene Voit for assisting in the coding of the

primary studies.

Correspondence should be addressed to Wolfgang Lenhard, Institute of Psychology,

Julius-Maximilians-University of Würzburg, Wittelsbacherplatz 1, 97074 Würzburg,

Germany. E-Mail: wolfgang.lenhard@uni-wuerzburg.de, Phone: +49 931 31 89791, FAX:

+49 931 31 84891


AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 2

A Meta-Analysis on the Longitudinal, Age-Dependent Effects of Violent Video Games


on Aggression

A large body of research exists on the causal relationship between violent video game
(VVG) consumption and aggression. However, no meta-analytic studies to date have
specifically examined the influence of age on these effects. We investigated the age-
dependent relationship between VVG consumption and subsequent physically
aggressive behavior in a meta-analytic procedure based on longitudinal studies. Further
exploratory analyses addressed the moderating effect of the year of data collection. A
random-effects meta-analysis with 30 effect sizes from 21 studies (N = 15,836) yielded
a significant and relevant positive effect of VVG on subsequent physically aggressive
behavior (r = .21). The effect was reduced but maintained at r = .11 when controlling
for aggression at T1. Polynomial regression showed a significant U-shaped age
trajectory with a peak in early adolescence at the age of 14. Year of data collection did
not correlate with the effect size. Consequently, we confirm the results of other meta-
analyses, with the added finding of a peak of effect sizes in early adolescence. The
empirical finding provide support for basing age ratings for game publication on
potential adverse behavioral effects.

Keywords: violent video games, aggressive behavior, age effects, meta-analysis


AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 3

Research on violent video games (VVG) usually considers the effects of these games

in relation to some form of aggressive behavior. Evidence from a large body of research –

including experimental, cross-sectional and longitudinal studies – indicates that VVG is

related to aggression (e.g., Anderson et al., 2010; Greitemeyer & Mügge, 2014; Prescott et al.,

2018). Experimental and longitudinal designs suggest a causal link, with higher amounts of

VVG consumption causing higher aggression levels, which can be explained by underlying

mechanisms of social cognition and hostile attribution (Crick, 1995; Möller & Krahé, 2009),

social learning theory as suggested by the General Aggression Model (Anderson & Bushman,

2002; 2018), operant reinforcement and associative networks (Bushman, 1998) and other

theories. The interpretation of the research findings and underlying mechanisms are still

debated (cf., Huesmann, 2010; Elson & Ferguson, 2014), which revolves around

methodological questions like publication bias.

Although scholars generally agree that VVG causally affect aggression, some aspects

remain under-researched. This paper will focus on two such neglected aspects, which have yet

to be studied in detail, namely age and the year of data collection. First, children, adolescents,

and adults alike play video games. Arguably, these different age groups fall into different

stages of development, potentially moderating the mechanisms and effects of violent media.

Consequently, policy makers base their decisions on age restrictions accordingly. Second,

considering the year of data collection, the games and the amount of exposure have

dramatically changed during the last decades, which in turn has most likely influenced the

mechanisms and magnitudes of the effects of playing VVG. Given that aggression depicted in

VVG is typically physical in nature (kicking, hitting, shooting, etc.), as opposed to verbal or

relational aggression, we focused on physically aggressive behavior.

The General Aggression Model

The most comprehensive model of the effects of VVG on aggressive behavior is the

General Aggression Model (GAM; Anderson & Bushman, 2002, 2018). The key features of
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 4

GAM are knowledge structures such as perceptual and person schemata, behavioral scripts,

beliefs and attitudes. The authors propose that these structures originate from experience.

They are connected to affective states and behavioral programs, and influence perception and

a person’s interpretation of their social and physical environment. Thus, the theory suggests

that VVG influence the players’ aggressive cognitions, affect and arousal, which in turn leads

to changes in their behavior (i.e., more frequent aggressive behavior).

Why Age could potentially moderate the Effects of VVG

Most research on the longitudinal effects of VVG has focused on people below the age

of 16 (Prescott et al., 2018). The sample of interest in many studies is thus subject to an age

span in which fundamental psychological and neuroendocrinological changes occur

(DePasque & Galván, 2017; Peters & Cone, 2017). These processes are an important catalyst

for the receptiveness to aggressive stimuli, which in turn might increase the effects of violent

media. Given the variation of puberty onset age, we focused on the mean onset age in the

current study (defined as gonadarche onset, Blakemore et al., 2010) and distinguished

between childhood, early adolescence as the period roughly between 11 and 13, middle and

late adolescence, and adulthood. The age developments can affect media preference,

aggressiveness, potential media effects and the interaction of these factors. Surveys conducted

among both German and American adolescents suggest that the amount of time spent playing

video games changes over the course of adolescence, peaking between the ages of 12 and 14

compared to both younger and older age groups (Greenberg et al., 2010; MPFS, 2017), with a

simultaneous shift in genre preference towards more violent games like first-person shooters

(Griffith et al., 2004; Quandt et al., 2013).

A comparable trajectory applies for the development of aggression over the lifespan

with research indicating a peak in aggression in early adolescence, although the exact age

range varies between studies (Lindeman, Harakka and Keltikangas-Järvinen, 1997; Loeber &

Stouthamer-Loeber, 1998). A growth in levels of gonadal and adrenal hormones in early


AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 5

adolescence seemingly relates to this increase in aggression (Carré et al., 2017). Hormonal

changes also amplify reward-sensitivity, an effect which is mediated by cerebral changes in

reward-seeking areas (i.e., the nucleus accumbens and dopaminergic pathways to the

prefrontal cortex; Blakemore et al., 2010). The heightened reward-sensitivity in adolescence

may facilitate learning by means of performance-related feedback (DePasque & Galván,

2017; Peters & Crone, 2017). In line with this assumption, research suggests that implicit-skill

learning changes depending on age: It is more effective until early adolescence than later in

life (Janacsek et al., 2012), potentially making these younger players more receptive to the

reward structures in VVG, which often reinforce the exhibition of aggressive behavior.

Steinberg (2010) adds in his dual systems model of adolescent risk-taking that while

reward-seeking behavior increases at the beginning of puberty, concurrently, impulse control

and self-regulation reach full maturation only later in life. Steinberg found evidence for this

development in his research, with self-reported impulsivity decreasing steadily from ages 10

and 11 onwards and self-reported reward-seeking increasing between ages 10 and 13 and then

decreasing. Because of this evident time lag between the two systems’ development,

Steinberg supposed a vulnerable period of risk-taking roughly during middle adolescence.

Kirsh (2003) relates these developmental changes to the GAM: Heightened aggression

in early adolescence might impact cognition, affect, and arousal, making aggressive

cognitions (e.g., scripts) more accessible, which also leads to them being reinforced more

frequently. Increased potential for aggressive affect facilitates aggressive responses. Lastly,

the heightened arousal in general (because of endocrine changes) and arousal elicited by

playing VVG cumulate. Together with the changes in risk-taking behavior during early

adolescence, this arousal increases the probability for impulsive (aggressive) action.

Empirical Findings on Violent Media Effects

A number of meta-analyses are available that have assessed magnitude of effects and

moderators of media violence on behavior and cognition. Notably, these results have often
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 6

been the subject of fierce debate with respect to methodological issues and the estimation of

the general effect size (e.g., Ferguson & Kilburn, 2010; Bushman et al., 2010; Elson &

Ferguson, 2014; Bushman & Huesmann, 2014; Rothstein & Bushman, 2015; Hilgard et al.,

2017). For example, Anderson et al. (2010) found that VVG exposure was significantly

related to higher levels of aggressive behavior in experimental, cross-sectional and

longitudinal studies and that this effect was largest for best-practice studies (r = .24)

compared to the full sample (r =.19). Prescott et al. (2018) focused exclusively on

longitudinal studies that investigated the effects of VVG on subsequent physical aggression.

Using standardized regression coefficients and controlling for covarying factors and pooling

these effect sizes in a random-effects meta-analysis yielded a smaller but still significant

result (r = .11). One of the most recent meta-analyses (Drummond, Sauer & Ferguson, 2020)

also focused on longitudinal studies with standardized effect sizes and yielded an overall

effect size of r = .059. Moreover, effect sizes decreased with study year.

Age was not a significant moderator of effect sizes in any of these analyses, although

(conflicting) descriptive results indicated both smaller effect sizes for older participants

(Anderson et al., 2010) and larger effect sizes for participants of age 13 and older (Prescott et

al., 2018). These analyses, however, only investigated linear relationships between age and

effects. To our knowledge, the (non-linear) moderating effect of age on the relationship

between VVG and aggression has yet to be a strong research focus.

Publication Year or Year of Data Collection as a Potential Moderator

The technology with which video games are created has become increasingly

elaborate over the last decade. These advances have allowed for more realistic and immersive

game design and, in extension, more realistic and immersive depictions of violence (Ivory &

Kalyanaraman, 2007; McGloin et al., 2013). Immersion can be defined as “a psychological

state characterized by perceiving oneself to be enveloped by, included in, and interacting with

an environment” (Witmer & Singer, 1998, p. 227). A higher level of immersion and greater
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 7

realism in games might affect players’ knowledge structures more gravely because the virtual

experiences seem more real and observational learning is facilitated to a greater extent.

Existing research already suggests that more immersive game play intensifies the short-term

effects of VVG on angry feelings (Lull & Bushman, 2016) and is related to higher cognitive

aggression (McGloin et al., 2013). Assuming that participants in more recent studies played

more recent games, effects should thus be stronger in more recent studies.

The representation of violence in the media has also increased over the last decade,

since the number of armed conflicts around the world has increased as well (Marshall &

Elzinga-Marshall, 2017). Through news outlets and especially social media, people—

specifically children—might be exposed to the idea and depiction of violence more frequently

than before. Research suggests that continued exposure to video game violence leads to

emotional desensitization to real-life aggression and violence, which in turn facilitates

aggressive behavior (Anderson & Bushman, 2018; Carnagey et al., 2007). The effect might be

bidirectional, that is, continued exposure to violence on the news and social media might

diminish the influence of VVG on the player because they are already more desensitized to

the content they consume. Therefore, the effects of VVG on aggression might even be weaker

today than they were 10 or 20 years ago. On the other hand, given that different media

consistently present aggressive content, the effect might add up or interact in complex way.

For instance, Gentile and Bushman (2012) showed that different risk factors, including

exposure to media violence, can have an additive effect on aggression. This additive effect

might extend to consuming video game violence and consuming violence via other media,

though testing an additive or interaction effect is beyond the scope of this manuscript.

Rationale

We hypothesized and analyzed two effects: first, as established by previous meta-

analytic research, playing VVG increases later physically aggressive behavior (H1). Second,

due to developmental aspect, game play in early adolescence might have a greater impact on
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 8

behavior than in late adolescence. Thus, we expected an inverse U-shape curvilinear

relationship between age and effect size: the magnitude of the effect investigated in H1 should

increase in childhood, peak in early adolescence (roughly between ages 11 and 13), and then

decline steadily through later adolescence into adulthood (H2). Based on the reviewed

literature, we assumed an effect of the year of data collection on the relationship between

VVG and aggression, but the direction of this effect can lead in either direction. Thus, the

relationship will be investigated in an exploratory analysis (H3).

Methods

Literature Search

We addressed the hypotheses by synthesizing existing research literature via meta-

analysis. The search for primary studies was conducted using the electronic databases

PsycINFO, PubMed, and Web of Science using the following keywords: (video game* OR

videogame* OR computer game* OR electronic game*) AND (longitudinal OR prospective)

AND (aggress* OR violen* OR delinquen*). The databases were searched in October 2019,

and again in March and August 2020. The initial results included studies published up to July

2020. We additionally applied a snowball strategy and searched the reference sections of

existing meta-analytic and literature reviews. Articles, book chapters and dissertations both

published and unpublished were eligible for inclusion as long as they appeared in the

databases and reviews searched. After removing duplicates, this search yielded a total of 181

results (see Fig. 1; list of included and excluded studies in ESM 1; analysis of the overlap

with other meta-analyses in ESM 3).


AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 9

Figure 1. Consort flow diagram of the selection of studies

Inclusion Criteria and Coding

To be included in the analysis, studies had to meet six inclusion criteria. (1) The study

had to include a measure of VVG exposure at the first time of measurement. Studies were

excluded if they exclusively measured overall video game exposure (including non-violent

games) or exposure to violent media in general (as opposed to VVG in particular) and did not

include a separate effect size for exposure in particular. Acceptable measures for video game

violence were self-reported ratings, expert ratings, and official ratings (e.g., assigned by the
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 10

Entertainment Software Rating Board, ESRB). (2) The study had to include a measure of

physically aggressive behavior. Acceptable measures were self-reports, peer-reports, parent-

reports or teacher-reports. If several different aggression measures and their correlations with

VVG exposure were reported, we used the best fitting outcome measure, for example,

“clinical aggression” as opposed to “civil aggression” in Fergusonet al. (2013). (3) When

experimental studies are not feasible for ethical reasons, as it is the case of VVG effects on

minors, longitudinal studies represent a potential way to provide the closest possible

approximation to causal inferences. Consequently, only studies using a longitudinal research

design were included. (4) Studies had to present zero-order correlations between the VVG

measure and the subsequent aggression measure(s). We additionally coded partial

correlations, controlling for aggression at T1 for subsequent analyses. Studies that exclusively

presented partial correlations or standardized regression coefficients were not included in the

analysis. If zero-order correlations were missing but the study was otherwise eligible for

inclusion, we contacted the study authors to obtain the missing coefficients. If studies

presented separate zero-order correlations for either boys and girls, different age groups or

different reports (self, other), the correlations were coded as separate effect sizes and included

in the analysis. If correlations for several different time intervals were reported, the interval

closest to the mean interval was chosen. In two cases (Greitemeyer 2019; Greitemeyer &

Sagioglou, 2017), the authors courteously provided the raw data and we computed the

coefficients for the age groups 25 years or younger. (5) The studies had to be written in

German or English. (6) We identified several studies that were based on the same or on

overlapping datasets. To avoid overrepresentation of data, we only included the study that had

been published first.

Studies were reviewed separately by two coders in relation to the inclusion criteria.

Age, year of publication as well as data collection, time interval between measurement points,

sample size and effect size were coded as well as the reason for possible exclusion of the
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 11

study (i.e., aggression measure, VVG measure and bivariate versus standardized effect size).

In case of discrepancies between the two independent datasets, the original papers were re-

screened to correct for possible errors. This procedure formed the basis for the final decision

on inclusion or exclusion, which was made by the lead author.

Even though Ferguson (2011), Ferguson et al. (2013), and von Salisch et al. (2011) all

used outcome variables that partly confounded verbal and physical aggression, the authors

argued that the relationships between those outcomes were high enough to justify a combined

use. After considering these arguments, the studies were included into the analysis even

though they did not completely fulfil the inclusion criteria.

After deciding which studies had met the inclusion criteria, the following study

characteristics were coded: effect size labels including authors and further identifiers, sample

size n, effect size based on the zero-order Pearson’s correlation between VVG consumption

and subsequent aggression, effect size for the partial correlation between VVG consumption

and subsequent aggression (controlling for initial aggression), mean sample age mage at the

first point of measurement (T1), year of data collection, time between measurements (lag).

The mean sample age was based on either direct age information from the study, or it was

estimated based on the participants’ school year. This procedure resulted in a final sample of

21 studies with k = 30 coded effect sizes and a total N of 15,836 participants. Since partial

correlations could only be obtained for a part of the primary studies, k = 13 effect sizes were

available for the controlled effect sizes.

Statistical Analysis

We used R version 3.6.1 (R Core Team, 2019) to conduct the analysis, using the

packages meta (Schwarzer, 2020), metafor (Viechtbauer, 2010), dmetar (Harrer et al., 2019)

and psych (Revelle, 2018). To calculate a pooled estimate, we used the effect size Pearson’s r

and standardized it using Fisher’s z transformation. The same was done for the partial

correlations. Anderson et al. (2010) point out that “the statistical theory underlying meta-
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 12

analytic procedures assumes that one is working with raw (zero-order) correlations”, while

Ferguson and Kilburn (2010) argue that “focusing on bivariate correlations is problematic”.

Although using partial correlations better represents longitudinal effects, the effect of prior

playing of VVG on aggression is not accounted for. In this way, the studies are less

comparable and tend to rather specify the lower bound of the longitudinal VVG effect. We

have tried to address the problem by reporting both approaches and specifying zero-order

correlations as the upper bound of the estimated true effect and partial correlations controlling

for T1 aggression as the lower bound.

An assessment of heterogeneity using Cochran’s Q, Higgin’s and Thompson’s I2 and

the Sidik-Jonkman estimator for τ2 indicated considerable heterogeneity across studies. Thus,

we chose to calculate a random-effects model. To further examine heterogeneity across

studies, we computed a Baujat plot (Baujat, Mahé, Pignon & Hill, 2002) and conducted both

an influence and an outlier analysis using the method proposed by Viechtbauer and Cheung

(2010) for the full sample of studies.

To test for publication bias, we first used the Egger’s test (Egger, Smith, Schneider &

Minder, 1997) to examine funnel plot asymmetry. In addition, we calculated an estimate for

the number of studies with nonsignificant results that would be needed to turn the pooled

effect size of this analysis nonsignificant, using Rosenthal’s fail-safe N (Borenstein et al.,

2009, pp. 282-286) for the full sample.

To test for the age-dependent effects of VVG on aggression, we calculated a

polynomial regression and predicted the effects (Cohen’s d) by mean sample age as the

predictor and sample size as weights. To decide on the statistical hypotheses, we drew on the

significance of the linear and quadratic effect in the polynomial regression. The exploratory

analysis of the relationship between year of data collection and effect size d was based on

Pearson’s product-moment correlation.


AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 13

Results

Descriptives

On average, the participants’ mean age was Mage = 13.5 years, ranging from mage = 8.9

to mage = 23.1 years for the full sample, and Mage = 15.1 years (mage = 8.9 to mage = 23.1 years)

for the partial effect studies. Sample sizes of the full set of primary studies ranged from n =

111 to n = 2,725, with a median of n = 302. The time of data collection was evenly distributed

between 1999 and 2017. Time interval between measurements also varied, ranging between 4

months and 2.5 years, with Mlag = 0.96 years (see Electronic Supplement ESM 2).

Analysis of Overall Effects

A random-effects meta-analysis yielded a pooled effect size of r = .21, 95% CI = (.16,

.25), z = 8.31, p < .001 as the upper bound of effects for the full sample. Drawing on the

pooled partial effects controlling for aggression at T1, the effect size was reduced to r = .11,

95% CI = (.06, .15), z = 4.92, p < .001, representing the lower bound of the estimated true

effect. Thus, the analysis revealed a significant, positive influence of VVG on subsequent

physically aggressive behavior. The results indicated significant heterogeneity, with a Q

statistic of χ2(29) = 161.98, p < .001 and I2 = 82.1% for the full sample (see Figure 2).
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 14

Figure 2. Forest plot of individual sample characteristics and the pooled effect size (full

sample)

Note. Distribution of the effect sizes with the source of the information (parent- or child-self-

report) and the total sample size n, mean sample age mage, the individual correlations in

relation to the pooled effect size estimate.

Omitting outliers as indicated by the Baujat plot (Baujat et al., 2002) and influence

analysis (Viechtbauer & Cheung, 2010) still resulted in significant heterogeneity with a Q

statistic of χ2(21) = 44.87, p = .002 and I2 = 53.2%. Due to this, as well as the high stability of

the overall results (routlier = .207, roriginal = .206), we decided not to exclude these studies to

avoid reducing statistical power and representativeness. No significant asymmetry was found
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 15

in the effects of the primary studies as indicated by a nonsignificant Egger’s test, t(29) = -

1.39, p = .177. Thus, the results were unlikely distorted from selective publication or

“missing” studies due to the file drawer problem. The number of nonsignificant missing

studies required to render the pooled effect size nonsignificant according to Rosenthal’s fail-

safe N = 6,656, which is also indicates no publication bias.

Analysis of Age-Dependent Effects

The polynomial regression of second degree (Figure 3) displayed an effect of age with

the expected trajectory, thereby descriptively supporting H2 of an inverse U-shaped

relationship between mean sample age and effect size. The effect sizes fitted by the model for

the complete sample of studies increased from childhood into adolescence, peaking between

the ages of 13 and 14, with the predicted effect size of dmax = .49 at mage = 13.80. Effect sizes

then continuously decreased into adulthood, reaching their minimum at mage = 23.15 with dmin

= .17. However, the effect could only be substantiated for the intercept, I = 0.45, t(29) =

10.15, p < .001. Conversely, repeating this analysis with the reduced sample of k = 13 effects

of the partial effect studies exhibited a significant quadratic effect in the expected direction,

but with a peak in later adolescence between the ages of 14 and 16 (a = -0.18, t(12) = -1.03, p

= .329; b = -0.38, t(12) = -2.62, p = .025).


AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 16

Figure 3. The distribution of effect sizes (Cohen’s d) among mean sample age (full sample)

Note. Line displays fitted values of a polynomial regression model of second degree.

Exploratory Analysis

The correlation between publication year and effect size d was significant, r = -.36,

t(28) = -2.07, p = .047, 95% CI (-.64, -.01), indicating that effect sizes decreased for data

collected between the years 1999 and 2017. To determine whether this was related to an

increase in the publication of nonsignificant results, we did a vote count on the significance of

the studies. We calculated a point biserial correlation between significance and year of data

collection and found a nonsignificant negative correlation of r = -.25, t(25) = -1.28, p = .213,

95% CI (-.57, .15). Although there appears to be a trend showing an increase in nonsignificant

results in recent years, the decrease in effect sizes cannot be attributed solely to the

publication of null effects. Repeating the analysis with the partial effects sample showed no
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 17

significant effect of year of data collection. Thus, the results are mixed and in sum, the

assumption of constancy of the effect over time cannot be rejected.

Discussion

We investigated the age-dependent relationship between VVG consumption and

subsequent aggression (defined as physically aggressive behavior) in a meta-analytic

procedure using zero-order correlations from longitudinal studies and a subsample of partial

correlations controlling for T1 aggression. First, we expected a significant positive

relationship in the overall analysis. Our random-effects meta-analysis supported this

hypothesis, yielding a significant pooled effect size at r = .21 and an effect of r = .11 for the

partial longitudinal effects. The first of the effect sizes is at a comparable level to that

described by Anderson et al. (2010), whereas the second is consistent with Prescott et al.

(2018). Given the high consistency of the results, we assume that the true effect is within this

range. Effects of this magnitude are common for social psychology (Richard, Bond Jr. &

Stokes-Zoota, 2003) and similar in size to other risk factors of aggression, for example, the

influence of poverty, substance abuse, or low IQ (Bushman et al., 2010). Our findings suggest

that playing VVG can affect players’ behavioral disposition by increasing the potential for

aggressive behavior to a similar extend, thus yielding comparable practical relevance.

Age-Dependent Effects

Secondly, we expected a maximum of the effect in early adolescence and a decrease

towards adulthood and hypothesized a vulnerable phase for the effects of VVG on aggression

in early adolescence. This assumption was based on research showing increased reward-

sensitivity during early adolescence due to facilitated learning by means of performance-

related feedback (DePasque & Galván, 2017; Peters & Crone, 2017) and the influence of the

neuro-development of reward-seeking and impulse control systems on risk-taking behavior

(Steinberg, 2010). Our assumption was supported descriptively by a polynomial regression of

second degree with a maximum between the ages of 13 and 14 using the full sample, whereas
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 18

the partial sample revealed a significant curvilinear relationship with a maximum between the

ages of 14 and 16.

Year of Data Collection and Effect Size

We assumed that video games played by participants in the oldest sample would differ

considerably from those played by participants more recently. Based on theoretical

considerations of immersion, methodological limitations and desensitization (Lull &

Bushman, 2016; Anderson & Bushman, 2018; Carnagey et al., 2007), we could not predict

the direction of the relationship between year and effect sizes. We found a significant

decrease of effect sizes in more recent studies for the full sample, but not for the partial

effects sample. We can rule out an increase in null effects studies within the last two decades.

Bearing these mixed results in mind, changes in effect sizes might be due to changes in study

methods over time and general changes in media consumption in society. The methodological

aspects relate to the way the effects are measured and analyzed, including the use of specific

questionnaires, the list of games being considered as violent, potential ceiling effects,

conservative testing and the granularity in the assessment of VVG consumption intensity and

type. To determine the influence of construct validity on the results would require an in-depth

analysis of the quality of the measures in each individual study.

As we have mentioned in our theoretical considerations, video games have changed

over the last decade. Higher immersion might facilitate learning, and the depiction of violence

is more graphic today than it was in the last century. Krcmar et al. (2011) found empirical

support for this observation by comparing Doom 1 and Doom 3 (released in 1993 and 2004

respectively). In their study, participants who played Doom 3 rated it to be more realistic and

experienced more presence compared to those who played Doom 1. The authors also found

physically aggressive intentions to be higher in those who played the newer, more realistic

Doom 3, compared to those who played the less realistic Doom 1. Notably, the graphics of

video games have become extremely more realistic since Doom 3, with the most recent Doom
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 19

(2016) being a prominent example. Unfortunately, this augmentation of violence is not

represented in the data because of the maximum rating for VVG (e.g., “USK 18” or the

“mature” rating). The rating system leads to a ceiling effect because the basis for the

comparison between games considered as being violent or not has changed. This confounding

factor might have led to a methodological artefact and an underidentification of potential

effects.

Behavioral consequences in the form of desensitization might also be a confounding

factor. Desensitization is a form of habituation that occurs because of repeated exposure to

(depictions of) violence. We have argued before that the daily exposure to violence in social

media and in the news has increased over the last decade. If people playing VVG are already

desensitized to violence because of these other sources, the games’ effects on their personality

and behavior will be smaller.

Limitations and Methodological Considerations

Some researchers argue that publication bias is an important issue in this field of

research (Ferguson & Kilburn, 2010). Our sample of studies showed no evidence of this file-

drawer bias. Studies with null or negative results or weak effect sizes were not significantly

underrepresented in our sample compared to studies with significant positive results. An

outlier analysis further indicated both a high stability of our overall results as well as

significant and considerable heterogeneity in our analysis, indicated by a significant

Cochran’s Q and a high I2 (comp. Higgins, 2003). Excluding outlier studies neither

completely removed heterogeneity nor changed the effect sizes. We therefore assume that

there are other potential variables that explain the variance among the qualified studies, for

example, sampling method, subjective versus objective ratings or the specific games

considered as being violent. Consequently, some caution is necessary when interpreting our

results. Our data could have subgroups of effect sizes because of additional variables, for
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 20

example, age, publication year, measures of VVG play and aggression measures, or author

subgroups.

A limitation which should be considered is the use of zero-order correlations between

VVG and aggression and its bidirectional path compared to partial effects, which control for

initial aggression. Arguably, the relevant outcome variable (aggression) could already have

influenced the amount of VVG play at the first measurement point in case of zero-order

correlations (Ferguson, 2015). That is, more aggressive individuals might choose to play more

VVG. Because of this bidirectional relationship (influence or causality), aggression at T1 is

often statistically controlled in longitudinal studies (e.g., see Bucolo, 2010; Ferguson, 2011).

This procedure might more accurately represent the specific influence of VVG on aggression

and provide a more conservative effect size. On the other hand, controlling for initial

aggression means ignoring the potential effects that VVG consumption prior to T1 might have

had on aggression. Therefore, we decided to present both full and partial results, and we

consider the zero-order correlations as an upper bound estimate of the “true” effect size and

the controlled effects as a conservative lower bound estimate. In the latter case, there is a high

plausibility that the value can be interpreted causally.

Notably, our meta-analysis was based on longitudinal rather than experimental

designs. Experiments on VVG would be better suited to directly investigate causality, but

they mostly examine short-term effects on aggression and instead, we wanted to concentrate

on the effects of sustained game play over a longer period of time.

Another point of discussion is the age effect. The effect was descriptively present in

both analyses and reached significance in the second case, although at a later age than we

expected. Possibly, as VVG play and aggression peak in early adolescence, the relationship

between the two remains the same. Although VVG might not have a more pronounced impact

on young adolescents’ aggression compared to children or adults, this age group might still

display more of both behaviors. To investigate this assumption, one might compare
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 21

aggression levels while controlling for the amount of VVG play between several studies with

different mean sample ages. In our theoretical considerations, we implied a relationship

between aggression and risk-taking behavior, the latter of which has been observed to

increase in early adolescence (Steinberg, 2010). However, this increase in risk-taking might

not result in increased aggression but in some other form of behavior, for example, the

consumption of alcohol or drugs or more physical activity. Including gender as a moderating

variable might also further increase our understanding of these age effects, since

developmental trajectories might differ.

Finally, the distribution of effect sizes from the studies that we selected poses another

limitation. In the studies with mean sample ages between 10 and 14 years, the effect sizes

vary considerably, for example, ranging from d = -.18 to d = .85 for the age of 13. This

variability makes an accurate prediction of fitted values in this age period more difficult,

which in turn limits the validity of the polynomial regression model. Figure 3 further

demonstrates the broad confidence intervals around our results, indicating that the true effects

could be distributed differently than in our specific sample of studies. These observations can

be explained by the density of available data points differing between age groups, with n = 19

samples in the younger age group, and only n = 8 samples in the older group. Only three

studies investigated the longitudinal effects of VVG on aggression in adults aged 18 or older.

Practical Implications

Given the statistical evidence of a curvilinear age effect, we want to discuss practical

consequences such a finding might have. Rating systems, for example, PEGI or ESRB,

already provide a form of intervention or prevention for children and adolescents because

they present an attempt to guide usage behavior and parenting. In the light of the age effect,

emphasis should be placed on how the ratings could be enforced in a more efficient way, for

example, discouraging younger players from obtaining and playing age-inappropriate games,

and whether the ratings (which are usually based on both violent and sexual content) properly
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 22

reflect the actual violence depicted. At the same time, the decrease in effects on older

adolescents (indicating less susceptibility to VVG effects than younger adolescents) justifies

relaxing usage control after the age of 16. For this reason, we explicitly do not advocate a "No

Never" policy and would – especially in pedagogical situations and in conflicts between

parents and their children – advocate offering adolescents the prospect of a later time of use

when they are older.

Conclusion

Our meta-analysis revealed a significant and meaningful positive effect of VVG on

subsequent physically aggressive behavior, which supports findings from prior research.

Effects increased from childhood progressively into early adolescence, reaching a peak

between the ages of 14 and 16, then steadily decreasing into adulthood. Our work might be a

starting point for future research and for an improvement of age rating criteria and

implementation. We hope more research on the effect of VVG (and other types of video

games as well) will lead to a better understanding of their influence on behavior and allow us

to better implement interventions to counter potential adverse effects. A concise

understanding of the effects will also lead to a more informed and nuanced discourse and thus

avoid stigmatizing gamers or creating a distorted perspective on video games in general. With

more knowledge, maybe the debates at the dinner table, in the news after yet another school

shooting, and among academics can become less heated and more constructive.
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 23

References

Anderson, C. A., & Bushman, B. J. (2002). Human Aggression. Annual Review of

Psychology, 53(1), 27–51. https://doi.org/10.1146/annurev.psych.53.100901.135231

Anderson, C. A., & Bushman, B. J. (2018). Media violence and the General Aggression

Model. Journal of Social Issues, 74(2), 386–413. https://doi.org/10.1111/josi.12275

Anderson, C. A., Shibuya, A., Ihori, N., Swing, E. L., Bushman, B. J., Sakamoto, A.,

Rothstein, H. R., & Saleem, M. (2010). Violent video game effects on aggression,

empathy, and prosocial behavior in Eastern and Western countries: A meta-analytic

review. Psychological Bulletin, 136(2), 151–173. https://doi.org/10.1037/a0018251

Baujat, B., Mahé, C., Pignon, J.-P., & Hill, C. (2002). A graphical method for exploring

heterogeneity in meta-analyses: Application to a meta-analysis of 65 trials. Statistics

in Medicine, 21, 2641–2652. https://doi.org/10.1002/sim.1221

Blakemore, S.-J., Burnett, S., & Dahl, R. E. (2010). The role of puberty in the developing

adolescent brain. Human Brain Mapping, 31(6), 926–933.

https://doi.org/10.1002/hbm.21052

Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009). Introduction to

meta-analysis. John Wiley & Sons.

Bucolo, D. (2010). Violent video game exposure and physical aggression in adolescence:

Tests of the general aggression model (Doctoral Dissertation). University of New

Hampshire Scholars Repository. https://scholars.unh.edu/dissertation/583

Bushman, B. J. (1998). Priming effects of media violence on the accessibility of aggressive

constructs in memory. Personality and Social Psychology Bulletin, 24(5), 537–545.

https://doi.org/10.1177/0146167298245009

Bushman, B. J., & Huesmann, L. R. (2014). Twenty-five years of research on violence in

digital games and aggression revisited: A reply to Elson and Ferguson (2013).

European Psychologist, 19(1), 47–55. https://doi.org/10.1027/1016-9040/a000164


AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 24

Bushman, B. J., Rothstein, H. R., & Anderson, C. A. (2010). Much ado about something:

Violent video game effects and a school of red herring: Reply to Ferguson and Kilburn

(2010). Psychological Bulletin, 136(2), 182–187. https://doi.org/10.1037/a0018718

Carnagey, N. L., Anderson, C. A., & Bushman, B. J. (2007). The effect of video game

violence on physiological desensitization to real-life violence. Journal of

Experimental Social Psychology, 43(3), 489–496.

https://doi.org/10.1016/j.jesp.2006.05.003

Carré, J. M., Ruddick, E. L., Moreau, B. J. P., & Bird, B. M. (2017). Testosterone and human

aggression. In P. Sturmey, The Wiley Handbook of Violence and Aggression (pp. 1–

14). John Wiley & Sons. https://doi.org/10.1002/9781119057574.whbva020

Crick, N. R. (1995). Relational aggression: The role of intent attributions, feelings of distress,

and provocation type. Development and Psychopathology, 7(2), 313–322.

https://doi.org/10.1017/S0954579400006520

Cuijpers, P. (2016). Meta-analysis in mental health: A practical guide. Pim Cuijpers

Uitgeverij. Retrieved from https://indd.adobe.com/view/5fc8f9a0-bf1e-49d3-bf5f-

a40bfe5409e0

DePasque, S., & Galván, A. (2017). Frontostriatal development and probabilistic

reinforcement learning during adolescence. Neurobiology of Learning and Memory,

143, 1–7. https://doi.org/10.1016/j.nlm.2017.04.009

Drummond, A., Sauer, J. D., & Ferguson, C. J. (2020). Do longitudinal studies support long-

term relationships between aggressive game play and youth aggressive behaviour? A

meta-analytic examination. Royal Society Open Science, 7(7), 200373.

https://doi.org/10.1098/rsos.200373

Egger, M., Smith, G. D., Schneider, M., & Minder, C. (1997). Bias in meta-analysis detected

by a simple, graphical test. BMJ, 315(7109), 629–634.

https://doi.org/10.1136/bmj.315.7109.629
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 25

Elson, M., & Ferguson, C. J. (2014). Twenty-five years of research on violence in digital

games and aggression: Empirical evidence, perspectives, and a debate gone astray.

European Psychologist, 19(1), 33–46. https://doi.org/10.1027/1016-9040/a000147

Ferguson, C. J. (2015). Do angry birds make for angry children? A meta-analysis of video

game influences on children’s and adolescents’ aggression, mental Health, prosocial

behavior, and academic performance. Perspectives on Psychological Science, 10(5),

646–666. https://doi.org/10.1177/1745691615592234

Ferguson, C. J., & Kilburn, J. (2010). Much ado about nothing: The misestimation and

overinterpretation of violent video game effects in Eastern and Western nations:

Comment on Anderson et al. (2010). Psychological Bulletin, 136(2), 174–178.

https://doi.org/10.1037/a0018566

Ferguson, C. J. (2011). Video games and youth violence: A prospective analysis in

adolescents. Journal of Youth and Adolescence, 40(4), 377–391.

https://doi.org/10.1007/s10964-010-9610-x

Ferguson, C. J., Garza, A., Jerabeck, J., Ramos, R., & Galindo, M. (2013). Not worth the fuss

after all? Cross-sectional and prospective data on violent video game influences on

aggression, visuospatial cognition and mathematics ability in a sample of youth.

Journal of Youth and Adolescence, 42(1), 109–122. https://doi.org/10.1007/s10964-

012-9803-6

Gentile, D. A., & Bushman, B. J. (2012). Reassessing media violence effects using a risk and

resilience approach to understanding aggression. Psychology of Popular Media

Culture, 1(3), 138–151. https://doi.org/10.1037/a0028481

Greenberg, B. S., Sherry, J., Lachlan, K., Lucas, K., & Holmstrom, A. (2010). Orientations to

video games among gender and age groups. Simulation & Gaming, 41(2), 238–259.

https://doi.org/10.1177/1046878108319930
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 26

Greitemeyer, T., & Mügge, D. O. (2014). Video games do affect social outcomes: A meta-

analytic review of the effects of violent and prosocial video game play. Personality

and Social Psychology Bulletin, 40(5), 578–589.

https://doi.org/10.1177/0146167213520459

Greitemeyer, T. (2019). The contagious impact of playing violent video games on aggression:

Longitudinal evidence. Aggressive Behavior, 45(6), 635–642.

https://doi.org/10.1002/ab.21857

Greitemeyer, T., & Sagioglou, C. (2017). The longitudinal relationship between everyday

sadism and the amount of violent video game play. Personality and Individual

Differences, 104, 238–242. https://doi.org/10.1016/j.paid.2016.08.021

Griffiths, M. D., Davies, M. N. O., & Chappell, D. (2004). Online computer gaming: A

comparison of adolescent and adult gamers. Journal of Adolescence, 27(1), 87–96.

https://doi.org/10.1016/j.adolescence.2003.10.007

Harrer, M., Cuijpers, P., Furukawa, T., & Ebert, D. D. (2019). dmetar: Companion R package

for the guide ‘Doing Meta-Analysis in R’ (Version 0.0.9000).

http://dmetar.protectlab.org

Higgins, J. P. T. (2003). Measuring inconsistency in meta-analyses. BMJ, 327(7414), 557–

560. https://doi.org/10.1136/bmj.327.7414.557

Hilgard, J., Engelhardt, C. R., & Rouder, J. N. (2017). Overstated evidence for short-term

effects of violent games on affect and behavior: A reanalysis of Anderson et al.

(2010). Psychological Bulletin, 143(7), 757–774. https://doi.org/10.1037/bul0000074

Huesmann, L. R. (2010). Nailing the coffin shut on doubts that violent video games stimulate

aggression: Comment on Anderson et al. (2010). Psychological Bulletin, 136(2), 179–

181. https://doi.org/10.1037/a0018567

Ivory, J. D., & Kalyanaraman, S. (2007). The Effects of Technological Advancement and

Violent Content in Video Games on Players’ Feelings of Presence, Involvement,


AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 27

Physiological Arousal, and Aggression. Journal of Communication, 57(3), 532–555.

https://doi.org/10.1111/j.1460-2466.2007.00356.x

Janacsek, K., Fiser, J., & Nemeth, D. (2012). The best time to acquire new skills: Age-related

differences in implicit sequence learning across the human life span. Developmental

Science, 15(4), 496–505. https://doi.org/10.1111/j.1467-7687.2012.01150.x.

Kirsh, S. J. (2003). The effects of violent video games on adolescents. Aggression and Violent

Behavior, 8(4), 377–389. https://doi.org/10.1016/S1359-1789(02)00056-3

Krahé, B., & Busching, R. (2015). Breaking the vicious cycle of media violence use and

aggression: A test of intervention effects over 30 months. Psychology of Violence,

5(2), 217–226. https://doi.org/10.1037/a0036627

Krahé, B., & Möller, I. (2010). Longitudinal effects of media violence on aggression and

empathy among German adolescents. Journal of Applied Developmental Psychology,

31(5), 401–409. https://doi.org/10.1016/j.appdev.2010.07.003

Krcmar, M., Farrar, K., & McGloin, R. (2011). The effects of video game realism on

attention, retention and aggressive outcomes. Computers in Human Behavior, 27(1),

432–439. https://doi.org/10.1016/j.chb.2010.09.005

Lindeman, M., Harakka, T., & Keltikangas-Järvinen, L. (1997). Age and gender differences

in adolescents’ reactions to conflict situations: Aggression, prosociality, and

withdrawal. Journal of Youth and Adolescence, 26(3), 339–351.

https://doi.org/10.1007/s10964-005-0006-2

Loeber, R., & Stouthamer-Loeber, M. (1998). Development of juvenile aggression and

violence: Some common misconceptions and controversies. American Psychologist,

53(2), 242–259. https://doi.org/10.1037/0003-066X.53.2.242

Lull, R. B., & Bushman, B. J. (2016). Immersed in violence: Presence mediates the effect of

3D violent video gameplay on angry feelings. Psychology of Popular Media Culture,

5(2), 133–144. https://doi.org/10.1037/ppm0000062


AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 28

Marshall, M. G., & Elzinga-Marshall, G. (2017). Global report 2017: Conflict, governance

and state fragility. Center for Systemic Peace. Retrieved from

http://www.systemicpeace.org/globalreport.html

McGloin, R., Farrar, K., & Krcmar, M. (2013). Video Games, Immersion, and Cognitive

Aggression: Does the Controller Matter? Media Psychology, 16(1), 65–87.

https://doi.org/10.1080/15213269.2012.752428

MPFS (2017). Jugend, Information, (Multi-) Media [Youth, Information (Multi-)Media].

Retrieved from https://www.mpfs.de/fileadmin/files/Studien/JIM/2017/JIM_2017.pdf

Möller, I., & Krahé, B. (2009). Exposure to violent video games and aggression in German

adolescents: A longitudinal analysis. Aggressive Behavior, 35(1), 75–89.

https://doi.org/10.1002/ab.20290

Peters, S., & Crone, E. A. (2017). Increased striatal activity in adolescence benefits learning.

Nature Communications, 8(1):1983. https://doi.org/10.1038/s41467-017-02174-z

Prescott, A. T., Sargent, J. D., & Hull, J. G. (2018). Metaanalysis of the relationship between

violent video game play and physical aggression over time. Proceedings of the

National Academy of Sciences, 115(40), 9882–9888.

https://doi.org/10.1073/pnas.1611617114

Quandt, T., Festl, R., Breuer, J., & Scharkow, M. (2013). Digitale Spiele: Stabile Nutzung in

einem dynamischen Markt. [Digital games: Stable patterns of use in a dynamic

market]. Media Perspektiven, 10, 483–492.

R Core Team (2019). R: A language and environment for statistical computing. R Foundation

for Statistical Computing, Vienna, Austria. https://www.R-project.org/

Revelle, W. (2018). psych: Procedures for Personality and Psychological Research.

Northwestern University, Evanston, Illinois, USA. https://CRAN.R-

project.org/package=psych
AGE EFFECTS, VIOLENT VIDEO GAMES, AND AGGRESSION 29

Richard, F. D., Bond, C. F., & Stokes-Zoota, J. J. (2003). One Hundred Years of Social

Psychology Quantitatively Described. Review of General Psychology, 7(4), 331–363.

https://doi.org/10.1037/1089-2680.7.4.331

Rothstein, H. R., & Bushman, B. J. (2015). Methodological and reporting errors in meta-

analytic reviews make other meta-analysts qngry: A commentary on Ferguson (2015).

Perspectives on Psychological Science, 10(5), 677–679.

https://doi.org/10.1177/1745691615592235

Schwarzer, G. (2020). meta: General Package for Meta-Analysis (4.15-1). https://CRAN.R-

project.org/package=meta

Steinberg, L. (2010). A dual systems model of adolescent risk-taking. Developmental

Psychobiology, 52(3), 216–224. https://doi.org/10.1002/dev.20445

Viechtbauer, W. (2010). Conducting meta-analyses in R with the metafor package. Journal of

Statistical Software, 36(3), 1-48. URL: http://www.jstatsoft.org/v36/i03/

Viechtbauer, W., & Cheung, M. W. -L. (2010). Outlier and influence diagnostics for meta-

analysis. Research Synthesis Methods, 1(2), 112–125. https://doi.org/10.1002/jrsm.11

von Salisch, M., Vogelgesang, J., Kristen, A., & Oppl, C. (2011). Preference for violent

electronic games and aggressive behavior among children: The beginning of the

downward spiral? Media Psychology, 14(3), 233–258.

https://doi.org/10.1080/15213269.2011.596468

Wickham, H. (2016). ggplot2: Elegant graphics for data analysis. Springer-Verlag New York.

https://ggplot2.tidyverse.org

Witmer, B. G., & Singer, M. J. (1998). Measuring presence in virtual environments: A

presence questionnaire. Presence: Teleoperators and Virtual Environments, 7(3),

225–240. https://doi.org/10.1162/105474698565686

You might also like