The contagious impact of playing violent video games on aggression: Longitudinal evidence

Abstract Meta‐analyses have shown that violent video game play increases aggression in the player. The present research suggests that violent video game play also affects individuals with whom the player is connected. A longitudinal study (N = 980) asked participants to report on their amount of violent video game play and level of aggression as well as how they perceive their friends and examined the association between the participant's aggression and their friends’ amount of violent video game play. As hypothesized, friends’ amount of violent video game play at Time 1 was associated with the participant's aggression at Time 2 even when controlling for the impact of the participant's aggression at Time 1. Mediation analyses showed that friends’ aggression at Time 1 accounted for the impact of friends’ amount of violent video game play at Time 1 on the participant's aggression at Time 2. These findings suggest that increased aggression in video game players has an impact on the player's social network.


| INTRODUCTION
Given its widespread use, the public and psychologists alike are concerned about the impact of violent video game play. In fact, a great number of studies have addressed the effects of exposure to violent video games (where the main goal is to harm other game characters) on aggression and aggression-related variables. Meta-analyses have shown that playing violent video games is associated with increased aggression in the player (Anderson et al., 2010;Greitemeyer & Mügge, 2014). The present longitudinal study examines the idea that violent video game play also affects the player's social network, suggesting that concern about the harmful effects of playing violent video games on a societal level is even more warranted.

| Theoretical perspective
When explaining the effects of playing violent video games, researchers often refer to the General Aggression Model (GAM) proposed by Anderson & Bushman (2002). According to this theoretical model, person and situation variables (sometimes interactively) may affect a person's internal state, consisting of cognition, affect, and arousal. This internal state then affects how events are perceived and interpreted. Based on this decision process, the person behaves more or less aggressively in a social encounter. For example, playing violent video games is assumed to increase aggressive cognition and affect, which in turn results in behavioral aggression. An extension of this model further assumes that increased aggression due to previous violent video game play may instigate an aggression escalation cycle in that the victim also behaves aggressively (cf. Anderson & Bushman, 2018, Figure 5). The present research tested key predictions derived from the GAM and its extension, that (a) violent video game play is associated with increased aggression in the player and that (b) individuals who are connected to the player will also become more aggressive.

| Effects of violent video game play on aggression
The relationship between violent video game play and aggression has been examined in studies employing cross-sectional, longitudinal, and experimental designs. Cross-sectional correlational studies typically show a positive relationship between the amount of violent video game play and aggression in real-world contexts (e.g., Gentile, Lynch, Linder, & Walsh, 2004;Krahé & Möller, 2004). Several longitudinal studies have been conducted, showing that habitual violent video game play predicts later aggression even after controlling for initial aggressiveness (e.g., Anderson, Buckley, & Carnagey, 2008). That violent video game play has a causal impact on aggression and related information processing has been demonstrated by experimental work (e.g., Anderson & Carnagey, 2009;Gabbiadini & Riva, 2018). Finally, meta-analyses corroborated that violent video game play significantly increases aggressive thoughts, hostile affect, and aggressive behavior (Anderson et al., 2010;Greitemeyer & Mügge, 2014). Some studies failed to find significant effects (e.g., McCarthy, Coley, Wagner, Zengel, & Basham, 2016). However, given that the typical effect of violent video games on aggression is not large, it is to be expected that not all studies reveal significant effects.

| The contagious effects of aggression
Abundant evidence has been collected that aggression and violence can be contagious (Dishion, & Tipsord, 2011;Huesmann, 2012;Jung, Busching, & Krahé, 2019). Indeed, the best predictor of (retaliatory) aggression is arguably previous violent victimization Goldstein, Davis, & Herman, 1975). However, even the observation of violence can lead to increased violence in the future (Widom, 1989). Overall, it is a well-known finding that aggression begets further aggression. Given that violent video game play increases aggression, it thus may well be that this increased aggression then has an impact on people with whom the player is connected.
Correlational research provides initial evidence for the idea that the level of people's aggression is indeed associated with how often their friends play violent video games (Greitemeyer, 2018). In particular, participants who did not play violent video games were more aggressive the more their friends played violent video games. However, due to the cross-sectional design, no conclusions about the direction of the effect are possible. It may be that violent video game players influence their friends (social influence), but it is also conceivable that similar people attract each other (homophily) or that there is some shared environmental factor that influences the behavior of both the players and their friends (confounding). That is, it is unclear whether indeed aggression due to playing violent video games spreads or whether the effect is reversed, such that aggressive people are prone to befriend others who are attracted to violent video game play. Moreover, it is possible that some third variable affected both, participants' reported aggression and their friends' amount of violent video game play. There is also the possibility that people are unsure about the extent to which their friends play violent video games. In this case, they may perceive their friends as behaving aggressively and then (wrongly) infer that the friends play violent video games. To disentangle these possibilities and to show that the effect of violent video game play (i.e., increased aggression in the player) indeed has an impact on the player's social network, relationships among variables have to be assessed over time while covarying prior aggression (Bond & Bushman, 2017;Christakis & Fowler, 2013). Verheijen, Burk, Stoltz, van den Berg, and Cillessen (2018) tested the idea that players of violent video games have a long-term impact on their social network. These authors found that participants' exposure to violent video games increased their friend's aggressive behavior 1 year later. However, given that the authors did not examine whether the violent video game player's increased aggression accounts for the impact on their friend's aggressive behavior, it is unknown whether violent video game play indeed instigates an aggression cycle. For example, players of violent video games may influence their friends so that these friends will also play violent video games. Any increases in aggression could then be an effect of the friends playing violent video games on their own.

| The present research
The present study examines the longitudinal association between the participant's aggression and their friends' amount of violent video game play, employing an egocentric networking approach (Stark & Krosnick, 2017). In egocentric networking analyses, participants provide self-reports but also report on how they perceive their friends. In the following, and in line with Greitemeyer (2018), the friends were treated as the players and the participant was treated as their friends' social network. Please note that ties between the participant's friends (i.e., whether friends also know each other) were not assessed (Greitemeyer, 2018;Mötteli & Dohle, 2019)

| Participants
Participants were citizens of the U.S. who took part on Amazon Mechanical Turk. Because it was unknown how many of the participants will complete both questionnaires, no power analyses were conducted a priori but a large number of participants was run.
At Time 1, there were 2,502 participants (1,376 females, 1,126 males; mean age = 35.7 years, SD = 11.8). Of these, 980 participants (522 females, 458 males; mean age = 38.9 years, SD = 12.5) completed the questionnaire at Time 2. Time 1 and Time 2 were 6 months apart. There were no data exclusions, and all participants were run before any analyses were performed. The questionnaire included some further questions (e.g., participant's perceived deprivation) that are not relevant for the present purpose and are reported elsewhere (Greitemeyer & Sagioglou, 2018). 1 Given that the questionnaire was relatively short, no attention checks were employed.

| Procedure and measures
Procedure and measures were very similar to Greitemeyer (2018), with the main difference that individuals participated at two time points (instead of one). After providing demographics, self-reported aggressive behavior was assessed. As in previous research (e.g., Krahé & Möller, 2010), participants indicated for 10 items how often they had shown the respective behavior in the past 6 months. Sample items are: "I have pushed another person" and "I have spread gossip about people I don't like" (5 items each address physical aggression and relational aggression, respectively). All items were rated on a scale from 1 (never) to 5 (very often), and scores were averaged.
Participants were then asked about their amount of violent video game play, employing one item: "How often do you play violent video games (where the goal is to harm other game characters)?" (1 = never to 7 = very often).
Afterwards, participants learned that they will be asked questions about people they feel closest to. These may be friends, coworkers, neighbors, relatives. They should answer questions for three contacts with whom they talked about important matters in the last few months. For each friend, they reported the level of aggression (αs between = 0.90 and 0.91) and the amount of violent video game play, employing the same questions as for themselves. Responses to the three friends were then averaged. Finally, participants were thanked and asked what they thought this experiment was trying to study, but none noted the hypothesis that their friend's amount of violent video game play would affect their own level of aggression. At Time 2, the same questions were employed. Reliabilities for how participants perceived the level of aggression for each friend were between 0.89 and 0.90.

| RESULTS
Descriptive statistics, intercorrelations, and internal consistencies of all measures are shown in Table 1.

| Time 1 (N = 2,502)
The relationship between the amount of violent video game play and reported aggression was significant, both for the participant and the friends. That is, violent video game play was associated with increased aggression in the player and participants perceived their friends who play more violent video games to be more aggressive than their less-playing friends. Participant's and friends' amount of violent video game play as well as their level of reported aggression, respectively, were also positively associated, indicating that participants perceived their friends to be similar to them. Most importantly, participant's aggression was significantly associated with friends' amount of violent video game play.   were assessed, the present study provides evidence for the hypothesized effect that violent video game play is associated with increased aggression in the player, which then instigates aggression in their social network. Importantly, the impact of the participant's amount of violent video game play was controlled for, indicating that the relationship between friends' amount of violent video game play and the participant's aggression is not due to the friends being similar to the participants. Moreover, the reverse effect that aggressive people will become attracted to others who play violent video games was not reliable. The present research thus documents the directional effects that violent video games is associated with increased aggression in the player and that this increased aggression then has an impact on people with whom the player is connected.
Overall, the present study provides comprehensive support for key hypotheses derived from the GAM and its extension (Anderson & Bushman, 2018). It shows that violent video game play is associated with increased aggression in the player and it documents that others who are connected to players might be also affected even when controlling for their own amount of violent video game play. To the best of my knowledge, this study is the first that shows that because violent video game players are more aggressive their friends will become aggressive, too. Previous research either employed a cross-sectional design and thus could not address the direction of the effect (Greitemeyer, 2018) or did not examine whether the effect of violent video game play (i.e., increased aggression) indeed spreads . As proposed by the GAM and its extension (Anderson & Bushman, 2018), increased aggression in violent video game players appears to instigate an aggression escalation cycle (cf. .  (Prescott, Sargent, & Hull, 2018), although the effect size was relatively small (β = 0.11) and thus single studies that produce nonsignificant results are to be expected. Importantly, in the present study, a single-item measure of violent video game play was employed. In contrast, previous research on the relationship between violent video game play and the player's aggression has often employed multi-item measurement scales that are typically more reliable and precise (for an overview, Busching et al., 2015). Hence, it may well be that due to the limitations of the single-item measure of the participant's amount of violent video game play the relationship between participants' violent game play and their aggressive behavior was artificially reduced.
Even though the longitudinal design allows ruling out a host of alternative explanations for the impact of violent video games on the player's social network, causality can only inferred by using an experimental design. Future research may thus randomly assign participants to play a violent or nonviolent video game (players) and assesses their aggression against new participants (partners). It can be expected that the partners suffer more aggression when the player had played a violent, compared to a nonviolent, video game.
Afterwards, it could be tested whether the partner of a violent video game player is more aggressive than a partner of a nonviolent video game player. Given that the partner is not exposed to any video games, firm causal conclusions could be drawn that violent video game play affects aggression in people who are connected to violent video game players. It could be also tested whether the partner of a violent video game player would not only be more likely to retaliate against the player, but also against a third party. In fact, previous research into displaced aggression has shown that people may react aggressively against a target that is innocent of any wrongdoing after they have been provoked by another person (Marcus-Newhall, Pedersen, Carlson, & Miller, 2000). It may thus well be that the effect of playing violent video games spreads in social networks and that even people who are only indirectly linked to violent video game players are affected.
An important limitation of the present egocentric network data is the reliance on the participant's perception of their social network, leaving the possibility that participants did not accurately perceive their friends. It is noteworthy that participants perceived their friends to be highly similar to them. In this regard, it is important to keep in mind that participants always provided self-ratings first, population than are college student samples (Paolacci & Chandler, 2014) and the pool of participants is geographically diverse. Moreover, MTurk participants appear to be more attentive to survey instructions than are undergraduate students (Hauser & Schwarz, 2016). Nevertheless, future research on the impact of violent video game play on the player's social network that employs other samples would improve the generalizability of the present findings.
In conclusion, violent video game play is not only associated with increased aggression in the player but also in the player's social network. In fact, increased aggression due to violent video game play appears to instigate further aggression in the player's social network.
This study thus provides suggestive evidence that not only players of violent video games are more aggressive, but also individuals become more aggressive who do not play violent video games themselves but are connected to others who do play.
2 Given that the measures of violent video game exposure and aggressive behavior violated the normal distribution, Spearman's ρ coefficients were also calculated. However, the pattern of finding was very similar (e.g., the crucial relationship between the participant's aggression and friends' amount of violent video game play was 0.18 [Pearson] and 0.17 [Spearman]). All these analyses can be obtained from the author upon request.