The social contagious effects of violent video game play: retaliatory, displaced, and the mere observation of aggression

ABSTRACT Aggression and violence have been shown to spread among connected individuals. The present experimental study aims to test whether exposure to violent video games increases aggression levels in players and whether this increased aggression spreads among their social network. Results showed that aggression after video game play is indeed contagious and that not only victims of the player’s aggression retaliate or are aggressive toward a neutral person (i.e., displaced aggression), but that even mere observers of the player’s aggression are aggressive toward a neutral person. Overall, our findings suggest that violent video game exposure may not only have an impact on the player’s level of aggression, but also on people who are connected to the player.

Since its beginning, the video game sector has been an ever-growing industry, with a population of players poised to reach 3 billion by 2023 (Clement, 2021).With global revenue surpassing that of the film industry (Witkoski, 2020), it comes to no surprise that researchers have taken an interest in the impact video games can have on us.Whilst there are many benefits, from a learning, health, or social stand point, that can come from playing (Granic et al., 2014;Greitemeyer, 2022b), much of the focus of this work has been turned to the negative impact of gameplay, especially in regards to the link between violent video game exposure (VVE) and aggression (Greitemeyer, 2022a).
Several meta-analyses documented that VVE leads to an increase in aggression (Anderson et al., 2010;Greitemeyer & Mügge, 2014) and that the effect is reliable in cross-sectional, longitudinal, and experimental work.Corroborating these previous meta-analyses, Prescott et al. (2018) found that, across 24 longitudinal studies measuring overt aggressive behaviors, VVE effects on behavioral aggression were significant, albeit of a relatively small effect size (~.11).It is important to note that some studies did not find a significant link between VVE and aggression (e.g., Kühn et al., 2018;McCarthy et al., 2016).However, given that the magnitude of the relationship between VVE and aggression is not large, it is not surprising that some studies do not find significant effects.

VVE and social networks
The aim of the present study is to examine whether VVE has not only an impact on players, but also on their social contacts.Previous research has shown that various psychological constructs, such as attitudes, emotions, and behaviors, can spread across social network ties.Indeed, an array of psychological phenomena have been demonstrated to spread in social networks, ranging from smoking (Christakis & Fowler, 2008) and health activities (Sepulvado et al., 2022) to voting behavior (Nickerson, 2008), student engagement in school (Mendoza & King, 2020), and even happiness (Bliss et al., 2012).Likewise, the spread of aggression is a well-known pattern (Herrmann et al., 2023;Jung et al., 2019).Some researchers (Bond & Bushman, 2017) have even compared aggression to a contagious disease.According to such a social contagion process, individuals can 'catch' certain behaviors or emotions from others through direct or indirect social interactions.For instance, if one person in a group becomes visibly angry or aggressive, it can trigger a chain reaction, causing others in the group to respond in a similar manner.As VVE should increase aggression, it is plausible that VVE can also indirectly affect the player's social network.Specifically, engaging in violent video games may contribute to increased aggression in the player, and this heightened aggression might subsequently lead to higher levels of aggression in people who are connected to the player.
Initial evidence for this idea was provided in a large cross-sectional study that found that individuals whose friends played violent video games were more aggressive, especially if they did not play violent games themselves (Greitemeyer, 2018).In a longitudinal study, Verheijen et al. (2018) assessed adolescent best friends' dyads' aggression after one year and found that male participants' VVE predicted their friends' aggression over time.Similarly, Greitemeyer (2019) showed that participants' friends' aggression level was a mediator between the friends' VVE and participants' aggression one year later.Delhove and Greitemeyer (2019) examined whether VVE would not only increase retaliatory aggression (against the player), but also aggression toward an innocent neutral target (i.e., displaced aggression; cf.Marcus-Newhall et al., 2000).Although the results were mixed (one study having failed to show VVE's effect on aggression), the authors found that VVE had a causal effect on both retaliatory and displaced aggression.
Even more recently, researchers have used longitudinal social network analysis to assess this effect.Verheijen et al. (2021) found that adolescents tend to befriend others similar to them in VVE and aggression and that they have a tendency to become more alike in terms of aggression (but not VVE) over time.However, they did not find support for the contagion hypothesis, as friends' VVE did not influence one's own aggressive behavior.Likewise, Delhove and Greitemeyer (2021) did not find a contagion effect, and contrary to Verheijen et al. (2021), they did not find signs of homophily either.Given these mixed findings, we deemed it important to provide another test of the idea that VVE related aggression spreads across social networks.
Our study also extends previous work in an important way.As noted above, it is a well-known finding that aggression spreads across a social network.It is also noteworthy that not only victims of aggression respond with retaliatory aggression, but even the mere observation of violence may increase aggression in subsequent encounters (Huesmann & Kirwil, 2007;Widom, 1989).That is, it is conceivable that simply observing that a violent video game player is aggressive toward someone else may suffice to instigate aggression in the neutral observer.Hence, we aimed to examine whether the effect of VVE (i.e., increased aggression in the player) even influences people who only observed the player's aggressive behavior.
Overall, we tested whether increased aggression after VVE would spread, in that victims of the player's aggression retaliate (i.e., are aggressive toward the player) and are aggressive toward a neutral person (i.e., displaced aggression) and that even mere observers of the player's aggression are aggressive toward a neutral person.If the latter is indeed the case, societal concern about the harmful effects of VVE is even more warranted.

The present research
We aimed at testing the spread of VVE-related aggression in an experimental setting.Our research was divided in two parts: first, participants were asked to play either a violent or a nonviolent video game and took part in a Cold-Water Task in order to measure their post-game aggression (see below).In a second laboratory session, a separate group of participants were subjected to the aggression of the Part 1 participants (or observed that a confederate was exposed to the aggression).Our hypotheses were as follows: H1 (tested in Part 1): Participants who played violent video games show higher levels of aggression than those who played nonviolent video games.H2 (tested in Part 2): Exposure to heightened aggression is expected to increase participants' levels of aggression, whether directed toward the aggressor (retaliation) or toward unrelated third parties (displaced aggression), and when they are merely witnesses of the aggressive behavior performed by the players.
The study received ethical approval from the Review Board for Ethical Questions of our university.The data are openly accessible at: https://osf.io/n9yqm/?view_only= 9e15e908815842b1bc59fbaba32ee158.

Methods
In our previous research on the contagious effects of VVE (Delhove & Greitemeyer, 2019), two-hundred video game players participated (an a priori power analysis determined that a sample of 187 participants was required to achieve a power of .80 given an expected effect size of r = .18and an α of .05).We aimed for a similar number of participants.In the end, one hundred ninety individuals (48% male, mean age = 22.95, SD = 4.62) participated for a monetary reward (8€).In order to hide our hypotheses, participants were told that we were pretesting materials for future studies interested in video game enjoyment.
First, participants were asked to complete an online questionnaire.Upon completion of this questionnaire, participants received a code they needed to bring to the laboratory for the gameplay part of the experiment, in order to pair their questionnaire data with their lab data.This questionnaire consisted of demographic questions followed by the Interpersonal Reactivity Index (IRI; Davis, 1983) and the Buss-Perry Aggression Questionnaire (BPAQ, Buss & Perry, 1992).
The IRI was used as a measure of empathy.The scale consists of 28 items.Each item is a statement, such as 'I often have tender, concerned feelings for people less fortunate than me.' that one needs to rate using a 5-point Likert scale ranging from 1 -'Does not describe me well' to 5 -'Describes me very well'.The scale proved reliable with a Cronbach's α = .84.
The BPAQ contains 29 items measuring trait aggression.Each item is a statement, such as 'Given enough provocation, I may hit another person.',rated with a 5-point Likert scale ranging from 1 -'Extremely uncharacteristic of me' to 5 -'Extremely characteristic of me'.With Cronbach's α = .82,this scale also proved reliable.
An attention check item was added to both scales.Participants who failed those checks were not invited to take part in the rest of the study.Once they had completed the questionnaire, participants were asked to take an appointment for the rest of the experiment.They were randomly assigned to play one of six different video games for 30 minutes.The games were selected to represent one of two genres (racing or shooter) and one of three violent content levels (nonviolent, mildly violent, or highly violent).To increase ecological validity and to allow more general conclusions, we employed two different game genres.However, we only expected the player's aggression to vary between the different violence levels, but not between the different genres.
All games were played on desktop computers using the game platform Steam.The three racing games, from least to most violent, were TrackMania Nations Forever, FlatOut 2, and Carmageddon: Max Damage.In TrackMania, the player is invited to complete race tracks as fast as possible, earning medals if they can beat a given time.FlatOut consists of races against other cars in which one can gain advantages by destroying the other racers' cars.Carmageddon is a racing game that rewards the player for destroying the opponents' cars and for running over pedestrians in effusions of blood.
We selected Portal as the nonviolent shooter game, Star Wars Battlefront II as the mildly violent one, and Call of Duty 4: Modern Warfare as the highly violent one.Portal is a puzzle-platform game where the players use a 'portal gun' to shoot teleportation doorways in order to solve increasingly difficult puzzles.Battlefront II is a third/first-person shooter set in the Star Wars universe.Players fight as one of the available factions, killing their enemies with lasers.Call of Duty 4 is a firstperson shooter game set in contemporary time in which players use realistic weapons to kill their foes.
After playing, participants were asked to rate how difficult, fun, frustrating, exciting, fast-paced, aggressive, realistic, and competitive the games were, as well as how well one could identify with the game character and how intuitive the controls were (using a 7-point Likert-scale ranging from 1 -'Not at all' to 7 -'Very').Following these game ratings, participants were asked to share the thoughts they had during the game session.
Finally, aggressive behavior was assessed by employing the Cold-Water Task (CWT) paradigm.This task has been successfully used in aggression research in the past (e.g., Pedersen et al., 2014;Vasquez et al., 2013).Participants were told to assign the duration of an unpleasant stimulus (i.e., keeping one's hand in painfully cold water) for a future participant, supposedly as a manipulation of attention.In order to give them a good understanding of how unpleasant the task was, participants were asked to keep their hand in the water themselves for ten seconds.The water was kept at 7°C, which is uncomfortable and can even be painful (Mitchell et al., 2004).Using a 7-point Likert scale, participants indicated how cold, distracting, unpleasant, and painful they felt the task was.These four items were combined into one scale assessing task perception (Cronbach's α = .78).Then, participants had to choose the time (between 0 and 60 seconds in 5 seconds intervals) that a future participant would have to keep their hand in the water.The scores were registered using a 13-point scale (1 = 0 second, 13 = 60 seconds).Finally, participants were briefed, paid, and thanked for their participation.

Results
First, we tested for differences between groups for IRI, BPAQ, and task perception.None of the tests proved significant, with all Fs(5, 184) < 1, ps > .05.Hence, the participants in each group did not differ from the others in regards to empathy, trait aggression, or how unpleasant the cold water was perceived to be, suggesting that random assignment was successful and that the perception of the cold water could not account for any differences between the experimental conditions on the assigned CWT duration.
We then compared all games on the different evaluations (Table 1).Most importantly, the highly violent video games were perceived as being more violent than the mildly violent video games, with the nonviolent video games perceived to be least violent.Hence, the manipulation check was successful.
A two (game type: racing vs shooter) by three (violent content: nonviolent vs mild vs high) analysis of variance (ANOVA) was conducted on the assigned CWT duration.Means and standard deviations for each experimental condition are reported in Table 2. Contrary to Hypothesis 1, the main effect of violent content was not significant, F(2, 184) = 1.31, p = .272,η 2 = .01.The interaction between game type and violent content was also not significant, F(2, 184) = 1.74, p = .178,η 2 = .02.
To test Hypothesis 2, we conducted a second session with a new group of participants.Our initial plan was to pair each participant from the first session with a new participant.Please recall that Hypothesis 2 aimed to examine whether heightened aggression following violent video game play would spread.That is, VVE can only have an impact on the player's social contacts if the players are more aggressive than players of nonviolent video games.Given that there was no significant main effect of violent content, the prerequisite for testing Hypothesis 2 was not met.However, a comparison of the mean CWT responses of the six experimental conditions revealed that participants playing the mildly violent shooter game (Battlefront) assigned significantly longer CWT duration (M = 5.13, SD = 2.36) than did participants playing the nonviolent shooter game (Portal, M = 3.66, SD = 1.15), t(60) = 3.16, p = .002.Hence, for Part 2, we chose to use only the participant's CWT responses from the Battlefront and Portal groups.Please note that there were several significant differences between the games on dimensions other than violent content (Table 1).However, Portal and Battlefront II did not significantly differ on any measure besides their aggressive rating.Hence, differences in aggression are likely due to differences in violent content between the two games.

Part 2
Our study investigated whether players of the violent video game Battlefront, who exhibited higher levels of aggression than players of the nonviolent game Portal, would cause greater aggression in three groups of participants: those who were directly victimized by the player and chose a punishment time for the player (retaliation condition), those who were victimized but chose a punishment time for a non-player (displaced aggression condition), and those who merely observed the player's aggressive behavior and chose a punishment time for a non-player (observation condition).

Methods
One hundred eighty-six volunteers (51% male, mean age = 22.35, SD = 4.29) participated in exchange for course credit or a monetary reward (3€).None of these individuals participated in Part 1.Each participant was randomly assigned to one of three conditions (retaliation, displaced aggression, or observation) and paired with one of the participants from Part 1.To increase statistical power and as we employed only two of the six experimental conditions from Part 1, each of the CWT score from the 62 selected participants from Part 1 was used three times, once for each of the three conditions.For example, if a Battlefront player in Part 1 chose 20 seconds for a future participant to keep their hand in the water, one participant in the retaliation condition and one participant in the displaced aggression condition had to keep the hand in the water for 20 seconds.In the observation condition, participants learned about the 20 seconds The means represent converted values.To get the actual duration in seconds, one must multiply these numbers by 5 and then subtract 5.
decision and observed another person keeping their hand in the water for this duration.All participants were informed of the duration allocated to them.In order to hide our hypothesis, participants were led to believe that we were studying the impact of physical distraction on memory capacity.Upon arrival at the lab, participants were asked to take place in front of a computer and to fill in a questionnaire.This questionnaire consisted of the same questions as that of Part 1 (i.e., demographics, the IRI, and the BPAQ).
Participants were then presented with the CWT, supposedly used as a distraction task.In the displaced aggression condition, participants were told that the time they had to keep their hand in the cold water was decided by a previous participant and we asked them to select the time to assign to the next one.
In the retaliation condition, a confederate was present at the start of the study but was then in a different room.Participants were told that they were selected to go first on the distraction task and that the confederate had chosen the time they had to keep their hand in the cold water.They were then asked to select the time the confederate had to keep the hand in the cold water.
In the observation condition, a confederate was also present but sat in the same room as the participant.Both were asked to draw one of two cards in order to decide who would be doing the task and who would be in a 'control' group.Unbeknownst to the participant, the confederate was always chosen to do the task.Before the confederate completed the task, the participant was asked to test the water with their own hand to realize how cold it was.They were then asked to choose the duration of the task for the next session.
After the CWT, participants were presented with a short bogus memory task, after which they were briefed, rewarded, and thanked for their participation.

Results
We ran a one-way ANOVA to look for differences between groups for IRI and BPAQ scores.Neither of the tests proved significant, with both Fs(5, 180) < 1.20, ps > .05,suggesting than random assignment was successful.We then examined whether participants' choice of CWT duration was influenced by the duration they had to keep their hand in the cold water.Indeed, there was a significant positive correlation, r(186) = .32,p < .001.
To test Hypothesis 2, a two (violent content: violent vs. nonviolent) by three (CWTcondition: displaced aggression vs retaliation vs observation) ANOVA was conducted on the assigned CWT duration (Table 3).The main effect of violent content was marginally significant, F(1, 180) = 3.76, p = .054,η 2 = .02.Participants in Part 2 paired with one of The means represent converted values.To get the actual duration in seconds, one must multiply these numbers by 5 and then subtract 5.
the participants that played the violent video game Battlefront in Part 1 tended to choose longer CWT duration (M = 4.53, SD = 2.69) than did participants paired with one of the participants that played the nonviolent video game Portal in Part 1 (M = 3.85, SD = 2.01).

Discussion
The present work aimed at further assessing the contagion effect of violent video game related aggression highlighted in recent works (e.g., Greitemeyer, 2018;Verheijen et al., 2021).Across all conditions, CWT responses from Part 1 and 2 were positively correlated.That is, both participants who had to keep their hand in cold water (in the retaliation condition and in the displaced aggression condition) and participants who observed the confederate keeping her hand in the water (in the observation condition) were more aggressive, the longer CWT times were assigned in Part 1.
Most importantly, participants paired with participants who had played a violent video game tended to exhibit higher levels of aggression than those paired with participants who had played a nonviolent game.Please note, however, that while the correlational effect was significant, the experimental effect was only marginally significant.A sensitivity analysis concerning the main effect of violent content (violent vs. nonviolent) showed that, given a sample size of 186, the study had 80% statistical power to detect an effect of d ≥0.37 (one-tailed), corresponding to small to medium sized (and larger) effects.Notably, the phenomenon of aggression contagion, where an individual's aggressive behavior triggers a similar response in those around them, is firmly established and aligns with classical models such as frustration-aggression (Dollard et al., 1939) and more contemporary cognitive neoassociation models (Berkowitz, 1989), which propose that aggression tends to escalate.Overall, our study contributes additional evidence to support the idea that aggression leads to further aggression.
The most notable finding of this study is that even mere observers of the player's aggression were more aggressive toward a neutral person.This suggests that exposure to violent video games can have a broader impact than previously thought and that it can contribute to a cycle of aggression and retaliation.Indeed, our findings are consistent with previous research demonstrating that previous violent victimization can lead to retaliatory aggression (Anderson et al., 2008), while also highlighting the effect of mere exposure to violence on increasing future aggression (Widom, 1989).
Concerning the effect of playing violent video games on aggression in the player, we expected more aggression from the player as the content becomes more violent.Surprisingly, the only effect we could find was between the nonviolent and mildly violent shooter.Some participants from the highly violent game groups highlighted in their comments that they were put off by the excessive brutality, which could have played a role in the absence of the expected effect for highly violent games.Future work may try to replicate and test possible explanations for a reduced effect of overly violent games on aggression in the player.
Notably, there was a difference between the CWT scores for both nonviolent games, with TrackMania players exhibiting higher levels of aggression when compared to Portal players (Welch's t(44.3)= 2.75, p < .01).Interestingly, participants rated TrackMania as more competitive and faster-paced, and both competitiveness (Adachi & Willoughby, 2011) and pace of action (Elson et al., 2015) have been shown to lead to increased aggression in the player.Surely, not only the violence level of a video game, but also other game characteristics contribute to whether playing a video game makes the player more aggressive.
Overall, the results of this experimental study suggest that aggression is contagious and that exposure to violent video games can have a profound impact not only on the player's level of aggression but also on those who are connected to them.It is important to keep in mind, however, that the overall effect of exposure to violent video games on aggression levels in players was not significant in Part 1.Previous meta-analyses have shown that VVE is positively related to increased aggression in the player (Anderson et al., 2010;Greitemeyer & Mügge, 2014;Prescott et al., 2018).However, as the effect of VVE on aggression is not large in its magnitude, there will also be studies (as our overall analysis in Part 1) that fail to find significant effects.
In this regard, it is important to note that we measured aggression toward a noncompeting, non-provoking other person.This is important insofar as previous research has predominantly examined whether VVE leads to aggression when there is some provocation either by the experimental set-up or by the partner.For example, in studies using the Competitive Reaction Time Task (e.g., Gabbiadini et al., 2022), arguably the task most commonly used to study whether VVE increases aggressive behavior, participants compete with an opponent who could also aggress toward them.Hence, it is conceivable that the absence of any provocation accounts for the relatively weak effect of the violent video game manipulation on the CWT assignment.

Limitations
While the findings of this study provide important insights into the impact of violent video games on aggression levels and the potential spread of aggression through social networks, there are several limitations to consider.One limitation of the study is that it was conducted in a laboratory setting, which may not accurately reflect real-world behavior.Participants in the study were also aware that they were being observed and may have modified their behavior accordingly, leading to potential demand characteristics.Furthermore, the study only examined a single outcome measure of aggression, and other forms of aggression, such as verbal aggression, were not assessed.
Another limitation is that the study only examined short-term effects of violent video game exposure.It is unclear whether these effects would persist over a more extended period or whether aggression would continue to spread beyond the immediate social network.In addition, the study did not examine the potential mitigating factors that may prevent the spread of aggression through social networks.Future research could explore potential interventions or protective factors that may reduce the spread of aggression.
Finally, while we expanded the literature by assessing the potential effect of mere observation of video-game induced aggression, we only assessed observers' aggression levels toward third parties, but not toward the perpetrator of the aggression.Testing whether individuals are more aggressive when given the opportunity to avenge the victim would be an interesting avenue for future research.
Despite these limitations, this study contributes to the abundant body of research on the impact of violent video games on aggressive behavior and highlights the importance of understanding the potential spread of aggression through social networks.

Conclusion
VVE-related aggression appears to be contagious.Advancing previous work, we found that increased aggression after VVE did not only lead to more pronounced retaliation and displaced aggression, but even the mere observation of aggression increased subsequent aggression.Overall, our study provided evidence for the notion that aggression begets further aggression.In contrast, less conclusive were the findings concerning the relationship between VVE and aggression.It thus appears that VVE has a relatively small impact on the player's later aggression.However, if a player becomes more aggressive, it is likely that the player's social network will become more aggressive as well.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
The work was supported by the Austrian Science Fund [P28913].

Table 1 .
Results from a one-way ANOVAs comparing games' evaluation, with means and standard deviation and Tukey Contrast (Part 1).

Table 2 .
Means and standard deviations (in parentheses) for assigned CWT duration (Part 1).

Table 3 .
Means and standard deviations (in parentheses) for assigned CWT duration (Part 2).