Gentile, D., Lynch, P., Ruh Linder, J., Walsh, D.  (2004)  'The effects of violent video game habits on adolescent hostility, aggressive behaviors, and school performance', in Journal of Adolescence, Vol 27, No 1: 5 - 22.

[This research follows closely the theoretical models and procedures developed in Anderson and Dill. Briefly, the underlying model, the GA(A)M, suggests that there are individual  'personological' factors, and situational factors involved in aggression. Both the Anderson and Dill piece and this one tends to look at ways in which these factors combine or add to produce aggression, although the model also suggests that one may moderate the other. This article does mention this possibility, although it is mostly convinced that addition takes place. Both papers also use statistical methods such as correlation and regression to establish associations between the variables, and both caution against the view that these correlations are the same as causal relations. The various additional techniques can be used to show the relative strength of different factors. As usual, then, despite all the care and efficiency with which this study is pursued, overall conclusions are much less certain. Nevertheless, these authors are also convinced that something needs to be done -- in this case, the policy that seems to emerge is that parents should play a much more active part in regulating both amount and type of video games playing. Finally, there's a great deal of 'argument by authority', where conclusions stretch beyond the actual evidence gathered, and the authors resort to citing other authorities who happen to agree with them].

The study argues there is a need to separate the amount of play and the content of games that are played in case they are independent of each other, especially when investigating impact on school behaviour [an issue left largely unexplored by Anderson and Dill]. Thus amount of play may affect school performance, while violent content might increase aggressive outcomes  (arguments with teachers and having physical fights).

Several studies seem to think that people who play lots of video games get poorer grades at school  [how good are these studies? Which way round does this association work?]. Gentile et al seem to think that there is a simple 'displacement hypothesis' at work here, where time spent playing video games is not spent doing something more useful such as  'reading, homework, or participating in creative activities' (7). Some people estimate that that  'as many as 89 per cent of games contain some violent content' (7), although clearly much depends on whose definitions these are. Further, a survey of research in the field has found agreement that  'Playing violent games increases aggressive behaviours, increases aggressive cognitions, increases aggressive emotions, increases physiological arousal, and decreases pro social behaviours. These effects are robust' (7). [These are all strongly causal words -- both this study and Anderson's and Dill's are much more cautious when it comes to discussing their own results. Regrettably, this kind of talking up of the problem seems to be unavoidable with this topic -- a kind of moral panic precedes the actual research, and sometimes follows it as well].

There may be some evidence for individual differences in response, however, or even for the view that personological factors such as  'trait hostility' can even moderate effects of playing violent games  (Anderson and Dill suggest this). There are also problems in deciding whether hostility outweighs the effect of playing violent games. Gentile et al prefer to stress one possibility in this general problem of interaction --  'the harmful effects of playing violent games may be even greater for children who are already at higher risk for aggressive behaviour' (7). Parental monitoring also seems to be important in moderating the effects of violent video games (and violent media in general). It is also associated with better academic performance. However, the data is ambiguous here, with parents reporting that they frequently monitor their children's game playing, but with teenagers themselves suggesting that monitoring is far less frequent. Video-game violence has also changed since the early 1980s. It is now  'more graphic and realistic' (8) [so, we are invited to imply, far more effective]. Nevertheless, these trends do indicate that even more research is needed.

[The GA(A)M model is discussed via Anderson and Dill, again with a rather pessimistic gloss in my view. Although the model permits the possibility of reflection and self monitoring in altering the chain being established between aggression, violent video games and aggressive personality, this possibility is hardly discussed. Instead far more speculative possibilities are outlined, such that there may be a cumulative amplification spiral involved, so that playing violent games leads to aggression which leads to demands for more games and so on].

Four hypotheses are to be tested in this study:
(1) exposure to violent video games is positively correlated with trait hostility
(2) exposure to violent games is positively correlated with aggression in naturalistic settings... arguments with teachers and physical fights
(3) trait hostility moderates the effects of violent video game exposure on aggressive behaviour
(4) trait hostility mediates effects of violent video game exposure on aggressive behaviour [that is, the effects have to run through trait hostility rather than acting directly].

607 students were recruited for the study, with a mean age of 14 years. 52 per cent were male, 87 per cent were Caucasian (self rated). Questionnaires were used to gather the data administered by classroom teachers [that is, rates of violence and game playing were also self rated].

A number of variables were isolated for testing:
(a) Violent video game exposure -- respondents were asked to nominate their favourite games and then rate them for violence. They were are also asked how much violence they prefer in their games.
(b) Amount of video game play -- converted to weekly amounts
(c) Trait hostility -- measured on a standard psychological instrument
(d) Parental limits -- how often parents put limits on time, how often parents check ratings, whether parents knew which gains were played and so on a
(e) Arguments with teachers -- how often they had taken place in the past year
(f) Grades -- respondents were asked for their average school grade  [any independent check on these?]
(g) Physical fights -- had these occurred in the last year.

The statistics were used to describe trends and patterns first, revealing that for example the average amount of time spent playing video games was nine hours a week, males played more frequently than females, more time was actually spent watching television and listening to music but less time reading for pleasure, video games playing was frequent  (although six per cent of the sample said they never played them). A moderate amount of violence was preferred, especially by boys. Of the favourite games nominated  '62 per cent... were rated as having some violence, and 37 per cent were rated as including violence... above the mid-point of the scale' (12). Again, boys preferred violent games compared to girls. Parents seemed not to be heavily involved in monitoring game playing -- for example 43 per cent said their parents never limited their play.

Turning to possible effects,  '23 per cent of children reported getting into arguments with their teachers  "almost weekly"  or  "almost daily", and 34 per cent reported getting into of physical fight within the past year' (12). Sex differences were apparent again. Girls also reported slightly higher average grades than boys  [although they all seemed to have done pretty well with mean grades of B+ and B respectively].

Turning to correlational techniques, the outcome measures and the predictor variables were intercorrelated  [among themselves that is -- this is not discussed any further, but I think it might be a problem, suggesting that there might be some common factor, or that the different factors actually overlap in some way?].

The main findings turn on whether video game habits are correlated in any way with any of the other factors. The usual complex picture emerges, but overall  'Exposure to violent video game content and amount of video game play were both positively associated with adolescents' trait hostility, the frequency with which they get into arguments with teachers, and whether or not they have been in a physical fight, and negatively associated with school grades' (13). Parental monitoring seems to have a negative association with arguments with teachers and trait hostility, and a positive one with school performance. However, higher trait hostility levels tended to lead to the consumption of more electronic media especially violent video games, and with lower parental limits  [in other words, hostile personalities might be attracted to violent games and not the other way around].

The next stage was to try to estimate whether video-game violence contributes any independent effect --  'Perhaps trait hostility is the only factor that matters' (14). The obvious [!] technique to use here is logistic regression. An outcome such as having physical fights is specified, and then the different variables are entered into an analysis which permits the researchers to see which ones have the most effect [roughly]. When you do this with trait hostility, exposure to violent games and their interaction, you can see how effective each one is, isolated and and combined. Together, these two factors explained 20 per cent of the variance in physical fights [which might look like low, but it is a statistically significant contribution nevertheless]. Taken separately, some interesting patterns appear -- students who scored low on both also had low instances of physical fights, and students who scored highly on both had high levels of involvement. What is interesting is the mixture in between, however. Here, students with low scores on hostility but high scores on video game violence had a higher incidence of physical fights than those with the opposite score profile. This is of course evidence that exposure to violent video games is a more important factor than having high scores for hostility.

A stricter statistical test was then pursued enabling the researchers to estimate the effect of each of a series of variables -- not only violent game playing, but also  'sex, trait hostility, and amount of video game play per week' (14). In the classic manner, these were tested first, and then violent video game exposure was entered into the analysis, and then finally parental involvement.  [I'm not statistician enough to comment here, but it is quite well known that the order in which you enter variables like this can also distort the analysis -- see Saunders. I gather that the most rigorous procedures involve entering the variables in a series of different orders to eliminate any distorting effects]. The results according to the authors indicate that  'Exposure to violent video games contributed a significant amount of the variance even when controlling for sex, trait hostility, and amount of play; and parental involvement contributed a significant amount of additional variance' (15). However, despite being statistically significant, it seems that these five variables together accounted for 24 per cent of the variance in physical fights  [which leads to a naive question -- what accounted for the remaining 76 per cent? Corresponding figures for the dependent variables below are 17 per cent of the variance accounted for when looking at arguments with teachers, and 14 per cent of the variance in school grades].

Similar analysis is then pursued, this time looking at arguments with teachers as the dependent outcome. Curiously, violent video game exposure  'contributed a nonsignificant amount of variance when controlling for sex, trait hostility, and amount of play' (15). When looking at school grades as the dependent variable, the predictors were entered in a different order, to highlight amount of video game play this time rather than violent content. Amount of play  'contributed a significant amount of variance even when controlling for sex, trait hostility, and violent game exposure'.

In a more complex exercise still, Gentile et al tried to model and test possible causal paths [depicted as a kind of flow chart]. Looking at the occurrence of physical fights, it seems that exposure to violent video games has both a direct association with physical fights and an indirect association, working through trait hostility as a mediator. Similar models were tested for school grades and arguments with teachers, but here trait hostility seemed to emerge as an equally significant variable, while  'violent video game exposure... was not directly associated with grades and was marginally directly associated with arguments with teachers' (16): it seemed to have an indirect effect when mediated through hostility, though.

The authors believe that these results support each of the hypotheses above. Thus exposure to video-game violence does seem to be positively correlated with trait hostility [although we still don't know the direction of causality here]. Exposure to video game violence does seem to be positively related to aggressive behaviours, especially with physical fights. The question of causality does arise, though:  'Are young adolescents more hostile and aggressive because they expose themselves to media violence, or do previously hostile adolescents prefer violent media? Due to the correlational nature of this study we cannot answer this question directly' (18). Some other studies do suggest a direct effect of violence on behaviour  [we have the usual problem with these studies though, because we don't know how good they are]. Some longitudinal research is cited which suggests that it might be early media violence consumption that predicts later aggressive behaviours and not the other way about [but this is a study carried out in 1972]. The evidence cited here does seem to indicate that violent game exposure has independent effects in predicting aggressive behaviour even when the other factors were controlled, and the comparisons of different profiles of scores seem to indicate that, at least, 'hostility is not the whole story', since even those with low scores for hostility but high scores for violent video games  'are more likely to have been involved in fights than high-hostile students who have the lowest exposure to violent video games' (18).

Personality traits such as hostility may moderate effects of media violence, as GAM permits. Thus it is 'possible that the people who are most affected by violent media are those who are most naturally aggressive, thus putting the most vulnerable at the greatest risk for increased aggression' (18)  [true, but what an odd way to put it! Another way to write this is to say that those with low hostility already are not affected as much by video game violence]. The present study seems to suggest that high hostility and video games violence add together to increase the risk of higher levels of aggression.

We should interpret these results in terms of assessing those risks. 'Clearly, media violence is not the sole cause of aggression. But it is likely that it is one of several causes leading to it' (19) [and lots of famous American academies are cited in support]. Surely, it should be that multiple risks are reduced as much as possible -- but [for example] 'boys are generally at greater risk of aggressive behaviours, and they compound that risk by playing more violent games for greater amounts of time than girls play' (19).

The evidence does seem to show that trait hostility is required to mediate the effects of violent video game exposure, but this may be partly an effect of using correlational studies [I don't understand this point]. It is also true that the measures used here are likely to underestimate effects -- for example the scale used to measure hostility covers attitudes and beliefs as well as actual actions, and this may have a distorting effect [I don't understand this either]. When looking at the amount of video game play, it seems this is negatively related to school performance, but less so to aggressive behaviour  [fights least of all]. There may be an indirect relation arising from the finding that 'hostile children play more and are also more likely to play violent games when they do play' (19). The drawbacks of self report data are noted, compared to being able to record episodes of actual violence. However, the authors feel that if anything self report underestimates aggressive behaviour. They also regret not being able to separate types of video games violence, especially since they suspect that realistic violence has greater impact

Parental involvement does appear 'as a protective factor' (19). Support for this arises from another study where parents were actively encouraged to limit the amount of TV and video used at home, or to check the ratings of the games their children play more frequently -- which indicates that parental involvement can even become a predictor, especially of physical fighting. However, such findings may simply indicate  'a broader pattern of more involved parenting styles. Future research on this issue is needed' (20).

Long-term effects cannot really be tested by studies that use correlations, and generally  'Inferences about causal direction should be viewed with caution' (20). Nevertheless, concern  'is not misplaced' (20).

back to key concepts page