Notes on Anderson, C., Dill, K.  (2000)  'Video Games and Aggressive Thoughts, Feelings, and Behavior in the Laboratory and in Life', in Journal of Personality and Social Psychology, vol. 78, no. 4: 772 - 790.

This article has been selected by the American Psychological Association as exemplary. It builds on the earlier pieces, including one by Anderson and Ford. The article offers a long report of two major studies on the effects of violent video game play, one involving correlations between real life examples, and the other involving the laboratory experiment. This file offers a fairly brief analysis of the procedure and the findings, but there is much to discuss in terms of the ingenious but flawed methodology, and I have reserved these more technical comments for another file.

The article begins rather controversially, with a reminder that the notorious deadly attack on Columbine High School was perpetrated by two people who enjoyed playing a violent video game  (Doom --  'a game licensed by the US military to train soldiers to effectively kill' (773)). There is also some support for the view that violent video games are having an effect on aggression generally in the US. However, the careful research that follows is much more cautious in its findings.

The Model
The authors have developed a more explicit model to explain how it is possible to explain aggressive behaviour, both in the short run and long run. They call this the GAAM  (general affective aggression model). The model is of course based on much earlier research on aggression. Basically, what it does is to describe a  'multi-stage process' whereby individual ('personological') and situational factors can lead to aggression. There is an excellent diagram of the model at the end of the article. Basically it involves 'input variables' such as aggressive personalities, or situational variables such as 'video game play' or 'provocation'. These can affect  'present internal states', not just affects or emotions, and arousal, but also  'cognitions' or  'aggression related knowledge structures' (776). These can involve, for example  'a behavioural script' telling people how to respond to aggression, or other ways to psychologically access  'aggressive cognitions'. The latter have been explored briefly in the earlier article through the notion of  'semantic priming', which seems to involve providing cues to access meaning. The authors quote here some research that suggests that seeing a weapon can increase aggressive thoughts. At the same time  'there is presently no empirical evidence on whether playing a violent video game increases accessibility of aggressive thoughts' (776) [so they intend to provide some].

The three internal states are thought to be interconnected. One important implication is that the arousal effect is not specific to violent video games  'but could occur with any game that happens to be very exciting' (777). This is a flaw with some early research, and one that is addressed in their own designs here. Before these internal states can be released in behaviour, people have also to go through an appraisal process, which can be  'automatic', or  'controlled'. The latter requires more cognitive resources and time. Apparently, appraisal processes can also activate  'behavioural scripts'--  'Well-learned scripts come to mind relatively easily and quickly and can be emitted fairly automatically' (778). People with aggressive personalities have lots of aggression scripts, and are also more ready to react with aggressive means. Anderson and Dill  'believe that video-game violence also primes aggressive thought, including aggressive scripts' (778).  [There are lots of writers who disagree with this  'script' metaphor, such as Plummer on sexuality, and who think that human beings are much more capable of modifying and writing their own scripts].

The authors go on to adapt the general model to try to pin down the long-term effects of video game violence. Basically, the argument [or hypothesis, but it  looks much more definite in the diagram] is that violent video games help people ' rehearse aggressive scripts that teach and reinforce vigilance for enemies... increase hostile perception, develop aggressive action against others, expectations that others will behave aggressively, positive attitudes towards use of violence, and beliefs that violent solutions are effective and appropriate' (778). The main effect operates through the individual variables, making people more aggressive, and the authors invoke a kind of labelling theory here to explain how personal aggression can get amplified as other people react adversely. It is also worth noting that the specific model, demonstrated in Figure 2 in the article, seems much more deterministic, with all the arrows pointing in one direction, leading from repeated violent game playing, through aggressive beliefs and behaviour scripts, to an increase in aggressive personality, and then a reinforcement of that personality: the appraisal processes seem to have been missed out altogether.

The first study
The work begins by reviewing the literature. Briefly, it seems to be ambivalent or poorly designed, especially in confusing the excitement generated by any game, and the specific effects of violent content. Their first study attempted to correlate various important measures to establish the strength of any particular relationships.  A sample of 227 psychology undergraduates was drawn from volunteers. Variables were measured such as the amount of exposure to video game violence, the time spent playing video games irrespective of content, and several  'outcome variables, namely aggressive behaviour, delinquency, academic achievement, and world-view' (782). The researchers also measured' individual difference variables related to aggression (trait aggression, irritability)' (782) in order to estimate the effects of individual differences. They also recorded gender and academic achievement. The actual study involved giving people a self report questionnaire so as to record data on individual differences, aggressive behaviour, delinquency and world-view [roughly, how unsafe and crime-ridden the USA is perceived to be]. These matters had to be defined and measured first, of course and the various scales that were constructed are themselves interesting and controversial  (see the accompanying file). Much is done to improve the technical consistency and reliability of these scales, which are often based on existing research instruments.

Then the sample was given another questionnaire about their video game habits. People were asked to name their 5 favourite games, and then rate them, reflecting how often they played the game and how 'violent the content and graphics of the game were'  (785). They were asked to categorise the game as well. The world-view of the sample was measured by asking them to estimate how dangerous life was in current America, in terms of crime likelihood and safety feelings. The data was then processed using correlational analyses of various kinds:

(a) First, the reliability inconsistency of each of the two scales was estimated by seeing if the items correlated with each other. They did. Then the variables were standardised in order to permit comparisons among them  [the normal statistical procedure, but one which may have implications]. In a brief description of the results, Anderson and Dill notes the relatively high level of video games playing declining from high school to college; the overall popularity of video game playing; the popularity of games such as Super Mario Brothers, Tetris, and Mortal Kombat. This gives the researchers an immediate problem, since only the third one appears to be very violent. However, they insist that  'Super Mario Brothers... [also involves]... considerable violence in the sense that the player typically spends a considerable amount of time destroying other creatures' (788). They also override the views of the players themselves for example the  'one person who listed Mortal Kombat... as a  "sports"  game' (788). By imposing their own classifications, they managed to make up the percentage of aggressive games in the categorisation exercise to 44 per cent.  [It is not at all clear if the results are calculated on this expanded category, or what they would look like if the lower estimates were taken].
(b) A far more technical analysis involving correlating the various items together is then described. For example, overall  'Aggressive delinquent behaviour was positively related to both trait aggressiveness and exposure to video game violence' (788), and measures of the strength of these correlations are given as 0.36, and 0.46 respectively. Similar but lower correlations were established between trait aggressiveness and exposure to video game violence and non-aggressive delinquent behaviour, and between exposure to video-game violence and aggressive personality. Gender was strongly related to several of the variables, especially  'perceived safety... video-game violence... and time spent playing video games' (789). In the strongest test of the relationship between the terms, Anderson and Dill pursued what they call  'destructive testing', which involves establishing a link between video-game violence and aggression, and then attempting to introduce more and more variables until the link disappears. [This is sometimes called organising a 'variable race', and it has its problems -- see accompanying file]. In short, video-game violence exposure persisted as a predictor of aggressive behaviour --'Thus, the link between video-game violence and aggressive behaviour is quite strong indeed' (790). Individual differences also had an effect though, and here, aggressive personality especially showed a link with aggressive behaviour. Anderson and Dill note a particular large effect produced by combining exposure to video game violence and aggressive personality factors --'the VGV [video games violence] effect on aggression occurred primarily among participants with high AP [aggressive personality] scores' (790), especially for men. Similar results can be obtained when considering aggressive delinquency, again with video games violence exposure being significantly related, and the only other significant factor being aggressive personality. When examining the effects on feelings of safety, the video-games violence variable seems slightly less important, and gender more important; when it comes to estimating crime likelihood, video games violence exposure seems to display no link with the tendency towards high estimates. Academic attainment seems to be more affected by time spent on video games rather than violent content.
(c) Summarising the results, Anderson and Dill say that  'Violent video game play and aggressive personality separately and jointly accounted for major portions of both aggressive behaviour and non-aggressive delinquency' (793). This finding is consistent with the predictions in the GAAM model  [note the caution here-- they're not saying that this establishes the GAAM model]. They are aware that correlations cannot confirm causes, but argue that  'a zero or negative correlation would have disconfirmed the hypothesis' (793). They remind us that their sample was drawn from college students, suggesting that this may have minimised the effect: 'Those with serious decrements in intellectual functioning or serious aggressive behaviour problems are not as frequently represented in college samples as would be the case in a high schools sample' (793).

Overall,
'concern about the deleterious effects of violent video games on delinquent behaviour, aggressive and non-aggressive is legitimate. Playing violent video games often may well cause increases in delinquent behaviours, both aggressive and non-aggressive. However, the correlational nature... [of this first study]... means that causal statements are risky at best. It could be that the obtained video-game violence links to aggressive and non-aggressive delinquency are wholly due to the fact that highly aggressive individuals are especially attracted to violent video games. Longitudinal work... would be very informative' (794).
The Second Study
This was laboratory based, and was better designed to attempt to pin down the causals. However, it tended to focus on short-term aggression. It was carefully pursued as usual:
(a) A pilot study helped to pick nonviolent and violent video games, and to try and focus on content rather than other relevant factors such as  'enjoyment, frustration level, and physiological arousal' (794). The researchers chose a video game (Wolfenstein 3D)  which they thought offered  'blatant violent content, realism, and human characters' '(794), and compared this to another violent game and to non-violent games  (Myst and Tetrix). A small sample of participants were exposed to these games in various orders, and their blood pressure and heart rate were measured. Participants also rated the game in terms of how difficult enjoyable frustrating and exciting they were. As a result, the researchers were hoping to choose two games that differed only in terms of 'amount of violence', but which scored exactly the same on all the other measures  (795). This was not that easy, since the chosen games (Wolfenstein and Myst) also reflected differences between genders, and excitement levels. As a result, participants in the main study were still asked to rate their games as a check.
(b) In the main study, 210 psychology undergraduates were compared in terms of their reaction to the different games. Aggressive thought, affect [emotions] and behaviour were measured, and the effects of gender and aggressive personality [only one measure of aggressive personality, in fact, the scale of 'irritability'] assessed. Participants were allowed to play the game three times, measuring different variables each time. To facilitate these comparisons, participants were selected who were particularly low and particularly high in irritability. The precise arrangements of the laboratory studies were designed to minimise the effects of the presence of the researcher or differences in terms of instructions. To prevent participants guessing the point of the experiment, they were given a cover story about learning in developing skills. Participants played for 15 minutes each time.
(c) ('State') hostile affects were measured by a particular scale developed in earlier research, which involved participants rating how 'angry' or 'mean'  they felt, and inviting agreement or disagreement with particular statements. Games were rated in terms of their difficulty etc using the instruments developed in the pilot study. Another 15 minutes play session ensued, then participants were tested for aggressive thinking by showing them particular words, including aggressive words ('e.g. murder') and 'anxiety words... escape words... and control words' (798), and measuring how quickly people were able to read them. After a pause of a week, participants came to play another for another 15 minutes, and aggressive behaviour was tested by measuring  'competitive reaction time', involving a competition to press a button faster than an opponent can, and punishing the loser by operating a blast of noise:  'Aggressive behaviour is operationally defined as the intensity and duration of noise blasts the participant chooses to deliver to the opponent' (799). This is apparently a widely used measure in psychology. The competition was in fact rigged so that each participant won 13 and lost 12 contests, and the noise blasts were in effect controlled by the computer. Care was taken to make sure participants knew how to work the equipment and to play the game. After the entire study, participants were debriefed and told the real purpose of the study.
(d) The results try to pin down the effects of the excitement levels of the game, and the difficulty in frustration that was caused. Game frustration did affect state hostility, and game difficulty  reaction time, but these effects could be controlled and 'did not appreciably alter the effects of most interest'  (200). The scales were tested for reliability by correlating the items, and unreliable items were deleted. This time, an analysis of variance was pursued, correlating the various elements with particular dependent variables: 
for state hostility overall irritability was a main factor, and so was gender. All other effects, including  'game type effect',  'were nonsignificant' (201). Gender was the main factor affecting the crime and safety ratings, with nothing else approaching significance.
accessibility to aggressive thoughts, as measured by the reactions to particular words  'did not produce a significant control word type by game type interaction'. However, when reactions to three types of words were combined into a single measure  [a pretty controversial step, especially when the resultant composite was called the  'Aggression Accessibility Index'], game type did seem to have a main effect in that those who have played a violent video game scored more highly than those who did not.  'In other words, the violent video game primed aggressive thoughts' [a highly dubious conclusion in my view]. Gender also had an effect, higher for men of course. However, this seemed to be no connection between aggression accessibility and the trait irritability measure [aggressive personality]. Anderson and Dill suggests that this may be because playing the games themselves is  'sufficient to temporarily override the usual differences between people high and low in irritability' (202)  [a puzzling and dubious remark, based on nothing really].
aggressive behaviour, measured by the noise delivery experiment, attempted to assess both the effects of the violent video game, and of other variables, such as the effects of losing a trial and wanting retaliation. These other variables may have had a results, including the effects of the instructions actually given. This might explain [!] their lack of any statistically significant effects of any of the independent variables (including game type) on the noise intensity settings. In terms of duration, gender seemed to have an effect, with women delivering more noise than men, as did the effect of losing a competition. Trait irritability had an effect. This time, so did game type:  'playing a violent video game increased the aggressiveness of participants after they had been provoked by their opponents noise blast' (203), although effects were only 'small to medium'. [NB this assumes that duration outweighs intensity as a measure of aggressiveness].
overall, violent video game types seem to have an effect on accessibility of aggressive thoughts and aggressive behaviour, but do not affect state hostility. This could be because cognitions are more affected than emotions. This too was tested and some support found for it.
Discussion of the Two Studies
Overall there is some evidence to suggest that violent video game play it is positively related to aggressive behaviour, both in and outside the laboratory. The combination of methods  'lends considerable strength to the main hypothesis that exposure to violent video games can increase aggressive behaviour' (204). [Lots to discuss here -- can the two methods be used like this to cancel out the faults of each one?]. The results of this study lends support to the results have prior work, especially those finding effects of TV violence. As a result,  'parents, educators and society in general should be concerned about the prevalence of violent video games in modern society' (205).

The effects of violent video games were moderated by individual differences in aggression, being stronger for those who were already high in trait aggressiveness. However, this study showed that low trait aggressiveness did not really modify the impact of video-game violence. Overall though it looks as if aggressive personality and video-game violence offer a  'long-term bidirectional causality effect in which frequent playing a violent video games increases aggressiveness, which in turn increases the desire and actual playing of even more violent video games' (205).

The usual effects of gender were modified here, since  'women displayed higher levels of state hostility and aggression than men' (205). This may reflect situational differences in provocation and how the genders react to them -- men may have been more familiar with video games, and women less happy to play them.

GAAM draws attention to cognitive, affective and arousal roots, but the experiments show that violent games increase the accessibility of the aggression related thoughts, but not feelings of hostility. Thus cognitive models of aggression seemed more important, even though it may be difficult to disentangle them from feelings. However,  'the danger in exposure to violent video games [assuming there is one] seems to be in the ideas they teach and not primarily in the emotions they incite in the player. The more realistic the violence, the more the players identify as with the aggressor' (206)  [I did not see any evidence in the study to support this view about realism -- it seems to be drawn from other work].

Academic achievement was not particularly related to violent video game play, but it was related to long-term exposure  [presumably by taking up lots of time]. The finding is still controversial, and some researchers  'found a positive relation between general video game play and IQ' (207). The sample is not designed to test this much further.

Thus 'concern about... playing violent video games is not misplaced (207). Video games may offer more dangers than television or movies. Other research indicates that identification with an aggressor increases aggression, and in video games  'the player assumes the identity of the hero... and usually sees the video game world through that characters' eyes' (207)  [there seem to be some major assumptions here, and again the authors are straying beyond the findings of their own study]. Active participation may also increase aggressive behaviour, since the video-game player chooses to aggress, helping  'the construction of a more complete aggressive script than would occur and the more passive role assumed him watching violent movies or TV shows' (208)  [an empiricist view of participation here]. Finally, video games have 'an addictive nature' (208), and more research is quoted here, including the view that  'one in five adolescents can be classified as pathologically dependent on computer games' (Anderson and Dill citing Griffiths and Hunt, page 208)  [we would need to look very carefully at definitions and correlations here]. Games do reinforce participation. Overall they  'provide a complete learning environment for aggression, with simultaneous exposure to modelling, reinforcement and rehearsal of behaviours' (208)  [again based on the work of other researchers].

Summary and Conclusions
The authors rehearse the findings, but in a rather cautious way -- for example arguing that  'If repeated exposure to violent video games does indeed lead to the creation and heightened accessibility of a variety of aggressive knowledge judges... the consequent changes in every day social interactions may also lead to consistent increases in aggressive affect' (208)  [my emphasis]. The piece ends with a controversial set of remarks about the Columbine murders, and the general worry about the increasing violence of video games. There is also a plea for more research, including longitudinal studies -- these are required to 'test the proposition that such exposure can produce stable changes in personality' (209)