I managed to find the citation:
Brady, S.S. & Matthews, K.A. (in press). Effects of media violence on health-related outcomes among young men. Archives of Pediatrics & Adolescent Medicine.
It's in press, so it'll be a few months before actual publication. I'm going to be keeping my eye on it, because the study sounds interesting.
On a more on topic note, I think these kind of beliefs are nonsense.
The methodology of the study seems pretty sound. I'll be able to provide a more detailed analysis once I can get my hands on the article, but there are some key features that I'd like to highlight.
Brady and Matthews had a group of 100 male undergraduates aged 18 to 21 play either Grand Theft Auto III or The Simpsons: Hit and Run. In the Simpsons game, players took the role of Homer Simpson and their task was to deliver daughter Lisa's science project to school before it could be marked late. In Grand Theft Auto III, players took the role of a criminal, and were instructed by the Mafia to beat up a drug dealer with a baseball bat.
The first primary factor to note is that the study was an actual experiment, not a post-hoc study, and so if the only distinction (besides any deviation caused by normal differences) is caused only by the difference in game selected by the experimentors. What this means is that given random assignment (which I would assume was built into the study, but won't take for sure until I see the method) of participants to the study and to each condition (i.e., each person had the same chance of being selected for either condition in the study), so the groups will generally have nonsignificant differences. It's also worthwhile to point out that, if they did the study would be worthless, but it also wouldn't be published without having been verified that they weren't significant.
Playing the violent game boosted young men's blood pressures, and appeared to have more of an effect on those who came from more violent homes or communities, the researchers report in the Archives of Pediatrics and Adolescent Medicine.
This is what we call a "mediating" variable. When analyzing the covariance of two variables, there are often tertiary variables that affect the relationship of the two we're interested in; mediating variables strengthen the relationship, while moderating variables reduce it. Statistical tests can be done to determine whether this is the case (piecewise regression is one such test), and what the strength of the mediating/moderating effect is, and whether the relationship is still significant having removed the mediator/moderator. (This is why I say stats are cool, to anyone who still doesn't believe me). And actually, it looks like that's exactly how they tested it.
They were also more likely to have permissive attitudes toward alcohol and marijuana use.
This is something I really want to see the study for. I'd have difficulty accepting this result unless it was pre-tested as well, although I imagine the publishers would too. But still, this finding is framed much differently than "They are more likely to drink alcohol and use marijuana."
Actually, this is how the study made the examination:
During the survey, participants completed a 13-item questionnaire to indicate their association with deviant peers,34 both for current (alpha=.83) and high school (alpha=.84) friends. The 2 scores (r=0.65; P<.001) were summed to create an overall composite for association with deviant peers (mean [SD], 35.6 [14.0]). Low scores indicated that few of a participant’s friends had ever engaged in any deviant behavior (eg, smoking cigarettes, hitting other people, cheating on tests).
So that's how the test made the pretest.
I haven't read the whole thing, but it seems the study was done very well.
