In 2008, the Journal of Media Psychology published the results of a longitudinal study on "Media Violence and Youth Violence", conducted by three German "anti-killergame" researchers (Werner Hopf, Günther Huber and Rudolf Weiß). The study is available here, for those who are interested : [url]http://www.vgmg.ch/forschung/media_violence_youth_violence_2008.pdf[/url]

The conclusions of this study are quite harsh for "violent games". In particular, the authors claim that [I][B]"playing violent electronic games is the strongest risk factor of violent criminality"[/B][/I] (sic) and that [B][I]"the more frequently children view horror and violence films during childhood and the more frequently they play violent electronic games at the beginning of adolescence the higher will these students' violence and delinquency be at the age of 14"[/I][/B] (re-sic). Given the current German context, it was predictable that this longitudinal study was used as an irrefutable evidence that a ban of "killer games" was urgent.

As I was intrigued (and a bit upset) by such conclusions, I decided to contact some researchers to have their opinion. Among them, Dr. Christopher Ferguson, who is well-known by GP readers. He was kind enough to give me his impressions on this German study, and I forward them here, with his approval :

[QUOTE="Christopher Ferguson]The problems with this study for me begin in the introduction section. The authors, frankly, are dishonest in the way that they present the research on media violence, asserting greater strength and certainty than actual exists in this field. They do not note many studies that contradict their beliefs, nor numerous criticisms and warnings about problems with the media violence field. This is a substandard literature review, one that is more akin to propaganda than science. If the authors present such a distorted view of the literature (and they're not the only ones who do this), that makes me worried about whether they are objectively analyzing their data or hunting for a particular outcome they are already invested in.

As for their analyses, their methodology section is poorly written and difficult to understand. However, they appear to select a number of poor-quality measures with little validity, when they could easily have chosen much better, well-researched measures. Perhaps the availability of measures in German was the issue, but they should have taken the time to translate effective measures into German and test them. They did not due this. As such their measures lack appropriate validity.

In the regression the "media effects" variable got the largest effect size, simply because they added it in early in the regression. Had they added the media effects variable in later in the regression, which would have been more appropriate as media effects is more "distal" to family and intrapersonal variables that are more "proximate". Thus their claims that media violence is the "biggest" risk factor is without merit as they didn't employ the proper destructive testing (putting media violence as the last step of the regression) to merit this claim. It's also hard to know from their writing whether they controlled for initial violent behavior in the regression (was T1 violent behavior a predictor? It should have been...but either it wasn't, or it wasn't significant...either bad methodology or a suspicious outcome that should have raised a red flag).

Their path analyses look like a mess. I'm not sure if it's just the way they present their graphs, but at times it looks like they threw variables in multiple times to the same path analysis which is not really appropriate (and at very least occurs without explanation here). There's no analyses of "goodness of fit" anywhere in this (which you'd get from SEM), so there's actually no evidence that these path models are theoretically sound. They seem more like a jumble of variables thrown together without any evidence that these are the most effective or parsimonious models. The authors do refer to some GPK "stability" statistic, but I've never seen this used before and the link they have to a reference on it doesn't seem to work.

Further their data is correlational, yet they attempt to assert causality which demonstrates the naive nature of this report. There is nothing you can do to correlational data to make it causal.

To me the report reads from the beginning as if the authors have an "axe to grind" and subsequently conducted a rather messy set of analyses using measures of questional validity, and statistical analyses of questionable sophistication. I certainly wouldn't enact public policy based on this.[/QUOTE]

I also forwarded this response to German website Stigma Videospiele ( [url]http://stigma-videospiele.de/wordpress/[/url] ) and French website Game in Society ( [url]http://www.gameinsociety.com/[/url] ).