There has been much talk about violent media and the effect on our children, or adults as some would say.  While our senators and governors play around on capital hill with accusations that are usually biased as hell (from both sides mind you) and point the finger in the wrong direction or at an illegitimate source, let us review some recent criminology findings.

This paper published by “Criminal Justice and Behavior” in 2008 compiled a more than adequate amount of research on the topic:  The Effects of Media Violence Exposure On Criminal Aggression: A Metal Analysis.  The authors did an excellent job of defining the current problems with our literature that we use, have looked at, and even replicate today.  These problems include faulty statistics, defunct methodology, lack of actual aggression links, the idea of Demand Effect, the psychological and behavioral difference in pressing a button in a confined space and actually hitting another individual, and the list actually goes on for a fair more amount of text including the idea of participants knowing they will not get into any legal trouble for particiapting in the study.

The authors, Savage and Yancey, continue on the path of the requirement for an actual definition of aggression:

“Although some argue that aggression is on a continuum, and thus, causes of minor aggression are likely to be the same as the causes of major aggression, we have made a significant effort to track down empirical sources that establish this link and find no empirical evidence to support this assumption.”

I appreciated that these authors were so candid about the matter as many researchers like to turn a blind eye to this.  Additionally Savage and Yancey point out the special problems that apply to a meta analysis:

Publication Bias: Only studies that have large effect sizes or significant results usually get published because of no findings or not the recorded findings wanted, this is considered the dark side of research

Mixed Quality: The quality of the methodology comes into question.  If the methods are not up to par and an effect size or significance is found, it can bias the research.

Statistical Reporting: Due to age of the studies (some are quite old) it is harder to derive effect sizes when interpreting the study via a new data analysis.

Post-Hoc Comparisons: When comparing post-hocs using multiple studies, it is possible to find significance where there is not any.  This is kind of a backwards “step-up” approach to statistics.  For example, if you run a MANOVA and do not find significance, some researchs (BAD researchers) will then look at the univariate table to see if there was any significance and report that.  During a meta-analysis, if the original study (think the univariate table) is not significant, but then gets put in with other studies of similar nature and statistics ran (MANOVA) there becomes a significant variable.

The procedure of the study was to recode all  of the past research into variables and then summarize them to attain one number for each category, age, sex, sample size, IVs, DVs, etc.  This is an acceptable manner of statistical procedure.  The researchers also took out the studies that used the same sample for multiple comparisons as this creates bias.

After analysis this is what the authors had to say about the studies:

“In the end, there is not one study that reports the comparison we would really like to see to satisfy our curiosity about the media violence–criminal aggression relationship. Such an analysis would use, as a dependent variable, serious criminal aggression or violent crime rates. It would use a measure of exposure to television violence that includes both an accurate assessment of exposure (how much time) and an independent rating of violence in the programs, and it would control for early-wave trait aggression using a reliable and valid measure of early childhood aggression. The analysis would also control for variables such as SES, parent education, parental violence, neglect, and intelligence, all of which are also associated with both exposure to TV violence and aggression. The Huesmann (1986) report approaches this—because criminal convictions are used as the outcome—but, unfortunately, it does not include enough detail about the analysis to judge the merits of the findings reported.”

It is also very important to point out that if you read the study, it points out subtly that criminal convictions are used as the outcome.  This does not include anyone that watches or plays violent media into the sample.  Interesting right?  It also furthers the case that Mental Health may be an issue that is not being considered.  Thoughts?