A recent article in Science magazine covered the story of a new research study (full study available here) linking violent video games and movies to aggression. The study claims the link is almost three times stronger than abusive parenting. Could this be true? Let’s take a look at the study and see.
The study is led by the sometimes controversial Iowa State group, who have long argued for links between violent media and aggression. This group has sometimes come into criticism, for example, for comparing the effects of violent media to smoking and lung cancer, or implying video games might play a role in mass shootings despite absence of evidence for such claims. One of the authors once participated in a murder trial where, in effect, his testimony appeared to help the defense shift blame from the (ultimately) convicted murder onto one of his teenage victims, arguing that a combination of video game playing and ADHD symptoms put the teenager at risk for violent behavior. I understand this must sound like a convoluted argument, but the actual murderer claimed he came upon the teenager murdering the teen’s own family, then had to defend himself from the video game-addled teen by beating him to death with a tire iron (the jury did not buy this argument). And this is not the first time that the Iowa State group has tried to argue that video games have more impact than abusive parenting.
Reading the article itself raises some immediate red flags. The authors engage in a practice called citation bias. This occurs when study authors only cite studies supporting their world view and ignore disconfirmatory evidence. This is generally regarded as bad practice, and some evidence suggests that this behavior is associated with inflated effects in video game studies. Citation bias can function as an indicator that study authors have strong beliefs… beliefs that can, unconsciously or consciously, lead them to nudge their study results one way or another. Granted, this is nothing remotely unique to media effects science, but it’s still a problem.
The study does benefit from one impressive aspect… the sample. 2100+ participants across seven countries ranging from the U.S. to China to Romania to Australia. That’s a pretty impressive collaboration.
Unfortunately, the impressive aspects of the study end there. I was less clear on exactly who the participants were or how they were recruited. They’re described as adolescents and young adults (mean age 21), but I wasn’t sure if these were mainly college students.
The measure of media use is a serious problem. The authors ask participants to rate their top three movies, TV shows and video games, then asked them to rate how violent each of these shows or games were. This approach is generally considered bad practice for some fairly obvious reasons. First, what one person considers violent may differ markedly from another. But more importantly, this approach pretty much advertises the hypotheses of the study. This is a phenomenon called “demand characteristics”… when participants can guess the purpose of a study they tend to give experimenters false responses in line with what they think the experimenters want them to say. So basically, because this study is entirely self-report, the authors asked, “How violent are the shows you watch, and, oh by the way, how violent are you?” This is very unfortunate as it can create all manner of spurious results that don’t reflect reality.
Questions about other factors such as abusive parenting or violent neighborhoods tended to be rather perfunctory, and had low reliability (one reason results for them may have come out weak). So, it would have been pretty obvious to participants this was a study attempting to link violent media to aggressiveness, given all the data are self-report.
The outcome measures for aggression and reduced empathy are ok, but all focus on hypothetical behaviors, not real ones. So, they’re not measures of actual aggressive behavior per se.
Indeed, the authors end by claiming that violent media has a bigger relationship (almost three times larger!) with aggression than does abusive parenting or victimization by peers or neighborhood crime. So, is it actually less bad to beat your kids than let them play violent video games?
Well, these results are obviously implausible. Even if you think violent video games or movies are bad, claiming they’re worse than abusive parenting is nonsense. The pie chart they provide is also a bit misleading as it makes it look as if violent media explains 23 percent of the variance in aggression. They did this by excluding unexplained variance... from their statistics, taken at face value, violent media only explained about 2.6 percent of the variance in aggression. In my interpretation, the study’s findings are the result of a crude correlational self-report study with obvious demand characteristics and a sub-par measure of violent media use. We’ve seen a lot of poor quality research attempting to link violent media to aggression over the years… unfortunately, this study, while having a few impressive aspects related to the sample, largely repeats some of the mistakes of the past. Nothing to see here folks, move along…