Every week I receive an email from the SafetyLit organisation summarizing new publications in the area of safety science that may have be of interest. Increasingly these have also included studies of video game playing. So I decided I would take time out from my usual PhD related reading (on Traffic Safety) and have a look at some of them and attempt to provide summaries of them here.
The idea being to give a slightly more in-depth examination of these papers than is typically given in a “study shows x” headline in a newspaper. This means providing a bit more information on the methodology and analysis, and therefore it may be a bit dry. You have been warned. Also if you want the full story then of course reading the article yourself is the best option (sorry they are all pay walled as far as I know - such is the way of academic papers at the moment).
Christopher J. Ferguson
Journal of Youth Adolescence, 2010
The paper opens with an introduction to the area and points out that the potential linkage between violence and computer games has been variously debated in public but lacks research consensus. While some research claims there are strong casual effects of violence in video games on real life violence, others claim only limited effects in specifically already high risk populations, whereas other researchers have found no effect.
Ferguson suggests that the reason why there are such diverging finding comes down to methodological issues which have also affected research on television and violence. Namely that the more validated a measure of aggression is then the chance of finding an effect becomes smaller (the same applies if you actually directly observe the behaviours in question), and that in correlational studies if other relevant variables such as socio-economic factors and personality are considered then effects often also disappear.
The article then continues to offer some critical assessments of several prospective studies (a study following a group (cohort) of people over time) of the effects of game violence (of articles that both find for and against such effects). Most of the critical comments being around the short time periods, small effect sizes (increases in aggression of 0.5% for instance), a lack of consideration for other potential contributing factors and what the author considers to be inappropriate aggression scales being used.
In summary Ferguson says that what is of real concern to him is if violence in computer games leads to pathological and serious aggression in wider society. A question that he believes is unanswered.
The paper then goes on to cover three theoretical views of the potential relationship between video game violence and serious aggressive behaviour in real life. Those being:
- Exposure to violence in video games leads, through learning, to serious aggression in other areas
- That individuals who already are highly aggressive are drawn to forms of entertainment which have violent content.
- The correlation between video game playing and serious aggression is due to other underlying variables (e.g. socio-economic conditions, family situation, etc).
In the subsequent discussion Ferguson likens the first, causal, relationship to worries expressed by Plato that Greek plays would cause “rebelliousness and licentiousness” in those young people who watched them. Which I guess shows how long arguments about the influence of media on people have been going on (and will continue to go on). He also seems to lean towards explanation number 2 personally, although he does a good job of presenting the other possible explanations.
So with all of the above in mind Ferguson moves on to describe his own experiment. In particular his experiment has concentrated on using measurement scales that relate only to serious, pathological, aggression. In brief, the experiment described involved an initial sample of 536 mostly Hispanic youths (selected via snowball sampling) from a city on the border of Mexico. They were aged between 10 and 14 when recruited for the study, and there was roughly the same number of boys and girls (51.3% boys in the sample). One year later, when the follow up questions were made, this sample had dropped to 302 children (52.3% female this time).
The study was a questionnaire study, and therefore did not observe actual behaviour but rather relied on self-reports of the children and their parents. During the initial phase they assessments of media violence exposure, negative life events, family environment, family violence, depressive symptoms, serious aggression, bullying (as instigators, not victims), and delinquent behaviour were made. Then during the follow up period 1 year later only the depressive symptoms, media exposure and last three outcome measures were collected.
The data gathered in the experiment was then analysed using hierarchical multiple regression equations. Power analysis was also carried out and suggested (r = .14) that the study had sufficient power to detect even trivial events. One great thing about this analysis is the author checks for the temporal direction of the results. That is he tries to see if being a violent person predicts playing violent video games, as well as checking to see if playing violent video games predicts being a violent person. This type of analysis is relatively rare, and nice to see.
So what are the results? Well there are many, but I will only mention the ones that seem most interesting in this forum. At the follow up period, 75% of the children reported played video games in the last month, 40% reported that the games they played were violent (in their own opinion) and 20.9% reported playing a ESRB M-rated game in the last month. In terms of correlations, boys were more likely to play the violent games, but playing violent games did not correlate with age or GPA (GPA was also not predicted by hours spent gaming).
As for violence in games, well there was only one correlation initially and that was for bullying (more reported bullying by kids who played violent games), although the r values found were small (.18). However it is worth mentioning that a p value of < 0.004 was chosen as the threshold for statistical significance. This is due a Bonferroni correction being used due to the multiple comparisons going on, but a p value of 0.004 is quite conservative.
This correlation disappeared however when the hierarchical multiple regressions were made with the results from the depressive scales being the strongest predictor of aggressive behaviour and rule breaking in the children studied (although all with somewhat small effect sizes). In other words, when the other factors were also considered then violent video gaming no longer had any correlation with any of the behaviours.
So, playing violent video games had no predictive power for aggression or rule breaking. However, being Male, reporting depressive symptoms, and playing video games when first initially questioned all significantly predicted playing video games one year later (again, small effect sizes).
In conclusion Ferguson says that his study shows no link between violent computer game playing and serious aggression and that future research should be more careful and conservative. He also suggests that perhaps violent content in computer games may have small, short term effects on low level aggression in individuals – but that these should not be generalised up to serious acts of aggression in society at large.
My personal final comment on this study is that it appears relatively methodologically sound (It isn’t a random sample, was limited to Hispanic youth, and could have missed other factors – but all of these limitations are clearly and openly acknowledged by the author). However, as always with pure questionnaire studies like this, it is a pity that no actual observation of behaviour was carried out although of course this is difficult to do with such a large sample size. However perhaps disciplinary records for the children involved could have also been examined?
Rani A. Desai, Suchitra Krishnan-Sarin, Dana Cavallo & Marc N. Potenza
Official Journal of the American Academy of Pediatrics, 2010
This paper attempts to focus not on the linkage between game playing and aggression, but rather on health outcomes. It starts much like the Ferguson article by pointing out that findings in this area are mixed, and mentions that while negative outcomes have been reported (e.g. gaming addiction, being involved in fights) by some studies others have reported positive indicators ranging from improved motor skills to improved social skills (from interacting with friends and family to play the game).
While the authors do spend more time here initially mentioning some of the reported positive impacts of game playing on health , they do also stress that the evidence for these positive effects are still often just as limited as the evidence for the negative effects.
The authors go on to suggest that was has been described as gaming addiction might be related to an impulse control disorder, much like pathological gambling is, were affected individuals have growing urges to take part in the behaviour, and start to feel increasingly tense if it is denied to them. They also note that the complaints of others about the behaviour (usually a decent indicator of addition), parents in particular, cannot always be taken as reliable indications of addictive behaviours in terms of teenagers and gaming, as conflict between parents and teenagers on many issues is typical.
The actual experiment reported again uses a survey methodology carried out by inviting 4 year (non-vocational or special education) public high school students in the state of Connecticut to participate. This resulted in 4028 respondents who provided data on video gaming.
The measures taken were gender, race, ethnicity, grade, family structure, GPA, extracurricular activities, some drug use questions (smoking, cannabis, alcohol, caffeine, and “other”), some depression related questions, involvement in fights that required medical attention, carrying a weapon to school, Body Mass Index, and time spent playing computer games. Those that indicated they did spend any time playing video games were then asked follow on questions asking if they had ever tried cutting back, if any family members had ever expressed concern about their game playing, if they had ever missed work or school because of it, if they thought they had a personal problem with playing too many video games, if they ever felt an irresistible urge to play video games, and if they ever felt an increasing tension or anxiety that could only be relieved by playing video games (the last three being indications of a possible impulse control disorder).
Once the survey data was collected correlations, and logistic regressions were used to examine the data. Again I will only report results I think are interesting or relevant.
Well, more boys (76.3%) admitted to playing video games (for at least 1 hour per week) than girls did. Then, if the total sample was examined being Asian or being in a lower grade level was related with playing video games.
The more interesting results however come when the sample is broken down into boys and girls. Here it was found that boys (45.8% of the sample) who reported playing games had a higher GPA, were more likely to report having never smoked or used Cannabis and had a higher reported caffeine consumption than the non-gamers.
However if the girls are examined then those who reported gaming were more likely to report smoking occasionally, having never used cannabis or alcohol, having a high level of caffeine use, having no history of depression, carrying weapons in school, getting into serious fights and having a slightly higher BMI (22.35 vs 21.94, which was statistically significant, but isn’t very significant in terms of health).
So those are the results using “playing at least 1 hour of video games a week” as an indication of a game player. If the time playing is examined in a little more detail then 61.1% reported playing 7 hours/week or less, 10.9% 20 hours or more. Also 4.9% of participants reported positively to the 3 items that indicated a possible impulse control disorder (with more boys (5.9%) reporting this than girls (3.0%)).
If only these particular impulse control issue gamers are examined, then amongst the boys they were more likely to be non-white or Asian, smoke regularly, depressed, get into fights and carry a weapon. In girls they were more likely to use other drugs, be depressed, and be involved in serious fights.
The authors do note though that many of these findings do not have particularly large effect sizes, but feel confident that problem gaming (that being those who answered positively to the three impulse control questions) is associated with a greater chance to smoke regularly, be depressed and be involved in serious fights.
So what do the authors conclude from all of this? Well first of all they state that there are no negative impacts on health of game playing for boys. They believe this is because video game playing in boys is normalised. Although they also say that it is perhaps boys are more likely to become problem game players, although in general the chance of this happening is very low anyway. In terms of girls they believe the results the found are due to the fact that while video games are attractive to boys generally across the whole population (or at least may have many other peers that also play), that there may be only a certain subtype of girls that are attracted to playing video games. They then suggest that this subtype of girl may already be aggressive, or may not have other female peers that play games. Similarly they suggest that the positive health effects, such as a lower level of smoking amongst boys who play games may be related to the type of peer group that male gamers might have.
However the authors do, and rightly so, admit that they cannot be sure of the causal direction of the relationships that they have found and that other unaccounted factors may be responsible for the effects they observed. The authors also note that because their definition of problematic gaming was quite strict (only those who answered all 3 impulse control questions) the sample size of problem gamers they gained was not really big enough to make any solid conclusions from. They also point out that their study is just a snapshot, and does not include any longitudinal evaluation of these behaviours over time. I would also personally add that it is a pity that no objective measures of behaviour could be included along with the subjective self-reported ones.
Kathleen Beullens, Keith Roe and Jan Van den Bulck
Accident Analysis and Prevention, 2011
This paper opens in the standard fashion for any paper that addresses young drivers, by restating the well established “young driver problem” - that being that young drivers all over the world are over represented in car accidents, and in fact traffic accidents are the number one cause of death amongst young people world wide.
The paper then moves on to talk about media exposure and the fact that young people play computer games, boys more so (they also are involved in more traffic accidents), and that the racing car genre is a popular one. The authors go on to then mention a couple of studies which have looked at playing racing games and their immediate effect on cognitions and intentions to perform risky traffic maneuvers, or self reported accident involvement. In comparison with the other two studies above no mention of mixed results is mentioned, and the few studies presented all showed harmful effects for game playing.
The authors however point out that these are only short term effects and propose a longitudinal study to see if they continue to occur over longer time periods. The authors then go on to outline their theoretical underpinning for the study and this is where they start to loose me personally.
The two theories they are using are cultivation theory, which suggest that media (such as TV and video games) are an important form of socialization and information and that these influence people’s attitudes (attitudes being what people say they think about things). Ok, fine. Now they move on to the Theory of Planned Behaviour. This is a very popular theory (perhaps one of the most popular in Social Psychology) and essentially states that attitudes can predict behaviour (through intention to perform that behaviour).
I dissagree with this theory. It essentially says that attitudes (what people tell you they think about something), social norms (what people tell you they think that other people think about something), perceived behavioural control (what someone tells you they think they can do about something) predicts intention to do something (what behaviour people say they are going to do) which then predicts what people actually do.
Now, even just the leap from what people say they are going to do = what they actually do is a difficult one for me to buy (see the “the road to hell is paved with good intentions”, etc and the Attitude-Behaviour gap) but also Theory of Planned behaviour research has an unfortunate tendency to rely only on self-report and never actually observe behaviour. This is a problem, because if the link between intention and behaviour is not well established (and as far as I am concernted it is not) then continuing to rely on the fact that attitude predicts intention and therefore behaviour is obviously troublesome.
Ok, but with that rant over let us get back to the paper. Again this is a questionnaire study, this time from 15 randomly selected secondary schools in Flanders, Belgium. The study was then carried out in two waves with one questionnaire delivered in school (to 2193 people, 65.2% male) and another coming two years later. There was quite a lot of participant loss in this time, and only those who had gained their drivers license between the first and 2nd survey were included (the licensing age is 18 in Belgium) so the final sample size was 354 people (62.7% male).
The measures taken were; ratings of how often they played racing games (e.g. racing sims like GT), how often they played ‘Drive’em up games’ (e.g. Driver, burnout, etc), a risk taking attitudes measure, a sensation seeking scale (which kind of taps innate/biological predisposition to risky behaviours), physical aggression, driving intensity, and measures of intention to speed, ‘fun ride’ and drink and drive, as well as self reported frequency of risk taking behaviours while driving.
The analysis was then carried out using structural equation modeling with risk taking attitudes, risky driving intentions, self reported driving recklessness, and racing and drive em up gaming as the main variables and self-reported driving intensity and aggression as potential confounding factors.
Ok, so the results. The authors report that Racing games were popular in their sample with 28.8% playing them at lease once a year, 12.8% a few times a week and 5.1% almost every day. The percentages for drive em up games were slightly lower, but still relatively common. In terms of self-reported risky driving behaviour over half of their participants (51.9%) reported sometimes driving fast for the thrill of it (only 8.4% said they did this often), and speeding was even more common with only 14.8% saying they never did it. Only 14.5% however admitted to drink driving.
In terms of the relationships between risky driving and video games, the authors report that self reported risky attitudes towards speeding appear to be significantly predicted by playing video games. It should be noted that they do not appear to specify if this is racing video games, drive em up video games or a combination of these. These attitudes in turn predicted self-reported intentions to speed which then goes on to predict self reported speeding. They do however state that if their model is broken down by gender then playing video games significantly only predicts self-reported speeding for males, and not for females.
Moving on, they also report that playing video games (again, not sure which, but I am guessing a combination of people to admitted to either racing or drive em up playing) significantly predicts attitude towards fun riding, which then predicts intention to fun ride, which then goes on to predict self-reported fun riding. It should be noted that in this case the predictive power of video games stayed for both genders.
Finally the authors report that there was no relationship between video game playing, attitudes towards drinking and driving, intention to drink and drive and self reported drinking and driving. This, the authors state, is in line with their predictions since computer games don’t feature drinking and driving (perhaps someone should tell them about GTAIV?), whereas they do feature ‘fun riding’ and speeding.
Based on this data the authors conclude that playing racing and drive em up games when young (before gaining a license) can predict risky driving behaviour. This is of course not an accurate statement of what their study shows. Based on their data a more accurate statement would be that playing these games appears to partly predict attitudes towards risky driving, which in turn appear to partly predict intentions to do certain risky driving behaviours, which in appears to partly predict people admitting doing those risky behaviours. This is of course not as an impressive a way of putting it, but it is more accurate.
The authors then conclude by briefly mentioning that their data doesn’t allow for causality to be established unambiguously and that only self-report measures where used. However these limitations are mentioned only in passing in the final paragraph. This is quite in contrast to the other two papers above which make sure to more clearly discuss their limitations.
Overall, out of the three papers I have summarized here, I have to say this last one was the hardest to follow in terms of presenting its data clearly and due to its use of the Theory of Planned Behaviour I was probably the toughest on it. It also happens to be the only one that showed only a really negative effect of playing games. This may lead to me being called biased. I hope however that I have been clear and fair with my comments and the summaries I have provided. Let me know what you think.