Longitudinal study on Canadian adolescents’ videogame play and aggression (Willoughby et al., 2011)

Just last week, I came back home to Montreal and went into a long hibernation.

On the subject of Canada, the Adolescent Development lab (Teena Willoughby, Brock University) have an in-press paper in Developmental Psychology on data from their longitudinal-sequential survey of Canadian adolescents. As far as I remember, this is the first Canadian longitudinal videogames study.


In the past 2 decades, correlational and experimental studies have found a positive association between violent video game play and aggression. There is less evidence, however, to support a long-term relation between these behaviors. This study examined sustained violent video game play and adolescent aggressive behavior across the high school years and directly assessed the socialization (violent video game play predicts aggression over time) versus selection hypotheses (aggression predicts violent video game play over time). Adolescents (N = 1,492, 50.8% female) were surveyed annually from Grade 9 to Grade 12 about their video game play and aggressive behaviors. Nonviolent video game play, frequency of overall video game play, and a comprehensive set of potential 3rd variables were included as covariates in each analysis. Sustained violent video game play was significantly related to steeper increases in adolescents’ trajectory of aggressive behavior over time. Moreover, greater violent video game play predicted higher levels of aggression over time, after controlling for previous levels of aggression, supporting the socialization hypothesis. In contrast, no support was found for the selection hypothesis. Nonviolent video game play also did not predict higher levels of aggressive behavior over time. Our findings, and the fact that many adolescents play video games for several hours every day, underscore the need for a greater understanding of the long-term relation between violent video games and aggression, as well as the specific game characteristics (e.g., violent content, competition, pace of action) that may be responsible for this association.

I was asked what my career goals are. Simple, we need Canadian representatives in the videogames effects research forum and someone to point out the academic importance of cultural differences in videogames (i.e. Canada, Quebec, Japan, U.S.A, Europe), starting with a content analysis.

This longitudinal study is actually part of a larger project examining adolescents’ lifestyle choices and videogames happens to be part of it. The authors pointed out some interesting limitations of prior longitudinal studies. First, prior studies have not examined sustained violent videogame play and its association to aggression. That is, whether adolescents played violent videogames throughout the study or they merely happened to play violent videogames when they were surveyed.

Tangent: This sounds quite interesting when I add in sustained marketing exposure to violent videogames, convincing adolescents (male ones) to play their (often violent) videogames, leaving the other nonviolent videogames in the mostly dark. It’s probably a good alternative to the selection hypothesis. Consider the huge marketing campaigns of the blockbusters (the Call of Duty vs. Battlefield series, Skyrim, Skyward Sword, etc.) and the retaining power of an online community (e.g. rankings, achievements, clans, matches, forums etc.). Even if the blockbusters games are not directly marketed towards minors, consider whether alternatives or more age appropriate marketing comes to their mind. Worse yet, the question is moot if there are no T-rated videogames that are immensely popular among adolescents, they would have went for the M-rated videogames no matter the amount of legislation on marketing towards minors. I don’t know if these premises are true, but it’s quite an interesting thought to explore.

A second limitation is the directional effects between violent videogame play and aggression. Most studies look at the effects in one direction. Two hypotheses are considered to explain and predict which variables move which ever variables in time. The socialization hypothesis posits that exposure to violent media causes individuals to be more aggressive, a unidirectional effect. The selection hypothesis posits that aggressive individuals are more likely to play violent videogames, another unidirectional effect. There are mixed support for the hypotheses, Willoughby and colleagues examined which hypothesis is the best answer.

A third limitation is the operationalization and assessment of violent videogame play. Using Moller and Krahe’s (2009) as an example, they critiqued the mathematics and the use of multiplications. The erroneous mathematics can lead someone who frequently (variable A) plays a lot of nonviolent videogames (variable B) looking like someone who played more violent videogames when averaged across a list of 40 games (the product of variable A and B) than someone infrequently play a lot of violent videogames.

Fourth limitation is the absence of nonviolent videogame play and the failure to directly test violent videogame play. The authors argued that if violent videogames has an inherent factor contributing to increases to aggression, then a nonsignificant finding between nonviolent videogames and aggression would bolster this argument… wait wha? As for the direct test, I’m not sure if I understand their argument. How can you directly test violent videogame play in isolation, surely an adolescent consume different sorts of media? I argue a gestalt approach (not sure if it’s the right concept) that our media experience is like soup, mixing television, internet, and videogames would change the taste of the soup, but not one element would stand. I need stronger arguments to suggest that videogames are the salient chunky parts of the media experience soup.

Finally, the “third/extraneous/confounding variable” limitation, an introduction textbook concept, that the relationship might be affected by a third or more variables. However, this does not necessarily mean that the videogames is blameless because of a confounding variable. It could be that the relationship between the variables are either moderated or mediated by this third variable. In any case, the longitudinal study (being a youth lifestyle study) has a bunch of interesting factors.


Participants: 1492 adolescents in Ontario, Canada were included in the analyses of this study. This is from the complete sample of 1771 adolescents. The reasoning for a partial sample is to include students who completed the survey at least twice in the study. Furthermore, they only included students who entered grade 9 in 2004 and completed the surveys each year until 2008 at grade 12. So, they have 4 waves or data time points at one year interval. Again, this is a partial sampling from the complete study where it originally started in 2003 and ended in 2008. The sample’s demographics are representative of Canada’s population, equal gender ratio, and the non-participation rate is mostly due to student absenteeism (e.g. illness).


Each measure was given the same questionnaire packet to the students each year.

Direct aggression: A composite of two scales: a 9-item 4-point scale measure. The other scale is a 4-item on a 5-point scale which was converted into a 4-point for statistical analysis and to create a composite score of direct aggression.

Violent videogame play: A “yes or no” if they played action (e.g. God of War) or fighting (e.g. Mortal Kombat) videogames. A ratio score is created by the number the participant reported to have played to the total number of waves he/she completed (e.g. participant reported playing violent videogames in all four waves is given a score of 1, whilst a participant who reported in one wave and has participated in three waves is given a score of 0.33) which they call a sustained play. Whilst I most agree that most of the violent videogames are within those genres, excluding RTS, RPG games seemed quite fair as they vary in violent content. However, I can speculate why this longitudinal project opted for a dichotomous measure: limited time for the high school students and space for other measures. I must say that a ‘yes or no’ assessment is very limiting in terms of accuracy and meaning, it would’ve been more meaningful if they’ve played violent videogames on a temporal basis (rarely to regularly). Furthermore, if this is a lifestyle study why did they not include television, internet or any other media consumption measures into their analyses? This is rather odd.

A second measure was included in the grade 11-12 where students were asked on a 5-point frequency scale for two items: “on a average day, how often do you play action games?” and “…do you play fighting games?”

Nonviolent videogame play: A “yes or no” if they played puzzle/art/building sims/quiz videogames. The scoring method is the same as the violent videogame play. The same for the second measure.

Overall videogame play: Two items on how the number of hours the student plays videogames on average on school days (1) and during the weekends (2). Okay, that might’ve addressed my earlier criticism, but how can I know that one student has spent 90% of videogame play time towards a puzzle game (say Portal)?

The following measures were entered into their analyses as the group of third variables:

Demographics: student’s sex, parent’s education, number of computers in home, at-risk background (e.g. learning disabilities, foster care, divorce, parents used/using drugs, etc.) on a yes or no basis.

Academic marks: self-report of grades on a 5-point scale.

Depressive symptoms: 20-items on a 4-point frequency scale from the classical Center for Epidemiological Studies Depression Scale (CES-D).

Delay of gratification: a 5-item on a 5-point frequency scale.

Peer deviance: a 10-item scale assessing whether their friends commit various deviant behaviours (e.g. drinking alcohol), assessed on a 5-point ratio (from none to all of them) scale.

Sports Involvement: a 2-item scale on a 5-point frequency scale asking: how often they participated in organized sports in school and outside school. A bit disappointed about this one, I was hoping which sports they would be involved in, like hockey, football soccer or perhaps wrestling.

Friendship quality: 18-items on a 4-point frequency scale from the Inventory of Parent and Peer Attachment (e.g. I like to get my friends’ point of video on things I’m concerned about”).

Parent-adolescent relationship quality: 17-items on a 4-point scale frequency scale from the Inventory of Parent and Peer Attachment (e.g. “My father helps me to talk about my difficulties”).

Parental control: 6-items on a 4-point frequency scale. E.g. “Do you need your parent’s permission to stay out late on a weekday evening?” This looks more like parental behavioural control.

School culture: 18-items on a 5-point agreement scale which assessed for perceptions of school involvement, peer behavioural values, and instructional management.


With some many variables in the analysis, the authors wrote many paragraphs explaining the statistical and mathematical steps leading to their conclusions. The statistical analysis is called a latent growth curve modeling, which is the appropriate approach in analysis their longitudinal data. There are so many details which the following is a simplification of their results.

The figures in the article are partials as I imagined the sheer number of lines between 26 variables would be incomprehensible. Quite interestingly, there were no gender differences or did not act as a significant moderator in their analyses. Perhaps they can release an interactive flash of the figure would be most instructive for the readers.

Their first analysis examined the long-term association between sustained violent videogame and aggression. Their results indicated that higher sustained violent videogame play leads to steeper increase in aggression than those who played less violent videogames. Nonviolent videogames play had no statistical significance. They further examined by adding overall videogame play and found that sustained violent videogame play is still predictive of aggression behaviour over time.

A third analyses by adding the group of third variables has shown that the predictive relationship still holds statistical significance.

Their second analysis tested which hypotheses, socialization and selection, are given support or not using cross-lagged models to examine the direction of effects in time. The authors started looking at the simple association between aggression and violent and nonviolent videogame play, they found that playing violent videogames predicted greater aggressive behaviours from grade 9 to 10, 10 to 11 (at a statistical trend), and 11 to 12, even after controlling the stability of aggression (as aggressive behaviours tend to be somewhat stable across ages) and when controlling for overall videogame play, it still holds significance from grade 9 to 10 and 11 to 12. They also found significance when aggressive behaviours predict later violent videogame play, from grade 9 to 10, 10 to 11 and when controlling for overall videogame play significance held only for grade 10 to 11. An interesting finding, which the abstract did not indicate, is that nonviolent videogame play predict lower aggression from grade 10 to 11 and grade 11 to 12, even controlling for overall videogame play. These findings lend support for the socialization hypothesis and some support for the selection hypothesis.

The statistical significances changed when the group of third variables are entered. The predictive significance of the socialization hypothesis remained where sustained violent videogame play predict later aggressive behaviours, from grade 9 to 10, and grade 11 to 12. Sustained nonviolent videogame play remained significant from grade 11 to 12. The statistical significances supporting the selection hypothesis went into nonsignificance. The authors argued that the socialization hypothesis is uniquely supported for the violent videogame, but not nonviolent. Wait… I don’t understand why they made such conclusion when there is a statistical significance for nonviolent videogame play lowering aggression.

Remember the second measure for videogame play for grade 11-12? Their findings made it more clear and meaningful in regards to socialization hypothesis. Violent videogame play in grade 11 predicted higher aggressive behaviours in grade 12, accounting for stability in aggression. Nonviolent videogame play, however, was non-significant and so were aggressive behaviours for later violent videogame play. When they added the group of third variables, the violent videogame play predictor held its significance.


The take home message is that violent videogame play would lead to later increases in aggressive behaviours in the long term of an adolescent high school life. In contrast, aggressive individuals are not likely to play more violent videogames. Being attracted to violent media is another issue, as stating an interest and actual behaviours are not always congruent. In the face of gender, parental education, risk factors, computers at home, academic performance, depressive symptoms, aggression, delay of gratification, peer deviance, sport involvement, friendship quality, parent quality, parental control and school culture, it still holds to be a significant, yet small and controllable, factor of aggressive behaviours.

Nevertheless, the authors argued that due to the results found from the nonviolent videogame play, it seemed that this is probably because nonviolent videogames differ in more than just content and, citing their experimental studies, other characteristics such as pace of action and competitiveness may be involved. I am constantly bothered that they kept writing that nonviolent videogames does not increase aggression, which seemed logical, but I am just concerned that they have not discussed the positive outcome and the support it lends to the prosocial videogames research thread (see Greitemeyer, 2011).

The authors reported their self-report of videogame play as an important limitation, they acknowledged the videogame play issue I ranted earlier. They also argued for more informants (i.e. peers, parents, teachers) in reporting videogame play use. This study started in 2003 and ended five years later, the measure they used (given my earlier frowning) would have been the best they can get. At the present time, videogame consoles and PC games connected to the net or to online services (e.g. Steam, Xbox Live, Playstation network) can provide a more objective means of measuring playtime. I am a bit familiar with Steam and it showed me the total number of hours I played on a particular game, when I last played and the number of hours play on the last two weeks. I’m not sure about the other online service, but I gather that the companies are holding valuable scientific data on players’ habits.

Returning to their examination of the socialization and selection hypotheses, I recall Michael Slater’s Downward Spiral Model. Slater et al. (2003) found similar findings in that violent media exposure led to increases in aggressiveness later. However, they argued that

The effects of that violent-media content on the youths who use it on aggressiveness is both concurrent and to a lesser extent prospective. Consistent with the downward spiral model, then, these effects can be viewed as mutually reinforcing (aggressiveness leading to violent-media use concurrently and violent-media use to aggressiveness both concurrently and prospectively), directional (leading both to increased violent-media content use and to increased aggressiveness), and, at least by inference, cumulative (these reinforcing effects can be expected to accumulate over time given continued, directional mutual influence).

Willoughby, Adachi and Good did not discuss their findings in relation to the Downward Spiral Model. I thought it would be interesting to hear their thoughts on that.

Willoughby, T., Adachi, P. J. C., & Good, M. (2011). A longitudinal study of the association between violent video game play and aggression among adolescents. Developmental Psychology . Doi: 10.1037/a0026046

1 thought on “Longitudinal study on Canadian adolescents’ videogame play and aggression (Willoughby et al., 2011)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s