Casual games for better cognition?

Are casual video games good “exercise for the brain”? TL;DR A study last year by Oei & Patterson (2013) found significant benefits after casual video game training on several measures of cognition, in areas of attention, cognitive control and working memory. But, whether you should pick up a new iPhone game just for brain “exercise” is still an open question.

With the current popularity of “brain-training” games (think of Lumosity’s ubiquitous ads), it is interesting to note that relatively little research has tested if video games, especially those targeting casual gamers, provide any cognitive benefits. The main exception would be for action games (first-person shooters), which have been found to produce improvements in visual attention (see Green & Bavelier’s 2003 Nature paper).

Recently, more studies have been published which have tested the cognitive benefits of a range of casual video games on wider range of cognitive skills. Last year, Oei and Patterson, of Nanyang Technological University in Singapore, published a study in PLoS ONE (available online for free). This was an interesting study, where participants completed baseline tests (of attention, spatial memory, working memory, and visual search). Afterwards, they randomly assigned participants to play one of 5 games (action, match-3, simulation, hidden-object and “memory matrix”). Each of the games was downloaded from iTunes, and participants played the games using their own iPod or iPhone. Over the following four weeks, they were asked to play their assigned game for 1 hour per day. Links to each of the games used are given below (this list is my best attempt to match the games based on the description in the paper to games published on the iTunes App Store).

After four weeks (and 20 hours of game play), participants were retested on all of the tasks.  According to the authors, they found that:

  • In the action game group, participants showed reduced attentional blink, better cognitive control and better performance when detecting changes in a large set of objects.
  • Several groups (match-3, spatial memory and hidden-object games) did better in visual search after the intervention.
  • Both the action and match-3 groups demonstrated higher complex verbal span (working memory) after game training.

Overall, this was an interesting study, and an interesting design. By using commercially available games, and having users play those games on their own devices, they have greatly improved teh external validity of the study (how well the research applies to regular people and real-life situations). And, while the effects of these games may be modest, this type of research gives us some hope that future research can replicate these effects, and begin to study how casual games can provide more cognitive benefits, while still appealing to users as entertainment.

After reading the manuscript, I did have a few concerns about the methodology and analyses used by the authors, which I think should be addressed by future research. The rest of this post will dig into these details, but overall I think these concerns are not critical (that is, from what I can see, the results cited above seem to be appropriate):

Figure 2 from Oei AC, Patterson MD (2013) Enhancing Cognition with Video Games: A Multiple Game Training Study. PLoS ONE 8(3): e58546. doi:10.1371/journal.pone.0058546

Tracking of video game experience, and adherence: The authors report that participants logged their use of the assigned video game on an online database, but they do not present any data showing that participants improved in the game that they were assigned to. It would be useful to know if participants who showed greater improvements in the assigned game also showed greater improvement in cognitive skills. Also problematic is that the authors do not provide any data on what types of games participants had experience with before the experiment, or other games they may have played during the four weeks of the intervention. I would feel much more comfortable if we have a better idea of what participants were doing in the 20 hours of training, and how much skill they developed in that time.

Choice of dependent variables: Some of the dependent variables used by Oei & Patterson have clear relationships to previous research (such as their measures of attentional blink). However, in one case the rationale for the choice of a dependent measure is not clear.  In the Filter Task, (shown in their Figure 2, above), they describe that: Participants were presented with an array of targets (red rectangles) and distractors (blue rectangles) in different orientations against a white background. Targets and distractors appeared in 10 different conditions of different quantities. These conditions varied from 2 to 8 targets as well as 2 to 6 distractors with the constraint of a maximum of 8 items in each array.”  

While participants were tested with what appears to have been 10 different conditions in the Filter Task, the authors only report analyses with 2 conditions: with 2 targets and 6 distractors, and with 8 targets (and no distractors).  They explain that these conditions were chosen to represent participants abilities to filter out distractors (in the 6 distractor condition) and to track multiple targets (in the 8 target condition). But, it is not clear if the other conditions were also examined, and why the results from those conditions were not reported. In this case, I would have appreciated a general description of the pattern seen in those data (even if they were not described in detail in the paper).

Analyzing group differences in the absence of a main effect or interaction: Most of the analyses reported used factorial ANOVAs to test how participants in each group improved (or did not) after the 20-hour intervention. An ANOVA allows us to test the influence of an independent variable (such as video game group) on a dependent variable (such as accuracy in the attentional blink task). When more than one independent variable is used (such as video game group and time (pre- versus post-test)), ANOVA allows us to measure the overall effect of each variable (the main effect of video game, or the main effect of time). If a main effect is significant, we believe that there is sufficient evidence to claim that at least one treatment group significantly differs from one other group. The ANOVA also lets us test the interaction between variables. If an interaction is signficant (say, for the interaction of time and video game group), this would indicate that the effect of one variable is not the same for every group in the second variable (so, that the amount of change from pre-test to post-test is not the same for every video game group).

For this research study, the most important prediction would seem to be for these interactions of video game group and time: if casual games have an effect on cognition, and different games have different effects, then the time x video game group interaction should be significant.  This was true for attentional blink (see Figure 5), the Filter Task (see Figure 6), visual search accuracy (see Figure 7) and complex span (see Figure 10). There were no main effects of group or significant interactions involving group for visual search reaction time and spatial memory accuracy, but even without these interactions, the authors went on to analyze each group independently for these measures (see Figures 8 and 9).

Figure 9 from Oei AC, Patterson MD (2013) Enhancing Cognition with Video Games: A Multiple Game Training Study. PLoS ONE 8(3): e58546. doi:10.1371/journal.pone.0058546

In general, I would not recommend this approach: if the interaction of video game group with time (and perhaps a third variable) is not significant, this indicates that we do not have a good reason to think that there were real differences in the effectiveness of the games. So, when we see differences between video game groups (see Figure 9, shown above), if the interaction is not significant, we assume that these differences are consistent with differences we would expect to see just based on random chance.

The results from the spatial memory task were not included in the abstract, but were referenced in the discussion of the article, and were given prominent figures in the results section of the article. Overall, this might mislead readers into believing that the group differences in those tasks were reliable and significant. I would have preferred that the authors indicate that there were no significant interactions for these measures (spatial memory accuracy and visual search reaction time), but that examined each group separately anyway, as exploratory analyses.