Some readers have pointed out an apparent discrepancy in my post about the NSF-funded Iowa State University study on VR and gender that I covered last week: The University's summary characterizes the gender difference in stark terms: The study found that the women test subjects ended a VR game early twice as often as men, and "reported a sickness intensity that was 40% higher".
However, the abstract of the peer-reviewed study is not as stark, saying: "We summarize the literature and conclude that women experience more cybersickness than do men, but that the size of the gender effect is modest."
How can a 40% sickness intensity be described as "modest"?
I checked back with Dr. Jon Kelly, who led the research. Asking him, is it fair to say that the gender difference is "modest" in terms of statistical analyses -- but is still a very important consideration for VR developers and companies when attempting to launch a mass market VR consumer product:
"Right," he replied. "I am talking about the effect in statistical terms, which considers the difference between the two groups (i.e., how much sicker are women than men on average) relative to the variation within the groups (more variation within each group corresponds to a smaller effect size). I mentioned previously that there are many women who experience zero cybersickness, even though the group average is above that of men. That is indicative of a moderate effect."
A gender difference in VR, as the report states, is more discernible when a study is done with a large, gender balanced number of volunteers -- which may also explain why a gender difference isn't noted in studies with a smaller pool of volunteers:
[A] researcher would need to enroll 258 participants (with equal numbers of men and women) to have a 95% chance of detecting a gender difference or 176 participants to have an 80% chance of detecting the effect. It is therefore unsurprising that there are many reports of non-significant gender comparisons in a field in which sample sizes of 20-40 participants are quite common. This also further emphasizes the need to rely on quantitative meta-analysis that considers sample size along with effect size.
The Iowa study was conducted with 150 participants, so on the higher side for studies of this kind. However, since VR devices and software target millions of consumers, the gender effect is likely to be even more clear and stark.
At minimum, it may be a good idea to follow Kelly's advice on minimizing cybersickness in VR:
“[S]ome of my research on the gender effect indicates that 1) the effect is relatively small, and 2) the effect is partially explained by differences in prior experience of visually-induced sickness (e.g., screen-based games, movies). It's certainly a topic worth investigating, but it's worth a reminder that there are vast individual differences in cybersickness susceptibility even within a given gender.”
Kelly's research is ongoing, so more soon-ish. Hopefully other researchers from the VR industry will finally join in.
Comments