Reply to Stoet and Geary: Effects of gendered behavior on testosterone, not sex differences, as research focus

In our report, we experimentally assessed gender-related modulation of testosterone (T) (1). Our focus—unlike the focus and basis for Stoet and Geary’s letter (2)—was not on sex difference. Stoet and Geary conflate sex with gender and with sex difference in their misread of our paper (1), resulting in confusion about the phenomena and the goal of the study. Indeed, scholars have repeatedly identified how commonly researchers and pundits assume all research on “sex” is on “sex difference,” and exclude gender and other valuable ways sex might be studied (3, 4).

Our statistical analyses (1) are robust: certainly so regarding Stoet and Geary’s (2) letter’s view of our figures. Our experiment was aimed at detecting effects of acting on T; we were not aiming to demonstrate preacting increases in T (which were, anyway, controlled for in our analyses) (1). We show a statistically significant effect of acting on T and a significant interaction of condition with gender/sex: women but not men showed the significant effect.

The null effect of acting on T in men is arguably the least-interesting finding of our experiment and, like any null finding, has many possible explanations. We go into several in our paper. For example, we acknowledge (1): “…our sample size was not large, which is a limitation….” Sample sizes differed for men and women, given our commitment to rigorous methods and excluding women with exogenously affected hormones (e.g., by hormonal contraceptives), something often ignored by hormone researchers. Our male sample size was larger than the female one and so an effect in men was more likely, if effect sizes were similar by gender/sex and sample size was the problem.

We have previously shown that the documentary is hormone-neutral (5, 6), and it is well known that T decreases with time (7, 8); the decrease in women’s T is thus most likely attributable to time, and not the documentary, as Stoet and Geary’s letter (2) suggests. Studies do not always show a decrease in T from neutral conditions (indeed, unlike in our laboratory, many hormone studies do not even include controls); instead, the key contrast is typically whether the experimental condition increased T relative to the neutral condition (7, 9, 10). In our report (1), this was the case for men, although not statistically significantly so. As such, it is certainly possible that the larger intra-assay coefficient of variation in the men added noise and obscured significant effects. It is not, however, plausible given that men actually showed smaller SEs than the women. Moreover, our analysis of percent change is a within-subject measure that statistically controls for assay technique.

“Gender socialization can contribute to variation in human testosterone levels” is one of our main points (1). In contrast to Stoet and Geary’s letter (2), we do not claim to prove this. We are excited that our science opens doors to new research questions about gender socialization and human physiology, and look forward to programs of research exploring this phenomenon.