Abstract: A cross-sex shift model of human sexual orientation differences predicts that homosexual men should perform or score in the direction of heterosexual women, and homosexual women in the direction of heterosexual men, in behavioral domains such as cognition and personality. In order to test whether homosexual men and women’s cognitive performance was closer to that of heterosexual men or that of heterosexual women (i.e., sex-atypical for their sex and closer to that of the opposite-sex), we conducted a multivariate meta-analysis based on data from our previous meta-analysis (Xu, Norton, & Rahman, 2017). A subset of this data was used and comprised 30 articles (and 2 unpublished datasets) and 244,434 participants. The multivariate meta-analysis revealed that homosexual men were sex-atypical in mental rotation (Hedges’ g = −0.36) and the water level test (Hedges’ g = −0.55). In mental rotation, homosexual men were in-between heterosexual men and women. There was no significant group difference on spatial location memory. Homosexual men were also sex-atypical on male-favoring spatial-related tasks (Hedges’ g = −0.54), and female-favoring spatial-related tasks (Hedges’ g = 0.38). Homosexual women tended to be sex-typical (similar to heterosexual women). There were no significant group differences on male-favoring “other” tasks or female-favoring verbal-related tasks. Heterosexual men and women differed significantly on female-favoring “other” tasks. These results support the cross-sex shift hypothesis which predicts that homosexual men perform in the direction of heterosexual women in sex differentiated cognitive domains. However, the type of task and cognitive domain tested is critical.
Keywords: Sexual orientation Meta-analysis Sex differences Cognition Spatial Verbal
Discussion
This analysis produced three main findings. First, homosexual men were sex-atypical in studies measuring mental rotations, the water level test, male-favoring spatial-related tasks, and female-favoring spatial-related tasks. That is, homosexual men’s cognitive performance was closer to that of heterosexual women than heterosexual men. Second, homosexual women were no different to heterosexual women, despite some tendency to be sex-atypical in certain domains (e.g., female-favoring verbal-related tasks). Third, there was considerable heterogeneity in the data as we found in our original meta-analysis.
The magnitude of the effect sizes revealed in the current multivariate meta-analysis was similar to that of our prior univariate meta-analysis. Once again, we found that homosexual men showed a cross-sex shift in male- and female-favoring spatial tasks, which is consistent with our prior demonstration that effect size was the highest for spatial tasks in men (Xu et al., 2017). The results for women were also consistent with previous work, suggesting that homosexual women are by and large sex-typical in most cognitive domains. However, given that the studies included in the current multivariate meta-analysis are a subsample of those from our prior study, the reduced number of studies may have contributed to the non-significant results found in women.
Our results should not be interpreted as indicating that homosexual men performed exactly the same as heterosexual women. In other words, we find little evidence of a complete sex inversion in this behavioral domain among homosexual men. Task type and cognitive domain are clearly critical. Traditionally, male-favoring spatial tasks (particularly mental rotation and spatial relations) appear to be most sensitive to sexual orientation differences. This is most likely due to the fact that they show robust general sex differences (Voyer et al., 1995) and that this domain provided the greatest number of studies. The cross-sex shifted pattern displayed by homosexual men is consistent with that found in several other behavioral domains such as sex-typed behavior and personality (Bailey et al., 2016). However, the effect sizes found here are much smaller than for other traits associated with sexual orientation, such as childhood gender nonconformity (Bailey et al., 2016).
In general, the body of work supports the prenatal androgen theory which predicts that homosexual men should show cross-sex shifts in sex differentiated behavioral domains in line with the atypical shift in their sexual partner orientation (Ellis & Ames, 1987). As the present study did not directly measure prenatal androgen levels, caution must be exercised in interpretation. However, some remarks regarding the patterns reported here and their relationship to the prenatal androgen model are worthwhile. The evidence for a cross-sex shift in cognition is inconsistent with research using putative markers of prenatal androgen exposure. For example, digit ratio (2D:4D) is a marker ascribed to the actions of prenatal androgen levels. However, nonheterosexual women have more masculine digit ratios (indicating greater exposure to prenatal androgens) than heterosexual women, but there is no significant difference in digit ratios between heterosexual and nonheterosexual men (Grimbos, Dawood, Burriss, Zucker, & Puts, 2010). Similarly, differences in handedness are a feature sometimes ascribed to the actions of prenatal testosterone acting on developing brain asymmetries. However, both nonheterosexual men and women are significantly more likely to be non-right-handed than heterosexual men and women rather than cross-sex shifted (Lalumière, Blanchard, & Zucker, 2000). As mentioned earlier, sexual orientation-related differences in sex-typed behavior (e.g., play and peer preferences), personality, and sexual orientation target preference itself (the preference for males or females as sexual and romantic partners) are much larger than cognitive differences (Bailey et al., 2016). Some of these traits (sex-typed behaviors) may show substantially larger sex and sexual orientation-related differences during childhood than other traits (cognition). Thus, it is possible that these discrepant findings where some traits show cross-sex shifts (cognition, sex-typed behavior) while others do not (somatic traits), or where cross-sex shifts are found in some traits in females (digit ratio) but not in males, point to a possible patterning of causal pathways by trait, sex, and developmental stage.
As mentioned before, the number and extent of critical periods for prenatal sex hormone actions might be important. There is a growing theoretical suggestion that males may have more than one critical period (e.g., prenatal, early postnatal, and pubertal), while females may have several but longer sustaining “sensitive periods” in which sex hormones and other developmental processes may act over a longer time period to influence behavioral outcomes (McCarthy et al., 2018). It is important to note that there are no longitudinal studies linking direct measures of prenatal androgens, such as amniotic levels of fetal testosterone, with later sexual orientation and cognition in humans. Such prospective studies would provide the critical test of the prenatal androgen model. Such studies will also need to control for important confounders or third factors such as genetics (e.g., genetic correlations between the traits in question over time). Such third factors might also be more important in the causal association between male sexual orientation and associated behavioral traits. One such factor is the well-known fraternal birth order effect (FBO; Blanchard, 2018). This refers to the robust finding that homosexual men have more older brothers than heterosexual men, an effect ascribed to maternal immune responses triggered by carrying successive male fetuses which affects sexual differentiation of the brain of later born males (Bogaert et al., 2018). One study has reported no significant association between FBO and spatial cognition in heterosexual and homosexual males (Rahman, 2005b; cf. Bogaert, 2003a, 2003b).
The current meta-analysis had several important limitations. Many of these are similar to those in our original meta-analysis so will not be repeated here. However, specific to the present analysis, we note that the heterogeneity between studies was high given the broad 95% confidence intervals. We have suggested that methodological variation (e.g., cognitive domain differences) is a significant contributor to this heterogeneity. Second, the number of studies for some cognitive domains included in the multivariate meta-analysis was small, which generally resulted in broad 95% confidence intervals (e.g., spatial location memory, female-favoring tasks, and male-favoring other tasks). Broad 95% confidence intervals indicate considerable uncertainty in effect sizes. Thus, more research with appropriate sample sizes is needed and this may change the conclusions. Finally, we were unable to find sufficient numbers of studies which reported within-group correlations between multiple cognitive tasks (only four studies reported the correlations). This latter point is of note for future research because having within-group correlations between tasks would permit the calculation of multivariate effect sizes (such as Mahalanobis D or other indices of multivariate distances). Such metrics would allow tests of the overall magnitude of sexual orientation differences where the groups differ along many variables of interest or where the construct is multidimensional (Del Giudice, 2013).
Posts
No comments:
Post a Comment