Analyzing brain data by sex: Are we asking the right question?

Abstract

The decades old hypothesis that sex effects on the brain result in ‘female’ and ‘male’ phenotypes governs conventional analyses by sex. In these (e.g., Student’s t-test), the null hypothesis is that males and females belong to a single population (or phenotype), and the alternative hypothesis is that they belong to two different populations/phenotypes. Yet, evidence that sex effects may be opposite under different conditions raises a third hypothesis – that both females and males may manifest each of the two phenotypes of a brain measure. Here we applied a mixture analysis, which can test this latter hypothesis, and Student’s t-test to 289 MRI-derived measures of grey and white matter from 23,935 human brains. Whereas Student’s t-test yielded significant sex/gender differences in 225 measures, the mixture analysis revealed that 282 brain measures were better described by the hypothesis that women and men sample from the same two phenotypes, and that, for the most part, they do so with quite similar probabilities. A further analysis of 41 brain measures for which there were a ‘female-favored’ and a ‘male-favored’ phenotype, revealed that brains do not consistently manifested the male-favored (or the female-favored) phenotype. Last, considering the relations between all brain measures, the brain architectures of women and men were remarkably similar. These results do not support the existence of ‘female’ and ‘male’ brain phenotypes but are consistent with other lines of evidence suggesting that sex category explains a very small part of the variability in human brain structure.