The Human Brain Is Best Described as Being on a Female/Male Continuum: Evidence from a Neuroimaging Connectivity Study

Abstract

Psychological androgyny has long been associated with greater cognitive flexibility, adaptive behavior, and better mental health, but whether a similar concept can be defined using neural features remains unknown. Using the neuroimaging data from 9620 participants, we found that global functional connectivity was stronger in the male brain before middle age but became weaker after that, when compared with the female brain, after systematic testing of potentially confounding effects. We defined a brain gender continuum by estimating the likelihood of an observed functional connectivity matrix to represent a male brain. We found that participants mapped at the center of this continuum had fewer internalizing symptoms compared with those at the 2 extreme ends. These findings suggest a novel hypothesis proposing that there exists a neuroimaging concept of androgyny using the brain gender continuum, which may be associated with better mental health in a similar way to psychological androgyny.

Beyond “Sex Prediction”: Estimating and Interpreting Multivariate Sex Differences and Similarities in the Brain

“…1A to those in Fig. 1B, which -as in other previous studies 14,30 -divides the originally continuous outcome into three, instead of two, equal segments. This simple modi cation reduces the %CC (Supplementary Tables 3A and 3B), thus illustrating how dichotomization leads to larger between-group differences through a suppression of the within-groups variation.…”

“…Therefore, it is apparent that -despite working with identical data from the same individuals-the different algorithms tested in the present study do not provide directly exchangeable outcomes or identify a single, coherent, and reproducible subset of brain features as the source of the males-females multivariate differences in GMVOL (neither in the raw dataset or in the PCP dataset). These observations are clearly consistent with the lack of agreement observed between the few studies that tried to identify the neuroanatomical features that could best distinguish the brains of females and males 4,10,12,14 (for a comparative review, see 41 ), and with evidence suggesting that ML algorithms rely on different brain features when classifying different subpopulations of females and males 44 . Together, these sources of empirical evidence directly challenge the binary sex views of human brains based on a global interpretation of %CC scores.…”

“…Consequently, %CC scores preclude describing the outcome's distribution, summarizing the individuals' scores or estimating the actual size of between-group differences. These limitations are overcome when the continuous output of ML algorithms is not discretized, and when females and males are treated as empirically-established distributions that spread at different probability levels within particular ranges of the outcome's continuous space 14,[32][33][34] . Therefore, we rst described how males and females mapped onto the continuous output provided by ML algorithms (the PCAM continuum) and then we assessed their distributional divergences to obtain proper estimates of the size of the male-female multivariate differences in GMVOL Figure 2 depicts the kernel density estimates (KDE) of the PCAM distributions yielded by each algorithm in each dataset.…”

“…When the output of ML algorithms is not discretized, multivariate information about the brains of females and males can be condensed in a single continuum 14,[32][33][34] . The present study shows that, by assessing how females and males differentially occupy this empirically-de ned unidimensional space, the size of their multivariate differences can be estimated on a standardized [0,1] scale or in ordinal/ distributional terms.…”

“…In the sex differences eld, most ML applications have focused on classi cation tasks. These studies have effectively shown that several predictive algorithms can exploit anatomical and/ or functional features to ascertain whether a brain belongs to a male or to a female with an ≈80-90% accuracy [2][3][4][5][6][7][8][9][10][11][12][13][14] . Because this kind of "prediction" only informs us about what is already known (to which sex category each particular sampled brain belongs), the interest and relevance of these ndings resides more in the brains' ability to be classi ed than in the classi cations obtained.…”