Neuroscience and Sex/Gender: Looking Back and Forward

Translating research from nonhuman mammals to humans

The influences of testosterone on brain structure and later behavior occur during critical periods of development. Consequently, testosterone must be present during a particular developmental window to have its effects, and the effects persist after the hormone is gone. In mammals, the critical periods correspond to times when testosterone is higher in male than female animals. These periods begin prenatally, when the SRY gene on the Y chromosome causes the gonads to differentiate as testes, and they begin to produce testosterone. In humans, the fetal testes become active at approximately week 7 of gestation, and testosterone is markedly higher in male than in female fetuses from approximately gestational weeks 8–16 or 24 (Smail et al., 1981). After birth, there is a second surge of testosterone that is larger in male than in female infants, particularly from approximately weeks 4–12 postnatal (Fig. 1) (Forest et al., 1973; Kuiri-Hanninen et al., 2011). These two periods are the presumed critical periods when testosterone might influence the development of sex differences in human brain and behavior.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Concentrations of serum testosterone in boys from conception to puberty.

The behaviors that are affected by early testosterone exposure in nonhuman mammals include reproductive behaviors, as well as other behaviors that differ on average for male and female animals. For example, male and female rats differ on average in juvenile sex-typed play, physical aggression, parenting behavior, and performance in spatial mazes, and all of these behaviors have been found to be influenced by early manipulations of testosterone (Hines, 2004). Similarly, the female offspring of rhesus macaques who were treated with testosterone during pregnancy show more male-typical patterns of juvenile play, as well as reproductive behaviors (Wallen, 2005).

Because early testosterone exposure influences characteristics that show sex differences, it was important to identify human behaviors that differ by sex. It also was important to determine the sizes of the differences because larger differences would be more likely to show effects of testosterone exposure than smaller differences. The metric typically used for size is Cohen's d, the difference between mean values for males and females divided by the SD (Cohen, 1988). Conventionally, d values of ∼0.2 are considered small, those of 0.5 medium, and those of 0.8 large. Decades of research, often using meta-analytic techniques that allow results of numerous studies to be combined to get reliable estimates of effect sizes, suggest that most sex differences in human behavior/psychology are small to negligible in size (Hyde, 2005). However, large sex differences in human behavior (d ≥ 0.8) have been documented for some characteristics. These include scores on some specific measures of empathy (higher in females), physical aggression, and social dominance, and for performance on a specific 3D mental rotation task (all three higher in males). The sizes of these sex differences, in SD units (d) (Cohen, 1988) are shown in Figure 2. This figure also shows the size of the sex difference in height (d = 2.0), providing a familiar comparator. Although these differences are large by Cohen's standards, they are only approximately half the size of the sex difference in height.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Effect size (d) values for human behaviors/psychological traits that show large sex differences. The effect size for the sex difference in height (d = 2.0) is included as a familiar comparator.

Figure 2 also shows the effect sizes for other behaviors that show sex differences as large, or larger, than that in height. For example, a population study using a questionnaire measure of children's gender-related play found a sex difference of d = 2.8 (Hines et al., 2002; Golombok et al., 2008). Children's sex-typed toy preferences (e.g., for dolls vs vehicles) also show large sex differences (d ≥ 1.8). In addition, sexual orientation and gender identity show large sex differences, with males more interested in female sexual partners and more likely to identify as men or boys, compared with females. Both of these sex differences are very large (d > 6.0 and d > 10.0, respectively) (Hines, 2015).

All three of the human psychological/behavioral characteristics that show particularly large sex differences (childhood sex-typed play, sexual orientation, and gender identity) have been found to relate to early testosterone exposure. Sex-typed play has been studied more extensively than any other human behavior in this context, and at least 10 independent research groups have reported a link to prenatal testosterone exposure (Hines, 2015). For example, girls with classic congenital adrenal hyperplasia (CAH), a genetic disorder, experience elevated testosterone exposure prenatally, and they have consistently been found to show increased male-typical, and reduced female-typical, play. These findings have been reported in studies that observed toy choices in a play room and in studies that used questionnaires or interviews. They also have been found in studies using unaffected female relatives as controls and in studies using matched controls. In addition, the severity of the CAH disorder in terms of either phenotype or genotype predicts the degree of behavioral change. Studies of the children of women who were treated with hormones during pregnancy also suggest androgenic influences on sex-typed play. Similarly, XY individuals with complete androgen insensitivity syndrome have functioning testes but a cellular inability to respond to testosterone, and they show female-typical play patterns. Thus, although this evidence comes from studies of clinical populations, rather than experiments involving random assignment to testosterone or placebo treatment, the findings converge on the conclusion that prenatal testosterone concentrations influence children's sex-typed play behavior. Regarding sexual orientation and gender identity, findings similarly suggest increased male-typical outcomes in females with CAH and reduced male-typical outcomes in XY females with complete androgen insensitivity syndrome. Similar evidence regarding other characteristics that show sex differences, including performance on mental rotation tasks, has been largely inconsistent, perhaps because these measures show smaller sex differences.

Although it is well established that children's sex-typed play is influenced by early testosterone exposure, there also is extensive evidence that the social environment and children's cognitive understanding of gender play a role in the same outcomes (Hines, 2015). For instance, parents, peers, teachers, and strangers reward children for playing with sex-typical toys and engaging in sex-typical activities. In addition, after children learn that they are girls or boys, they value engaging in the activities that they identify as appropriate for their sex, and they self-socialize based on social information as to what is sex-appropriate behavior.

These findings raise the question of whether external socialization or self-socialization is altered for girls with CAH. When they are observed in a laboratory playroom, with access to a range of sex-typed and sex-neutral toys, parents of girls with CAH encourage them to engage with female-stereotyped toys, such as dolls, more than they do their unaffected daughters (Pasterski et al., 2005). Parents of girls with CAH also report, however, that they encourage them to engage in male-typical activities in their day-to-day life (Wong et al., 2012). This probably occurs because parents tend to encourage their children to engage in the activities the children enjoy, and girls with CAH like male-typical activities. Nevertheless, this encouragement may further masculinize the girls' behavior. Self-socialization also has been found to be altered in girls with CAH. They are less likely than other girls to model the behavior of other females, and they are less likely to engage with toys that they have been taught are “for girls” (Hines et al., 2016). These findings suggest that prenatal androgen exposure may influence children's toy preferences not only by influencing brain development prenatally, but also by changes in postnatal socialization, including parental socialization, and children's self-socialization, of sex-typed behavior.

Developmental systems perspective

Developmental scientists view outcomes, such as gender-related behavior, as the product of a developmental system that involves numerous factors interacting over time to produce stability or change. In the case of gender, the influences of interest include sex chromosome genes, early testosterone exposure, socialization by external forces (such as parents and broader society), and self-socialization based on cognitive understanding of gender (Hines, 2015). Some current research aims to evaluate how different elements of the system related to sex/gender differences in brain and behavior interact. The studies of parental socialization and self-socialization in girls with CAH are examples of this approach. Individuals with CAH are not numerous, however, and it would be useful to assess individual variability in early testosterone exposure in typically developing individuals. For instance, some studies have related testosterone during the early postnatal surge, sometimes called mini-puberty, to later behavior. Mini-puberty is of particular interest because it provides an opportunity to take repeated measures directly from developing individuals, at a time of rapid brain development, as well as social influence.

One study found that testosterone measured in repeated urine samples obtained approximately monthly across the first 6 months postnatal (during mini-puberty) predicted later sex-typed play (Lamminmaki et al., 2012). Another study found that penile growth during the first 6 months postnatal, which correlates with testosterone during mini-puberty, predicted later sex-typed play in boys, even when prenatal testosterone exposure, as measured by anogenital distance at birth, was controlled (Pasterski et al., 2015). Testosterone during mini-puberty has also been found to relate negatively to sex-typed language development in boys and in girls (Kung et al., 2016), and positively to 3D mental rotation performance in boys (Constantinescu et al., 2018). In contrast, parental attitudes to gender, specifically disapproval of cross-gendered behavior, negatively predicted 3D mental rotation performance in girls (Constantinescu et al., 2018). If independently replicated, these findings could provide a foundation for studies of interactions between hormonal influences and socialization during early development as they operate within the developmental system that shapes human sex/gender development.

Sex/gender and the human brain

Sex differences in behavior necessitate sex differences in the brain. Perhaps the most obvious sex difference in the human brain involves its size. Men's brains are ∼11% larger, on average, than women's brains (Pakkenberg and Voigt, 1964; Luders et al., 2005; Ruigrok et al., 2014). As noted above, men are also taller than women, however, and a larger brain may be needed to run a larger body.

Based on the findings regarding the SDN-POA in rodents, researchers also looked for a similar sex difference in the human brain, using autopsy material. Several independent research groups reported that the third interstitial nucleus of the anterior hypothalamus (INAH-3) was larger in men than in women and appeared to be the human equivalent of the SDN-POA (Allen et al., 1989; LeVay, 1991; Byne et al., 2000, 2001). INAH-3 also is larger in heterosexual men than in men who are not heterosexual, a finding that also has been independently replicated (LeVay, 1991; Byne et al., 2001).

Similarly, the central region of the BST (BSTc) has been reported to show a sex difference and has been related to gender identity. An initial study reported that BSTc was smaller in control women and in a group of 6 trans women than in control men (Zhou et al., 1995). A subsequent study evaluated the BSTc across the lifespan, including during fetal development (Chung et al., 2002). Surprisingly, the sex difference did not appear until after puberty. In contrast, animal models of sex differences, such as the SDN-POA, find that the differences are present from early in life. In addition, most trans people report thinking from childhood that they were assigned to the wrong sex. The findings may suggest, therefore, that the sex- and gender identity-related differences in the human BSTc result from experience, rather than causing variability in gender identity. Alternatively, the sex difference may not have been evident in the brains of children because of small sample sizes.

The emergence and growing access to technologies that allow visualization of the living human brain shifted researchers' focus from subcortical structures, such as INAH-3 and BSTc, to structures that could be visualized using these technologies. These studies are less closely connected to the animal models of testosterone and brain organization than are studies of INAH-3 or the BST. The broad hypothesis that testosterone influences brain structures that differ by sex, however, contributed to a search for reliable sex differences in the living human brain. One difficulty in attempting to identify such sex differences relates to the sex difference in brain size. The studies that identified INAH-3 as the human equivalent of the rodent SDN-POA adjusted INAH-3 volume by calculating its size as a ratio to total brain volume. This ratio procedure mitigated the possibility that INAH-3 was larger in males only because it was scaled to the larger male than female brain. This ratio procedure worked in this case, where a brain region remained larger in males after adjusting for their larger overall brain size. It can be problematic, however, because it can cause brain regions to appear larger in female than male brains, simply because of adjusting for the larger male brain by calculating a ratio. Currently, researchers generally agree that it is appropriate to adjust for the sex difference in brain size, as well as to report unadjusted values. They have, however, found adjustment procedures, such as including total brain volume as a covariate, that do not create spurious differences (Sanchis-Segura et al., 2019).

In addition to the problem of brain size, early research on sex differences in the human brain was hampered by the use of small samples and a tendency to publish significant, but not nonsignificant, results. Recent years have seen larger and more systematic studies, however, involving hundreds or even thousands of participants and examining sex differences across the whole brain, using nondistorting procedures to adjust for brain size. These studies have generally reported that most brain regions are similar in size in males and females, and that, where differences are seen, they tend to be small.

For example, the largest of these studies to date (Ritchie et al., 2018) involved several thousand participants, 45–75 years of age. Like previous studies, it reported a large (d = 1.41) sex difference in total brain volume. After using nondistorting procedures to adjust for the larger male brain, sex differences in gray matter volume, area, and cortical thickness were observed in some of the 68 subregions examined. These differences sometimes favored females and sometimes favored males, and they were generally small, though significant in this large sample. Regarding white matter, the study reported that, for all 22 fiber tracts evaluated, men averaged higher values for fractional anisotropy (average d = 0.19), a measure that is thought to relate to white matter integrity, whereas females averaged higher values for orientation dispersion (average d = 0.30), a measure thought to relate to white matter complexity. The greater values for fractional anisotropy in men were reduced by adjustment for total brain volume, and this adjustment produced a significantly higher fractional anisotropy score for women in one white matter tract. In contrast, the brain volume adjustment produced few changes to the sizes of the orientation dispersion differences. They continued to be of moderate size in the 22 tracts.

The use of larger sample sizes, more complete reporting of results, and nondistorting procedures to adjust for brain size represents progress. Other issues could be addressed, too. One problem that has received relatively little attention is sampling bias, particularly related to the sex difference in volunteering: women are more likely than men to volunteer to participate in research (Rosnow and Rosenthal, 1976; Senn and Desmarais, 2001). Other factors, including education and socioeconomic status, also relate positively to volunteering for research (Rosnow and Rosenthal, 1976). Thus, somewhat different selection biases may lead men versus women to participate. For instance, male volunteers may be more educated or intelligent than female volunteers. Consistent with this possibility, Ritchie et al. (2018) found, to their surprise, that their male participants averaged higher scores than their female participants on a measure of verbal-numerical reasoning. Although this sex difference was small (d = 0.18), it is similar in size to the brain differences observed. It also suggests that the differences observed in this study, and perhaps others that did not measure or match for intelligence, could relate to intelligence differences instead of sex differences. One approach to this problem, particularly when looking at sex differences in the cerebral cortex, might be to match male and female participants for intelligence.

Studies of children are likely to be less affected by the associations between sex and volunteering. Because parents volunteer their children to participate, and because they have male and female children, they are probably equally likely to volunteer boys and girls. Several studies have examined sex differences in the brains of children younger than 2 years. An initial study of 74 infants found that the male brain is already larger than the female brain in the first few weeks postnatal (d = 0.75), a difference similar in size to the sex difference in birth weight (d = 0.71) (Gilmore et al., 2007). Another study of a larger sample of neonates also found a similar sex difference in intracranial volume and body weight, and a positive correlation between intracranial volume and body weight (Knickmeyer et al., 2017). A review of sex differences in the brain across childhood, adolescence, and young adulthood concluded that patterns of sex differences are not static but appear to differ at different ages, often in a nonlinear manner (Kaczkurkin et al., 2018).

Other research suggests that, in regard to brain characteristics that show sex differences, the adult human brain is not uniformly male-typical or female-typical. As suggested by research in rodents, where numerous tissue-specific molecular mechanisms are involved in masculinization of different brain regions, most people have brains that are a mixture of characteristics that are generally more typical of men and characteristics that are generally more typical of women. For example, a study of >1400 adult brains found that, for the brain regions that showed the largest sex differences, most brains did not show a consistent sex-typical pattern in all regions (Joel et al., 2015). Similarly, the same study found that brain connectivity was never consistently male-typical or female-typical in any of the brains. They concluded that brains are not simply male brains or female brains.

In addition to studying brain structure, researchers have used imaging technologies to explore sex differences in brain function. These functional studies add new dimensions that complicate interpretation of results, such as whether participants are resting or completing a task while their brains are imaged. In addition, it has been suggested that functional studies have not used sufficiently similar procedures to allow for replication of specific findings (Carp, 2012). Another challenge relates to interpretation of outcomes. For example, an early study found that male and female brains functioned differently when engaging in a language task (Shaywitz et al., 1995). Male and female performance on the language task was similar, however, suggesting that male and female brains may function differently to produce similar behavior. This finding coincides with a suggestion that sex differences in the brain exist to make male and female behavior more similar, rather than different, given the different hormone milieus of men and women (De Vries and Boyle, 1998). In other words, the brain sex differences may compensate for sex differences caused by gonadal steroids.

Why study sex/gender and the brain

Why do we care about sex differences in the brain? Many researchers in this area want to identify sex differences in the human brain, and their relations to sex differences in behavior, to increase understanding of why men and women differ behaviorally, and why many psychiatric diagnoses show unequal sex ratios. For example, autistic spectrum disorder, attention deficit/hyperactivity disorder, and conduct disorder are more commonly diagnosed in males, whereas eating disorders, major depressive disorder, and generalized anxiety disorder are more commonly diagnosed in females (American Psychiatric Association, 2013). Many neurological disorders (Hanamsagar and Bilbo, 2016; Sohrabji et al., 2016) and chronic pain disorders (Fillingim et al., 2009; Mogil, 2012) are also more common in one sex than the other. Perhaps knowledge of how male and female brains differ could provide insights that would help people with these disorders. Others want to know whether there are brain differences that relate to behavioral sex differences, such as those in 3D mental rotation performance. The ultimate aim is to use understanding of links between sex differences in behavior and sex differences in the brain to facilitate changes in behavior and reduce psychopathology, as well as other types of disorders, related to the brain. In this context, it is important to note that the sex differences that have been reported in the living human brain, with the exception of that in overall brain size, appear to be substantially smaller than the sex differences in some human behaviors, including those in gender identity, sexual orientation, or children's play. In addition, although the subcortical regions INAH-3 and BSTc have been linked to sexual orientation and gender identity, the causes of these links are not known. Also, no other sex differences in brain structure have been linked to human behaviors or psychological characteristics that show reliable and robust sex differences.

Some researchers, and members of society more broadly, may also think that the existence of sex differences in the human brain suggests that sex differences in behavior are inborn, and therefore resistant to change. This perspective reflects a misunderstanding, however. Although we now know that there are sex differences in human brain structure, we do not know what causes them. The existence of sex differences in behavior necessitates sex differences in the brain, and the factors that influence sex differences in behavioral development are likely to also influence sex differences in the brain. So, genes on the sex chromosomes; hormones prenatally, during mini-puberty or at adolescent puberty; socialization by parents, peers, and others; and self-socialization, based on cognitive understanding of gender, are all likely to contribute to sex differences both in behavior and the brain.

Where do we go from here?

Research on sex and the brain has come a long way since 1969. The 1959 report of permanent influences of early androgen exposure on later behavior, as well as reports of dramatic sex differences in the avian and rodent brain, corresponded to a historical time when the social roles of men and women differed dramatically. The narratives of that time often referenced inborn systems and hard wiring of the brain. More recent research has demonstrated greater neural plasticity than was imagined 50 years ago, and interactions between hormones and environmental factors in shaping the human brain and behavior. In addition, developmental scientists have moved on from discussions of nature versus nurture to developmental systems formulations that conceptualize sex/gender development as involving many types of influences interacting over time.

Future research is likely to be aided by improved methods of visualizing the living human brain. Research in nonhuman mammals reported sex differences in cell groups, often of only a few hundred cells, that cannot be visualized in the living brain using current technologies. The small magnitudes of sex differences reported to date in the living human brain may partly reflect an inability to look more closely, particularly at small subcortical cell groups. In addition, thus far, human research has looked largely at relatively gross characteristics, such as regional volumes and fiber tracts. Animal research, however, has found that many more subtle characteristics, such as neurochemical phenotypes, dendritic branching, and synaptic densities, to name just a few, show sex differences (McCarthy et al., 2015). Analyzing these types of characteristics may increase the explanatory power of human research. In addition, researchers developing animal models and those studying humans have not always communicated as well as they might have. It would be interesting to know, for instance, which of the small sex differences that have been seen consistently in the living human brain are also seen consistently in other species, and whether these differences in other species are influenced by early androgen exposure. Similarly, researchers examining early hormone effects on human brain and behavior might benefit from closer attention to the specific hypotheses suggested by the large body of relevant research in other species.

Recent work has benefitted from attention to issues of reliability, sample size, and statistical procedures. Additional attention is needed to sampling biases as well. More information on sex differences over the life span, beginning at birth, would also be useful. Early life is a time when interventions may have maximal impact. Future research might also explore how different types of factors, such as early testosterone exposure and parental socialization, work together in the developmental system that produces sex/gender differences in brain and behavior. Measures of testosterone and parental behaviors during the first few months of infancy (mini-puberty) might be useful in pursuing this goal.