From a biological perspective, the terms ‘male’ and ‘female’ relate to two reproductive types, or sex categories, that are distinguished by the gametes (the cells that fuse together in the process of sexual reproduction) they produce. By definition, females produce the larger gametes (eggs) and males produce the smaller gametes (sperm). In humans, eggs and sperm are produced, respectively, by ovaries and testes. Thus, sex category can be defined by the gonads, and it is only in extremely rare cases (~1:100,000, Blackless et al., 2000; Lee et al., 2016) that a baby is born with gonads that are not clearly either ovaries or testes. Currently, gonads can be removed, but ovaries cannot be changed into testes or testes into ovaries.

The development of a person’s gonads and genital organs occurs in utero and is governed by hormones and genes. Sex-related hormones, the best known of which are testosterone, estrogen and progesterone, are also responsible later in life for the development and functioning of these organs as well as for the development of secondary sex characteristics (e.g., breasts, facial hair), and affect other morphological features (e.g., height, fat and muscle distribution). While often referred to as ‘female’ versus ‘male’ hormones, in fact they do not belong to two distinct categories, as the gonads do. Rather, these hormones are found in all humans, in levels that overlap to a greater or lesser degree, depending on the hormone, age and other factors (reviewed in Hyde et al., 2019; Joel and Yankelevitch-Yahav, 2014). In addition, the levels of these hormones are very dynamic within each individual, changing across the life span (e.g., prenatally, neonatally, childhood, adolescence, menopause, advanced age) as well as in response to internal and external conditions and stimuli (e.g., pregnancy, menstrual cycle, stress, competition, parenting, e.g., Geniole et al., 2017; van Anders et al., 2015). Sex-related genes are not well defined, but many of the genes responsible for the formation of the gonads reside on autosomes, and only a few are found on the X and Y chromosomes (Richardson, 2013). Thus, unlike sex categories, sex as a biological system is a dynamic and often overlapping multi-dimensional variable (Hyde et al., 2019; Joel and Yankelevitch-Yahav, 2014).

Although currently sex-related genes are not medically modified, medically induced alterations of the levels of sex-related hormones are very common, from reproductive-aged females who use birth control pills, to medications for prostate or breast cancer. There are also pharmacological interventions that can promote the typical average hormone levels of reproductive-aged females, or males, including as part of gender-affirming hormonal treatment.

Being a human female is very strongly associated with external sex characteristics in the form of female external genital organs (vulva and clitoris), while being male is tightly associated with external sex characteristics in the form of male external genital organs (scrotum and penis) (Blackless et al., 2000; Lee et al., 2016). For this reason, one can rely on the form of the external genitalia to identify sex category at birth in the vast majority of cases. Nonetheless, both naturally occurring and medically induced variation in primary sex characteristics means that there is diversity in phenotypic sex. For example, approximately 0.02% of human neonates have external sex characteristics that are not clearly female or male (Blackless et al., 2000; Hull, 2003). Currently available surgical and pharmacological interventions can alter or remove many primary sex characteristics (e.g., a womb, a penis), for different purposes, including gender-affirming medical treatment.

If understood to include additional morphological as well as neural characteristics, phenotypic sex becomes even more diverse in its manifestation. As noted above, sex-related hormones are responsible for the development of secondary sex characteristics (e.g., breasts, facial hair), and affect other physical characteristics and capacities, like height, shoulder breadth and upper body strength. In contrast to the genitalia, some of these physical characteristics are affected not only by sex-related hormones but, to a larger or lesser degree, also by gender norms. Also unlike the genitalia, where clear male and female forms exist and intermediate forms are rare, there are varying degrees of overlap between females and males for all other physical measures. Thus, not all males lack breasts and are taller and stronger than all females, and not all females lack facial hair and are shorter and weaker than all males. However, these morphological and physical capacities are strongly intercorrelated (Carothers and Reis, 2013), meaning the presence of one in a form more common in males than in females predicts the presence of other characteristics in a form more common in males than in females, and vice versa. Relatedly, one’s sex category can often be accurately predicted on the basis of these morphological features, even if the genitalia are covered.

The effects of hormonal interventions on physical capacities provide a burgeoning area for research (e.g., Jordan-Young and Karkazis, 2019; Scharff et al., 2019; Wiik et al., 2020). Hormonal and surgical interventions can change some physical characteristics to a level typical of members of the other sex (e.g., facial hair, haemoglobin levels), but, if provided post-puberty, appear to have limited or no effect on others (e.g., height, shoulder breadth, muscle mass and strength, or pelvic width, Harper et al., 2021; Hembree et al., 2017; Hilton and Lundberg, 2021; Spanos et al., 2020).

Diversity and overlap are even more pronounced in the brain. Current findings reveal that sex effects on the brains of both females and males arise out of interactions between genetic, hormonal and environmental factors (for a recent review see, Joel et al., 2020). As a result, there is high variability in the degree of ‘maleness’/’femaleness’ of different features within a single brain (Joel, 2011, 2021; McCarthy and Arnold, 2011). Indeed, studies of the human brain reveal group-level differences in specific brain measures, but brains are often composed of mosaics of both female-typical and male-typical measures (Alon et al., 2020; Joel et al., 2015, 2020). Moreover, unlike other facets of phenotypic sex (that is, genitals and secondary sexual characteristics), an individual’s sex category provides little information regarding their brain structure and function and how these will be similar or different from someone else’s brain (Alon et al., 2020; Eliot et al., 2021; Joel et al., 2018).

Small-scale studies of the brains of trans women and men before any medical intervention reveal a complex picture of group-level similarities and differences between the transgender groups and the corresponding same or other sex-at-birth cis groups (for review see, Nguyen et al., 2019). The same complexity is evident in studies reporting the effects of hormonal treatments on brain structure and function. The average change is typically small and is only sometimes towards the average score in the corresponding cis group, while the variability of the changes between individuals is very high (for review see, Nguyen et al., 2019). Taken together with the observation that the brains of most humans are unique mosaics of female-typical and male-typical measures, it seems very likely that while one’s brain structure can be expected to be different after hormone treatment, it is still likely to continue to take a mosaic form. Indeed, a recent co-analysis of several hypothalamic measures (that show large sex/gender differences) revealed that most of the trans women in the sample possessed a mosaic brain (Joel et al., 2020).

‘Sex’ can refer to sex category (male or female), sex as a system (genes and hormones), or phenotypic sex (in particular primary sex characteristics, but potentially also conceptualized as including secondary sex characteristics, body morphology, and the brain). Phenotypic sex, particularly post-pubescence, often allows an accurate prediction of gonadal sex, thus giving rise to what might be referred to as social sex (i.e., whether one is perceived by others as female, male or not readily classified as either). The latter is also affected by many non-biological characteristics, such as hair style, apparel, vocal modulation, interactional styles, behavioral manners, and name use (e.g., Levitt, 2019; Morgenroth and Ryan, 2021; Tate et al., 2014).

These very different usages of the term ‘sex’ may help to explain a portion of some persistent disagreements in these debates, such as whether or not ‘sex’ is binary, or can be changed. For example, sex categories are binary (only ~1:100,000 are born with gonads that are not clearly either ovaries or testes) and cannot be reversed (i.e., from ovaries to testes or from testes to ovaries), while there is more overlap, dimensionality and mutability among the characteristics of phenotypic sex and sex as a system (see Table 1 for a summary of the different meanings of sex and the corresponding answers to whether ‘sex’ is binary and can be changed).

Although they may vary across time and place, there are group-level behavioral differences between males and females. Some of these are substantial (e.g., sexual interest in females versus males), many others are negligible to moderate (e.g., reading comprehension, extraversion; Hyde, 2005), and some are rare in both sexes while still being more common in one than the other (e.g., sexual violence). However, these characteristics do not create two distinct categories at the individual level. Rather, people possess unique combinations of both feminine (i.e., more common in girls/women than in boys/men) and masculine characteristics (e.g., Joel et al., 2015). In other words, average differences at the population level do not typically allow for accurate predictions regarding the specific preferences, attitudes, and behaviors of an individual on the grounds of their sex category, or for useful generalizations as to ‘what women (or men) are like’.

The inclusion of sex as a protected attribute in discrimination law marks recognition of the fact that one’s registered or social sex category structures one’s social interactions, across all or most social contexts, due to gender norms, stereotypes and social institutions (e.g., Lippert-Rasmussen, 2014; Ridgeway, 2011). Gender norms include laws that restrict property, reproductive or legal rights, or that restrict educational or economic opportunities, on the basis of sex category, but also include informal gender norms that, by leading to social sanctions for those who violate them, serve to facilitate conformity to feminine and masculine codes. Gender stereotypes culturally attribute particular masculine and feminine psychological and behavioral attributes to males and females, respectively, facilitating discriminatory behavior. Gender norms and stereotypes may also shape behavior via internalization. Gendered social institutions, such as the media, the family, workplaces and government help to produce and sustain inequalities between the sexes (such as media that reproduce gender stereotypes, or tax and transfer policies that encourage a gendered division of paid and unpaid labor).

The content and effects of these gender norms, stereotypes and institutions are affected by race, class, age, nationality, religion, ability status and so on (e.g., Crenshaw, 1989; Hall et al., 2019; Maroto et al., 2019), as well as transgender status, which we address specifically in a later section. The overall outcome of the gender system is a consistent pattern of female material, sociocultural and political disadvantage, relative to comparable males, albeit that some issues predominantly affect boys and men (for discussions, see Benatar, 2012; Robeyns, 2003). Even in countries in which the sexes have equal political rights, such as the United States, compared to comparable men, cis women have lower wages and wealth (Chang, 2010; Semega et al., 2020), less leisure time (Charmes, 2022), lower representation in leadership positions (Rhode, 2017), and are at greater risk of sexual and gender-based harassment and violence (Fitzgerald and Cortina, 2018; Tjaden and Thoennes, 1998). For, inter alia, mothers, women of color, working class women, and women with disabilities, these inequalities are often particularly severe or appear in unique forms (e.g., Chang, 2010; Crenshaw, 1989; Hooks, 1984).

Decades of research across multiple disciplines has elucidated and documented the mechanisms by which these inequalities arise. Such research has identified mechanisms that operate at the level of the individual (e.g., effects of gender socialization on self-concept, interests, and skills), interpersonal dynamics (e.g., via descriptive and prescriptive stereotypes, and social networks), organizational/institutional norms and policies (e.g., ‘ideal worker’ norms that marginalize primary caregivers in both pay and career progression, and women’s greater unpaid care labor in the household), and/or mechanisms operating at the societal level (e.g., devaluing or deprioritizing of females within religion or legislation) (e.g., Chang, 2010; Ferrant et al., 2014; Jones, 2019; Pearse and Connell, 2016; Ridgeway, 2011; Rudman and Glick, 2008; Unger, 1979).

The term “gender identity” was coined in the 1960s to refer to a person’s stable and immutable sense of themselves as belonging to a particular sex. The concept was initially discussed in the context of people with atypical sex development or among transgender people (then referred to as transsexuals) expressing a strong desire to change sex (see Meyerowitz, 2002). There has been little agreement about how gender identity (in this sense) develops, with theories throughout the decades including: the action of a sex-linked biological factor in utero, awareness of one’s genital anatomy, or gender-specific socialization (for reviews see Diamond, 2004; Gülgozlgoz et al., 2019; Martin and Ruble, 2004; Person and Ovesey, 1983). The available evidence suggests that no one factor is decisive (e.g., Erickson-Schroth, 2013; Gooren, 2006; Gülgozlgoz et al., 2019; Jordan-Young, 2010; Olson et al., 2015; Voracek et al., 2018).

In recent years, the conception of gender identity has broadened beyond simply the sense of one’s sex as male or female. This is evident in the multiplicity of gender identity labels. For example, James et al. (2016, p. 44) reported that 12% of respondents identified with a gender identity term not among the 25 listed in their U.S. Transgender Survey, and provided “more than 500 unique gender terms with which they identified”. Gender identities have been conceptualized as “self-categorization into a gender group” that may be within traditional categories (male or female), beyond them (e.g., ‘gender blender’) or outside them (e.g., ‘agender’) (Tate et al., 2014, p. 303), and as “constellations of personal qualities (such as behaviors, attitudes, feelings, and interactional styles) associated with physiological sex within a given culture”, such as ‘Southern Belle femininity’ (Levitt, 2019, pp. 276–77). Accordingly, analyses of transgender narratives reveal that individuals choose a gender identity label that authentically reflects their inner sense of self (e.g., Levitt, 2019; Tate et al., 2014), and this label may change as new gender identity labels become culturally available (e.g., non-binary).

Moreover, in transgender people, even identification with being a man or a woman no longer necessarily entails gender dysphoria or a strong desire for social, medical or legal transition. As noted earlier, this has been accompanied by a shift to the conceptually broader term ‘transgender’, to better encompass the identities of those who, unlike transsexual individuals, ‘adopt and embrace fluid, shifting, and ambiguous gender identifications, which seek to combine attributes of masculinity and femininity rather than to “switch” from one gender identity to the other’ (Diamond et al., 2011, p. 633). For example, the 2015 U.S. Transgender Survey (with nearly 28,000 respondents) found that 21% of respondents were living as neither a man nor a woman, and 15% lived sometimes as one gender, and other times as another (James et al., 2016). Similarly, while the dominant narrative of transsexual individuals was, historically, one of ‘crossing over’, more recent transgender narratives also describe oscillations between genders, deliberately ambiguous gender presentation, or an explicitly political project designed to help dismantle gender dichotomies and hierarchies (Ekins and King, 1999; Whittle, 2006).

With this more recent understanding of gender identity, as Morgenroth and Ryan (2021, p. 1118) note, the binary of man versus woman ‘falls short of describing the gender-identity experiences of many individuals’, including those who do not identify as transgender. In line with this, gender identity experiences that transcend the binary have also been reported in children and adults who do not self-label as transgender, with individuals reporting feeling sometimes as both genders or as neither (Jacobson and Joel, 2019; Joel et al., 2013; Martin et al., 2017). For example, a survey that included nearly 5,000 such adults, both female and male, found that many responded in a non-binary fashion to questions about always feeling like a woman and never as a man (or vice versa), feeling as both genders, in-between genders, and neither gender (Jacobson and Joel, 2019). This is consistent with the aforementioned findings that individuals generally possess combinations of masculine and feminine attributes. In contrast, body dysphoria (dislike of one’s body because of its female or male form and the desire to have the body of the other sex) was relatively distinctive of (although not exclusive to, or universal in) participants identifying as transgender and, to a lesser degree, those identifying as gender diverse.

As Stock (2021) has noted, the concept of ‘gender identity’ therefore refers to a multiplicity of phenomena with extremely broad, and often divergent, subjectivities (see also Whittle, 2006). For example: occasional cross-dressing versus medical transition; fluid versus permanent; dysphoric versus political; binary versus non-binary.

The first set of policies seeks to promote girls’ and women’s privacy, safety and other dimensions of wellbeing. One subset of such policies pertains to public (or state) intimate spaces such as toilets, changing-rooms, refuges and shelters, dormitories or prisons by provisioning them with their own facilities, from which men (and boys over a certain age) are excluded. With respect to privacy, this recognizes that in certain contexts (such as communal accommodation or a person employed as a changing room attendant), the need for privacy justifies single-sex provision (e.g., UK’s Equality Act, 2010). Here, it is the physical variables of phenotypic sex or, more specifically, social sex, that are most relevant to these policies. However, it is worth noting that there is variation across individuals and populations in how important this kind of privacy is, and in which contexts. For example, mixed-sex nudity (e.g., in saunas) is commonplace in some countries.

With respect to safety, this exclusion is based on the fact that males comprise the vast majority of the perpetrators of sexual objectification, sexual harassment, and sexual violence. This aspect of the policy is therefore based on the expression of certain masculine behaviors and, to some degree, a male advantage in physical strength that is part of phenotypic sex. Like other exclusions based on statistical group differences (e.g., minimum age for driving, or mandatory retirement), it is a form of imperfect line-drawing that is both over- and under-inclusive: it excludes all males (not just the minority who pose a genuine threat to females); and it does not exclude dangerous females. Members of excluded groups may regard such policies as denigrating (e.g., implying that they are not competent to drive or work, or are potential assailants), but these expressive costs are considered justified, as weighed against the practical benefits and, in these instances, enhanced safety for females.

A ‘softer’ subset of these policies is the institution of formal groups and associations (e.g., a menopause support group). One underlying reason for the provision of women-only groups/associations of this kind is to create a particular kind of emotionally safe or supportive environment based on sex- and/or gender-based shared experiences. A related, further subset of policies is women-only services, such as a national screening program for cervical cancer, assistance with breastfeeding for new mothers, or a women-only group counselling service for victims of sexual assault. The underlying reason for the provision of women-only services is that there is little or no demand among men for the service, or that those in need of the service will be better served by gender-specialized deliveries. (In the latter case, a women-only group may be seen as equivalent to a specialist mental health group for individuals of a specific age range, such as young adults. Even though age is a protected attribute, it is usually considered justified to discriminate on the basis of age for this purpose.) For some women, for groups or services involving discussion of particularly private, intimate or distressing information (e.g., rape recovery), part of creating such an environment may be the absence of individuals at risk of expressing certain unwanted masculine behaviors, and/or whose social sex is male.

A second set of policies takes the form of affirmative action (e.g., gender quotas on boards; women-only prizes, fellowships, shortlists or competitions; targeted recruitment, workshops, training or networking events) that seeks to reduce (and eventually eliminate) the material, political and/or sociocultural disadvantage experienced by girls and women as a group. These policies seek to create substantive equality between women and men, based on argumentation that this is not achieved by equal treatment (formal equality) where there is a background context of unequal conditions.

Although only benefiting some members of the disadvantaged group, affirmative action measures seek to break the nexus between group membership and disadvantage at the system level (Khaitan, 2015). Forms of affirmative action that temporarily reserve an opportunity or resource for women are controversial. Their legitimacy rests on the assumption that the measure will have positive effects for the group more broadly, and that the beneficiaries will help to change institutions structurally, socially or culturally in ways that will help undo lingering effects of male dominance (see Fine, 2021). The proposed mechanisms by which this is hoped to come about may be quite complex and context-specific (e.g., as a role model to younger women, representing women’s interests and concerns based on group identification and/or shared sex- or gender-based experiences, social network effects, creating a ‘critical mass’ of women to change group dynamics and discussion, challenging norms about the ideal incumbent of the position, and so on). It follows that different aspects of sex and gender may be relevant for different affirmative action policies. For example, social sex may be the most relevant aspect for policies with an equality promoting purpose achieved mainly by providing a role model for other women or challenging stereotypes that males are more suited for particular roles. In contrast, repeated experiences of having social interactions structured by one’s (social) sex may be most relevant for policies that seek to promote equality primarily by challenging androcentric policies, practices, products or services (Fine et al., 2020).

A specific form of affirmative action is sex-segregated competitive sporting activities. These provide females with access to the benefits of meaningful competition and the possibility of competitive success, given that most popular sports require physical skills and attributes (a facet of phenotypic sex) for which post-pubescent males, on average—and particularly at the highest levels—are greatly advantaged (Coleman, 2017). As for other kinds of affirmative action, while elite competitions ultimately benefit only a relatively small number of individual female athletes, they bring system-level benefits by challenging stereotypes about females’ sporting abilities, providing role models, and raising the status of women through the visibility of sporting celebrities.

A third set of policies takes the form of transparency measures and reporting requirements (e.g., female representation on publicly listed boards of companies, gender pay gap data, crime statistics). Some of these measures can be seen as a form of affirmative action (Khaitan, 2015), as they ideally contribute to systemic change by motivating action (e.g., by ‘naming and shaming’). Such measures provide data that can document group-based disadvantage, monitor change over time, and assist in the planning and delivery of services (Sullivan, 2020). The purpose of such data is thus, in part, to track the continuing effects of gender as a system on female versus male outcomes.

As this overview indicates, the underlying purpose(s) of the many policies currently being contested do not simply relate to one attribute, such as sex category, phenotypic sex, or gender identity. Nor does identification of the relevant attribute provide ‘the answer’ to a policy’s inclusion criteria, because there are other considerations to take into account. Alongside practical issues of feasibility, there is also the important matter of how a change in policy will affect all stakeholders.