Testosterone mediates life-history trade-offs in female mammals.

The development of physiological and morphological sex differences in humans begins early in the first trimester, when expression of the SRY gene from the Y chromosome in males initiates a cascade of events leading to higher prenatal and postnatal testosterone levels in this sex (Rey, Josso & Racine,  2020 ). This process involves expression of genes with either ‘pro‐female’ and ‘anti‐male’ effects in both sexes (e.g. FOXL2 , WNT4 , RSPO1 , FST and CTNNB1 ) or ‘pro‐male’ and ‘anti‐female’ effects (e.g. SOX9 , and AMH ) (reviewed in Crespi & Dinsdale,  2023 ). The outcome of these effects is divergence in the developmental trajectories of females and males, mediated by organizational effects of testosterone on growth and differentiation of androgen‐sensitive tissues. In addition to the divergence between males and females in sexually dimorphic, testosterone‐mediated traits, beginning early in utero , both females and males exhibit considerable, naturally occurring within‐sex variation in levels of prenatal and postnatal testosterone, which is due to heritable genetic as well as environmental factors (Harris, Vernon & Boomsma,  1998 ; Coviello et al .,  2011 ; Barrett et al .,  2019 ; Ruth et al .,  2020 ). With regard to prenatal effects, this variation is demonstrated by quantitative effects of testosterone levels on developmental elongation of the perineum (the region between the genitalia and the anus). Length of this region is quantified as anogenital distance (AGD), from the anus to either the posterior fourchette (called AGD‐AF) or clitoris (called AGD‐AC) (in women), or to the base of the scrotum (in males). In humans, AGD is about twofold longer in adult males than in adult females, and among females, it varies from about 20 to 60 mm (for AGD‐AF), in association with a broad range of factors that have been linked, experimentally and observationally, to androgen levels and activity (Dean & Sharpe,  2013 ; Thankamony et al .,  2016 ; Schwartz et al .,  2019 ). AGD has been well validated as a strong correlate of early prenatal androgen exposure and its effects, as evidenced by studies of diverse mammal species especially including humans and rodents (Dean & Sharpe,  2013 ; Swan & Kristensen, 2018 ). In support of this assertion, among human females, shorter AGD has been associated, for example, with: ( i ) lower serum testosterone levels in adulthood (Mira‐Escolano et al .,  2014a ; Wu et al .,  2017 ); ( ii ) lower numbers of ovarian follicles (Mendiola et al .,  2012 ); ( iii ) fewer menstrual‐cycle irregularities in one's mother (Mira‐Escolano et al .,  2014a , b ); and ( iv ) lower body mass index in reproductive‐aged individuals (Mendiola et al .,  2016 ; Fabregues et al .,  2018 ; Crestani et al .,  2020 ; Peters et al .,  2020 ; Liu et al .,  2023 ). Shorter AGDs have also been linked with higher risk for premature ovarian insufficiency (Dural et al .,  2021 ; Tan et al .,  2023 ), apparently as a maladaptive extreme of high, early loss of oocytes and ovarian follicles (Crespi & Dinsdale,  2023 ). The second to fourth digit length ratio (2D:4D digit ratio) has also been used as an indicator of prenatal testosterone levels in many studies, but this biomarker is notably less well validated experimentally than AGD (Richards, Browne & Constantinescu,  2021 ), and measures of AGD and digit ratio show variable, mixed evidence of correlations with one another, generally in the expected direction, in the only study to date with such data (Barrett, Parlett & Swan,  2015 ). Interpretation of digit ratio data can be alleviated, in part, by replication and by evidence from other sources of data that show convergent effects, though such data are seldom available. In women, androgens are produced predominantly from the ovaries and the adrenal glands, and converted among their various forms in diverse tissues, where they exert their effects via activation of the androgen receptor (Walters, Allen & Handelsman,  2008 ; Walters,  2015 ) (Fig.  1 ). In the ovaries, folliculogenesis, the process of oocyte maturation, progresses under the influence of testosterone released from theca cells, which exerts stimulating effects in early folliculogenesis, and inhibitory effects in its later stages (Astapova, Minor & Hammes,  2019 ). Relatively low levels of ovarian testosterone during folliculogenesis tend to cause apoptosis of developing oocytes (Ono et al .,  2014 ), but especially high levels of testosterone result in arrest of follicular development at the small antral stage and anovulation or oligo‐ovulation, as exemplified by polycystic ovary syndrome (PCOS) (where such small, arrested follicles are mistakenly referred to as ‘cysts’) (Franks & Hardy,  2018 ). These effects of testosterone during the follicular stage of the menstrual cycle, coupled with variation among women in menstrual cycle length and regularity, which are also mediated by testosterone (van Anders & Watson,  2006 ; Wei et al .,  2010 ), exert strong impacts on the timing and numbers of oocytes produced that are available for fertilization and implantation. As such, they regulate the nature, timing and intensity of investment in reproduction. Organizational and activational effects of testosterone among female mammals in development, survival, and reproduction. Androgens also play key roles in normal and abnormal functioning of the endometrium, the tissue that subserves implantation, placentation, and fetal development (Simitsidellis, Saunders & Gibson,  2018 ). These roles include: ( i ) conversion of androstenedione to testosterone during the early secretory phase, a process that facilitates decidualization (Das et al .,  2009 ; Gibson, Simitsidellis & Saunders,  2016 ); ( ii ) local regulation of endometrial inflammation by androgens, which is important because embryo implantation is an inflammatory process whereby levels that are too high or too low inhibit the process (Dekel et al .,  2010 ; Huang et al .,  2021b ; Velez, Seldin & Motta,  2021 ); and ( iii ) direct effects on endometrial breakdown and on the repair of endometrial tissue after menstruation (Cousins et al ., 2016 ). These findings indicate that, as for the ovaries, variation in levels of testosterone in the endometrium plays key roles in successful reproduction, notably through effects on implantation and early embryo loss, which have important impacts on human fertility (e.g. Jarvis,  2016 ). Testosterone and other androgens have been demonstrated, across hundreds of studies, to exert diverse effects on aspects of behaviour in women (e.g. Heany et al .,  2016 ; Hampson,  2018 ). This work shows, most broadly, that testosterone is important in the regulation of adaptive fitness‐related behaviour in women, via mechanisms that involve short‐ and long‐term changes in serum testosterone levels, coupled with effects from variation in the densities and distributions of the androgen receptors that control the behavioural effects of testosterone. As regards the testosterone trade‐off hypothesis, the primary prediction regarding behaviour is that elevated testosterone in women should be associated with higher levels of competition, dominance, and aggression, expressed in specific contexts and individuals where such behaviours are expected to be adaptive; higher testosterone should also involve increased muscle mass and strength, and associated physical abilities. In any given situation, the benefits of testosterone‐associated phenotypes should link with fitness‐related gains, such as increased food, a better mate, higher social status, or other benefits. Moreover, the benefits of these traits should be associated with costs in terms of ability to produce offspring. In contrast to these effects, lower testosterone in women should be associated with reduced success in arenas of direct and indirect competition among women, as well as with benefits to components of reproduction.

As applied to humans, the testosterone trade‐off hypothesis predicts that relatively low (but not pathologically low) prenatal and postnatal testosterone levels should be associated with higher reproduction, and with a consequent higher degree of mate preference by males, but also with reduced dominance, strength, aggression, and survival; for higher testosterone levels the opposite patterns are predicted. These predictions apply most directly to natural‐fertility human populations, especially hunter–gatherers, horticulturalists, and traditional farmers, but the physiological mechanisms involved, which relate testosterone levels to other phenotypes, should be largely conserved across populations with diverse ecologies. AGD has not been studied in relation to offspring production in healthy women, but longer (more ‘female‐biased’) 2D:4D digit ratios, which have been considered to indicate lower prenatal testosterone levels, have been associated with higher fecundity (numbers of children), and with being married versus unmarried, in several European populations (Manning et al .,  2000 ; Manning & Fink,  2008 ; Klimek et al .,  2016 ). Evidence salient to the effects of prenatal testosterone on female reproduction also comes from studies that compared the fitness effects of developing in utero with a female co‐twin versus a male co‐twin. Lummaa, Pettay & Russell ( 2007 ) and Butikofer et al . ( 2019 ) both reported that females with a male co‐twin, who are expected to be subject to testosterone transfer from them, showed significantly lower fecundity, measured as numbers of children, than did females with a female co‐twin; by contrast, Medland et al . ( 2008 ) did not find a difference between females in these two groups. Adult serum testosterone level has been associated with many correlates of reproduction in healthy women. Thus, lower testosterone level has been linked with: ( i ) an earlier age of menarche (Apter & Vihko,  1983 ), which predicts earlier first reproduction, more‐regular cycles, and higher fecundability (conceptions per cycle), more‐successful parturition, and higher numbers of offspring produced, across many studies (Udry & Cliquet,  1982 ; Sandler, Wilcox & Horney,  1984 ; Mulder,  1989 ; Apter & Vihko,  1990 ; Komura et al .,  1992 ; Anai et al .,  2001 ; Helle,  2008 ; Hochberg et al .,  2011 ; Guldbrandsen et al .,  2014 ; Zhang et al .,  2017 ; Lassek & Gaulin,  2021 ); ( ii ) shorter, more‐regular menstrual cycles, which are associated with a reduced incidence of anovulation or oligoovulation (Anai et al .,  2001 ; Sjaarda et al .,  2018 Bloom et al .,  2021 ) and higher offspring production (Wesselink et al .,  2017 ; Zhang et al .,  2017 ); ( iii ) reduced rates of nulliparity (Nagata et al .,  2011 ); and ( iv ) lower risks of recurrent miscarriage (Okon et al .,  1998 ; Cocksedge et al .,  2008 ). Lower prenatal testosterone level, as inferred from having a female compared to a male co‐twin, has also been linked with earlier menarche by Kaprio et al . ( 1995 ) and Jahanfar & Walters ( 2019 ). Relatively low maternal testosterone levels, or low testosterone relative to oestradiol, during gestation have also been associated with higher offspring birth mass, either across both sexes (Carlsen, Jacobsen & Romundstad,  2006 ; Svensson et al .,  2019 ; Huang et al .,  2021a ), or only in males (Voegtline et al .,  2013 ; Palm et al .,  2024 ). A genetic basis to this effect is indicated by an association of high genetically predicted testosterone levels with lower offspring birth mass (Zheng et al .,  2022 ). These findings are important because lower birth mass is associated with higher infant mortality in a more or less linear fashion, for birth masses below about 3.5 kilograms (Ma & Finch,  2010 ). Low serum or salivary testosterone level has also been reported among women in life‐history situations that involve higher levels of non‐competitive reproductive effort (compared to mate‐acquisition effort, or investment in maintenance and survival) (van Anders & Watson,  2007 ; Roney & Gettler,  2015 ). These situations include: ( i ) recent and current romantic love (van Anders & Goldey,  2010 ; Sorokowski et al .,  2019 ); ( ii ) marriage or stable cohabitation with a male versus being single (Edelstein, Chopik & Kean  2011 ; Barrett et al .,  2013 ; Edelstein et al .,  2014 ); and ( iii ) being engaged in rearing young children (Kuzawa et al .,  2010 ; Barrett et al .,  2013 ). Lower testosterone levels in women thus appear to facilitate and reflect increased investment in monogamous relationships and in child‐rearing (see also Deady et al .,  2006 ), which reinforces its associations with higher reproduction. Two key correlates of testosterone levels in women, with regard to aspects of reproduction, are waist‐to‐hip ratio (WHR) and body mass index (BMI). WHR and BMI are thus associated with female fitness through some combination of direct, naturally selected effects on reproduction, and effects via sexual selection by male mate choice and male contributions to female lifetime reproduction. Lower WHR, corresponding to a narrow waist and wide hips, is generally indicative of high reproductive value (in terms of future reproduction) and high levels of reproductive (gluteofemoral) fat deposition, which serves as a store of energy to support the high costs of gestation and lactation (Wells,  2007 ; Lassek & Gaulin,  2008 ; Wells, Griffin & Treleaven,  2010 ; Chiappa & Singh,  2017 ). Reproductive‐aged women with lower WHRs exhibit lower levels of serum testosterone (Evans et al .,  1983 ; de Ridder et al .,  1990 ; Sowers et al .,  2001 ; van Anders & Hampson,  2005 ), high serum testosterone combined with low oestradiol has also been linked with high WHR (Mondragon‐Ceballos et al .,  2015 ), and metrics of wider hips and narrower waist predict earlier age of menarche (Lassek & Gaulin,  2007 ). By contrast, WHR is not clearly associated with 2D:4D digit ratio, with two studies reporting evidence of associations of lower WHR with higher digit ratios (i.e. lower prenatal testosterone level) (Manning et al .,  2000 ; Zurawiecka et al .,  2019 ), but two studies showing a lack of association between these traits (Fink, Neave & Manning,  2003 ; Swami et al .,  2019 ). Relatively low WHR confers benefits to females in the contexts of higher fecundability, increased offspring birth mass, and higher levels of serum DHA (docosahexaenoic) fatty acid that are crucial for lactation and offspring early brain development (e.g. Zaadstra et al .,  1993 ; Jasieńska et al .,  2004 ; Weeden & Sabini,  2005 ; Cashdan,  2008 ; Lassek & Gaulin,  2008 ; Butovskaya et al .,  2017 ). In a life‐history context, low WHRs may also signal high expected total future reproduction, rather than current high rate of offspring production per se (Lassek & Gaulin,  2019 ). In contrast to WHR, BMI represents a measure of body mass in relation to height, such that it serves primarily as a measure of obesity versus thinness overall (Müller et al ., 2016 ). Among women, BMI is positively correlated with WHR, reflecting the effects of variation in visceral obesity (and waist circumference) on both measures. In contrast to gluteofemoral fat, which represents ‘reproductive fat’, visceral fat can be considered as ‘survival fat’ (Wells,  2023 ), given that it functions for both energy storage (as a buffer against starvation) and as a source of immune system components such as cytokines (used in fighting infection). These functions are important because, across evolutionary timescales, malnutrition and infectious disease represent two of the most important selective agents affecting human survival (Kuzawa,  2010 ; Karlsson, Kwiatkowski & Sabeti,  2014 ). BMI is strongly positively associated with serum testosterone levels in women (Taponen et al .,  2003 ; Stanikova et al .,  2019 ). Physiologically, higher visceral fat, insulin resistance, and increased LH production appear to mediate the strong links of BMI with serum testosterone levels, although the roles of specific metabolic and reproductive factors in this process remain unclear (Valderhaug et al .,  2015 ). Women with higher BMI also show longer AGDs (indicating higher prenatal testosterone levels) (Mendiola et al .,  2016 ; Fabregues et al .,  2018 ; Crestani et al .,  2020 ; Peters et al .,  2020 ; Liu et al .,  2023 ), which implicates in utero programming effects; however, BMI does not show associations with digit ratios (Fink, Manning & Neave,  2006 ; Swami et al .,  2019 ). Fecundability generally decreases with increasing BMI, from a maximum at around 20–21, and with an especially notable decline at BMIs above about 25 (Jensen et al .,  1999 ; Hassan & Killick,  2004 ; Gesink Law, Maclehose & Longnecker, 2007 ; Yilmaz et al .,  2009 ; Wise et al .,  2010 ; McKinnon et al .,  2016 ). Findings that lower BMI is associated with lower testosterone levels, and with enhanced reproductive ability, are subject to the caveat that very low BMI values (under about 19) are associated with reduced reproduction due to insufficient adipose tissue to support gestation and lactation, which typically leads to anovulation (e.g. Rich‐Edwards et al .,  2002 ). Similarly, very low levels of testosterone can be deleterious to fertility, given that they are associated with premature ovarian failure (Gleicher et al .,  2013 ; Soman et al .,  2019 ). In this context, it is also noteworthy that especially short AGDs, and thus lower prenatal testosterone levels, have also been associated with premature ovarian failure (Dural et al .,  2021 ), endometriosis (Crespi,  2024 ), and reduced levels of AMH (Fabregues et al .,  2018 ), which are indicative of low ovarian reserve (Fabregues et al .,  2018 ; Sánchez‐Ferrer et al .,  2019 ). Among non‐human mammals including primates, higher testosterone level in females has been associated with higher dominance and social rank (Stockley & Bro‐Jørgensen, 2011 ; French et al .,  2013 ), which, in turn, contributes to higher survival rates, longer reproductive periods and higher offspring survivorship (Altmann & Alberts,  2003 ; Nelson et al .,  2010 ; Campos et al .,  2020 ). The testosterone trade‐off hypothesis predicts that in humans, women with higher testosterone levels will exhibit higher survival rates (and higher survival of their offspring), in association with enhanced ability to acquire resources due to elevated dominance, social status, and physical abilities; they should also tend to show reduced and delayed reproduction. In adult, reproductive‐aged women, levels of testosterone have shown evidence of positive association with diverse measures of competitiveness (Hahn et al .,  2016 ), dominance (Dabbs & Hargrove  1997 ; Grant & France, 2001 ;Vermeersch et al .,  2010 ; Cobey et al .,  2015 ) and aggression [reviewed in Denson et al . ( 2018 ) and Bushell & Crespi ( 2024 )]. Higher adult testosterone level has also been linked with muscle mass and athletic performance (for many sports) in healthy women (reviews in Handelsman, Hirschberg & Bermon,  2018 ; Hirschberg,  2020 ), and PCOS, which centrally involves high testosterone levels and reduced ovulation rates, has been reported at high rates among female athletes (Dadgostar et al .,  2009 ; Hagmar et al .,  2009 ; Coste et al ., 2011 ). Reduced ovulation rates have also been described among female athletes with high levels of testosterone but without diagnoses of PCOS (Rickenlund et al .,  2003 , 2004 ; Awdishu et al .,  2009 ). A recent review (Taylor et al .,  2023 ) reported an overall lack of association of total testosterone level with muscle mass or strength (although free testosterone level was positively associated with lean muscle mass); however, most of the studies included in the review involved post‐menopausal women. Higher levels of athleticism, and higher handgrip strength, have also been reported consistently among women with lower 2D:4D digit ratios, which are considered to be indicative of higher prenatal testosterone levels (see Eklund et al .,  2020 ; Pasanen et al .,  2022 ; Bushell & Crespi,  2024 ). Given archaeological and anthropological evidence of high strength and muscularity among women in small‐scale societies (e.g. Eshed et al .,  2004 ; Macintosh, Pinhasi & Stock,  2017 ; Miller et al .,  2018 ; reviewed in Bushell & Crespi,  2024 ), these findings suggest that higher testosterone levels can provide benefits to women as regards physical traits that contribute to survival. The specific contexts of such benefits remain unclear, but may include enhanced food acquisition through hunting, physically demanding gathering, and other forms of activity that involve athleticism, endurance, and strength (Macintosh et al .,  2017 ; Miller et al .,  2018 ; Anderson et al .,  2023 ). Robust tests of these ideas require measurements of testosterone and its correlates, in relation to dominance, social status, muscularity, and physical performance among women in small‐scale societies. Variation in lifetime fitness among women is also expected to be related to survival ability during periods of food scarcity, and ability to resist death from infectious diseases. These components of fitness have not been evaluated with regard to variation in testosterone levels among women. However, the high positive correlations of testosterone levels with BMI and levels of visceral fat in women, across the full range of BMIs (Janssen et al .,  2010 ; Tin Tin, Reeves & Key,  2020 ), combined with evidence for links of visceral fat with enhanced survival during food restriction (Kuzawa,  2010 ), suggest that higher testosterone levels may enhance survival in this context. These expectations contrast with those for males, for whom testosterone, as the main hormone mediating investment in reproduction, demonstrates an inverse relationship with both BMI and immune function (Muehlenbein et al .,  2023 ). The testosterone trade‐off hypothesis predicts that, given associations of lower testosterone with higher female reproduction, males should prefer to mate and pair with females who show traits that are indicative of relatively low testosterone levels. As such, higher testosterone level in females would engender benefits from higher dominance, strength, endurance and social status, but also costs with regard to reduced benefits from preference by, and investment from, males. The hypothesis that males prefer to mate with lower‐testosterone females has been supported by data from a suite of non‐human animals (Fig.  2 ) and by data from humans. The most direct evidence relating human female testosterone levels to attractiveness to males comes from studies by Probst, Bobst & Lobmaier ( 2016 ) and Żelaźniewicz et al . ( 2021 ), both of which showed associations of higher female facial attractiveness with lower serum testosterone levels. Female faces that are morphologically ‘more female’, as defined morphometrically (e.g. with smaller jaws and elevated eyebrows), have also been shown to be preferred by males, in both westernized and small‐scale populations (Lee et al .,  2014 ; Marcinkowska et al .,  2014 ; Scott et al .,  2014 ; Kleisner et al .,  2017 ; Kočnar, Saribay & Kleisner,  2019 ) (reviews in Kościński,  2007 ; Little, Jones & DeBruine, 2011 ). Females with higher facial attractiveness to males have also been shown to exhibit higher fecundity (measured as number of children) across two studies (Jokela,  2009 ; Pflüger et al .,  2012 ), although not in a third (Pawłowski et al .,  2008 ). Human female bodily attractiveness has also been associated, across many studies, with relatively low WHR and BMI (Furnham, Petrides & Constantinides,  2005 ; Singh et al .,  2010 ; Singh & Singh,  2011 ; Bremser & Gallup,  2012 ; Cloud & Perilloux,  2014 ; Grillot et al .,  2014 ; Wang et al .,  2015 ; Andrews et al .,  2017 ; Del Zotto & Pegna,  2017 ; Lassek & Gaulin,  2019 ), which also represent indicators of high reproductive value (Wang et al .,  2015 ; Lassek & Gaulin,  2018 ). Both WHR and BMI are positively correlated with serum testosterone levels in women, as described above. Whether or not WHR and BMI per se are the traits perceived as more or less attractive remains unknown; for example, Rilling et al . ( 2009 ) showed that low abdominal depth and waist circumference were stronger predictors of attractiveness than WHR and BMI, which suggests that levels of visceral fat were the trait subject to sexual selection in this context. Male preferences for relatively high‐pitched voices in females have been demonstrated across a large set of studies (Collins & Missing,  2003 ; Feinberg et al .,  2008 ; Borkowska & Pawłowski,  2011 ; Valentova et al .,  2019 ; Barkat‐Defradas, Raymond & Suire,  2021 ; Suire, Raymond & Barkat‐Defradas,  2021 ). In turn, higher pitch has been linked with lower testosterone levels (Abitbol, Abitbol & Abitbol,  1999 ; Hannoun et al .,  2011 ; Hamdan et al .,  2018 ), and treatment of women with endometriosis with the synthetic androgen danazol causes deepening of pitch, in up to about 10% of cases (Pattie et al .,  1998 ). Atkinson et al . ( 2012 ) also demonstrated that females with higher‐pitched voices had more children, in a population of indigenous women in Namibia. Finally, if overall ‘attractiveness’ of females to males is associated with low testosterone levels, then female attractiveness‐related phenotypes should be positively associated with one another. This pattern is demonstrated by links of facial with vocal attractiveness (Collins & Missing,  2003 ; Wheatley et al .,  2014 ), lower WHR with higher vocal attractiveness (Hughes, Harrison & Gallup Jr,  2009 ), lower BMI with higher facial attractiveness [Hu et al . ( 2019 ) at the genetic‐correlation level], and WHR and BMI with facial shape (Pisanski et al .,  2016 ; Mayer et al .,  2017 ).

The testosterone trade‐off hypothesis, as originally proposed by Packer et al . ( 1995 ) and elaborated here, predicts that levels of this hormone mediate trade‐offs between aspects of reproduction and dominance in female mammals. From the available data on non‐human animals as well as humans, its predictions generally appear to be met. Thus, a relatively low testosterone level tends to involve earlier onset of reproduction, higher reproductive rates, and higher attractiveness to males, whereas relatively high testosterone levels involve increased aggression, dominance, strength, and access to fitness‐related resources, which are expected to result in higher survivorship of a female and her offspring. The testosterone trade‐off hypothesis is quite general in its application, but social and ecological context are central to how it applies in any given species. Thus, the strength and nature of this trade‐off is predicated on the degree and forms of female–female competition over resources, which can include ability to secure a mate or a better mate, ability to secure defendable or difficult‐to‐obtain resources for oneself or one's offspring, and ability to survive and reproduce under elevated levels of ecological or social stress, such as when female densities are relatively high. The testosterone trade‐off hypothesis thus applies in situations where females engage in levels of intrasexual competition sufficiently high to select for levels of testosterone that are high enough to interfere with optimal reproductive–physiological functioning. As such, the trade‐off will not manifest among mammalian species where females do not compete with one another or are not selected for elevated testosterone in some other context; moreover, in many species, levels of competition among females are expected to vary substantially across time and space, such that the presence and strength of the trade‐off will fluctuate due to variability in socioecological conditions. Like the classic ‘cost of reproduction’ trade‐offs in life‐history theory (Reznick,  1985 ), the core of the testosterone trade‐off hypothesis centres on compromises between reproduction and survival, where both of these fitness components make substantial contributions to lifetime reproductive success. Such compromises are enacted at the level of individual females, due to variation in their genes and environments, and selection on testosterone‐associated traits takes place at the level of populations. A key aspect of the hypothesis is that it identifies a specific hormonal cause of the trade‐off, level of testosterone, that is easily (though seldom) measured and rarely tested for linkage to fitness components among females, and that has well‐recognized functions during prenatal deveopment, in ovaries, endometrium, adipose, muscle and neural tissues (Fig.  1 ). Trade‐offs mediated directly by hormones differ somewhat from those centred on energy and time, where the fitness‐related currencies are straightforward and interchangeable across the life‐history traits subject to compromises. Indeed, hormonal trade‐offs are based on a lack of extensive adult tissue‐specific plasticity and a lack of functional independence of the physiological systems that could optimize the effects of testosterone across female tissues on short‐term timescales. The primary factors that mitigate against such potential high levels of plasticity are twofold (see Hau & Wingfield,  2011 ). First, the HPO axis is subject to developmental programming by levels of prenatal testosterone, as indicated, for example, by associations of AGD with adult testosterone levels (Mira‐Escolano et al .,  2014a ; Wu et al .,  2017 ), BMI (Mendiola et al .,  2016 ; Crestani et al .,  2020 ; Peters et al .,  2020 ; Liu et al .,  2023 ), AMH (Fabregues et al .,  2018 ), premature ovarian insufficiency (Dural et al .,  2021 ; Tan et al .,  2023 ), dysmenorrhea (Hewitt,  2020 ; Saputra et al .,  2023 ), and risks of endometriosis and PCOS (Crespi,  2024 ). To the extent that the functional capacities of the HPO axis are permanently programmed by organizational effects of prenatal testosterone, and represent reaction norms constrained by their effects, they cannot be dynamically optimized across time and tissues to meet short‐term exigencies. Second, the genetic basis of HPO axis regulation shows evidence of extensive pleiotropy: for example, a specific haplotype of the key HPO axis gene FSH‐B has been associated with higher FSH (follicle‐stimulating hormone), lower LH, and lower testosterone levels, earlier menarche, shorter menstrual cycles, earlier age of first birth, higher numbers of offspring, higher rate of twinning, earlier menopause, higher risk of endometriosis, and reduced risk of PCOS (Dinsdale, Nepomnaschy & Crespi,  2021 ). The existence of such loci provides evidence that female life histories and reproductive traits are programmed to develop as a suite of coadapted phenotypes that reflect variation in key variables such as GnRH (gonadotropin‐releasing hormone) pulse frequencies and HPO regulatory mechanisms mediated by androgens and oestrogens. More broadly, to the extent that testosterone in females has evolved to coordinate adaptively across ovarian, endometrial, muscle, adipose, and neural phenotypes, it is expected to exhibit relatively low levels of long‐term, trait‐specific adult physiological plasticity. For traits to evolve under the selective pressures that underlie life‐history theory, they must exhibit a genetic basis and be heritable; moreover, trade‐offs should reflect negative genetic correlations (Stearns,  1989 ). Testosterone levels are highly heritable and mediated by many alleles each of small effect in adult reproductive‐aged women (Harris et al ., 1998 ; Coviello et al .,  2011 ; Ruth et al .,  2020 ), as well as being genetically uncorrelated with testosterone levels in males (Ruth et al .,  2020 ; Leinonen et al .,  2023 ), indicating independent genetic control in the two sexes. There is thus considerable natural variation in testosterone levels among adult women that is associated with diverse factors including genetic and environmental variation, as well as considerable variation within women over time, associated with causes including age and menstrual cycle stage. Free testosterone level shows a significant negative genetic correlation with number of live births in women, using data from FinnGen (Leinonen et al .,  2023 ), but further analyses are needed for robust tests of the genetic basis of testosterone‐related trade‐offs. Such tests are complicated by the observation that prenatal testosterone levels, which program the reproductive system as indicated by links of AGD with adult traits (including adult serum testosterone level) in women, are determined in part by testosterone‐related phenotypes of the mother, including her age, parity, BMI, and levels of exposure to stress (Barrett & Swan, 2015 ; Barrett et al .,  2019 ). As a result of such maternal influences, testosterone‐linked phenotypes in women, including for example their risk of PCOS, can exhibit transgenerational, hormonally mediated effects that are more or less independent of allelic variation (Risal et al .,  2023 ). The causes of variation in testosterone levels in adult women are thus complex, involving both heritable and diverse environmental effects, and remain incompletely understood. Studies testing for testosterone‐mediated trade‐offs should take account not just of adult testosterone and its activational effects, but also measure AGDs as robust indicators of prenatal, organizational testosterone effects, that integrate prenatal hormonal influences from both the mother and her offspring. Studies of heritability for testosterone levels are also needed for females of non‐human mammals, as very few such data are currently available. Studies on red deer ( Cervus elaphus ) (Pavitt et al .,  2014 ), and nutria rodents ( Myocastor coypus ) (Fishman et al .,  2024 ), both showed significant heritabilities. As described here, data salient to the testosterone trade‐off hypothesis comes from two main domains of evidence: non‐human animals and women (Figs  3 and 4 ). These data provide convergent, independent lines of evidence that support most predictions of the hypothesis, even though none of the information was collected to test it explicitly. Thus, the data from animals demonstrate broad generality in the presence of the predicted fitness‐related effects of testosterone across diverse mammals, and the data from healthy women provides extensive evidence of how testosterone levels affect reproductive phenotypes and components of lifetime fitness. Overview of the testosterone trade‐off hypothesis and points salient to its main predictions. Fecundability refers to conceptions per cycle. Evidence showing convergent effects of prenatal and postnatal testosterone among rodents and humans (the taxa with most data available), for a suite of traits relevant to the predictions of the testosterone trade‐off hypothesis. See text for citations. 0M, 1M, and 2M indicate exposure to 0, one or two adjacent male fetuses, respectively, during gestation; T, testosterone. In addition to mediating trade‐offs, testosterone, like other hormones in females, coordinates the co‐expression of different phenotypes across tissues. Thus, according to the hypothesis, conditions of high ecological or social stress in humans should favour women subject to effects of higher prenatal and postnatal testosterone because they generate a set of coadapted phenotypes including delayed or reduced reproduction, higher investment in survival via preferential deposition of visceral fat, and higher dominance that fosters success in among‐female competition (Volk,  2023 ). By contrast, beneficial socioecological conditions and high resource availability should favour earlier and higher fertility and reproduction, more deposition of gluteofemoral fat, and avoidance of the costs of intrasexual competition. In such favourable situations, trade‐offs involving testosterone should indeed be ameliorated or eliminated, such that low levels may be universally beneficial. Presumably, levels of social and ecological stresses have, in humans, varied across ecological and evolutionary timescales in ways that maintain variation in testosterone levels in females, leading to the reaction norms and physiological patterns observed in contemporary human populations. As such, short‐term, contextually adaptive variation in levels of adult testosterone in females, representing organizational effects, is expected (e.g. van Anders, Steiger & Goldey,  2015 ), but its range of influences should be constrained or limited by programming during the prenatal stage. Despite the range of evidence consistent with the predictions of the testosterone trade‐off hypothesis, the results are subject to numerous limitations and caveats. First, much of the animal data comes from experimental, laboratory studies, many of which involve supraphysiological levels of testosterone or procedures that do not reflect key aspects of natural, field conditions, such as food limitation or direct competition between females. Moreover, almost all relevant animal studies involve increased levels of testosterone, such that the key effects of experimentally reduced levels remain largely unknown. More‐robust tests of the hypothesis using animal data will require study of natural or semi‐natural populations (e.g. of rodents), where females are measured for AGD and monitored as regards testosterone levels, competitive behaviour, reproduction, and survival, as well as being experimentally manipulated with regard to testosterone levels. Do dominant, high‐testosterone females in mammalian social groups have higher lifetime fitness, in spite of moderately reduced reproduction? How important is survival, compared to reproduction, as a determinant of lifetime fitness in most mammals? How does the presence and strength of testosterone‐mediated trade‐offs in female mammals vary across and within species? Second, little of the data salient to the testosterone trade‐off hypothesis from healthy, reproductive‐aged women has been collected and analysed in the context of life‐history theory or fitness‐related trade‐offs. Thus, most physiological and medical studies have analysed testosterone levels in relation to specific aspects of infertility or disease in women, and most behavioural‐ecological and psychological studies have sought to connect testosterone levels with correlates of female attractiveness or ability to produce offspring, or with correlates of female strength or dominance. Robust tests of the hypothesis require drawing links between ( i ) specific phenotypes (e.g. AGD, WHR), ( ii ) testosterone levels, ( iii ) components of reproductive fitness (e.g. age‐specific fecundability), and ( iv ) components of survival‐related fitness (e.g. strength, dominance, status, and survivorship of self and offspring), all in the same population. Better data are also needed on survivorship itself (of a mother and her offspring) in relation to testosterone and its correlates, among animals. The hypothesis can also be evaluated further through collection of relevant data on correlates of reproduction and survival in women with endometriosis and PCOS, given that these two disorders appear to involve maladaptive extremes of trade‐off‐related to effects from low versus high prenatal and postnatal testosterone levels (Dinsdale et al .,  2021 ; Dinsdale & Crespi,  2021 ; Crespi & Dinsdale,  2023 ; Crespi,  2024 ). Given the primacy of trade‐offs in the evolution of life histories among humans and other taxa, and the key roles of steroid hormones in regulating them, collection of additional data on testosterone effects in women should be important to progress in evolutionary endocrinology.

(1) Testosterone has diverse functions in female mammals, but they have yet to be comprehensively analysed with regard to life‐history trade‐offs involving reproduction and survival. (2) Lower testosterone level among female mammals is associated, across many species including mice, rats, gerbils, voles, hamsters, rabbits, sheep, cows, monkeys, and humans, with higher reproduction, but decreased aggression and dominance. (3) Males prefer to mate with females subject to lower levels of prenatal and postnatal testosterone, apparently due to the higher rates of offspring production by such females. (4) Overall, the findings support the hypothesis that testosterone in female mammals mediates trade‐offs involving components of reproduction in relation to dominance and survival. Testosterone has diverse functions in female mammals, but they have yet to be comprehensively analysed with regard to life‐history trade‐offs involving reproduction and survival. Lower testosterone level among female mammals is associated, across many species including mice, rats, gerbils, voles, hamsters, rabbits, sheep, cows, monkeys, and humans, with higher reproduction, but decreased aggression and dominance. Males prefer to mate with females subject to lower levels of prenatal and postnatal testosterone, apparently due to the higher rates of offspring production by such females. Overall, the findings support the hypothesis that testosterone in female mammals mediates trade‐offs involving components of reproduction in relation to dominance and survival.

Hormones coordinate physiological systems. As such, they typically exhibit a set of properties that integrates information from the environment with information on internal physiological processes, resulting in dynamic optimization of physiological and behavioural states to maximize condition‐dependent fitness (lifetime fitness, which is a product of survival and reproduction) (McGlothlin & Ketterson, 2008 ). Such integration fundamentally involves trade‐offs, which represent compromises between beneficial but incompatible traits (Sinervo & Svensson,  1998 ; Ellison,  2017 ). Trade‐offs typically involve time and allocation of energy and resources to different bodily functions, especially aspects of growth, maintenance and reproduction. Although hormones are often considered to have specific functions, they may more accurately be construed as proteins that regulate trade‐offs across both short‐term (physiological), and long‐term (developmental) timescales. In these contexts, the magnitude and form of trade‐offs depends upon aspects of the environment, especially those related to resource limitations and other ecological and social selective pressures that impact life histories (Sgrò & Hoffman,  2004 ; Roff & Fairbairn,  2007 ). Among female mammals, studies of life‐history trade‐offs, and their endocrine bases, have focused mainly on the timing and levels of investment in reproduction, as well as their potential effects on survival. These emphases on reproduction follow from the increased energetic costs of reproduction for females, compared to males. In particular, most research on female reproductive traits and trade‐offs in mammals involves analyses of the effects of three hormones: ( i ) oestrogens, which mainly fuel growth and development of reproductive tissues; ( ii ) progesterone, which mainly supports pregnancy; and ( iii ) prolactin, which mainly controls lactation. These hormones thus centrally mediate life‐history transitions from growth to reproduction, and from one reproductive cycle to another, via shifts in investment patterns (Hill & Kaplan,  1999 ). The primary endocrine‐based trade‐offs under this paradigm include those between growth, survival, and reproduction (mediated by the timing of menarche), and those between reproduction and maintenance (mediated by survival‐related costs of reproduction) (Reznick,  1985 ). Such trade‐offs are typically validated using tests for negative phenotypic correlations between traits expected to trade off, quantitative genetic data on genetic correlations that can demonstrate a genetic basis to trade‐offs, and experiments designed to parse alternative causal hypotheses concerning trade‐off presence and strength (Reznick,  1985 ; Bolund,  2020 ). We argue that this trade‐off‐based framework for understanding and analysing female life histories is substantively incomplete because it neglects the key roles of testosterone in female development, physiology, reproductive functions, behaviour and, ipso facto , trade‐offs. This hypothesis has its origins in a paper on female baboons ( Papio cynocephalm anubis ) at Gombe (Packer et al .,  1995 ). The authors reported that female baboons of high rank showed evidence of reproductive benefits, including shorter interbirth intervals, enhanced infant survival, and earlier maturation of daughters, all traced to enhanced food availability due to dominance. However, highly ranked females also suffered higher costs in components of reproduction, namely higher rates of miscarriage, and reduced number of births apparently due to anovulation. Packer et al . ( 1995 ) suggested that both the benefits of high rank, and its costs, were linked to relatively high testosterone levels, such that dominance traded off with ability to produce offspring. This trade‐off represented a direct consequence of female–female competition, which selects for higher testosterone levels which have both benefits and costs. Despite the high prevalence of socioecologically based competition among females in mammals, including humans (Hrdy,  1999 ; Hsiung,  2005 ; Clutton‐Brock & Huchard, 2013 ), and considerable evidence that relatively high testosterone in females involves both costs and benefits, the hypothesis developed by Packer et al . ( 1995 ) has seldom been discussed or evaluated in subsequent work. In this article, we generalize and extend their insight, by testing the hypothesis that testosterone mediates centrally important life‐history trade‐offs in female mammals, including humans. We do so using three main domains of evidence. First, we provide a brief description of the roles of testosterone in female mammalian development, reproductive functions, and behaviour. Second, we present data from a suite of non‐human mammals on associations of prenatal and postnatal testosterone with reproductive, behavioural and life‐history phenotypes. These findings allow critical consideration of the hypothesis that levels of testosterone mediate fitness‐related trade‐offs in female mammals, such that lower levels tend to enhance reproduction, while higher levels are associated with increased dominance, rank, resource acquisition, and survival but lower reproductive rates. Third, we describe evidence for such trade‐offs from studies of healthy women. Taken together, these domains of evidence thus provide new insights into the evolution of female life histories and how they are mediated by hormone‐driven trade‐offs.