The evolution of same-sex sexual behaviour in mammals

The paper addresses why and how same-sex sexual behaviour (SSSB)—defined as any attempted sexual activity between members of the same sex—evolves and persists in mammals despite apparent direct reproductive costs. SSSB has been reported in over 1500 animal species across major groups and is especially prevalent in nonhuman primates; it occurs in both sexes and in wild and captive settings, and is common in humans across societies. The behaviour can incur energetic, time, disease, and injury costs and entails an opportunity cost by not producing offspring when occurring instead of different-sex sexual behaviour, framing a Darwinian paradox. Non-adaptive hypotheses propose SSSB arises from mistaken identity, limited access to opposite-sex individuals, sexual frustration, or as a byproduct of selection on traits like high sexual responsiveness. A recent null hypothesis posits indiscriminate mating (both same- and different-sex interactions) as ancestral in sexually reproducing animals. In contrast, adaptive hypotheses suggest SSSB can function in social bonding, alliance formation, dominance negotiation, and conflict mitigation. The study’s purpose is to map the phylogenetic distribution of SSSB across Mammalia, reconstruct ancestral states, and test whether its prevalence is associated with sociality and/or lethal intraspecific aggression (adulticide).

Prior work documents widespread SSSB across taxa and emphasizes its prevalence in primates. Non-adaptive explanations include mistaken identity (e.g., in feral cats), scarcity of opposite-sex mates, frustration after rejection, and byproducts of selection on sexual responsiveness. The indiscriminate mating hypothesis proposes SSSB as part of an ancestral, non-discriminating mating strategy. Adaptive case studies suggest SSSB can serve social functions: facilitating reconciliation and reducing intra-group competition in bonobos and Japanese macaques, reinforcing alliances in male bottlenose dolphins, and strengthening dominance hierarchies in American bison. However, other systems find evidence for non-adaptive causes, indicating multiple evolutionary routes may underlie SSSB and motivating a broad, phylogenetically informed test across mammals.

Data compilation: The authors compiled records of SSSB (courtship, mounting, genital contact, copulation, and pairing between same-sex individuals) across mammals, noting context (wild vs. captive), life stage, and frequency. They built four analysis subsets to control for research effort and context: (I) all species with any data (261 SSSB, 1470 without recorded SSSB), (II) excluding species with SSSB only in captivity (209 SSSB, 1470 without), (III) species with reproductive/sexual behaviour profusely studied (205 SSSB, 252 without), and (IV) species with high overall behavioural research intensity based on citation counts (154 SSSB, 238 without). Research intensity was quantified via Google Scholar citation counts (search: species name AND behaviour). Additional covariates included sociality (solitary vs. group-living; compiled for 2546 species), adulticide (presence/absence of lethal aggression by sex), and body size from PanTheria, EltonTraits, Animal Diversity Web. Phylogenetic framework: Analyses used mammalian phylogenies with repeated sampling to incorporate tree uncertainty (e.g., 30–100 randomly chosen trees; some analyses up to 100). Phylogenetic signal (D) for binary SSSB (male/female) was computed using caper (phylo.signal), comparing observed patterns to random (D=1) and Brownian (D=0) expectations. Male–female SSSB correlation was tested with diversitree (fitMk; uncorrelated vs. correlated models accounting for phylogenetic pseudoreplication). Ancestral state reconstructions (ACE; ARD model) estimated node probabilities of SSSB presence; significance vs. p=0.5 assessed via z/t-tests. Stochastic character mapping (phytools make.simmap/make.simmulti) with 1000 iterations estimated numbers of independent gains/losses. Node ages with inferred presence/absence of SSSB were compared via t-tests using null distributions from tip-state reshuffling. Hypothesis testing: To test whether sociality and adulticide predict SSSB while controlling for research effort, four complementary approaches were used: (1) phylogenetic logistic regression (phylolm) with SSSB (binary) as response and sociality, sex-specific adulticide, and sampling effort as covariates (bootstrapped parameters; alpha for phylogenetic signal; applied to subsets I and II); (2) modelling research effort as weighting/measurement structure using MCMCglmm for subsets I and II; (3) restricting analyses to species with profusely studied sexual behaviour (subset III) with phylogenetic logistic regressions; and (4) restricting to species with high overall behavioural research intensity (subset IV) with phylogenetic logistic regressions. Directional (dependent) evolution tests: Pagel’s test (phytools fitPagel, ARD) compared four models for each pair of traits (SSSB vs. sociality; SSSB vs. adulticide, sex-specific): independence, non-directional interdependence, and two directional models (sociality/adulticide depends on SSSB vs. SSSB depends on sociality/adulticide). Likelihood ratio tests compared independence vs. interdependence; AIC weights assessed relative support; tree uncertainty addressed with repeated analyses on randomly selected trees.

- Coverage and contexts: SSSB was recorded in 261 mammalian species (~4–5% of species), spanning 62 families (~50%) and 12 orders (~63%). Behavioural forms included mounting/genital contact (67% of species), courtship (27%), and pair bonding (24%). SSSB was mostly recorded in adults (81 species in adults; 12 species in juveniles) and occurred in wild or semi-wild conditions for 209 species (83% of those with SSSB). About 40% of species exhibited SSSB at moderate to frequent levels during mating seasons. - Phylogenetic signal: Significant phylogenetic signal indicated non-random distribution across the mammalian tree. For females, D ranged ~0.4–0.9 (p < 0.0001); for males, D ~0.63–0.83 (p < 0.001 except in one subset). D values exceeded Brownian expectations (D > 0), implying that closely related species need not share the trait. SSSB was common in Cetartiodactyla, Carnivora, Diprotodontia (kangaroos/wallabies), Rodentia, and especially Primates. - Ancestral states: The most recent common ancestor (MRCA) of Mammalia had equivocal likelihood for SSSB in both sexes (likelihood ~0.5). The MRCA of placentals likely lacked SSSB: female likelihood ~0.08–0.32; male likelihood ~0.10–0.40, both significantly <0.5. Stochastic mapping suggested multiple independent gains and losses of SSSB throughout mammalian evolution. - Temporal pattern: Ancestral nodes inferred to exhibit SSSB tended to be younger than nodes inferred to lack SSSB, consistent with relatively recent origins in many lineages. - Associations with sociality and adulticide: Across control methods for research intensity, SSSB in both sexes was positively associated with sociality (e.g., phylogenetic logistic coefficients commonly positive and highly significant; representative estimates for females ~0.65 to >2.2; males ~1.04 to >1.7). Male SSSB was positively associated with male adulticide across several models (e.g., coefficients ~1.6–2.63; often significant), whereas female SSSB generally did not correlate with female adulticide. - Directional evolution tests: For both sexes, models of interdependence outperformed independence for SSSB vs. sociality, with strongest support for SSSB evolution depending on sociality (i.e., shifts to sociality precede SSSB). For SSSB vs. adulticide, male SSSB showed support for dependence on male adulticide; for females, support for dependence was weak or absent. Overall, independence models had higher AICs compared to dependent models, and AIC weights favored SSSB depending on sociality (and, in males, on adulticide).

The findings indicate that SSSB is relatively common across mammalian families and orders but unevenly distributed phylogenetically. Contrary to the indiscriminate mating-as-ancestral hypothesis for all animals, ancestral reconstructions suggest SSSB is not ancestral in mammals and likely originated multiple times, often more recently in evolutionary history, indicating convergent evolution potentially driven by selection. The robust positive association between SSSB and sociality, along with directional tests showing SSSB evolution contingent on shifts to social living, supports adaptive roles of SSSB in establishing, maintaining, and strengthening social bonds, alliances, and facilitating post-conflict reconciliation. The male-specific association with adulticide suggests an additional adaptive role in mitigating male–male aggression and intra-sexual conflict, consistent with patterns where lethal aggression reflects male intrasexual competition. While many species-level studies corroborate these roles, the heterogeneity across taxa and occasional evidence for non-adaptive mechanisms underscore that multiple pathways (adaptive and non-adaptive) likely contribute to the evolution and maintenance of SSSB. The study advances understanding by integrating broad phylogenetic analyses with controls for research effort, but acknowledges that underreporting likely depresses observed prevalence and may affect some inferences.

This study synthesizes the phylogenetic distribution of SSSB across mammals, reconstructs its evolutionary history, and tests adaptive hypotheses at a broad taxonomic scale. It concludes that SSSB is not ancestrally mammalian but has evolved repeatedly, particularly in social lineages, consistent with adaptive functions in social bonding and conflict mitigation, with a male-specific link to adulticide. These insights help resolve the apparent evolutionary paradox of SSSB by highlighting its potential fitness-relevant social benefits. Future research should: (1) expand behavioural sampling across undersurveyed mammalian taxa to reduce false negatives; (2) refine measures of sociality and aggression (e.g., gradations of group structure, dominance, coalitionary behaviour) to parse mechanisms; (3) integrate ecological, life-history, and neuroendocrine data; (4) perform comparable phylogenetic analyses in other vertebrate and invertebrate groups to test generality; and (5) explore genomic and developmental correlates of SSSB to identify proximate mechanisms.

Key limitations include incomplete behavioural data for many species and likely underreporting/false negatives for SSSB, potential biases due to variable research intensity across species, and phylogenetic uncertainty. Although the study used multiple subsets, weighting, and repeated analyses across alternative trees to mitigate these issues, residual biases could affect ancestral state estimates, association strengths, and inference about the timing and frequency of gains/losses. The reliance on binary categorizations of sociality and adulticide and cross-species comparative methods limits causal inference, and unmeasured ecological or life-history covariates may confound observed associations.