PROCEEDINGS B royalsocietypublishing orgjournal/rsph Research 8 ® Gite this artic: Brindle M,Fergson-Gow B, Wiliamson J Thorsen SommerV. 2023 The evolution of masturbation Is asacited with postopulatry selection and pathogen ‘dance in primates, FcR So. 8290: 0230061. htpsdo.org/t01098/sph.2023.0061 Received: 3 February 2023, Accepted: 4 May 2023, Subject Category: olution Subject Areas: behaviour, ecology, evolution Keywords: masturbation, postopuatory seal selection, pathogen aoidence, Bayesian phylogenetis ‘Author for correspondence: Matilda Brindle ‘ema: maildabrindle@goaglemallcam leone supplementary mae is avaiable cali a htpsiéo.eg/16084m9 sare, 653124 THE ROYAL SOCIETY PUBLISHING The evolution of masturbation is associated with postcopulatory selection and pathogen avoidance in primates Matilda Brindle’, Henry Ferguson-Gow?, Joseph Williamson’, Ruth Thomsen’ and Volker Sommer! "Deparment of Antwaplogy, Univesty Clge Landon, 4 Tavton Suet Landon WCIH OB, UK he nsttte af Zuogy, Zoli Sct of Lando, Regent’ Pak, London NE 715, UK este fr Bodivesy and Envanent Reseach, Departmen of Genetics, elton and Envanment, Unietsty Coleg Londan, Gover Set, London VTE 68, UK “Seta! of logical and Cel Scenes, Queen ary Univers of Londo, Landon E1415, UK (© is, co0-c03 2232-481; HG, cooo-oom7151-6421; IN, 000-003 451653, 1, ono onoe7s-32 ‘Masturbation occurs throughout the animal kingdom. At first lance, how= ever, the fitness benefits of this self-directed behaviour are unclear. ‘Regardless, several drivers have been proposed. Non-functional hypotheses pposit that masturbation is either a pathology, or a byproduct of high under- lying sexual arousal, whereas functional hypotheses argue an adaptive ‘benefit. The Postcopulatory Selection Hypothesis states that masturbation ads the chances of fertilization, while the Pathogen Avoidance Hypothesis sates that masturbation helps reduce host infection by flushing, pathogens from the genital tract, Here, we present comprehensive new data document- {ng masturbation across the primate order and use these in conjunction with phylogenetic comparative methods, to reconstruct the evolutionary path- ways and correlates of masturbation. We find that masturbation is an ancient trait within the primate onder, becoming a more common aspect of the haplorrhine behavioural repertoire after the split from tarsiers. Our ana~ Iyses provide support for both the Posteopulatory Selection and Pathogen ‘Avoidance Hypotheses in male primates, suggesting that masturbation ‘may be an adaptive tat, fmctioning, at a macroevolutionary scale. 1. Background Autosexual behaviour oF masturbation, is common across the animal kingcom, but appears to be partculaely prevalent in the primates [12]. There is lite sys- tematie comparative rseazch into this behaviour and its evolutionary history is unclear. The proximate driver of masturbation isnot difficult to establish, since genital stimulation is reinforced unough the same hedonic feedback [3,4] repaniless of whether its ato- or allosexual. However, at a superficial level, rasturbation poses a problem for evolutionary theory. It does not directly increase survival prospects and, by definition, occurs tothe exclusion of eprom ductive partners, while incurring costs in terms of time, attention and energy. ‘Consequently, masturbation has historically boen consicened, a worst, a patho- logical behaviour eartied out by aberrant, typically captive, indivicuals and, at best, a sexual outlet necessitated by high libido [5]. However, the Pathology Hypothesis fails to account for antosewuality in wik-living primates. The Outlet Hypothesis may not account for masturbation immediatly pre- oF post- coptlation, or in the presence of willing partners, particularly én males of species with an extended refractory period (Le. the period of time after ejacula- tion, where sexual arousal and activity is reduced [3). In any case, the idea that masturbation may serve as a sexual outlet is not mutually exclusive with © 10% The hats, Pied by the Roper the tems he este Coons Auton Liens hp:fortivcermensoriesey/, hi pes unesticed we, pve the cgi autor rd source eastadaptive explanations, since arousal can be understood. as the proximate mechanism of the indirect ultimate causes {Gc functions), ‘The Posteopulatory Selection Hypothesis [1,26] predicts that masturbation aids the chances of fertilization anc is composed of two constituent hypotheses: the Sexual Arou- sal Hypothesis and the Sperm Quality Hypothesis, The Sexual Arousal Hypothesis posits that non-ejaculatory male ‘masturbation may speed up subsequent ejaculation, benefit- ing low-ranking males in danger of being outcompeted by rivals (eg. in marine iguanas, Amblyrlaynchus crstatus (7 and Japanese macaques, Macacn fuscata [8)), or increase ejaculate quality (eg, in humans, Homa sapiens, [9). Similarly, the Sperm Quality Hypothesis. states that ejaculatory precopalatory male masturbation may expel inferior sperm, improving subsequent ejaculate quality. (eg. in humans HIDE Japanese macaques IS; and Rhesus macaques, Macaca ‘mulata 111). The Sexual Arousal Hypothesis could also help explain female masturbation, since sexual arousal and ‘orgasm may facilitate cryptic female choice. Female arousal increases vaginal pH, creating a more hospitable environ- ‘ment for sperm [12]. The vaginal transudate associated swith arousal filters out inferior sperm, while facilitating the transfer of high-quality sperm towards the uterus (13). Sim- larly, the contractions associated with female orgasm may enhance the passajge of sperm through the uterine cavity, and associated secretions of prolactin capacitate sperm 113,14], While litle is known about the storage and survival durations of sperm within the female reproductive tract in ron-human primates, in humans, sperm can be viable for greater than 5 days (15,161. It seems likely, therefore, th female primates could use pre- oF postcopulatory masturba- ion asa strategy to increase their chances of being fertilized by a given male. Alternatively, masturbation could also serve as a form of precopulatory display or courtship be- haviour in both sexes [2,17], similar to precopulatory ‘penile displays’ in chimpanzees [18] ‘The Pathogen Avoidance Hypothesis predicts that mas- turbation is a form of postcopulatory genital grooming helping to prevent sexually transmitted infections (STIs). Be- havioural strategies are often the frst line of defence against infection, followed by physiological and immune responses [19]. Postcopulatory grooming, strategies, such as oral self- cleansing oF urination, have been adopted! by various mam- mals (iruit bats, Cynoplerus sphinx [20]; cestain callitichit and strepsirrhine primates [19]; humans [21]; anc murine rodents [22), An altemative strategy is used by male Cape ground squirrels (Xerus inauris), who masturbate after ‘mating, cleansing the reproductive tract with ejaculate 171 We employ Bayesian phylogenetic Markov chain Monte Carlo (MCMC) models and hierarchical modelling to recon- struct the evolutionary history of masturbation and test the basic predictions of two functional hypotheses (the Postcopu- latory Selection Hypothesis and Pathogen Avoidance Hypothesis) within a comparative framework for the first time, using a large-scale dataset charting masturbation across the primate order. The Postcopulatory Selection Hypothesis predicts that masturbation should be associated with high postcopulatory selection pressure, os measured through the proxies of multi-male mating systems and rela- five testes. mass. The Pathogen Avoidance Hypothesis predicts that taxa are more likely to masturbate if exposed to a higher pathogen load, as measured via the proxies of ‘geographic region, mean annual precipitation and environ= mental harshness, as well as the more direct measure of STI load. 2. Results (a) Data assembly We assembled the largest dataset of masturbation occurrence to date, compiled from nearly 4 sources, including 246 jrublictiors and 150.quictionaine reponse and peioral ‘communications from primatoogists and zoo-keepers [12 We combined this dataset with a posterior distribution of 10000 molscular phylogenies spanning the primate order, from the 10k Trees Project [23], Our data covered 105/272 species (386%), 51/67 genem (8.6%) and 18/19 (subliami- lies (47%) within our phylogeny. The proportion of species from each major radiation for which we have data varies, and likely reilees a eseatch bins torenrds larger bodied primates, Thus, data were available for 197% of lemurs, 33.3% of lores and_galagos, 200% of tarsiers, 4185 of plaiprhine monkeys, 404% of entarthine monkeys and 682% of apes “The data cover both sexes, with data for ferales covering 67.2% of goners (Min 19), and males covering 76.1% of Aenea (Myeies=83}. This allowed vs to ran all analyses separately for females and males, to account for potential sex differences. th il ots patna re copies Th ost dataset, though reports of masturbation are more common for captive primates (n=298) than wild individuals ( 105): Howeven given that many mpeties are more comma, studied in captivity, the absolute number of reports is not necessarily informative, A-more appropriate measure is to compare how many studies on captive versus wild primates report masturbation, or the lack thereof. In our dataset, mas- turbation is reported to be present in 74.5% of stucties on captive females, and 874% of studies on captive male, versus 25.4% of studies on wil females, and 73.3% of studios on wild males. (b) Phylogenetic signal and ancestral state reconstructions Both female and male masturbation showed phylogenetic signal indicating that species with more recent shared ancestry have more similar trait values (electronic supplementary material, table SI) Ancestral state reconstructions of feniale masturbation (electronic supplementary material, table $2; figure 1; Baye- sian MCMC ancestral state reconstructions; hyn =49) indicate that masturbation was present in ancestral (non- tarsiex) haplorthines (node D; mean probability =0.76), eatar- rhines (node F: mean probability ~0.91), apes (node G; mean 195) and catarthine monkeys (node H; mean 87). Other nodes were reconstructed with less confidence, with the model indicating that masturbation vas probably present in the ancestral haplorthine (including tarsiers; node C; mean probability =0.65) and platyrrhine monkey (node E; mean probability = 0.63), but absent in the ancestral strepsirrhine (node B; mean probability =0.67), ‘The estimation of female masturbation occurrence at the root strong \90nez02 062 8 205 y ony adsyjeunolfioSousndhoposeias ACatanthine monkeye Platyrhine monkeys Tarsiers Lemurs Lorises & Galagos Figure 1. Phylogeny of 67 gener ofthe primate ore, stating reonsruted acest nodes and rats of extant genera in female and male primates eecroic supplementary matt, able 5) Reported presence (back squares) or absence (arty squres) of masturbation in extant ems (9) and males (in les one specs ofthat genus sindcate atthe tps ofthe we. Empty tips inate alk of data. At the ade ofthe we, lack shading inates masturbation presence, re shang indicates masturbation absence (bth mean probability greater than 0), and white indicates equivocal recast (meen probity les than 40), Ancestal state recess showed tal the same pate in feral and mal priates, withthe exceptions of rede wih indicate masturbation was preset in eal (eft half of ce) but was equnor or mae ight half of cic), nd nde B, which nde that masurkaon was absent in females (kta of dc) but was equal for males gt af of cide). Maximum dade ced ee ces rama sample of 1.000 molecular phylogenies fom the Were project [1 ‘was equivocal (node A; mean probability of masturbation presence =0.55) Ancestral state reconstructions of male masturbation (le tronic supplementary material, table 2; figure 1; Bayesian FIMCMC ancestral state reconstructions; tepcies =83) followed te same broad pattern. Masturbation was present in ancestral (non-tarsicr) haplorrhines (node D; mean probability = 0.74), catarrhines (node F; mean probability =0.92), apes (nodle C; mean probability = 0.95) and catarthine monkeys (node H; ‘mean probability =0.97), The ancestral platyrrhine monkey node was reconstructed with less confidence, but masturba- tion was likely present (node E; mean probability =0.67) ‘Compared to females, ancestral male haplorthines (including tarsiers; node C; mean probability of masturbation presence = (058), and strepsirthines (node B; mean probability of masturbation presence =0.49), reconstructed with lower confidence, In keeping with females, the estimation of male masturbation occurrence at the root was equivocal (node A; mean probability of masturbation presence = 0.56). (0 Postcopulatory selection hypothesis Different mating systems vary inthe leve! of postcopulatory selection and associated behavioural and morphological phe notypes they generate [24] For example, multimale mating, systems produce high levels of postcopulatory male-male competition, and are associated ‘with lange testes to body mass ratio [525] and elongated bacula [24], We therefore ayy cating tend teed ina CG ak by soe) as proxies for postcopulitory selection, using the compilation of data in Brindle & Opie (24) We found support for coevolution between masturbation presence and mating system in male, but not female, pri- nates, providing support for the Postcopulatory- Selection snr 62 6 24 ay abominations ElFigure 2. Coelution between masturbation occurence (present versus absent left-hand side of cds in schematic and mating system (ingle versus mul- mal, ght-hand sie f cides in seat). Tais present in a ien tat ar stown in Hack, and thas absent ar shown ing. Z dente the pretae of tnaels where a tanstin didnot ocur,utch indicates how Hy a tanstion i to have taken place. Bak aowsconespond to tanstons wih 2 very ow poster petty of being zero (e,vxy common tstn), dak are aro toa opti pobabity of being ze (Le, common tanstions, and light orey arows to 2 high posterior probability of being zero (ie. ae tans) Hypothesis in males (electwnic supplementary material, table $3; figure 2; FMCMC models of correlated versus independent evolution: tzpcin=299; male masturbation and mating system, log BF=2.44; female masturbation and ‘mating system, log BF = ~2.40). Examination of evolutionary transitions revealed that, in male primates, shifts from mas- turbation absence to presence occurred in both single- and ‘multi-male mating systems (Z = 18.8% and 0.2%, respectively where Z is the percentage of models where a transition did not take place, indicating the likelihood of a given transition) though more frequently in multismale mating sy tems. However, masturbation was also lost frequently in single male mating systems (Z=0.0%), but almost never in multi-male mating, systems (Z=98.7%). This inclicates thal—while masturbation is a very labile trait in single-male mating, systems—in multi-male mating systems, once male ‘masturbation has evolved, it persists, We dic not find evidence fora relationship between mastur- potion and adult male testes mass in either female or male primates (electronic supplementary material, table S4; figure 3). (d) Pathogen avoidance hypothesis Many parasites require hot, humid conditions or water, complete their life cycles. This may explain why global para- site diversity tends to increase towards the equator, and is closely linked to climatic variables (26,271. For this reason we employ (primate geographical region temperate ves tropic!) (2) mean atmual. poepliation, and (i) environmental harshness (the degree of climatic prediiabil- ty) as a proxy for pathogen load. We also invoke the more dlircct measure of sexually transmitted pathogen occurrence in wild populations, gleaned from the Global Primate Pasa- site Database (GPPD) [2], which was eatogorized as either present or absent for a given species. We found very stong evidence for coevolution between masturbation and pathogen occurence in males, but not females, providing support for the Pathogen Avoidance ‘Hypothesis in males (electronic supplementary material, lable $3 figure 4; fMICMC models of correlated versus inde pendent ovalution; yaa=299; male masturbation and pathogen occurence, log BF=1.25; female masturbation ancl pathogen occurrence log, BF=~ 122). Evolutionary path analyses indicate that, in male primates, shifts from masturs- tion absence to presence occurred regardless of whether pathogens were absent or prosent (Z=06% and. 267%, respectively). However, masturbation was los at a very high rate when pathogens were absent (Z=0.0%) but almost never when they were present (Z=99.8%), This indicates that masturbation isa labile trait when pathogens are absent, bout when pathogens are present masturbation is retained. By contrast, our analyses with respect to environmental predictors of pathogen risk failed to find any patterns. We found no evidence in support of coevalution between masturbation and geographical region in female oF male pri- mates (electronic supplementary materia, table $3). Similarly, there was no support fora relationship between masturbation sin at a a” eeavenlrtaapilonaeta.ec a a # 5 standarttedestiate Fir 3, Sond ests citer) mation pesos infer rd male pias th at make ees mas (man atl pe Cffaton iad eierent hanes tor, os nate seer mea i atl nese 9 Been cele te), nd hn fro sine 98 Bs Figure 4. Coevluton between masurbation occurence (present versus absent left-hand side of ccs in sthematc) and pathogens (presen versus absent; right hand side of cles in schematic). Figure 2 fr furber deals To confirm that our results were not an artefact of poten tial coevolution between pathogen presence and mating system, we also checked these variables for coevolution, but and mean annual precipitation or environmental harshness, neither females or males (electronic supplementary material table S4; figure 3),didnot find support for this model (electronic supplementary ‘material, table S3). 3. Discussion While some consider masturbation 0 be pathological (ex. 29) or a byproduct of neuroendocrinological specialization for high sexual arousal [5], others stress its adaptive role in postcopulatory selection [1,230] or STI reduction [17]. The data presented here highlight the inadequacy of the Pathology Hypothesis to explain autosexuality in primates since, in our dataset, masturbation is reported in over a thin of studies fon wild females, and over twoethin’s of reports on wild males (Our analyses demonstrate that masturbation has a strong Phylogenetic signal and is an ancient trait within the primate conier. We also find that both posteopulatory selection ancl pathogen presence have been associated with masturbation ‘across the course oF evolution in male, but not female, primates With regants to adaptive explanations, we found support forthe Postcopulatory Selection Hypothesis in male primates, swith masturbation covarying with mating system, but not {esios mass. Male masturbation presence is unstable in single-male mating systems, but evolves and is maintained in muli-male mating systems. We also found strong support for the Pathogen Avoidance Hypothesis in males, with mastur- bation coevolving with pathogen presence. Similar to the results for male masturbation andi mating system, when patho- gens are absent, male masturbation presence is labile, but ‘when they are present, masturbation evolves and is retained. ‘There are several explanations for why masturbation could coevoive with pathogen presence in males but not females, One ofthe key functions of seminal plasma isto pro- ‘vide antibacterial defence for sperm once it reaches the female reproductive tract [31], ut these bactericidal proper- ties also protect hosts from STIs [32]. In males, ejaculate is forcefully expelled through the urethra, whereas during female arousal, vaginal transudate passes gradually through the vaginal wal, coalescing to form a film which is eventually secreted [33]. Since the urethra is the primary site of infection for many STIs [21], the powerful expulsion of hacteriidal ja- culate through the urethral opening is likely to bo a more effective means of avoiding infection than the slow secretion of transudate from the vaginal canal, which does not necess- arly cleanse the urethra. Furthermore, unlike in males, the Posteopulatory Selection Hypothesis ancl Pathogen Avoid+ ance Hypothesis may be mutually exclusive in females Under normal conditions, when a female is not aroused, ‘vaginal pH is moderately acidic, in order to defend against infection [121 However, according to the Posteopulatory Selection Hypothesis, one mechanism of cryptic female choice is that, during. arousal, vaginal pH increases to become more neuiral, While this creates a more hospitable environment for sperm, allowing females to differentially favour certain males, its also less hostile towards pathogens, leaving females vulnerable to infection [12]. ‘What is more surprising is that we did not find support for the Postcopulatory Selection Hypothesis in females, Neither female nor male masturbation showed an association with adult male testes mass, in contrast to previous findings [5], and female masturbation did mot eoevolve with mating system. The discrepancy between our results and previous gh research may be due to our introduction of a phylogenetic framework, in conjunction with a larger sample size, high- lighting the importance of using rigorous phylogenetic methods for comparative rescarch. Regardless, this docs not explain why male masturbation has been associated with multimale mating systems across the course of evolution, but female masturbation has not. It is important to note that there are far fewer reports for masturbation in female pri- ‘mates in our dataset. This is in part because female arousal and masturbation can be less conspicuous than that of males, but also reflects a broacler paucity of information on female sexual behaviour and morphology in the biological sciences. While it is possible, therefore, that female masturba- tion is not associated with a high degree of postcopulatory selection pressure, it is important to gather more data to fully assess this hypothesis. ‘The results presented here are consistent with the idea that primates have evolved behavioural strategies, in this ‘ease masturbation, to reduce the risk oF STIs. However, a pre- vious study 1341 found no evidence that species at greater risk ‘of infection (Le. primates that mate with multiple partners, as defined by relative testes mass and oestrus duration) were more likely to employ postcopulatory strategies to cleanse the reproductive tract: oral-genital grooming and urination BA], Tnstead—while most strepsirrhines and_callitrichids show stereotypic oral-genital grooming and urination after ‘copulation—such behaviour is rarely seen in langer primates (catarrhines and non-callitrichid platyrrhine monkeys), per haps because they are less able to reach their genitals ‘orally. This may indicate that these traits are not envionmen- tally driven, but reflect derived phenotypes. Interestingly, strepsirhines and calltrichids also engage in scent-marking behaviours, including urination and anogenital rubbing [19], while masturbation is rare in these clades. It therefore ‘seems possible that strepsirrhines and callitrchids are predis- posed to urinate and orally groom their genitals, rather than ‘masturbate, in order to mitigate STI risk. By contrast, in the absence of scent-marking and oral-genital grooming, larger primates may use masturbation to reduce the risk of STIs. If masturbation had already evolved in this group alongside increased postcopulatory selection pressure it may have sub- sequently been @xapted for pathogen avoidance (and rmainlained via the combinod influence of postcopulatory selec: tion and selection for pathogen avoidance) Indeed, ancestal sate roconsinictions indicate that masturbation vvas present in ancestral haplorshines—after the split from (arsiers—and por- Sted in ancestral calasthines ane! all subsequent nodes (apes and catarshine monkeys). I is also likely that ancestral platyr- shine monkeys masturbated, although this node was less well solved, particularly in females. However, the models suggested! that masturbation was likely absent in ancestral female strepsirines, although they were equivocal regarding males, Phylogenetic approaches can only capture a glimpse of the past, and the view can be incomplete or misleading ‘when based only on present day information (35). Our data- set contains less information for strepsirrhines compared with haplorthines and, as such, the ancestral state of mastur= bation occurrence in the strepsirrhines was less well resolved than for clades with more data. Similarly, of the non-tarsier haplorshines, there are fewest dats for platyrrhine monkeys and, corespondingly, their ancestral state is reconstructed Lo00tzo 062 # 205 y 2m qdsyjewnol Sue Suusyqndiapeseko,‘with a lower probability than other nodes. It may be that the lack of data in these taxa reflets the fnet that absence data aze less likely to be reported. Regardless, the uncertainty sur- rounding the strepsirthine ancestral state has likely had a knock-on effect on estimates of the root of the phylogeny, ‘which are also equivocal. The lack of certainty in these recon- structions is also likely to reflect the fact that the basal primate is the most phylogenetically distant from extant taxa, which has an adverse effect on ancestral state recon structions. Gathering more data on strepsirthines, therefore, would be an important step towards resolving deeper nodes within the phylogenies. While there was strong support for the Pathogen Avoidance Hypothesis when examining, pathogen load, no support was found when indirect measures of STI risk were used, namely geographical region, mean annual rainfall and environmental IRarshness. The most likely reason for this is that environmental predictors are poor indicators of STI risk. Although there is ‘300 evidence that such environmental variables are coreated ‘with pathogen diversity (eg [2627D, this may be driven by pro- tozoan parasites [26), which are primarily transmitted via insect vectors. Singe STIs are passed on via direct contact between indi ‘viduals [36], the environmental measures employed as proxies Ihere may not have adequately captured regional variation in STI risk. It is also possible that puthogen load is associated with other variables. that may influence masturbation. For example, there is a positive relationship between host group size and the prevalence of direcly transmitted parasites across taxa [37] It would therefore be instructive for future research to explore the potentially complex relationships between mastur- bation, pathogen load and other factors associated with ‘group-iving in primates. It is also worth noting that, although we found strong support for coevolution between male masturbation and pathogen pressure, categorizing pathogen load as ‘present’ or ‘absent’ in a given species is highly reductive. Similarly, ‘masturbation rate and occurrence vary both within- and between-species, so that some individuals masturbate far more frequently, and some species far more ubiquitously, ‘than others. In reality, both pathogen load and masturbation ‘occurrence fall along a broad continuum. It would be useful {to examine whether primates witha higher pathogen load aze ‘more likely to masturbate. With additional data, this question may be addressed in a more nuanced light. ‘Our findings provide a promising indication tat mastu- bation may serve an adaptive function. We highlight important avenues of future enquiry regarding the evolution of primate masturbation that likely requite empirical ‘ground, truthing’ via observational studies in a range of species, in addition to further comparative analyses. For example, to test between the Postcopulatory Selection Hypothesis and the Pathogen Avoidance Hypothesis in a given species, itis important to know whether masturbation occurs before (osteopulatory Selection Hypothesis) or after (Pathogen Avoidance Hypothesis) copulation. Similarly, support for the Postcopulatory Selection Hypothesis in male primates does not allow us to distinguish the relative importance of ‘two of the hypotheses it is underpinned by, namely, the Sexual Arousal Hypothesis and the Sperm Quality Hypoth~ sis. To test between these, we need to establish whether ‘masturbation occurs with ejaculation Sperm Quality Hypothesis) or without (Gexual Arousal Hypothesis). Finally, although we did not find evidence in support of the Pathology Hypothesis, it would be informative to compare wwithin-species masturbation rates in wild and captive indi viduals, to assess the relative effect of captivity on ‘masturbation rate. Here, we illustrate that masturbation is not simply a ‘pathological behaviour, and is unlikely to solely be a bypro- duct of high sexual arousal. We provide the first evidence that both postcopulatory selection pressure and pathogen avoidance may influence this common, but little understood, sexual behaviour at a macroevolutionary scale, Considering the behavioural and socioecological complexity of primate societies, it is likely that primates employ masturbation as a flexible strategy according to the circumstances they find themselves in 4, Methods While the taxonomic clasifcation of primates is subject to constant revision (eg. 138), the laxa in Our analyses reflect the fetant species prsunt in the l0kLrw Genbank phylogenies [23]. In rare circumstances when data were given to subspecies level, these wer grouped as a single species. (a) Masturbation data We define masturbation as the selftimulaton ofthe anogenital cor breast nigh, cared ont with an individual's own body parts fr extemal tools [1.2] This operational definition eas internal motivations as a black box, albelt without ascribing to the stance that mental causes are absent Data were compiled fom the Iiterature, as well as question- aire responses. Particular care was taken to solicit information from primatologists and zo keepers who had extensive expert cexce with taxa for which published Iterature provided litle or 10 information about mxisturbation (deals in (12. Species ‘with a single confirming scientific eport were considered to be masturbatory, Thus, eports of masturbation presence overrode those of masturbation absence, regardless of the number of times absence was meported (Le 10 people said a species did not masturbate, but one individual confirmed they di this species would he cased as masturbtory). Conversely a species ‘was not imply casi as non-masturbalory if there were a lack of reports on the behaviour, because behavioural studs of many primate species are limited, and masturbation may be ore, 50 anwedotal incidences are cnlikely to be published Instead, we only inchided cases which explicitly state that the Tehaviour was not observed (b) Postcopulatory selection data Primate specios were clasiied acconing to sehether their ‘mating system and anatomy’ were likely to be associated with high or low posteopulatory competition, Inthe fist case, species with singlemale mating systems (monogamy, polygyny) were considered as having low postcopulatory competition, while species with multimale mating systems (polygynandry, polyandry) were considered to have high posteopulatory compe- tition, following Brindle & Opie [24], Species with intraspecitic vriabiity that routinely spans both sides ofthe binary grouping were discarded from the analyses. Inthe second case, species with larger adult male testes mass (while controlling, for body size) were identified as experiencing higher posteopulatory seletion (25). Mating system (single-male versus mul-male; Maps =181) and estes MASS pane 75), were therefore cermployed as proxies for postcoptlatory selection pressure These data were taken fom Brindle & Ope [2 L90nez02 062 g 205 y ony adsyjeunolfioSousndhoposeias A(¢) Pathogen load data and environmental proxies ‘Two different methods were adopted to evaluate the risk of infec- ton from sexually transmitted pathogens: a direct approach, in which we assessed the fnquency with which species swere recorded to carry sexually transmitted pathogens; and an indirect approach, in which we assessed environmental corelates of pathogen abundance. Data on environmental mansures were available for a wide sample of primate species [39], while the ‘sample for sexually transmitted pathogens was less extensive [28] In the first approach, a measure of sexually transmitted pathogen load in wild populations was gleaned from the Global Mammal Parasite Database v. 2.0 (CMPD; note that ‘hore ‘parasite’ also rofers to viral, bacterial and fungal pathogens op. Data were filtered to include only sewually transmitted pathogens, and only those primate species included in the 10k ‘Trees GenBank phylogenies. Data in which primates were only labelled to: genus were also romoved. While itis unlikely that pathogens are truly absent across a species, for the purposes of ‘our analyses, we marked them 9s such if all reported pathogen screens were negative in at least 10 individuals (see electronic supplementary material for further discussion), In the second approach, given the limited availability of data fon pathogen load, we assessed the environmental correlates of ppathogen infection risk. In the light of previous stucles (eg, 12627), we identified three potential predictors. These comprised () geographical ragion (temperate versus tropical jee covered = 7 and 117, respectively), (i) mean annual precipitation (geo V- cered!=147) and (id environmental harshness Utes covered = 147) as proxies for pathogen load. These data ‘wer’ collected from Bote ef al 391 who maasure precipitation as the mean annual precipitation averaged acts a species’ geographical ‘ange, and environmental harshness as the amount of exposure to dec, less productive environments, with colder, and more vari able annual temperatures (with higher scores indicating harsher environments). Only data for primate species included in the 10k “Troes GenBank phylogenies were collated from the database. Ana- Iyses of the environmental correlates. were restricted to only include species for whieh data swore available for all variables ‘within a model (d) Phylogenies Compartve analyses must incorporate phylogenetic infor mation 1 comet or the confounding efiecsof shared Frasca pada gic casts ala ot ‘ical independent 41. We therfore emducied phylogenstic dralgies acsw a ptrior diarbuton of 10000 olor Phylegevesfllowing Gordan teonory, om the 10 Tras Popa ¥. 40 Bal. Where» single toe wee Tete, wre gered a mani cae ety tee om our dist triton of 10090 molecular ullrmetie phylogenies, wing TrevAnotator (pat af the BEAST software package LD. (e) Phylogenetic signal The phylogenetic signal of female and male masturbation was caleulated using the “phylod” function within the R package ‘caper’ 13,4 (f) Ancestral state reconstructions ‘We used Thyesian MCMC framework to reconstruct the eval wionary history of masturbation across the primate order for females and males, sing, the “MulliState’ function in BayesTeaits v. 30 (45). The analyses were cazriad out at the species level, with missing data included as such, so that nodes cold take one of two states (masturation present versas mas furketion absent) reveriblejamp hyperprior approach was employes, seeding, from an exponential distribution with a range of 0.00-0.05, Hyperpriors allow the details of the prior dis- tribution to be estimated from the data, rather than an @ priori assumption, which is useful where investigators have Title or ‘0 information about the mean and vaziance of rate coefficients 6), Each model was run three times, and the results wore chwcked for equivalence. We then selected the model with the median log marginal likelihood (following, (47). Mosils were rn for 3000000 iterations, with a burn in of 50000 iterations. Ancestral states of interest were reconstructed via the ‘Add MCA’ function in BayesTraits v. 20 HS}. The nodes recon- structed wore the most recent common ancestors of (A) all primates, (B) strepsirrhines, (©) haplorrhines, (D) haplorehines ‘excluding tarsiers, (2) platyrchine moaleys, (F) catarhines, (@) apes and (H) catarrhine monkeys. Using this approach, the probability of a given state at any one node iS generated. We ‘consider probabilities >0.70 as providing high confidence of a ‘ven state, >0.60 as low confidence and <0.60 as oquivocal (Coftowing (47D. (g) Correlated evolution Towobinay tits uch 8 masturaion presence vers absence, nd singlemale verse ml-mae mating systems) can prosce four divorent combinations of observable sates within a species across a phylageny. If traits have coevalved, the rate of change tetcen two sates should depend on the background sate of the other [6 If the tits followed an independent pattern of evolution, the rte of change of one state should be independent th othe. By combining information on anceiral states with Posterior rate distributions Ge. tition rate it is possible fisoern the probable cnevolutonary pathway two tits have taken [4, We tested for coevoution between () masturbation and mating system and (4 masturbation and pathogen presence, in both female and male primates In addition, we checked for ccovoliton between mating system and pathogen presence, t9 fide the possibility that the independent vores were ints Gn stir “Todothis we employed reversiblejump Marko chain Monte (Caro MCMC) techniques in Bayesratsv. 3.0 6) to imal tancously test for convolution between binary varabis and «stmt traston mites beeen sats This metho fs contin foustime Markov mods to discrete tris allowing them 10 transition to another sae al any fine. The MCMC isis each ml in proportion to ts posterior probabil searching for the testing model to describe the coevolution of two tits across a phylogeny. We cimpared theft ofa dependent model of eval viion a which tana mss wn fee to vary swching ‘cnt Sand‘ to that of am independent meal when ramsiton Totes ewan state: were consimined to be eal, Tamiton tates sere estima Hy taking the moan postr ate etn cross all posterior models for which a given tststion was SeEns Whale eal thn portions a which toch parame was etm fo be eno (ie. when 9 tain twas “sched of?) within the rMCME mod “Ta ensure these estimates were accurate a sable, with ite rumsto-un variation cach chain wos run thre times, the chains converged well, the model with the median log marginal iki hood veas chosen (allowing [471 Markov chains were run for 5000000 iterations, after a burnin of 500000 somples. We employed a reveriblejump hyperprior approach, seeing fom fan exponential distribution of 0-05. Modal ft was established ting Lng Bayes Factors (BF) = 2imp-my). Natural lg marginal likethonls f the depencent (np) and dependent mn) moxtels wore estimated using a sieppingstone sampler oot with 100 stones, each run fo 10000 erations, and the default parameters ff a= 04 and f= 1. BFS were interpreted following Kast & Raftery [SSF 0-2, minimal sepport 2-0, postive sepports 10, steong export rena tcs 10, vary esong eupport L9D0EZOT “067 § 205 yoy —Qdsu/jeUsNo(/fi0-BurysyqndAaposyekas &(h) Phylogenetic logistic regressions ‘We built Bayesian phylogenetic binary logistic regression models using the ass package in R [4449], We modelled masturbation pprasence asa binary response (present versus abnent) separately for two different environmental fxed effects (mean asia! ‘precipitation and environmental harshness), and one posteopula- tory fixed effect (testes: mass). All analyses were run for female and male primates separately, esiting in four environmental models and two postcopulatory models, We accounted for phy- logenetic relationships in all models by inchuding a variance: covariance matrix of phylogenetic distance as a random efic. We includes adult male body mass a8 3 fixed etoct in or post copulatory models to contol for the effets of body size om testes ‘mass, rather than using residuals from a separate regression between the two variables as data, which ean lead to biased ‘parameter estimates 150) We ron four chains for 10000 iterations, with a burn-in of 5000 iterations using default priors. Delta was adapted 10 (0.9999 anki maximum tree depth set to 12. Model performance ‘was evaluated by visually assessing trace plots as well 3s exam- ining Rehat varies (with those <.01 considered to have mived well) and effective sample sizes (ul BSS should be greater than 100 times the number of chains, so in our case gueater that 00), We considered statistical differences to be substantial References 1. Bile M, 2022 Aatsnal behav in pte: 1 fm, pylageny and funon. Dol thes, Uniesty olege Londen, London, UK 2. Semmes Thaen RB M202 Masson in pes. In The Conse habt of actinay 12 especies a seul pyc eT Shae, mb, UA: Ganbie Uniety Fes. 3. Dicom AF. 2012 Primate seule compare ste ofthe prosiios, mans apes, ond ua bins. utd, UE: Oxford Unesty Press. 4. ft, en PFs. 222 Masurbaton inthe 14, imal Kinga, Ser es $9, 11. (dot 108 omzpp.am220446) Dusan Af, Andes M2004 Seal behavior, 15, repute ploy and sperm crpetitian in {4.106 physbe 200 08.027) 6. Bind W208 Pscaplxr selection. in 3o-et ngeopd of nial ogaitn and Whar eds) 17 Wk, T Shaded, pp. 1-4 Cham, Sten Springer itemsionalPubishing 7, Wl M, Bae S. 1886 Pc BSc 8 263, pone 012060) 459-484, (410.108.1996 0066) 18 ison AF, Mundy M1984 Sera behav, sexual seling and pele ecuton in chimpanzees (an 27 trodes) Ah Se Beh 23, 267-28, 400. & Thomsen Sis J. 2004 Male masurbaton in fre ranging apanese macaques. it. Pinot 25, 1033-104, (dit. 123/8:U0R NODS. 7569789) ® 9, Pound Ned MH, ube Shah MA, ot Yelle AP 2002 Out of sens cus pris semen paametes fr maura acts, Phyl Sew 76, 685-685. (0.1016 0031- sooo) Pres arrests) 10, Baker RR, Bells MA, 1993 Hunan spe 20. Tan, Jones 6, Zhu 6 Ye, Hon , oS, hang 6, Thang | 209 Feo by ft bats plang option time. oS ONE 4 e555 (S10.13747 30. Thr R, Sos, Tetscher C2003 Sperm jul goe00755) competition: ate ajstment by males a the anon of massbain. Am. Boh 6, 1-H, (10.106 19%.1271 Dubuc C Coyne SP, MosipeD. 2073 Elst of 2 mating acy and dine ark on male masturbation among teeing male esis macaques. aogy 118, 106-1013, Meson C200 The phephpilogial assent 22 offend seal futon. J. Sr Eb Theopy 25, 616, (det0soeno61457620007107829) Suarez 8S, Pay AA. 2005 Sper tnspor inthe ‘ele reproductive tat Hum Rep. Updte 2, 2-2, (dot 10.1093hurspldms7) Pats DA, Danood K 2006 The evn of fmale game: aapaion or byproduct Ten Rs. um. Gent 9, 467-42. Gok 1OI3TSwn9.3467) Johason M, Evi 1588 seman. Cif, UK: Blakvl Sei Publications. rmle mara. Pil, lay, 3, 361-371. 16, Martine . 1889 Sperm tamper inthe haan ‘ere repute tact. Ot fer ped Bi §, 25 ateman JM, 2010 The adaptive fucian of mason in 3 promisaos Aicn gad 25, Nunn CL, ie SM, Sch Caring Ad Squire, FONE 5, e13060. ok 10.137 jour Nunn Cater S. 2004 Sera selection, Deavaut and sonlly tse dens In 28 Sea secon in rites ew ond compart perspectives eds PM Kael, Can Shai, pp. 117-130 Cambie, Ut: Cambie Univesity if the 95% Bayesian credible intervals did not overlap 7er0, and ‘moderate if 50% Bayesian credible intervals did not overlap 2ero, Dia acessity. The data and code ore provided in tl supplementary material [51 Autor’ contin. MLB. conceptualization, data. curation, formal analysis, funding acquisition, investigation, methodology, projct administration, validation, vicualization, writing—original draft, wnitingreview and editing: H.P-G: methodology, writing-review. and editing: JV. formal analysis, writing —review and exiting: RT: data curation; VS: conceptualizatio, data curation, superv- sion, writing review and editing, ‘Al authors gave final approval for publication and ogres to be hel accountable fr the work performed thers, aft frees dedaatin. We declare we have no competing interests. funding. This work was supported by the Natural Environment ‘Research Counel (grant no. NE/L002485/1) ‘Aconledgenerts. We would like to thank Guy Cowlishaw for is invaluable support, guidance and academic insight. Many thanks to Paul Vasey and Helen Chatterice for their thoughts and sugnes- tions. We would also like to thank veo anonymous teviewers and the editor for thei helpful and eonstructve fellsck on this man script, This resoareh would aot hove hoon possible without the kine assistance of sollaagies who responded lo questionnaies oF provided addtional inlormation ena 8.210 The pena of human fers to aid seal ansible dees pas and preset Pt 2 sees sed dingo ae St Team Ine, 16, 8-93 610 136(8.76288) Hat Bl, Hofiek E, Bean 1987 Pseopustoy geil gouing in male a: preertin of sewualy tasted ets. Phys, bw. 41, 321-325, (10. 1016/031936487)90385.7) Ao Maite un CL 2010 The Tes webste 2 naw online source fr primate Dhlogny. ! Antopol.tses News and 19, 14-118. a 10.1002eran 20951) inde M, Ope. 206 Petcqulatry seal Seleaion infueces baculum evolution In primates and cares. Pr. RS. 8288, 2016736, (do 1028rsph2016.736) zou A, Haney Laon SG, Shot MU 198 Tess woh body went and being tem n nes, Nae 298, 55-7. (010883550) 205 atl gadiens of parse species ries in primates. ies. Dit. 19, 29-256, (doc to14,136-95162005.01603) Gueer Vy, Hocbers ME, Guégan F, Harney P 2 Flog dives the wos dition of Iunan oseases oS Bik 2 18, (i.1371 our pbioao2041) nn Ler SM, 205 The gba mama paste dab: an one rare fr hecias ‘ese ed in wd primes i Aahop ses News 14, 1-2 (kt 1004002047) 2B itch L168 Bnbatrungen zur Masuraton bel Siugetcren in Zologicen ate, elope Anziger 10, 298-312 competion and the function of male masutaton L9D0EZOT “067 § 205 yoy —Qdsu/jeUsNo(/fi0-BurysyqndAaposyekas fo3. 2 in rofuran pines. In Sul elon ant ‘prt conpetien in prime: ea pepe and dens, Sl Tepes in Pity o 3 eC oe), p 35-83, 1 Americ Sect cf Pimaogss. esi A, Moi, ia 8, Mae JA Gesan h Mn Hl A, Sen OF 2008 Thema tcl atty of uma sernallasma is zee ad dei fom fogmentatn te eno marl 18, 383-30. kA jenn 53813) Schulze lop U, ter F Herman ¢ edna ier lang Maik 6.220 Atta defers in al a Bw sere: a ptt ek to the mictbiome ard etucve sate? Teron 157, 335-0, E1006. thegeion 20200703) evo. 2003 Theis and outs of aga Lubin. Seva! eos Thy 18 59-51, 180748895031000785859) ann CL. 203 Sehavual dete aga seul wasted dese i tes Am Be. 6, 37-18, E10 ove 2032130) ann L200 The compare pach n elton cnt ad Bika. Cag, Univesity of chiage Pes Gevitaman A abi, Come Tyg Sk 201 feng of seal tive v % », 4 a Ifecions. n Sway tasted nfetiosan seal wonsted diseases eds C Gos, § K Tying), pp. 13-3. Mew York, NY: Springer. Paton JE Rudshl KE. 2073 Paste nein and ost gup sit a metal even. Parstloy 14, 803-873, (Ak10.1017 ons 1e201200225) Rome, yess 2017 A the wor primates Chateown, Rl: acpis Pres, ateo A, Dor , Mian Ch, San RL. 2004 Enviormetl hanes i poste coed vith ntaspetc diverge in maramal ad bis ofa, 23, 28-268 do:10.1 Ve 12572) Steers P eta 2017 Gaba! Marna Paste Dsabseveson 20, clgy 98, 476-1975. Felsenstein J 1985 Phylogeis and the ‘ompatve method. An. Na 125, 1-15, (0 ossr325) Drummond A, ambaut A. 207 EAST: ayesan volun ana by sampling tes, 2 Eek Bol 7, 1-8 (dito ri6)1471-2087218) (me 0, Fcsten A, Tamas, Peo T, Ft 5, cM, Pease 2018 Cpe coat ara af pages and nln npg eon 1, See hms AR pec adage. Coe Team, 200 fe lnguae and entearet fo tis! corp. Vea, has: Foundaion for Stats Computing. fou. Repojectorg 5 a sh eae A, Pagel M2017 Bayes compute padage fr ans of tat tin eran 3. See hpi eoton main Bassas tn Pagel M, Meade A 2005 Boyesn nas of conte erin of dscete characters by rereslejump Naa can Monte (Glo, ém. Na 167, 808-825. (1010867 saa444) Jordan FU, Gray RD, Gres I, Mace R. 2008 Nutro edence sane in Astonsan sacs, Prac R Soc. B 276, 1957-196, (oi 10. 1088p, 200908) 4s RE Rar AE 1995 Bayes facts. 1 Ae Stat, sw, 90, 775-85. doe ONOBUONEES, 199510876572) Asher PC. 2017 bs an 8 plage or Bean level models using San St Sot, 80, 1-28, (t10.1837 00101) Fredleton RP 2002 On the mis of rsa in colony: regen of resus ws, mille ragessin-1. Ann. Eo 71, 542-55. oo 16.145 2656200206183) Bande M,Ferguso-Gam H, llamo J, Thamsen , Sommer ¥. 2023 The ercion uf masurbatn ascot wth solr seletion an pathogen aidan n imate. Fish. (Uo: 10.60e4 stare sim) L9D0EZOT “067 § 205 yoy —Qdsu/jeUsNo(/fi0-BurysyqndAaposyekas B