Social and ecological drivers of reproductive seasonality in geladas

Social and ecological drivers of reproductive seasonality in geladas Abstract Many nonseasonally breeding mammals demonstrate some degree of synchrony in births, which is generally associated with ecological factors that mediate fecundity. However, disruptive social events, such as alpha male replacements, also have the potential to affect the timing of female reproduction. Here, we examined reproductive seasonality in a wild population of geladas (Theropithecus gelada) living at high altitudes in an afro-alpine ecosystem in Ethiopia. Using 9 years of demographic data (2006–2014), we determined that, while females gave birth year-round, a seasonal peak in births coincided with peak green grass availability (their staple food source). This post-rainy season “ecological peak” in births meant that estimated conceptions for these births occurred when temperatures were at their highest and mean female fecal glucocorticoid concentrations were at their lowest. In addition to this ecological birth peak, we also found a separate birth peak that occurred only for females in groups that had experienced a recent replacement of the dominant male (i.e., a takeover). Because new dominant males cause abortions in pregnant females and kill the infants of lactating females, takeovers effectively “reset” the reproductive cycles of females in the group. This “social birth peak” was distinct from the ecological peak and was associated with higher rates of cycling and conceptions overall and higher glucocorticoid levels immediately following a takeover as compared to females that did not experience a takeover. These data demonstrate that social factors (in this case, male takeovers) can contribute to population-level reproductive seasonality above and beyond group-level reproductive synchrony. INTRODUCTION The ways in which ecology interacts with reproductive timing can be characterized as a continuum, reflecting various species-specific energy-use strategies and patterns (Negus and Berger 1972; Drent and Daan 1980; Brockman and van Schaik 2005; Janson and Verdolin 2005). In the strictest sense, income breeders, at one end of the continuum, rely on external cues (e.g., photoperiod) to time conceptions so that vulnerable stages of the reproductive cycle, such as births or weaning periods, coincide with optimal ecological conditions. As a result, income breeders are typically strict seasonal breeders, with a discrete breeding season and birth season each year (though there may be some flexibility in birth timing in certain cases, e.g., Tecot 2010; Carnegie et al. 2011). Capital breeders, at the other end of the continuum, rely on internal cues of condition (e.g., energy balance, or the ratio between energy intake and energy expenditure: Valeggia and Ellison 2004), which allows for reproduction at any point in the year, as long as energetic requirements are met. Therefore, the timing of births for capital breeders depends on the seasonality and predictability of the environment: in a highly seasonal and predictable environment, capital breeders may demonstrate a strict birth season, much like income breeders do. On the other hand, in a less seasonal or less predictable environment, capital breeders may exhibit no seasonal patterns in births, or they may exhibit some degree of birth seasonality (where births cluster around the same time each year) or birth synchrony (where births cluster around certain times within a year: Heideman and Utzurrum 2003) following periods of energetic abundance. At a proximate level, birth seasonality is shaped by ecological factors that impact a female’s ability to “finance” a reproductive event across critical stages, from ovulation to conception and through parturition. Energy balance can impact a female’s ability to initiate (e.g., ovulation and conception: Bronson 1989; Wade and Schneider 1992; McCabe and Thompson 2013) or sustain (e.g., gestation and lactation: Roberts et al. 1985; Clutton-Brock et al. 1989; Wade and Schneider 1992; Beehner et al. 2006; Maestripieri and Georgiev 2016) a reproductive event. Birth seasonality (i.e., either a birth peak, when a high proportion of births are concentrated at a particular time of year, or a birth valley, when the proportion of births drops significantly at a particular time of year: Lancaster and Lee 1965; Janson and Verdolin 2005; Tecot 2010; Carnegie et al. 2011; Erb et al. 2012) can arise when maternal condition, specifically maternal energy balance, varies consistently across seasons. Energy balance is a function of energy intake and energy expenditure, and both are known to shape birth seasonality across many mammalian species (reviewed in: Di Bitetti and Janson 2000; Brockman and van Schaik 2005; Janson and Verdolin 2005). Because directly assessing energy intake can be methodologically challenging, many studies have successfully used food availability and/or rainfall as proxies for energetic condition in a number of studies (reviewed in Martínez-Mota et al. 2016). Both high food availability (e.g., in insectivorous mouse-eared bats, Myotis blythii: Arlettaz et al. 2001) and low food availability (e.g., in a survey of neotropical mammals: Dubost and Henry 2017) have been associated with either birth peaks or birth valleys, respectively. Energy expenditure can vary according to a number of different factors, from day range length to thermoregulation. In mammals, thermoregulatory demands due to both extreme heat and cold are known to hinder female reproduction (e.g., in owl monkeys, Aotus azarai: Fernandez-Duque et al. 2002; in black-and-white snub-nosed monkeys, Rhinopithecus bieti: Xiang and Sayers 2009; reviewed in Bronson 1985; 1989; Loudon and Racey 1987; Manning and Bronson 1990). For example, in captive conditions, the energetic demands of thermoregulation during extremely cold months were shown to impact reproduction even when energy intake was not restricted (e.g., in captive hamadryas baboons, Papio hamadryas: Polo and Colmenares 2016). Moreover, cold stress is known to be compounded by the hypoxic (i.e., low oxygen) conditions of high altitude where thermogenic capacity is even more constrained (e.g., Hayes 1989; Ward M, Milledge PJS, West JB. 1995; Chappell and Hammond 2004; Cheviron et al. 2013). One way to identify the harsh conditions that hamper female reproduction is to measure energy constraints (or their proxies), as outlined above. Another informative approach is to measure the internal physiological state of the organisms living through these adverse conditions. In particular, glucocorticoids (and glucocorticoid metabolites, or GCMs) are steroid hormones that lend themselves to this endeavor: first, GCMs rise in response to the energetic demands that accompany ecological and/or social challenges, such as food scarcity (e.g., Pride 2005; Gesquiere et al. 2008; Foerster et al. 2012), extreme temperatures (e.g., Weingrill et al. 2004, Beehner and McCann 2008; reviewed in Jessop et al. 2016), or the threat of infanticide from an incoming male (e.g., Beehner et al. 2005). Second, GCM concentrations in fecal, urine, and hair samples have been shown to increase with an individual’s recent exposure to ecological and/or social challenges (e.g., Fardi et al. 2017; reviewed in Dantzer et al. 2014; Beehner and Bergman 2017). Activation of the hypothalamic-pituitary adrenal axis (HPA-axis), which results in an increase in GCM secretion, has also been repeatedly associated with reproductive suppression (via the inhibition of pulsatile leutenizing hormone: Wasser 1996; Landys et al. 2006; Breen et al. 2007). Therefore, GCM profiles used in conjunction with measurements of challenging conditions (ecological or social) can help identify circumstances where reproductive constraints might be greatest. Group-living animals must contend with challenging social environments in addition to any harsh climatological conditions; correspondingly, the social environment can also alter the timing of female reproduction. For example, reproductive synchrony among females (where births cluster together within a year) can arise in response to social as well as ecological forces (reviewed in Ims 1990). One salient social threat to female reproduction known to promote reproductive synchrony is infanticide by males (reviewed in Agrell et al. 1998). Infanticide occurs when nonsire males kill the dependent offspring of females to expedite reproductive cycling and mating receptivity in the mother (Sugiyama 1965; Hrdy 1974; reviewed in: Hrdy 1979; Fedigan 2003; Palombit 2015). Perhaps in response to the threat of infanticide, lactating females often prematurely wean their dependent offspring (i.e., accelerated weaning: e.g., vervet monkeys, Chlorocebus pygerythrus: Fairbanks and McGuire 1987; siamangs, Symphalangus syndactylus: Morino and Borries 2017; reviewed in Smuts and Smuts 1993). Similarly, nonsire males can cause the death of a fetus in utero directly (i.e., sexually-selected feticide, Zipple et al. 2017) or indirectly (i.e., the Bruce effect: Bruce 1959; e.g., prairie voles, Microtus ochrogaster: Fraser-Smith 1975; wild horses, Equus caballus: Berger 1983; geladas, T. gelada: Roberts et al. 2012). Therefore, the arrival of nonsire males represents a challenging social environment that is known to “reset” the reproductive cycles of females. In social species, such as primates, the arrival of a nonsire male poses an infanticidal threat primarily when he takes over the dominant male position of a group (i.e., a takeover: Teichroeb and Jack 2017). Takeovers impact all females in a group simultaneously—effectively synchronizing their estrous cycles in the months that follow (e.g., Packer and Pusey 1983; Colmenares and Gomendio 1988; Ims 1990) and altering the timing of subsequent births. As a result, male takeovers have the potential to alter reproductive timing to such an extent that ecological patterns of reproductive seasonality is disrupted, which may result in fitness costs for affected females. In some species, females may need to delay conceiving until ecological conditions improve, while in others, females may conceive immediately following the death of their infant, resulting in births that fall outside any seasonal birth peak. If such “mis-timed” births occur during periods of negative energy balance, then it stands to reason that females will suffer downstream costs (e.g., neonatal loss, slow infant development and/or a prolonged interbirth interval) over and above the costs of infanticide alone. For example, in recent decades many species are experiencing rapid changes in their ecology due to climate change; these changes are causing slight shifts in birth timing that can result in mis-timed and, as a result, less successful births (e.g., in cattle, Bos taurus L.: Burthe et al. 2011; reviewed in Bronson 2009; Campos et al. 2017). Moreover, if male takeovers are seasonal (e.g., in white-faced capuchins, Cebus capucinus: Schoof and Jack 2013; reviewed in Teichroeb and Jack 2017), then takeovers themselves may produce a distinct “social birth peak” in addition to (or instead of) the usual “ecological birth peak.” Here, we examined the impact of male takeovers on reproductive seasonality in a population of wild primates, the gelada (T. gelada), living in the Simien Mountains National Park, Ethiopia. Geladas are an ideal species for this inquiry for a number of reasons. First, the social structure of geladas allows us to study dozens of groups, or “reproductive units,” at once. Reproductive units (hereafter, “units”) comprise the core groups of gelada multi-level society, and are composed of a dominant leader male, 1–12 adult females and their offspring, and possibly one or more subordinate follower males (Dunbar 1980; Snyder-Mackler et al. 2012b). Bachelor males form peripheral “all-male groups,” and challenge dominant leader males for reproductive control of the females within a unit (Dunbar and Dunbar 1974; Dunbar 1984). Second, male takeovers (i.e., when a bachelor male defeats a dominant leader male), are a frequent occurrence in this population (0.32 takeovers/unit/year, Beehner and Bergman 2008) and have severe consequences on female reproduction. Following takeovers, incoming males are known to commit infanticide (killing up to half of unit infants: Beehner and Bergman 2008), they trigger pregnancy termination (via the Bruce effect in 80% of females: Roberts et al. 2012), and they may cause lactating females to begin cycling sooner than they would otherwise (Dunbar 1980). Moreover, because male takeovers in this population follow a seasonal pattern themselves (with a peak between February to April each year, Pappano and Beehner 2014), takeovers have the potential to not only disrupt reproductive patterns for females that experience a takeover but to also produce their own seasonal peak in reproduction for these same females. Third, geladas represent an opportunity to examine the separate ecological effects of rainfall and temperature (as proxies for green grass availability and thermoregulatory constraints, respectively) on reproductive timing. Availability of green montane grass, the staple food for geladas, fluctuates throughout the year and peaks following the rainy season (Jarvey et al., In press). However, this optimal period of green grass availability also coincides with the least optimal temperatures for this region, when temperatures routinely fall below freezing (Iwamoto and Dunbar 1983; Ohsawa and Dunbar 1984). Previous research has demonstrated that male geladas in this same population exhibited the highest fecal GCMs during these cold-wet months (Beehner and McCann 2008), however, seasonal patterns of GCMs in females have never been examined. Females geladas (like males) should also be sensitive to thermoregulatory constraints, but females (unlike males) are expected to be sensitive to fluctuations in green grass availability due to the energetic demands of reproduction (Crook and Gartlan 1966; reviewed in Schülke and Ostner 2012). Indeed, for this same population of geladas, Dunbar (1980) identified a seasonal pattern to births that included not one, but 2 distinct birth peaks—one during the dry season and another during the wet season. He suggested that the 2 peaks reflected a compromise between the constraints of food and temperature: females were more likely to conceive when food availability was high, but also avoided giving birth to vulnerable infants during the cold-wet season. Our first aim in this study is to confirm the nature of reproductive seasonality in geladas with a larger dataset across nearly a decade (9 years) in conjunction with seasonality in rainfall and temperature. Our second aim is to use fecal hormone profiles (GCMs) to test 2 separate predictions about seasonal causes of metabolic stress. If green grass availability represents a significant metabolic constraint for females, then we expect to see the highest female GCMs during the dry season, when green grass is scarce. Alternatively, or additionally, if thermoregulatory pressures represent a significant metabolic constraint for females, then we expect to see the highest GCMs during the late rainy season, when temperatures are lowest. We predicted that, as capital breeders, the onset of female gelada reproduction would coincide with periods of low metabolic demands as indicated by low GCM profiles overall. Our third aim is to assess the extent to which male takeovers, with their known effects on female reproduction, explain seasonal variation in birth patterns. We predicted that male takeovers would be associated with higher GCM profiles and within-unit reproductive synchrony for affected females. Moreover, due to the seasonality of takeovers, we predicted that takeovers would produce some degree of across-unit reproductive seasonality. Finally, to consider the costs of giving birth outside of the ecological birth peak, we compared infant survival for births within versus outside of the ecological birth peak. METHODS Study site and subjects The data for this study were collected between 2006 and 2014 from a population of wild geladas living in the Simien Mountains National Park, in northern Ethiopia (13°13.5′ N latitude). The Simien Mountains Gelada Research Project (SMGRP – formerly called the University of Michigan Gelada Research Project) has collected behavioral, demographic, genetic, and hormonal data from individuals since January 2006. All gelada subjects are habituated to human observers on foot and are individually recognizable. Here, we used longitudinal data from 167 adult females comprising 25 reproductive units. All adult females had known or estimated birth dates from which we calculated age (mean age = 11.4 years; range = 4.3–27.0 years). Most estimated birth dates (N = 51) were calculated by subtracting the mean female age at major life-history milestones from known dates for each milestone (e.g., maturation, first birth: Roberts et al. 2017). For a subset of older females (N = 60), date of birth was estimated based on age of oldest offspring, or number of known offspring. Reproductive events We recorded all known births (N = 341; “births dataset”) and then assigned an estimated conception date for each birth based on the mean gestation length for this population (183 days; Roberts et al. 2017; “conceptions dataset”). For many of these births, the day of birth was known within 7 days (n = 291), but in cases where it was not, the birth date was estimated for the missed observation days as the mid-point of the period from which we last saw the mother until the time we saw the mother again with the new infant (n = 50 births; mean range = 37 ± 5.3 SE days; see Beehner and Bergman 2008 for more detail on assigning birth dates). The return to cycling for postpartum females was assigned as the first day we observed sexual swellings (i.e., swollen, bead-like vesicles surrounding a patch of exposed skin on the chest and neck, Roberts et al. 2017; “return to cycling” dataset) following the period of postpartum amenorrhea. The return to cycling dataset was independently observed from the conceptions and births datasets; and not all conceptions and births are included in the return to cycling dataset (e.g., females conceiving for the first time, females that were never observed to resume postpartum sexual swellings). By contrast, the conceptions dataset is not independent from the births dataset (i.e., the conception dates are derived from the birthdates). Nevertheless, we analyzed these 2 datasets separately to understand how the climatological data that correspond to each event contribute (or not) to the timing of female reproduction. In this way, we were able to separately examine the weather surrounding conceptions from the weather surrounding births. We have no cases where we included a conception that did not result in a live birth (since, for this dataset, live birth was how we estimated the date of conception). With respect to examining the costs of birth seasonality, we selected infant survival to 2 years as our fitness measure because the mean age at weaning in this population is approximately 1.5 years (Roberts et al. 2017) and thus the 2-year mark is sure to include the majority of weaned infants in our analyses. We opted against using additional measures that overly restricted our dataset (e.g., interbirth-intervals) or did not adequately account for infant losses prior to weaning (e.g., infant survival prior to 2 years). Weather data As part of our long-term climatological monitoring, we recorded daily cumulative rainfall and maximum and minimum temperature. Seasonal patterns of green grass availability in this area are known to be positively correlated with rainfall from the previous 90 days (Jarvey et al., In press). Therefore, because we did not directly collect food abundance data (or intake rates) across the entire study period, we used the total rainfall over the 90 days prior to each reproductive event (or hormone sample) as our proxy for green grass availability (e.g., Hill et al. 2000; McFarland et al. 2014). The physiological effects of temperature on reproduction are more direct than rainfall (via changes in core body temperatures). Therefore, as our proxy for thermoregulatory constraints, we calculated the mean maximum and minimum daily temperatures for the 30 days preceding each reproductive event or hormone sample (e.g., Dunbar et al. 2002). Takeovers We recorded the dates of all observed male takeovers (n = 72) of known reproductive units. Most takeovers were recorded within days of occurrence (n = 62; range = 0–7 days). For takeovers that were not directly observed (n = 18), we were able to assign the day of takeover to within a mean of 30.7 ± 5.3 SE days of occurrence. These takeovers were assigned the mid-point of the missing observation period. Because we were primarily concerned with how takeovers might alter the timing of reproductive events (e.g., due to infanticide, the Bruce effect, or accelerated weaning: Beehner and Bergman 2008; Roberts et al. 2017), we labeled any return to cycling and conception dates that occurred during the 3 months following a takeover as takeover return to cycling/conception dates, and those that did not follow a takeover we labeled nontakeover return to cycling/conception dates. The 3-month window was chosen based on the observation that most known or suspected infanticides occur during this 3-month window (in rare cases, infanticide may occur up to 9 months after a takeover, but we wanted to reflect the immediate impact of takeovers on female reproduction in the seasonality analysis: Beehner and Bergman 2008). Similarly, we labeled all births that occurred between 6–9 months following a takeover as takeover births (this time period reflects all births that resulted from conceptions by the new dominant male during his first 3 months as leader male, plus the period of gestation, 183 days, or just over 6 months; Roberts et al. 2017). Births that did not occur between 6–9 months after a takeover were labeled nontakeover births. Hormone collection and analysis We collected fecal samples from 148 known adult females between 2006 and 2014 (N =3,841 hormone samples; mean = 26 samples per female; range: 1–150 samples per female). Fecal samples were collected using noninvasive methods developed by the SMGRP for hormone extraction and preservation under field conditions (Beehner and Whitten 2004; Beehner and McCann 2008). In brief, we mixed the full fecal samples prior to placing a small aliquot in 3 ml of a methanol:acetone solution (4:1). Samples were vortexed and later filtered and extracted using a solid-phase cartridge. All samples were washed with 2.0 ml of 0.1% sodium azide (NaN3) solution, placed in a sterile Whirl-pak bag with a silica desiccant, and stored frozen until shipment to J. Beehner’s endocrine laboratory at the University of Michigan for radioimmunoassay (RIA). Dry fecal weights from all samples were obtained to the nearest ± 0.0001 g, and hormones values were calculated as ng/g dry feces. At the University of Michigan, all samples were assayed for glucocorticoid metabolites (GCMs) using reagents from the ImmuChemTM double antibody corticosterone 125I RIA kit (MP Biomedicals, LLC, Orangeburg, NY). This antibody has been validated both analytically and biologically for use in geladas (Beehner and McCann 2008). The primary antibody in this kit cross-reacts 100% with corticosterone, 0.34% with desoxycorticosterone, 0.1% with testosterone, 0.05% with cortisol, 0.03% with aldosterone, and 0.02% with progesterone. We ran all standards, controls, and samples in duplicate. We used a low (~20% binding) a mid- (~50% binding) and a high (~80% binding) fecal pool control in all assays. The respective interassay coefficients of variation (CVs) were: low fecal pool: 20.69%; mid fecal pool: 21.43%; high fecal pool: 20.93%; high kit control: 24.03%; and low kit control: 18.85% (N = 177 assays). Our intra-assay CV for a high and low fecal pool was 6.3% (N = 10 assays) and 8.7% (N = 10 assays), respectively, with a high kit control CV of 6.3% (N = 12 assays) and a low kit control of 4.1% (N = 12 assays). Data analyses We conducted 5 sets of models (described in detail below) that correspond to each of the 5 outcome variables (GCMs, return to cycling, conceptions, births, infant deaths). Model averaging For each of the first 4 sets of models (GCMs, return to cycling, conceptions, and births), we took an information theoretic approach based on Akaike’s Information Criterion (using AIC for the GCMs model; AICc, corrected for small sample sizes, for the return to cycling, conceptions, and births models; Anderson and Burnham 2002) to mean all candidate models for each outcome variable (Table 1; Johnson and Omland 2004). Candidate models represent all combinations of the predictor variables of interest and their interactions (mean maximum temperature, mean minimum temperature, cumulative rain, and takeover, Table 1. Model averaging was done using the Mu-MIn package (version 1.15.6: Barton 2016) in R (R Core Team 2016: Version 3.3.2) to produce estimates of predictors within an averaged model. We considered predictors to have a meaningful effect on the outcome variable if the 95% confidence intervals of the averaged effect size did not overlap zero (i.e., “reliable” predictors: Table 3, though see Supplementary Table S1 for full averaging results). Glucocorticoid seasonality To assess the effects of ecological and social seasonality on GCMs, we first log-transformed GCM values to approximate a normal distribution, and then modeled logGCMs as a function of the following candidate predictors (including interaction terms; see Table 1 for descriptions): 1) mean maximum temperature, 2) mean minimum temperature, 3) cumulative rainfall, and 4) takeover (yes/no). We controlled for the repeated measures of individual identity, unit, and year as random effects, as well as the known effects of age and reproductive state on GCMs (Beehner and Bergman 2017) by including an interaction term between reproductive state (pregnant, cycling, or lactating) and age (both as a linear and as a quadratic term; see Supplementary Figure S1 for full averaged model results). We constructed 41 candidate linear mixed-effects models (including an intercept model and a model controlling for age and reproductive state alone) with the lme4 package in R (version 1.1–12: Bates et al. 2015). The candidate models represent all combinations of predictor variables and their interactions (mean maximum temperature, mean minimum temperature, cumulative rain, and takeover), and the model fits were compared using AIC and Akaike model weights (Anderson and Burnham 2002). Initiating reproduction To assess how ecological (i.e., rainfall, temperature) and social (i.e., male takeover) factors were associated with the initiation of reproduction, we considered 2 different events that could indicate the start of a reproductive event for gelada females: 1) the resumption of cycling from postpartum amenorrhea (based on observational data on each female), and 2) the date of conception for each birth (estimated backwards from each observed birth). We conducted 2 separate analyses using binomial general linear mixed models (GLMMs). We coded the reproductive events as binomial variables by month, where females either returned to cycling or not, or conceived or not. Because we were interested in population-level patterns of reproductive seasonality (since populations, and not individuals, exhibit birth peaks), we analyzed reproductive seasonality across the entire population, not at the level of the unit. Therefore, the dependent binomial variable for the first model was Return to Cycling (or the number of postpartum females in the population that had resumed cycling in a given month out of the total females in the study population that month). The dependent binomial variable for the second model was Conceptions (or the number of females in the population that had conceived in a given month out of the total females in the study population that month). We also coded these reproductive events according to whether or not the event occurred within the context of a takeover. Using conceptions as an example, in a given month we calculated how many of all of our known females conceived in that month out of the total number of known females. This variable was calculated separately for females that did not experience a takeover in the previous 3 months (nontakeover conceptions) and those that did (takeover conceptions). Therefore, each month of the study is featured twice in the dependent variable column, once for reproductive events following takeovers and once for reproductive events not following takeovers. For each outcome variable (the number of Return to Cycling and the number of Conception events), we created 40 models (including an intercept model) predicting the number of reproductive events for each month based on a set of candidate predictor variables and their interactions (Table 1). Specifically, we considered the following predictors (as well as all interactions between them): 1) mean maximum daily temperature (across the previous 30 days), 2) mean minimum daily temperatures (across the previous 30 days), 3) cumulative rainfall (across the previous 90 days), and 4) the categorical predictor of whether or not a takeover had occurred prior (yes or no). Because each month was represented twice in our dataset, all models also included month and year as random effects to control for any potential monthly differences that were unrelated to the predictor variables. Table 1 Description of outcome variables and predictors used in model selection (lme4) Outcome variable Main effects Interaction effects Random effects Number of candidate models A. Log glucocorticoid metabolites: Linear Mixed-Effects Models logGCMs Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Individual ID, Unit, Year 41 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 30 days B. Return to cycling: Binomial General Linear Mixed Models Total number of monthly postpartum return to cycling out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days C. Conceptions: Binomial General Linear Mixed Models Total number of monthly conceptions out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days D. Births: Binomial General Linear Mixed Models Total number of monthly births out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 270 days Outcome variable Main effects Interaction effects Random effects Number of candidate models A. Log glucocorticoid metabolites: Linear Mixed-Effects Models logGCMs Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Individual ID, Unit, Year 41 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 30 days B. Return to cycling: Binomial General Linear Mixed Models Total number of monthly postpartum return to cycling out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days C. Conceptions: Binomial General Linear Mixed Models Total number of monthly conceptions out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days D. Births: Binomial General Linear Mixed Models Total number of monthly births out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 270 days View Large Table 1 Description of outcome variables and predictors used in model selection (lme4) Outcome variable Main effects Interaction effects Random effects Number of candidate models A. Log glucocorticoid metabolites: Linear Mixed-Effects Models logGCMs Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Individual ID, Unit, Year 41 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 30 days B. Return to cycling: Binomial General Linear Mixed Models Total number of monthly postpartum return to cycling out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days C. Conceptions: Binomial General Linear Mixed Models Total number of monthly conceptions out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days D. Births: Binomial General Linear Mixed Models Total number of monthly births out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 270 days Outcome variable Main effects Interaction effects Random effects Number of candidate models A. Log glucocorticoid metabolites: Linear Mixed-Effects Models logGCMs Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Individual ID, Unit, Year 41 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 30 days B. Return to cycling: Binomial General Linear Mixed Models Total number of monthly postpartum return to cycling out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days C. Conceptions: Binomial General Linear Mixed Models Total number of monthly conceptions out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days D. Births: Binomial General Linear Mixed Models Total number of monthly births out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 270 days View Large Birth seasonality To identify fluctuations in birth rates across the year due to either ecological or social factors (e.g., birth peaks and/or birth valleys), we first characterized the degree of seasonality in births qualitatively. We used cut-offs to distinguish between moderate seasonality (where 33–67% of births occur in a 3-month period), strict seasonality (where more than 67% of births cluster in a 3-month period), or no seasonality (where births occur all year with no clear peak; see van Schaik et al. 1999). We also characterized the degree to which these patterns changed (i.e., became more or less seasonal, or shifted in timing) when we considered takeover versus nontakeover births separately. Second, we constructed a third set of binomial GLMMs (with the lme4 package version 1.1–12: Bates et al. 2015) to assess when females were most likely to give birth (Table 1). For the births model, the dependent binomial variable was the number of females in the population that gave birth in a given month out of the total females in the study population that month. As with the cycling and conceptions models, we calculated this outcome variable as a function of whether or not each female had experienced a takeover, but rather than a 3-month window, this time we considered the 6–9 months prior to the birth (to identify births resulting from conceptions in the 3 months following a takeover plus the ~6 months of gestation). Again, we created 40 models (including an intercept model) predicting the number of births each month based on the same set of candidate predictor variables and their interactions used in the Return to Cycling and Conception models, reflecting hypotheses about which ecological and social variables influenced birth timing (Table 1). Specifically, we considered the following predictors (as well as all interactions between them): 1) mean maximum daily temperature (across the previous 30 days), 2) mean minimum daily temperatures (across the previous 30 days), 3) cumulative rainfall (across the previous 90 days), and 4) the categorical predictor of whether or not a takeover had occurred prior (yes or no). Again, because each month was represented twice in our dataset (once for births following takeovers, and once for births that did not follow takeovers), all models also included month and year as random effects to control for any potential monthly differences that were unrelated to the predictor variables. Costs to birth timing To assess whether there was a cost to giving birth at a certain time of the year, we considered the effect of birth timing on survival to 2 years of age (extending beyond the mean age of weaning in this population, 1.5 years of age, to be sure to include all weaned infants; Roberts et al. 2017). Specifically, because we identified a birth peak (see Birth seasonality in Results), we were interested in the potential costs of giving birth outside of this peak (which we defined as a 3-month period where 33–67% of births occurred: van Schaik et al. 1999). We constructed a series of binomial GLMMs to assess whether being born during the birth peak predicted infant survival to 2 years of age (yes/no). For each model, our predictor was whether the infant had been born “in-peak” or “off-peak” according to the seasonality analysis described above. First, we considered births and deaths for all infants until they reached 2 years of age (births = 306; deaths = 69; note that this dataset is slightly reduced from the one used in the seasonality analysis. To assess survival to 2 years of age we could only include births prior to 2014). Second, to focus solely on infant deaths due to “ecological reasons” we removed all suspected infanticide deaths (n = 33) and constructed another binomial GLMM with births (n = 273) and noninfanticide deaths (n = 36). Third, because takeovers force females to shift their reproductive cycles (see Birth seasonality in Results), we conducted a third binomial GLMM removing all births following takeovers to determine whether there was a cost to nonpeak births outside of the potential influence of takeovers (nontakeover births: n = 195; noninfanticide deaths: n = 26). For each model, we controlled for the repeated effects of birth year and the identity of the mother. RESULTS Ecological and social seasonality The climate in the Simien Mountains National Park can be broadly divided into 3 distinct seasons: a cold-dry season, a hot-dry season, and a cold-wet season (Figure 1). The cold-dry season typically occurred between October to January and featured the lowest minimum temperatures and very little rainfall (mean daily minimum temperature = 7.02 °C ± 1.39 SD; mean daily maximum temperature = 16.46 °C ± 1.64 SD; mean daily precipitation = 1.25 mm ± 4.61 SD). The hot-dry season typically occurred between February to May, and featured the warmest temperatures seen throughout the year with low (but variable) levels of rainfall (daily minimum temperature = 9.03 °C ± 1.39 SD; mean daily maximum temperature = 20.59 °C ± 2.19 SD; mean daily precipitation = 2.21 mm ± 6.64 SD). Finally, the cold-wet season typically occurred between June to September and featured the lowest maximum temperatures with the highest levels of daily precipitation (mean daily minimum temperature = 8.34 °C ± 1.38 SD; mean daily maximum temperature = 15.58 °C ± 2.42 SD; mean daily precipitation = 13.03 mm ± 17.35 SD). Typically, peak rainfall occurred between June to August, which corresponds with a peak in green grass availability between October to November (Figure 2a). Figure 1 View largeDownload slide Seasonality of ecological (temperature and rainfall) and social (takeovers) predictors. Mean maximum and minimum daily temperatures by month ± 95% confidence intervals (maximum = downward triangles; minimum = upward triangles; solid line); mean total rainfall by month (grey squares; dashed line) ± 95% confidence intervals; total number of takeovers observed by month (white bars). Background colors indicate season: orange = hot-dry season; light blue = cold-wet season; pale yellow = cold-dry season. Figure 1 View largeDownload slide Seasonality of ecological (temperature and rainfall) and social (takeovers) predictors. Mean maximum and minimum daily temperatures by month ± 95% confidence intervals (maximum = downward triangles; minimum = upward triangles; solid line); mean total rainfall by month (grey squares; dashed line) ± 95% confidence intervals; total number of takeovers observed by month (white bars). Background colors indicate season: orange = hot-dry season; light blue = cold-wet season; pale yellow = cold-dry season. Figure 2 View largeDownload slide (a) Cumulative seasonality predictors in relation to peak green grass availability, indicated by the shaded region (July to November). All models included the following variables as predictors: mean maximum (downward triangles; solid line) and minimum (upward triangles; solid line) daily temperature across the previous 30 days as a proxy for thermoregulatory requirements and cumulative rainfall (grey squares; dashed line) across the previous 90 days (as a proxy for green grass availability). (b) Seasonal patterns of mean residual log-transformed glucocorticoid metabolites (logGCMs) ± 95% confidence intervals (after controlling for the effects of reproductive state, age, and experiencing a takeover 30 days prior to sample collection). Warm and cold months were determined by taking the average of the mean maximum and mean minimum daily temperatures over the previous 30 days for each month (see Figure 2a; note that these categories were used for visualization purposes only). Background shades indicate the 4 warmest months (March to June) and the 4 coldest months (August to September; December to January). (c) Mean residual logGCMs ± standard error (after controlling for the effects of reproductive state, age, and temperature) for females that did not experience a takeover in the previous 30 days (gray box) vs. females that did (white box). (d) Mean residual logGCMs ± standard error (after controlling for the effects of reproductive state, age, and takeover) for females during the 4 warmest months (March to June; shaded box labeled "HOT" in Figure 2b) and during the 4 coldest months (August to September and December to January; shaded boxes labeled "COLD" in Figure 2b). For both (c) and (d), mean = solid line; standard error = box outline; 95% confidence intervals = whiskers. Figure 2 View largeDownload slide (a) Cumulative seasonality predictors in relation to peak green grass availability, indicated by the shaded region (July to November). All models included the following variables as predictors: mean maximum (downward triangles; solid line) and minimum (upward triangles; solid line) daily temperature across the previous 30 days as a proxy for thermoregulatory requirements and cumulative rainfall (grey squares; dashed line) across the previous 90 days (as a proxy for green grass availability). (b) Seasonal patterns of mean residual log-transformed glucocorticoid metabolites (logGCMs) ± 95% confidence intervals (after controlling for the effects of reproductive state, age, and experiencing a takeover 30 days prior to sample collection). Warm and cold months were determined by taking the average of the mean maximum and mean minimum daily temperatures over the previous 30 days for each month (see Figure 2a; note that these categories were used for visualization purposes only). Background shades indicate the 4 warmest months (March to June) and the 4 coldest months (August to September; December to January). (c) Mean residual logGCMs ± standard error (after controlling for the effects of reproductive state, age, and temperature) for females that did not experience a takeover in the previous 30 days (gray box) vs. females that did (white box). (d) Mean residual logGCMs ± standard error (after controlling for the effects of reproductive state, age, and takeover) for females during the 4 warmest months (March to June; shaded box labeled "HOT" in Figure 2b) and during the 4 coldest months (August to September and December to January; shaded boxes labeled "COLD" in Figure 2b). For both (c) and (d), mean = solid line; standard error = box outline; 95% confidence intervals = whiskers. In total, we observed 72 takeovers during the study period, which demonstrated a moderately seasonal pattern (a subset of these data were already analyzed and reported in Pappano and Beehner 2014): 45.8% of all takeovers occurred during a 3-month period (February to March, n = 33, Figure 1), and the highest number of takeovers out of all months occurred in March (n = 16). Seasonality of glucocorticoid metabolites Log glucocorticoid metabolites (logGCMs) were highly seasonal: high temperatures predicted low logGCMs (Figure 2b,d). LogGCMs decreased during the hot-dry season, reaching a nadir from April to July, and increased in the colder months, peaking twice: September to October and December to January (Figure 2b). Mean maximum and minimum temperatures were included as predictors in the top 4 models, which together contributes 90% of model weight and performed substantially better than the intercept only null model (Table 2). In the averaged model, mean maximum temperature (estimate = −0.0378, SE = 0.119, z-value = 3.179; Table 3) and mean minimum temperature (estimate = −0.0368, SE = 0.008, z-value = 4.563; Table 3) were strong and reliable negative predictors of logGCMs (i.e., 95% confidence intervals of mean estimates did not overlap zero, Figure 3a). This effect size corresponds to a 1.4% or a 2.9% decrease in GCMs for every 1 °C increase in mean maximum or mean minimum temperature, respectively (see Figure 2d for a comparison between mean residual GCMs in hot versus cold months, categories which were used for visualization purposes only). Rainfall, on the other hand, was not a strong or reliable predictor of logGCMs in the averaged model (estimate = −0.0075, SE = 0.011, z-value = 0.660; Table 3), although it was included as a predictor in one of the top 4 models (Table 2). Together, these results suggest that females are cold-stressed, and that thermoregulation rather than green grass availability may limit female energy balance. Seasonality of postpartum return to cycling We identified a seasonal peak in when females returned to cycling following postpartum amenorrhea, which mirrored the seasonality of logGCMs: 51.7% of all nontakeover cycling events occurred within a 3-month period during the hot-dry season (76 out of 147 total, March to May; Figure 4a: Return to cycling). Mean maximum temperature was included as a predictor in the top model (which comprised 19.1% of the model weight, Table 2), and was highly reliable based on 95% confidence intervals in the overall averaged model (Figure 3b). Mean maximum temperature significantly predicted the number of postpartum returns to cycling observed (estimate = 0.641, SE = 0.202, z-value = 3.175; Table 3), corresponding to a 35.6% increase in the number of females that returned to cycling for every 1 °C increase in maximum temperature. Table 2 Akaike’s information criterion (AIC/AICc) model comparison results, showing the model components for the null model and the top models for each analysis (where ∆AIC or ∆AICc ≤ 6) A.Log glucocorticoid metabolites Degrees of freedom Log-likelihood ∆AIC Weight Maximum temperature, minimum temperature, takeover 16 −487.37 0.00 0.47 Maximum temperature, minimum temperature, rain, takeover 17 −486.85 0.96 0.29 Minimum temperature, takeover, maximum temperature × rain 18 −486.79 2.83 0.11 Maximum temperature, minimum temperature 15 −490.67 4.60 0.05 Intercept (null model) 5 −638.97 281.19 0.00 B.Return to cycling Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, takeover 4 −248.893 0.00 0.191 Maximum temperature, minimum temperature × takeover 6 −246.831 0.14 0.179 Maximum temperature × takeover 5 −248.573 1.48 0.091 Maximum temperature, rain × takeover 6 −247.708 1.89 0.074 Maximum temperature, rain, takeover 5 −248.834 2.00 0.070 Maximum temperature, minimum temperature, takeover 5 −248.883 2.10 0.067 Maximum temperature, rain, minimum temperature × takeover 7 −246.791 2.22 0.063 Takeover, maximum temperature × rain 6 −247.916 2.31 0.060 Rain, maximum temperature × takeover 6 −248.532 3.54 0.033 Minimum temperature, maximum temperature × takeover 6 −248.569 3.61 0.031 Takeover, maximum temperature × minimum temperature 6 −248.612 3.70 0.030 Maximum temperature, minimum temperature, rain × takeover 7 −247.708 4.06 0.025 Intercept (null model) 2 −280.993 60.04 0.000 C.Conceptions Degrees of freedom Log-likelihood ∆AICc Weight Rain × takeover 5 −282.827 0.00 0.470 Minimum temperature, rain × takeover 6 −282.680 1.85 0.186 Maximum temperature, rain × takeover 6 −282.827 2.14 0.161 Maximum temperature, minimum temperature, rain × takeover 7 −284.500 3.61 0.077 Intercept (null model) 2 −304.483 37.03 0.000 D.Births Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, minimum temperature, rain × takeover 7 −289.742 0.00 0.851 Minimum temperature, rain × takeover 6 −292.897 4.14 0.108 Intercept (null model) 2 −328.858 67.62 0.000 A.Log glucocorticoid metabolites Degrees of freedom Log-likelihood ∆AIC Weight Maximum temperature, minimum temperature, takeover 16 −487.37 0.00 0.47 Maximum temperature, minimum temperature, rain, takeover 17 −486.85 0.96 0.29 Minimum temperature, takeover, maximum temperature × rain 18 −486.79 2.83 0.11 Maximum temperature, minimum temperature 15 −490.67 4.60 0.05 Intercept (null model) 5 −638.97 281.19 0.00 B.Return to cycling Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, takeover 4 −248.893 0.00 0.191 Maximum temperature, minimum temperature × takeover 6 −246.831 0.14 0.179 Maximum temperature × takeover 5 −248.573 1.48 0.091 Maximum temperature, rain × takeover 6 −247.708 1.89 0.074 Maximum temperature, rain, takeover 5 −248.834 2.00 0.070 Maximum temperature, minimum temperature, takeover 5 −248.883 2.10 0.067 Maximum temperature, rain, minimum temperature × takeover 7 −246.791 2.22 0.063 Takeover, maximum temperature × rain 6 −247.916 2.31 0.060 Rain, maximum temperature × takeover 6 −248.532 3.54 0.033 Minimum temperature, maximum temperature × takeover 6 −248.569 3.61 0.031 Takeover, maximum temperature × minimum temperature 6 −248.612 3.70 0.030 Maximum temperature, minimum temperature, rain × takeover 7 −247.708 4.06 0.025 Intercept (null model) 2 −280.993 60.04 0.000 C.Conceptions Degrees of freedom Log-likelihood ∆AICc Weight Rain × takeover 5 −282.827 0.00 0.470 Minimum temperature, rain × takeover 6 −282.680 1.85 0.186 Maximum temperature, rain × takeover 6 −282.827 2.14 0.161 Maximum temperature, minimum temperature, rain × takeover 7 −284.500 3.61 0.077 Intercept (null model) 2 −304.483 37.03 0.000 D.Births Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, minimum temperature, rain × takeover 7 −289.742 0.00 0.851 Minimum temperature, rain × takeover 6 −292.897 4.14 0.108 Intercept (null model) 2 −328.858 67.62 0.000 View Large Table 2 Akaike’s information criterion (AIC/AICc) model comparison results, showing the model components for the null model and the top models for each analysis (where ∆AIC or ∆AICc ≤ 6) A.Log glucocorticoid metabolites Degrees of freedom Log-likelihood ∆AIC Weight Maximum temperature, minimum temperature, takeover 16 −487.37 0.00 0.47 Maximum temperature, minimum temperature, rain, takeover 17 −486.85 0.96 0.29 Minimum temperature, takeover, maximum temperature × rain 18 −486.79 2.83 0.11 Maximum temperature, minimum temperature 15 −490.67 4.60 0.05 Intercept (null model) 5 −638.97 281.19 0.00 B.Return to cycling Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, takeover 4 −248.893 0.00 0.191 Maximum temperature, minimum temperature × takeover 6 −246.831 0.14 0.179 Maximum temperature × takeover 5 −248.573 1.48 0.091 Maximum temperature, rain × takeover 6 −247.708 1.89 0.074 Maximum temperature, rain, takeover 5 −248.834 2.00 0.070 Maximum temperature, minimum temperature, takeover 5 −248.883 2.10 0.067 Maximum temperature, rain, minimum temperature × takeover 7 −246.791 2.22 0.063 Takeover, maximum temperature × rain 6 −247.916 2.31 0.060 Rain, maximum temperature × takeover 6 −248.532 3.54 0.033 Minimum temperature, maximum temperature × takeover 6 −248.569 3.61 0.031 Takeover, maximum temperature × minimum temperature 6 −248.612 3.70 0.030 Maximum temperature, minimum temperature, rain × takeover 7 −247.708 4.06 0.025 Intercept (null model) 2 −280.993 60.04 0.000 C.Conceptions Degrees of freedom Log-likelihood ∆AICc Weight Rain × takeover 5 −282.827 0.00 0.470 Minimum temperature, rain × takeover 6 −282.680 1.85 0.186 Maximum temperature, rain × takeover 6 −282.827 2.14 0.161 Maximum temperature, minimum temperature, rain × takeover 7 −284.500 3.61 0.077 Intercept (null model) 2 −304.483 37.03 0.000 D.Births Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, minimum temperature, rain × takeover 7 −289.742 0.00 0.851 Minimum temperature, rain × takeover 6 −292.897 4.14 0.108 Intercept (null model) 2 −328.858 67.62 0.000 A.Log glucocorticoid metabolites Degrees of freedom Log-likelihood ∆AIC Weight Maximum temperature, minimum temperature, takeover 16 −487.37 0.00 0.47 Maximum temperature, minimum temperature, rain, takeover 17 −486.85 0.96 0.29 Minimum temperature, takeover, maximum temperature × rain 18 −486.79 2.83 0.11 Maximum temperature, minimum temperature 15 −490.67 4.60 0.05 Intercept (null model) 5 −638.97 281.19 0.00 B.Return to cycling Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, takeover 4 −248.893 0.00 0.191 Maximum temperature, minimum temperature × takeover 6 −246.831 0.14 0.179 Maximum temperature × takeover 5 −248.573 1.48 0.091 Maximum temperature, rain × takeover 6 −247.708 1.89 0.074 Maximum temperature, rain, takeover 5 −248.834 2.00 0.070 Maximum temperature, minimum temperature, takeover 5 −248.883 2.10 0.067 Maximum temperature, rain, minimum temperature × takeover 7 −246.791 2.22 0.063 Takeover, maximum temperature × rain 6 −247.916 2.31 0.060 Rain, maximum temperature × takeover 6 −248.532 3.54 0.033 Minimum temperature, maximum temperature × takeover 6 −248.569 3.61 0.031 Takeover, maximum temperature × minimum temperature 6 −248.612 3.70 0.030 Maximum temperature, minimum temperature, rain × takeover 7 −247.708 4.06 0.025 Intercept (null model) 2 −280.993 60.04 0.000 C.Conceptions Degrees of freedom Log-likelihood ∆AICc Weight Rain × takeover 5 −282.827 0.00 0.470 Minimum temperature, rain × takeover 6 −282.680 1.85 0.186 Maximum temperature, rain × takeover 6 −282.827 2.14 0.161 Maximum temperature, minimum temperature, rain × takeover 7 −284.500 3.61 0.077 Intercept (null model) 2 −304.483 37.03 0.000 D.Births Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, minimum temperature, rain × takeover 7 −289.742 0.00 0.851 Minimum temperature, rain × takeover 6 −292.897 4.14 0.108 Intercept (null model) 2 −328.858 67.62 0.000 View Large Table 3 Model-averaging results for significant predictors of interest (full model-averaging results appear in Supplementary Table S1) Predictors Importance Number of models 2.5% CI 97.5% CI Estimate Adjusted standard error z value P-value A. Log glucocorticoid metabolites Maximum temperature 1.00 26 −0.0611 −0.0145 −0.0378 0.0119 3.179 **0.0015 Minimum temperature 1.00 26 −0.0526 −0.0210 −0.0368 0.0081 4.563 ***5.0 × 10−06 Takeover (Y) 0.93 26 0.0131 0.1066 0.0598 0.0238 2.509 **0.0121 B. Return to cycling Takeover (Y) 1.00 26 0.7490 1.5406 1.1450 0.2019 5.669 ***<2.0 × 10−16 Maximum temperature 0.99 26 0.2454 1.0371 0.6412 0.2020 3.175 **0.0015 C. Conceptions Rain 0.98 26 −0.6158 −0.1363 −0.3761 0.1223 3.074 **0.00211 Takeover (Y) 1.00 26 0.4935 1.0706 0.7821 0.1472 5.312 ***<1.0 × 10−7 Rain × takeover (Y) 0.89 4 0.0203 0.7936 0.4070 0.1973 2.063 *0.03911 D. Births Takeover (Y) 1.00 26 0.7171 1.3147 1.0160 0.1525 6.663 ***<2.0 × 10−16 Rain 1.00 26 0.2131 0.7366 0.4749 0.1336 3.556 ***0.0004 Rain × takeover (Y) 1.00 4 −0.9984 −0.3002 −0.6493 0.1781 3.646 ***0.0003 Minimum temperature 0.96 26 −0.5716 −0.0540 −0.3128 0.1320 2.369 *0.0178 Predictors Importance Number of models 2.5% CI 97.5% CI Estimate Adjusted standard error z value P-value A. Log glucocorticoid metabolites Maximum temperature 1.00 26 −0.0611 −0.0145 −0.0378 0.0119 3.179 **0.0015 Minimum temperature 1.00 26 −0.0526 −0.0210 −0.0368 0.0081 4.563 ***5.0 × 10−06 Takeover (Y) 0.93 26 0.0131 0.1066 0.0598 0.0238 2.509 **0.0121 B. Return to cycling Takeover (Y) 1.00 26 0.7490 1.5406 1.1450 0.2019 5.669 ***<2.0 × 10−16 Maximum temperature 0.99 26 0.2454 1.0371 0.6412 0.2020 3.175 **0.0015 C. Conceptions Rain 0.98 26 −0.6158 −0.1363 −0.3761 0.1223 3.074 **0.00211 Takeover (Y) 1.00 26 0.4935 1.0706 0.7821 0.1472 5.312 ***<1.0 × 10−7 Rain × takeover (Y) 0.89 4 0.0203 0.7936 0.4070 0.1973 2.063 *0.03911 D. Births Takeover (Y) 1.00 26 0.7171 1.3147 1.0160 0.1525 6.663 ***<2.0 × 10−16 Rain 1.00 26 0.2131 0.7366 0.4749 0.1336 3.556 ***0.0004 Rain × takeover (Y) 1.00 4 −0.9984 −0.3002 −0.6493 0.1781 3.646 ***0.0003 Minimum temperature 0.96 26 −0.5716 −0.0540 −0.3128 0.1320 2.369 *0.0178 View Large Table 3 Model-averaging results for significant predictors of interest (full model-averaging results appear in Supplementary Table S1) Predictors Importance Number of models 2.5% CI 97.5% CI Estimate Adjusted standard error z value P-value A. Log glucocorticoid metabolites Maximum temperature 1.00 26 −0.0611 −0.0145 −0.0378 0.0119 3.179 **0.0015 Minimum temperature 1.00 26 −0.0526 −0.0210 −0.0368 0.0081 4.563 ***5.0 × 10−06 Takeover (Y) 0.93 26 0.0131 0.1066 0.0598 0.0238 2.509 **0.0121 B. Return to cycling Takeover (Y) 1.00 26 0.7490 1.5406 1.1450 0.2019 5.669 ***<2.0 × 10−16 Maximum temperature 0.99 26 0.2454 1.0371 0.6412 0.2020 3.175 **0.0015 C. Conceptions Rain 0.98 26 −0.6158 −0.1363 −0.3761 0.1223 3.074 **0.00211 Takeover (Y) 1.00 26 0.4935 1.0706 0.7821 0.1472 5.312 ***<1.0 × 10−7 Rain × takeover (Y) 0.89 4 0.0203 0.7936 0.4070 0.1973 2.063 *0.03911 D. Births Takeover (Y) 1.00 26 0.7171 1.3147 1.0160 0.1525 6.663 ***<2.0 × 10−16 Rain 1.00 26 0.2131 0.7366 0.4749 0.1336 3.556 ***0.0004 Rain × takeover (Y) 1.00 4 −0.9984 −0.3002 −0.6493 0.1781 3.646 ***0.0003 Minimum temperature 0.96 26 −0.5716 −0.0540 −0.3128 0.1320 2.369 *0.0178 Predictors Importance Number of models 2.5% CI 97.5% CI Estimate Adjusted standard error z value P-value A. Log glucocorticoid metabolites Maximum temperature 1.00 26 −0.0611 −0.0145 −0.0378 0.0119 3.179 **0.0015 Minimum temperature 1.00 26 −0.0526 −0.0210 −0.0368 0.0081 4.563 ***5.0 × 10−06 Takeover (Y) 0.93 26 0.0131 0.1066 0.0598 0.0238 2.509 **0.0121 B. Return to cycling Takeover (Y) 1.00 26 0.7490 1.5406 1.1450 0.2019 5.669 ***<2.0 × 10−16 Maximum temperature 0.99 26 0.2454 1.0371 0.6412 0.2020 3.175 **0.0015 C. Conceptions Rain 0.98 26 −0.6158 −0.1363 −0.3761 0.1223 3.074 **0.00211 Takeover (Y) 1.00 26 0.4935 1.0706 0.7821 0.1472 5.312 ***<1.0 × 10−7 Rain × takeover (Y) 0.89 4 0.0203 0.7936 0.4070 0.1973 2.063 *0.03911 D. Births Takeover (Y) 1.00 26 0.7171 1.3147 1.0160 0.1525 6.663 ***<2.0 × 10−16 Rain 1.00 26 0.2131 0.7366 0.4749 0.1336 3.556 ***0.0004 Rain × takeover (Y) 1.00 4 −0.9984 −0.3002 −0.6493 0.1781 3.646 ***0.0003 Minimum temperature 0.96 26 −0.5716 −0.0540 −0.3128 0.1320 2.369 *0.0178 View Large Figure 3 View largeDownload slide Model-averaged coefficient estimates ± 95% confidence intervals for averaged models for each outcome variable: (a) log glucocorticoid metabolites (here showing only predictors of interest; each model also controlled for reproductive status and age-squared: see Supplementary Figure S1), (b) return to cycling, (c) conceptions, and (d) births. Coefficient estimates less than zero result in a lower than expected outcome while those greater than zero result in a higher than expected outcome. Note that these plots depict the direction and reliability of each estimate (i.e., predictors are considered reliable when the 95% confidence intervals do not cross zero), not the relative effect size. Predictors are displayed in descending order of relative importance assigned by the MuMIn comparison. Figure 3 View largeDownload slide Model-averaged coefficient estimates ± 95% confidence intervals for averaged models for each outcome variable: (a) log glucocorticoid metabolites (here showing only predictors of interest; each model also controlled for reproductive status and age-squared: see Supplementary Figure S1), (b) return to cycling, (c) conceptions, and (d) births. Coefficient estimates less than zero result in a lower than expected outcome while those greater than zero result in a higher than expected outcome. Note that these plots depict the direction and reliability of each estimate (i.e., predictors are considered reliable when the 95% confidence intervals do not cross zero), not the relative effect size. Predictors are displayed in descending order of relative importance assigned by the MuMIn comparison. Figure 4 View largeDownload slide Seasonal patterns of reproductive events. (a) No takeover events: postpartum return to cycling where a takeover did not occur within the previous 3 months; conceptions where a takeover did not occur within the previous 3 months; and births where a takeover did not occur within the previous 6–9 months. (b) Reproductive events following takeovers: post-partum return to cycling observed within 3 months of a takeover; conceptions where a takeover did occur within the previous 3 months; and births where a takeover did occur within the previous 6–9 months. In all panels: left axis = total observed (bars); right axis = mean rates ± standard error (i.e., the number of reproductive events per female per month; black squares). Background shades indicate season: hot-dry (February to May); cold-wet (June to September); cold-dry (October to January). Figure 4 View largeDownload slide Seasonal patterns of reproductive events. (a) No takeover events: postpartum return to cycling where a takeover did not occur within the previous 3 months; conceptions where a takeover did not occur within the previous 3 months; and births where a takeover did not occur within the previous 6–9 months. (b) Reproductive events following takeovers: post-partum return to cycling observed within 3 months of a takeover; conceptions where a takeover did occur within the previous 3 months; and births where a takeover did occur within the previous 6–9 months. In all panels: left axis = total observed (bars); right axis = mean rates ± standard error (i.e., the number of reproductive events per female per month; black squares). Background shades indicate season: hot-dry (February to May); cold-wet (June to September); cold-dry (October to January). Seasonality of conceptions The number of conceptions also exhibited a moderate peak at the beginning of the hot-dry season, with 35.9% of nontakeover conceptions occurring between February and April (89 out of 248, Figure 4a: Conceptions). Rainfall was included in the top 3 models based on AICc comparison, which all performed better than the intercept only null model and together contributes over 90% of the model weight (ΔAICc = 37.03, Table 2). In our averaged model, rainfall (estimate = −0.3761, SE = 0.1223, z-value = 3.074; Figure 3c; Table 3) reliably predicted fewer conceptions, and every additional 1 mm of rain corresponded to a 0.07% decrease in the number of conceptions. Effect of takeovers on reproductive seasonality Females that experienced a takeover demonstrated 6.2% higher GCMs than females that did not experience a takeover (estimate = 0.0598, SE = 0.0238, z-value = 2.509, Figure 2c; Table 3). Takeovers also resulted in significantly different seasonal patterns for both postpartum return to cycling and conceptions (Figure 4b). After takeovers, the seasonal pattern of returns to cycling was more distinct than that observed outside of takeovers, with 57.6% of all takeover cycling events occurring within the same 3-month period (34 out of 59 total, March to May, Figure 4b: Return to cycling). However, females that experienced a takeover were also significantly more likely to return to cycling than females that did not experience a takeover, regardless of the time of year (estimate = 1.145, SE = 0.2019, z-value = 5.669; Table 3). This translated to a more than 200% increase in the number of females that returned to cycling post-takeover compared to the number of nontakeover females that returned to cycling. Females that experienced a takeover conceived later in the year than females that did not experience a takeover, and takeover conceptions demonstrated a clear peak at the start of the wet season, from June to August (52.8% of all takeover conceptions, 38 out of 72 total, Figure 4b: Conceptions). Takeovers also had a reliably positive effect on the number of conceptions that occurred in the following 3 months, regardless of the season (estimate = 0.7821, SE = 0.1472, z-value = 5.312, Figure 3c). Females that experienced a takeover were 118.6% more likely to conceive than females that did not experience a takeover. Furthermore, the interaction between rainfall and takeovers was highly reliable (estimate = 0.407, SE = 0.1973, z-value = 2.063, Figure 3c), reflecting a shift in conceptions by 4 months (from March to July, Figure 4a: Conceptions) if a takeover had occurred. In other words, if a female experienced a takeover, she was 50.2% more likely to conceive during the cold-wet season, when temperatures were at their lowest, than during the hot-dry season. Birth seasonality We found 2 distinct birth patterns depending on whether or not births followed a takeover. First, while female geladas gave birth throughout the year (mean monthly birth rate = 0.03 births per female ± 0.03 SD), 37.4% of all nontakeover births occurred between August and October (102 out of 273 total, mean birth rate = 0.05 births per female ± 0.03 SD, Figure 4a: Births). This nontakeover peak was tightly coupled with seasonal patterns in both rainfall and mean minimum temperature, and both were included as predictors in the top 2 birth models (which together represent over 90% of the model weight, Table 2). Rainfall was the strongest predictor of the number of births (estimate = 0.4739, SE = 0.1336, z-value = 3.556, Table 3; Figure 3d): an increase of 1mm of rainfall corresponded to a 0.1% increase in births. In contrast, mean minimum temperature was negatively associated with births (estimate = −0.3128, SE = 0.132, z-value = 2.369, Table 3; Figure 3d); an increase of 1 °C corresponded to a 35.7% decrease in births. Nontakeover births peaked at the end of the cold-wet season, which also corresponded with peak food availability (Figure 4a: Births). On the other hand, takeover births peaked later in the year: 54.4% of all takeover births occurred between December and February, during the cold-dry season (37 out of 68 total, mean birth rate = 0.14 births per female ± 0.09 SD, Figure 4b: Births). In our averaged model, takeovers positively predicted births (estimate = 1.016, SE = 0.1525, z-value = 6.663, Table 3), and experiencing a takeover resulted in 176.2% more births 6–9 months later than observed for those females that did not experience a takeover. Because takeovers themselves followed a seasonal pattern, we observed a “birth valley” between April and July, immediately following the takeover season, when only 7.4% of takeover births occurred (5 out of 68, Figure 4b: Births). However, birth rates recovered sharply, and during the takeover birth peak, birth rates were double those observed during the nontakeover birth peak (e.g., the mean takeover birth rate in January = 0.13 births per female ± 0.07 SD, while the mean nontakeover birth rate in September = 0.06 births per female ± 0.03 SD). This increase likely reflects the synchronizing effect that takeovers had on the reproductive timing of females within the same unit. Finally, the interaction between takeovers and rain was a reliable negative predictor of the number of births (estimate = −0.6493, SE = 0.1781, z-value = 3.646, Table 3; Figure 3d), reflecting the shift in the takeover birth peak to the cold-dry season, which is when we see minimal rainfall. In other words, the takeover birth peak is characterized by a 47.8% increase in births during the driest months of the year. Costs to birth timing We found no evidence that there were costs associated with giving birth outside the birth peak. Of the 306 births observed (from our reduced dataset of infants born prior to 2014), 69 of the infants subsequently died before reaching 2 years of age (rate = 0.225 infant deaths per birth). Approximately half of these deaths were attributed to infanticide and half were attributed to other factors (n = 33 possible infanticide deaths; n = 36 noninfanticide deaths; Supplementary Figure S2). Being born during the birth peak (August to October) versus outside of the birth peak (November to July) did not significantly predict whether an infant died before reaching 2 years of age (in-peak: estimate = −0.021, SE = 0.309, z-value = −0.069). When we removed infanticide deaths from the analysis, we still found no effect of being born in-peak on infant survival (in-peak: estimate = −0.405, SE = 0.470, z-value = −0.862). Finally, when we completely removed the potential influence of takeover (i.e., births that followed a takeover and deaths due to infanticide), we still found no effect of being born in-peak (in-peak: estimate = 0.021, SE = 0.966, z-value = 0.022). DISCUSSION Our results highlight the effect that male takeovers can have on the timing of female reproduction: in addition to the direct effects from infanticide (reviewed in Hrdy 1979; Fedigan 2003; Palombit 2015) and from the Bruce effect and/or accelerated weaning (reviewed in Smuts and Smuts 1993), we demonstrated that takeovers can disrupt reproductive seasonality within a population across years. Specifically, we identified 2 distinct seasonal birth patterns: one that we believe is shaped by energetic constraints related to thermoregulation (“ecological birth peak”) and one shaped by the timing of male takeovers (“social birth peak”). These results demonstrate the separate effects that ecological and social variables can have on seasonal patterns of female reproduction. Although female geladas are not seasonal breeders, births follow a moderately seasonal pattern (as defined by van Schaik et al. 1999), with 37.4% of all nontakeover births occurring at the end of the cold-rainy season (August to October). This ecological birth peak corresponds with a peak in conceptions that occurs in the middle of the hot-dry season (February to April). The hot-dry season is also when we observed the lowest concentrations of glucocorticoid metabolites and a peak in the number of females that returned to cycling. The seasonal variation in female glucocorticoid metabolites mirrors the pattern previously identified in cold-stressed males (Beehner and McCann 2008). These data indicate that, for some females in this population, thermoregulatory requirements due to cold temperatures may be a significant barrier to the onset of reproduction. Although food scarcity has been associated with poor energetic condition and reproductive performance in a number of species (reviewed in Di Bitetti and Janson 2000 and Brockman and van Schaik 2005; though see: Weingrill et al. 2004), we were unable to detect a relationship between the initiation of reproduction and the availability of green grass—the staple food source of geladas (Jarvey et al., In press). However, we acknowledge that green grass availability may be a poor proxy for energy intake in this population. Although the hot-dry season is when the availability of green grass is at its lowest, we also know that geladas readily switch to fallback foods during this time (e.g., underground storage organs and tubers: Jarvey et al., In press), which may provide sufficient energy to offset any additional costs associated with foraging for them (Hunter 2001). Furthermore, although females are able to give birth at other points in the year, the ecological birth peak aligns with the peak in green grass availability (ecological birth peak = August to October; peak green grass = July to November), suggesting that females that give birth at this time may be able to maintain better body condition throughout the energetically-costly period of lactation. Therefore, although our data suggest that the constraints of temperature are more significant than the constraints of green grass availability, it remains likely that food availability may still shape gelada reproductive patterns. Addressing this important question will require a cross-year assessment that includes periods of intense scarcity (e.g., in yellow baboons, Papio cynocephalus: Lea et al. 2015). By contrast, the birth pattern following takeovers was characterized by a decrease in the number of births during the months immediately following a takeover (e.g., a birth valley: Lancaster and Lee 1965), followed by a birth peak 6–9 months after the “takeover season.” This social birth peak differed from the ecological birth peak in 2 ways. First, it occurred later in the year than the ecological birth peak (in the cold-dry season instead of the cold-rainy season). Second, it resulted in a significantly higher birth rate than that recorded for the ecological birth peak, indicating that births following a takeover were more synchronized than births solely tracking fluctuations in temperature and/or rainfall. From an evolutionary perspective, reproductive synchrony among females is hypothesized to be a counterstrategy to male reproductive tactics such as coercion, male monopolization, or infanticide. Some of the benefits of synchronous mating may include, for example, a higher degree of female mate choice (Ostner et al. 2008; Roberts et al. 2014), higher rates of extrapair fertilizations (Stutchbury and Morton 1995), and a reduced risk of harassment and/or infanticide (Boness et al. 1995; Gilchrist 2006; Hodge et al. 2011; Riehl 2016). Although the exact mechanism causing reproductive synchrony may vary by species, the end result is thought to be increased female fitness. These potential fitness benefits do not explain the results we report here for the gelada system: geladas live in polygynous groups where female mate choice is limited (Snyder-Mackler et al. 2012a) and synchrony appears to result from, rather than prevent, the fitness costs associated with male takeovers (i.e., due to infanticide and/or the Bruce effect). At the proximate level, females show higher rates of postpartum resumptions of cycling only after experiencing a takeover, regardless of their reproductive state at the time of takeover (cycling, lactating, or pregnant) or the time of year. For most of these females, we were unable to distinguish “deceptive” signals of fertility (i.e., nonovulatory and/or nonconceptive swellings), a known female counterstrategy to infanticide (e.g., Zinner and Deschner 2000), from “true fertility.” Nevertheless, although we expect that some postpartum returns to cycling were indeed deceptive, male takeovers were also associated with high conception rates in the following months, suggesting that the majority of females truly returned to cycling after a takeover. Two additional lines of evidence suggest that male takeovers drive reproductive synchrony in geladas and not the reverse. First, experiencing a takeover (even during the ecological peak in conceptions) actually delayed the conception peak by 4–6 months for the majority of females in these units as compared to females unaffected by takeovers. Second, most of the females that return to cycling following a takeover were pregnant or lactating at the time of takeover. We have hormonal evidence that 80% of females terminate pregnancies after a takeover (Roberts et al. 2012); and we have statistical evidence that incoming males kill nearly half of the infants of lactating females (Beehner and Bergman 2008). Although we have not examined reproductive changes across individual females, the overall pattern indicates that female receptivity increases after, and not before, male takeovers. In addition, because takeovers themselves are seasonal in this population (this manuscript; see also Pappano and Beehner 2014), the post-takeover reproductive synchrony described here also produced a distinct pattern of birth seasonality. Why are male takeovers seasonal? In some systems, seasonal periods of female cycling and receptivity appear to attract male takeovers and/or influxes of males (e.g., Sugiyama and Ohsawa 1974; Borries 2000; Cords 2000; Morelli et al. 2009; Zhao et al. 2011; Hongo et al. 2016). However, most of these examples come from strict seasonal breeders, where females only conceive during a narrow window of time due to ecological constraints, and males target that seasonal peak in reproductive activity. Although we do not yet have a complete answer as to why takeovers cluster during one time of the year, we do not believe male geladas target a conception season in quite the same way, in part because geladas are not strict seasonal breeders. In addition, our finding that takeovers alter subsequent patterns of female fertility (i.e., the takeover precedes the increase in females that return to cycling and delays the birth peak) suggests that the timing of male takeovers does not simply “mirror” female reproductive seasonality. Finally, previous research in this population has suggested that relative male body condition is one important factor determining the outcome of male takeovers (Pappano and Beehner 2014). Therefore, we suggest that the timing of male takeovers is more or less independent from the timing of female reproductive seasonality, and we are currently conducting a more fine-grained analysis in order to determine the causal factors involved. Together, the evidence we report here for reproductive synchrony and seasonality following male takeovers is in line with the neuroendocrine literature on chemosensory mechanisms that stimulate or inhibit female reproduction. Specifically, in rodents, we see male-mediated resumption of cycling in females (the Whitten effect: Whitten et al. 1968), male-mediated female maturation (the Vandenbergh effect: Vandenbergh et al. 1972) and male-mediated pregnancy termination (the Bruce effect: Bruce 1959), in addition to sexually selected infanticide (Hrdy 1979). While we have yet to identify the mechanism(s) at work in geladas, our results suggest that such male-mediated proximate mechanisms shape gelada birth patterns. Further research will help untangle the evolutionary significance of such a response, which requires an analysis of individual females to see whether those that “reset” have a reproductive advantage over those that do not. We suspect that this kind of male-mediated birth seasonality could help explain seasonal reproductive patterns in species that experience infanticide, feticide, and/or the Bruce effect, or where alpha male replacements themselves (or their equivalent) are seasonal. For example, in white-faced capuchins, male takeovers are seasonal, and often result in infanticide (Fedigan 2003; Schoof and Jack 2013). Births show a seasonal peak in March, but also increase roughly 5 months after the peak takeover season (Carnegie et al. 2011). Here, we found no evidence that infants born outside of the ecological birth peak were more likely to die before 2 years of age. As likely capital breeders, this result is perhaps unsurprising: reproduction in capital breeders can only occur once females reach a certain condition threshold, which is thought to represent the appropriate energetic stores required to successfully carry a pregnancy to term. As such, any costs associated with reproduction should be paid upfront via infertility and/or miscarriage. We were unable to identify pregnancy loss in our dataset (we did not have hormones for most of these pregnancies) to address this directly. However, we expect that infertility is the primary "bottleneck" for reproduction in this population because females routinely give birth during off-peak periods. In this study, we were unable to identify pregnancy loss in the absence of hormonal monitoring because females continue to produce sexual swellings well into pregnancy (Roberts et al. 2017). Nevertheless, we do not suspect that pregnancy loss is a significant cost to breeding out-of-peak in this population, mainly because we routinely record births during off-peak periods. Instead, birth seasonality is likely due to periods of infertility. Still, there may be more subtle costs associated with birth timing. That is, births that occur out-of-peak (and not during the period of maximum food availability) may carry costs relating to maternal condition or to infant growth and development. Because our analysis was at the population-level, we did not examine individual-level characteristics that may influence infant survival. Therefore, the next questions are: 1) whether there is variability in infant survival if we factor in individual characteristics of the mother such as age, dominance rank, or unit size; and 2) whether there are any life history or developmental costs to giving birth outside of the ecological birth peak beyond the one we use here (infant survival to 2 years of age). For example, a female that gives birth outside of the peak may have a longer subsequent interbirth interval or suffer decreased longevity herself. Birth timing may also impact the availability of weaning foods at critical developmental periods for the infant (e.g., Koenig et al. 1997), or show overall effects on infant growth and development. The 2 birth peaks described here are reminiscent of those first described by Dunbar (1980). However, we give them very different interpretations. First, Dunbar (1980) described a dry season birth peak (November to January) that he attributed to a prospective strategy for females to avoid giving birth to vulnerable infants during the cold-wet season. By contrast, our dry season birth peak is driven by post-takeover females that give birth about 4 months after the population-wide seasonal birth peak. Second, Dunbar (1980) attributed his early wet-season birth peak (June to August) to a capital breeding strategy for females to exploit the plentiful food following the short rains (rain that we did not detect in our dataset). By contrast, while our late cold-wet season birth peak (August to October) falls within the window of peak green grass availability, it is also tightly associated with warm weather and low glucocorticoid metabolites at the time of conception. Therefore, our current working hypothesis is that reproduction in female geladas is limited by the energetics of thermoregulation in a cold, hypoxic environment. Testing this hypothesis further will require physiological data on energy balance. Finally, although the takeover birth peak helps explain some of the variation in birth timing observed, we still do not know why some females give birth outside of either birth peak. It may be that females that are shifted from the general ecological pattern (e.g., as the result of a takeover) have difficulties shifting themselves back to the ecological birth peak again. If costs to giving birth outside of the ecological peak are low (as our results suggest), then perhaps the time it takes to delay a reproductive event is a higher cost than just giving birth at a less opportune time. Future exploration of developmental costs associated with off-peak births will help elucidate whether our data are consistent with this hypothesis. SUPPLEMENTARY MATERIAL Supplementary data are available at Behavioral Ecology online. FUNDING This work was supported by: National Science Foundation (grant numbers BCS-0715179, IOS-1255974, BCS-1340911); Leakey Foundation; National Institutes of Health (grant number K99AG051764); National Geographic Society (grant number 8100–06, 8989-11); Sigma Xi; Wildlife Conservation Society; and the University of Michigan. We would like to thank the Ethiopian Wildlife Conservation Authority (EWCA) and the wardens and staff of the Simien Mountain National Park for their permission and on-going support for our long-term research project. Additionally, we are grateful to our excellent field team across the years, most especially E. Jejaw, A. Fanta, S. Girmay, J. Jarvey, and M. Gomery for their assistance with field data collection. We also owe thanks to A. Marshall, B. Dantzer, and J. Mitani for providing valuable feedback on data analyses. Finally, we would like to thank the 2 anonymous reviewers and the editors of Behavioral Ecology for their helpful feedback and comments on an earlier draft. Data accessibility: Analyses reported in this article can be reproduced using the data provided by Tinsley Johnson et al. (2018). REFERENCES Agrell J , Wolff JO , Yloenen H . 1998 . Counter-strategies to infanticide in mammals: costs and consequences . Oikos . 83 : 507 – 517 . Google Scholar CrossRef Search ADS Anderson DR , Burnham KP . 2002 . Avoiding pitfalls when using information-theoretic methods . J Wildl Manage . 66 : 912 – 918 . Google Scholar CrossRef Search ADS Arlettaz R , Christe P , Lugon A , Perrin N , Vogel P . 2001 . Food availability dictates the timing of parturition in insectivorous mouse-eared bats . Oikos . 95 : 105 – 111 . Google Scholar CrossRef Search ADS Bartoń K . 2016 . MuMIn: Multi-model inference. R package version 1.15.6 [cited 2018 February 10] . Available from: https://cran.r-project.org/web/packages/MuMIn. Bates D , Maechler M , Bolker B , Walker S . 2015 . Fitting linear mixed-effects models using lme4 . J Stat Softw . 67 : 1 – 48 . Google Scholar CrossRef Search ADS Beehner JC , Bergman TJ . 2008 . Infant mortality following male takeovers in wild geladas . Am J Primatol . 70 : 1152 – 1159 . Google Scholar CrossRef Search ADS PubMed Beehner JC , Bergman TJ . 2017 . The next step for stress research in primates: To identify relationships between glucocorticoid secretion and fitness . Horm Behav . 91 : 68 – 83 . Google Scholar CrossRef Search ADS PubMed Beehner JC , Bergman TJ , Cheney DL , Seyfarth RM , Whitten PL . 2005 . The effect of new alpha males on female stress in free-ranging baboons . Anim Behav . 69 : 1211 – 1221 . Google Scholar CrossRef Search ADS Beehner JC , McCann C . 2008 . Seasonal and altitudinal effects on glucocorticoid metabolites in a wild primate (Theropithecus gelada) . Physiol Behav . 95 : 508 – 514 . Google Scholar CrossRef Search ADS PubMed Beehner JC , Onderdonk DA , Alberts SC , Altmann J . 2006 . The ecology of conception and pregnancy failure in wild baboons . Behav Ecol . 17 : 741 – 750 . Google Scholar CrossRef Search ADS Beehner JC , Whitten PL . 2004 . Modifications of a field method for fecal steroid analysis in baboons . Physiol Behav . 82 : 269 – 277 . Google Scholar CrossRef Search ADS PubMed Berger J . 1983 . Induced abortion and social factors in wild horses . Nature . 303 : 59 – 61 . Google Scholar CrossRef Search ADS PubMed Boness DJ , Bowen WD , Iverson SJ . 1995 . Does male harassment of females contribute to reproductive synchrony in the grey seal by affecting maternal performance ? Behav Ecol Sociobiol . 36 : 1 – 10 . Google Scholar CrossRef Search ADS Borries C . 2000 . Male dispersal and mating season influxes in hanuman langurs living in multi-male groups . In: Kappeler PM , editor. Primate males: causes and consequences of variation in group composition . Cambridge : Cambridge University Press . p. 146 . Breen KM , Oakley AE , Pytiak AV , Tilbrook AJ , Wagenmaker ER , Karsch FJ . 2007 . Does cortisol acting via the type II glucocorticoid receptor mediate suppression of pulsatile luteinizing hormone secretion in response to psychosocial stress ? Endocrinology . 148 : 1882 – 1890 . Google Scholar CrossRef Search ADS PubMed Brockman DK , van Schaik CP . 2005 . Seasonality and reproductive function . In: Brockman DK , van Schaik CP , editors. Seasonality in primates: studies of living and extinct human and non-human primates . Vol. 44 . Cambridge : Cambridge University Press . p. 269 – 305 . Google Scholar CrossRef Search ADS Bronson FH . 1985 . Mammalian reproduction: an ecological perspective . Biol Reprod . 32 : 1 – 26 . Google Scholar CrossRef Search ADS PubMed Bronson FH . 1989 . Mammalian reproductive biology . Chicago (IL) : University of Chicago Press . Bronson FH . 2009 . Climate change and seasonal reproduction in mammals . Philos Trans R Soc Lond B Biol Sci . 364 : 3331 – 3340 . Google Scholar CrossRef Search ADS PubMed Bruce HM . 1959 . An exteroceptive block to pregnancy in the mouse . Nature . 184 : 105 . Google Scholar CrossRef Search ADS PubMed Burthe S , Butler A , Searle KR , Hall SJ , Thackeray SJ , Wanless S . 2011 . Demographic consequences of increased winter births in a large aseasonally breeding mammal (Bos taurus) in response to climate change . J Anim Ecol . 80 : 1134 – 1144 . Google Scholar CrossRef Search ADS PubMed Campos FA , Morris WF , Alberts SC , Altmann J , Brockman DK , Cords M , Pusey A , Stoinski TS , Strier KB , Fedigan LM . 2017 . Does climate variability influence the demography of wild primates? Evidence from long-term life-history data in seven species . Glob Chang Biol . 23 : 4907 – 4921 . Google Scholar CrossRef Search ADS PubMed Carnegie SD , Fedigan LM , Melin AD . 2011 . Reproductive seasonality in female capuchins (Cebus capucinus) in Santa Rosa (Area de Conservacion Guanacaste), Costa Rica . Int J Primatol . 32 : 1076 – 1090 . Google Scholar CrossRef Search ADS Chappell MA , Hammond KA . 2004 . Maximal aerobic performance of deer mice in combined cold and exercise challenges . J Comp Physiol B . 174 : 41 – 48 . Google Scholar CrossRef Search ADS PubMed Cheviron ZA , Bachman GC , Storz JF . 2013 . Contributions of phenotypic plasticity to differences in thermogenic performance between highland and lowland deer mice . J Exp Biol . 216 : 1160 – 1166 . Google Scholar CrossRef Search ADS PubMed Clutton-Brock TH , Albon SD , Guinness FE . 1989 . Fitness costs of gestation and lactation in wild mammals . Nature . 337 : 260 – 262 . Google Scholar CrossRef Search ADS PubMed Colmenares F , Gomendio M . 1988 . Changes in female reproductive condition following male take-overs in a colony of hamadryas and hybrid baboons . Folia Primatol (Basel) . 50 : 157 – 174 . Google Scholar CrossRef Search ADS PubMed Cords M . 2000 . The number of males in guenon groups . In: Kappeler PM , editor. Primate males: causes and consequences of variation in group composition . Cambridge : Cambridge University Press . p. 84 – 96 . Crook JH , Gartlan JS . 1966 . Evolution of primate societies . Nature . 210 : 1200 – 1203 . Google Scholar CrossRef Search ADS PubMed Dantzer B , Fletcher QE , Boonstra R , Sheriff MJ . 2014 . Measures of physiological stress: a transparent or opaque window into the status, management and conservation of species ? Conserv Physiol . 2 : cou023 . Google Scholar CrossRef Search ADS PubMed Di Bitetti MS , Janson CH . 2000 . When will the stork arrive? Patterns of birth seasonality in neotropical primates . Am J Primatol . 50 : 109 – 130 . Google Scholar CrossRef Search ADS PubMed Drent RH , Daan S . 1980 . The prudent parent: energetic adjustments in avian breeding . Ardea . 68 : 225 – 252 . Dubost G , Henry O . 2017 . Seasonal reproduction in neotropical rainforest mammals . Zool Stud . 56:2 . Dunbar RIM . 1980 . Demographic and life history variables of a population of gelada baboons (Theropithecus gelada) . J Anim Ecol . 49 : 485 – 506 . Google Scholar CrossRef Search ADS Dunbar RIM . 1984 . Reproductive decisions: an economic analysis of gelada baboon social strategies . Princeton : Princeton University Press . Dunbar RIM , Dunbar P . 1974 . Social dynamics of gelada baboons . Anim Behav . 22 : 203 . Google Scholar CrossRef Search ADS PubMed Dunbar RIM , Hannah-Stewart L , Dunbar P . 2002 . Forage quality and the costs of lactation for female gelada baboons . Anim Behav . 64 : 801 – 805 . Google Scholar CrossRef Search ADS Erb WM , Borries C , Lestari NS , Hodges JK . 2012 . Annual variation in ecology and reproduction of wild simakobu (Simias concolor) . Int J Primatol . 33 : 1406 – 1419 . Google Scholar CrossRef Search ADS Fairbanks LA , McGuire MT . 1987 . Mother-infant relationships in vervet monkeys: response to new adult males . Int J Primatol . 8 : 351 – 366 . Google Scholar CrossRef Search ADS Fardi S , Sauther ML , Cuozzo FP , Jacky IAY , Bernstein RM . 2017 . The effect of extreme weather events on hair cortisol and body weight in a wild ring-tailed lemur population (Lemur catta) in southwestern Madagascar . Am J Primatol. e22731 . Fedigan LM . 2003 . Impact of male takeovers on infant deaths, births and conceptions in Cebus capucinus at Santa Rosa, Costa Rica . Int J Primatol . 24 : 723 – 741 . Google Scholar CrossRef Search ADS Fernandez-Duque E , Rotundo M , Ramirez-Llorens P . 2002 . Environmental determinants of birth seasonality in night monkeys (Aotus azarai) of the Argentinean Chaco . Int J Primatol . 23 : 639 – 656 . Google Scholar CrossRef Search ADS Foerster S , Cords M , Monfort SL . 2012 . Seasonal energetic stress in a tropical forest primate: proximate causes and evolutionary implications . PLoS One . 7 : e50108 . Google Scholar CrossRef Search ADS PubMed Fraser-Smith AC . 1975 . Male-induced pregnancy termination in the prairie vole, Microtus ochrogaster . Science . 187 : 1211 – 1213 . Google Scholar CrossRef Search ADS PubMed Gesquiere LR , Khan M , Shek L , Wango TL , Wango EO , Alberts SC , Altmann J . 2008 . Coping with a challenging environment: effects of seasonal variability and reproductive status on glucocorticoid concentrations of female baboons (Papio cynocephalus) . Horm Behav . 54 : 410 – 416 . Google Scholar CrossRef Search ADS PubMed Gilchrist JS . 2006 . Female eviction, abortion, and infanticide in banded mongooses (Mungos mungo): implications for social control of reproduction and synchronized parturition . Behav Ecol . 17 : 664 – 669 . Google Scholar CrossRef Search ADS Hayes JP . 1989 . Field and maximal metabolic rates of deer mice (Peromyscus maniculatus) at low and high altitudes . Physiol Zool . 62 : 732 – 744 . Google Scholar CrossRef Search ADS Heideman PD , Utzurrum RC . 2003 . Seasonality and synchrony of reproduction in three species of nectarivorous Philippines bats . BMC Ecol . 3 : 11 . Google Scholar CrossRef Search ADS PubMed Hill RA , Lycett JE , Dunbar RIM . 2000 . Ecological and social determinants of birth intervals in baboons . Behav Ecol . 11 : 560 – 564 . Google Scholar CrossRef Search ADS Hodge SJ , Bell MB , Cant MA . 2011 . Reproductive competition and the evolution of extreme birth synchrony in a cooperative mammal . Biol Lett . 7 : 54 – 56 . Google Scholar CrossRef Search ADS PubMed Hongo S , Nakashima Y , Akomo-Okoue EF , Mindonga-Nguelet FL . 2016 . Female reproductive seasonality and male influxes in wild mandrills (Mandrillus sphinx) . Int J Primatol . 37 : 416 – 437 . Google Scholar CrossRef Search ADS Hrdy SB . 1974 . Male-male competition and infanticide among the langurs (Presbytis entellus) of Abu, Rajasthan . Folia Primatol (Basel) . 22 : 19 – 58 . Google Scholar CrossRef Search ADS PubMed Hrdy SB . 1979 . Infanticide among animals: a review, classification, and examination of the implications for the reproductive strategies of females . Ethol Sociobiol . 1 : 13 – 40 . Google Scholar CrossRef Search ADS Hunter CP . 2001 . Ecological determinants of gelada ranging patterns (Theropithecus gelada) . Ph.D. dissertation, The University of Liverpool, Liverpool. Ims RA . 1990 . The ecology and evolution of reproductive synchrony . Trends Ecol Evol . 5 : 135 – 140 . Google Scholar CrossRef Search ADS PubMed Iwamoto T , Dunbar RIM . 1983 . Thermoregulation, habitat quality and the behavioural ecology of gelada baboons . J Anim Ecol . 52 : 357 – 366 . Google Scholar CrossRef Search ADS Janson C , Verdolin J . 2005 . Seasonality of primate births in relation to climate . In: Brockman DK , van Schaik CP , editors. Seasonality in primates: studies of living and extinct human and non-human primates . Vol. 44 . Cambridge : Cambridge University Press . p. 307 . Google Scholar CrossRef Search ADS Jarvey JC , Low BS , Pappano DJ , Bergman TJ , Beehner JC . In press. Graminivory and fallback foods: annual diet profile of geladas (Theropithecus gelada) living in the Simien Mountains National Park, Ethiopia . Int J Primatol . Jessop TS , Lane ML , Teasdale L , Stuart-Fox D , Wilson RS , Careau V , Moore IT . 2016 . Multiscale evaluation of thermal dependence in the glucocorticoid response of vertebrates . Am Nat . 188 : 342 – 356 . Google Scholar CrossRef Search ADS PubMed Johnson JB , Omland KS . 2004 . Model selection in ecology and evolution . Trends Ecol Evol . 19 : 101 – 108 . Google Scholar CrossRef Search ADS PubMed Koenig A , Borries C , Chalise MK , Winkler P . 1997 . Ecology, nutrition, and timing of reproductive events in an Asian primate, the Hanuman langur (Presbytis entellus) . J Zool . 243 : 215 – 235 . Google Scholar CrossRef Search ADS Lancaster JB , Lee RB . 1965 . The annual reproductive cycle in monkeys and apes . In: Devore I , editor. Primate behavior . New York (NY) : Holt, Rinehart and Winston . p. 486 – 513 . Landys MM , Ramenofsky M , Wingfield JC . 2006 . Actions of glucocorticoids at a seasonal baseline as compared to stress-related levels in the regulation of periodic life processes . Gen Comp Endocrinol . 148 : 132 – 149 . Google Scholar CrossRef Search ADS PubMed Lea AJ , Altmann J , Alberts SC , Tung J . 2015 . Developmental constraints in a wild primate . Am Nat . 185 : 809 – 821 . Google Scholar CrossRef Search ADS PubMed Loudon ASI , Racey PA . 1987 . Reproductive energetics in mammals: The proceedings of a symposium held at the Zoological Society of London on 10th and 11th April 1986. Oxford: Published for the Zoological Society of London by Clarendon Press . Maestripieri D , Georgiev AV . 2016 . What cortisol can tell us about the costs of sociality and reproduction among free-ranging rhesus macaque females on Cayo Santiago . Am J Primatol . 78 : 92 – 105 . Google Scholar CrossRef Search ADS PubMed Manning JM , Bronson FH . 1990 . The effects of low temperature and food intake on ovulation in domestic mice . Physiol Zool . 63 : 938 – 948 . Google Scholar CrossRef Search ADS Martínez-Mota R , Righini N , Palme R . 2016 . Fluctuations in daily energy intake do not cause physiological stress in a Neotropical primate living in a seasonal forest . Oecologia . 182 : 973 – 984 . Google Scholar CrossRef Search ADS PubMed McCabe GM , Thompson ME . 2013 . Reproductive seasonality in wild Sanje mangabeys (Cercocebus sanjei), Tanzania: relationship between the capital breeding strategy and infant survival . Behaviour . 150 : 1 – 31 . Google Scholar CrossRef Search ADS McFarland R , Barrett L , Boner R , Freeman NJ , Henzi SP . 2014 . Behavioral flexibility of vervet monkeys in response to climatic and social variability . Am J Phys Anthropol . 154 : 357 – 364 . Google Scholar CrossRef Search ADS PubMed Morelli TL , King SJ , Pochron ST , Wright PC . 2009 . The rules of disengagement: Takeovers, infanticide, and dispersal in a rainforest lemur, Propithecus edwardsi . Behaviour . 146 : 499 – 523 . Google Scholar CrossRef Search ADS Morino L , Borries C . 2017 . Offspring loss after male change in wild siamangs: the importance of abrupt weaning and male care . Am J Phys Anthropol . 162 : 180 – 185 . Google Scholar CrossRef Search ADS PubMed Negus NC , Berger PJ . 1972 . Environmental factors and reproductive processes in mammalian populations . In: Velardo JT , Kasprow BA , editors. Biology of reproduction: basic and clinical studies . New Orleans (LA) : Third Pan American Congress on Anatomy . p. 89 – 98 . Ohsawa H , Dunbar RIM . 1984 . Variations in the demographic-structure and dynamics of gelada baboon populations . Behav Ecol Sociobiol . 15 : 231 – 240 . Google Scholar CrossRef Search ADS Ostner J , Nunn CL , Schülke O . 2008 . Female reproductive synchrony predicts skewed paternity across primates . Behav Ecol . 19 : 1150 – 1158 . Google Scholar CrossRef Search ADS PubMed Packer C , Pusey AE . 1983 . Male takeovers and female reproductive parameters: a simulation of oestrous synchrony in lions (Panthera leo) . Anim Behav . 31 : 334 – 340 . Google Scholar CrossRef Search ADS Palombit RA . 2015 . Infanticide as sexual conflict: coevolution of male strategies and female counterstrategies . Cold Spring Harb Perspect Biol . 7:a017640 . Pappano DJ , Beehner JC . 2014 . Harem-holding males do not rise to the challenge: androgens respond to social but not to seasonal challenges in wild geladas . R Soc Open Sci . 1 : 140081 . Google Scholar CrossRef Search ADS PubMed Polo P , Colmenares F . 2016 . Seasonality of reproductive events and early mortality in a colony of hamadryas baboons (Papio hamadryas hamadryas) over a 30-year period: capital breeding and life history patterns in a food-provisioned population seasonally thermally stressed . Am J Primatol . 78 : 1149 – 1164 . Google Scholar CrossRef Search ADS PubMed Pride RE . 2005 . Optimal group size and seasonal stress in ring-tailed lemurs (Lemur catta) . Behav Ecol . 16 : 550 – 560 . Google Scholar CrossRef Search ADS R Core Team . 2016 . R: A language and environment for statistical computing. Vienna (Austria): R Foundation for Statistical Computing [cited 2018 February 10]. Available from: https://www.R-project.org/ . Riehl C . 2016 . Infanticide and within-clutch competition select for reproductive synchrony in a cooperative bird . Evolution . 70 : 1760 – 1769 . Google Scholar CrossRef Search ADS PubMed Roberts SB , Cole TJ , Coward WA . 1985 . Lactational performance in relation to energy intake in the baboon . Am J Clin Nutr . 41 : 1270 – 1276 . Google Scholar CrossRef Search ADS PubMed Roberts EK , Lu A , Bergman TJ , Beehner JC . 2012 . A Bruce effect in wild geladas . Science . 335 : 1222 – 1225 . Google Scholar CrossRef Search ADS PubMed Roberts EK , Lu A , Bergman TJ , Beehner JC . 2017 . Female reproductive parameters in wild geladas (Theropithecus gelada) . Int J Primatol . 38 : 1 – 20 . Google Scholar CrossRef Search ADS Roberts S-J , Nikitopoulos E , Cords M . 2014 . Factors affecting low resident male siring success in one-male groups of blue monkeys . Behav Ecol . 25 : 852 – 861 . Google Scholar CrossRef Search ADS van Schaik CP , van Noordwijk MA , Nunn CL . 1999 . Sex and social evolution in primates . In: Lee PC, editor. Comparative Primate Socioecoogy. Cambridge: Cambridge University Press. p. 204 – 231 . Schoof VA , Jack KM . 2013 . The association of intergroup encounters, dominance status, and fecal androgen and glucocorticoid profiles in wild male white-faced capuchins (Cebus capucinus) . Am J Primatol . 75 : 107 – 115 . Google Scholar CrossRef Search ADS PubMed Schülke O , Ostner J . 2012 . Ecological and social influences on sociality . In: Call J , Kappeler PM , Palombit RA , Silk JB , Mitani JC , editors. The evolution of primate societies . Chicago : The University of Chicago Press . p. 195 – 219 . Smuts BB , Smuts RW . 1993 . Male aggression and sexual coercion of females in nonhuman primates and other mammals: evidence and theoretical implications . Adv Study Behav . 22 : 1 – 63 . Google Scholar CrossRef Search ADS Snyder-Mackler N , Alberts SC , Bergman TJ . 2012a . Concessions of an alpha male? Cooperative defence and shared reproduction in multi-male primate groups . Proc Biol Sci . 279 : 3788 – 3795 . Google Scholar CrossRef Search ADS Snyder-Mackler N , Beehner JC , Bergman TJ . 2012b . Defining higher levels in the multilevel societies of geladas (Theropithecus gelada) . Int J Primatol . 33:1054-1068. Stutchbury BJ , Morton ES . 1995 . The effect of breeding synchrony on extra-pair mating systems in songbirds . Behaviour . 132 : 675 – 690 . Google Scholar CrossRef Search ADS Sugiyama Y . 1965 . On the social change of hanuman langurs (Presbytis entellus) in their natural condition . Primates . 6 : 381 – 418 . Google Scholar CrossRef Search ADS Sugiyama Y , Ohsawa H . 1974 . Population dynamics of Japanese macaques at Ryozenyama, Suzuka Mts. I. General view . Jap J Ecol . 24 : 50 – 59 . Tecot SR . 2010 . It’s all in the timing: Birth seasonality and infant survival in Eulemur rubriventer . Int J Primatol . 31 : 715 – 735 . Google Scholar CrossRef Search ADS Teichroeb JA , Jack KM . 2017 . Alpha male replacements in nonhuman primates: variability in processes, outcomes, and terminology . Am J Primatol . 79:e22674 . Tinsley Johnson ET , Snyder-Mackler N , Lu A , Bergman TJ , Beehner JC . 2018 . Data from: social and ecological drivers of reproductive seasonality in geladas . Dryad Digital Repository . http://dx.doi.org/10.5061/dryad. Valeggia C , Ellison PT . 2004 . Lactational amenorrhoea in well-nourished Toba women of Formosa, Argentina . J Biosoc Sci . 36 : 573 – 595 . Google Scholar CrossRef Search ADS PubMed Vandenbergh JG , Drickamer LC , Colby DR . 1972 . Social and dietary factors in the sexual maturation of female mice . J Reprod Fertil . 28 : 397 – 405 . Google Scholar CrossRef Search ADS PubMed Wade GN , Schneider JE . 1992 . Metabolic fuels and reproduction in female mammals . Neurosci Biobehav Rev . 16 : 235 – 272 . Google Scholar CrossRef Search ADS PubMed Ward M, Milledge PJS, West JB. 1995. High altitude physiology and medicine. London (UK): Chapman & Hall Medical. Wasser SK . 1996 . Reproductive control in wild baboons measured by fecal steroids . Biol Reprod . 55 : 393 – 399 . Google Scholar CrossRef Search ADS PubMed Weingrill T , Gray DA , Barrett L , Henzi SP . 2004 . Fecal cortisol levels in free-ranging female chacma baboons: relationship to dominance, reproductive state and environmental factors . Horm Behav . 45 : 259 – 269 . Google Scholar CrossRef Search ADS PubMed Whitten WK , Bronson FH , Greenstein JA . 1968 . Estrus-inducing pheromone of male mice: transport by movement of air . Science . 161 : 584 – 585 . Google Scholar CrossRef Search ADS PubMed Xiang ZF , Sayers K . 2009 . Seasonality of mating and birth in wild black-and-white snub-nosed monkeys (Rhinopithecus bieti) at Xiaochangdu, Tibet . Primates . 50 : 50 – 55 . Google Scholar CrossRef Search ADS PubMed Zhao Q , Borries C , Pan W . 2011 . Male takeover, infanticide, and female countertactics in white-headed leaf monkeys (Trachypithecus leucocephalus) . Behav Ecol Sociobiol . 65 : 1535 – 1547 . Google Scholar CrossRef Search ADS Zinner D , Deschner T . 2000 . Sexual swellings in female hamadryas baboons after male take-overs: ‘Deceptive’ swellings as a possible female counter-strategy against infanticide . Am J Primatol . 52 : 157 – 168 . Google Scholar CrossRef Search ADS PubMed Zipple MN , Grady JH , Gordon JB , Chow LD , Archie EA , Altmann J , Alberts SC . 2017 . Conditional fetal and infant killing by male baboons . Proc Biol Sci . 284:20162561 . © The Author(s) 2018. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Behavioral Ecology Oxford University Press

Social and ecological drivers of reproductive seasonality in geladas

Loading next page...
 
/lp/ou_press/social-and-ecological-drivers-of-reproductive-seasonality-in-geladas-xvIwPpmW9F
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
1045-2249
eISSN
1465-7279
D.O.I.
10.1093/beheco/ary008
Publisher site
See Article on Publisher Site

Abstract

Abstract Many nonseasonally breeding mammals demonstrate some degree of synchrony in births, which is generally associated with ecological factors that mediate fecundity. However, disruptive social events, such as alpha male replacements, also have the potential to affect the timing of female reproduction. Here, we examined reproductive seasonality in a wild population of geladas (Theropithecus gelada) living at high altitudes in an afro-alpine ecosystem in Ethiopia. Using 9 years of demographic data (2006–2014), we determined that, while females gave birth year-round, a seasonal peak in births coincided with peak green grass availability (their staple food source). This post-rainy season “ecological peak” in births meant that estimated conceptions for these births occurred when temperatures were at their highest and mean female fecal glucocorticoid concentrations were at their lowest. In addition to this ecological birth peak, we also found a separate birth peak that occurred only for females in groups that had experienced a recent replacement of the dominant male (i.e., a takeover). Because new dominant males cause abortions in pregnant females and kill the infants of lactating females, takeovers effectively “reset” the reproductive cycles of females in the group. This “social birth peak” was distinct from the ecological peak and was associated with higher rates of cycling and conceptions overall and higher glucocorticoid levels immediately following a takeover as compared to females that did not experience a takeover. These data demonstrate that social factors (in this case, male takeovers) can contribute to population-level reproductive seasonality above and beyond group-level reproductive synchrony. INTRODUCTION The ways in which ecology interacts with reproductive timing can be characterized as a continuum, reflecting various species-specific energy-use strategies and patterns (Negus and Berger 1972; Drent and Daan 1980; Brockman and van Schaik 2005; Janson and Verdolin 2005). In the strictest sense, income breeders, at one end of the continuum, rely on external cues (e.g., photoperiod) to time conceptions so that vulnerable stages of the reproductive cycle, such as births or weaning periods, coincide with optimal ecological conditions. As a result, income breeders are typically strict seasonal breeders, with a discrete breeding season and birth season each year (though there may be some flexibility in birth timing in certain cases, e.g., Tecot 2010; Carnegie et al. 2011). Capital breeders, at the other end of the continuum, rely on internal cues of condition (e.g., energy balance, or the ratio between energy intake and energy expenditure: Valeggia and Ellison 2004), which allows for reproduction at any point in the year, as long as energetic requirements are met. Therefore, the timing of births for capital breeders depends on the seasonality and predictability of the environment: in a highly seasonal and predictable environment, capital breeders may demonstrate a strict birth season, much like income breeders do. On the other hand, in a less seasonal or less predictable environment, capital breeders may exhibit no seasonal patterns in births, or they may exhibit some degree of birth seasonality (where births cluster around the same time each year) or birth synchrony (where births cluster around certain times within a year: Heideman and Utzurrum 2003) following periods of energetic abundance. At a proximate level, birth seasonality is shaped by ecological factors that impact a female’s ability to “finance” a reproductive event across critical stages, from ovulation to conception and through parturition. Energy balance can impact a female’s ability to initiate (e.g., ovulation and conception: Bronson 1989; Wade and Schneider 1992; McCabe and Thompson 2013) or sustain (e.g., gestation and lactation: Roberts et al. 1985; Clutton-Brock et al. 1989; Wade and Schneider 1992; Beehner et al. 2006; Maestripieri and Georgiev 2016) a reproductive event. Birth seasonality (i.e., either a birth peak, when a high proportion of births are concentrated at a particular time of year, or a birth valley, when the proportion of births drops significantly at a particular time of year: Lancaster and Lee 1965; Janson and Verdolin 2005; Tecot 2010; Carnegie et al. 2011; Erb et al. 2012) can arise when maternal condition, specifically maternal energy balance, varies consistently across seasons. Energy balance is a function of energy intake and energy expenditure, and both are known to shape birth seasonality across many mammalian species (reviewed in: Di Bitetti and Janson 2000; Brockman and van Schaik 2005; Janson and Verdolin 2005). Because directly assessing energy intake can be methodologically challenging, many studies have successfully used food availability and/or rainfall as proxies for energetic condition in a number of studies (reviewed in Martínez-Mota et al. 2016). Both high food availability (e.g., in insectivorous mouse-eared bats, Myotis blythii: Arlettaz et al. 2001) and low food availability (e.g., in a survey of neotropical mammals: Dubost and Henry 2017) have been associated with either birth peaks or birth valleys, respectively. Energy expenditure can vary according to a number of different factors, from day range length to thermoregulation. In mammals, thermoregulatory demands due to both extreme heat and cold are known to hinder female reproduction (e.g., in owl monkeys, Aotus azarai: Fernandez-Duque et al. 2002; in black-and-white snub-nosed monkeys, Rhinopithecus bieti: Xiang and Sayers 2009; reviewed in Bronson 1985; 1989; Loudon and Racey 1987; Manning and Bronson 1990). For example, in captive conditions, the energetic demands of thermoregulation during extremely cold months were shown to impact reproduction even when energy intake was not restricted (e.g., in captive hamadryas baboons, Papio hamadryas: Polo and Colmenares 2016). Moreover, cold stress is known to be compounded by the hypoxic (i.e., low oxygen) conditions of high altitude where thermogenic capacity is even more constrained (e.g., Hayes 1989; Ward M, Milledge PJS, West JB. 1995; Chappell and Hammond 2004; Cheviron et al. 2013). One way to identify the harsh conditions that hamper female reproduction is to measure energy constraints (or their proxies), as outlined above. Another informative approach is to measure the internal physiological state of the organisms living through these adverse conditions. In particular, glucocorticoids (and glucocorticoid metabolites, or GCMs) are steroid hormones that lend themselves to this endeavor: first, GCMs rise in response to the energetic demands that accompany ecological and/or social challenges, such as food scarcity (e.g., Pride 2005; Gesquiere et al. 2008; Foerster et al. 2012), extreme temperatures (e.g., Weingrill et al. 2004, Beehner and McCann 2008; reviewed in Jessop et al. 2016), or the threat of infanticide from an incoming male (e.g., Beehner et al. 2005). Second, GCM concentrations in fecal, urine, and hair samples have been shown to increase with an individual’s recent exposure to ecological and/or social challenges (e.g., Fardi et al. 2017; reviewed in Dantzer et al. 2014; Beehner and Bergman 2017). Activation of the hypothalamic-pituitary adrenal axis (HPA-axis), which results in an increase in GCM secretion, has also been repeatedly associated with reproductive suppression (via the inhibition of pulsatile leutenizing hormone: Wasser 1996; Landys et al. 2006; Breen et al. 2007). Therefore, GCM profiles used in conjunction with measurements of challenging conditions (ecological or social) can help identify circumstances where reproductive constraints might be greatest. Group-living animals must contend with challenging social environments in addition to any harsh climatological conditions; correspondingly, the social environment can also alter the timing of female reproduction. For example, reproductive synchrony among females (where births cluster together within a year) can arise in response to social as well as ecological forces (reviewed in Ims 1990). One salient social threat to female reproduction known to promote reproductive synchrony is infanticide by males (reviewed in Agrell et al. 1998). Infanticide occurs when nonsire males kill the dependent offspring of females to expedite reproductive cycling and mating receptivity in the mother (Sugiyama 1965; Hrdy 1974; reviewed in: Hrdy 1979; Fedigan 2003; Palombit 2015). Perhaps in response to the threat of infanticide, lactating females often prematurely wean their dependent offspring (i.e., accelerated weaning: e.g., vervet monkeys, Chlorocebus pygerythrus: Fairbanks and McGuire 1987; siamangs, Symphalangus syndactylus: Morino and Borries 2017; reviewed in Smuts and Smuts 1993). Similarly, nonsire males can cause the death of a fetus in utero directly (i.e., sexually-selected feticide, Zipple et al. 2017) or indirectly (i.e., the Bruce effect: Bruce 1959; e.g., prairie voles, Microtus ochrogaster: Fraser-Smith 1975; wild horses, Equus caballus: Berger 1983; geladas, T. gelada: Roberts et al. 2012). Therefore, the arrival of nonsire males represents a challenging social environment that is known to “reset” the reproductive cycles of females. In social species, such as primates, the arrival of a nonsire male poses an infanticidal threat primarily when he takes over the dominant male position of a group (i.e., a takeover: Teichroeb and Jack 2017). Takeovers impact all females in a group simultaneously—effectively synchronizing their estrous cycles in the months that follow (e.g., Packer and Pusey 1983; Colmenares and Gomendio 1988; Ims 1990) and altering the timing of subsequent births. As a result, male takeovers have the potential to alter reproductive timing to such an extent that ecological patterns of reproductive seasonality is disrupted, which may result in fitness costs for affected females. In some species, females may need to delay conceiving until ecological conditions improve, while in others, females may conceive immediately following the death of their infant, resulting in births that fall outside any seasonal birth peak. If such “mis-timed” births occur during periods of negative energy balance, then it stands to reason that females will suffer downstream costs (e.g., neonatal loss, slow infant development and/or a prolonged interbirth interval) over and above the costs of infanticide alone. For example, in recent decades many species are experiencing rapid changes in their ecology due to climate change; these changes are causing slight shifts in birth timing that can result in mis-timed and, as a result, less successful births (e.g., in cattle, Bos taurus L.: Burthe et al. 2011; reviewed in Bronson 2009; Campos et al. 2017). Moreover, if male takeovers are seasonal (e.g., in white-faced capuchins, Cebus capucinus: Schoof and Jack 2013; reviewed in Teichroeb and Jack 2017), then takeovers themselves may produce a distinct “social birth peak” in addition to (or instead of) the usual “ecological birth peak.” Here, we examined the impact of male takeovers on reproductive seasonality in a population of wild primates, the gelada (T. gelada), living in the Simien Mountains National Park, Ethiopia. Geladas are an ideal species for this inquiry for a number of reasons. First, the social structure of geladas allows us to study dozens of groups, or “reproductive units,” at once. Reproductive units (hereafter, “units”) comprise the core groups of gelada multi-level society, and are composed of a dominant leader male, 1–12 adult females and their offspring, and possibly one or more subordinate follower males (Dunbar 1980; Snyder-Mackler et al. 2012b). Bachelor males form peripheral “all-male groups,” and challenge dominant leader males for reproductive control of the females within a unit (Dunbar and Dunbar 1974; Dunbar 1984). Second, male takeovers (i.e., when a bachelor male defeats a dominant leader male), are a frequent occurrence in this population (0.32 takeovers/unit/year, Beehner and Bergman 2008) and have severe consequences on female reproduction. Following takeovers, incoming males are known to commit infanticide (killing up to half of unit infants: Beehner and Bergman 2008), they trigger pregnancy termination (via the Bruce effect in 80% of females: Roberts et al. 2012), and they may cause lactating females to begin cycling sooner than they would otherwise (Dunbar 1980). Moreover, because male takeovers in this population follow a seasonal pattern themselves (with a peak between February to April each year, Pappano and Beehner 2014), takeovers have the potential to not only disrupt reproductive patterns for females that experience a takeover but to also produce their own seasonal peak in reproduction for these same females. Third, geladas represent an opportunity to examine the separate ecological effects of rainfall and temperature (as proxies for green grass availability and thermoregulatory constraints, respectively) on reproductive timing. Availability of green montane grass, the staple food for geladas, fluctuates throughout the year and peaks following the rainy season (Jarvey et al., In press). However, this optimal period of green grass availability also coincides with the least optimal temperatures for this region, when temperatures routinely fall below freezing (Iwamoto and Dunbar 1983; Ohsawa and Dunbar 1984). Previous research has demonstrated that male geladas in this same population exhibited the highest fecal GCMs during these cold-wet months (Beehner and McCann 2008), however, seasonal patterns of GCMs in females have never been examined. Females geladas (like males) should also be sensitive to thermoregulatory constraints, but females (unlike males) are expected to be sensitive to fluctuations in green grass availability due to the energetic demands of reproduction (Crook and Gartlan 1966; reviewed in Schülke and Ostner 2012). Indeed, for this same population of geladas, Dunbar (1980) identified a seasonal pattern to births that included not one, but 2 distinct birth peaks—one during the dry season and another during the wet season. He suggested that the 2 peaks reflected a compromise between the constraints of food and temperature: females were more likely to conceive when food availability was high, but also avoided giving birth to vulnerable infants during the cold-wet season. Our first aim in this study is to confirm the nature of reproductive seasonality in geladas with a larger dataset across nearly a decade (9 years) in conjunction with seasonality in rainfall and temperature. Our second aim is to use fecal hormone profiles (GCMs) to test 2 separate predictions about seasonal causes of metabolic stress. If green grass availability represents a significant metabolic constraint for females, then we expect to see the highest female GCMs during the dry season, when green grass is scarce. Alternatively, or additionally, if thermoregulatory pressures represent a significant metabolic constraint for females, then we expect to see the highest GCMs during the late rainy season, when temperatures are lowest. We predicted that, as capital breeders, the onset of female gelada reproduction would coincide with periods of low metabolic demands as indicated by low GCM profiles overall. Our third aim is to assess the extent to which male takeovers, with their known effects on female reproduction, explain seasonal variation in birth patterns. We predicted that male takeovers would be associated with higher GCM profiles and within-unit reproductive synchrony for affected females. Moreover, due to the seasonality of takeovers, we predicted that takeovers would produce some degree of across-unit reproductive seasonality. Finally, to consider the costs of giving birth outside of the ecological birth peak, we compared infant survival for births within versus outside of the ecological birth peak. METHODS Study site and subjects The data for this study were collected between 2006 and 2014 from a population of wild geladas living in the Simien Mountains National Park, in northern Ethiopia (13°13.5′ N latitude). The Simien Mountains Gelada Research Project (SMGRP – formerly called the University of Michigan Gelada Research Project) has collected behavioral, demographic, genetic, and hormonal data from individuals since January 2006. All gelada subjects are habituated to human observers on foot and are individually recognizable. Here, we used longitudinal data from 167 adult females comprising 25 reproductive units. All adult females had known or estimated birth dates from which we calculated age (mean age = 11.4 years; range = 4.3–27.0 years). Most estimated birth dates (N = 51) were calculated by subtracting the mean female age at major life-history milestones from known dates for each milestone (e.g., maturation, first birth: Roberts et al. 2017). For a subset of older females (N = 60), date of birth was estimated based on age of oldest offspring, or number of known offspring. Reproductive events We recorded all known births (N = 341; “births dataset”) and then assigned an estimated conception date for each birth based on the mean gestation length for this population (183 days; Roberts et al. 2017; “conceptions dataset”). For many of these births, the day of birth was known within 7 days (n = 291), but in cases where it was not, the birth date was estimated for the missed observation days as the mid-point of the period from which we last saw the mother until the time we saw the mother again with the new infant (n = 50 births; mean range = 37 ± 5.3 SE days; see Beehner and Bergman 2008 for more detail on assigning birth dates). The return to cycling for postpartum females was assigned as the first day we observed sexual swellings (i.e., swollen, bead-like vesicles surrounding a patch of exposed skin on the chest and neck, Roberts et al. 2017; “return to cycling” dataset) following the period of postpartum amenorrhea. The return to cycling dataset was independently observed from the conceptions and births datasets; and not all conceptions and births are included in the return to cycling dataset (e.g., females conceiving for the first time, females that were never observed to resume postpartum sexual swellings). By contrast, the conceptions dataset is not independent from the births dataset (i.e., the conception dates are derived from the birthdates). Nevertheless, we analyzed these 2 datasets separately to understand how the climatological data that correspond to each event contribute (or not) to the timing of female reproduction. In this way, we were able to separately examine the weather surrounding conceptions from the weather surrounding births. We have no cases where we included a conception that did not result in a live birth (since, for this dataset, live birth was how we estimated the date of conception). With respect to examining the costs of birth seasonality, we selected infant survival to 2 years as our fitness measure because the mean age at weaning in this population is approximately 1.5 years (Roberts et al. 2017) and thus the 2-year mark is sure to include the majority of weaned infants in our analyses. We opted against using additional measures that overly restricted our dataset (e.g., interbirth-intervals) or did not adequately account for infant losses prior to weaning (e.g., infant survival prior to 2 years). Weather data As part of our long-term climatological monitoring, we recorded daily cumulative rainfall and maximum and minimum temperature. Seasonal patterns of green grass availability in this area are known to be positively correlated with rainfall from the previous 90 days (Jarvey et al., In press). Therefore, because we did not directly collect food abundance data (or intake rates) across the entire study period, we used the total rainfall over the 90 days prior to each reproductive event (or hormone sample) as our proxy for green grass availability (e.g., Hill et al. 2000; McFarland et al. 2014). The physiological effects of temperature on reproduction are more direct than rainfall (via changes in core body temperatures). Therefore, as our proxy for thermoregulatory constraints, we calculated the mean maximum and minimum daily temperatures for the 30 days preceding each reproductive event or hormone sample (e.g., Dunbar et al. 2002). Takeovers We recorded the dates of all observed male takeovers (n = 72) of known reproductive units. Most takeovers were recorded within days of occurrence (n = 62; range = 0–7 days). For takeovers that were not directly observed (n = 18), we were able to assign the day of takeover to within a mean of 30.7 ± 5.3 SE days of occurrence. These takeovers were assigned the mid-point of the missing observation period. Because we were primarily concerned with how takeovers might alter the timing of reproductive events (e.g., due to infanticide, the Bruce effect, or accelerated weaning: Beehner and Bergman 2008; Roberts et al. 2017), we labeled any return to cycling and conception dates that occurred during the 3 months following a takeover as takeover return to cycling/conception dates, and those that did not follow a takeover we labeled nontakeover return to cycling/conception dates. The 3-month window was chosen based on the observation that most known or suspected infanticides occur during this 3-month window (in rare cases, infanticide may occur up to 9 months after a takeover, but we wanted to reflect the immediate impact of takeovers on female reproduction in the seasonality analysis: Beehner and Bergman 2008). Similarly, we labeled all births that occurred between 6–9 months following a takeover as takeover births (this time period reflects all births that resulted from conceptions by the new dominant male during his first 3 months as leader male, plus the period of gestation, 183 days, or just over 6 months; Roberts et al. 2017). Births that did not occur between 6–9 months after a takeover were labeled nontakeover births. Hormone collection and analysis We collected fecal samples from 148 known adult females between 2006 and 2014 (N =3,841 hormone samples; mean = 26 samples per female; range: 1–150 samples per female). Fecal samples were collected using noninvasive methods developed by the SMGRP for hormone extraction and preservation under field conditions (Beehner and Whitten 2004; Beehner and McCann 2008). In brief, we mixed the full fecal samples prior to placing a small aliquot in 3 ml of a methanol:acetone solution (4:1). Samples were vortexed and later filtered and extracted using a solid-phase cartridge. All samples were washed with 2.0 ml of 0.1% sodium azide (NaN3) solution, placed in a sterile Whirl-pak bag with a silica desiccant, and stored frozen until shipment to J. Beehner’s endocrine laboratory at the University of Michigan for radioimmunoassay (RIA). Dry fecal weights from all samples were obtained to the nearest ± 0.0001 g, and hormones values were calculated as ng/g dry feces. At the University of Michigan, all samples were assayed for glucocorticoid metabolites (GCMs) using reagents from the ImmuChemTM double antibody corticosterone 125I RIA kit (MP Biomedicals, LLC, Orangeburg, NY). This antibody has been validated both analytically and biologically for use in geladas (Beehner and McCann 2008). The primary antibody in this kit cross-reacts 100% with corticosterone, 0.34% with desoxycorticosterone, 0.1% with testosterone, 0.05% with cortisol, 0.03% with aldosterone, and 0.02% with progesterone. We ran all standards, controls, and samples in duplicate. We used a low (~20% binding) a mid- (~50% binding) and a high (~80% binding) fecal pool control in all assays. The respective interassay coefficients of variation (CVs) were: low fecal pool: 20.69%; mid fecal pool: 21.43%; high fecal pool: 20.93%; high kit control: 24.03%; and low kit control: 18.85% (N = 177 assays). Our intra-assay CV for a high and low fecal pool was 6.3% (N = 10 assays) and 8.7% (N = 10 assays), respectively, with a high kit control CV of 6.3% (N = 12 assays) and a low kit control of 4.1% (N = 12 assays). Data analyses We conducted 5 sets of models (described in detail below) that correspond to each of the 5 outcome variables (GCMs, return to cycling, conceptions, births, infant deaths). Model averaging For each of the first 4 sets of models (GCMs, return to cycling, conceptions, and births), we took an information theoretic approach based on Akaike’s Information Criterion (using AIC for the GCMs model; AICc, corrected for small sample sizes, for the return to cycling, conceptions, and births models; Anderson and Burnham 2002) to mean all candidate models for each outcome variable (Table 1; Johnson and Omland 2004). Candidate models represent all combinations of the predictor variables of interest and their interactions (mean maximum temperature, mean minimum temperature, cumulative rain, and takeover, Table 1. Model averaging was done using the Mu-MIn package (version 1.15.6: Barton 2016) in R (R Core Team 2016: Version 3.3.2) to produce estimates of predictors within an averaged model. We considered predictors to have a meaningful effect on the outcome variable if the 95% confidence intervals of the averaged effect size did not overlap zero (i.e., “reliable” predictors: Table 3, though see Supplementary Table S1 for full averaging results). Glucocorticoid seasonality To assess the effects of ecological and social seasonality on GCMs, we first log-transformed GCM values to approximate a normal distribution, and then modeled logGCMs as a function of the following candidate predictors (including interaction terms; see Table 1 for descriptions): 1) mean maximum temperature, 2) mean minimum temperature, 3) cumulative rainfall, and 4) takeover (yes/no). We controlled for the repeated measures of individual identity, unit, and year as random effects, as well as the known effects of age and reproductive state on GCMs (Beehner and Bergman 2017) by including an interaction term between reproductive state (pregnant, cycling, or lactating) and age (both as a linear and as a quadratic term; see Supplementary Figure S1 for full averaged model results). We constructed 41 candidate linear mixed-effects models (including an intercept model and a model controlling for age and reproductive state alone) with the lme4 package in R (version 1.1–12: Bates et al. 2015). The candidate models represent all combinations of predictor variables and their interactions (mean maximum temperature, mean minimum temperature, cumulative rain, and takeover), and the model fits were compared using AIC and Akaike model weights (Anderson and Burnham 2002). Initiating reproduction To assess how ecological (i.e., rainfall, temperature) and social (i.e., male takeover) factors were associated with the initiation of reproduction, we considered 2 different events that could indicate the start of a reproductive event for gelada females: 1) the resumption of cycling from postpartum amenorrhea (based on observational data on each female), and 2) the date of conception for each birth (estimated backwards from each observed birth). We conducted 2 separate analyses using binomial general linear mixed models (GLMMs). We coded the reproductive events as binomial variables by month, where females either returned to cycling or not, or conceived or not. Because we were interested in population-level patterns of reproductive seasonality (since populations, and not individuals, exhibit birth peaks), we analyzed reproductive seasonality across the entire population, not at the level of the unit. Therefore, the dependent binomial variable for the first model was Return to Cycling (or the number of postpartum females in the population that had resumed cycling in a given month out of the total females in the study population that month). The dependent binomial variable for the second model was Conceptions (or the number of females in the population that had conceived in a given month out of the total females in the study population that month). We also coded these reproductive events according to whether or not the event occurred within the context of a takeover. Using conceptions as an example, in a given month we calculated how many of all of our known females conceived in that month out of the total number of known females. This variable was calculated separately for females that did not experience a takeover in the previous 3 months (nontakeover conceptions) and those that did (takeover conceptions). Therefore, each month of the study is featured twice in the dependent variable column, once for reproductive events following takeovers and once for reproductive events not following takeovers. For each outcome variable (the number of Return to Cycling and the number of Conception events), we created 40 models (including an intercept model) predicting the number of reproductive events for each month based on a set of candidate predictor variables and their interactions (Table 1). Specifically, we considered the following predictors (as well as all interactions between them): 1) mean maximum daily temperature (across the previous 30 days), 2) mean minimum daily temperatures (across the previous 30 days), 3) cumulative rainfall (across the previous 90 days), and 4) the categorical predictor of whether or not a takeover had occurred prior (yes or no). Because each month was represented twice in our dataset, all models also included month and year as random effects to control for any potential monthly differences that were unrelated to the predictor variables. Table 1 Description of outcome variables and predictors used in model selection (lme4) Outcome variable Main effects Interaction effects Random effects Number of candidate models A. Log glucocorticoid metabolites: Linear Mixed-Effects Models logGCMs Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Individual ID, Unit, Year 41 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 30 days B. Return to cycling: Binomial General Linear Mixed Models Total number of monthly postpartum return to cycling out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days C. Conceptions: Binomial General Linear Mixed Models Total number of monthly conceptions out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days D. Births: Binomial General Linear Mixed Models Total number of monthly births out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 270 days Outcome variable Main effects Interaction effects Random effects Number of candidate models A. Log glucocorticoid metabolites: Linear Mixed-Effects Models logGCMs Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Individual ID, Unit, Year 41 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 30 days B. Return to cycling: Binomial General Linear Mixed Models Total number of monthly postpartum return to cycling out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days C. Conceptions: Binomial General Linear Mixed Models Total number of monthly conceptions out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days D. Births: Binomial General Linear Mixed Models Total number of monthly births out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 270 days View Large Table 1 Description of outcome variables and predictors used in model selection (lme4) Outcome variable Main effects Interaction effects Random effects Number of candidate models A. Log glucocorticoid metabolites: Linear Mixed-Effects Models logGCMs Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Individual ID, Unit, Year 41 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 30 days B. Return to cycling: Binomial General Linear Mixed Models Total number of monthly postpartum return to cycling out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days C. Conceptions: Binomial General Linear Mixed Models Total number of monthly conceptions out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days D. Births: Binomial General Linear Mixed Models Total number of monthly births out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 270 days Outcome variable Main effects Interaction effects Random effects Number of candidate models A. Log glucocorticoid metabolites: Linear Mixed-Effects Models logGCMs Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Individual ID, Unit, Year 41 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 30 days B. Return to cycling: Binomial General Linear Mixed Models Total number of monthly postpartum return to cycling out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days C. Conceptions: Binomial General Linear Mixed Models Total number of monthly conceptions out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 90 days D. Births: Binomial General Linear Mixed Models Total number of monthly births out of total females Maximum temperature Mean maximum daily temperature across the previous 30 days Maximum temperature × minimum temperature Maximum temperature × rain Minimum temperature × rain Maximum temperature × takeover Minimum temperature × takeover Rain × takeover Month-year 40 Minimum temperature Mean minimum daily temperature across the previous 30 days Rain Cumulative precipitation across the previous 90 days Takeover (Y/N): whether a takeover occurred in the previous 270 days View Large Birth seasonality To identify fluctuations in birth rates across the year due to either ecological or social factors (e.g., birth peaks and/or birth valleys), we first characterized the degree of seasonality in births qualitatively. We used cut-offs to distinguish between moderate seasonality (where 33–67% of births occur in a 3-month period), strict seasonality (where more than 67% of births cluster in a 3-month period), or no seasonality (where births occur all year with no clear peak; see van Schaik et al. 1999). We also characterized the degree to which these patterns changed (i.e., became more or less seasonal, or shifted in timing) when we considered takeover versus nontakeover births separately. Second, we constructed a third set of binomial GLMMs (with the lme4 package version 1.1–12: Bates et al. 2015) to assess when females were most likely to give birth (Table 1). For the births model, the dependent binomial variable was the number of females in the population that gave birth in a given month out of the total females in the study population that month. As with the cycling and conceptions models, we calculated this outcome variable as a function of whether or not each female had experienced a takeover, but rather than a 3-month window, this time we considered the 6–9 months prior to the birth (to identify births resulting from conceptions in the 3 months following a takeover plus the ~6 months of gestation). Again, we created 40 models (including an intercept model) predicting the number of births each month based on the same set of candidate predictor variables and their interactions used in the Return to Cycling and Conception models, reflecting hypotheses about which ecological and social variables influenced birth timing (Table 1). Specifically, we considered the following predictors (as well as all interactions between them): 1) mean maximum daily temperature (across the previous 30 days), 2) mean minimum daily temperatures (across the previous 30 days), 3) cumulative rainfall (across the previous 90 days), and 4) the categorical predictor of whether or not a takeover had occurred prior (yes or no). Again, because each month was represented twice in our dataset (once for births following takeovers, and once for births that did not follow takeovers), all models also included month and year as random effects to control for any potential monthly differences that were unrelated to the predictor variables. Costs to birth timing To assess whether there was a cost to giving birth at a certain time of the year, we considered the effect of birth timing on survival to 2 years of age (extending beyond the mean age of weaning in this population, 1.5 years of age, to be sure to include all weaned infants; Roberts et al. 2017). Specifically, because we identified a birth peak (see Birth seasonality in Results), we were interested in the potential costs of giving birth outside of this peak (which we defined as a 3-month period where 33–67% of births occurred: van Schaik et al. 1999). We constructed a series of binomial GLMMs to assess whether being born during the birth peak predicted infant survival to 2 years of age (yes/no). For each model, our predictor was whether the infant had been born “in-peak” or “off-peak” according to the seasonality analysis described above. First, we considered births and deaths for all infants until they reached 2 years of age (births = 306; deaths = 69; note that this dataset is slightly reduced from the one used in the seasonality analysis. To assess survival to 2 years of age we could only include births prior to 2014). Second, to focus solely on infant deaths due to “ecological reasons” we removed all suspected infanticide deaths (n = 33) and constructed another binomial GLMM with births (n = 273) and noninfanticide deaths (n = 36). Third, because takeovers force females to shift their reproductive cycles (see Birth seasonality in Results), we conducted a third binomial GLMM removing all births following takeovers to determine whether there was a cost to nonpeak births outside of the potential influence of takeovers (nontakeover births: n = 195; noninfanticide deaths: n = 26). For each model, we controlled for the repeated effects of birth year and the identity of the mother. RESULTS Ecological and social seasonality The climate in the Simien Mountains National Park can be broadly divided into 3 distinct seasons: a cold-dry season, a hot-dry season, and a cold-wet season (Figure 1). The cold-dry season typically occurred between October to January and featured the lowest minimum temperatures and very little rainfall (mean daily minimum temperature = 7.02 °C ± 1.39 SD; mean daily maximum temperature = 16.46 °C ± 1.64 SD; mean daily precipitation = 1.25 mm ± 4.61 SD). The hot-dry season typically occurred between February to May, and featured the warmest temperatures seen throughout the year with low (but variable) levels of rainfall (daily minimum temperature = 9.03 °C ± 1.39 SD; mean daily maximum temperature = 20.59 °C ± 2.19 SD; mean daily precipitation = 2.21 mm ± 6.64 SD). Finally, the cold-wet season typically occurred between June to September and featured the lowest maximum temperatures with the highest levels of daily precipitation (mean daily minimum temperature = 8.34 °C ± 1.38 SD; mean daily maximum temperature = 15.58 °C ± 2.42 SD; mean daily precipitation = 13.03 mm ± 17.35 SD). Typically, peak rainfall occurred between June to August, which corresponds with a peak in green grass availability between October to November (Figure 2a). Figure 1 View largeDownload slide Seasonality of ecological (temperature and rainfall) and social (takeovers) predictors. Mean maximum and minimum daily temperatures by month ± 95% confidence intervals (maximum = downward triangles; minimum = upward triangles; solid line); mean total rainfall by month (grey squares; dashed line) ± 95% confidence intervals; total number of takeovers observed by month (white bars). Background colors indicate season: orange = hot-dry season; light blue = cold-wet season; pale yellow = cold-dry season. Figure 1 View largeDownload slide Seasonality of ecological (temperature and rainfall) and social (takeovers) predictors. Mean maximum and minimum daily temperatures by month ± 95% confidence intervals (maximum = downward triangles; minimum = upward triangles; solid line); mean total rainfall by month (grey squares; dashed line) ± 95% confidence intervals; total number of takeovers observed by month (white bars). Background colors indicate season: orange = hot-dry season; light blue = cold-wet season; pale yellow = cold-dry season. Figure 2 View largeDownload slide (a) Cumulative seasonality predictors in relation to peak green grass availability, indicated by the shaded region (July to November). All models included the following variables as predictors: mean maximum (downward triangles; solid line) and minimum (upward triangles; solid line) daily temperature across the previous 30 days as a proxy for thermoregulatory requirements and cumulative rainfall (grey squares; dashed line) across the previous 90 days (as a proxy for green grass availability). (b) Seasonal patterns of mean residual log-transformed glucocorticoid metabolites (logGCMs) ± 95% confidence intervals (after controlling for the effects of reproductive state, age, and experiencing a takeover 30 days prior to sample collection). Warm and cold months were determined by taking the average of the mean maximum and mean minimum daily temperatures over the previous 30 days for each month (see Figure 2a; note that these categories were used for visualization purposes only). Background shades indicate the 4 warmest months (March to June) and the 4 coldest months (August to September; December to January). (c) Mean residual logGCMs ± standard error (after controlling for the effects of reproductive state, age, and temperature) for females that did not experience a takeover in the previous 30 days (gray box) vs. females that did (white box). (d) Mean residual logGCMs ± standard error (after controlling for the effects of reproductive state, age, and takeover) for females during the 4 warmest months (March to June; shaded box labeled "HOT" in Figure 2b) and during the 4 coldest months (August to September and December to January; shaded boxes labeled "COLD" in Figure 2b). For both (c) and (d), mean = solid line; standard error = box outline; 95% confidence intervals = whiskers. Figure 2 View largeDownload slide (a) Cumulative seasonality predictors in relation to peak green grass availability, indicated by the shaded region (July to November). All models included the following variables as predictors: mean maximum (downward triangles; solid line) and minimum (upward triangles; solid line) daily temperature across the previous 30 days as a proxy for thermoregulatory requirements and cumulative rainfall (grey squares; dashed line) across the previous 90 days (as a proxy for green grass availability). (b) Seasonal patterns of mean residual log-transformed glucocorticoid metabolites (logGCMs) ± 95% confidence intervals (after controlling for the effects of reproductive state, age, and experiencing a takeover 30 days prior to sample collection). Warm and cold months were determined by taking the average of the mean maximum and mean minimum daily temperatures over the previous 30 days for each month (see Figure 2a; note that these categories were used for visualization purposes only). Background shades indicate the 4 warmest months (March to June) and the 4 coldest months (August to September; December to January). (c) Mean residual logGCMs ± standard error (after controlling for the effects of reproductive state, age, and temperature) for females that did not experience a takeover in the previous 30 days (gray box) vs. females that did (white box). (d) Mean residual logGCMs ± standard error (after controlling for the effects of reproductive state, age, and takeover) for females during the 4 warmest months (March to June; shaded box labeled "HOT" in Figure 2b) and during the 4 coldest months (August to September and December to January; shaded boxes labeled "COLD" in Figure 2b). For both (c) and (d), mean = solid line; standard error = box outline; 95% confidence intervals = whiskers. In total, we observed 72 takeovers during the study period, which demonstrated a moderately seasonal pattern (a subset of these data were already analyzed and reported in Pappano and Beehner 2014): 45.8% of all takeovers occurred during a 3-month period (February to March, n = 33, Figure 1), and the highest number of takeovers out of all months occurred in March (n = 16). Seasonality of glucocorticoid metabolites Log glucocorticoid metabolites (logGCMs) were highly seasonal: high temperatures predicted low logGCMs (Figure 2b,d). LogGCMs decreased during the hot-dry season, reaching a nadir from April to July, and increased in the colder months, peaking twice: September to October and December to January (Figure 2b). Mean maximum and minimum temperatures were included as predictors in the top 4 models, which together contributes 90% of model weight and performed substantially better than the intercept only null model (Table 2). In the averaged model, mean maximum temperature (estimate = −0.0378, SE = 0.119, z-value = 3.179; Table 3) and mean minimum temperature (estimate = −0.0368, SE = 0.008, z-value = 4.563; Table 3) were strong and reliable negative predictors of logGCMs (i.e., 95% confidence intervals of mean estimates did not overlap zero, Figure 3a). This effect size corresponds to a 1.4% or a 2.9% decrease in GCMs for every 1 °C increase in mean maximum or mean minimum temperature, respectively (see Figure 2d for a comparison between mean residual GCMs in hot versus cold months, categories which were used for visualization purposes only). Rainfall, on the other hand, was not a strong or reliable predictor of logGCMs in the averaged model (estimate = −0.0075, SE = 0.011, z-value = 0.660; Table 3), although it was included as a predictor in one of the top 4 models (Table 2). Together, these results suggest that females are cold-stressed, and that thermoregulation rather than green grass availability may limit female energy balance. Seasonality of postpartum return to cycling We identified a seasonal peak in when females returned to cycling following postpartum amenorrhea, which mirrored the seasonality of logGCMs: 51.7% of all nontakeover cycling events occurred within a 3-month period during the hot-dry season (76 out of 147 total, March to May; Figure 4a: Return to cycling). Mean maximum temperature was included as a predictor in the top model (which comprised 19.1% of the model weight, Table 2), and was highly reliable based on 95% confidence intervals in the overall averaged model (Figure 3b). Mean maximum temperature significantly predicted the number of postpartum returns to cycling observed (estimate = 0.641, SE = 0.202, z-value = 3.175; Table 3), corresponding to a 35.6% increase in the number of females that returned to cycling for every 1 °C increase in maximum temperature. Table 2 Akaike’s information criterion (AIC/AICc) model comparison results, showing the model components for the null model and the top models for each analysis (where ∆AIC or ∆AICc ≤ 6) A.Log glucocorticoid metabolites Degrees of freedom Log-likelihood ∆AIC Weight Maximum temperature, minimum temperature, takeover 16 −487.37 0.00 0.47 Maximum temperature, minimum temperature, rain, takeover 17 −486.85 0.96 0.29 Minimum temperature, takeover, maximum temperature × rain 18 −486.79 2.83 0.11 Maximum temperature, minimum temperature 15 −490.67 4.60 0.05 Intercept (null model) 5 −638.97 281.19 0.00 B.Return to cycling Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, takeover 4 −248.893 0.00 0.191 Maximum temperature, minimum temperature × takeover 6 −246.831 0.14 0.179 Maximum temperature × takeover 5 −248.573 1.48 0.091 Maximum temperature, rain × takeover 6 −247.708 1.89 0.074 Maximum temperature, rain, takeover 5 −248.834 2.00 0.070 Maximum temperature, minimum temperature, takeover 5 −248.883 2.10 0.067 Maximum temperature, rain, minimum temperature × takeover 7 −246.791 2.22 0.063 Takeover, maximum temperature × rain 6 −247.916 2.31 0.060 Rain, maximum temperature × takeover 6 −248.532 3.54 0.033 Minimum temperature, maximum temperature × takeover 6 −248.569 3.61 0.031 Takeover, maximum temperature × minimum temperature 6 −248.612 3.70 0.030 Maximum temperature, minimum temperature, rain × takeover 7 −247.708 4.06 0.025 Intercept (null model) 2 −280.993 60.04 0.000 C.Conceptions Degrees of freedom Log-likelihood ∆AICc Weight Rain × takeover 5 −282.827 0.00 0.470 Minimum temperature, rain × takeover 6 −282.680 1.85 0.186 Maximum temperature, rain × takeover 6 −282.827 2.14 0.161 Maximum temperature, minimum temperature, rain × takeover 7 −284.500 3.61 0.077 Intercept (null model) 2 −304.483 37.03 0.000 D.Births Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, minimum temperature, rain × takeover 7 −289.742 0.00 0.851 Minimum temperature, rain × takeover 6 −292.897 4.14 0.108 Intercept (null model) 2 −328.858 67.62 0.000 A.Log glucocorticoid metabolites Degrees of freedom Log-likelihood ∆AIC Weight Maximum temperature, minimum temperature, takeover 16 −487.37 0.00 0.47 Maximum temperature, minimum temperature, rain, takeover 17 −486.85 0.96 0.29 Minimum temperature, takeover, maximum temperature × rain 18 −486.79 2.83 0.11 Maximum temperature, minimum temperature 15 −490.67 4.60 0.05 Intercept (null model) 5 −638.97 281.19 0.00 B.Return to cycling Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, takeover 4 −248.893 0.00 0.191 Maximum temperature, minimum temperature × takeover 6 −246.831 0.14 0.179 Maximum temperature × takeover 5 −248.573 1.48 0.091 Maximum temperature, rain × takeover 6 −247.708 1.89 0.074 Maximum temperature, rain, takeover 5 −248.834 2.00 0.070 Maximum temperature, minimum temperature, takeover 5 −248.883 2.10 0.067 Maximum temperature, rain, minimum temperature × takeover 7 −246.791 2.22 0.063 Takeover, maximum temperature × rain 6 −247.916 2.31 0.060 Rain, maximum temperature × takeover 6 −248.532 3.54 0.033 Minimum temperature, maximum temperature × takeover 6 −248.569 3.61 0.031 Takeover, maximum temperature × minimum temperature 6 −248.612 3.70 0.030 Maximum temperature, minimum temperature, rain × takeover 7 −247.708 4.06 0.025 Intercept (null model) 2 −280.993 60.04 0.000 C.Conceptions Degrees of freedom Log-likelihood ∆AICc Weight Rain × takeover 5 −282.827 0.00 0.470 Minimum temperature, rain × takeover 6 −282.680 1.85 0.186 Maximum temperature, rain × takeover 6 −282.827 2.14 0.161 Maximum temperature, minimum temperature, rain × takeover 7 −284.500 3.61 0.077 Intercept (null model) 2 −304.483 37.03 0.000 D.Births Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, minimum temperature, rain × takeover 7 −289.742 0.00 0.851 Minimum temperature, rain × takeover 6 −292.897 4.14 0.108 Intercept (null model) 2 −328.858 67.62 0.000 View Large Table 2 Akaike’s information criterion (AIC/AICc) model comparison results, showing the model components for the null model and the top models for each analysis (where ∆AIC or ∆AICc ≤ 6) A.Log glucocorticoid metabolites Degrees of freedom Log-likelihood ∆AIC Weight Maximum temperature, minimum temperature, takeover 16 −487.37 0.00 0.47 Maximum temperature, minimum temperature, rain, takeover 17 −486.85 0.96 0.29 Minimum temperature, takeover, maximum temperature × rain 18 −486.79 2.83 0.11 Maximum temperature, minimum temperature 15 −490.67 4.60 0.05 Intercept (null model) 5 −638.97 281.19 0.00 B.Return to cycling Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, takeover 4 −248.893 0.00 0.191 Maximum temperature, minimum temperature × takeover 6 −246.831 0.14 0.179 Maximum temperature × takeover 5 −248.573 1.48 0.091 Maximum temperature, rain × takeover 6 −247.708 1.89 0.074 Maximum temperature, rain, takeover 5 −248.834 2.00 0.070 Maximum temperature, minimum temperature, takeover 5 −248.883 2.10 0.067 Maximum temperature, rain, minimum temperature × takeover 7 −246.791 2.22 0.063 Takeover, maximum temperature × rain 6 −247.916 2.31 0.060 Rain, maximum temperature × takeover 6 −248.532 3.54 0.033 Minimum temperature, maximum temperature × takeover 6 −248.569 3.61 0.031 Takeover, maximum temperature × minimum temperature 6 −248.612 3.70 0.030 Maximum temperature, minimum temperature, rain × takeover 7 −247.708 4.06 0.025 Intercept (null model) 2 −280.993 60.04 0.000 C.Conceptions Degrees of freedom Log-likelihood ∆AICc Weight Rain × takeover 5 −282.827 0.00 0.470 Minimum temperature, rain × takeover 6 −282.680 1.85 0.186 Maximum temperature, rain × takeover 6 −282.827 2.14 0.161 Maximum temperature, minimum temperature, rain × takeover 7 −284.500 3.61 0.077 Intercept (null model) 2 −304.483 37.03 0.000 D.Births Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, minimum temperature, rain × takeover 7 −289.742 0.00 0.851 Minimum temperature, rain × takeover 6 −292.897 4.14 0.108 Intercept (null model) 2 −328.858 67.62 0.000 A.Log glucocorticoid metabolites Degrees of freedom Log-likelihood ∆AIC Weight Maximum temperature, minimum temperature, takeover 16 −487.37 0.00 0.47 Maximum temperature, minimum temperature, rain, takeover 17 −486.85 0.96 0.29 Minimum temperature, takeover, maximum temperature × rain 18 −486.79 2.83 0.11 Maximum temperature, minimum temperature 15 −490.67 4.60 0.05 Intercept (null model) 5 −638.97 281.19 0.00 B.Return to cycling Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, takeover 4 −248.893 0.00 0.191 Maximum temperature, minimum temperature × takeover 6 −246.831 0.14 0.179 Maximum temperature × takeover 5 −248.573 1.48 0.091 Maximum temperature, rain × takeover 6 −247.708 1.89 0.074 Maximum temperature, rain, takeover 5 −248.834 2.00 0.070 Maximum temperature, minimum temperature, takeover 5 −248.883 2.10 0.067 Maximum temperature, rain, minimum temperature × takeover 7 −246.791 2.22 0.063 Takeover, maximum temperature × rain 6 −247.916 2.31 0.060 Rain, maximum temperature × takeover 6 −248.532 3.54 0.033 Minimum temperature, maximum temperature × takeover 6 −248.569 3.61 0.031 Takeover, maximum temperature × minimum temperature 6 −248.612 3.70 0.030 Maximum temperature, minimum temperature, rain × takeover 7 −247.708 4.06 0.025 Intercept (null model) 2 −280.993 60.04 0.000 C.Conceptions Degrees of freedom Log-likelihood ∆AICc Weight Rain × takeover 5 −282.827 0.00 0.470 Minimum temperature, rain × takeover 6 −282.680 1.85 0.186 Maximum temperature, rain × takeover 6 −282.827 2.14 0.161 Maximum temperature, minimum temperature, rain × takeover 7 −284.500 3.61 0.077 Intercept (null model) 2 −304.483 37.03 0.000 D.Births Degrees of freedom Log-likelihood ∆AICc Weight Maximum temperature, minimum temperature, rain × takeover 7 −289.742 0.00 0.851 Minimum temperature, rain × takeover 6 −292.897 4.14 0.108 Intercept (null model) 2 −328.858 67.62 0.000 View Large Table 3 Model-averaging results for significant predictors of interest (full model-averaging results appear in Supplementary Table S1) Predictors Importance Number of models 2.5% CI 97.5% CI Estimate Adjusted standard error z value P-value A. Log glucocorticoid metabolites Maximum temperature 1.00 26 −0.0611 −0.0145 −0.0378 0.0119 3.179 **0.0015 Minimum temperature 1.00 26 −0.0526 −0.0210 −0.0368 0.0081 4.563 ***5.0 × 10−06 Takeover (Y) 0.93 26 0.0131 0.1066 0.0598 0.0238 2.509 **0.0121 B. Return to cycling Takeover (Y) 1.00 26 0.7490 1.5406 1.1450 0.2019 5.669 ***<2.0 × 10−16 Maximum temperature 0.99 26 0.2454 1.0371 0.6412 0.2020 3.175 **0.0015 C. Conceptions Rain 0.98 26 −0.6158 −0.1363 −0.3761 0.1223 3.074 **0.00211 Takeover (Y) 1.00 26 0.4935 1.0706 0.7821 0.1472 5.312 ***<1.0 × 10−7 Rain × takeover (Y) 0.89 4 0.0203 0.7936 0.4070 0.1973 2.063 *0.03911 D. Births Takeover (Y) 1.00 26 0.7171 1.3147 1.0160 0.1525 6.663 ***<2.0 × 10−16 Rain 1.00 26 0.2131 0.7366 0.4749 0.1336 3.556 ***0.0004 Rain × takeover (Y) 1.00 4 −0.9984 −0.3002 −0.6493 0.1781 3.646 ***0.0003 Minimum temperature 0.96 26 −0.5716 −0.0540 −0.3128 0.1320 2.369 *0.0178 Predictors Importance Number of models 2.5% CI 97.5% CI Estimate Adjusted standard error z value P-value A. Log glucocorticoid metabolites Maximum temperature 1.00 26 −0.0611 −0.0145 −0.0378 0.0119 3.179 **0.0015 Minimum temperature 1.00 26 −0.0526 −0.0210 −0.0368 0.0081 4.563 ***5.0 × 10−06 Takeover (Y) 0.93 26 0.0131 0.1066 0.0598 0.0238 2.509 **0.0121 B. Return to cycling Takeover (Y) 1.00 26 0.7490 1.5406 1.1450 0.2019 5.669 ***<2.0 × 10−16 Maximum temperature 0.99 26 0.2454 1.0371 0.6412 0.2020 3.175 **0.0015 C. Conceptions Rain 0.98 26 −0.6158 −0.1363 −0.3761 0.1223 3.074 **0.00211 Takeover (Y) 1.00 26 0.4935 1.0706 0.7821 0.1472 5.312 ***<1.0 × 10−7 Rain × takeover (Y) 0.89 4 0.0203 0.7936 0.4070 0.1973 2.063 *0.03911 D. Births Takeover (Y) 1.00 26 0.7171 1.3147 1.0160 0.1525 6.663 ***<2.0 × 10−16 Rain 1.00 26 0.2131 0.7366 0.4749 0.1336 3.556 ***0.0004 Rain × takeover (Y) 1.00 4 −0.9984 −0.3002 −0.6493 0.1781 3.646 ***0.0003 Minimum temperature 0.96 26 −0.5716 −0.0540 −0.3128 0.1320 2.369 *0.0178 View Large Table 3 Model-averaging results for significant predictors of interest (full model-averaging results appear in Supplementary Table S1) Predictors Importance Number of models 2.5% CI 97.5% CI Estimate Adjusted standard error z value P-value A. Log glucocorticoid metabolites Maximum temperature 1.00 26 −0.0611 −0.0145 −0.0378 0.0119 3.179 **0.0015 Minimum temperature 1.00 26 −0.0526 −0.0210 −0.0368 0.0081 4.563 ***5.0 × 10−06 Takeover (Y) 0.93 26 0.0131 0.1066 0.0598 0.0238 2.509 **0.0121 B. Return to cycling Takeover (Y) 1.00 26 0.7490 1.5406 1.1450 0.2019 5.669 ***<2.0 × 10−16 Maximum temperature 0.99 26 0.2454 1.0371 0.6412 0.2020 3.175 **0.0015 C. Conceptions Rain 0.98 26 −0.6158 −0.1363 −0.3761 0.1223 3.074 **0.00211 Takeover (Y) 1.00 26 0.4935 1.0706 0.7821 0.1472 5.312 ***<1.0 × 10−7 Rain × takeover (Y) 0.89 4 0.0203 0.7936 0.4070 0.1973 2.063 *0.03911 D. Births Takeover (Y) 1.00 26 0.7171 1.3147 1.0160 0.1525 6.663 ***<2.0 × 10−16 Rain 1.00 26 0.2131 0.7366 0.4749 0.1336 3.556 ***0.0004 Rain × takeover (Y) 1.00 4 −0.9984 −0.3002 −0.6493 0.1781 3.646 ***0.0003 Minimum temperature 0.96 26 −0.5716 −0.0540 −0.3128 0.1320 2.369 *0.0178 Predictors Importance Number of models 2.5% CI 97.5% CI Estimate Adjusted standard error z value P-value A. Log glucocorticoid metabolites Maximum temperature 1.00 26 −0.0611 −0.0145 −0.0378 0.0119 3.179 **0.0015 Minimum temperature 1.00 26 −0.0526 −0.0210 −0.0368 0.0081 4.563 ***5.0 × 10−06 Takeover (Y) 0.93 26 0.0131 0.1066 0.0598 0.0238 2.509 **0.0121 B. Return to cycling Takeover (Y) 1.00 26 0.7490 1.5406 1.1450 0.2019 5.669 ***<2.0 × 10−16 Maximum temperature 0.99 26 0.2454 1.0371 0.6412 0.2020 3.175 **0.0015 C. Conceptions Rain 0.98 26 −0.6158 −0.1363 −0.3761 0.1223 3.074 **0.00211 Takeover (Y) 1.00 26 0.4935 1.0706 0.7821 0.1472 5.312 ***<1.0 × 10−7 Rain × takeover (Y) 0.89 4 0.0203 0.7936 0.4070 0.1973 2.063 *0.03911 D. Births Takeover (Y) 1.00 26 0.7171 1.3147 1.0160 0.1525 6.663 ***<2.0 × 10−16 Rain 1.00 26 0.2131 0.7366 0.4749 0.1336 3.556 ***0.0004 Rain × takeover (Y) 1.00 4 −0.9984 −0.3002 −0.6493 0.1781 3.646 ***0.0003 Minimum temperature 0.96 26 −0.5716 −0.0540 −0.3128 0.1320 2.369 *0.0178 View Large Figure 3 View largeDownload slide Model-averaged coefficient estimates ± 95% confidence intervals for averaged models for each outcome variable: (a) log glucocorticoid metabolites (here showing only predictors of interest; each model also controlled for reproductive status and age-squared: see Supplementary Figure S1), (b) return to cycling, (c) conceptions, and (d) births. Coefficient estimates less than zero result in a lower than expected outcome while those greater than zero result in a higher than expected outcome. Note that these plots depict the direction and reliability of each estimate (i.e., predictors are considered reliable when the 95% confidence intervals do not cross zero), not the relative effect size. Predictors are displayed in descending order of relative importance assigned by the MuMIn comparison. Figure 3 View largeDownload slide Model-averaged coefficient estimates ± 95% confidence intervals for averaged models for each outcome variable: (a) log glucocorticoid metabolites (here showing only predictors of interest; each model also controlled for reproductive status and age-squared: see Supplementary Figure S1), (b) return to cycling, (c) conceptions, and (d) births. Coefficient estimates less than zero result in a lower than expected outcome while those greater than zero result in a higher than expected outcome. Note that these plots depict the direction and reliability of each estimate (i.e., predictors are considered reliable when the 95% confidence intervals do not cross zero), not the relative effect size. Predictors are displayed in descending order of relative importance assigned by the MuMIn comparison. Figure 4 View largeDownload slide Seasonal patterns of reproductive events. (a) No takeover events: postpartum return to cycling where a takeover did not occur within the previous 3 months; conceptions where a takeover did not occur within the previous 3 months; and births where a takeover did not occur within the previous 6–9 months. (b) Reproductive events following takeovers: post-partum return to cycling observed within 3 months of a takeover; conceptions where a takeover did occur within the previous 3 months; and births where a takeover did occur within the previous 6–9 months. In all panels: left axis = total observed (bars); right axis = mean rates ± standard error (i.e., the number of reproductive events per female per month; black squares). Background shades indicate season: hot-dry (February to May); cold-wet (June to September); cold-dry (October to January). Figure 4 View largeDownload slide Seasonal patterns of reproductive events. (a) No takeover events: postpartum return to cycling where a takeover did not occur within the previous 3 months; conceptions where a takeover did not occur within the previous 3 months; and births where a takeover did not occur within the previous 6–9 months. (b) Reproductive events following takeovers: post-partum return to cycling observed within 3 months of a takeover; conceptions where a takeover did occur within the previous 3 months; and births where a takeover did occur within the previous 6–9 months. In all panels: left axis = total observed (bars); right axis = mean rates ± standard error (i.e., the number of reproductive events per female per month; black squares). Background shades indicate season: hot-dry (February to May); cold-wet (June to September); cold-dry (October to January). Seasonality of conceptions The number of conceptions also exhibited a moderate peak at the beginning of the hot-dry season, with 35.9% of nontakeover conceptions occurring between February and April (89 out of 248, Figure 4a: Conceptions). Rainfall was included in the top 3 models based on AICc comparison, which all performed better than the intercept only null model and together contributes over 90% of the model weight (ΔAICc = 37.03, Table 2). In our averaged model, rainfall (estimate = −0.3761, SE = 0.1223, z-value = 3.074; Figure 3c; Table 3) reliably predicted fewer conceptions, and every additional 1 mm of rain corresponded to a 0.07% decrease in the number of conceptions. Effect of takeovers on reproductive seasonality Females that experienced a takeover demonstrated 6.2% higher GCMs than females that did not experience a takeover (estimate = 0.0598, SE = 0.0238, z-value = 2.509, Figure 2c; Table 3). Takeovers also resulted in significantly different seasonal patterns for both postpartum return to cycling and conceptions (Figure 4b). After takeovers, the seasonal pattern of returns to cycling was more distinct than that observed outside of takeovers, with 57.6% of all takeover cycling events occurring within the same 3-month period (34 out of 59 total, March to May, Figure 4b: Return to cycling). However, females that experienced a takeover were also significantly more likely to return to cycling than females that did not experience a takeover, regardless of the time of year (estimate = 1.145, SE = 0.2019, z-value = 5.669; Table 3). This translated to a more than 200% increase in the number of females that returned to cycling post-takeover compared to the number of nontakeover females that returned to cycling. Females that experienced a takeover conceived later in the year than females that did not experience a takeover, and takeover conceptions demonstrated a clear peak at the start of the wet season, from June to August (52.8% of all takeover conceptions, 38 out of 72 total, Figure 4b: Conceptions). Takeovers also had a reliably positive effect on the number of conceptions that occurred in the following 3 months, regardless of the season (estimate = 0.7821, SE = 0.1472, z-value = 5.312, Figure 3c). Females that experienced a takeover were 118.6% more likely to conceive than females that did not experience a takeover. Furthermore, the interaction between rainfall and takeovers was highly reliable (estimate = 0.407, SE = 0.1973, z-value = 2.063, Figure 3c), reflecting a shift in conceptions by 4 months (from March to July, Figure 4a: Conceptions) if a takeover had occurred. In other words, if a female experienced a takeover, she was 50.2% more likely to conceive during the cold-wet season, when temperatures were at their lowest, than during the hot-dry season. Birth seasonality We found 2 distinct birth patterns depending on whether or not births followed a takeover. First, while female geladas gave birth throughout the year (mean monthly birth rate = 0.03 births per female ± 0.03 SD), 37.4% of all nontakeover births occurred between August and October (102 out of 273 total, mean birth rate = 0.05 births per female ± 0.03 SD, Figure 4a: Births). This nontakeover peak was tightly coupled with seasonal patterns in both rainfall and mean minimum temperature, and both were included as predictors in the top 2 birth models (which together represent over 90% of the model weight, Table 2). Rainfall was the strongest predictor of the number of births (estimate = 0.4739, SE = 0.1336, z-value = 3.556, Table 3; Figure 3d): an increase of 1mm of rainfall corresponded to a 0.1% increase in births. In contrast, mean minimum temperature was negatively associated with births (estimate = −0.3128, SE = 0.132, z-value = 2.369, Table 3; Figure 3d); an increase of 1 °C corresponded to a 35.7% decrease in births. Nontakeover births peaked at the end of the cold-wet season, which also corresponded with peak food availability (Figure 4a: Births). On the other hand, takeover births peaked later in the year: 54.4% of all takeover births occurred between December and February, during the cold-dry season (37 out of 68 total, mean birth rate = 0.14 births per female ± 0.09 SD, Figure 4b: Births). In our averaged model, takeovers positively predicted births (estimate = 1.016, SE = 0.1525, z-value = 6.663, Table 3), and experiencing a takeover resulted in 176.2% more births 6–9 months later than observed for those females that did not experience a takeover. Because takeovers themselves followed a seasonal pattern, we observed a “birth valley” between April and July, immediately following the takeover season, when only 7.4% of takeover births occurred (5 out of 68, Figure 4b: Births). However, birth rates recovered sharply, and during the takeover birth peak, birth rates were double those observed during the nontakeover birth peak (e.g., the mean takeover birth rate in January = 0.13 births per female ± 0.07 SD, while the mean nontakeover birth rate in September = 0.06 births per female ± 0.03 SD). This increase likely reflects the synchronizing effect that takeovers had on the reproductive timing of females within the same unit. Finally, the interaction between takeovers and rain was a reliable negative predictor of the number of births (estimate = −0.6493, SE = 0.1781, z-value = 3.646, Table 3; Figure 3d), reflecting the shift in the takeover birth peak to the cold-dry season, which is when we see minimal rainfall. In other words, the takeover birth peak is characterized by a 47.8% increase in births during the driest months of the year. Costs to birth timing We found no evidence that there were costs associated with giving birth outside the birth peak. Of the 306 births observed (from our reduced dataset of infants born prior to 2014), 69 of the infants subsequently died before reaching 2 years of age (rate = 0.225 infant deaths per birth). Approximately half of these deaths were attributed to infanticide and half were attributed to other factors (n = 33 possible infanticide deaths; n = 36 noninfanticide deaths; Supplementary Figure S2). Being born during the birth peak (August to October) versus outside of the birth peak (November to July) did not significantly predict whether an infant died before reaching 2 years of age (in-peak: estimate = −0.021, SE = 0.309, z-value = −0.069). When we removed infanticide deaths from the analysis, we still found no effect of being born in-peak on infant survival (in-peak: estimate = −0.405, SE = 0.470, z-value = −0.862). Finally, when we completely removed the potential influence of takeover (i.e., births that followed a takeover and deaths due to infanticide), we still found no effect of being born in-peak (in-peak: estimate = 0.021, SE = 0.966, z-value = 0.022). DISCUSSION Our results highlight the effect that male takeovers can have on the timing of female reproduction: in addition to the direct effects from infanticide (reviewed in Hrdy 1979; Fedigan 2003; Palombit 2015) and from the Bruce effect and/or accelerated weaning (reviewed in Smuts and Smuts 1993), we demonstrated that takeovers can disrupt reproductive seasonality within a population across years. Specifically, we identified 2 distinct seasonal birth patterns: one that we believe is shaped by energetic constraints related to thermoregulation (“ecological birth peak”) and one shaped by the timing of male takeovers (“social birth peak”). These results demonstrate the separate effects that ecological and social variables can have on seasonal patterns of female reproduction. Although female geladas are not seasonal breeders, births follow a moderately seasonal pattern (as defined by van Schaik et al. 1999), with 37.4% of all nontakeover births occurring at the end of the cold-rainy season (August to October). This ecological birth peak corresponds with a peak in conceptions that occurs in the middle of the hot-dry season (February to April). The hot-dry season is also when we observed the lowest concentrations of glucocorticoid metabolites and a peak in the number of females that returned to cycling. The seasonal variation in female glucocorticoid metabolites mirrors the pattern previously identified in cold-stressed males (Beehner and McCann 2008). These data indicate that, for some females in this population, thermoregulatory requirements due to cold temperatures may be a significant barrier to the onset of reproduction. Although food scarcity has been associated with poor energetic condition and reproductive performance in a number of species (reviewed in Di Bitetti and Janson 2000 and Brockman and van Schaik 2005; though see: Weingrill et al. 2004), we were unable to detect a relationship between the initiation of reproduction and the availability of green grass—the staple food source of geladas (Jarvey et al., In press). However, we acknowledge that green grass availability may be a poor proxy for energy intake in this population. Although the hot-dry season is when the availability of green grass is at its lowest, we also know that geladas readily switch to fallback foods during this time (e.g., underground storage organs and tubers: Jarvey et al., In press), which may provide sufficient energy to offset any additional costs associated with foraging for them (Hunter 2001). Furthermore, although females are able to give birth at other points in the year, the ecological birth peak aligns with the peak in green grass availability (ecological birth peak = August to October; peak green grass = July to November), suggesting that females that give birth at this time may be able to maintain better body condition throughout the energetically-costly period of lactation. Therefore, although our data suggest that the constraints of temperature are more significant than the constraints of green grass availability, it remains likely that food availability may still shape gelada reproductive patterns. Addressing this important question will require a cross-year assessment that includes periods of intense scarcity (e.g., in yellow baboons, Papio cynocephalus: Lea et al. 2015). By contrast, the birth pattern following takeovers was characterized by a decrease in the number of births during the months immediately following a takeover (e.g., a birth valley: Lancaster and Lee 1965), followed by a birth peak 6–9 months after the “takeover season.” This social birth peak differed from the ecological birth peak in 2 ways. First, it occurred later in the year than the ecological birth peak (in the cold-dry season instead of the cold-rainy season). Second, it resulted in a significantly higher birth rate than that recorded for the ecological birth peak, indicating that births following a takeover were more synchronized than births solely tracking fluctuations in temperature and/or rainfall. From an evolutionary perspective, reproductive synchrony among females is hypothesized to be a counterstrategy to male reproductive tactics such as coercion, male monopolization, or infanticide. Some of the benefits of synchronous mating may include, for example, a higher degree of female mate choice (Ostner et al. 2008; Roberts et al. 2014), higher rates of extrapair fertilizations (Stutchbury and Morton 1995), and a reduced risk of harassment and/or infanticide (Boness et al. 1995; Gilchrist 2006; Hodge et al. 2011; Riehl 2016). Although the exact mechanism causing reproductive synchrony may vary by species, the end result is thought to be increased female fitness. These potential fitness benefits do not explain the results we report here for the gelada system: geladas live in polygynous groups where female mate choice is limited (Snyder-Mackler et al. 2012a) and synchrony appears to result from, rather than prevent, the fitness costs associated with male takeovers (i.e., due to infanticide and/or the Bruce effect). At the proximate level, females show higher rates of postpartum resumptions of cycling only after experiencing a takeover, regardless of their reproductive state at the time of takeover (cycling, lactating, or pregnant) or the time of year. For most of these females, we were unable to distinguish “deceptive” signals of fertility (i.e., nonovulatory and/or nonconceptive swellings), a known female counterstrategy to infanticide (e.g., Zinner and Deschner 2000), from “true fertility.” Nevertheless, although we expect that some postpartum returns to cycling were indeed deceptive, male takeovers were also associated with high conception rates in the following months, suggesting that the majority of females truly returned to cycling after a takeover. Two additional lines of evidence suggest that male takeovers drive reproductive synchrony in geladas and not the reverse. First, experiencing a takeover (even during the ecological peak in conceptions) actually delayed the conception peak by 4–6 months for the majority of females in these units as compared to females unaffected by takeovers. Second, most of the females that return to cycling following a takeover were pregnant or lactating at the time of takeover. We have hormonal evidence that 80% of females terminate pregnancies after a takeover (Roberts et al. 2012); and we have statistical evidence that incoming males kill nearly half of the infants of lactating females (Beehner and Bergman 2008). Although we have not examined reproductive changes across individual females, the overall pattern indicates that female receptivity increases after, and not before, male takeovers. In addition, because takeovers themselves are seasonal in this population (this manuscript; see also Pappano and Beehner 2014), the post-takeover reproductive synchrony described here also produced a distinct pattern of birth seasonality. Why are male takeovers seasonal? In some systems, seasonal periods of female cycling and receptivity appear to attract male takeovers and/or influxes of males (e.g., Sugiyama and Ohsawa 1974; Borries 2000; Cords 2000; Morelli et al. 2009; Zhao et al. 2011; Hongo et al. 2016). However, most of these examples come from strict seasonal breeders, where females only conceive during a narrow window of time due to ecological constraints, and males target that seasonal peak in reproductive activity. Although we do not yet have a complete answer as to why takeovers cluster during one time of the year, we do not believe male geladas target a conception season in quite the same way, in part because geladas are not strict seasonal breeders. In addition, our finding that takeovers alter subsequent patterns of female fertility (i.e., the takeover precedes the increase in females that return to cycling and delays the birth peak) suggests that the timing of male takeovers does not simply “mirror” female reproductive seasonality. Finally, previous research in this population has suggested that relative male body condition is one important factor determining the outcome of male takeovers (Pappano and Beehner 2014). Therefore, we suggest that the timing of male takeovers is more or less independent from the timing of female reproductive seasonality, and we are currently conducting a more fine-grained analysis in order to determine the causal factors involved. Together, the evidence we report here for reproductive synchrony and seasonality following male takeovers is in line with the neuroendocrine literature on chemosensory mechanisms that stimulate or inhibit female reproduction. Specifically, in rodents, we see male-mediated resumption of cycling in females (the Whitten effect: Whitten et al. 1968), male-mediated female maturation (the Vandenbergh effect: Vandenbergh et al. 1972) and male-mediated pregnancy termination (the Bruce effect: Bruce 1959), in addition to sexually selected infanticide (Hrdy 1979). While we have yet to identify the mechanism(s) at work in geladas, our results suggest that such male-mediated proximate mechanisms shape gelada birth patterns. Further research will help untangle the evolutionary significance of such a response, which requires an analysis of individual females to see whether those that “reset” have a reproductive advantage over those that do not. We suspect that this kind of male-mediated birth seasonality could help explain seasonal reproductive patterns in species that experience infanticide, feticide, and/or the Bruce effect, or where alpha male replacements themselves (or their equivalent) are seasonal. For example, in white-faced capuchins, male takeovers are seasonal, and often result in infanticide (Fedigan 2003; Schoof and Jack 2013). Births show a seasonal peak in March, but also increase roughly 5 months after the peak takeover season (Carnegie et al. 2011). Here, we found no evidence that infants born outside of the ecological birth peak were more likely to die before 2 years of age. As likely capital breeders, this result is perhaps unsurprising: reproduction in capital breeders can only occur once females reach a certain condition threshold, which is thought to represent the appropriate energetic stores required to successfully carry a pregnancy to term. As such, any costs associated with reproduction should be paid upfront via infertility and/or miscarriage. We were unable to identify pregnancy loss in our dataset (we did not have hormones for most of these pregnancies) to address this directly. However, we expect that infertility is the primary "bottleneck" for reproduction in this population because females routinely give birth during off-peak periods. In this study, we were unable to identify pregnancy loss in the absence of hormonal monitoring because females continue to produce sexual swellings well into pregnancy (Roberts et al. 2017). Nevertheless, we do not suspect that pregnancy loss is a significant cost to breeding out-of-peak in this population, mainly because we routinely record births during off-peak periods. Instead, birth seasonality is likely due to periods of infertility. Still, there may be more subtle costs associated with birth timing. That is, births that occur out-of-peak (and not during the period of maximum food availability) may carry costs relating to maternal condition or to infant growth and development. Because our analysis was at the population-level, we did not examine individual-level characteristics that may influence infant survival. Therefore, the next questions are: 1) whether there is variability in infant survival if we factor in individual characteristics of the mother such as age, dominance rank, or unit size; and 2) whether there are any life history or developmental costs to giving birth outside of the ecological birth peak beyond the one we use here (infant survival to 2 years of age). For example, a female that gives birth outside of the peak may have a longer subsequent interbirth interval or suffer decreased longevity herself. Birth timing may also impact the availability of weaning foods at critical developmental periods for the infant (e.g., Koenig et al. 1997), or show overall effects on infant growth and development. The 2 birth peaks described here are reminiscent of those first described by Dunbar (1980). However, we give them very different interpretations. First, Dunbar (1980) described a dry season birth peak (November to January) that he attributed to a prospective strategy for females to avoid giving birth to vulnerable infants during the cold-wet season. By contrast, our dry season birth peak is driven by post-takeover females that give birth about 4 months after the population-wide seasonal birth peak. Second, Dunbar (1980) attributed his early wet-season birth peak (June to August) to a capital breeding strategy for females to exploit the plentiful food following the short rains (rain that we did not detect in our dataset). By contrast, while our late cold-wet season birth peak (August to October) falls within the window of peak green grass availability, it is also tightly associated with warm weather and low glucocorticoid metabolites at the time of conception. Therefore, our current working hypothesis is that reproduction in female geladas is limited by the energetics of thermoregulation in a cold, hypoxic environment. Testing this hypothesis further will require physiological data on energy balance. Finally, although the takeover birth peak helps explain some of the variation in birth timing observed, we still do not know why some females give birth outside of either birth peak. It may be that females that are shifted from the general ecological pattern (e.g., as the result of a takeover) have difficulties shifting themselves back to the ecological birth peak again. If costs to giving birth outside of the ecological peak are low (as our results suggest), then perhaps the time it takes to delay a reproductive event is a higher cost than just giving birth at a less opportune time. Future exploration of developmental costs associated with off-peak births will help elucidate whether our data are consistent with this hypothesis. SUPPLEMENTARY MATERIAL Supplementary data are available at Behavioral Ecology online. FUNDING This work was supported by: National Science Foundation (grant numbers BCS-0715179, IOS-1255974, BCS-1340911); Leakey Foundation; National Institutes of Health (grant number K99AG051764); National Geographic Society (grant number 8100–06, 8989-11); Sigma Xi; Wildlife Conservation Society; and the University of Michigan. We would like to thank the Ethiopian Wildlife Conservation Authority (EWCA) and the wardens and staff of the Simien Mountain National Park for their permission and on-going support for our long-term research project. Additionally, we are grateful to our excellent field team across the years, most especially E. Jejaw, A. Fanta, S. Girmay, J. Jarvey, and M. Gomery for their assistance with field data collection. We also owe thanks to A. Marshall, B. Dantzer, and J. Mitani for providing valuable feedback on data analyses. Finally, we would like to thank the 2 anonymous reviewers and the editors of Behavioral Ecology for their helpful feedback and comments on an earlier draft. Data accessibility: Analyses reported in this article can be reproduced using the data provided by Tinsley Johnson et al. (2018). REFERENCES Agrell J , Wolff JO , Yloenen H . 1998 . Counter-strategies to infanticide in mammals: costs and consequences . Oikos . 83 : 507 – 517 . Google Scholar CrossRef Search ADS Anderson DR , Burnham KP . 2002 . Avoiding pitfalls when using information-theoretic methods . J Wildl Manage . 66 : 912 – 918 . Google Scholar CrossRef Search ADS Arlettaz R , Christe P , Lugon A , Perrin N , Vogel P . 2001 . Food availability dictates the timing of parturition in insectivorous mouse-eared bats . Oikos . 95 : 105 – 111 . Google Scholar CrossRef Search ADS Bartoń K . 2016 . MuMIn: Multi-model inference. R package version 1.15.6 [cited 2018 February 10] . Available from: https://cran.r-project.org/web/packages/MuMIn. Bates D , Maechler M , Bolker B , Walker S . 2015 . Fitting linear mixed-effects models using lme4 . J Stat Softw . 67 : 1 – 48 . Google Scholar CrossRef Search ADS Beehner JC , Bergman TJ . 2008 . Infant mortality following male takeovers in wild geladas . Am J Primatol . 70 : 1152 – 1159 . Google Scholar CrossRef Search ADS PubMed Beehner JC , Bergman TJ . 2017 . The next step for stress research in primates: To identify relationships between glucocorticoid secretion and fitness . Horm Behav . 91 : 68 – 83 . Google Scholar CrossRef Search ADS PubMed Beehner JC , Bergman TJ , Cheney DL , Seyfarth RM , Whitten PL . 2005 . The effect of new alpha males on female stress in free-ranging baboons . Anim Behav . 69 : 1211 – 1221 . Google Scholar CrossRef Search ADS Beehner JC , McCann C . 2008 . Seasonal and altitudinal effects on glucocorticoid metabolites in a wild primate (Theropithecus gelada) . Physiol Behav . 95 : 508 – 514 . Google Scholar CrossRef Search ADS PubMed Beehner JC , Onderdonk DA , Alberts SC , Altmann J . 2006 . The ecology of conception and pregnancy failure in wild baboons . Behav Ecol . 17 : 741 – 750 . Google Scholar CrossRef Search ADS Beehner JC , Whitten PL . 2004 . Modifications of a field method for fecal steroid analysis in baboons . Physiol Behav . 82 : 269 – 277 . Google Scholar CrossRef Search ADS PubMed Berger J . 1983 . Induced abortion and social factors in wild horses . Nature . 303 : 59 – 61 . Google Scholar CrossRef Search ADS PubMed Boness DJ , Bowen WD , Iverson SJ . 1995 . Does male harassment of females contribute to reproductive synchrony in the grey seal by affecting maternal performance ? Behav Ecol Sociobiol . 36 : 1 – 10 . Google Scholar CrossRef Search ADS Borries C . 2000 . Male dispersal and mating season influxes in hanuman langurs living in multi-male groups . In: Kappeler PM , editor. Primate males: causes and consequences of variation in group composition . Cambridge : Cambridge University Press . p. 146 . Breen KM , Oakley AE , Pytiak AV , Tilbrook AJ , Wagenmaker ER , Karsch FJ . 2007 . Does cortisol acting via the type II glucocorticoid receptor mediate suppression of pulsatile luteinizing hormone secretion in response to psychosocial stress ? Endocrinology . 148 : 1882 – 1890 . Google Scholar CrossRef Search ADS PubMed Brockman DK , van Schaik CP . 2005 . Seasonality and reproductive function . In: Brockman DK , van Schaik CP , editors. Seasonality in primates: studies of living and extinct human and non-human primates . Vol. 44 . Cambridge : Cambridge University Press . p. 269 – 305 . Google Scholar CrossRef Search ADS Bronson FH . 1985 . Mammalian reproduction: an ecological perspective . Biol Reprod . 32 : 1 – 26 . Google Scholar CrossRef Search ADS PubMed Bronson FH . 1989 . Mammalian reproductive biology . Chicago (IL) : University of Chicago Press . Bronson FH . 2009 . Climate change and seasonal reproduction in mammals . Philos Trans R Soc Lond B Biol Sci . 364 : 3331 – 3340 . Google Scholar CrossRef Search ADS PubMed Bruce HM . 1959 . An exteroceptive block to pregnancy in the mouse . Nature . 184 : 105 . Google Scholar CrossRef Search ADS PubMed Burthe S , Butler A , Searle KR , Hall SJ , Thackeray SJ , Wanless S . 2011 . Demographic consequences of increased winter births in a large aseasonally breeding mammal (Bos taurus) in response to climate change . J Anim Ecol . 80 : 1134 – 1144 . Google Scholar CrossRef Search ADS PubMed Campos FA , Morris WF , Alberts SC , Altmann J , Brockman DK , Cords M , Pusey A , Stoinski TS , Strier KB , Fedigan LM . 2017 . Does climate variability influence the demography of wild primates? Evidence from long-term life-history data in seven species . Glob Chang Biol . 23 : 4907 – 4921 . Google Scholar CrossRef Search ADS PubMed Carnegie SD , Fedigan LM , Melin AD . 2011 . Reproductive seasonality in female capuchins (Cebus capucinus) in Santa Rosa (Area de Conservacion Guanacaste), Costa Rica . Int J Primatol . 32 : 1076 – 1090 . Google Scholar CrossRef Search ADS Chappell MA , Hammond KA . 2004 . Maximal aerobic performance of deer mice in combined cold and exercise challenges . J Comp Physiol B . 174 : 41 – 48 . Google Scholar CrossRef Search ADS PubMed Cheviron ZA , Bachman GC , Storz JF . 2013 . Contributions of phenotypic plasticity to differences in thermogenic performance between highland and lowland deer mice . J Exp Biol . 216 : 1160 – 1166 . Google Scholar CrossRef Search ADS PubMed Clutton-Brock TH , Albon SD , Guinness FE . 1989 . Fitness costs of gestation and lactation in wild mammals . Nature . 337 : 260 – 262 . Google Scholar CrossRef Search ADS PubMed Colmenares F , Gomendio M . 1988 . Changes in female reproductive condition following male take-overs in a colony of hamadryas and hybrid baboons . Folia Primatol (Basel) . 50 : 157 – 174 . Google Scholar CrossRef Search ADS PubMed Cords M . 2000 . The number of males in guenon groups . In: Kappeler PM , editor. Primate males: causes and consequences of variation in group composition . Cambridge : Cambridge University Press . p. 84 – 96 . Crook JH , Gartlan JS . 1966 . Evolution of primate societies . Nature . 210 : 1200 – 1203 . Google Scholar CrossRef Search ADS PubMed Dantzer B , Fletcher QE , Boonstra R , Sheriff MJ . 2014 . Measures of physiological stress: a transparent or opaque window into the status, management and conservation of species ? Conserv Physiol . 2 : cou023 . Google Scholar CrossRef Search ADS PubMed Di Bitetti MS , Janson CH . 2000 . When will the stork arrive? Patterns of birth seasonality in neotropical primates . Am J Primatol . 50 : 109 – 130 . Google Scholar CrossRef Search ADS PubMed Drent RH , Daan S . 1980 . The prudent parent: energetic adjustments in avian breeding . Ardea . 68 : 225 – 252 . Dubost G , Henry O . 2017 . Seasonal reproduction in neotropical rainforest mammals . Zool Stud . 56:2 . Dunbar RIM . 1980 . Demographic and life history variables of a population of gelada baboons (Theropithecus gelada) . J Anim Ecol . 49 : 485 – 506 . Google Scholar CrossRef Search ADS Dunbar RIM . 1984 . Reproductive decisions: an economic analysis of gelada baboon social strategies . Princeton : Princeton University Press . Dunbar RIM , Dunbar P . 1974 . Social dynamics of gelada baboons . Anim Behav . 22 : 203 . Google Scholar CrossRef Search ADS PubMed Dunbar RIM , Hannah-Stewart L , Dunbar P . 2002 . Forage quality and the costs of lactation for female gelada baboons . Anim Behav . 64 : 801 – 805 . Google Scholar CrossRef Search ADS Erb WM , Borries C , Lestari NS , Hodges JK . 2012 . Annual variation in ecology and reproduction of wild simakobu (Simias concolor) . Int J Primatol . 33 : 1406 – 1419 . Google Scholar CrossRef Search ADS Fairbanks LA , McGuire MT . 1987 . Mother-infant relationships in vervet monkeys: response to new adult males . Int J Primatol . 8 : 351 – 366 . Google Scholar CrossRef Search ADS Fardi S , Sauther ML , Cuozzo FP , Jacky IAY , Bernstein RM . 2017 . The effect of extreme weather events on hair cortisol and body weight in a wild ring-tailed lemur population (Lemur catta) in southwestern Madagascar . Am J Primatol. e22731 . Fedigan LM . 2003 . Impact of male takeovers on infant deaths, births and conceptions in Cebus capucinus at Santa Rosa, Costa Rica . Int J Primatol . 24 : 723 – 741 . Google Scholar CrossRef Search ADS Fernandez-Duque E , Rotundo M , Ramirez-Llorens P . 2002 . Environmental determinants of birth seasonality in night monkeys (Aotus azarai) of the Argentinean Chaco . Int J Primatol . 23 : 639 – 656 . Google Scholar CrossRef Search ADS Foerster S , Cords M , Monfort SL . 2012 . Seasonal energetic stress in a tropical forest primate: proximate causes and evolutionary implications . PLoS One . 7 : e50108 . Google Scholar CrossRef Search ADS PubMed Fraser-Smith AC . 1975 . Male-induced pregnancy termination in the prairie vole, Microtus ochrogaster . Science . 187 : 1211 – 1213 . Google Scholar CrossRef Search ADS PubMed Gesquiere LR , Khan M , Shek L , Wango TL , Wango EO , Alberts SC , Altmann J . 2008 . Coping with a challenging environment: effects of seasonal variability and reproductive status on glucocorticoid concentrations of female baboons (Papio cynocephalus) . Horm Behav . 54 : 410 – 416 . Google Scholar CrossRef Search ADS PubMed Gilchrist JS . 2006 . Female eviction, abortion, and infanticide in banded mongooses (Mungos mungo): implications for social control of reproduction and synchronized parturition . Behav Ecol . 17 : 664 – 669 . Google Scholar CrossRef Search ADS Hayes JP . 1989 . Field and maximal metabolic rates of deer mice (Peromyscus maniculatus) at low and high altitudes . Physiol Zool . 62 : 732 – 744 . Google Scholar CrossRef Search ADS Heideman PD , Utzurrum RC . 2003 . Seasonality and synchrony of reproduction in three species of nectarivorous Philippines bats . BMC Ecol . 3 : 11 . Google Scholar CrossRef Search ADS PubMed Hill RA , Lycett JE , Dunbar RIM . 2000 . Ecological and social determinants of birth intervals in baboons . Behav Ecol . 11 : 560 – 564 . Google Scholar CrossRef Search ADS Hodge SJ , Bell MB , Cant MA . 2011 . Reproductive competition and the evolution of extreme birth synchrony in a cooperative mammal . Biol Lett . 7 : 54 – 56 . Google Scholar CrossRef Search ADS PubMed Hongo S , Nakashima Y , Akomo-Okoue EF , Mindonga-Nguelet FL . 2016 . Female reproductive seasonality and male influxes in wild mandrills (Mandrillus sphinx) . Int J Primatol . 37 : 416 – 437 . Google Scholar CrossRef Search ADS Hrdy SB . 1974 . Male-male competition and infanticide among the langurs (Presbytis entellus) of Abu, Rajasthan . Folia Primatol (Basel) . 22 : 19 – 58 . Google Scholar CrossRef Search ADS PubMed Hrdy SB . 1979 . Infanticide among animals: a review, classification, and examination of the implications for the reproductive strategies of females . Ethol Sociobiol . 1 : 13 – 40 . Google Scholar CrossRef Search ADS Hunter CP . 2001 . Ecological determinants of gelada ranging patterns (Theropithecus gelada) . Ph.D. dissertation, The University of Liverpool, Liverpool. Ims RA . 1990 . The ecology and evolution of reproductive synchrony . Trends Ecol Evol . 5 : 135 – 140 . Google Scholar CrossRef Search ADS PubMed Iwamoto T , Dunbar RIM . 1983 . Thermoregulation, habitat quality and the behavioural ecology of gelada baboons . J Anim Ecol . 52 : 357 – 366 . Google Scholar CrossRef Search ADS Janson C , Verdolin J . 2005 . Seasonality of primate births in relation to climate . In: Brockman DK , van Schaik CP , editors. Seasonality in primates: studies of living and extinct human and non-human primates . Vol. 44 . Cambridge : Cambridge University Press . p. 307 . Google Scholar CrossRef Search ADS Jarvey JC , Low BS , Pappano DJ , Bergman TJ , Beehner JC . In press. Graminivory and fallback foods: annual diet profile of geladas (Theropithecus gelada) living in the Simien Mountains National Park, Ethiopia . Int J Primatol . Jessop TS , Lane ML , Teasdale L , Stuart-Fox D , Wilson RS , Careau V , Moore IT . 2016 . Multiscale evaluation of thermal dependence in the glucocorticoid response of vertebrates . Am Nat . 188 : 342 – 356 . Google Scholar CrossRef Search ADS PubMed Johnson JB , Omland KS . 2004 . Model selection in ecology and evolution . Trends Ecol Evol . 19 : 101 – 108 . Google Scholar CrossRef Search ADS PubMed Koenig A , Borries C , Chalise MK , Winkler P . 1997 . Ecology, nutrition, and timing of reproductive events in an Asian primate, the Hanuman langur (Presbytis entellus) . J Zool . 243 : 215 – 235 . Google Scholar CrossRef Search ADS Lancaster JB , Lee RB . 1965 . The annual reproductive cycle in monkeys and apes . In: Devore I , editor. Primate behavior . New York (NY) : Holt, Rinehart and Winston . p. 486 – 513 . Landys MM , Ramenofsky M , Wingfield JC . 2006 . Actions of glucocorticoids at a seasonal baseline as compared to stress-related levels in the regulation of periodic life processes . Gen Comp Endocrinol . 148 : 132 – 149 . Google Scholar CrossRef Search ADS PubMed Lea AJ , Altmann J , Alberts SC , Tung J . 2015 . Developmental constraints in a wild primate . Am Nat . 185 : 809 – 821 . Google Scholar CrossRef Search ADS PubMed Loudon ASI , Racey PA . 1987 . Reproductive energetics in mammals: The proceedings of a symposium held at the Zoological Society of London on 10th and 11th April 1986. Oxford: Published for the Zoological Society of London by Clarendon Press . Maestripieri D , Georgiev AV . 2016 . What cortisol can tell us about the costs of sociality and reproduction among free-ranging rhesus macaque females on Cayo Santiago . Am J Primatol . 78 : 92 – 105 . Google Scholar CrossRef Search ADS PubMed Manning JM , Bronson FH . 1990 . The effects of low temperature and food intake on ovulation in domestic mice . Physiol Zool . 63 : 938 – 948 . Google Scholar CrossRef Search ADS Martínez-Mota R , Righini N , Palme R . 2016 . Fluctuations in daily energy intake do not cause physiological stress in a Neotropical primate living in a seasonal forest . Oecologia . 182 : 973 – 984 . Google Scholar CrossRef Search ADS PubMed McCabe GM , Thompson ME . 2013 . Reproductive seasonality in wild Sanje mangabeys (Cercocebus sanjei), Tanzania: relationship between the capital breeding strategy and infant survival . Behaviour . 150 : 1 – 31 . Google Scholar CrossRef Search ADS McFarland R , Barrett L , Boner R , Freeman NJ , Henzi SP . 2014 . Behavioral flexibility of vervet monkeys in response to climatic and social variability . Am J Phys Anthropol . 154 : 357 – 364 . Google Scholar CrossRef Search ADS PubMed Morelli TL , King SJ , Pochron ST , Wright PC . 2009 . The rules of disengagement: Takeovers, infanticide, and dispersal in a rainforest lemur, Propithecus edwardsi . Behaviour . 146 : 499 – 523 . Google Scholar CrossRef Search ADS Morino L , Borries C . 2017 . Offspring loss after male change in wild siamangs: the importance of abrupt weaning and male care . Am J Phys Anthropol . 162 : 180 – 185 . Google Scholar CrossRef Search ADS PubMed Negus NC , Berger PJ . 1972 . Environmental factors and reproductive processes in mammalian populations . In: Velardo JT , Kasprow BA , editors. Biology of reproduction: basic and clinical studies . New Orleans (LA) : Third Pan American Congress on Anatomy . p. 89 – 98 . Ohsawa H , Dunbar RIM . 1984 . Variations in the demographic-structure and dynamics of gelada baboon populations . Behav Ecol Sociobiol . 15 : 231 – 240 . Google Scholar CrossRef Search ADS Ostner J , Nunn CL , Schülke O . 2008 . Female reproductive synchrony predicts skewed paternity across primates . Behav Ecol . 19 : 1150 – 1158 . Google Scholar CrossRef Search ADS PubMed Packer C , Pusey AE . 1983 . Male takeovers and female reproductive parameters: a simulation of oestrous synchrony in lions (Panthera leo) . Anim Behav . 31 : 334 – 340 . Google Scholar CrossRef Search ADS Palombit RA . 2015 . Infanticide as sexual conflict: coevolution of male strategies and female counterstrategies . Cold Spring Harb Perspect Biol . 7:a017640 . Pappano DJ , Beehner JC . 2014 . Harem-holding males do not rise to the challenge: androgens respond to social but not to seasonal challenges in wild geladas . R Soc Open Sci . 1 : 140081 . Google Scholar CrossRef Search ADS PubMed Polo P , Colmenares F . 2016 . Seasonality of reproductive events and early mortality in a colony of hamadryas baboons (Papio hamadryas hamadryas) over a 30-year period: capital breeding and life history patterns in a food-provisioned population seasonally thermally stressed . Am J Primatol . 78 : 1149 – 1164 . Google Scholar CrossRef Search ADS PubMed Pride RE . 2005 . Optimal group size and seasonal stress in ring-tailed lemurs (Lemur catta) . Behav Ecol . 16 : 550 – 560 . Google Scholar CrossRef Search ADS R Core Team . 2016 . R: A language and environment for statistical computing. Vienna (Austria): R Foundation for Statistical Computing [cited 2018 February 10]. Available from: https://www.R-project.org/ . Riehl C . 2016 . Infanticide and within-clutch competition select for reproductive synchrony in a cooperative bird . Evolution . 70 : 1760 – 1769 . Google Scholar CrossRef Search ADS PubMed Roberts SB , Cole TJ , Coward WA . 1985 . Lactational performance in relation to energy intake in the baboon . Am J Clin Nutr . 41 : 1270 – 1276 . Google Scholar CrossRef Search ADS PubMed Roberts EK , Lu A , Bergman TJ , Beehner JC . 2012 . A Bruce effect in wild geladas . Science . 335 : 1222 – 1225 . Google Scholar CrossRef Search ADS PubMed Roberts EK , Lu A , Bergman TJ , Beehner JC . 2017 . Female reproductive parameters in wild geladas (Theropithecus gelada) . Int J Primatol . 38 : 1 – 20 . Google Scholar CrossRef Search ADS Roberts S-J , Nikitopoulos E , Cords M . 2014 . Factors affecting low resident male siring success in one-male groups of blue monkeys . Behav Ecol . 25 : 852 – 861 . Google Scholar CrossRef Search ADS van Schaik CP , van Noordwijk MA , Nunn CL . 1999 . Sex and social evolution in primates . In: Lee PC, editor. Comparative Primate Socioecoogy. Cambridge: Cambridge University Press. p. 204 – 231 . Schoof VA , Jack KM . 2013 . The association of intergroup encounters, dominance status, and fecal androgen and glucocorticoid profiles in wild male white-faced capuchins (Cebus capucinus) . Am J Primatol . 75 : 107 – 115 . Google Scholar CrossRef Search ADS PubMed Schülke O , Ostner J . 2012 . Ecological and social influences on sociality . In: Call J , Kappeler PM , Palombit RA , Silk JB , Mitani JC , editors. The evolution of primate societies . Chicago : The University of Chicago Press . p. 195 – 219 . Smuts BB , Smuts RW . 1993 . Male aggression and sexual coercion of females in nonhuman primates and other mammals: evidence and theoretical implications . Adv Study Behav . 22 : 1 – 63 . Google Scholar CrossRef Search ADS Snyder-Mackler N , Alberts SC , Bergman TJ . 2012a . Concessions of an alpha male? Cooperative defence and shared reproduction in multi-male primate groups . Proc Biol Sci . 279 : 3788 – 3795 . Google Scholar CrossRef Search ADS Snyder-Mackler N , Beehner JC , Bergman TJ . 2012b . Defining higher levels in the multilevel societies of geladas (Theropithecus gelada) . Int J Primatol . 33:1054-1068. Stutchbury BJ , Morton ES . 1995 . The effect of breeding synchrony on extra-pair mating systems in songbirds . Behaviour . 132 : 675 – 690 . Google Scholar CrossRef Search ADS Sugiyama Y . 1965 . On the social change of hanuman langurs (Presbytis entellus) in their natural condition . Primates . 6 : 381 – 418 . Google Scholar CrossRef Search ADS Sugiyama Y , Ohsawa H . 1974 . Population dynamics of Japanese macaques at Ryozenyama, Suzuka Mts. I. General view . Jap J Ecol . 24 : 50 – 59 . Tecot SR . 2010 . It’s all in the timing: Birth seasonality and infant survival in Eulemur rubriventer . Int J Primatol . 31 : 715 – 735 . Google Scholar CrossRef Search ADS Teichroeb JA , Jack KM . 2017 . Alpha male replacements in nonhuman primates: variability in processes, outcomes, and terminology . Am J Primatol . 79:e22674 . Tinsley Johnson ET , Snyder-Mackler N , Lu A , Bergman TJ , Beehner JC . 2018 . Data from: social and ecological drivers of reproductive seasonality in geladas . Dryad Digital Repository . http://dx.doi.org/10.5061/dryad. Valeggia C , Ellison PT . 2004 . Lactational amenorrhoea in well-nourished Toba women of Formosa, Argentina . J Biosoc Sci . 36 : 573 – 595 . Google Scholar CrossRef Search ADS PubMed Vandenbergh JG , Drickamer LC , Colby DR . 1972 . Social and dietary factors in the sexual maturation of female mice . J Reprod Fertil . 28 : 397 – 405 . Google Scholar CrossRef Search ADS PubMed Wade GN , Schneider JE . 1992 . Metabolic fuels and reproduction in female mammals . Neurosci Biobehav Rev . 16 : 235 – 272 . Google Scholar CrossRef Search ADS PubMed Ward M, Milledge PJS, West JB. 1995. High altitude physiology and medicine. London (UK): Chapman & Hall Medical. Wasser SK . 1996 . Reproductive control in wild baboons measured by fecal steroids . Biol Reprod . 55 : 393 – 399 . Google Scholar CrossRef Search ADS PubMed Weingrill T , Gray DA , Barrett L , Henzi SP . 2004 . Fecal cortisol levels in free-ranging female chacma baboons: relationship to dominance, reproductive state and environmental factors . Horm Behav . 45 : 259 – 269 . Google Scholar CrossRef Search ADS PubMed Whitten WK , Bronson FH , Greenstein JA . 1968 . Estrus-inducing pheromone of male mice: transport by movement of air . Science . 161 : 584 – 585 . Google Scholar CrossRef Search ADS PubMed Xiang ZF , Sayers K . 2009 . Seasonality of mating and birth in wild black-and-white snub-nosed monkeys (Rhinopithecus bieti) at Xiaochangdu, Tibet . Primates . 50 : 50 – 55 . Google Scholar CrossRef Search ADS PubMed Zhao Q , Borries C , Pan W . 2011 . Male takeover, infanticide, and female countertactics in white-headed leaf monkeys (Trachypithecus leucocephalus) . Behav Ecol Sociobiol . 65 : 1535 – 1547 . Google Scholar CrossRef Search ADS Zinner D , Deschner T . 2000 . Sexual swellings in female hamadryas baboons after male take-overs: ‘Deceptive’ swellings as a possible female counter-strategy against infanticide . Am J Primatol . 52 : 157 – 168 . Google Scholar CrossRef Search ADS PubMed Zipple MN , Grady JH , Gordon JB , Chow LD , Archie EA , Altmann J , Alberts SC . 2017 . Conditional fetal and infant killing by male baboons . Proc Biol Sci . 284:20162561 . © The Author(s) 2018. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Behavioral EcologyOxford University Press

Published: Feb 17, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off