Is there evidence for the Bruce effect in white-faced capuchins?

Abstract

Infanticide by males is an extreme form of sexual conflict that can increase male reproductive success at a cost to female reproductive success. Females have evolved a variety of strategies to reduce the occurrence and the cost of infanticide, including the termination of pregnancy after nontraumatic exposure to non-sire males, known as the Bruce effect. A recent model (Zipple 2020) proposed that the Bruce effect will evolve in populations if the risk of infanticide is high and alpha male replacements occur routinely but infrequently relative to gestation length. We tested this model using 38 years of demographic data from a population of wild white-faced capuchins (Cebus imitator) in Sector Santa Rosa of the Área de Conservación Guanacaste, Costa Rica. We found that this population has a high rate of infanticide associated with alpha male replacements that occur every 2.9 years on average. Applying the Zipple (2020) model to our population leads to the prediction that capuchins should exhibit the Bruce effect, as the associated reproductive costs would be lower than the expected costs of future infanticide. However, we did not find evidence of any type of male-mediated prenatal loss in this species: female birth rates after alpha male replacements were not lower compared to stable periods. Possible explanations include that white-faced capuchins do not respond to extreme events with reproductive inhibition, or that other female strategies such as allonursing better mitigate the costs of infanticide. Finally, the evolution of female reproductive strategies may not be labile enough that it can be predicted from population-specific social and demographic patterns without regard to phylogenetic constraints. More studies are needed to understand the conditions that determine the occurrence and evolution of the Bruce effect in wild mammalian populations.

Introduction

Sexual conflict is an antagonistic coevolutionary process between the sexes over the control of reproduction (Parker 1979; 2006; Lessells 2006; Chapman 2006). For some species, it can manifest in a variety of behavioral, physiological, genetical, and anatomical adaptations and counter-adaptations that function to increase the reproductive success of individuals of one sex while decreasing the reproductive success of individuals of the opposite sex (Chapman et al. 2003; Arnqvist & Row 2005). In group-living mammals, where the reproductive interests of the sexes diverge through a skew in the time and energy costs associated with parental investment (Trivers 1972; Stumpf et al. 2011), a widespread form of sexual conflict involves infanticide by males (Hrdy 1974; 1979; Hausfater & Hrdy 1984; van Schaik & Janson 2000; Lukas & Huchard 2014). Male infanticide (hereafter, infanticide) is the killing of infants by unrelated males. It typically occurs in the context of a replacement of the primary breeding male (Beehner & Bergman 2008; Pavé et al. 2012; Brasington et al. 2017), is more common in mammalian mating systems where males have a high potential to monopolize breeding opportunities (i.e., high reproductive skew) (Lukas & Huchard 2014), and is a major source of infant mortality in many populations where it occurs (Pusey & Packer 1994; Palombit 2012). According to the sexual selection hypothesis for the occurrence of infanticide (Hrdy 1974, 1979), by removing a dependent infant, the mother ends lactational amenorrhea and resumes ovarian cycling more rapidly, making it possible for the infanticidal male to reproduce sooner with that female. Infanticide, in some species, is an intrasexually selected trait that increases the reproductive success of the infanticidal male and simultaneously reduces the reproductive success of other males. Additionally, infanticide creates intersexual conflict as females that lose infants to infanticide have decreased lifetime reproductive success due to the time and energy lost to dead-end parental investment, especially in taxa that have slow life-histories (e.g., nonhuman primates, van Schaik 2000; Stumpf et al. 2011; Palombit 2015; Fedigan et al. 2021).

Females have evolved a variety of adaptations to counter infanticide. Across mammals, the evolution of infanticide is associated with subsequent evolutionary transitions to multi-male mating by females (Wolff & Macdonald 2004; Lukas & Huchard 2014). Beyond polyandry, females can decrease paternity certainty through strategies including extended receptivity, concealed or unpredictable ovulation, non-conceptive mating, situational receptivity, and deceptive signaling (Hrdy 1979; Packer & Pusey 1983; Agrell et al. 1998; Ebensperger 1998; van Schaik et al. 1999; Palombit 2015; Davidian et al. 2022). These strategies will vary in effectiveness in relation to how long an infanticidal male has been in a group, his mating history with a female, and female reproductive status. However, while these mating strategies may reduce the occurrence of infanticide, they may not eliminate it. Thus, females may also evolve strategies to minimize reproductive costs when the threat of infanticide is high (e.g., accelerated weaning, Teichroeb & Sicotte 2008). One physiological adaptation to mitigating the costs of infanticide is the Bruce effect, whereby females block pregnancy through the prevention of embryo implantation or terminate an existing pregnancy in response to nontraumatic sensory exposure to a non-sire male (Bruce 1959; 1960). The Bruce effect was initially identified in captive rodents (Labov 1981), but has been reported in a variety of taxa in field settings, including alpine marmots (Marmota marmota, Hackländer & Arnold 1999), domestic horses (Equus ferus caballus, Berger 1983; Bartoš et al. 2011), lions (Panthera leo, Bertram 1975), and primates (Theropithecus gelada, Roberts et al. 2012; Papio hamadryas, Amann et al. 2017). The Bruce effect is distinct from feticide; while both are considered forms of male-mediated prenatal loss (Zipple et al. 2019), feticide is the termination of pregnancy in response to aggression from males, whereas the Bruce effect only requires exposure to the male. Feticide is a sexually selected trait that functions similarly to how infanticide apportions benefits to males and costs to females, and may even be more pervasive than infanticide across mammals (Zipple et al. 2019). In fact, it has been suggested that the Bruce effect may be a female adaptation to either potential infanticide or feticide (Zipple 2020).

The Bruce effect is likely adaptive for females because it permits the reallocation of reproductive investment (gestation and lactation) from at-risk offspring into future offspring sired by the potentially infanticidal male, as these offspring will be more likely to survive (Hrdy 1979; Schwagmeyer 1979; Labov 1981; Zipple 2020). Additionally, terminating an at-risk embryo or fetus may reduce the potential physical and physiological costs to the female (e.g., injury or physiological stress response) that are associated with infanticidal behavior by males (Packer & Pusey 1983; Le Boeuf & Mesnick 1991; Beehner et al. 2005; Engh et al. 2006; Schneider-Crease et al 2020). Despite its benefits, the Bruce effect may not be present in all taxa that exhibit infanticide or feticide, suggesting that there may be scenarios where this potential cost-mitigating strategy is more or less effective. Zipple (2020) formalized verbal models (Schwagmeyer 1979; Labov 1981; Zipple et al. 2019) by developing an analytical and individual-based model to test the hypothesis that the Bruce effect could evolve as a female adaptation to counter infanticide and/or feticide, and to determine under what conditions it is likely to do so. He found that the Bruce effect is likely to evolve when the risk of infanticide is high enough, such that the reproductive costs of potential infanticide outweigh the costs of terminating a pregnancy. The benefits of the Bruce effect are enhanced in taxa with shorter gestation lengths and an increased probability of maternal death following infanticide, although it can still evolve with no maternal death risk if the probability of infanticide is high enough. He found that regardless of how high infanticide risk is, frequent male takeovers of the primary breeding position (or here, alpha male replacements, AMRs) reduce the benefits of the Bruce effect, as the costs associated with terminating a pregnancy increase if AMRs occur at intervals shorter than gestation length. Finally, the evolution of the Bruce effect as a feticide avoidance strategy will not benefit females unless the risk of both feticide and maternal death following feticide is high; consequently, the presence of the Bruce effect and feticide are expected to be mutually exclusive in a population.

The Zipple model has previously been supported empirically with data from three species of papionin primates: chacma baboons (Papio ursinus), yellow baboons (P. cynocephalus), and geladas (Theropithecus gelada). Despite infanticide being the primary cause of infant mortality in the three study populations, and the species’ similar gestation lengths, the model accurately predicted the presence of the Bruce effect in geladas and its absence in chacma and yellow baboons. The difference is likely due to the differential risks of infanticide (higher in geladas) and the frequency of male takeovers (more frequent in chacma and yellow baboons). The Zipple (2020) model is a useful tool as it provides clear theoretical predictions that it has translated into quantifiable variables (monthly probabilities of infanticide p_inf, feticide p_fet, male take-over p_take, and maternal death p_death) that can be used to determine the relative costs and benefits of the Bruce effect and its presence in a population.

Here, we provide the first examination of potential for the Bruce Effect in a species of platyrrhine primates, white-faced capuchins (Cebus imitator). White-faced capuchins are an excellent species to examine for this female reproductive strategy using predictions generated by Zipple’s (2020) model. They live in social groups with an average of 15 individuals (range 3–40) and approximately equal numbers of adult males and females (Fedigan & Jack 2013; Perry 2012). Females are philopatric, whereas males disperse from their natal group between 4.5–7.5 years old and subsequently transfer among social groups every ~4 years (Jack & Fedigan 2004a, 2004b; Perry 2012; Perry et al. 2017). Females give birth to their first offspring around the age of 6–7 years, gestate for 5.5 months, and have interbirth intervals of roughly 2 years throughout their adult lives (Carnegie et al. 2005; Fedigan et al. 2008; 2021; Perry 2012). They exhibit moderate reproductive seasonality, with an ecological birth peak that coincides with the early wet season (May to June), although births can occur at any time of year (Carnegie et al. 2011a; Brasington et al. 2022). Compared to other primates of similar body size, they have a relatively late weaning age (as early as 12 months but as late as 23 months, Carnegie et al. 2011a; Sargeant et al. 2015). Females mate with multiple males, but alpha males nearly monopolize reproduction and sire the majority of offspring (80%–100%, Jack & Fedigan, 2006; Muniz et al. 2006; 2010; Godoy et al. 2016; Wikberg et al. 2017). Male contest competition is more common during times of group instability, i.e. during an AMR, whereas male-male aggression is rare when groups are stable (Fedigan 1993; 2003; Perry 1998; Jack 2003; Brasington et al. 2017). However, AMRs can be aggressive or peaceful, and can involve extra-group or co-resident males (Teichroeb & Jack 2014; Brasington et al. 2017). Infanticide is the primary cause of infant mortality, occurs in the context of all forms of AMRs, and rarely occurs when groups are stable (Brasington et al. 2017; Perry 2012, but see Schoof et al. 2014; Nishikawa et al. 2020; Kulick et al. 2021). AMRs take place most often in the dry season, with a wave of infanticides occurring in quick succession; this timing leads to a subsequent birth peak in AMR-affected groups that is shifted to earlier in the year relative to the birth peak of stable groups (Brasington et al. 2022). Female white-faced capuchins exhibit traits suggested to decrease the occurrence of infanticide, including multi-male mating, concealed ovulation, nonconceptive mating, situational receptivity, and, in rare cases, female dispersal (reviewed in Fedigan & Jack 2013). Despite this, infanticide still occurs at a very high rate; on average, females exposed to high infanticide risk (e.g., during AMRs) lose half of their infants, compared to 1 in 5 infants during stable periods (Perry 2012; Brasington et al. 2017; Kalbitzer et al. 2017; Fedigan et al. 2021). Infanticide is costly to female reproductive success as females that experience infant loss due to infanticide experience increased energetic and temporal costs (Fedigan et al. 2021). Consequently, it may benefit females to mitigate these costs of infanticide with other strategies, such as the Bruce effect.

In this study, we use 38 years of long-term demographic data on wild white-faced capuchins in the Santa Rosa Sector of the Área de Conservación Guanacaste, Costa Rica to address the following questions:

1. Are capuchins predicted to evolve the Bruce effect? Following the Zipple (2020) model, we calculate the monthly probability of infanticide (p_inf) to determine whether the maternal costs of the Bruce effect are lower than the costs of infanticide, as expected if the Bruce effect is predicted to evolve in a species. Specifically, we test whether p_inf in capuchins exceeds 0.1, which was the threshold found by Zipple (2020) for females to accrue benefits of the Bruce effect. We also determine the monthly probability of an AMR (p_take) as the Bruce effect is more likely to evolve when AMRs are not common (p_take≤0.15). We specifically focus on the benefits the Bruce effect may offer in relation to infanticide as we do not have comparable data on the monthly probability of feticide (p_fet) in this population.

2. Do capuchins exhibit male-mediated prenatal loss? We use indirect evidence from birth patterns to determine if male-mediated prenatal loss occurs in response to AMRs. Specifically, we examine birth rates in the 18 months surrounding an AMR among females who could have experienced male-mediated prenatal loss. If the Bruce effect is present in this population, we predict that birth rates immediately after the AMR (i.e., given a 5.5 month gestation length, infants born in this time period would be sired by the previous alpha) will be lower than birth rates in the 6 months before or 6–12 months after the AMR. While the Bruce effect has been broadly defined to occur in response to a non-sire male (Zipple et al. 2019), it is more narrowly defined as a response to a novel male (Bruce 1960). If the Bruce effect occurs only in response to novel males in our population, we may find that birth rates will be lower in the months immediately after an extragroup AMR compared to periods before or 6–12 months after, but that the pattern will not be evident for co-resident AMRs. We also examine birth rates in the 18 months surrounding an AMR compared to birth rates of stable groups (groups that did not experience an AMR) in the same time periods. If male-mediated loss occurs in response to AMRs in general, we predict that, compared to stable groups, birth rates will be lower in the months immediately after an AMR, and birth rates will increase in the 6–12 months after an AMR (i.e., females that exhibit prenatal loss will then conceive and give birth by the new male). If male-mediated prenatal loss does not occur in response to AMRs, we predict that this decrease and subsequent increase in births will not be apparent in AMR groups and birth rates in AMR groups will not differ from birth rates in stable groups in any month after the AMR (or during the same time period in stable groups). While we do not have the data to distinguish between the different forms of male-mediated prenatal loss (i.e., the Bruce effect or feticide), they are expected to be mutually exclusive with feticide more commonly found in populations with lower rates of infanticide (Zipple et al. 2019; Zipple 2020). Therefore, we predict that any form of male-mediated loss that is found in this population will be due to the Bruce effect given capuchins’ established high rates of infanticide and the general observation that male to female aggression is infrequent (Fragaszy et al. 2004).

Methods

Study site and population

We collected data over 38 years between January 1986 and August 2024 from eight groups of wild white-faced capuchins in the Santa Rosa Sector of the Área de Conservación Guanacaste, Costa Rica (Table 1). This study population has been followed nearly continuously since 1983 (Melin et al. 2020).

Table 1.

Study groups, study periods, the number of AMRs each group experienced, tenure lengths of alpha males, and the mean (range) number of adult females during the study period.

Group	Years studied	# Months observed	# AMRs (N=40)	Alpha tenure length, months, mean (range) (N=32)	# Adult females, mean (range)
AD	2013–2024	142	4	35.9 (32.1–40.8)	6.7 (4–8)
BC	2020–2024	57	3	22.9 (4.8–41)	4.2 (3–6)
CP1/RM2/S
S	1986–2024	471	6	82.8 (16.2–178)	6.2 (2–12)
EX	2007–2016	114	3	30.9 (19.9–52.2)	3.2 (1–4)
FF	2023–2024	9	1	NA	6.7 (6–8)
GN3	2007–2024	215	6	30.9 (7.2–81.2)	9.1 (5–12)
LV	1991–2024	410	14	28.1 (3.3–64.5)	5.8 (3–10)
SE	1986–1993	89	3	23.6 (22.9–24.3)	2.7 (1–4)

¹CP fissioned in 2013 into AD and RM

²RM fissioned in 2023 into SS and FF

³GN fissioned in 2019 into GN and BC

Data collection

All study groups were censused at least twice per month and researchers noted all instances of demographic change (births, deaths, emigrations, immigrations, and AMRs). Between 1991 and 2010, there were eight gaps in data collection due to researcher absences from the field site, each lasting 3–8 months. We assume no successful AMRs occurred during these gaps as all alpha males were the same before and after the gaps, and we excluded these gaps in our analysis of birth rates (see below). AMRs can occur by extragroup or co-resident males, although the rate of infanticide does not differ between them (Brasington et al. 2017). For each AMR, we recorded the date of the previous alpha’s loss of status, the length of instability in the male hierarchy (if relevant), the date when a new male was definitively recorded as alpha as determined by behavioral observation, and whether the new male was a co-resident or extragroup male. Alpha males are readily discernible based on increased size, piloerection, increased vigilance, and specialized behavior directed towards them by conspecifics (Gould et al. 1997; Jack et al. 2014; Jack & Fedigan 2018).

Here we report on the births and deaths of N=379 infants in relation to AMRs (N=40 involving 37 distinct adult males) during the study period. Individuals are considered infants from birth until 12 months of age. Given infant dependency on mothers in the first year of life (Fragaszy et al., 2004; Sargeant et al., 2015), and the youngest age of confirmed emigration in our population (20 months, Jack & Fedigan 2004a), we consider any missing infant to be deceased. If infants were <12 months old at the end of the study period for their group (n=3), they were not considered in mortality analyses. Any infants born during data gaps were not included in birth analyses (see below).

Data analysis

Aim 1: Are capuchins predicted to evolve the Bruce effect?

Following the calculations from the analytical and individual-based models in Zipple (2020), we calculated the monthly probability of infanticide (p_inf), expected maternal cost of infanticide (c_inf), maternal cost of the Bruce effect (c_Bruce), and the probability of an AMR (p_take). For these analyses, we used the full dataset of AMRs (N=40).

We first determined risk periods and stable periods for each group in relation to infanticide risk. Following others (e.g. Brasington et al. 2017; Kalbitzer et al. 2017; Fedigan et al. 2021), risk periods were calculated from the date of the AMR up to 5.5 months after (to account for gestation for infants conceived by the former alpha male). In some instances, male hierarchies were unstable for a period of time between the prior alpha’s rank loss and when the new alpha’s position stabilized (e.g., Jack & Fedigan 2018; Perry 2012). In these cases, we extended the risk period to include these unstable periods (beginning from the loss of prior alpha) plus the 5.5 months after stabilization of the new alpha male. To calculate p_inf, we only considered infants born during risk periods (versus infants that experienced risk periods during their infancy but were born before the risk period began), as their mortality rates represent the biologically relevant variable to consider when assessing the expected maternal cost of infanticide relative to the maternal cost of the Bruce effect. We calculated for each month of life, up to 12 months, the mortality rate of infants born into risk periods (N=60) and the mortality rate for infants born into stable periods (N=247). For example, month 1 = the proportion of newborns who die in the first month of life, month 2 = the proportion of one-month olds who die in the second month of life, and so on up to 12 months. We then calculated each monthly $p_{\mathit{\text{inf}},n}$ as:

$$p_{\mathit{\text{inf}},n}={\mathrm{m}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{h}}_{\mathrm{n}}{\mathrm{m}\mathrm{o}\mathrm{r}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y}\mathrm{i}\mathrm{n}\mathrm{r}\mathrm{i}\mathrm{s}\mathrm{k}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{o}\mathrm{d}\mathrm{s}-\mathrm{m}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{h}}_{\mathrm{n}}\mathrm{m}\mathrm{o}\mathrm{r}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y}\mathrm{i}\mathrm{n}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{b}\mathrm{l}\mathrm{e}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{o}\mathrm{d}\mathrm{s}$$

The expected maternal cost of infanticide across all months prior to weaning ($c_{\mathit{\text{inf}}}$) is the sum over all months of (maternal investment × probability of infanticide × probability infanticide has not previously occurred):

$$c_{inf}={\sum }_{n=1}^L\left[(G+n-1)\left(p_{in{f}_n}\right){\prod }_{k=0}^{n-1}\left(1-p_{in{f}_k}\right)\right]$$

where L is lactation length (in capuchins, the onset of weaning is L=12 months, Carnegie et al. 2011a), G is gestation length (in capuchins, G=5.5 months), and n represents the infants’ age in months, and p_inf,0 is defined as 0. The estimated maternal cost of the Bruce effect (c_Bruce) is considered by Zipple (2020) to be halfway through gestation (G/2), assuming females experience an AMR on average halfway through their pregnancy.

To determine the monthly probability of an AMR (p_take), we first calculated p_take for each study group (Table 1) as the total number of AMRs divided by the total number of months of observation. We calculated the final p_take as the mean p_take across all groups.

Aim 2: Do capuchins exhibit male-mediated prenatal loss?

To determine whether capuchins exhibit male-mediated prenatal loss, we used indirect evidence from birth patterns in the 18 months surrounding AMRs. We only considered births from females who could have exhibited the Bruce effect, i.e. we excluded births from females whose previous infant died during an AMR period and then gave birth again in the same 18-month period (n=26 births). For each AMR, we set the date of the AMR to month0. We then determined the number of births that occurred in each month from 6 months before the AMR to 12 months after the AMR. Every AMR time period was scaled, such that if the AMR occurred on 12 March, month-1 was 12 February, month+1 was 12 April, and so on. When there was an unstable period before the stabilization of a new alpha male, we considered the start date of the risk period (e.g., the former alpha’s loss of status, see above) as month0 as it may be expected that the Bruce effect could be exhibited upon the commencement of group instability as males compete for the alpha position. In all cases where there was this extended risk period, the next alpha was present in the group, either through prior immigration that started the instability or by being co-resident. There were some instances in which a group experienced more than one AMR in close succession, resulting in overlapping 18-month windows around each AMR. In these cases, to avoid muddying our month-specific analyses (e.g., a single month that could be defined as +2 relative to the first AMR and – 2 relative to the second AMR), we excluded months in overlapping windows, but we included months in each 18-month AMR time period that did not overlap with the other. We also excluded months where there were gaps in data collection or that were right-censored when the 18-month period had not concluded before the end of data collection. Out of a total possible 720 months across all 40 AMRs, our dataset includes births from 588 months.

We first fit a Bayesian generalized linear mixed model (GLMM) with a Poisson error distribution using the brms R package (Bürkner 2017) to determine how AMR Type (co-resident or extragroup male) influenced the number of births across the 18 month time period (month-6 to month+12). To control for a different number of females in each group, we included an offset of log(number of adult females in the group in each month). This functionally models births as the number of births per female. For random effects, we fit AMR ID as a random intercept to account for multiple months of birth data for each AMR, and additionally fit a random slope of AMR Type within month with a correlated intercept. The latter two terms allow us to assess if co-resident and extragroup AMRs differ from each other in their birth patterns in each month relative to the global effect of AMR type, as well as if any of the months show a significant elevation in the likelihood of giving birth relative to the global average. We used the default priors in brms (weakly informative), and ran the model for 2,000 iterations including 1,000 warmup iterations. We assessed model convergence with $\widehat{◻}$ which were all 1.

To control for the potential for environmental effects on prenatal loss, we ran a second analysis (similar to Roberts et al. (2012) and Amann et al. (2017)). We paired each AMR group to different groups that were in a stable period for at least part of the same 18-month time periods surrounding the AMR. When matching group-months, we excluded months that were right-censored or encompassed data collection gaps, and if a particular matched group became unstable during the period, only the stable group-months were used for pairing. We were able to find appropriate stable group pairs for N=33 AMRs (mean 1.9, range 1–4 pairs per AMR) for a total of 63 stable time periods across 7 groups. This resulted in n=491 months from AMR time periods and n=996 months from stable time periods.

We fit a Bayesian GLMM with a Poisson error distribution to determine how Group Type (stable or AMR) influenced the number of births across the 18 month time period. To control for a different number of females in each group, we included an offset of log(number of adult females in the group in each month). For random effects, we fit AMR ID as a random intercept to account for multiple months of birth data for each AMR. We also fit Group-Period as a random intercept to account for multiple months of birth data from each paired group during that period. Finally, we fit a random slope of Group Type within month with a correlated intercept to assess if AMR and stable groups differ from each other in their birth patterns in each month relative to the global effect of Group Type, and if any of the months show a significant elevation in the likelihood of giving birth relative to the global average. We used the default priors in brms (weakly informative), and ran the model for 2,000 iterations including 1,000 warmup iterations. We assessed model convergence with $\widehat{◻}$ which were all 1.

All analyses were completed in R version 4.2.2 (R Core Team 2022).

Ethical note

This study complied with protocols approved by Tulane University IACUC (Protocol #0399, 0810, and 1612) and the University of Calgary (ACC Protocols AC15–0161 and AC 20–0148). Fieldwork in Santa Rosa was purely observational and did not interfere with capuchin behavior. All animal protocols followed the ABS/ASAB guidelines for the use of animals in research.

Results

Aim 1: Are capuchins predicted to evolve the Bruce effect?

In the entire dataset (N=40 AMRs), overall infant mortality was 37% (140/376), and 52% (67/140) of these deaths occurred during risk periods. Infant mortality for all infants in risk periods was 56% (73/129) compared to 27% (67/247) in stable periods. Consequently, the excess mortality rate that can be attributed to AMR-related infanticide beyond the baseline death rate is 29%. More specifically, infant mortality for infants born into risk periods was 68% (41/60). The monthly age-based probability of mortality for these infants born into risk periods ranged from 0–0.400 compared to a range of 0–0.085 for infants born into stable periods, resulting in a p_inf range of 0–0.315 across the first 12 months of life (Table 2). The highest mortality was for infants born into risk periods in the first month of their life. p_inf exceeds 0.1 for the first three months of life.

Table 2.

Infant mortality for infants born into risk periods, born into stable periods, and the calculated probability of infanticide (p_inf). Mortality for each month of life was calculated as the proportion of newborns who died in their first month of life, the proportion of 1-month olds who died in their 2nd month of life, etc. p_inf is calculated as mortality in risk periods - mortality in stable periods for each month. If this value was negative, p_inf was considered 0.

Infant age (months)	Infant mortality, risk periods (N=60)	Infant mortality, stable periods (N=247)	pinf
1	0.400	0.085	0.315
2	0.167	0.027	0.140
3	0.167	0.018	0.148
4	0.120	0.032	0.088
5	0.000	0.010	0.000
6	0.045	0.039	0.007
7	0.000	0.040	0.000
8	0.000	0.016	0.000
9	0.048	0.021	0.026
10	0.050	0.016	0.034
11	0.000	0.006	0.000
12	0.000	0.000	0.000

Given p_inf across a 12 month lactation period (L), the expected maternal cost of infanticide (c_inf = 3.8 months) is 1.4x larger than the expected maternal cost of the Bruce effect (c_Bruce = 2.75 months). Because the costs of the Bruce effect would be exceeded by the expected costs of future infanticide, Zipple’s model predicts that the Bruce effect should have evolved in capuchins.

AMRs occurred on average every 2.86 years (range 0.76–6.54, n=8 groups), corresponding to a mean monthly probability of an AMR (p_take) of 0.04 (0.01–0.11, n=7). Of the 40 AMRs for which we had end dates (N=32), the mean alpha male tenure length was 3.1 years (range 3.3 months - 14.8 years). 3 of the 32 AMRs had tenure lengths shorter than the 5.5 month gestation length, and 29 AMRs had tenure lengths longer than gestation length (Table 1).

Aim 2: Do capuchins exhibit male-mediated prenatal loss?

Co-resident AMRs (N=22, n=100 births) and extragroup AMRs (N=18, n=46 births) did not differ in birth rates when considering the entire 18 month time period (fixed effect AMR Type: posterior mean estimate = −0.41, 95% CI = [−1.07, 0.11]). Further, when looking at the effect of AMR Type at each individual month (random slope of AMR Type across months), the 95% credible intervals of the posterior estimates all contain 0 (Figure 1b); in combination with the lack of a global effect of AMR type, this suggests that there is no significant effect of AMR Type on birth rates in any month. Since births are not lower in the months immediately after an AMR (all 95% CIs of random intercepts of month contain zero; Fig. 1a), nor differ by the new alpha male being co-resident or extragroup, this suggests that females may not exhibit male-mediated prenatal loss during either type of AMR. We considered both types of AMRs together in the subsequent analysis.

Figure 1.

Bayesian GLMM posterior estimates of the number of births in the months surrounding AMRs for extragroup and co-resident AMRs ± 80%, 95%, and 99% credible intervals for the a) random intercept of month (representing baseline number of births in co-resident AMRs) and b) random slope of AMR Type across months (representing the number of births in extragroup AMRs relative to the number of births in co-resident AMRs). AMR groups (N=33 time periods, n=129 births); stable groups (N=63 time periods, n=294 births). Extragroup AMRs (N=18, n=46 births); co-resident AMRs (N=22, n=100 births).

In our comparisons between groups that experience AMRs and groups that were stable during the same time periods, AMR groups (N=33, n=129 births) and stable groups (N=63, n=294 births) did not differ in birth rates across the entire 18-month time period (fixed effect Group Type, posterior mean estimate = −0.03, 95% CI = [−0.40, 0.33]). When looking at the effect of Group Type at each individual month (random slope of Group Type across months), the 95% CIs of the posterior estimates of all months but month-5 contained 0 (Figure 2b), suggesting that there are no differences in birth rates between AMR groups and stable groups in any month. The 80% CI of the random intercept the 80% and 95% CI of the random slope of month-5 does not contain 0 (although the 99% CIs do), suggesting that month-5 had a higher baseline number of births, with AMR groups having fewer less than baseline, however, only months after the AMR (or associated time period in stable groups) are biologically relevant to assessing the Bruce effect. Together, this adds support to our first analysis that capuchins do not exhibit the Bruce effect in response to AMRs.

Figure 2.

Bayesian GLMM posterior estimates of the number of births in the months surrounding AMRs (or paired time periods in stable groups) ± 80%, 95%, and 99% credible intervals for the a) random intercept of month (representing baseline number of births in stable groups) and b) random slope of Group Type across months (representing the number of births in AMR groups relative to the number of births in stable groups). AMR groups (N=33 time periods, n=129 births); stable groups (N=63 time periods, n=294 births).

Discussion

We assessed the potential for the Bruce effect in a species of platyrrhine primate, white-faced capuchins. We tested a recent model (Zipple 2020) that uses infanticide risk and alpha male replacement rates to predict the presence of different forms of male-mediated prenatal loss in mammals. We found that the model predicts that white-faced capuchins will exhibit the Bruce effect due to their high rates of infanticide and infrequent AMRs. However, we found no evidence of the Bruce effect in this population, as birth rates in groups that experience AMRs i) are not lower immediately after the AMR compared to the 6 months before and 6–12 months after the AMR; and ii) are not lower immediately after the AMR compared to stable groups in the same time period.

The Zipple (2020) model seeks to determine the thresholds at which the Bruce effect may provide benefits to females in relation to the risk of infanticide or feticide, i.e. are the costs of the Bruce effect lower than the expected costs of future infanticide if pregnancy is maintained. An increased risk of infanticide will increase the relative benefits of the Bruce effect, with moderate to high levels (p_inf ≥ 0.1) of infanticide yielding a net benefit of the Bruce effect. Here, consistent with previous studies on this population (Brasington et al. 2017; Kalbitzer et al. 2017; Fedigan et al. 2021), we found that white-faced capuchins experience a high rate of infanticide: over half of the infants born during infanticide risk periods are likely to die in their first year of life compared to only a quarter of infants that die during periods of group stability. This corresponds to a monthly probability of infanticide (p_inf) that ranges from 0–0.315 across the first 12 months of life, with the first three months of life exceeding the proposed 0.1 threshold where benefits of the Bruce effect begin to accrue. However, frequent AMRs can increase the costs of the Bruce effect even if a population experiences a high infanticide rate because if AMRs happen at intervals similar to the gestation period, terminating pregnancy will not benefit overall female reproduction as they may never have a successful pregnancy. The Zipple (2020) model found that when p_take ≥ 0.15 (AMRs every ~6.6 months or less given a 6 months gestation period), the Bruce effect will be too costly to overcome its benefits to countering infanticide. Here we found that capuchin AMRs occur roughly every 2.9 years (p_take = 0.04), which far exceeds the 5.5 month gestation (and the 2.1 year inter-birth interval, Fedigan et al. 2021), and therefore should not surmount the potential benefits of the Bruce effect in this population. The increasing risk of maternal death after infanticide can further increase the benefits of the Bruce effect. However, the probabilities of infanticide and AMR found for capuchins here predict a net benefit of the Bruce effect even in the absence of maternal death (Zipple 2020), which is uncommon in the study population (unpublished data). Together, the analytical and individual-based models in Zipple (2020) predict that there are net benefits to capuchin females in this population and they are predicted to have evolved the Bruce effect as a counterstrategy to infanticide. However, we did not find support for these models: indirect evidence from 38 years of birth patterns across 8 social groups suggests that groups that experience AMRs do not exhibit decreased birth rates in the immediate months after an AMR, as is predicted if females experience any form of male-mediated prenatal loss.

Since no male-mediated form of prenatal loss was evident, this eliminates the need to separate out whether fetal loss was driven by feticide or the Bruce effect. However, feticide and the Bruce effect are likely mutually exclusive phenomena and feticide may be a more common male reproductive strategy when infanticide rates are lower (Zipple et al. 2019; Zipple 2020). For example, geladas experience both high rates of infanticide and the Bruce effect, but lack feticide as evidenced by a lack of disproportionate male aggression towards pregnant females upon immigration (Schneider-Crease et al. 2020). Given the high rate of infanticide in this population, feticide is not likely to have evolved in capuchins, thus the remainder of our discussion is centered on why capuchins may not have evolved the Bruce effect as a female strategy to mitigate the costs of male infanticide.

Of the nonhuman primates where male-mediated prenatal loss has been explored in relation to AMRs and infanticide, geladas (Roberts et al. 2012) and hamadryas baboons (Amann et al. 2017) exhibit the Bruce effect, whereas yellow baboons (Zipple et al. 2017), chacma baboons (Zipple 2020), olive baboons (P. anubis, Bailey et al. 2021), Hanuman langurs (Semnopithecus entellus, Sommer 1987; 1994; Agoramoorthy et al. 1988), and golden snub-nosed monkeys (Rhinopithceus roxellana, Xiang et al. 2022) exhibit feticide, and crested macaques (Macaca nigra) appear to show no male-mediated prenatal loss (Kerhoas et al. 2014). The Zipple (2020) model correctly predicts the presence of the Bruce effect in geladas and its absence in yellow and chacma baboons, as well as the presence of feticide in the yellow baboons. It is surprising then that capuchins, which are similar to gelada and hamadryas baboons in their probabilities of infanticide and AMRs, do not exhibit the Bruce effect as predicted, as it appears it would provide reproductive benefits to females. This is especially true considering that most infants that are subject to infanticide are those that are born in the first months following the AMR (Brasington et al. 2017; this study), i.e. the infants that would have been fetal losses if the Bruce effect was exhibited.

Phylogenetic constraints may explain this discrepancy as this is the first study to examine male-mediated prenatal loss in a platyrrhine primate. While infanticide is present across the primate order (Lukas & Huchard 2014; Xiang et al. 2023), male-mediated prenatal loss has only been examined in catarrhine primates. Infanticide was likely present in the ancestor of all catarrhines, whereas infanticide evolved multiple times in platyrrhines with the infanticide found in white-faced capuchins stemming from the ancestor of the family Cebidae (Lukas & Huchard 2014). If the Bruce effect evolves in response to infanticide, it may be possible that it has not evolved in the cebids. However, the presence of infanticide and the Bruce effect are not correlated across rodents (Blumstein 2000; Stokes & Sandal 2019), further emphasizing that the evolution of the Bruce effect can be independent from the evolution of infanticide, and may have evolved for other purposes (e.g., mate choice, Vodjerek et al. 2024); therefore it may not be expected in all clades with infanticide. It will be important to assess the presence of both forms of male-mediated prenatal loss in other platyrrhine species, to understand the relationships among infanticide, the Bruce effect, and feticide outside of catarrhines.

Additionally, it may be that the reproductive biology of capuchins may restrict them from disrupting pregnancy once they have conceived, especially early in pregnancy. For example, during an 18-month period of prolonged extreme drought conditions at this field site, most females still gave birth, although nearly 100% of infants died (Campos et al. 2020). That said, some visibly pregnant females (i.e., ~ 4.5 months gestation) were never observed with infants, suggesting very late pregnancy loss or unobserved neonatal death. In contrast, the sympatric spider monkey (Ateles geoffroyi) stopped reproducing altogether (e.g., terminated pregnancies or anovulatory cycles) as evidenced by no births (Campos et al. 2020), suggesting different responses to extreme environmental stressors in capuchins and spider monkeys. Stress can disrupt pregnancy across a variety of animals (de Catanzaro & Macniven 1992; Nakamura et al. 2008), and while anecdotally it appears that capuchin females may lose pregnancies in response to environmental stressors (although possibly not in early pregnancy when it would most benefit them), they do not appear to lose pregnancies in response to AMRs (Carnegie et al., 2011b). Perhaps social conditions are better buffered compared to environmental conditions that impact access to food and water resources (Campos et al. 2020) which may have a stronger effect on reproductive function (Harcourt 1989; Carnegie et al. 2011a). Alternatively, since the capuchins still reproduced during the extreme drought, it is possible that while some late pregnancy losses may have occurred, they were not common. Capuchins may not have the reproductive physiology to facilitate pregnancy loss even in the face of extreme infant mortality and its associated energetic/time costs to female reproductive success.

Social and mating systems impact what reproductive strategies evolve and co-evolve. While infanticide is present across both polygynous and polygynandrous species (Lukas & Huchard 2014), the Bruce effect may be a more beneficial strategy to mitigate the costs of infanticide in polygynous groups where paternity certainty is higher (Pillay & Kinahan 2009; Zipple et al. 2019). For example, in populations of vlei rats (Otomys irroratus), only polygynous populations exhibit the Bruce effect (Pillay & Kinahan 2009); this may be because polygynandrous vlei rats are able to minimize the occurrence of infanticide by decreasing paternity certainty through multi-male mating. However, despite capuchins mating polygynandrously, they exhibit high reproductive skew and still experience high rates of infanticide, which suggests that this strategy is not entirely effective at removing infanticide as it may be in some rodent populations. Another potential strategy females may use to avoid infanticide, even in different social systems, is female transfer between groups (e.g., Thomas langurs, Presbytis thomasi: Sterck 1997; golden snub-nosed monkeys: Xiang et al 2022). Though uncommon for this female philopatric species, we have previously found that females are more likely to disperse in association with AMRs (Jack & Fedigan 2009). Finally, females may mitigate the costs of potential infanticide with high rates of allonursing. Capuchins have long lactation periods; while infants can be weaned at 12 months, weaning can approach 23 months (Carnegie et al. 2011a, Sargeant et al. 2015). It is possible that late weaning combined with allonursing may decrease the individual maternal costs associated with lactation and/or facilitate earlier resumption of cycling (Brasington 2020), as has been similarly proposed for lions (Pusey & Packer 1987) and rodents (Hayes 2000).

Finally, it may be that despite their high infanticide rate associated with all types of AMRs (Brasington et al. 2017), capuchin AMRs do not elicit the Bruce effect in females due to the variation in how AMRs occur (Teichroeb & Jack 2017) or the variable tenure lengths of alpha males (3.5 months - 14.8 years). While just over half of AMRs occur with the immigration of a novel male (e.g., take-overs and waltz-ins), the other AMRs are undertaken by co-resident males (e.g., rank reversals, succession, fission) (Brasington et al. 2017). The Bruce effect, more strictly defined, may only occur in response to exposure to a novel male, not simply a non-sire male (Bruce 1959; 1960; Zipple et al. 2019). While we did not find a significant difference in births after AMRs between co-resident and extragroup AMRs, if the male ‘cue’ to miscarry is not consistent, the variation in AMR type in capuchins may have precluded the evolution of the Bruce effect. Further, the costs of the Bruce effect are only one month less than the cost of infanticide in this population (2.7 vs 3.8 months, respectively). It may be that the selection pressure has not been strong enough for the Bruce effect to evolve based on these reproductive benefits. Other aspects of male tenure beyond frequency (i.e., variability in tenure length, modes of dominance acquisition) may play an important role in the evolution of the Bruce effect and should be considered in future efforts to model its evolution.

It may be possible that our analysis is not fine-grained enough to detect the Bruce effect. Since capuchins do not show obvious visual signs of early pregnancy, such as changes in anogenital coloration (e.g., baboons, Beehner et al. 2006), we used an indirect approach of assessing birth rates before and after AMRs. A preferable method is to track female ovulatory cycles by measuring reproductive hormones to determine pregnancy and any potential losses. While we have done this on shorter time scales (e.g., Carnegie et al. 2005; Schoof et al., 2014; Buehler 2024), a lack of long-term hormonal data precludes this approach. However, the only study on wild primates to track the Bruce effect hormonally also found indirect evidence through drops in birth rate immediately after AMRs (geladas, Roberts et al. 2012), suggesting this method should be able to detect the phenomenon if it is present. While our sample size may preclude finding strong evidence of the Bruce effect if it is relatively weak or sporadic, it is notable that births after AMRs occur regularly in capuchins, in contrast to other primate populations in which the Bruce effect has been documented. In geladas (Roberts et al. 2012) and hamadryas baboons (Amann et al. 2017), births after AMRs are very rare and are 18x and 4x higher, respectively, in stable groups compared to AMR groups in the 6 months after an AMR. It is clear that a comparably stark Bruce effect does not exist in capuchins. Demographic data from long-lived animals takes many years to accumulate, as evidenced from our limited sample size despite 38 years of longitudinal data, and are likely why phenomena like male-mediated prenatal loss are difficult to identify and report on beyond anecdotes in wild populations.

Ultimately, the costs and benefits of different female counter strategies to infanticide are likely to differ across species, especially when considering the independent evolution of infanticide and variation in reproductive physiology and social systems across the mammalian kingdom. While the Zipple (2020) model attempts to understand the variation in the presence of the Bruce effect through infanticide/feticide rates, male transfer frequency, and risk of maternal death, the presence and effectiveness of other female strategies is important to consider. While the model accurately predicts the presence of the Bruce effect and feticide in geladas and baboons, respectively, it did not accurately predict the apparent absence of any male-mediated prenatal loss in our population of white-faced capuchins. More studies on the presence and absence of male-mediated prenatal loss are needed to test the robusticity of the predictions found in the Zipple (2020) model and determine its applications, limitations, and expansions.