Independent fitness consequences of group size variation in Verreaux's sifakas

The study addresses how intraspecific variation in group size influences direct fitness components—reproductive success and survival—in Verreaux's sifakas. While group living has evolved repeatedly and group size varies widely, general patterns linking group size to fitness remain unclear and heterogeneous across taxa. Previous work has yielded mixed results across mammals, birds, and primates, and suggests potential roles for environmental variability and social composition (e.g., adult sex ratio) in shaping optimal group size. The authors aim to test whether reproduction and survival are maximized at intermediate group sizes (optimal group size hypothesis) or show no group-size effect (as suggested by earlier indirect fitness proxies in the same population). They explicitly evaluate effects of group size, adult sex ratio (ASR), interannual rainfall (as a food/condition proxy), age, and number of co-resident adult females on female birth rates and individual survival over 25 years.

The paper reviews extensive evidence showing heterogeneous relationships between group size and fitness across taxa. In social carnivores (wild dogs, hyenas) and cooperative breeders (meerkats), group size effects on survival and reproduction vary by study and context. Among rodents and other mammals (marmots, degus, prairie voles), survival and reproductive metrics often peak at intermediate sizes or show independence from group size. Primates often show fitness costs in larger groups (e.g., longer interbirth intervals, reduced fertility and juvenile survival). Studies incorporating environmental factors reveal interactions where wetter or drier years alter group-size benefits (e.g., superb starlings, anis), and ASR can shape optimal group size (prairie voles, ostriches). Theoretical expectations often posit stabilizing selection for intermediate group sizes, but local ecological and social factors and differing interests among age/sex classes can shift optima and maintain variation. Gaps remain due to diverse covariates and fitness measures used across studies, motivating focused, long-term analyses in additional taxa.

Study site and species: Long-term study of Verreaux's sifakas (Propithecus verreauxi) in Kirindy Forest, Western Madagascar—a seasonal dry deciduous forest with known predators (fosa, Harrier hawks). Sifakas are diurnal, arboreal, frugi-folivorous, can live up to 25 years (median 12), and can produce a single infant annually from age 4 (mean age at first reproduction 5.6 years). Since 1994, all animals in ~60 ha were individually marked (nylon/radio collars). Dataset includes 279 individuals in up to 10 adjacent groups, with 236 births (41 mothers; female lifetime reproductive success 1–16, mean 5.75) recorded between 1994 and 2021. Death classification: 105 confirmed dead (observed predation, remains found, or <8 months at disappearance). Additional 162 individuals disappeared (unknown status: emigration vs death). Twenty individuals were alive at data acquisition.

Predictors and definitions: Group size was the total number of group members excluding dependent infants at the time of birth/death. Age was measured in years. Adult sex ratio (ASR) was the proportion of adult males among adults (adults ≥5 years). Rainfall was the total in the 12 months prior to the July birth season (external estimates). For birth models, the number of co-resident adult females was included due to suspected inhibitory effects. Non-linearities were assessed via quadratic terms for group size, age, and number of adult females.

Statistical analysis: Binomial GLMMs with logit link were used to model (1) probability a female gave birth (yes/no) and (2) probability an individual died (yes/no). Fixed effects for birth: group size (+ quadratic), female age (+ quadratic), number of adult females (+ quadratic), rainfall. A separate birth model replaced number of adult females with ASR (due to collinearity). Random effects: individual ID and group ID with maximal random slopes (group size, quadratic group size, age, quadratic age, number of adult females and its quadratic, rainfall) within individual and group; correlations between random slopes and intercepts were dropped for convergence. Additional reduced models assessed dominance status effects on birth in groups with ≥2 adult females. Survival analyses: three GLMMs with different death classifications: (a) confirmed deaths only (N=105), (b) confirmed deaths + unknown females (N=173), (c) confirmed deaths + unknown females + males ≤4 years (N=216). Fixed effects: group size (+ quadratic), age (+ quadratic), ASR, rainfall; random effects: individual and group with random slopes. In model (a), age terms were removed from random slopes for convergence. Model selection used AIC; models with ΔAIC < 2 are presented in main text. R version 4.3.2 was used; predictors were log-transformed (age, rainfall) and standardized (mean 0, SD 1). Significance was assessed by LRTs comparing full vs null (random-only) models. Parametric bootstrap (lme4::bootMer, 1000 replicates) provided CIs. VIFs were checked (car package). Model stability was verified by leave-one-level-out of random effects.

• Group size distribution: 2–10 individuals (mean 6.26, SD 1.96; N=352 group-years). Mean composition: adult females 1.93 ± 0.80, adult males 2.13 ± 0.86, juveniles 2.20 ± 1.34.
• Reproductive success (probability of giving birth): Best-supported by linear-terms model (AIC linear 425.99 vs quadratic 434.96). Total group size had no significant effect (LRT full-null: χ^2=9.26, df=3, p>0.001; N=352). The number of co-resident adult females significantly reduced a female’s probability of giving birth (reproductive inhibition). Female age significantly affected birth probability, with older females exhibiting the lowest rates. Rainfall in the prior year had no effect. In the alternative model including ASR, results were similar and ASR was not significant (LRT full-null: χ^2=22.18, df=5, p>0.001). Dominance status did not impede birth rates in analyses restricted to multi-female groups.
• Survival (probability of dying): The quadratic-terms model was best (AIC 764.28 vs 818.68 for linear). Individuals in larger groups had significantly higher mortality risk (lower survival) (LRT full-null: χ^2=55.02, df=5, p<0.001; N up to 1022 person-years; main model c: N=216 deaths classification sample). Age affected mortality, with younger individuals showing higher mortality. Neither ASR nor prior-year rainfall predicted mortality. Models using alternative death classifications yielded similar effects.

Findings indicate asymmetric fitness consequences of group size in Verreaux's sifakas rather than a single intermediate optimum: survival decreases with increasing group size, while reproduction is constrained by the number of co-resident adult females rather than overall group size. These results imply that different fitness components (survival vs reproduction) are influenced by distinct aspects of social organization and can vary independently across group sizes. Environmental variability (rainfall) and ASR did not modulate birth or death rates, suggesting intrinsic (age) and social (female–female competition) factors predominate in this population.

The negative association between co-resident adult females and female birth probability suggests subtle female competition over reproduction among largely philopatric, often related females, without overt dominant suppression; mechanisms may be behavioral or physiological. The survival costs of larger groups could reflect increased detectability by predators, within-group competition, or intrinsic disease/parasite transmission risks, consistent with broader comparative findings that larger group size does not necessarily confer longevity benefits. Group composition and sexual conflict may further shape realized group sizes: dominant males benefit from more females (high reproductive skew), but female dominance and costs of larger groups to females may bias outcomes toward smaller groups. Together, these factors explain why net fitness benefits accrue at smaller rather than intermediate group sizes in this system.

Contrary to optimal group size predictions, Verreaux's sifakas do not maximize both reproduction and survival at intermediate group sizes. Survival is highest in smaller groups, and female reproduction is inhibited by more co-resident adult females, with overall group size having no direct effect on birth probability. Environmental (rainfall) and ASR effects were negligible. The study highlights that specific components of group size (e.g., number of adult females) and sexual conflict over group composition can drive fitness outcomes independently of total group size. Future work should elucidate mechanisms of female reproductive competition, assess whether smaller groups advance age at first reproduction, shorten inter-birth intervals, or improve infant survival, and further dissect how ecological variation and social network structure mediate disease and energetic costs across group sizes.

• Observational, single-population study at one site may limit generalizability to other habitats or populations.
• Death classification includes categories of unknown status; alternative classifications were modeled, but some uncertainty about emigration vs mortality remains.
• Collinearity between number of adult females and ASR necessitated separate birth models, precluding simultaneous estimation of their effects.
• Some model convergence constraints required simplifying random-effects structures (e.g., removing correlations; omitting age terms from random slopes in one survival model), which may affect variance partitioning.
• Rainfall was used as a proxy for resource availability/condition; finer-grained measures of food abundance or quality were not included.
• Mechanisms underlying female reproductive competition were not identified (behavioral vs physiological pathways remain unresolved).