Variable social organization is ubiquitous in Artiodactyla and probably evolved from pair-living ancestors

Abstract

Previous studies to understand the evolution of interspecific variation in mammalian social organization (SO; composition of social units) produced inconsistent results, possibly by ignoring intraspecific variation. Here we present systematic data on SO in artiodactyl populations, coding SO as solitary, pair-living, group-living, sex-specific or variable (different kinds of SOs in the same population). We found that 62% of 245 populations and 83% of species (83/100) exhibited variable SO. Using Bayesian phylogenetic mixed-effects models, we simultaneously tested whether research effort, habitat, sexual dimorphism, breeding seasonality or body size predicted the likelihood of different SOs and inferred the ancestral SO. Body size and sexual dimorphism were strongly associated with different SOs. Contingent on the small body size (737 g) and putative sexual monomorphism of the earliest fossil artiodactyl, the ancestral SO was most likely to be pair-living (probability = 0.76, 95% CI = 0-1), followed by variable (p = 0.19, 95% CI = 0-0.99). However, at body size values typical of extant species, variable SO becomes the dominant form (p = 0.74, 95% CI = 0.18-1.00). Distinguishing different kinds of 'variable' highlights transitions from SOs involving pair-living to SOs involving solitary and/or group-living with increasing body size and dimorphism. Our results support the assumption that ancestral artiodactyl was pair-living and highlight the ubiquity of intraspecific variation in SO.

Keywords: Artiodactyla; intraspecific variation; phylogenetic mixed-effects model; social organization.

Figure 1.

Representation of animal social systems, including the four components (SO, mating system, social structure, care system). Double arrows indicate that the four components shape the entire social system and can be shaped by other components. Adapted from [1,6]. (Online version in colour.)

Figure 2.

Phylogeny of extant artiodactyls with data on SO, with the coloured boxes at the tips of the phylogeny show SOs observed in different populations of the same species. As a reference, the five possible states (solitary, pair-living, sex-specific, group-living, variable) are plotted above and below the phylogeny in this order. Using the same colours, the inset figures show the probabilities (from Model 1) of each SO (sol, solitary; pair, pair-living; SS, sex-specific; grp, non-variable group-living; var, variable) when body size and dimorphism are at the levels known from the fossil record (ancestral state), or the phylogenetically controlled averages estimated from extant species (phylogenetic mean); number of habitats and studies as well as breeding seasonality are kept at their baselines of 1 and non-seasonal, respectively. The scale bar shows million years before present. For the same figure with variable split into different categories (Model 2), see electronic supplementary material, figure S1. (Online version in colour.)

Figure 3.

Illustrating evolutionary transitions in SO as a function of the predictors (from Model 1). Columns show (from left to right) the probability of solitary, pair-living, sex-specific, group-living and variable SO—using the same colours as in figure 2—while rows show (from top to bottom) predicted changes in those probabilities as a function of female body size (a–e), sexual dimorphism (f–j), number of habitats (k–o), number of studies (p–t) and breeding seasonality (u–y). The numbers in the legends are the posterior probabilities (PP), i.e. the proportion of the posterior distribution that supports a given association; these were not available for solitary, as this was the reference category. Within each row, all other predictors were held at their baseline value, except for body size; for pair-living, body size was kept at the ancestral state, for all others it was kept at the phylogenetic mean for better visibility. Solid black lines are the predicted means, thin coloured lines are 100 random samples drawn from the posterior to illustrate uncertainty. For breeding seasonality (u–y), points are predicted means and lines are 95% CIs. (Online version in colour.)