Use of social network models to understand play partner choice strategies in three primate species.

Although play is seen in many species, its evolutionary function is still largely unknown. Several relevant, proposed hypotheses (such as the training for the unexpected, self-assessment, social skills, and dominance hierarchy hypotheses) make predictions about how animals should optimally choose their play partners based on their familiarity or other demographic variables. We used a social network approach to analyze focal sample data on brown capuchins (Cebus apella), hamadryas baboons (Papio hamadryas), and diademed sifaka (Propithecus diadema) to understand how these species choose their play partners with respect to demographic variables. Using exponential random graph models (ERGMs), we found that sifaka and capuchins generally tended to play with animals who were similar to them. The baboons were only sensitive to age differences in the formation of strong play relationships. Our data most strongly support the training for the unexpected hypothesis, as according to predictions all species preferred to play with animals who were their close social partners, decreasing the possibility of cheating during play. Through the first application (to our knowledge) of ERGMs to primate behavior, we were able to compare the effects of many demographic variables on the complex, interdependent social structure of primates. Applying this tool to additional groups and species will provide further insight into evolutionary mechanisms of play behavior across taxa.