Humans are innately social animals, and social interactions are among the single most biologically important type of behavior in which humans and non-human primates engage. Life outcomes such as mental and physical health, status, material well-being, and reproductive success depend on an individual’s ability to successfully navigate social interactions by deploying appropriate combinations of behavior. Direct observation of the fine structure of social interactions in free-ranging non-human primates can provide valuable insights into the neurobiological processes underlying social behavior by making it possible to quantify social phenotypes in terms of how they deploy particular combinations of behavior in social interactions. However, the complexity of social interactions makes the estimation of such phenotypes computationally and statistically challenging. I develop computational methods for modeling behavioral phenotypes in terms of the combinations of behaviors that characterize social interactions, and apply these methods to behavioral observations of rhesus macaques. I find that modeling the way animals combine social behaviors reveals novel information about animal social phenotypes that is lost when behaviors are considered individually, and I extend the model to directly incorporate biological and environmental covariates.