Disentangling nonrandom assortment, indirect effects, and joint plasticity as causes of phenotypic (dis)similarity between social partners.

Social partners frequently resemble each other. These correlations between the phenotypes of interacting individuals (e.g., social partners, group members, etc.) can be caused by multiple processes. These processes include joint plasticity in response to shared environments, plasticity in response to partner phenotype, and genetic similarity arising from nonrandom assortment due to clustered relatives, spatiotemporal stratification, and partner choice. Although social plasticity and nonrandom assortment can influence evolutionary dynamics, these two processes have most often been studied separately, and disentangling the causes of partner resemblance in observational datasets can be challenging. Furthermore, standard statistical models of social plasticity do not allow for potential social feedback between partners' phenotypes, and estimating joint plasticity to shared environmental effects requires environmental data that is rarely available. We assessed the performance of several statistical models to estimate nonrandom assortment and social plasticity in observational datasets, using simulations of a socially monogamous species, in which nonrandom mating, social plasticity (with or without feedback) and joint plasticity occurred alone or simultaneously. Standard "variance-partitioning approaches" retrieved biased estimates except when the process they aimed to estimate occurred on its own. By contrast, a recently proposed statistical model explicitly including social plasticity as a dynamic process generating feedback between partners' phenotypes (the so-called social animal model) performed best even in scenarios with multiple co-occurring processes. While we recommend empiricists use this latter approach, we also highlight the importance of appropriate sampling designs given the study question and system, and using simulations to assess model performance in realistic scenarios.