Dimensionality of sexual isolation during reinforcement and ecological speciation in Timema cristinae stick insects

Background: Speciation can involve variation in the dimensionality of population divergence (defined as variation in the number of independent ecological variables, phenotypic traits, or genes differing between populations). Recent work indicates that reproductive isolation between populations is multidimensional, but that this multidimensionality has an upper limit. A remaining question is how ecological and geographic factors affect the dimensionality of reproductive isolation. We address this question here, focusing on sexual isolation. Methods: The data used stem from previously published mating trials from 12 populations (comprising 66 pairwise population comparisons) of Timema cristinae stick insects that vary in host use and geographic arrangement (allopatry versus parapatry). We estimated the dimensionality of sexual isolation by calculating independent axes of divergence that represent combinations of (unmeasured) phenotypic traits and mating preferences that determine degree of sexual isolation. We then examined how populations varied along these axes. Results and conclusions: We found that despite the large number of population comparisons examined, only two dimensions were required to explain observed variation in sexual isolation. Furthermore, patterns of divergence illuminated ecological and geographic effects on the dimensionality of reproductive isolation. Males and females from allopatric populations were tightly clustered along one axis, consistent with sexual selection balanced by natural selection towards a single optimum. In contrast, parapatric males were widely dispersed across both axes, as would occur if reinforcement in parapatry were driving divergence in latent traits that remain constant in allopatry. Populations clustered to some extent by host plant use, although the effect was much weaker than that of geographic arrangement. We discuss our results in terms of the factors affecting speciation. In particular, if sexual isolation tends to have a strong upper limit to its dimensionality, as observed in our dataset, levels of divergence sufficient for speciation may require multiple forms of reproductive isolation.