Evaluating the Performance of Diversification Models for Reconstructing Evolutionary History

Abstract

Understanding the processes that generate diversity is key to interpreting the patterns we see in the present. New developments in modelling these processes have promised unprecedented prospects for unravelling the evolutionary past. However, the empirical behaviour of these models in many of their practical applications is not well understood. This thesis investigates the influence of diversification models in a variety of contexts. First, I consider the Generalised Mixed Yule-Coalescent (GMYC) method for molecular species delimitation. This method identifies transition points between species- and population-level diversification processes on a time-resolved evolutionary tree. I show that this method is sensitive to the choice of mitochondrial marker used, and that the best marker can vary widely across study groups. Next, I investigate the influence of diversification models used to place prior distributions on time-resolved trees in molecular dating. Specifically, I look at the influence of the tree prior in analysing data sets with multiple individuals per species. These data sets can arise by accident where species boundaries are not well understood, and violate the assumptions of both population- and species-level tree priors. I use simulation to show that molecular date estimates can be seriously affected by the choice of tree prior in some circumstances, but are remarkably robust in general. Finally, I extend the analysis of tree prior sensitivity to new methods for dating the origins of human language families. I show that these methods are also robust to the choice of tree prior, and that speciation priors are preferred for language data sets regardless of taxonomic scale. My work will contribute to an improved understanding of the role of diversification models in empirical studies and will increase confidence in these methods across multiple realms of enquiry.