Oversplitting and Bayesian Inference

Abstract In this paper I will offer a detailed analysis of Bayesian species delimitation methods and the oversplitting problem. I will argue that when discussing oversplitting, two issues should be kept separate: the inference problem concerning the discovery of population structure and the labelling problem surrounding the classification of different lineages as distinct species. Understanding oversplitting as a methodological problem of Bayesian inference is the consequence of not separating the inferential problem from the labelling problem. While there is no intrinsic flaw with Bayesian inference of population structure revealed by oversplitting, treatments of the inference problem can benefit from some of the lessons learned in earlier discussions of Bayesian phylogenetic inference.