Detecting correlated evolution on phylogenies: a general method for the comparative analysis of discrete characters

Abstract

I present a new statistical method for analysing the relationship between two discrete characters that are measured across a group of hierarchically evolved species or populations. The method assesses whether a pattern of association across the group is evidence for correlated evolutionary change in the two characters. The method takes into account information on the lengths of the branches of phylogenetic trees, develops estimates of the rates of change of the discrete characters, and tests the hypothesis of correlated evolution without relying upon reconstructions of the ancestral character states. A likelihood ratio test statistic is used to discriminate between two models that are fitted to the data: one allowing only for independent evolution of the two characters, the other allowing for correlated evolution. Tests of specific directional hypotheses can also be made. The method is illustrated with an application to the Hominoidea.