What drives results in Bayesian morphological clock analyses?

Abstract

Recently, approaches that estimate species divergence times using fossil taxa and models of morphological evolution have exploded in popularity. These methods incorporate diverse biological and geological information to inform posterior reconstructions, and have been applied to several high-profile clades to positive effect. However, there are important examples where morphological data are misleading, resulting in unrealistic age estimates. While several studies have demonstrated that these approaches can be robust and internally consistent, the causes and limitations of these patterns remain unclear. In this study, we dissect signal in Bayesian dating analyses of three mammalian clades. For two of the three examples, we find that morphological characters provide little information regarding divergence times as compared to geological range information, with posterior estimates largely recapitulating those recovered under the prior. However, in the cetacean dataset, we find that morphological data do appreciably inform posterior divergence time estimates. We supplement these empirical analyses with a set of simulations designed to explore the efficiency and limitations of binary and 3-state character data in reconstructing node ages. Our results demonstrate areas of both strength and weakness for morphological clock analyses, and help to outline conditions under which they perform best and, conversely, when they should be eschewed in favour of purely geological approaches.