Bayesianism, convergence and molecular phylogenetics

Abstract

Bayesian methods are very popular in molecular phylogenetics. At the same time there is concern among biologists and philosophers regarding the properties of this methodology. In particular, there is concern about the lack of objectivity of evidential statements in Bayesian confirmation theory due to the role of prior probabilities. One standard reply to be found in the Bayesian literature is that as data size grows larger differences in prior probability assignments will “wash out” and there will be convergence of opinion among different agents. This paper puts the “washing out of priors” argument to the test in the context of phylogenetic inference. I argue that the role of nuisance parameters in molecular phylogenetics prevents the application of convergence arguments typically found in the literature on Bayesianism.