Strongest Evidence: Maximum apparent phylogenetic signal as a new cladistic optimality criterion

Abstract

A new method for phylogenetic inference, Strongest Evidence (SE), is described. In this method, a character's support for a phylogenetic hypothesis, its apparent phylogenetic signal, is greatest when the amount of implied homoplasy is most remarkably small given background knowledge alone. Because evolutionary rates are not assumed to be slow, background expectations for character length can be derived through modeling complete dissociation between branching pattern and character state assignments. As in unweighted parsimony, SE holds that fewer required evolutionary steps in a character indicates stronger support for a tree. However, in SE, the relationship between steps and support differs by unlabeled tree topology and character state distribution. Strongest evidence is contrasted in detail with both unweighted parsimony and Goloboff's method of implied weights. An iterative process is suggested for incrementally resolving a phylogenetic hypothesis while conducting cladistic analyses at increasingly local levels. © 1999 The Willi Hennig Society.