Perspective: Parsimony, explanatory power, and dynamic homology testing

Abstract

Listen

Translate article

Parsimony remains one of the most widely used optimality criteria in phylogenetic systematics. There are many practical reasons for this, including parsimony's intuitive simplicity, applicability to diverse character data, computational speed, and unique statistical properties as a nonparametric method. The epistem‐ological argument for parsimony as maximising explanatory power by minimising ad hoc hypotheses of homoplasy (MAH) has been accepted by many workers. However, in addition to philosophical problems, this justification of parsimony can lead to contradictory results under dynamic homology analysis. Because of the decoupling of homoplasy and tree length under dynamic homology analysis, trees of the same length but different amounts of homoplasy would have different parsimony scores, while trees of different lengths but the same amount of homoplasy would be considered equally parsimonious. An alternative epistemological justification of parsimony based on the anti‐superfluity principle (ASP) leads to the minimisation of hypothesised transformation events, thereby treating trees of the same length as equally parsimonious regardless of the homoplasy they entail. In response, it has been charged that the arguments against the MAH justification are incorrect and that the ASP justification is based on a tautology. We show that the defence of MAH employs logical errors and contradictions and fails to validate that justification of parsimony. Further, we show that the ASP justification is not tautological and that this charge is due to a misconception of the dynamic nature of hypothesis testing in phylogenetic systematics.