Testing Phylogenetic Methods with Tree Congruence: Phylogenetic Analysis of Polymorphic Morphological Characters in Phrynosomatid Lizards

Abstract

Congruence between trees from separately analyzed data sets is a powerful approach for assessing the performance of phylogenetic methods but has been applied primarily to the analysis of molecular data. In this study, different methods for treating polymorphic characters were compared using morphological data from phrynosomatid lizards. Clades were identified that are both traditionally recognized and supported by recent molecular analyses, and species were sampled from these clades to make three “known” phylogenies of eight species each. The ability of different methods to estimate these “known” phylogenies with a finite sample of characters was tested. The phylogenetic methods included eight parsimony methods for coding polymorphism, three distance approaches (UPGM A, neighbor joining, and Fitch– M argoliash) applied to two genetic distance measures (Nei's and the modified Cavalli-Sforza and Edwards chord distance), and continuous maximum likelihood. The effects of excluding polymorphic characters and character weighting (a priori and successive) were also tested. Among the different parsimony approaches, the fixed-only method (excluding all polymorphic characters) performed relatively poorly, whereas the frequency method (including all polymorphic characters) performed relatively well. However, frequency-based distance methods consistently outperformed parsimony, especially with a small sample size (n = 1 individual per species). These results agree closely with those from recent simulation studies of polymorphic data and argue against the common practices of excluding polymorphic morphological characters, ignoring the frequencies of traits within species, and the exclusive use of parsimony to analyze morphological data.