Testing Species Phylogenies and Phylogenetic Methods with Congruence

We assessed the utility of congruence and multiple data sets to test species relationships and the accuracy of phylogenetic methods. The ongoing controversy about whether to combine data sets for phylogenetic analysis was evaluated against the naturalness of different types of data (as com­ monly recognized by systematists) and character independence. We defend the recommendation that independent data sets (defined in terms of process partitions; sensu Bull et aI., 1993, Syst. BioI. 42:384-397) should rarely be combined but should be kept separate for phylogenetic analysis because their independence increases the significance of corroboration. Trees of natural taxa, well supported by many independent lines of evidence, should be used in the same way as the known phylogenies of simulations and of certain laboratory and domesticated groups, i.e., as standards for evaluating the accuracy of different phylogenetic methods. Although compromised by their imperfect reliabilities, such tests using well-supported trees of wild taxa provide important reality checks on the conclusions of the other two approaches by encompassing more of the complexity and diversity of natural systems and their evolutionary processes. In this way, a combination of testing with the well-supported trees of natural groups, with simulations, and with those laboratory and domesticated taxa with known phylogenies is most likely to prove effective in establishing the strengths, weaknesses, and assump­ tions of different phylogenetic methods. (Accuracy; taxonomic congruence; character congruence; process partitions; character independence; well-supported trees; phylogenetic methods.)