ROBUST PHYLOGENIES AND ADAPTIVE RADIATIONS - A CRITICAL EXAMINATION OF METHODS USED TO IDENTIFY KEY INNOVATIONS

摘要

A critical examination of a recent article by Brooks and McLennan has led to the conclusion that their phylogeny of parasitic Platyhelminthes is based on ''synapomorphies'' that are clearly wrong or superficial similarities whose homology is unlikely. Using characters likely to be homologous and DNA analyses, it has been shown that Temnocephalida is not the sister group of the parasitic Platyhelminthes (Neodermata), Udonellidea is unlikely to be the sister group of the other Neodermata, the sister group of the Neodermata is unknown but likely to be a large taxon comprising several turbellarian groups, no evidence indicates that the Arthropoda are the original hosts of the parasitic Platyhelminthes, the possibility cannot be ruled out that species poverty of some groups of parasitic Platyhelminthes is an artifact, and ''key innovations'' supposedly explaining the great species richness of the parasitic Platyhelminthes and some taxa within them are based on insufficient or faulty evidence. Based on an examination of the recent literature, I suggest the following steps to identify key innovations: first, establish phylogenetic trees using characters likely to be homologous on the basis of established homology criteria; second, examine whether trees are ''significantly'' unbalanced using null models; finally, use multiple testing to tentatively identify putative key innovations. Retrospective studies of processes that have led to radiations in the past will in many cases be impossible, and the best one can hope for is a tentative identification of putative key innovations by multiple testing. if multiple testing cannot be done, then even a tentative identification is impossible, and one is left with a ''plausible suggestion.'' Concerning the practice of phylogenetic systematics, it is emphasized that phylogenetic analyses using a few characters whose homology has been made likely by careful assessment are more likely to reveal phylogenetic relationships than analyses using large numbers of unassessed characters. [References: 88]