PERSPECTIVE Testing Phylogenetic Approaches with Empirical Data, as Illustrated with the Parsimony Method1

In the absence of certainty, well-supported phylogenies stand as our best estimates of the correct evolutionary relationships for a group (Miyamoto and Cracraft 199 1). Such phylogenies provide a solid foundation for understanding the origins and history of biological diversity and the forces responsible for them. However, what remains less obvious is that they can also serve as standards against which the results of different tree-building procedures can be compared to determine their accuracy. Approaches that lead to solutions congruent with well-supported topologies should be preferred over those that rarely do so. Thus, well-supported phylogenies fill the role of model trees in simulation studies, as the basis for comparison of phylogenetic reliability (Nei 1991; also see Mickevich 1978). In the present study, the evolutionary relationships of lipotyphlan insectivores (class Mammalia, infraclass Eutheria) are investigated with new mitochondrial DNA (mtDNA) sequences of the 12s ribosomal RNA (rRNA) gene. A single phylogeny based on parsimony analyses of these sequences is accepted as well supported according to different criteria, although an exception to this conclusion is noted. This exception forms the basis for an investigation of why an incorrect solution is obtained by the parsimony method in this particular case. The 12s rRNA gene sequences for the lipotyphlans Amblysomus hottentotus ( AHO; African golden mole, family Chrysochloridae ) , Atelerix albiventris ( AAL; hedgehog, Erinaceidae), and Blarina brevicauda (BBR, short-tailed shrew, Soricidae) were determined by a combination of the polymerase chain reaction (PCR) amplification and direct dideoxy sequencing of the asymmetrical products (Allard et al. 199 1) . These new sequences were aligned to each other and to outgroup representatives from three other eutherian orders: Bos taurus [ BTA; domestic cow, Artiodactyla (Anderson et al. 1982)], Homo sapiens [ HSA; human, Primates (Anderson et al. 198 1 )], and Rattus norvegicus [ RNO; rat, Rodentia (Gadaleta et al. 1989)]. These orthologues were chosen to provide alternative mammalian outgroups which have had a widespread use in comparative molecular studies. In the final alignment of the study group and outgroups (fig. 1)) gaps were included only when they saved two or more substitutions (Kraus and Miyamoto 199 1). Throughout the present study, gaps were treated as single differences regardless of their lengths. Parsimony analyses of the 12s rRNA sequences were conducted with the computer