Support for a monophyletic lemuriformes: overcoming incongruence between data partitions.

The Malagasy lemurs (Lemuriformes), represented today by 32 extant species (Mittermeier et al. 1994), are thought to have diverged approximately 60 MYA from their last common ancestor with the Lorisiformes in Africa and Asia (Yoder et al. 1996). Despite a great deal of morphological and molecular data, the family-level relationships—and even the monophyly—of the Lemuriformes remain controversial (e.g., Tattersall and Schwartz 1974; Cartmill 1975; Szalay 1975; Dene et al. 1976; Sarich and Cronin 1976; Dene, Goodman, and Prychodko 1980; Schwartz and Tattersall 1985; MacPhee and Cartmill 1986; Adkins and Honeycutt 1994; Yoder 1994; Del Pero et al. 1995; Porter et al. 1995; Yoder et al. 1996; Yoder, Vilgalys, and Ruvolo 1996; Stanger-Hall 1997). In this paper, we focus on resolving a remarkable pattern of incongruence between codon positions found in two mitochondrial DNA genes, cytochrome b (Cyt b) (Yoder, Vilgalys, and Ruvolo 1996) and cytochrome c oxidase subunit II (COII) (Adkins and Honeycutt 1994). To test the hypothesis that first and second codon positions of these genes are misleading in their support for a diphyletic Lemuriformes (Yoder, Vilgalys, and Ruvolo 1996), we collected partial sequences of the mitochondrial large ribosomal subunit (16S) from 17 species and subspecies of the Lemuriformes. Our sequences represented the 39 portions of the gene (33%), which are known to be far more conserved than the 59 portions (Hillis and Dixon 1991). This study represents the largest taxon sampling to date of the Lemuriformes for any gene. We chose the 16S gene, because, as a mitochondrial gene, it is guaranteed to have had the same history as the two protein-coding genes, but, as a ribosomal gene, it presumably operates under different selective constraints. As many taxa as possible were analyzed in an effort to resolve the long deep branches found in previous analyses with fewer taxa. After analyzing the 16S gene by itself, we used an iterative procedure to investigate whether the incongruence between data partitions is aggravated by inadequate phylogenetic reconstruction methods (Cunningham 1997). Three different phylogenetic reconstruction models were applied to the data using PAUP* 4.0d63 (written by D. L. Swofford, Smithsonian Institution): the best-fit model under maximum likelihood (ML), equally weighted parsimony, and six-parameter parsimony. The