Reevaluation of the largest published morphological data matrix for phylogenetic analysis of Paleozoic limbed vertebrates
Marjanović et al
The largest data matrix for phylogeny of early limbed vertebrates (Ruta M, Coates MI. 2007. J. Syst. Palaeont. 5:69–122) has supported controversial hypotheses; e.g., it has recovered Seymouriamorpha, Diadectomorpha and (in some trees) Caudata as paraphyletic and found the “temnospondyl hypothesis” on the origin of Lissamphibia (TH) to be one step more parsimonious than the “lepospondyl hypothesis” (LH). Scrutiny of the matrix reveals thousands of suboptimal scores (many clearly due to typographic and similar errors) as well as logically linked (redundant) characters, characters with only one described state, and even cases where taxa were scored after presumed relatives. Moreover, all characters – even obviously continuous ones – were unordered, effects of ontogeny were not sufficiently taken into account, and the authors mostly excluded data published after 2001, even their own. Our revised version – we document and justify all changes – yields much longer trees with a different topology, e.g. monophyletic Caudata, Diadectomorpha and (sometimes) Seymouriamorpha, Ichthyostega more rootward than Acanthostega, Anthracosauria more rootward than Temnospondyli, and the LH, which is 10 steps more parsimonious than the TH and 15 more than the “polyphyly hypothesis” (PH). Bootstrap values, though, are low, and few of the topologies are statistically distinguishable. For another set of analyses, we added 48 OTUs to the original 102. This destabilizes parts of the tree, e.g. the relationships of Anthracosauria and Temnospondyli. However, many of the added taxa have a fully resolved position or nearly so; this concerns the well-known Chroniosaurus (sister to a clade containing Solenodonsaurus, Seymouriamorpha, Diadectomorpha, Amniota and Amphibia), but also isolated lower-jaw material from the Devonian and Carboniferous. Despite the addition of Gerobatrachus, Micropholis and Tungussogyrinus and the extremely peramorphic salamander Chelotriton, the difference between LH and TH only shrinks to 9 steps, that between LH and PH to 13 steps. The “lepospondyl” Brachydectes is neither found as sister to Lissamphibia nor in the “microsaur” grade. Bootstrap values plummet, though, and all three hypotheses become statistically indistinguishable at p = 0.05. We then duplicated all analyses after coding all losses of bones as irreversible. Anthracosauria is then consistently placed more rootward than Temnospondyli; given the original taxon sample, the LH is 12 steps shorter than the “temnospondyl hypothesis” and 17 steps shorter than the PH, while the expanded taxon sample makes the LH 10 steps shorter than the TH and only 12 steps shorter than the PH. More robust results could likely be obtained by adding the many characters used in other analyses or discussed in the literature. We discuss phylogeny, approaches to coding, and certain character complexes, in particular the supposed middle ear of temnospondyls.