The Origin(s) of Modern Amphibians: A Commentary

Abstract

Anderson(2008)recentlyreviewedthecontroversialtopicofextant amphibian origins, on which three (groups of)hypotheses exist at the moment. Anderson favors the‘‘polyphyly hypothesis’’ (PH), which considers the extantamphibians to be polyphyletic with respect to many Paleo-zoic limbed vertebrates and was most recently supported bythe analysis of Anderson et al. (2008). Another is the‘‘temnospondyl hypothesis’’ (TH—lissamphibians nestedwithintemnospondyls),mostrecentlysupportedbyRutaandCoates (2007). We prefer the ‘‘lepospondyl hypothesis’’(LH—lissamphibians nested within ‘‘lepospondyls’’; mostrecently supported by Vallin and Laurin 2004 and Marja-novic´ and Laurin 2008a). We would like to clarify importantpoints that were not discussed in Anderson’s review, or forwhich crucial arguments were left out.Anderson (2008) argues that most molecular dates favorthe PH because they suggest a Devonian or Early Car-boniferous diversification of Lissamphibia. This is inac-curate, since the confidence intervals of the dates obtainedby Hugall et al. (2007) range from Early Carboniferous toMiddle Permian, and our own molecular dating suggests aPermian origin. Indeed, three methods (molecular dating, apaleontological supertree and a confidence interval on thestratigraphic range of Lissamphibia) all hint at a Permian or(less likely) a Late Carboniferous origin of Lissamphibia(Marjanovic´ and Laurin 2007, 2008b).Citing Schoch and Milner (2004), Anderson (2008,p.234) argues that the LH is mainly supported by loss char-acters, and that this is problematic ‘‘given the relative easethat these losses can arise via paedomorphosis, whichappears to evolve repeatedly.’’ This is especially surprisingbecause we count (Supplementary Table 1) about fifty losscharacters in the matrix by Anderson et al. (2008)—morethan one out of five characters—, including several thatdescribethelossofbones thatossified lateintheontogenyofbranchiosaurids (Schoch 1992) and/or the ai¨stopod Phlege-thontia (Anderson 2002) and are absent in lissamphibians.Furthermore, Anderson’s remark amounts to criticizingthe use of loss characters simply because they could behomoplastic. Yet, Anderson (2008) emphasizes develop-ment characters such as digit development and skull ossi-fication order, which are known to be homoplastic. Forinstance, under Anderson’s version of the PH, the similaritybetween the digit development orders of the branchiosauridtemnospondyl Apateon and the urodeles is either conver-gent, or homologous between these two taxa but reversed inanurans; indeed, Johanson et al. (2007) suggest that thedigits of tetrapods are homologous to the ‘‘radials’’ of othersarcopterygians and find the ‘‘radials’’ of the Australianlungfish to develop independently of the rest of the forelimb(pectoral fin), like in urodeles and Apateon (and unlike inanurans and amniotes, where the limb chondrifies in a strictproximal-to-distal sequence), strongly suggesting that theurodele-Apateon pattern is plesiomorphic—regardless ofwhether the PH, the TH, or the LH is (closest to) correct.Anderson (2008, p. 242) furthermore mentions that ‘‘thepattern of cranial ossification […] has compared very clo-sely with the sequence of cranial ossification seen in sala-manders’’ (making explicit on the next page that these