70 



overall matrix can accommodate them with a minimal amount of extra homoplasy (see 

 Overall conclusions - possible effects of polymorphic data below). 

 Despite these minimal increases in length, some clear patterns did arise in this analysis 

 (Tab. 3). Of the paired constraint trees, those advancing a diphyletic Pinnipedia always 

 resulted in a longer and, as measured by the corresponding bootstrap frequency, more 

 weakly supported solution (see also Tab. 2). Additional support for a lutrine affinity for 

 the pinnipeds under this data matrix is provided by the observation that the lutrines always 

 form the sister group to the phocids and/or pinnipeds as a whole (Fig. 13), unless other 

 sister taxa were specifically constrained for. Numbers of most parsimonious trees were 

 again very low for such a large data set (Tab. 3), once again roughly indicative of good 

 resolving power (but see Overall Parsimony Analysis; Hillis & Huelsenbeck 1992). 

 More importantly, relationships within the phocids were identical with those of the overall 

 solution in most cases, or, at most, only slightly altered within the Phocini (plus 

 Erignathus). Only those constraint trees supporting an ursid ancestry for the pinnipeds 

 ("ursid - monophyly", "ursid - diphyly", and "ursid - odobenid") produced major disrup- 

 tions within the phocids. Again, this was limited to the Phocini (plus Erignathus), and 

 amounted to a basal shift of Histriophoca and Pagophilus to form successive sister taxa 

 to the clade of Erignathus, Phoca spp., and Pusa spp. Thus, the possibility of a 

 symplesiomorphic relationship between Histriophoca and Pagophilus (see de Muizon 

 1982a) only appears to arise under the assumption of an ursid affinity for the phocids. 

 Of the individual solutions (see Fig. 13), the constraint of a non-monophyletic Otarioidea 

 - as advocated by Wyss (1987), Berta (1991), Wyss & Flynn (1993), and Berta & Wyss 

 (1994) - converged on the same solution as that resulting from a forced Odobenus-^\\oc\d. 

 pairing. However, while this common solution might point to some affinity between 

 Odobenus and the phocids [it only required an extra 400 steps (six corrected steps) over 

 that of the overall solution], such a pairing has very weak support in the data matrix 

 (bootstrap frequency of 6%). 



Altogether, the negligible increases in length resulting from the constraint of the various 

 alternative outgroup relationships, coupled with their minimal effects on internal phocid 

 phylogeny point to the potential bias from assuming one outgroup taxon over another for 

 the phocids as being very small. The apparently inherently less stable Phocini (plus 

 Erignathus) notwithstanding (see Statistical Tests section), the selection of any major 

 arctoid lineage (e.g., lutrines, mustelids, otarioids, or ursids) will apparently all give 

 roughly the same set of internal relationships for the phocids, as was also claimed for the 

 phocines by Perry et al. (1995). This finding might ensue from the early history of the 

 arctoids, whereby the fact that all of these lineages (including the phocids) were diverging 

 at about the same time (Sarich 1976; Wayne et al. 1989; CA. Repenning, pers. comm.) 

 largely renders the designation of sister taxa as irrelevant, or even erroneous. Thus, the 

 supposition herein of a lutrine affinity for the pinnipeds might be artif actual (i.e., a 

 consequence of this particular biased data set), and an artificial resolution of a real 

 polytomy. Although an intriguing possibility, and preliminarily substantiated by Perry et 

 al. (1995), this question should remain open until more paleontological and/or molecular 

 evidence is accumulated. 



