58 



Successive approximations (Figs. 9 and 10) 



The a posteriori weighting of the characters based on one of three goodness-of-fit statistics 

 (CI, RI, or RC) converged on either of two solutions (Fig. 9). Each solution, when 

 constrained with the original set of inversely weighted characters, was only very slightly 

 longer than the overall solution at 69,842 steps (= 7 steps / 1 corrected step longer). 

 Topological differences between these two solutions, and between either and the overall 

 solution were found solely within the Phocini (plus Erignathus) and limited to the 

 interrelationships between Phoca spp. and Pusa spp. This conflict is reflected by the strict 

 consensus tree of both solutions [Fig.lOA; the majority rule solution (Fig.lOB) is the same 

 as that produced by either the RI or RC] where the species of both genera are the sole 

 members of a completely unresolved polytomy. 



Support analysis (Fig. 11) 



Interpreting the results of the support analysis varies according to the form of consensus 

 tree that is viewed, with the strict and majority rule consensus algorithms treating 

 conflicting solutions relatively more severely and more forgivingly respectively (see 

 Methods and Materials). 



The strict consensus trees show a steady decrease in resolution as increasingly 

 homoplasious solutions are retained. At only one corrected step longer than the most 

 optimal solution (Fig. IIA), resolution within the Phocini (plus Erignathus) is almost com- 

 pletely lost. Resolution is also lost for the lutrines, reflecting the equally large tendency 

 for these two taxa to form a monophyletic sister group to the pinnipeds (see above). An 

 increase of an additional corrected step (Fig. IIB) shows a partial degradation of 

 lobodontine relationships, which becomes complete at the next step (Fig.llC). At this 

 latter step (three corrected steps longer), resolution is completely lost for the Phocini (plus 

 Erignathus) as well, and the integrity of the monachines is also lost. Finally, at the limits 

 of the analysis (Fig. 1 ID), almost all structure within the phocids is lost. Only the clades 

 of Mirounga spp. and Monachus spp. (and within Monachus) retain unanimous support. 

 As well, structure is lost for the pinnipeds as a whole, with the otarioids and phocids. 

 although still distinct clades, forming a polytomy with the lutrines. Another polytomy was 

 also formed between Procyon, Martes, and the lutrine-pinniped clade. 

 In contrast, the majority rule consensus trees show virtually complete and unaltered 

 resolution, even at the limits of the analysis. Only a progressive degeneration of Phocini 

 (plus Erignathus) structure was observed, although the support within the phocids as a 

 whole (as visualized by the percentage of solutions supporting each node) gradually 

 decreased with increasing length (except for most strongly supported species clusters). 

 Among outgroup relationships, only support for the Lwfra-pinniped and Martes-\\xtnnt 

 pairings were observed to decrease, albeit only slightly, with increasing length. 



Overall conclusions 



Altogether, the findings of these statistical tests largely corroborate those of the parsimony 

 analysis conducted in the previous section. This could be an artifact of all these tests 



