175 



Therefore, if the desired signal in a phylogenetic analysis is the "overall signal" (see 

 Statistical Tests section: although the "'phylogenetic signal" could equally be one of the 

 "regional signals"), then phylogenies based largely on a set of characters from a single 

 localized region should be avoided. Although the morphologies of any characters obtained 

 from this one region will likely be strongly correlated with one another, this correlation 

 can very often arise from some non-phylogenetic process [e.g., the general resemblance 

 of the snout in Cystophora and Mirounga spp.. which is a manifestation of the 

 morphological requirements of possessing a (convergently derived) inflatable proboscis]. 

 Again, of the two solutions mentioned above, the more practical at this point is to examine 

 a wide range of characters, possibly as a first step towards a total evidence approach. 

 However, this should be done with some caution. The very different signals from the 

 various, supposedly discrete anatomical regions beg the statistical question of whether 

 these separate matrices should have been "pooled" in the first place (see Bull et al. 1993), 

 a major limitation of the total evidence approach. Therefore, one would ideally want to 

 generate a set of randomly distributed characters, as we have argued previously (see 

 Statistical Tests section), but it is difficult to fathom how the degree of randomness of 

 such a set could be ascertained. In any case, when properly applied, either method should 

 increase the probability of generating a data matrix where the influence of the various 

 non-phylogenetic signals is minimized, allowing the presumably more widespread 

 phylogenetic signal to dominate and determine the resultant outcome. 

 It remains then for the final section to present a short summary of the phylogeny of the 

 phocids indicated in this study, including identifying the areas of stronger and weaker 

 support. Various selected implications of this phylogeny will also be examined, as will 

 the lines of research still required in the area of phocid systematics. 



DISCUSSION AND CONCLUSIONS 



We here recap this study by first presenting an overall summary of the phylogenetic 

 findings in light of the results of the many statistical tests performed and comparative 

 tools employed in the Statistical Tests and Comparative Tools sections. Possible sources 

 of error affecting the validity of any conclusions thereby reached will be analyzed, before 

 the remainder of this section is given to such miscellaneous topics as the taxonomic 

 implications of our proposed phylogeny of the phocids, concordance of our proposed 

 phylogeny with other lines of evidence (primarily biogeographical and fossil), and. finally, 

 possible future lines of research suggested by this study. 



Summary of results 



Virtually all of the analyses conducted in the Overall Parsimony Analysis, Statistical 

 Tests and Comparative Tools sections point to a single common pattern, which Faith 

 (1992) would equate with a form of Popperian corroboration. With respect to the overall 

 solution (Fig.5B), outgroup relations clearly possess the strongest support within the 

 cladogram, demonstrating alterations in their topology only when specifically forced to 

 do so. However, as indicated by the constraint analyses (see Comparative Tools section), 



