77 



the requirements of very different topologies would require a larger amount of extra 

 homoplasy. But, in view of the unexpectedly low number of equally most parsimonious 

 solutions (and slightly less than most parsimonious; see skewness analysis in the 

 Statistical Tests section) entailed by this supposedly more flexible data matrix, the effects 

 of polymorphic data on parsimony analyses need to be investigated further. 



Missing taxa (Fig. 15) 



The impetus for this analysis initially stemmed from taxonomic considerations within the 

 Phocini (plus Erignathus). In a preliminary parsimony analysis, all of the constituent 

 genera except for Phoca (sensu stricto) were monophyletic. However, the presence of 

 Phoca largha in this study is contingent on the somewhat debatable species status granted 

 it here. Subordinating P. largha as a subspecies of Phoca vitulina (as advocated by Scheffer 

 1958; Burns 1970; Shaughnessy 1975; Baram et al. 1991) would effectively render the 

 now monotypic Phoca monophyletic. But, as Arnold (1981) suggests that the absence of 

 some taxa might seriously affect the outcome in a low level cladistic analysis, we desired 

 to investigate the effects of the removal of P. largha from the analysis. Subsequent analyses 

 also saw the individual removal of Cystophora, Erignathus, Lobodon, and Ommatophoca 

 in order to view the effects of their deletion. Cystophora was chosen due to its basal 

 position within the phocines, combined with its strong tendency to join the monachines. 

 Erignathus was likewise selected due to the novel position indicated for it here, and 

 because of its position within the labile Phocini clade (as with Phoca largha). The removal 

 of Lobodon provided an insight into the effects of the deletion of a rather topologically 

 unspectacular and undistinguished taxon. Finally, Ommatophoca was deleted as it appears 

 to be one the more unstable taxa in the otherwise fairly robust Monachinae (see 

 Unweighted analysis below). 



As would be expected, the removal of each taxon yielded a shorter solution than that 

 obtained when all taxa were present. Single most parsimonious solutions were the norm, 

 with only the matrix lacking Erignathus producing dual equally most parsimonious 

 solutions. However, the degree of shortening in each case was much greater than would 

 be achieved by merely "pruning" the single species branch in question from the tree. For 

 example, the removal of Phoca largha, which effected virtually no topological changes, 

 resulted in a solution some 1,452 steps (22 corrected steps) shorter than the overall 

 solution. As the length of the branch leading to P. largha was only 383 steps (seven 

 unweighted steps), the removal of taxa must decrease homoplasy elsewhere in the tree. 

 This is demonstrated by the generally reduced branch lengths around the region that the 

 removed taxon formerly occupied (compare Figs.5C and 15 for all removed taxa). Branch 

 lengths within the outgroups were virtually unchanged. As well, the various goodness-of- 

 fit statistics are slightly altered to reflect this decrease in the overall level of homoplasy. 

 Although major changes in topology were generally not evident in this analysis, the 

 removal of taxa again demonstrated how the topology of one region of a tree can affect 

 the topology of another, supposedly distinct region. This was especially evident with the 

 removal of either of the two monachines. The deletion of either Lobodon or Ommatophoca 

 produced minor, if any, topological changes within the monachines, but generated more 

 substantial changes within the phocines (Figs.l5D and E). Again, these latter alterations 



