33 



constraint trees as tested are presented in Figs. 3 and 4 respectively. Two additional ingroup 

 trees, corresponding to the solutions of the unweighted and condensed analyses (see 

 below), were also tested (Fig.40 and P). 



As with Bremer support, constraint analyses suffer from a lack of any explicit statement 

 on how much longer a tree must be for it to be rejected on statistical or other grounds 

 [but again, see Cavender (1978, 1981)]. Thus, we again employ this analysis more in a 

 relative fashion, using it to distinguish between more weakly and more strongly supported 

 solutions. Whenever possible, bootstrap frequencies for the clade being examined were 

 also added as a second, albeit similarly approximate, line of evidence. 

 All searches employed the heuristic search pattern described above, except for the ingroup 

 constraint trees "de muizon" and ■"condense", which were analyzed using exhaustive and 

 branch and bound search algorithms respectively (both collapsing zero-length branches), 

 both of which guarantee an optimal solution. 



Missing taxa 



On the suggestion of Arnold (1981) that missing taxa may have a detrimental effect on 

 low level cladistic analyses, five phocid taxa (Cystophora, Erignathus, Lobodon, 

 Ommatophoca, and Phoca largha) were selectively deleted and the analysis re-run in order 

 to view the effects of their individual removal, if any. These taxa were selected on the 

 basis of their topological position, various tendencies elucidated by other analyses, or for 

 historical considerations (see Comparative Tools section for full details). As the removal 

 of these taxa alters the intrinsic properties of the data matrix, comparisons with other 

 results are primarily limited to comparisons of gross topological changes and various 

 goodness-of-fit-statistics. All searches were conducted using the heuristic search pattern 

 described above. 



Condensed analysis 



Historical considerations of phocid phylogeny show a strong tendency to assume the 

 monophyly of some higher taxa. However, as this is the first species-level cladistic analysis 

 of the entire family to be performed, some of these assumptions of monophyly will not 

 have been rigorously tested before now. Thus, in order to view the possible historical 

 effects of assumed monophyly on phocid phylogeny, we re-ran the analysis with several 

 species collapsed into one of four higher taxa: the genera Mirounga, Monachus, and Phoca 

 (sensu Burns & Fay 1970), and the tribe Lobodontini. 



Although the paraphyly of three of these four taxa has been suggested, it has not been 

 widely accepted in each case. Certainly, the strongest and almost indisputable case is for 

 Phoca (sensu Burns & Fay); however, this taxon is almost universally recognized today. 

 Only Wyss (1988a) has provided evidence for a paraphyletic Monachus, something that 

 has not been re-examined to date [although it is endorsed by Berta & Wyss (1994)]. It 

 has been hinted that the Lobodontini may be polyphyletic, or even paraphyletic. but only 

 with the inclusion of fossil taxa (Hendey 1972; McLaren 1975; Ray 1976a; Berta & Wyss 



