I 



47 



for this line of reasoning lies in the relatively limited number of synapomorphies 

 supporting most nodes in this region, in addition to the polytomy, which is more typical 

 in regions of a cladogram where speciation has occurred via a rapid adaptive radiation 

 (Wagner 1992). The pattern may also be obscured somewhat by a large amount of parallel 

 evolution within the Phocini. 



Although the number of proposed systematic alterations for the monachines is greater than 

 for the phocines, this does not seem to present as great a problem. For the most part, this 

 is because the monachines have not been as well studied, possibly due in large measure 

 to the remoteness of most species with respect to the primarily northern hemisphere 

 population of phocid researchers, combined with the seemingly more intuitive relationships 

 of, and tribal allotment within, this subfamily. This seems to be especially true for the 

 lobodontines, whose internal relationships have never been studied in detail, and whose 

 monophyly (despite their obvious morphological differences) has seemingly never been 

 questioned due to their common and distinctive geographic range, coupled with a similar 

 lack of detailed examination. In the end, the novel monachine relationships advocated here 

 may be a consequence of the freedom allowed all species to form the most parsimonious 

 set of pairings, as opposed to previous studies which tended to constrain the monophyly 

 of one or more of the monachine tribes (see Condensed analysis in the Comparative 

 Tools section), particularly the Lobodontini. 



Finally, in contrast to the assertions of Wyss (1988a) and Berta & Wyss (1994), 

 convergences appear to be much more common than reversals in phocid evolution (see 

 Appendix E and Character Analysis), a pattern that holds even under ACCTRAN 

 optimization, where reversals are favoured. Of the homoplastic characters (and excluding 

 within terminal changes), 49 / 85 were convergent (numbers given as ACCTRAN / 

 DELTRAN), 16/5 were reversals, and 79 / 55 displayed both. Thus, reversals, when 

 present, were typically found together with convergences (although convergent incidences 

 of reversals only accounted for 45 / 29 characters of this subset). Nor was there a 

 discernible pattern of homoplasy [in contrast to Wyss (1988a) and Berta & Wyss (1994) 

 who indicate a distinct pattern of retrogression for the phocines], with convergences and 

 reversals spread throughout the phocids. 



Support for the overall solution 



Various indicators point to the "good resolving power" of the data set as a whole (see also 

 Statistical Tests and Comparative Tools sections). On a purely empirical basis, the data 

 set ran surprisingly "cleanly" (only two solutions) and quickly for such a large matrix (27 

 taxa and 168 characters). In part, this can be traced to the use of inverse character 

 weighting. While similar runs using unweighted characters (see Comparative Tools 

 section) produced only four solutions, analysis times were considerably longer, due to a 

 greater number of slightly less than most parsimonious solutions that needed to be searched 

 through. However, the common perception that a low number of most parsimonious 

 solutions implies a good quality to the data set may be an unsubstantiated claim, as some 

 studies suggest that this number is dependent upon the number of characters (and how 

 many states each possesses) and the number of taxa (Hillis & Huelsenbeck 1992; Lamboy 



