165 



possible, preference was given to those sources employing growth curves (e.g., McLaren 

 1993) or statistics (e.g., Ralls 1976), as they presumably would be less prone to sampling 

 effects than isolated descriptions. As well, lengths were preferentially used to judge size 

 rather than the more commonly used, and often more appropriate, measure of mass. As 

 noted by McLaren (1993), weights are often not recorded for pinnipeds, and the high 

 seasonal variation in pinniped blubber stores makes mass a less reliable criterion for 

 judging size in these animals. 



The plesiomorphic condition among the Caniformia is for the male to be larger than the 

 female. This is found universally among all outgroups and is retained into the basal 

 members of each phocid subfamily. Beyond this, the trend within each subfamily is for 

 parallel derivations of monomorphism. This occurs for the phocine clades Erignathus, 

 Histriophoca, plus Pagophilus; and Pusa hispida plus Pusa sibirica. As well, it 

 characterizes the lobodontines plus Monachus spp., with Hydrurga and Ommatophoca 

 convergently deriving the condition whereby the female is the larger sex. 



Summary 



Several features of the character set presented above (and, indeed, the data matrix as a 

 whole) need to be stressed. This data matrix is one of the most comprehensive ever 

 compiled (with respect to the number of taxa and morphological characters) to answer the 

 question of phocid phylogeny. To our knowledge, the only other comparable matrix is that 

 of Berta & Wyss (1994). But, beyond its sheer size, another advantageous feature of the 

 matrix lies in the wide range of osteological characters that were employed, originating 

 from virtually all regions of the organism. Most of these characters appear to be 

 phylogenetically informative (but see below), including many of the 28 that were excluded 

 from the analysis. In most cases, these latter characters were excluded as they were deemed 

 to be redundant when considered in the light of other characters. This primarily reflects 

 presence of "character pairs", where the first character examined for the presence of a 

 feature, while the second detailed the morphology of that feature. However, in some cases, 

 redundant characters reflected an inferior coding scheme (characters #18, 32-36, and 100). 

 Only five characters were excluded a priori because we felt that they were of a dubious 

 nature (characters #1,2, 90, 124, and 129). However, of those apparently phylogenetically 

 informative characters that remained, many could be improved still further (i.e., have their 

 information content increased) through recoding to remove the apparent a posteriori cases 

 of homoplasy (i.e., non-homologous similarity) within them as advocated by Hennig 

 (1966). (It should be noted that very few characters, discounting within terminal changes, 

 can be seen to possess a "clean" distribution free of any homoplasy.) Largely, this involves 

 a refinement of the coding scheme to distinguish between morphologically very similar, 

 but non-homologous character states (e.g., see character #116). 



Several discrepancies can be noted between historical observations of some characters and 

 our observations. No doubt, this can be traced, in part, to the unusually high intraspecific 

 variation among cranial features in pinnipeds. Thus, both sets of observations might well 

 be accurate, but only with respect to the specimen(s) that they were obtained from [and 

 within the bounds of the subjective judgement of different researchers for many qualitative 



