Giles: Multivariate Analysis of Pleistocene and Recent Coyotes 383 



groups to which the specimen might belong (table 3). The differences between the 

 results for the single specimen and those for one of the groups are squared and 

 summed. If this sum is not significant at the 5 per cent level when entered in a 

 c/ii-square table, the single specimen might belong to this group. If the same pro- 

 cedure with the other group indicates that the single specimen might belong to it 

 also, a discriminant function must be set up. The value midway between those of 

 the two groups is used in a new linear equation derived from the groups, pre- 

 suming them to represent populations of the same size. The value for the specimen 

 will fall on one side or the other of this value when substituted into the equation, 

 indicating the specimen's affinities. 



DISCUSSION 



The possibility of varying rates of evolution always makes the amount of morpho- 

 logical difference an uncertain indicator of evolutionary divergence, which is the 

 basis for classification. Yet since the coyote remains examined in this study repre- 

 sent a very short span of geologic time and since stratigraphy offers no particular 

 aid in pinpointing their evolutionary divergence, to draw inferences from morpho- 

 logical differences is both plausible and necessary. The possibility that similarities 

 may be the result of evolutionary convergence from a polyphyletic origin is present 

 in every taxonomic analysis, but the coyotes arouse no suspicion that this is true 

 of them. 



Stirton (1955:28) has modified Mayr, Linsley, and Usinger's (1953) definition 

 of subspecies to include paleontological specimens: subspecies are "geographically 

 and/or stratigraphically defined aggregates of local populations which differ mor- 

 phologically from other such subdivisions of a species." The question arises, of 

 course, as to the amount of separation necessary in this purely morphological 

 scheme to justify specific or higher ranking. There is no valid reason to consider 

 a statistical difference between two groups as taxonomic per se. The boundaries of 

 the species may be the most objective in taxonomy, but they are by no means un- 

 equivocal even in neozoological systematics, where the number of characters and 

 the number of specimens from which the inference is made are usually much 

 greater than in paleontology. 



Rollins (1953) has suggested that the pertinence of possible collocations be 

 judged by reference to well-defined units within the group under consideration; 

 in other words, by use of what he calls a "species standard." When available, such 

 a standard seems preferable to a procedure that rather inflexibly assigns sub- 

 specific rank to groups having characters that overlap and specific rank to those 

 that do not. Mayr (1942:172) has observed that the lower taxonomic groupings 

 grade almost insensibly into those above and below. Stirton (1955) has demon- 

 strated that the often-used criterion of tooth-row length is not sufficient to dis- 

 tinguish established species of the white-footed deer mouse Peromyscus because 

 there is overlap among species. Colbert and Hooijer (1953) utilized a species- 

 standard approach in classifying fossil remains of the bamboo rat Bhyzomys from 

 China. 



Recent coyotes can be used as a "subspecies standard" in guiding taxonomic as- 

 signment of fossil specimens. The morphological distance between fossil forms 



