Sorex Mensural Discrimination 23 



(1.58, 16.90, and -5.49, respectively), sum the three products, and 

 add a constant (-17.25). The resulting value is the specimen's discriminant 

 score. If the score is greater than zero, the specimens in assigned 

 to S. cinereus, otherwise it is assigned to S. longirostris. The average 

 discriminant score for S. cinereus is 3.06, and the average for S. longirostris 

 is -3.18. 



Discriminant analysis using cranial and external measurements 

 for the two species correctly classified all specimens with two characters, 

 LU34 and tail length, and this bivariate comparison can be used to 

 identify new specimens without transformation (Fig. la). Similarly, 

 using only external characters, all specimens can be identified to species 

 with a bivariate comparison of body and tail lengths (Fig. lb). Using 

 only cranial characters, specimens can be correctly classified with a 

 high degree of probability (99.5%) with a bivariate comparison of LU34 

 and BM2 (Fig. lc). 



VALIDATION 



Validation of the model separating the species showed that the 

 results were stable, as 193 of 200 trial runs produced 100% correct 

 classification. The seven trails producing errors had one misclassification 

 each; therefore, of 7,579 individuals classified in the validation process, 

 only seven were classified incorrectly. All errors were the misclassification 

 of S. cinereus specimens. 



We validated the utility of this model based upon six test specimens 

 of each species from localities not used in developing the model and 

 with a more conventionally available measuring device (i.e., dial calipers 

 and dissecting microscope). The discriminant analysis was sufficiently 

 robust that all specimens were correctly identified. 



CONCLUSION 



Sorex cinereus and S. longirostris can be distinguished on the 

 basis of any one of three bivariate plots using untransformed data 

 (Fig. 1) or by discriminant analysis. The results of our univariate compari- 

 sons are similar to those of French (1980c) and Greer (1989); we 

 observed a high degree of overlap in all but one mensural character 

 (tail length). Possibly, the separation of the two species by cranial 

 characters in our study is a reflection of the finer scale of measurement 

 permitted by computer assisted video imaging. We should note that 

 regional differences in the morphology of both S. cinereus and S. longirostris 

 might limit the effectiveness of the methods and characters used by 

 us in mensural discrimination of these two species in areas other than 

 the Southeast. 



