Sorex Mensural Discrimination 21 



Very little of the total variance could be attributed to the repeated 

 measures (range of to 24%, x = 5.5%), suggesting that with careful 

 calibration the video system provided highly repeatable measurements. 



In univariate comparisons of the sexes, only the length of unicuspids 

 3 and 4 (LU34) in S. longirostris differed significantly (P = 0.031). 

 Males and females averaged 0.71 mm (SE = 0.0064, n = 49) and 

 0.69 mm (SE = 0.0091, n = 20), respectively (Rice 1989). When this 

 character was examined within regions, sexes did not differ significantly 

 (southern sample, P = 0.23; northern sample, P = 0.10). 



In a few cases, differences among age classes within species, 

 regions, and sexes were individually, but not collectively, significant 

 (a < 0.05). The only consistently significant (P < 0.01) difference 

 among age groups was length of first unicuspid (LU1) which tended 

 to decrease in magnitude with increasing age. 



In univariate analysis of morphological variation, all characters 

 except body length, breadth of third unicuspid (BU3), and breadth 

 of fourth unicuspid (BU4) differed significantly (P < 0.001) between 

 species. For all characters that showed significant differences, except 

 breadth across second molars (BM2), S. cinereus was larger than S. 

 longirostris. For character BM2, the size of S. longirostris exceeded 

 S. cinereus. In all cases except tail length, however, the range of 

 measurements for both species overlapped (Table 1). 



Multivariate analysis using cranial measurements was successful 

 in correctly identifying all specimens of the two species. However, 

 the geographic origin of only 79% of the northern and southern specimens 

 could identified correctly. For S. longirostris, 11 southern and 11 northern 

 specimens were incorrectly classified into the opposite geographic group; 

 for S. cinereus, 7 southern and 12 northern specimens were incorrectly 

 classified. Such a measure of geographic variation, perhaps clinal, 

 was expected. 



Discriminant analysis using cranial measurements for the two 

 species correctly classified all specimens with three characters, LU34, 

 BM2, and CBL, in order of inclusion into the model. Standardized 

 canonical discriminant function coefficients were 0.60, 0.72, and 

 -0.73 for LU34, BM2, and CBL, respectively. Pooled within-groups 

 correlations between discriminating variables and canonical discriminant 

 functions, variables ordered by size of correlation within function, were 

 0.69, -0.25, and 0.53 for these characters, respectively. 



Unknown specimens can be identified to species with this latter 

 model using unstandardized canonical discriminant function coefficients 

 for the three variables. To do so, measure the unknown specimen for 

 CBL, LU34, and BM2, then multiply each measurement by its coefficient 



