Mensural Discrimination of Peromyscus 113 



RESULTS 



In univariate tests, we found significant differences between the 

 sexes in P. gossypinus for body length (P<0.02), RL (P<0.02), and 

 NL (P<0.05); in P. leucopus for foot (P<0.()3) and PFL (P<0.04); in 

 P. maniciilatus for SL (P<0.05), PB (P<0.04), and PFL (P<0.01); 

 and in P. polionotus for body length (P<0.01). Although these differences 

 were individually significant, there was considerable overlap in character 

 ranges, and none was significant when we applied the Bonferroni correction 

 (table-wide significance began at P< 0.003). The differences between 

 the sexes of P. maniciilatus approached significance (P<0.07), but 

 none was significantly different (P<0.05) when subjected to two group 

 (i.e., male vs. female) discriminant analysis. We included gender in 

 the discriminant analysis of all characters, but its effect was not significant, 

 and it did not enter the final stepwise model. Table 1 contains means, 

 ranges, and standard errors for all characters. 



Univariate analyses were marginally successful in identifying the 

 four species, but no single measurement unambiguously separated them. 

 Most characters separated the large P. gossypinus from the small P. 

 polionotus, but six of 17 characters showed overlapping distributions. 

 Tail length greater or less than 55 mm is the simplest method to separate 

 these two species. No single character could separate P. gossypinus 

 from P. leucopus or P. maniculatus, but anterior palatal foramen length 

 5.4 mm identified most (67%) P. gossypinus. Tail length 83 mm separated 

 81% of P. maniciilatus from the other three species, but four P. gossypinus 

 had tails longer than 83 mm. There was no overlap in the tail lengths 

 of P. polionotus and P. maniciilatus. No single character separated P. 

 leucopus from P. maniculatus. 



Multivariate analyses using external and skull measurements were 

 successful in identifying the four species. Stepwise discriminant analysis 

 correctly classified all specimens using measurements of 13 characters 

 (in order of inclusion into model: Tail, SL, MTL, Foot, RL, OC, PFL, 

 Body, TTL, PB, BNL, PL, PW). The three axes accounted for 55.51, 

 37.16, and 7.34% of the variance (Fig. 3a). After a varimax rotation, 

 the variables most highly correlated with the first discriminant function 

 were TTL (0.87), SL (0.85), BNL (0.74), PL (0.69), RL (0.58), PFL 

 (0.58), ZB (0.57), MTL (0.53), and NL (0.52); those highly correlated 

 with the second function were BD (0.82), PB (0.40), and OC (0.37); 

 and those highly correlated with the third function were PFL (0.49), 

 RL (0.45), PB (0.23), and OC (-0.21). 



Discriminant analysis using only skull measurements correctly 

 classified at most 90% of the specimens with 10 characters (in order 

 of inclusion into model: SL, BD, MTL, RL, PFL, OC, TTL, BNL, PB, 



