Mensural Discrimination of Peromyscus 109 



to the problem has been necessary because several of the species, 

 particularly P. leucopus and P. maniculatus, have a high degree of 

 intraspecific variation in morphology. 



In the southeastern United States the ranges of four species overlap 

 (Fig. 1). It is difficult to correctly identify these species using available 

 taxonomic keys (e.g., Golley 1962, 1966; Blair et al. 1968; Hall 1981) 

 based only on pelage features and/or cranial measurements. The four 

 species usually can be distinguished based on collection location, 

 habitat, and morphological data. Populations of Peromyscus maniculatus 

 in this region are referred to as P. m. nubiterrae and are typically 

 found in mesic forests at elevations higher than 900 m, and P. maniculatus 

 usually has a sharply bicolored tail that is longer than the head and 

 body. Peromyscus gossypinus is generally found in hardwood river 

 bottoms and coastal oak-palmetto (Quercus sp. and Serenoa repens) 

 forests and is the largest and heaviest of the four species. Peromyscus 

 polionotus is generally found in areas of sandy soil and has a very 

 short, distinctly bicolored tail. Peromyscus leucopus leucopus is generally 

 found at elevations below 900 m in relatively xeric woodlands. Its 

 tail is shorter than the head and body, and it is smaller and lighter in 

 mass than P. gossypinus. A plot of principal component scores generated 

 from the correlation structure of three standard external measurements 

 (body, tail, and hind foot lengths) illustrates the overlap in measurements 

 from specimens collected in the Southeast and graphically illustrates 

 the difficulty in separating these four species based on these features 

 (Fig. 2). 



For museum personnel that acquire poorly curated public or private 

 collections, or who desire to reexamine their holdings, identification 

 of specimens from regions where ranges overlap may be difficult. 

 The objective of this study is to examine the effectiveness of statistical 

 procedures to distinguish these species in the southeastern United 

 States without the use of collection-location information and without 

 using statistically unsound ratios (Humphries et al. 1981). To do this, 

 we generate discriminant functions from both external and skull measurements 

 and from skull measurements alone. 



METHODS 



We used univariate and multivariate statistics to examine 460 

 Peromyscus museum specimens collected in the southeastern United 

 States for variation in 17 morphometric characters. We selected sample 

 sites based on the availability of large numbers of adult specimens 

 from throughout the region. Sample sites were selected to reduce 

 potential for incorrect a priori species identification by eliminating, 

 to some degree, consideration of localities where ranges overlap. These 

 criteria resulted in the distribution of sample sites in Figure 1. 



