68 Christopher W. Brown and Carl H. Ernst 



occur in quadrants one and four, and C. h. horridus in quadrants two 

 and three (Fig. 5A). Factor 1 (adult size and postocular stripe), then, 

 appears to differentiate C. horridus into two forms. Of those specimens 

 of C. h. atricaudatus occurring in the second and third quadrants, only 

 one lacks a postocular stripe, and all are under 1000 mm snout-vent 

 length, the smallest individuals of their group. For example, the speci- 

 men of C. h. atricaudatus having the largest negative factor 1 score is 

 only 765 mm snout-vent length (probably a subadult). Similarly, those 

 specimens of C. h. horridus lying in the first and fourth quadrants are 

 the physically largest individuals of their group. Such large or small 

 individuals, though not typical of their group, can be expected. Factor 

 2, which includes character R, demonstrates that the ratio of tail length 

 to snout-vent length as an indicator of size is not as reliable as the 

 lengths themselves (factor 1) in distinguishing the two groups. For 

 instance, a large snake having a correspondingly large tail could have 

 the same ratio as a smaller snake, or even one of a different species. 



Factor 1 (abscissa) versus factor 3 (ordinate, Fig. 5B) produces a 

 scatterplot much like plot 5A, indicating that factor 3 (dorsal scale 

 rows), like factor 2 (adult tail measurements), is relatively unimportant 

 in differentiating the two subspecies. Factor 1 (adult size) again pro- 

 duces good separation of the two groups in plot 5B with the same individ- 

 uals lying far to the left or right of their respective groups as seen in plot 

 5A. 



Since factor 2 (adult tail measurements) and factor 3 (dorsal scale 

 rows) have been shown to be unimportant in distinguishing the two 

 groups, the plot of factor 2 versus factor 3 would be expected to demon- 

 strate no clustering, and this is observed in Figure 5C. 



In the plot of factor 2 (abscissa) versus factor 4 (ordinate), cluster- 

 ing occurs along the vertical axis, with most specimens of C. h. atricau- 

 datus in the first two quadrants and those of C. h. horridus in the last 

 two (Fig. 5D). A similar plot occurs for factor 3 versus factor 4, as 

 would be expected (Fig. 5E). Factor 4 (middorsal stripe) therefore 

 appears to differentiate the specimens into two groups: those possessing 

 a distinct middorsal stripe (factor 4 greater than zero, which corre- 

 sponds to C. h. atricaudatus) and those possessing an indistinct middor- 

 sal stripe or none at all (factor 4 less than zero, which corresponds to C. 

 h. horridus). 



Specimens with factor scores outside the normal range of variation 

 for their group were examined more closely in plots 5D and 5E to 

 determine why they clustered with the "wrong" group. Those few C. h. 

 horridus that possessed a distinct middorsal stripe (factor 4 greater than 

 zero) were all from localities in North Carolina and Georgia where 

 intergradation might occur. None, in other words, came from localities 

 well to the north of the C. h. atricaudatus range. Those few specimens 

 of C. h. atricaudatus having a large negative factor 4 score all possessed 

 a faint middorsal stripe, rather than lacked one entirely, which classified 



