Variation in Qrotalus horridus 71 



above). For example, in their study mean body bands ranged from 

 23.14 to 25.05 among all groups. In our study, body crossbands aver- 

 aged from only 9.14 to 12.54 (Table 4). The differences in mean tail 

 bands is similar. In our first analysis these characteristics discriminated 

 between the two subspecies, whereas they did not in Pisani et al. (1973); 

 however, their method may have measured an entirely different charac- 

 ter variable than the one they intended. 



In the discriminant analysis of all 19 characters, 2 of the 6 added 

 characters discriminated. These were HL and MS. Because head length 

 is probably indicative of the total length of the adult snake (Klauber 

 1938, 1972), it appears that adult size and pattern are important in dis- 

 criminating between the two subspecies. Dorsal scale rows and number 

 of ventrals also discriminated here, as in Pisani et al. (1973), except that 

 the discrimination was by ADS instead of DSM. Characters BCB and 

 TCB did not discriminate in the second analysis. 



In comparing the two analyses, we find that all 19 characters 

 allowed better discrimination. This is evident in the larger eigenvalue 

 (2.989 vs. 1.088), the larger canonical correlation (0.866 vs. 0.722), the 

 smaller Wilks' lambda (0.251 vs. 0.479), and better classification of indi- 

 viduals into the two groups (7% incorrectly classified vs. 16%). In the 

 canonical variable histograms (Fig. 6), separation of the two groups is 

 much better in the analysis of all 19 characters, again showing the 

 importance of size and pattern. 



Comparison of the discriminating characters ADS, VS, and HL 

 between nonintergrades and intergrades (Tables 4 and 5) shows that, as 

 expected, the mean character values of the intergrade specimens lie 

 between the mean character values of C. h. horridus and C. h. atricau- 

 datus, regardless of sex. Since the remaining characters other than MS 

 did not discriminate, their mean values for the intergrade specimens are 

 not expected to be intermediate or even different from the mean values 

 of either horridus or atricaudatus. Interestingly, all intergrades pos- 

 sessed a distinct middorsal stripe, much like intergrades reported from 

 western localities (Gloyd 1940, Smith 1961, Webb 1970). 



Of the three new pattern characters tested in this study (MS, GC, 

 POS; Table 1), POS and, especially, MS were important. None, how- 

 ever, was completely free from subjectivity in measurement. In some 

 cases, the distinction between light- and dark-colored or indistinct and 

 clearly visible was a fine line. Use of old museum specimens, many 

 faded by preservatives, may have induced too much subjectivity, how- 

 ever unintentional. Some dark-colored specimens had a middorsal stripe 

 that had apparently faded to an almost white color, making the stripe 

 unusually conspicuous. Had the specimens been living, the stripe may 

 have been inconspicuous. In other specimens the ground color was 

 faded and difficult to determine. Another problem is the inadequacy of 

 the coding scheme for GC. Gloyd (1940) and others (Wright and Wright 

 1957, Conant 1975) mentioned that C. h. horridus has two color 



