142 
T. C. Kane, T. C. Barr, Jr., and G. E. Stratton 
Genotype frequencies at variable loci in each population were 
tested statistically for fit to expectations under Hardy-Weinberg Equili- 
brium (HWE). Three statistical tests were used to test correspondence of 
the data to HWE: Chi-square goodness of fit with Levene’s (1949) 
correction for small sample size, an exact probability test analogous to 
Fisher’s exact test, and — in cases where three or more electromorphs 
occurred in a population — a Chi-square test with pooling. Because 
sample sizes were relatively small (i.e. almost always < 50 per population), 
and the Chi-square test is less reliable when expected values for some 
classes are small (Sokal and Rohlf 1981), we chose to reject the null 
hypothesis of HEW only when all tests applied indicated statistically 
significant ( P < 0.05) deviation from HWE. Sixteen populations were 
variable at the PGI locus, and in all cases at least one test was not 
significant ( P > 0.05), suggesting that none deviates from HWE. 
Two populations of the bowlingi subgroup (CG and WP; Fig. 1) 
were found to deviate significantly from HWE expectations at the CAH 
locus by all three tests. For the remaining 1 1 variable populations, 
however, HWE at the CAH locus could not be rejected by one or more 
statistical tests. Two separate collections, one in 1983 and the other in 
1984, were made at both the CG and WP sites. Further, the two 
collections were made at slightly different, though proximate, microhabitats 
within each site. In contrast, the HE and RC bowlingi locations were 
sampled on a single date and at single sites in each location. Thus 
deviation from HWE at the CAH locus for the CG and WP locations 
may reflect temporal and / or microspatial heterogeneity in gene frequency 
in these populations. However, the fact that no such genetic heterogeneity 
is evident at the PGI locus for these two locations argues against this 
contention. Further, because 27 of the 29 cases of variable loci in 
populations meet our conservative requirements for fit to HWE, it is 
difficult to ascribe much significance to these two exceptions from 
available data. 
Variation among populations within subgroups was examined using 
F-statistics and a Chi-square contingency analysis (Workman and 
Niswander 1970) (Table 4). Significant heterogeneity in gene frequency 
( P < 0.05) occurs among populations in the bowlingi, tusquitee, and 
vandykei subgroups at both the CAH and PGI loci. The six pisgahensis- 
subgroup populations, which are monomorphic for the same allele at 
the CAH locus, show significant heterogeneity in allele frequency at the 
PGI locus. For those subgroups in which significant heterogeneity in 
allele frequencies was observed, genetic differentiation can be described 
as slight (F sx < 0.05) to moderate (0.05 < F sx <0.15) (Table 4). Rogers’ 
Genetic Similarity (S) calculated over all loci produces values of 
