770 
Fishery Bulletin 96(4), 1998 
ity. This value represents the probability that any 
two individuals drawn at random will differ in 
mtDNA haplotype. Intrapopulational nucleotide se- 
quence diversity within sample localities also was 
estimated after Nei and Tajima (1981). This value 
represents the average nucleotide sequence differ- 
ence between any two individuals drawn at random 
from a given sample or locality. 
Homogeneity testing of mtDNA haplotype frequen- 
cies among sample localities was carried out by us- 
ing 1) a randomization (bootstrap) procedure devel- 
oped by Roff and Bentzen (1989), 2) a log-likelihood 
( G ) test (Sokal and Rohlf, 1969), and 3) V tests that 
employed arcsine, square-root-transformed haplotype 
frequencies (DeSalle et al., 1987). Significance lev- 
els for multiple tests were adjusted after Rice (1989). 
F st values, a measure of the variance in mtDNA 
haplotype frequencies, were calculated after Weir and 
Cockerham (1984) by using algorithms described in 
Weir ( 1990). Significance of E gT values was tested by 
the randomization procedure in version 1.4 of the 
analysis of molecular variance ( AMOVA) program of 
Excoffier et al. ( 1992). The latter program (AMOVA) 
was used primarily to examine the distribution of 
variance in mtDNA haplotypes. AMOVA analysis 
generates estimates of (genetic) variance components 
and a set of hierarchical E-statistic analogs (<7> sta- 
tistics) that are tested for significance through ran- 
dom permutation methods. The permutation ap- 
proach avoids the parametric assumptions of normal- 
ity and independence normally not met by molecu- 
lar distance measures (Excoffier et al., 1992). Sample 
localities were nested into regional groupings, i.e. 
Gulf and Atlantic, that were input into AMOVA. In 
one AMOVA run, the sample from the Florida Keys 
was included with the Gulf group, whereas in a sec- 
ond run it was included in the Atlantic group. 
Restriction-site presence and absence matrices for 
individual mtDNA haplotypes were constructed with 
the GENERATE program in REAP by inferring restric- 
tion site gains and losses for each enzyme. Maximum- 
parsimony analysis of restriction-site presence and 
absence matrices representing all haplotypes employed 
MULPARS and CONTREE options in version 3.1 of 
the Phylogenetic Analysis Using Parsimony (PAUP) 
program of Swofford (1991). Autapomorphic and 
symplesiomorphic characters were removed prior to 
PAUP runs. Nucleotide-sequence divergence values 
among sample localities (interpopulational divergence) 
were generated following Nei and Tajima (1981) and 
Nei and Miller ( 1990). MtDNA-based similarity among 
sample localities was assessed with the neighbor join- 
ing method (Saitou and Nei, 1987) and employed the 
NEIGHBOR program in Phylogenetic Inference Pack- 
age (PHYLIP), version 3.4 of Felsenstein (1989). 
The spatial distribution of mtDNA haplotypes was 
investigated by means of spatial autocorrelation 
analysis (SAAP; Wartenberg, 1989). This analysis 
determines whether haplotype frequencies at any 
sample locality are independent of haplotype frequen- 
cies at neighboring sample localities. Correlograms 
that plot autocorrelation coefficients (Moran’s I val- 
ues) as a function of geographic distance between 
pairs of localities were used to summarize patterns 
of geographic variation of haplotype frequencies. To 
minimize “noise” generated by low-frequency 
haplotypes, Moran’s 7 values were calculated only for 
haplotypes occurring in eight or more individuals ( 10 
haplotypes total). These included haplotypes 1-4, 6, 
11, 13, 23, 37, and 48. 
Results 
Single digestions of mtDNA molecules from the 444 
individuals surveyed produced variable fragment 
patterns for the 16 restriction enzymes. The major- 
ity of fragment patterns observed are given in Ap- 
pendix Table A2 of Richardson and Gold (1993). New 
fragment patterns revealed in our study are as fol- 
lows (fragment sizes in base pairs; asterisks repre- 
sent fragments assumed to exist but not covered by 
the mtDNA probe): EcoRI, pattern C (8700, 5025, 
3175*); Hpal, pattern C (8800, 5550, 2550); Ncol, 
pattern C (11050, 5850); Sstl, pattern C (11700, 3450, 
1750); Neal, pattern C (7500, 3600, 3100, 2700) and 
pattern D (10000, 3600, 3300); Smal, pattern D 
(10700, 4250, 1500, 450); Spe I pattern D (7900, 5800, 
1200, 1200*, 800); Sspl, pattern C (6600, 6200, 4100); 
and Xbal, pattern D (7300, 5000, 4600) and pattern 
E (7300, 4600, 2600, 2275, 125*). The mean genome 
size of all apparently complete digestion patterns was 
16.9 ±0.2 kilobase pairs. No evidence of mtDNA size 
variation or heteroplasmy was observed. All frag- 
ment patterns for each restriction enzyme were con- 
sistent with the hypothesis of single nucleotide sub- 
stitutions. A total of 72 unique restriction sites was 
inferred from the digestion patterns. 
A total of 49 mtDNA haplotypes was identified 
among the 444 individuals surveyed (Table 2). Four 
haplotypes (8, 14, 16, and 18) are not listed in Table 
2; these were listed in Richardson and Gold (1993) 
and were identified by three restriction enzymes 
(Clal,Nsil, and Puul) not employed in our study. Four 
haplotypes (1, 4, 6, and 13) were abundant, occur- 
ring in 77, 91, 54, and 70 individuals, respectively. 
Two haplotypes (2 and 3) occurred in 20 and 22 indi- 
viduals, respectively, whereas the remainder oc- 
curred in 10 or fewer individuals. Twenty haplotypes 
were found in only one individual each. Estimates of 
