774 
Fishery Bulletin 96(4), 1 998 
long terminal branches to individual samples, but short 
intemodal branches linking samples. Samples did not 
necessarily cluster geographically, and in no case did 
geographically proximate samples form a cluster. 
Spatial autocorrelation analyses generated 40 
Moran’s I values in each run (10 haplotypes x four 
distance classes). Four significant (P<0.05) values 
were obtained when equal numbers of pairwise com- 
parisons were used: two were positive and occurred 
in the first and third distance classes, and two were 
negative and occurred in the third and fourth dis- 
tance classes. Two significant values (P<0.05) were 
obtained when equal geographic distances 
were used: one in the first distance class 
was positive, and one in the third distance 
class was negative. Mean I values were 
negative in all four distance classes in both 
runs and did not differ significantly from 
expected values of I in the absence of 
autocorrelation (Fig. 4). 
Discussion 
Variation of mtDNA in greater amberjack 
is consistent with the hypothesis that one 
subpopulation (stock) exists in the north- 
ern Gulf of Mexico (Gulf), whereas a sec- 
ond subpopulation (stock) exists along the 
U.S. southeast Atlantic coast (Atlantic). 
The latter (Atlantic) subpopulation in- 
cludes the sample from the Florida Keys. 
Evidence supporting this hypothesis in- 
cludes 1) significant heterogeneity in the 
distribution of pooled haplotypes (and of 
haplotype 6) from the Gulf versus those 
from the Atlantic (and the Florida Keys); 
2) a significant (nonzero) ^ST value in the 
same comparison; 3) a highly significant 
(P<0.001) between-region component (d> CT 
value) in AMOVA, when regions were 
defined as the Gulf versus the Atlan- 
tic (and the Florida Keys); and 4) com- 
parison of pairwise <2> ST “distances” 
where samples from the Gulf or from 
the Atlantic were at least three times 
more similar to one another than were 
samples from the Gulf versus those from 
the Atlantic (and the Florida Keys). 
Neither maximum-parsimony (MP) 
analysis of individual mtDNA haplo- 
types nor neighbor-joining analysis of 
a matrix of interpopulational (inter- 
sample) nucleotide-sequence distances 
revealed phylogeographic patterns in- 
dicative of population structuring. A 
few, well-supported clades of indi- 
vidual haplotypes were detected in MP 
analysis, but in no case were haplo- 
types within individual clades from the 
same or geographically proximate 
a 
3 
c r 
o> 
Haplotype 6 
Haplotype 13 
Figure 2 
Frequency of mtDNA haplotypes 6 and 13 among greater amberjack 
(Seriola dumerili) sampled from 11 localities in the northern Gulf of Mexico 
and southeastern (U.S.) Atlantic. Acronyms are defined in Table 1. 
Pensacola, FL 
Sarasota, FL 
— New Smyrna, FL 
Panama City, FL 
Port Fourchon, LA 
— Florida Keys 
Freeport, TX 
St. Petersburg, FL 
■ Port Aransas, TX 
South Carolina 
Gulfport, MS 
Figure 3 
Neighbor-joining topology generated from matrix of pairwise, interpop- 
ulational ( mtDNA) nucleotide sequence divergence among samples of greater 
amberjack, Seriola dumerili. 
