Ward et a I.: Population structure of Thunnus albacares 
571 
Pairwise comparisons of GPI-A 
Table 4 
* allele frequencies (P above, chi square below) 
. GOM = Gulf of Mexico. 
GOM 
Sri Lanka 
Philippines 
Coral Sea 
Kiribati 
Hawaii 
California 
Mexico 
GOM 

<0.001 
0.156 
0.468 
0.129 
0.769 
<0.001 
<0.001 
Sri Lanka 
21.700 
— 
0.001 
<0.001 
<0.001 
0.001 
0.031 
0.214 
Philippines 
5.938 
24.115 
— 
0.264 
0.965 
0.291 
<0.001 
<0.001 
Coral Sea 
1.586 
28.644 
4.870 
— 
0.270 
0.237 
<0.001 
<0.001 
Kiribati 
3.905 
22.586 
1.196 
2.630 
— 
0.203 
<0.001 
<0.001 
Hawaii 
0.493 
19.130 
4.635 
2.824 
3.280 
— 
<0.001 
<0.001 
California 
24.850 
6.047 
46.871 
35.021 
37.324 
25.374 
— 
0.289 
Mexico 
34.934 
3.579 
51.193 
46.401 
44.495 
33.366 
2.526 
— 
Table 5 
Pairwise comparisons of mtDNA haplotype frequencies ( P above, chi square below). GOM = Gulf of Mexico. 
GOM 
Seychelles 
Sri Lanka 
Philippines 
Coral Sea 
Kiribati 
Hawaii 
California 
Mexico 
GOM 

0.277 
0.009 
0.089 
0.025 
0.009 
0.384 
0.285 
0.076 
Seychelles 
14.079 
— 
0.549 
0.062 
0.188 
0.212 
0.200 
0.478 
0.707 
Sri Lanka 
23.110 
9.697 
— 
0.269 
0.369 
0.171 
0.111 
0.025 
0.612 
Philippines 
19.823 
22.529 
14.013 
— 
0.223 
0.025 
0.428 
0.207 
0.597 
Coral Sea 
18.085 
16.020 
7.785 
15.432 
— 
0.087 
0.053 
0.092 
0.683 
Kiribati 
21.523 
17.480 
11.862 
22.368 
15.117 
— 
0.086 
0.093 
0.253 
Hawaii 
13.801 
18.574 
18.760 
15.533 
19.300 
18.754 
— 
0.667 
0.378 
California 
11.888 
11.682 
13.905 
17.165 
14.055 
16.316 
8.888 
— 
0.383 
Mexico 
15.622 
9.718 
5.017 
10.188 
7.203 
12.214 
13.080 
8.473 
— 
A chi-square test (Table 3) showed 
that the mtDNA haplotype variation 
across all nine regions was just signifi- 
cant (a=0.05, P=0.048, with the stan- 
dard 2,000 replicates, and P=0.045, 
with 10,000 replicates). Genetic diver- 
sity analysis (Table 3) gave a result bor- 
dering on significance (P=0.071, with a 
“true” G st of about 1%). All collections 
were compared pairwise with chi- 
square tests (Table 5) to determine 
which collections contributed most to 
the marginal chi-square differentiation. 
Although some pairs appeared signifi- 
cantly different (e.g. Gulf of Mexico ver- 
sus Sri Lanka, P=0.009; Gulf of Mexico 
versus Kiribati, P=0.009), none was sig- 
nificant after Bonferroni adjustments 
for table-wide comparisons. 
Two UPGMA dendrograms were es- 
timated. One, based on mtDNA haplo- 
type frequencies alone (Fig. 3A), showed 
the maximal genetic-distance estimates among col- based on percent sequence divergence (Fig. 3B), con- 
lections to be about 0.05 — much less than the ma- firmed the high degree of similarity among the col- 
jor GPI-A* split of nearly 0.30 (Fig. 2). The second, lections. After correcting for within-collection nucle- 
0.30 0.20 0.10 0.00 
Genetic distance 
Figure 2 
UPGMA phenogram for the GPI-A* locus constructed from Nei’s (1978) 
unbiased genetic distance. 
