Ward et al.: Population structure of Thunnus albacares 
573 
from the Bismarck Sea in the western Pacific were 
very similar to our western Pacific Ocean frequen- 
cies (Ward et al., 1994), supporting the separation of 
western and eastern Pacific stocks. Another allozyme 
study (Fujino, 1970) failed to find differences between 
Hawaiian and eastern Pacific fish for an esterase and 
for transferrin, although the esterase was nearly 
monomorphic. 
We interpret the GPI-A* differentiation as being 
indicative of stock differences, resulting from re- 
stricted gene exchange between the four identified 
regions. However, the alternative explanation, that 
of differential selection in the presence of gene flow, 
cannot be ruled out. Indeed, the very limited mtDNA 
differentiation observed could be held to support this 
interpretation. Microsatellite analysis, currently 
underway, may help to resolve this question. Selec- 
tion acting on these noncoding genetic markers is 
presumed to be minimal or nonexistent; therefore 
microsatellite differentiation paralleling the GPI-A* 
differentiation would suggest drift of neutral GPI- 
A* alleles, whereas lack of microsatellite differen- 
tiation would indicate significant gene flow and 
thereby implicate selection as the cause of the GPI- 
A* differentiation. Pogson et al. ( 1995) have recently 
suggested that the highly heterogeneous distribution 
of anonymous nuclear RFLP markers among popu- 
lations of cod, Gadus morhua, reflects limited gene 
flow and that the much more homogeneous distribu- 
tion of allozyme alleles reflects stabilizing selection 
rather than extensive gene flow. Such an argument 
applied to yellowfin tuna data would interpret the 
GPI-A* heterogeneity as indicative of limited gene flow, 
and the ADA*, FH*, and GPI-B* homogeneity as in- 
dicative of stabilizing selection at these three loci. 
Differences between collections in mtDNA was only 
just significant (P=0.048), with a “true” G ST value 
across all nine collections of around 1%. When col- 
lections were pooled within oceans, i.e. the three 
groups (Atlantic, Indian, and Pacific), significant dif- 
ferentiation was detected (P=0.009), although the 
“true” G sr was only of the order of 0.5%. All three 
pairwise ocean comparisons were statistically signifi- 
cant. However, because collections within oceans did 
not always pool together in the distance dendrograms 
(Fig. 3), possibly because of limited sample sizes, it 
would clearly be useful to have more data to confirm 
(or refute) this evidence of interoceanic differentia- 
tion. When collections were pooled into the four pu- 
tative stocks indicated by the GPI-A* data, limited 
but significant heterogeneity in mtDNA haplotype 
frequencies was apparent (P=0.024), but there were 
no significant pairwise comparisons. 
Scoles and Graves (1993) were unable to detect 
significant mtDNA differentiation between Pacific 
and Atlantic yellowfin tuna, whereas the probability 
of homogeneity in our tests of these two oceans was 
only 0.017. However, they adopted a different test 
strategy. Instead of examining relatively large num- 
bers of fish (our study: Pacific fish, n=561; Atlantic 
fish, n- 94) with relatively few restriction enzymes 
(n= 2, but known to detect polymorphic sites), they 
chose to examine relatively few fish (Pacific fish, 
re=100; Atlantic fish, t?= 20) with a relatively large 
number of restriction enzymes (n- 12, which included 
the two enzymes we used). Given that the common 
12-enzyme haplotype in Scoles and Graves’ study 
comprised 52 fragments or 304 bp and that the com- 
mon 2-enzyme haplotype in our study comprised 7 
fragments or 42 bp (see Ward et al., 1994) and that 
the mean size of the yellowfin tuna mtDNA genome 
is about 16,702 bp (Scoles and Graves (1993] esti- 
mate= 16,549; Ward et al. [ 1994]= 16,856), Scoles and 
Graves surveyed about 1.8% of the mtDNA genome, 
whereas we surveyed only about 0.3%. However, al- 
though it is of course true that had we surveyed more 
restriction enzymes, we would have uncovered many 
additional haplotypes, the two enzymes that we did 
select revealed most of the mtDNA diversity shown 
by Scoles and Graves (1993). For example, the 
(pooled) 12-enzyme haplotype diversity of 0.840 of 
Scoles and Graves was not much larger than our 
(pooled) 2-enzyme diversity of 0.677. Four of the en- 
zymes used by Scoles and Graves showed no varia- 
tion at all in the 120 fish and therefore were of no 
use for population discrimination. Twenty of the 34 
12-enzyme haplotypes detected by Scoles and Graves 
(1993) among their 120 fish were seen only once, 
whereas only four of the 22 2-enzyme haplotypes in 
our 655 Atlantic and Pacific fish were seen only once: 
such rare haplotypes are of extremely limited use in 
population studies. Given that mtDNA heterogene- 
ity among regions is very limited, it is not surprising 
that the approach of screening large numbers of fish 
for a small number of sequences known to be vari- 
able should be more powerful than screening small 
numbers of fish for a larger number of sequences, 
many of which are relatively invariant. 
MtDNA data from another tuna, the albacore, 
Thunnus alalunga, showed a somewhat more pro- 
nounced separation of Atlantic Ocean and Pacific 
Ocean collections than did data for yellowfin tuna, 
but again no intraoceanic heterogeneity was detected 
(Chow and Ushiama, 1995). 
The limited mtDNA differentiation among yellow- 
fin tuna sampled throughout their range contrasts 
with the marked population subdivision revealed by 
the GPI-A* locus. Mitochondrial DNA has an effec- 
tive population size only one quarter that of nuclear 
DNA (Birky et al., 1989) and evolves more rapidly 
