134 
Fishery Bulletin 99(1 ) 
western Kamchatka and eastern Sakhalin Island face 
each other across the Sea of Okhotsk but have different 
allele frequencies at PEPD-2*100 [0.71 ±0.02 (SE) for 
western Kamchatka versus 0.61 ±0.01 for eastern Sakha- 
lin), GR*100 (0.90 ±0.01 versus 0.97 ±0.005), PEPB*100 
(0.91 ±0.01 versus 0.79 ±0.01), and PGDH*100 (0.76 
0.02 versus 0.82 ±0.01). Therefore, it is unlikely that 
there is substantial gene flow between these regions. If 
large movements of spawning pink salmon occur (fluctu- 
ating stock hypothesis; e.g. Zhivotovsky and Glubokovsky, 
1989; Shuntov et al. 1994), our results suggest that they 
are restricted to within-region movements for even-year 
pink salmon. Moreover, the relative homogeneity observed 
within regions does not necessarily indicate large numbers 
of strays. Relatively small exchanges between drainages 
(say 10 fish or so per generation) can arrest genetic diver- 
gence (e.g. Gharrett, 1994). This level of straying does not, 
however, provide demographic insurance against overfish- 
ing or environmental catastrophes in the short run (Hop 
and Gharrett, 1989). 
The contemporary genetic structure of even-year pink 
salmon has practical implications for fishery scientists. 
Genetic differences among populations can be used as 
markers for stock identification. It is often necessary to 
manage a species that has as many small populations as 
pink salmon as a regional assemblage. The regional basis 
for genetic structure observed in Asian pink salmon lends 
itself to stock separation analyses (Hawkins et al., 1998). 
Most of the genetic variability observed in even-year 
pink salmon is attributable to within-population variabil- 
ity (e.g. Gharrett et al., 1988; Beacham et al., 1988). How- 
ever, low levels of divergence among populations do not 
preclude hierarchical population structure. In our study 
we saw little or no temporal structure among Sakhalin 
Island populations. This finding does not necessarily con- 
flict with observations of temporal structure observed by 
Altukhov et al. (1983) who used much larger numbers of 
fish (but fewer alleles), by McGregor et al. (1998) who used 
multiple years of data and numerous loci, and by Brykov 
et al. ( 1999) who used highly variable mitochondrial DNA 
haplotypes, because their tests had much greater statisti- 
cal power. In addition, allozymes may not be appropriate 
for detecting some kinds of population structure because 
a very low level of gene flow can “homogenize” frequen- 
cies of neutral or nearly neutral loci. Note, for example, 
the genetic component observed for time of return within 
a spawning season (Smoker et al., 1998) and the persis- 
tence of a genetic marker for time of spawning (Lane et al., 
1990). 
The strong regional structure we observed in Asian 
even-year pink salmon populations is a stark contrast to 
the nearly complete absence of structure reported for odd- 
year Kamchatka pink salmon (Varnavskaya and Beacham, 
1992; Shaklee and Varnavskaya, 1994). However, the lat- 
ter surveys covered smaller geographic ranges and in- 
cluded neither Sakhalin Island nor Japanese pink salmon 
populations. These combined studies involving numerous 
allozyme loci confirm the work of Zhivotovsky et al. (1989) 
who, using four allozyme loci, also recognized the brood- 
year differences; and who estimated that 1.6% (we esti- 
mated 1.75%) and 0.6% of the total genetic variability 
were attributable to interregional divergence for even- 
and odd-year broods, respectively. 
The differences observed between Asian and North 
American even-year pink salmon are not surprising; they 
have been reported previously for both even-year (Zhivo- 
tovsky et al., 1989) and odd-year broodlines (Varnavska- 
ya and Beacham, 1992; Shaklee and Varnavskaya, 1994). 
However, the strong coherence of populations within each 
of the major North Pacific basins (Sea of Okhotsk, Bering 
Sea, and Gulf of Alaska) is striking. Each basin includes 
a range of habitats and climates that suggests that the 
genetic similarity among populations within a geographic 
region does not result from convergent or homogeneous 
selection. The differences among populations of different 
basins are also reflected by the different average hetero- 
zygosities. The apparent directional change in heterozy- 
gosities could be interpreted in terms of differences in 
effective population size, age of the populations, or the ex- 
tent of environmental variation. However, speculation is 
probably not warranted because three different variables 
can be arranged in a monotonic pattern in two of six dif- 
ferent possible orders. 
A more evocative explanation of the genetic structure 
combines geographic and oceanographic influences, as well 
as recent geologic history. Populations in the contemporary 
Sea of Okhotsk, Bering Sea, and Gulf of Alaska are sep- 
arated geographically by land masses and oceanographi- 
cally by the different currents that flow into or between 
the oceanic basins and that influence migration routes of 
the fish (Royce et al., 1968). The geographic separation was 
greatly exacerbated by the limits and effects of late Pleisto- 
cene glaciation. In their northern range, pink salmon pop- 
ulations experienced increased isolation between the ma- 
rine basins as a result of lower sea level, loss of freshwater 
habitat to increased ice cover, and more extensive sea ice. 
Just as recent fluctuations in salmon productivity have 
resulted from relatively minor climate changes (Mantua 
et al., 1997), less favorable freshwater and marine envi- 
ronmental conditions undoubtedly decreased the sizes and 
numbers of populations dramatically, further isolating the 
remaining populations in this region. 
During the past several 100,000 years, there have been 
periodic changes in climate, sea level, glacial extent, and 
oceanographic conditions. It is important to realize that 
our modern, interglacial conditions are an extreme (San- 
cetta and Silvestri, 1986) and that the oceanic record of 
global ice volumes (from the d 18 0 record in marine sedi- 
ments) and geologic evidence (Mann and Hamilton, 1995) 
from the north Pacific realm during that period indicate 
that lower sea levels, more extensive glaciation, colder sea 
surface temperatures, and more extensive and southerly 
sea ice were typical (Bartlein et al., 1991; Rohling et ah, 
1998). The relative proportion of the 18 0 isotope (<5 18 0) in a 
stratum of a core is related to the portion of the earth’s wa- 
ter tied up in ice at the time corresponding to the stratum, 
and consequently the sea level in relation to the modern 
sea level. 
At the last glacial maximum (LGM: ca. 14,000-20,000 
years before the present [BP] ), the paleogeography of the 
