538 
Fishery Bulletin 99(4) 
that at some time the haplotypes were disseminated either 
through gene flow or colonization. Our data, are generally 
consistent with the known distributions and abundances of 
North Pacific coho salmon, that is, coho salmon is a species 
composed of many small spawning populations that have 
low levels of gene flow. The low geneflow rates estimated 
within several geographic ranges indicate that a finer scale 
study of coho salmon population structure is warranted. 
Superimposed on this “snapshot” of population struc- 
ture is the mtDNA “gene tree,” which carries information 
about the demographic history of coho salmon. Nested 
clade analysis of the geographic distribution of haplotypes 
and clades of related haplotypes reveal recent isolation by 
distance, particularly for the geographical distribution of 
A-cluster haplotypes. There is also evidence of past popu- 
lation fragmentation in the distribution of E-cluster hap- 
lotypes, and the interior of the tree has evidence of range 
expansion. The topology of the tree also indicates recent 
demographic expansion preceded by stationary, or more 
likely, declining populations. Radiation of haplotypes to 
form a star-like pattern occurs when populations expand 
(Slatkin and Hudson, 1991), such as the haplotypes that 
surround haplotypes A and E. Whereas, in stationary or 
declining populations, the haplotype composition of a popu- 
lation eventually tends toward a single, evolutionarily re- 
lated (monophyletic) set as a result of stochastic processes 
(Neigel and Avise, 1986), such as the A- or E-cluster. 
It is likely that the demographic history of coho salmon 
is closely tied to Pleistocene events. During the Pleistocene 
Epoch, glaciers periodically advanced and retreated in re- 
gions bordering the northeastern Pacific Ocean. These ad- 
vances were accompanied by a drop in sea level (exceed- 
ing 100 m), decreased sea surface temperatures, increased 
coverage of sea ice, and reduced marine surface water 
productivity (e.g. Porter, 1989; Bartlein et al., 1991; de Ver- 
nal and Pedersen, 1997). Much of the present day fresh- 
water range was physically covered with ice (Hamilton, 
1986). During the last glacial maximum, coho salmon must 
have been displaced from most of British Columbia and 
the Gulf of Alaska coast, now the center of their range. 
During each glacial advance, it is likely that marine and 
freshwater habitats of Pacific salmon populations were 
greatly reduced, especially for species that require freshwa- 
ter rearing, such as coho salmon. Alteration and reduction 
of natural ranges during these advances probably result- 
ed in the declines or extirpation of many populations. The 
patterns observed in the analysis of coho salmon mtDNA 
variation resulted from major demographic changes involv- 
ing the range of drainages sampled and are consistent with 
the effects that would be expected as a result of glacial ad- 
vances and retreats. For most of the last 500,000 years, the 
environment was much harsher than now, as indicated by 
the sea level ( a measure of the amount of the world’s wa- 
ter tied up in ice), which was more than 50 m lower than 
today during most of that period (Porter, 1989; Bartlein et 
al., 1991). Peaks of recent interglacial periods during which 
the environment probably approached modern conditions 
have occurred about every 100 thousand years, approxi- 
mately 120 thousand years ago (120 ka) , 200 ka, 330 ka, 
and 400 ka (Porter, 1989; Petit et al., 1999). 
A linkage between geologic record and molecular evolu- 
tion requires application of a “molecular clock” to relate 
observed nucleotide divergences to a mutation rate. The 
calibration of mtDNA clocks is contentious. A rate of 2% 
nucleotide substitution per million years (Brown et al., 
1982), referred to as the “standard clock,” has been broadly 
applied. However, the rates for salmonids may be slower, 
and rates of 1% per million years or less (Smith, 1992) may 
be more appropriate. 
An interpretation of our results based on the “standard 
clock” would be that divergence between clusters began 
136 ka at the beginning of the last interglacial period and 
that divergence within the clusters began 43 ka, at the 
end of the last interglacial period. The small number of 
haplotypes within each cluster is consistent with a rela- 
tively small effective number of females (N e( ^) within refu- 
gia followed by rapid expansion or the existence of several 
isolates within each refugium. Following the last glacial 
maximum, dispersion from one or more glacial refugia 
was followed by incomplete inter-refugial introgression. 
The objection to interpretations based on the “standard 
clock” is that the divergence rate is much faster than the 
rate generally accepted for salmonids. However, mtDNA 
clocks are calibrated from interspecies comparisons, which 
occur over times measured in millions of years (Thomas 
and Beckenbach, 1989; Martin et al., 1992; Bentzen et al., 
1993; Phillips and Oakley, 1997). The mutation rate is cer- 
tainly not homogeneous over the mtDNA genome, so it is 
possible that mutational hot spots dominate the rate in 
shorter time spans but are saturated and much less im- 
portant in estimates over long time spans. 
Although there are unquestionably strong demographic 
signals in the molecular evolutionary record of coho salm- 
on, we can not match them unequivocally to specific Pleis- 
tocene events. To make such a match, several pieces of 
information are needed. First, the data available for cali- 
brating a clock for the entire salmonid mtDNA genome 
are limited. Although there are numerous studies of par- 
ticular mtDNA regions, rate differences among regions 
compromise their utility. Second, short-term, intraspecific 
clocks based on hot spots must be developed and compared 
with long-term, interspecific clocks. We are now examin- 
ing data from other Pacific salmon species that share the 
northern Pacific Ocean range with coho salmon. Concor- 
dance among those results interpreted from the well-de- 
scribed Pleistocene environmental history may provide us 
with independent records of Pleistocene influences on the 
demography of Pacific salmon, which can be used to cali- 
brate intraspecific mtDNA clocks. 
The haplotype compositions of the populations that we 
studied leave us with questions about patterns of post- 
glacial colonization and influences of the Cascadian and 
Beringian refugial stocks. Acquisition and analysis of sam- 
ples from native populations in the Pacific Northwest 
should resolve questions about the composition of coho 
salmon in the Wisconsin Cascadian refugium. And, an in- 
tensive study of populations skirting the Alaska Peninsula 
and eastern Bering Sea should provide information about 
local colonization patterns and the composition of Berin- 
gian coho. 
