Noll et al.: Analysis of genetic structure of Asian and western Oncorhynchus gobuscha 
131 
'\ 
PWS(17D) s °c? 
• - • - PWS(17C) 
\ 
\ 
v 
PWS(17B) 
Hokkaido E. Kamchatka 
till 1 i ■ — i — — i — i i — i 1 1— — a 
0 0.01 0.02 0.03 0 04 0.05 0 06 0 07 
Genetic distance 
Figure 2 
Neighbor-joining tree of even-year pink salmon broodlines from Japan, Russia, and 
Prince William Sound based on Cavalb-Sforza and Edwards (1967) chord distances 
from allele frequencies at 36 loci. Significance between nodes or for collections joined 
at nodes were tested within regions by using log-likelihood ratios (G-tests; Sokal and 
Rohlf, 1995) of the 36 loci to test homogeneity of branches joined at a node, and 
between basins by tests of homogeneity among the populations of the two adjacent 
regions. The numbers following the population names correspond to their locations in 
Figure 1. PWS = Prince William Sound. 
clustered with the eastern and western Kamchatka collec- 
tions (not shown). The arrangement of the other clusters 
remained as in Figure 2. 
The variation at 36 loci common to all collections 
was partitioned by using gene diversity analysis (G ST ’ s; 
Chakraborty and Leimar, 1987) and analysis of variance 
(0 ST ’s; Weir and Cockerham, 1984; Weir, 1996) (Table 3). 
Analysis of variance accounts for bias inherent in analyses 
that use finite samples from and small numbers of popula- 
tions. Differences between the G sr ’ s and d ST ’s are most ap- 
parent at upper levels of the hierarchy where fewer groups 
are compared. Although biased, the G ST ’ s have been wide- 
ly used, so we summarize results of partitioning variation 
with G st ’s but included 0 ST ’s parenthetically. On average, 
variation within streams accounted for 97.2% (95.57%) of 
the total, with 0.63% (0.05%) attributable to differences 
among streams within regions, 0.65% (1.29%) to differences 
among regions within continents, and 1.52% (3.09%) to 
differences between continents. 
We combined our data with available data (Altukhov et 
al., 1983; Salmenkova and Omelchenko, 1983; Zhivitovsky 
et al., 1989; Kartavtsev, 1991; Kartavtsev et al., 1992) to 
conduct a log-likelihood ratio analysis to test for homo- 
geneity within and among Asian regions. The primary 
(variable) loci for which data were available were PGDH* 
(omitting collections missing the *, 95 allele), G3PDH-1 *, 
and MDH-B1,2*. These data included collections from Pri- 
morie, western Sakhalin Island, and the Kuril Islands 
in addition to the regions we reported (Table 4). There 
was no overall heterogeneity within regions and only 
a single test suggested heterogeneity. However, PGDH* 
and MDH-B1,2* exhibited strong heterogeneity (P<10 4 ) 
among regions. 
Data from an earlier study of genetic diversity in Alaska 
pink salmon (Gharrett et al., 1988) were included to pro- 
duce an unrooted neighbor-joining tree based on 21 loci 
common (Appendix 1) to that study and the present one 
(Fig. 3). Data from all collections were condensed by re- 
gion, except for Japan, where the collections were hetero- 
geneous. Some regions were represented by a single collec- 
tion; data for the Aleutian Islands collections were pooled 
because they exhibited no heterogeneity, but as a group 
differed from other Western Alaska regions (Gharrett et 
al., 1988). The Magadan collections were omitted from our 
analysis because data from several of the 21 loci were un- 
available. As in the previous analysis, the tree showed a 
clear geographic basis for genetic variation. A cluster con- 
taining the Japan and Sakhalin Island collections adjoined 
the western Kamchatka group. A longer span joined the 
latter with a cluster consisting of the eastern Kamchat- 
