622 



Fishery Bulletin 94(4). 1996 



whether the survey has used protein electrophoretic, 

 mitochondrial DNA, or minisatellite DNA variation. 



In the current study, DNA bands were grouped in 

 bins or "alleles" on the basis of the estimated size of 

 the DNA fragment. The precision of estimated frag- 

 ment size was based on analyzing a single standard 

 fish on all gels, on obtaining repeated estimates of 

 fragment size, and on determining a relationship 

 between estimated fragment size and precision or 

 standard deviation of the estimate (Galbraith et al., 

 1991; Taylor et al., 1994). The bin width was set at 8 

 to 10 SD, less than the 16 SD suggested by Galbraith 

 et al. (1991) as necessary before the chances of as- 

 signing a band to the wrong bin falls below 0.05. 

 However, because most of the source of variation in 

 the Galbraith et al. (1991) study was due to digitiz- 

 ing bands that were manually marked on an acetate 

 overlay, the computerized estimation system em- 

 ployed in the present study should result in greater 

 precision of estimation of fragment size. Taylor et al. 

 ( 1994) used a bin width of 6 SD, based upon the mea- 

 surement precision of the fragment size of a stan- 

 dard fish, but a more conservative bin width of 8 to 

 10 SD was used in the current study. 



The variation observed at the minisatellite loci 

 examined in the current study was probably less than 

 that actually present at the loci. Given the level of 

 precision of the measurements of fragment size, 



Table 6 



Estimated percentage composition of 100-fish mixtures of 

 Yukon River chum salmon stocks resolved with a 5-stock 

 baseline in which distributions of the allele frequencies or 

 band counts in the baseline stock were fixed or resampled 

 as outlined in Table 4. Standard deviations are given in 

 parentheses. 



Distribution (%) 



Stock 



Actual Fixed Resampled 



Single-stock mixtures 



Andreafsky 100 100.0(0.0) 97.0(3.3) 



Sheenjek 100 98.2(1.7) 97.5(3.1) 



Fishing Branch 100 96.5(2.0) 95.1(4.1) 



Tatchun 100 100.0(0.0) 98.7(3.4) 



Kluane 100 100.0(0.0) 95.9(4.6) 



Multistock mixture 



Kluane 10 9.3(0.8) 10.0(2.4) 



Tatchun 20 20.3(0.7) 19.8(3.7) 



Sheenjek 30 30.3(0.4) 30.8(4.0) 



Fishing Branch 40 39.1(1.1) 38.8(4.5) 



closely spaced true alleles could not be distinguished 

 reliably and were therefore combined in the binning 

 process. It was assumed that there was no consis- 

 tent bias in the assignment of fragments to specific 

 bins. Additionally, differences in DNA sequence varia- 

 tion of two alleles of the same size were not deter- 

 mined. It is possible, for example, that an allele oc- 

 curring in both Japanese and North American popula- 

 tions will show the same length for both populations 

 but will reveal a considerably different DNA sequence. 



The simulations conducted for mixtures contain- 

 ing localized stocks (Japan, Yukon River, and Fraser 

 River) suggested that the minisatellite DNA varia- 

 tion surveyed may provide relatively accurate and 

 precise estimates of individual stocks in the mixtures. 

 Although the simulations were encouraging, the per- 

 formance of the minisatellite DNA variation for esti- 

 mation of stock composition needs to be verified by 

 actual applications in field situations, where fish in 

 the mixture can originate from stocks not found in 

 the baseline, and where novel genotypes can be ob- 

 served. In the present study, sample sizes of the 

 baseline stocks were usually less than 50 fish; there- 

 fore, increased sampling of individuals in the stocks 

 would be required in order to obtain representative 

 distributions of allele frequencies or band counts. 



The simulations suggested that accurate estimates 

 of stock composition of individual stocks may be pos- 

 sible for mixtures of either Fraser River or Yukon 

 River chum salmon. A previous survey of variation 

 at seven protein-coding loci indicated that there was 



