206 



Fishery Bulletin 90(1). 1992 



Table 2 



Discriminant function analysis results of different Atlantic herring samples from known spawning grounds. 



Sample 



1983 discriminant function 

 construction sample 

 (N = N, + N2 = 100) 

 Jeffries' Ledge fNi = 50) 

 Trinity Ledge (N2=50) 



1983 Sample (N = Ni + N2 = 100) 

 Jeffries' Ledge (Ni = 50) 

 Trinity Ledge (N2 = 50) 



1984 Sample (N = Ni + N, = 198) 

 Jeffries' Ledge (Ni = 99) 

 Trinity Ledge (N, = 99) 



Number from each spawning 

 ground classified as: 



Percent from each spawning 

 ground classified as: 



Jeffries' Ledge Trinity Ledge Jeffries' Ledge Trinity Ledge 



Overall correct 

 classification 



87% 



77% 



54% 



accounted for most of the between-group variation in 

 the 1983 sample. The percent correct classification by 

 spawning group of three sets of samples (two from 

 1983, one from 1984) separated by the derived discrimi- 

 nant function is found in Table 2. Overall misclassifica- 

 tion of fish collected in 1983 was 18%, while that of 

 fish collected in 1984 was 46%. 



The phenotypic variation of the unadjusted measure- 

 ments was partitioned similarly within each morpho- 

 metric character, except AH (Table 3). The partition- 

 ing of the phenotypic variation averaged across all 

 characters is found in Table 4. None of the variation 

 was explained by differences between spawning 

 groups, while approximately one-half was partitioned 

 within a spawning group between years. The remain- 

 der of the variation was within a sample. One-way 

 ANOVA of between-year differences within a spawn- 



ing group showed that within the TL group the means 

 of all the characters, except AH (;?< 0.02), were highly 

 significantly different (p< 0.0001) between 1983 and 

 1984. In contrast, within the JL group three char- 

 acters—distance between the insertions of the pelvic 

 fins (PVI), AH, and PCI— were not significantly dif- 

 ferent between years. The remaining characters were 

 significantly different (p<.05) between years. 



Electrophoresis 



Allelic frequencies within each sample for the four loci 

 chosen for analysis are found in Table 5. Other enzyme 

 systems were also investigated, but few specimens ex- 

 pressed enzyme activity at these loci (Safford 1985). 

 We chose these loci because they had previously been 

 shown to be polymorphic and to follow Mendelian in- 



