FISHERY BULLETIN: VOL. 87. NO. 3, 1989 



ratios (G, Table 7). For five of the 11 loci, allelic 

 frequencies are significantly nonindependent of 

 locality; significant heterogeneity of allelic fre- 

 quencies corresponds to Fst values ^0.019. 

 Considering just the 11 polymorphic loci, Nei's 

 (1978) unbiased genetic identity and distance 

 statistics for 36 pairs of population samples aver- 

 age 0.992 ± 0.001 and 0.008 ± 0.001, respec- 

 tively; over all 39 loci these statistics average 

 0.998 ± 0.0003 and 0.002 ± 0.0003, respectively. 

 Further analyses show that some of the heter- 

 ogeneity in allehc frequencies is geogi'aphically 



Table 7. — F-statistics and log-likelihood ratio (G) tests of 

 allelic frequency x locality Independence for polymorphic loci 

 in nine population samples of northern anchovy. 



'Probability levels for significance of G-lests are ' = 0.01 < P 

 < 0.05; •• = 0,001 < P < 0.01 ; ■" = P < 001 



patterned, some is associated with genetic dif- 

 ferences between the sexes and among age 

 classes, but that most is not associated with 

 any obvious environmental or biological factor. 

 Allehc frequencies at two of the five variable 

 loci, Pg»i and Xdh, are significantly correlated 

 with latitude (Fig. 4). An attempt was made to 

 examine the dependence of gene frequencies on 

 sex and age as well as locahty, but analyses of 

 cross-classified data are made difficult by small 

 sample sizes and uneven distributions of sexes 

 and age classes over localities. In tests of three- 

 factor (locality X SEX X ALLELIC FREQUENCY) 



log-linear models for Hbdh-2, Lgg, Pgm and 

 Xdh, sex was found to be associated with 

 ALLELIC FREQUENCY ouly foT Hbdh-2, for which 

 a fully saturated model with all three pairwise 

 interactions among factors appeared to be the 

 best fit. In 15 tests of log-hnear models of age x 

 SEX X ALLELIC FREQUENCY withiu individual 

 population samples, only four could not be fit by 

 the model of complete factor independence 

 (Hbdh-2 in population 9, Lgg in population 4, 

 Pgm in populations 3 and 4). Age and allelic 

 FREQUENCY Were associated only for Pgm in the 

 inshore Santa Cruz sample, for which, again, a 

 fully saturated model was the best fit. Interest- 

 ingly, interactions of age and sex independent of 

 (or only conditionally associated with) allelic 

 FREQUENCY Were significant in all four cases; an 

 association of older males with younger females 

 appeared to be responsible for this. 



0.55 0.56 0.57 0.58 0.59 0.60 



LATITUDE 



(SINE) 



Figure 4. — Frequencies of the common alleles for Pgm and Xdh (arcsine- 

 square-root transformed) plotted against the sine of latitude of collection site. 

 Spearman rank correlation coefficients for the two loci are rg = -0.783, t = 

 -3.330 for Pgm and rg = 0.812, t = 3.681 for Xdh. 



664 



