772 



Fishery Bulletin 101(4) 



at north-south of the 15-20°N range, spatial homogene- 

 ity was tested at the following levels: overall samples (O), 

 among longitudinal regions (L; coastal, intermediate, and 

 offshore localities), and latitudinal regions (N; north-south 

 of the 15-20° range). The Gulf of California sample was 

 excluded from this analysis because the large variation 

 found in this sample (reflected in the significant differ- 

 ences shown in allele frequency homogeneity from pairwise 

 comparisons) would not allow an accurate assessment of 

 whether longitudinal or latitudinal differentiation exists or 

 not. Significance levels were adjusted for multiple testing 

 through the Bonferroni sequential method (Rice, 1989). 



Population subdivision was estimated by using the 

 Weir and Cockerham (1984) method through the TFPGA 

 program. Standard error and confidence intervals were 

 obtained through jackknife and bootstrapping procedures, 

 respectively, with Fgj, Pro 1.0 (Weir, 1990). Estimates of pop- 

 ulation subdivision were partitioned into the following lev- 

 els: O, over all the samples; L, longitudinal regions (coastal, 

 intermediate and offshore); and N, latitudinal regions. 



We used the 6 statistic to estimate gene values between 

 sample pairs (Slatkin, 1993) that are defined as Mg. An 

 "isolation by distance" model was evaluated from the cor- 

 relation between the distance between localities measured 

 as geographic separation in nautical miles (nmi), and the 

 M(j values by means of the Mantel test (Hellberg, 1994) in 

 both allozymes and RAPDs. 



The patterns of the amplification products resulting from 

 the RAPD analysis were subjected to the same analyses 

 as allozymes with the procedures described in Lynch and 

 Milligan (1994). RAPD fragments were interpreted under 

 the following assumptions: 1) fragments were considered 

 to behave as dominant genes (Williams et al., 1990); 2) 

 every polymorphic fragment was considered derived from 

 a two-allele locus; 3) the equilibrium of Hardy- Weinberg 

 was assumed for all genotypes, and 4) each fragment was 

 considered to be an independent locus. 



Only those fragments clearly defined and having consis- 

 tent intensity were recorded. Because of this, the Michoacan 

 sample, with poor consistency in the banding patterns, was 

 excluded from the RAPD analysis. The allele frequency of 

 every fragment was calculated on the basis of the inferred 

 homozygous recessive genotypes. Because of the dominant 

 nature of the alleles, and in order to correct the bias originat- 

 ed by calculating the recessive allelic frequencies, we chose 

 the estimation based on the Taylor expansion (Kendall and 

 Stuart, 1977, cited m Le Corre et al., 1997) as implemented 

 in TFPGA program. This reduction on the bias is based on 

 the equation resulting from the second order expansion 

 of Taylor (see details in Lynch and Milligan, 1994). 



Results 



Allozymes 



eight polymorphic loci detected are shown in Table 1. After 

 adjusting levels of significance by the Bonferroni procedure, 

 significant deviation of genotypic frequencies from those 

 expected under Hardy-Weinberg was found in the loci Lgg 

 and Pap-F* for the Gulf of California sample and in the Aat- 

 S* locus for two localities — west of Revillagigedo Islands, 

 and the Gulf of California (P<0.0006, Table 1). Deviations 

 displayed for both locations corresponded to a heterozygous 

 deficit (P<0.0006, after Bonferroni correction). 



Comparison of allozyme allele frequencies among all 

 collections (overall) by the exact probability test revealed 

 significant heterogeneity at loci Glud, La, and Lgg, after 

 Bonferroni adjustment (P<0.006). Pairwise comparisons 

 among samples to test allele homogeneity gave signifi- 

 cant differences for nine of 45 comparisons after correc- 

 tion for multiple tests (P<0.001), seven of which involved 

 comparisons with the Gulf of California (GC) sample. The 

 remaining significant differences were between Guerrero- 

 Nayarit, and Cape San Lucas-Nayarit comparisons (Table 

 2), resulting from significant heterogeneity at Glud, La, 

 and Lgg loci. 



In general, allozyme analysis displayed low levels of dif- 

 ferentiation. The 9 value over all loci was different from 

 zero (P<0.05) and showed that 4.8% of the variance was 

 attributable to differences among samples (Table 3). 



Individual loci showed d values ranging between 0.0037 

 and 0.27. Highly significant values at loci La (9=0.13 

 ±0.089) and Lgg (0=0.27 ±0.253) evidently resulted from 

 their weak polymorphism in some samples (Table 3). 



Allele frequency homogeneity was tested among coastal, 

 intermediate, and offshore regions (Table 4). Significant 

 heterogeneity was found by exact test between coastal 

 and offshore comparisons (P=0.0043) but was found to be 

 not significant between coastal and intermediate regions 

 (P=0.0632). Subdivision as measured by among coastal, 

 intermediate, and offshore localities was not different from 

 zero. However, for the Gpi-F* locus the population subdivi- 

 sion among regions (0.0058) was twice as large as that noted 

 among samples (0=0.003), but neither value was significant 

 (Table 3). No latitudinal differentiation by the exact test or 

 population subdivision estimated by the 9 index was found 

 between north and south regions (data not shown). 



The gene flow values (Mg) were high (mean 24.8 mi- 

 grants per generation). A lack of correlation between gene 

 flow estimations and geographic distance by means of the 

 Mantel test was observed (r'-=-0.144; P=0.22), resulting in 

 a rejection of the isolation by distance model. 



For the Gpi-F* locus, paired tests of significance (data 

 not shown) showed discrepancies in the Gpi-F* 175 allele 

 frequencies among localities from the coast with those 

 located at the CYRA limits (Fig. 1, Table 5). Four of eight 

 comparisons of coastal and offshore samples revealed 

 differences of allele frequencies at this locus (P<0.05), 

 although none of these differences was significant after 

 correction for multiple testing (P>0.001). 



Of the 28 analyzed allozyme loci, eight (28.5%) showed 

 polymorphism under the 0.95 criterion. The observed het- 

 erozygosities per sample over all allozyme loci ranged from 

 0.027 to 0.083 (mean 0.052). Allozyme frequencies for the 



RAPDs 



The primer OPF-10 produced 11 amplified fragments, with 

 sizes from 200 to 600 bp (base pairs). Four of the fragments 



