Gold et al ; Genetic studies of Scomberomorus cavalla in Florida 



499 



In general, results of the three approaches to homo- 

 geneity testing were fairly consistent, with one notable 

 exception. At Sca-37. probability values from the exact 

 test and the Roff-Bentzen procedure were 0.009 and 

 0.001 in the comparison of Atlantic, Florida Keys, and 

 Gulf samples, respectively, whereas the probability that 

 0[.j. differed from zero was 0.794 (Table 5). We examined 



this discrepancy further by carrying out "V" tests of ho- 

 mogeneity (DeSalle et al. 1987) for each allele at Sca-37. 

 Significant heterogeneity (P<0.05) was found only at 

 Sca-37*12: this allele was found only in the March 1997 

 sample from the Florida Keys (KEY''), where it occurred 

 at a frequency of 6.9'7f (Appendix Tables 2 and 3). Because 

 there were far fewer alleles at Sca-37 sampled from the 

 Florida Keys (258) than from either the Atlantic (884) or 

 Gulf (872), the disproportionate frequency of this allele 

 within the Florida Keys likely accounts for the signifi- 

 cance encountered in the exact test and the Roff-Bentzen 

 procedure. Given the absence of this allele in two of the 

 three samples from the Florida Keys, we do not believe 

 the significant heterogeneity detected at Sca-37 is mean- 

 ingful biologically. 



Although homogeneity testing of pooled samples in- 

 dicated that samples from the Atlantic differed from 

 samples from the Gulf at Sca-14 and Sca-23, and that 

 samples from the Florida Keys differed from the other two 

 at Sca-44. the allele-frequency differences were small and 

 accounted for only a fraction of the overall genetic vari- 

 ance. Results of AMOVA for the comparison of Atlantic with 

 Gulf samples revealed that on average 99.74% of the total 

 genetic variance at the seven microsatellites occurred 

 within samples, as compared to only 0.19% between re- 

 gions. For the comparison of Atlantic, Florida Keys, and 

 Gulf samples, 99.78'7( of the genetic variance on average 

 occurred within samples, whereas only 0.11% occurred 

 among regions. For both comparisons, the proportion of 

 the variation among samples within regions accounted 

 for the remainder of the genetic variance, and for both 

 comparisons, this proportion was small and statistically 

 nonsignificant. 



Finally, homogeneity tests were used to examine the 

 temporal stock boundaries currently used in management 

 of the king mackerel resource by classifying each of the 20 

 samples as either Atlantic or Gulf stock (Table 1). No sig- 

 nificant heterogeneity at any of the seven microsatellites 

 was found, providing no genetic evidence for existence of 

 temporal boundaries dividing Atlantic and Gulf migratory 

 units (stocks). 



Neighbor joining of Cavalli-Sforza's chord distances 

 between pairs of samples yielded little evidence of geo- 

 gi-aphic structure among the 20 samples. With few excep- 

 tions, samples from the same or geographically proximate 

 localities did not cluster together, and most nodes in the 

 topology (available from the first author) were supported 

 by well less than 50% of bootstrap proportions. Spatial au- 

 tocorrelation (SAAP) analysis also indicated the absence 

 of a relationship between allele frequency and geographic 

 distance. Initially, SAAP analysis employed both equal 

 geographic distances between each of five distance classes 

 and equal numbers of pairwise comparisons in each dis- 

 tance class. Analysis involving equal geographic distances 

 between distance classes generated an uneven number of 

 pairwise comparisons among distance classes, i.e. 18, 14, 

 15, 5, and 3 pairwise comparisons in distance classes 1-5, 

 respectively, resulting in a high variance in Moran's I val- 

 ues among alleles in distance classes 4 and 5. Accordingly, 

 the analysis was restricted to equal numbers of pairwise 



