374 



Fishery Bulletin 105(3) 



The estimated ts/tv ratio was 2.1:1. As was the case in 

 comparisons using microsatellites, the mtDNA distance 

 among species (D^) was lowest in comparisons between 

 species within small-scaled and large-scaled groups 

 (Table 2). 



Hierarchical likelihood ratio tests indicated that the 

 K81uf-i-I-(-r model (Kimura, 1981) was the most ap- 

 propriate model of sequence evolution at the control 

 region locus. This model also had the second highest 

 AIC weight, at 0.23. The KSluf-i-I-i-r model approxi- 



mates the rate of heterogeneity across sites (a property 

 which can simply be described as unequal probabilities 

 of mutations among sites) by using a gamma distribu- 

 tion, after variant sites have been removed. Haplotype 

 phylogenies generated with both the M-L method and 

 the N-J cluster method indicated three divergent lin- 

 eages of menhadens (Fig. 4), and these lineages were 

 identical in haplotype membership between the two 

 tree-building methods. The first lineage had 99'7f boot- 

 strap support and included haplotypes contributed from 

 both B. patronus and B. tyrannus. This main 

 group of large-scaled haplotypes (n = 59) was 

 characterized primarily by transitions and 

 had minimal (<7) transversions among hap- 

 lotypes. A second, small group of divergent 

 Atlantic menhaden occurred in the haplotype 

 phylogeny, with 100% bootstrap support. 

 This second group comprised the remaining 

 eight large-scaled haplotypes and differed 

 from the first group by 14-i- transversions 

 per haplotype, likely the result of mutation 

 saturation. Plots of transitions and trans- 

 versions indicated that this data set was 

 saturated; the rate of transversions begins to 

 rival that of transitions at approximately a 

 Kimura distance of 0.10 (Fig. 5A). Although 

 saturation is indicated in cases where di- 

 vergent haplotype clades are compared, 

 saturation is not indicated in comparisons 

 involving within-clade comparisons (Fig. 5, 

 B-D). Thus, the main effect of saturation on 

 this data set was on longer (more divergent) 

 branches. A third lineage on the maximum 

 likelihood tree had 100% bootstrap support 

 and consisted of haplotypes contributed by 

 B. gunteri and B. smithi in paraphyly. Hap- 

 lotypes in this clade indicated incomplete 

 sorting of mtDNA haplotypes within the 

 small-scaled menhadens. However, there was 

 clear evidence for genetic divergence between 

 this group and both large-scaled clades, and 

 100% bootstrap support for monophyly of the 

 small-scaled haplotypes. 



Discussion 



The morphological differences among the four 

 North American menhaden species have been 

 examined previously at egg, larval, juvenile, 

 and adult stages (Dahlberg, 1970 [and ref- 

 erences therein]; Hettler, 1984; Ahrenholz, 

 1991; Tolan and Newstead, 2004), and previ- 

 ous classifications have the species divided 

 between small-scaled and large-scaled groups 

 (Dahlberg, 1970). The results presented here 

 reinforce the conventional hypothesis that B. 

 smithi and B. gunteri share recent common 

 ancestry as do B. patronus and B. tyran- 

 nus. Both classes of genetic markers showed 



