Recently, much attention has focused on the patterning of intraspecific genetic variation among solitary, primitively eusocial, and advanced eusocial Hymenoptera. Berkelhamer (1983), after an analysis of electrophoretic data from 51 hymenopteran species, concludes that primitively eusocial species are less genetically variable than are either solitary or advanced eusocial species. She believes this pattern supports the idea that eusociality arose in species with relatively inbred populations, and she considers the implications of inbreeding for genetic models of the origin of hymenopteran eusociality. We have discovered two serious errors in Berkelhamer's analysis: 1) several species are assigned an incorrect social classification, and 2) her statistical comparisons of mean heterozygosity values (H) of solitary, primitively eusocial, and advanced eusocial groups fail to control for possible taxonomic biases in genetic variability. If these errors are corrected, a re-analysis of the same data yields results that differ markedly from those presented by Berkelhamer. In Table 2, Berkelhamer incorrectly classifies the four wasp species of the genus Polistes as advanced eusocial species. Polistes societies are universally classified by insect sociobiologists as primitively eusocial relative to the advanced eusocial species Berkelhamer considers: Polistes species exhibit all of the characteristics Berkelhamer herself (p. 543) views as diagnostic of primitive eusociality (i.e., relatively small nest size, little or no morphological caste differentiation, and behavioral domination of the workers by the queen; Evans, 1958; Evans and Eberhard, 1970; Wilson, 1971 p. 8; Spradbery, 1973; Akre, 1982). Conversely, the advanced eusocial wasp species, Vespula vulgaris, which displays large, sometimes perennial, colonies, pronounced morphological caste differentiation, and chemical queen control, is erroneously categorized as primitively eusocial. These errors have a large effect on Berkelhamer's results since the Polistes species have relatively high H values and the Vespula species has an H of 0.000. Berkelhamer's statistical analyses of the data she presents in Table 2 are vulnerable to taxonomic biases. She made pairwise comparisons of the 1H values of three groups: advanced eusocial species (N = 24; 18 ant, 2 bee, and 4 wasp species), primitively eusocial species (N = 9; 8 bee and 1 wasp species), and solitary species (N = 18; 10 bee, 7 wasp, and 1 sawfly species). Clearly, differences in H among these groups could reflect taxonomic differences, instead of social level differences, in genetic variability. For example, ants are disproportionately represented in the advanced eusocial group, and bees are overrepresented in the primitively eusocial group. If ants were more genetically variable than bees for reasons unrelated to differences in social level (e.g., because of differences in population structure or ecology), then one would erroneously conclude from a simple pairwise comparison that advanced eusocial species are more genetically variable than are primitively eusocial species. Indeed, we found that ft (0.045) of all the (advanced eusocial) ant species is significantly greater than ft (0.011) of the two advanced eusocial bee species (P = 0.05, two-tailed Mann Whitney test), a result that is consistent with a possible taxonomic bias. One simple way to reduce a potential taxonomic bias without a two-way analysis of variance procedure is to make statistical comparisons only within taxonomic groups. Since eusociality likely arose independently in ants, bees, and wasps (Wilson, 1971), we believe the latter procedure is especially well-suited for providing independent tests of a putative association between genetic variability and level of sociality. We have corrected Berkelhamer's classification errors, and, using the data she presents in Table 2, we have compared the ft values of solitary, primitively eusocial, and advanced eusocial species within bees and within aculeate wasps (no comparisons were made within ants since all ant species are advanced eusocial). Within the wasps, primitively eusocial species exhibit the highest H (0.062), a trend contrary to that reported by Berkelhamer, but this value is not significantly different from H (0.052) of solitary wasp species (P > 0.10, twotailed Mann Whitney test). The single advanced eusocial wasp species (Vespula vulgaris) has a reported H of 0.000, but meaningful statistical comparisons with the other two groups are not possible. Within bees, solitary species have a higher H (0.032) than do primitively eusocial bees (H = 0.013), but the difference is not significant (P > 0. 10, two-tailed Mann Whitney test). The two advanced eusocial bee species exhibit the lowest H (0.01 1) but, again, insufficient sample size prevents meaningful comparisons. We therefore find no evidence that primitively eusocial species tend to be less genetically variable than solitary species as Berkelhamer claims. We
Read full abstract