TRADEOFFS IN PERFORMANCE ON DIFFERENT HOSTS: EVIDENCE FROM WITHIN‐ AND BETWEEN‐SITE VARIATION IN THE BEETLE DELOYALA GUTTATA

Host races of phytophagous insects are sympatric populations that have different host preferences and between which gene flow is restricted because of the difference in host preference. Sympatric, host-associated sibling species are sympatric populations that use different host plants and that do not interbreed because of the presence of isolating mechanisms not related to host preference (Jaenike, 1981; see Mayr, 1970, and Bush, 1969, for slightly different definitions of these terms). Although the existence of host races and host-associated sibling species of phytophagous insects has been suspected for over 50 years (e.g., Thorpe, 1930), the evolutionary mechanism of host race formation remains controversial. In particular, there is disagreement about the importance of sympatric divergence in generating host races and species. White (1978) has argued that sympatric speciation must be invoked to explain, in part, the great diversity of specialized herbivorous insects, while Bush (1974, 1975) has forcefully advocated the operation of sympatric divergence in the creation of sympatric host races of tephritid flies. By contrast, Futuyma and Mayer (1980; see also Futuyma, 1983a) have argued that there is no reliable experimental evidence to support these claims and that models of sympatric speciation are based on assumptions that are probably not met by most phytophagous insects. Several authors have developed formal models of sympatric divergence and speciation (Maynard Smith, 1966; Dickinson and Antonovics, 1973; Caisse and Antonovics, 1978; Pimm, 1979; Felsenstein, 1981). Although these models do not account for the evolution of differences in host plant preference associated with purported sympatric speciation in phytophagous insects, Bush and Diehl (1983) and Rausher (1984) have suggested how modification of these models could remedy this problem. A common assumption of all these models is that fitness on one host is negatively correlated with fitness on a second. This type of negative correlation is manifested in these models in the assumption that genetic variation exists at loci that exhibit a "crossing" genotype x host plant interaction, i.e., at loci at which some genotypes have high fitness on one host but low fitness on a second host while other genotypes have low fitness on the first but high fitness on the second host. This negative correlation seems to be necessary for sympatric speciation because it permits linkage disequilibrium to be established between loci influencing, say, viability on different hosts and loci influencing host preference. The resulting coadapted preference-viability gene complexes represent incipient host races or species. Moreover, the breakdown of these coadapted gene complexes by recombination provides the selection pressure that improves reproductive isolation, whether isolation is achieved via mating on the host (e.g., Bush, 1974, 1975) or by a separate assortative mating locus in linkage disequilibrium with loci affecting preference and fitness (e.g., Felsenstein, 1981). Without a crossing interaction, one homozygote genotype would have maximal fitness on both host species. Consequently, unless reproductive isolation were achieved instantaneously by a mutation