We suggest several possible explanations for the markedly different levels of intraspecific variation within glaucus and canadensis. 1) The two species differ in population structure such that gene flow is more restricted in glaucus than canadensis. 2) P. canadensis originated as a very small population, perhaps as a result of Pleistocene isolation in some periglacial refugium (Scriber 1988), and low variation is a result of the associated genetic bottleneck. 3) P. canadensis is of such recent evolutionary origin that populations have not yet differentiated, but will in the future. 4) There is less variation in canadensis host use; perhaps Populus tremuloides is the prevalent host for all canadensis popuiations, and other host-use records are rare "oviposition mistakes" that contribute only insignificantly to selective pressures shaping the nutritional physiology of canadensis. 5) The boreal forest is characterized by low phytochemical diversity relative to that of temperate forests, the most ubiquitous boreal hosts being restricted to four genera in two families: Populus, Salix, Betula, and Alnus. Consequently, natural selection has favored the development of a single detoxification system that is optimal for most hosts, i.e. canadensis detoxification systems have been canalized such that mutations and recombination seldom lead to phenotypic variation.

Hypothesis 1 can be tested by comparing the divergence in allozyme frequencies between populations within species. If it holds true, the fixation index (Fst) of glaucus populations should be greater than that of canadensis populations separated by the same geographic distance. The available data are limited, but they indicate extensive gene flow within both species (Hagen 1990); the estimated genetic distance between P. glaucus from Ohio and Florida was no greater than that between P. canadensis from Alaska and Michigan: Nei's genetic identities equalled 0.99 in both cases (Hagen and Scriber 1990). Hypotheses 2 and 3 cannot account for divergence between canadensis populations in traits other than nutritional physiology such as egg size, adult size, and temperature responses. Hypothesis 2 would be supported by lower levels of allozyme heterozygosity in canadensis than in glaucus (Bonnell and Selander 1974). Hypothesis 3 predicts that other recently derived swallowtail taxa (as estimated from allozyme and mtDNA sequence divergence) would be similarly invariant in their nutritional physiology. Hypothesis 4 questions our view of P. canadensis as a polyphagous herbivore and requires that suitable hosts go generally unused. It can be tested through quantification of oviposition patterns in the field. We presently favor hypothesis 5, which predicts that the various secondary metabolites present in boreal hosts are detoxified by canadensis using only one or a few biochemical systems. It further predicts that other polyphagous herbivores of the boreal forest will similarly exhibit low variation in nutritional physiology, and will be consistently less variable than related species of polyphagous herbivores from temperate and tropical forests (Mattson et al. 1988, Mattson et al. in this volume). An important correlate is that physiological preadaptation to alternative hosts will be more common in boreal forests than elsewhere. That is, boreal herbivores challenged with sympatric nonhost plant species should be more likely to successfully detoxify that foliage than temperate or tropical herbivores challenged with nonhost species from their environment.