The regional dynamics of the Bay checkerspot butterfly (Euphydryas editha bayensis) were studied in a 15 x 30 km region in Santa Clara County, Calif., USA, in which the butterfly forms discrete populations on scattered patches of serpentine grassland. Analysis of the butterfly's distribution on 60 patches of potential habitat showed that the ten populations extant in 1987 formed a metapopulation, linked by extinction and mutual recolonization. Eighteen vacant habitats were as suitable for the butterfly, in a composite measure of habitat quality, as the ten habitats which supported populations; the former were apparently unoccupied because of their distance from a source of colonists. The metapopulation is shown to have mainland and island" dynamics, in which a single large population is the source of nearly all colonization. Models of the dynamics of this metapopulation, based on the butterfly's distribution and additional evidence, predicted an equilibrium configuration fairly similar to the present one. Field experiments, including within-habitat and out-of-habitat releases, showed that the Bay checkerspot butterfly has relatively low capacities for interhabitat dispersal, habitat finding, and establishment in vacant habitat. Earlier work on this species had already shown that it has a low propensity to disperse from its habitat. The Bay checkerspot's evident lack of the adaptations associated with colonizing ability, despite its proneness to local extinction, demonstrates that its regional persistence depends on the existence of one or several large populations. Simulation modeling of metapopulation dynamics quantified the importance of independent local environments to regional species persistence. In a standard metapopulation model with binary local dynamics, introducing correlated temporal fluctuations among the component populations in their chances of becoming extinct reduced total persistence time. However, this effect was directly proportional to the values of three parameters: number of populations, mean extinction rate, and variance in the extinction rate. Environmental independence vs. correlation is therefore likely to affect persistence to the greatest degree in large metapopulations of short-lived organisms (such as many insects) which are highly prone to local extinction. The effect of repeated destructive sampling for scientific study on two Bay checkerspot butterfly populations at Jasper Ridge Biological Preserve (Stanford University, San Mateo County, Calif.) was analyzed through "bootstrap" computer simulations. Mean expected sizes of the populations were reduced by 80% by the sampling performed during 23 years of study (1961-1983). However, this effect was not statistically detectable at any time during the study period, due to the strong natural fluctuations in population size. The sampling significantly increased the risk of extinction for the two populations, but this effect was considerably diminished when density depended ..."