The interaction of population dynamics and movement among two habitat types (toxic transgenic fields and nontoxic refuge fields) on the evolution of insecticide resistance was examined in two different simulation models. The two models were developed to test the hypothesis that increasing habitat grain from fine-grained to coarse-grained, and the resultant increase in nonrandom mating, would increase the rate of local adaptation, here the evolution of resistance. The first model, a complex, stochastic spatially explicit model, altered habitat grain by varying adult dispersal rates between habitat patches. In contrast to the expectation that increasing patch isolation and increasing the coarseness of the habitats would increase the rate of resistance evolution, intermediate levels of dispersal actually delayed resistance by as much as fivefold over the range of dispersal levels observed. Source-sink dynamics related to ovipositional patterns and the related population dynamics appear to explain the results. A simple deterministic model was developed to abstract out the separate impacts of mating and ovipositional behaviors. This model showed qualitatively the same results, although under similar assumptions it predicted much longer delays in resistance evolution. In this model, nonrandom mating alone always increased the rate at which insects adapted to transgenic crops, but nonrandom mating in combination with nonrandom oviposition could significantly delay resistance evolution. Differences between the two models may be due to the population regulation incorporated in the spatially explicit model. The models clearly suggest that resistance management programs using untreated refuges should not over-emphasize random mating at the cost of making the habitat too fine-grained.