Refuges of non-Bacillus thuringiensis (Bt) cotton, Gossypium hirsutum L., are used to delay Bt resistance in pink bollworm, Pectinophora gossypiella (Saunders) (Lepidoptera: Gelechiidae), a pest that eats cotton seeds. Contamination of refuges by Bt transgenes could reduce the efficacy of this strategy. Previously, three types of contamination were identified in refuges: 1) homozygous Bt cotton plants, with 100% of their seeds producing the Bt toxin Cry1Ac; 2) hemizygous Bt plants with 70–80% of their seeds producing Cry1Ac; and 3) non-Bt plants that outcrossed with Bt plants, resulting in bolls with Cry1Ac in 12–17% of their seeds. Here, we used laboratory bioassays to examine the effects of Bt contamination on feeding behavior and survival of pink bollworm that were resistant (rr), susceptible (ss), or heterozygous for resistance (rs) to Cry1Ac. In choice tests, rr and rs larvae did not differ from ss in preference for non-Bt versus Bt seeds. Survival of rr and rs also did not differ from ss on artificial outcrossed bolls (a mixture of 20% Bt and 80% non-Bt cotton seeds). On artificial hemizygous Bt bolls (70% Bt seeds) and homozygous Bt bolls (100% Bt seeds), rr had higher survival than ss, although rs and ss did not differ. In a simulation model, levels of refuge contamination observed in the field had negligible effects on resistance evolution in pink bollworm. However, in hypothetical simulations where contamination conferred a selective advantage to rs over ss individuals in refuges, resistance evolution was accelerated.