Models of habitat selection often assume that organisms choose habitats based on their intrinsic quality, regardless of the position of these habitats relative to low-quality habitats in the landscape. We created a habitat matrix in which high-quality (predator-free) aquatic habitat patches were positioned adjacent to (predator-associated) or isolated from (control) patches with single or two species of caged predators. After 16 days of colonization, larval insect abundance was reduced by 50% on average in both the predator and predator-associated treatments relative to isolated controls. Effects were largely similar among predator treatments despite variation in number of predator species, predator biomass, and whether predators were native or nonnative. Importantly, the strength of effects did not depend on whether predators were physically present. These results demonstrate that predator cues can cascade with equal strength across ecological boundaries, indirectly altering community assembly via habitat selection in intrinsically high-quality habitats.