General predictions of community dynamics require that insights derived from local habitats can be scaled up to explain phenomena across geographic scales. Across these larger spatial extents, adaptation can play an increasing role in determining the outcome of species interactions. If local adaptation is common, then our ability to generalize measures of species interaction strength across communities will be limited without an additional understanding of the genetic variation underlying interaction traits. In the context of predator–prey interactions, prey individuals commonly are expected to reduce risky foraging behaviors and subsequent growth under predation threat. However, rapid growth into a large body size can defend against gape-limited predators, creating a tradeoff between increased predation risk due to elevated foraging activity and decreased predation risk due to large size. Here I combine field observations, natural selection experiments, and common garden assays to understand potential adaptations of spotted salamander Ambystoma maculatum larvae to gape-limited and gape-unconstrained predators. Field observations and natural selection trials suggested antagonistic selection on prey body size among ponds dominated by gape-limited predator salamanders A. opacum and gape-unconstrained beetle larvae Dytiscus. In common garden experiments, prey from sites with high gape-limited predation risk grew larger than those from other sites, suggesting the evolution of rapid growth into a prey size refuge. Larvae from all sites grew to a large size when exposed to the gape-limited N. viridescens predator's kairomones. Hence, induced rapid growth into a size refuge may be an adaptive response to gape-limited predation risk. Results point to an important role for cross-community generalizations based on functional classifications of predators by their gape constraints and inter-site genetic variation in prey growth rates and behaviors.