Inducible defenses of prey and inducible offenses of predators are examples of adaptive phenotypic plasticity. Although evolutionary ecologists have paid considerable attention to the adaptive significances of these strategies, they have rarely focused on their evolutionary consequences. Because the functional phenotypes of predator and prey determine strength of interactions between the species, the inducible plasticity can modify selective pressure on trait distribution and, ultimately, trait evolution in the interacting species. We tested this hypothesis in a predator–prey system composed of salamander larvae (Hynobius retardatus) and frog tadpoles (Rana pirica) capable of expressing antagonistic inducible offensive or defensive traits, an enlarged gape in the salamander larvae and a bulgy body in the tadpoles, when predator–prey interactions are strong. We examined selection strength on the tadpole's defensive trait by comparing survival rates of tadpoles with different defensive levels under predation pressure from offensive or non-offensive salamander larvae in a lab-experiment. Survival rates of more-defensive tadpoles were greater than those of less-defensive tadpoles only when the tadpoles were exposed to offensive salamander larvae; thus, the predator's offensive phenotype could select for an amplified defensive phenotype in their prey.