Predators can shape genetic correlations in prey by altering prey perception of risk. We manipulated perceived risk to test whether such non-consumptive effects tightened behavioural trait correlations in wild-caught stickleback from high- compared to low-risk environments due to genetic variation in plasticity. We expected tighter genetic correlations within perceived risk treatments than across them, and tighter genetic correlations in high-risk than in low-risk treatments. We identified genetic variation in plasticity, with genetic correlations between boldness, sociality, and antipredator morphology, as expected, being tighter within treatments than across them, for both of two populations. By contrast, genetic correlations did not tighten with exposure to risk. Tighter phenotypic correlations in wild stickleback may thus arise because predators induce correlational selection on environmental components of these traits, or because predators tighten residual correlations by causing environmental heterogeneity that is controlled in the laboratory. Our study places phenotypic integration firmly into an ecological context.