Several lines of indirect evidence suggest that selection imposed by predators may favor certain combinations of prey coloration and behavior at the expense of other combinations, but this hypothesis has never been tested experimentally. We manipulated color pattern and behavior of pygmy grasshoppers (Tetrix subulata) and exposed them to predation from domestic chickens. We painted grasshoppers either uniformly black or striped and manipulated their behavior by changing the ambient temperature. We found that the striped pattern enhanced grasshopper survival when reaction distance was short and jumping performance poor, but it decreased survival when reaction distance was long and performance high, relative to uniformly black individuals. To our knowledge, this is the first experimental demonstration that selection mediated by visual predators acts on the combination of prey color pattern and behavior. Further studies are necessary, however, to clarify how widespread correlational selection is in coevolved predator-prey relationships. Correlational selection may result in genetic coupling between traits, influence the dynamics of polymorphisms, and promote the evolution of sexual dichromatism in animals exhibiting sexual differences in behavior. Our results also illustrate the potential importance of visual illusions created by moving objects and suggest that it may be dangerous to make inferences about the relative survival value of different color patterns from the outcome of experiments that do not take into account prey behavior.