Phenotypic plasticity has important ecological consequences because the strengths of species interactions can change with the behavior and morphology of interacting individuals. Evolutionary studies of plasticity can predict conditions under which shifts in phenotypes will occur and, therefore, may modify species interactions. We studied evolutionary mechanisms maintaining an induced response to predators in Triturus newt larvae, which are among many taxa in freshwater habitats exhibiting predator-induced plasticity. When exposed to caged (nonlethal) Aeshna dragonfly larvae, newts of two species (T. alpestris and T. helveticus) spent more time hiding in the leaf litter, had darker pigmentation in the tail fin, and developed larger heads and larger tails relative to their body size, in comparison with newts in predator-free ponds. The two phenotypes faced a performance trade-off across environments with and without odonates: the predator-induced phenotype survived twice as well as the no-predator phenotype when exposed to free dragonflies, but the predator-induced phenotype of both species grew more slowly until just before metamorphosis. For Triturus alpestris, a direct comparison of performance between phenotypes was complicated because predator-induced newts emerged later in the summer but at a larger body size. Nonrandom mortality imposed by hunting dragonflies caused selection favoring increasing tail size, but we found no selection on specific traits in predator-free ponds. Head shape was not subject to selection in either environment; we suspect that head shape is involved in consuming different prey in the presence and absence of predators and is unrelated to predator escape. Triturus in 25 natural populations from which we collected quantitative samples in 1997 and 1998 exhibited extreme spatial variation in predation regime (density of large predators ranged from 0 to 24 individuals/m2). Variation among populations in head shape was exactly as predicted by experimental results (Triturus of both species had relatively large heads when exposed to predators), but results for tail shape were consistent with the experiments in only one of the two years. The evolutionary mechanisms maintaining plasticity in Triturus and other amphibian larvae should apply to many organisms inhabiting freshwater ponds, so trait-mediated indirect effects seem especially likely to occur in these habitats.