Novel environments often impose directional selection for a new phenotypic optimum. Novel environments, however, can also change the distribution of phenotypes exposed to selection by inducing phenotypic plasticity. Plasticity can produce phenotypes that either align with or oppose the direction of selection. When plasticity and selection are parallel, plasticity is considered adaptive because it provides a better pairing between the phenotype and the environment. If the plastic response is incomplete and falls short of producing the optimum phenotype, synergistic selection can lead to genetic divergence and bring the phenotype closer to the optimum. In contrast, non-adaptive plasticity should increase the strength of selection, because phenotypes will be further from the local optimum, requiring antagonistic selection to overcome the phenotype-environment mismatch and facilitate adaptive divergence. We test these ideas by documenting predator-induced plasticity for resting metabolic rate and growth rate in populations of the Trinidadian guppy (Poecilia reticulata) adapted to high and low predation. We find reduced metabolic rates and growth rates when cues from a predator are present during development, a pattern suggestive of adaptive and non-adaptive plasticity, respectively. When we compared populations recently transplanted from a high-predation environment into four streams lacking predators, we found evidence for rapid adaptive evolution both in metabolism and growth rate. We discuss the implications for predicting how traits will respond to selection, depending on the type of plasticity they exhibit.