On exposure to ultraviolet radiation (UV), many plant species both reduce stem elongation and increase production of phenolic compounds that absorb in the UV region of the spectrum. To demonstrate that such developmental plasticity to UV is adaptive, it is necessary to show that the induced phenotype is both beneficial in inductive environments and maladaptive in non-inductive environments. We measured selection on stem elongation and phenolic content of seedlings of Impatiens capensis transplanted into ambient-UV and UV-removal treatments. We extended the range of phenotypes expressed, and thus the opportunity for selection in each UV treatment, by pretreating seedlings with either a low ratio of red:far-red wavelengths (R:FR), which induced stem elongation and reduced phenolic concentrations, or high R:FR, which had the opposite effect on these two phenotypic traits. Reduced stem length relative to biomass was advantageous for elongated plants under ambient UV, whereas increased elongation was favored in the UV-removal treatment. Selection favored an increase in the level of phenolics induced by UV in the ambient-UV treatment, but a decrease in phenolics in the absence of UV. These results are consistent with the hypotheses that reduced elongation and increased phenolic concentrations serve a UV-protective function and provide the first explicit demonstration in a wild species that plasticity of these traits to UV is adaptive. The observed cost to phenolics in the absence of UV may explain why many species plastically upregulate phenolic production when exposed to UV, rather than evolve constitutively high levels of these compounds. Finally, pretreatment with low R:FR simulating foliar shade did not exacerbate the fitness impact of UV exposure when plants had several weeks to acclimate to UV. This observation suggests that the evolution of adaptive shade avoidance responses to low R:FR in crowded stands will not be constrained by increased sensitivity to UV in elongated plants when they overtop their neighbors.