There is a critical need to understand patterns and causes of intraspecific variation in physiological performance in order to predict the distribution and dynamics of wild populations under natural- and human-induced environmental change. However, the usual explanation for trait differences, local adaptation, fails to account for the small-scale phenotypic and genetic divergence observed in fishes and other species with dispersive early life stages. We tested the hypothesis that local-scale variation in the strength of selective mortality in early life mediates the trait composition in later life stages. Through in situ experiments, we manipulated exposure to predators in the coral reef damselfish Dascyllus aruanus and examined consequences for subsequent growth performance under common garden conditions. Groups of 20 recently-settled D. aruanus (mean ± SD standard length = 8.53 ± 0.450 mm) were outplanted to experimental coral colonies in Moorea lagoon and either exposed to natural predation mortality (52% mortality in three days) or protected from predators with cages for three days. Predator-exposed groups were shorter, while those with lower survival were in better condition, suggesting that predators removed the larger, thinner individuals. Growth of both treatment groups was subsequently compared under common conditions. We did not detect consequences of predator exposure for subsequent growth performance: growth over the following 37 days was not affected by the prior predator treatment or survival. Genotyping at 10 microsatellite loci did indicate, however, that predator exposure significantly influenced the genetic composition of groups. We conclude that post-settlement mortality did not have carry-over effects on the subsequent growth performance of cohorts in this instance, despite evidence for directional selection during the initial mortality phase.