To examine the effectiveness of larval antipredator mechanisms and to test the assumption that diet breadth and chemistry are important predictors of predation responses, I offered 70 species of lepidopteran larvae to the predatory ant Paraponera clavata (Hymenoptera: Formicidae). Prey chemistry and diet breadth were significant predictors of rejection, as was plant chemistry but to a lesser degree. Specialist caterpillars were better protected than generalists, and prey with unpalatable extracts were frequently rejected by P. clavata, while prey with palatable extracts were rarely rejected. Specialists with unpalatable host plant extracts were more likely to be rejected than specialists that had host plants with more palatable extracts. Other significant predictors included morphology, behavior, and interactions between developmental stage and prey chemistry and between developmental stage and diet breadth. I concluded that predation could be a substantial selective force in the evolution of narrow diet breadth and that plant chemistry could be the mechanism whereby specialists are better defended.