Summary. Field observations indicated that hornworms select feeding sites non-randomly on tobacco. We tested the hypotheses that differences in feeding site locations of larvae of Manducasexta L. and Manducaquinquemaculata (Haworth) (Lepidoptera: Sphingidae) on tobacco could be explained by differential nicotine concentrations within plants and leaves, species-specific responses to nicotine, or pressure exerted by natural enemies. Results showed that third-instar larvae of M. sexta fed more proximally and centrally on the leaf, whereas M. quinquemaculata fed more distally. Within-plant selection of leaves did not differ; both species selected leaves in the middle region of the plant. Nicotine concentrations in a high nicotine genotype, NC95, varied within each leaf, increasing 23 fold from the basal to apical portion of the leaf, and within each plant, increasing 710 fold from the first fully expanded leaf to the twelfth (lowest) leaf. In laboratory bioassays, both Manduca species responded to nicotine as a feeding deterrent. Electrophysiological studies demonstrated that gustatory organs of both species responded to nicotine at concentrations found in tobacco leaves and that M. quinquemaculata generally showed a less vigorous response to nicotine than M. sexta. Field mortality of M. sexta due to parasitism by Cotesiacongregata (Say) and to parasitism and predation combined differed among feeding sites; predation alone did not. Results suggest that although nicotine concentration and species specific responses to nicotine play a role in determining feeding site locations, pressure exerted by natural enemies, especially parasitism by C. congregata, is more important.