Omnivorous natural enemies of herbivores consume plant-based resources and may elicit induced resistance in their host plant. A greater induction threshold for damage produced by omnivorous predators than for strict herbivores might be expected if omnivore performance is enhanced on noninduced plants, allowing them to reduce future levels of herbivory. Currently, it is not known if a plant responds to feeding by omnivorous predators and by herbivores similarly. To examine this question, we chose herbivore and omnivore species that produce the same kind of quantifiable damage to cotton leaves, enabling us to control statistically for the intensity of plant damage, and ask whether plant responses differed depending on the identity of the damaging species. We first compared changes in plant peroxidase activity, gossypol gland number and density, and leaf area in response to feeding by the spider mite Tetranychus turkestani (Ugarov and Nikolski) (an herbivore) and by one of the mite's principal natural enemies, the western flower thrips Frankliniella occidentalis (Pergande) (an omnivore). Both species increased the activity of peroxidase, but when we controlled for the amount of damage, the peroxidase activity of mite-damaged plants was higher than that of thrips-damaged plants. We also found that thrips, but not spider mites, increased the density of gossypol glands in the second true leaf. In a second experiment we included an additional herbivore, the bean thrips Caliothrips fasciatus (Pergande), to see if the different responses of cotton to thrips and mite herbivory we first observed were attributable to differences in trophic function (herbivore versus omnivore) or to other differences in feeding generated by thrips versus mites. Cotton plants exhibited the same pattern of induced responses (elevated peroxidase, increased number of glands, reduced leaf area) to herbivory generated by the bean thrips (an herbivore) and western flower thrips (an omnivore), suggesting that trophic function was not a key determinant of plant response. Thrips-damaged plants again showed a significantly higher density of gossypol glands than did mite-damaged plants. Overall, our results suggest that (1) an omnivorous predator systemically induces resistance traits in cotton and (2) whereas there is evidence of taxonomic specificity (thrips versus mites), there is little support for trophic specificity (herbivorous thrips versus omnivorous thrips) in the elicitation of induced responses.