Many plants employ induced defenses that are activated after attack rather than constitutive defenses that are always present. Current theory predicts that one benefit of induced defenses is to optimize the plant's allocation to defense, growth, and reproduction. According to this reasoning induced defenses are expected under the following conditions: (1) Attack is unpredictable and not fatal. (2) Defense is costly; when herbivores are not present, the plant can save resources by not being defended and use them for growth and reproduction. (3) When herbivores are present, activation of the defense increases plant growth and reproduction. This study tested the allocation hypothesis by examining the effects of early—season damage to a desert shrub, Gossypium thurberi, in environments that either included or did not include its most abundant herbivore, Bucculatrix thurberiella. A companion paper demonstrated that induction occurred. Plants damaged early in the season supported fewer miners than controls that were not previously induced. If there was a benefit of induced resistance, I expected that survival, growth, and seed production would be greater for plants that had been induced experimentally by caterpillars early in the season compared to those that had not. I found no evidence that plants benefited by responding to damage early in the season, despite reductions in the populations of herbivores on damaged plants. If there were costs of induced resistance, I predicted that survival, growth, and seed production would be lower for induced plants than for controls when herbivores were excluded from both treatments. I found no evidence that induced resistance cost the plant anything in terms of survival, growth, or seed production. Finally, to determine if plant crowding was costly and if crowding interacted with induced resistance, survival, growth, and seed production were compared for plants that had no neighbors within 50 cm vs. those experiencing endemic densities of neighbors. Plants that experienced reduced density grew more and produced twice as many seeds as controls that experienced endemic densities of neighbors. No interactions between damage and plant density were observed. There are many different ways to interpret these negative results. I measured only a subset of the plant's lifetime fitness and the statistical power of my tests may have been insufficient. My experiment to test the benefits of induced resistance was much weaker than my tests of the costs of damage and of crowding. Herbivory may be unimportant in this system, although indirect evidence suggests otherwise. This and other empirical results suggest that the cost—benefit model may not be appropriate to explain the advantages of induced resistance over constitutive resistance. I discuss three alternative hypotheses to explain the existence of induced resistance.