Previous explanations for the evolution of induced resistance of plants to herbivory emphasized arguments based on saving costs when allocations to defense were not needed; these models met with limited empirical support. We offer a novel explanation based on induced resistance providing increased variability in defense. As long as maximal levels of defense are constrained, variability will increase the effectiveness of a given level of investment in defense. We show that variability can decrease herbivore performance if herbivore performance is a concave function of the level of resistance. In particular, if herbivores can choose among different plants and plant tissues, then variability created by induced resistance may benefit plants under attack and hence may be favored by selection. The key assumptions of this model are broadly supported by empirical data from many plant–herbivore systems.