We use a common framework to compare three models of plant strategies to confront herbivory: constitutive defense, optimal inducible defense, and the ''moving target.'' Plants with constitutive defenses retain a fixed defensive phenotype. Plants with optimal inducible defenses respond to attack by increasing defenses. Plants following the moving target strategy respond to attack by altering phenotype. The constitutive and optimal inducible defense models, unlike the moving target model, require that plant phenotypes can be arrayed along an axis representing the trade-off between cost and defense. We argue that the evidence for such an axis is not yet convincing. Our models indicate that constitutive defenses are favored when herbivory is relatively constant or when multiple herbivores attack and costs of defense or defensive machinery are high. Optimal inducible defenses are favored when herbivory rates vary, costs are not too high, and plant phenotypes can be arrayed along a defense axis. The moving target strategy is favored when costs are not too high, herbivory rates vary, and plant phenotypes effective against one herbivore are ineffective against others. We conclude that the moving target strategy might be a robust response to unpredictable and uninformative environments.