Almost forty years ago, Ehrlich and Raven (1964) hypothesized that the great diversity of plants and the herbivores that feed on them arose from a process of coevolution. Plants do possess an amazing diversity of traits that are easily imagined as having arisen from an antagonistic interaction between plants and herbivores. Two basic assumptions lie at the root of most theories of coevolution between plants and their herbivores. First, herbivores are agents of natural selection on plant resistance traits. Second, plants incur a significant fitness cost for possessing these resistance traits. An ecological genetic approach can provide rigorous evidence for these coevolutionary assumptions. In this paper, I present new experimental work on the subject of costs of resistance and review and discuss my own previous work bearing directly on these questions. Using both field experiments on natural populations of the mouse-ear cress (Arabidopsis thaliana) and laboratory experiments using genetically modified plants, I demonstrate that herbivores are exerting selection on both a chemical and physical resistance trait and that there are significant fitness costs to possessing these two traits. These results provide direct confirmation that our current models of the evolution of plant defenses are appropriate.