Coevolution, the reciprocal evolutionary change in interacting species driven by natural selection, has an enormous importance on ecological and evolutionary theory. However, there are very few empirical studies demonstrating coevolution. In particular for plant-herbivore interactions, even though there are many adaptations that seem to be the result of the interaction, just a few empirical evidences on the processes leading to these adaptations exist. I used a highly integrative approach to study coevolution in an ideal plant-herbivore system: the alkaloid bearing legume Crotalaria pallida and its seed predator, the arctiid moth Utetheisa ornatrix. To unravel coevolutionary dynamics and to demonstrate reciprocal coevolutionary selection a through story of the specific interaction is necessary. In my dissertation I started to develop a through story of the coevolution between U. ornatrix and C. pallida. First, I investigated how the history of a hostiiiintroduction may affect coevolutionary dynamics by showing that the native herbivore U. ornatrix has much higher fitness on the introduced host C. pallida than on a native congeneric C. incana. Second, I used a geographic approach to investigate patterns of local adaptation. I showed local adaptation of the moth to its host plant populations at a continental scale (populations from Brazil and Florida), but not at a local scale (populations ca. 150 Km apart). Third, I investigated how genetic differentiation among plant and moth populations may affect the patterns of local adaptation. I found genetic differences among populations in plant defense traits and moth differentiation on neutral microsatellite loci even at the local scale. Finally, because a thoroughly understanding of coevolutionary dynamics depends on a careful examination of the adaptations related to the interaction for both interacting species, I studied the counteradaptation of the moth to the host-plant defensive alkaloids. I combined chemical ecology techniques with extensive laboratory experiments to show that sequestration of alkaloids has no fitness costs. I discuss how these unexpected and novel results have many important implications to our understanding of coevolution.