Interactions between insect herbivores and host plants can involve herbivorehost pairs that are evolutionarily ancient or only recently associated. Novel herbivore-host species pairs are continually being formed via host shifts, dispersal, and increasingly via anthropogenic introductions. Conceptual models of enemy-victim coevolution (specifically, the evolution of plant tolerance and of insect virulence) suggest that the impact of an herbivore on its novel host should, at least at first, be more intense than its impact on its ancestral host. We tested this hypothesis for the specialist gallmaking caterpillar Gnorimoschema gallaesolidaginis (Lepidoptera: Gelechiidae) on its ancestral and novel hosts, Solidago altissima and S. gigantea. We measured aboveground ramet mass for paired attacked and unattacked ramets of each species at two sites (Fredericton, NB, and Toronto, ON, Canada), and also measured allocation of tissue mass to stems, leaves, and flowers in galled and ungalled ramets. G. gallaesolidaginis attack reduced ramet growth considerably more on S. gigantea, the novel host, consistent with the coevolutionary hypothesis. We were unable to detect reallocation of tissues in galled ramets as a mechanism for tolerance, and found no intraspecific difference in the impact of gallmaking on allocation patterns. Herbivore host shifts between alternative native hosts will provide an excellent opportunity to understand the evolutionary history of novel herbivore-host associations, particularly (as in the Solidago system) when multiple insect herbivores have host-shifted across the same plant pair.