I argue that nonequilibrium allele frequency dynamics due to coevolution can drive the evolution of specialized host races in parasites capable of host choice-for example, herbivorous insects or parasitoids. The proposed mechanism does not require genetic trade-offs in performance on different host species. It is based on the premise that the ability of the parasite to overcome the resistance of different host species is to a large degree genetically independent-that is, controlled by different loci. The intuitive rationale is that the genetic lineage of a parasite that evolves host preference becomes more consistently exposed to selection for performance on its preferred host. Such a choosy lineage can thus coevolve faster in response to evolving host defenses than a generalist lineage distributed among several host species. Given genetic variation in host preference, an initially generalist parasite population evolves toward specialized host races, each choosing one host species. This idea is supported by a series of multilocus models of coevolution between a parasite and two host species, in which the parasite virulence on each host is affected by a different set of loci and an additional locus or two loci control host choice.