Abstract. Vector-borne parasites are commonly predicted to be less virulent to the vector than to the definitive host as the parasite gains little by harming its main route of transmission. Here we assess the empirical evidence from systems in which insects are vectors for vertebrate, plant, and invertebrate parasites. The body of evidence supports lower (but nonzero) parasite virulence to vectors than to plant or invertebrate hosts, but not to vertebrate hosts. We consider why this might be by assessing evolutionarily stable strategies for an insect parasite that can infect both predator and prey (or vector and definitive host) and can have distinct virulences in these two potential hosts. In a homogeneous environment, the parasite is predicted to be equally virulent to predator and prey. However, in a patchy environment it is predicted to become benign toward the more mobile of the two potential hosts, provided interpatch movement of free parasites is low and competitive displacement among strains in a patch is weak. This prediction meets reality in that the vector is usually more mobile between patches than is the definitive host in plant and invertebrate systems, but not necessarily in vertebrate hosts.