Question: How does cannibalism in the host alter the evolution of a parasitoid’s oviposition strategy? Can differences in cannibalism risk between parasitized and healthy hosts alter the stage-specific foraging of parasitoids? Can host-specific differences in cannibalistic behaviour explain why parasitoids vary in what host stages they attack? Mathematical methods: We examined the evolutionary dynamics of a stage-structured host–parasitoid model using two complementary approaches: (1) individual-based numerical simulations of evolutionary dynamics, and (2) the theory of adaptive dynamics focusing on evolutionarily stable strategies (ESSs). Assumptions: Cannibalism in the host is assumed to be stage structured, with larger stages consuming smaller stages. The consumption of parasitized hosts also results in killing of the parasitoid’s offspring. Vulnerability to cannibalism of parasitized versus healthy hosts was allowed to vary. The parasitoid’s preference for attacking early versus late host stages was the trait under selection and allowed to evolve. Results: When cannibalism rates increase relative to the parasitoid’s attack rates, the ESS of the parasitoid shifts from attacking only early host stages to attacking only late host stages. This shift occurs at lower cannibalism rates when parasitized hosts are more susceptible to cannibalism than healthy hosts. Under equilibrium conditions, a small boundary area exists between these two regions where attacking only early or only late host stages are alternative ESSs. The threshold and alternative stable ESSs are the result of cannibalism, which creates a positive feedback between the parasitoid’s oviposition rate and its own mortality. Intermediate strategies, where parasitoids evolve to attack both stages, occur only when host populations exhibit large population oscillations or when generalist parasitoids that attack both stages have a foraging advantage