Two potential defences against brood parasitism by the cuckoo Cuculus canorus were compared experimentally between British populations of reed warblers Acrocephalus scirpaceus that are parasitised at different rates. (1) Rates of rejection of model cuckoo eggs were lower at two unparasitised populations which did not have resident cuckoos, than at a rarely parasitised population which had cuckoos nearby, and at a regularly parasitised population. (2) Reed warblers from an unparasitised population showed a slightly weaker response to taxidermic mounts of cuckoos and, unlike a parasitised population, did not differentiate between mounts of a cuckoo, sparrowhawk Accipiter nisus and jay Garrulus glandarius. Differences in exposure to real predators may explain the differences in responses to mounted predators between populations, as specific aggressive responses to predators are likely to have been learned. Although evidence from dispersal and population turnover data suggests that there is likely to be gene flow between reed warbler populations in Britain, the hypothesis that the population differences reflect genotypic differences could not be ruled out. An alternative explanation of phenotypic plasticity in defences could also explain the population differences. Phenotypic plasticity in defences would be favoured in environments where the risk of parasitism fluctuates, if those defences are costly to unparasitised reed warblers.