Both short-term and long-term variation in predation risk can affect the behaviour of prey, thus affecting growth, reproduction, survival and population dynamics. Inferences about the strength of such 'risk effects' in the wild have been limited by a lack of studies that relate antipredator responses to the magnitude of direct predation, measure responses of prey to risk from complete predator guilds, and quantify risk in more than one way. Here, we quantify behavioural responses of a complete ungulate prey guild to long-term and short-term variation in risks from all of the large predators in Liuwa Plain National Park, with known patterns of direct predation. Our analysis allows the first direct test for interaction between responses to long-term and short-term risk in the wild, and reveals that prey vigilance responds strongly to locations with high long-term risk when short-term risk is high, but not when short-term risk is low. This result has broad ramifications for the design and interpretation of field studies of antipredator behaviour, its costs and its consequences for population dynamics.