Abstract
Inducible defences against predators evolve because they reduce the rate of direct predation, but this benefit is offset by the cost (if any) of defence. If antipredator responses carry costs, the effect of predators on their prey is partitioned into two components, direct killing and risk effects. There is considerable uncertainty about the strength of risk effects, the factors that affect their strength, and the mechanisms that underlie them. In some cases, antipredator responses are associated with a glucocorticoid stress response, and in other cases they are associated with trade‐offs between food and safety, but there is no general theory to explain this variation. Here, I develop the control of risk ( COR ) hypothesis, predicting that proactive responses to predictable and controllable aspects of risk will generally have food‐mediated costs, while reactive responses to unpredictable or uncontrollable aspects of predation risk will generally have stress‐mediated costs. The hypothesis is grounded in laboratory studies of neuroendocrine stressors and field studies of food‐safety trade‐offs. Strong tests of the COR hypothesis will require more studies of responses to natural variation in predation risk and the physiological consequences of these responses, but its explanatory power can be illustrated with existing case studies.