In many systems, the number of prey killed by predators increases with prey density. This in turn generates higher levels of the indirect signals that prey use to assess predation risk. A model developed by Peacor (2003) showed that prey that respond to predator cues without accounting for conspecific density will consistently over- or under-estimate risk and therefore invest improperly in anti-predator defense. We tested this model using Rana temporaria tadpoles as prey and Aeshna cyanea dragonfly larvae as predators. As assumed by the model, prey reduced risky activity with increasing concentrations of predator kairomones and increased activity at high prey density. However, prey did not react to changes in cue or density if the ratio of cue-to-density remained constant. Prey therefore monitored their per capita risk, strongly supporting Peacor's model.