The evolution of prey warning colouration is, literally, a text-book example of Darwinian adaptive evolution by natural selection. The cornerstone of this evolutionary process is a predation event, the dynamics of which are poorly understood. Aposematic (warningly-coloured) prey are relatively unpalatable and their conspicuous appearance should enable predators to avoid them, but such is not always the case. It has been assumed, based on models of conditioned learning, that the number of aposematic prey that a predator will attack as it learns to avoid such prey should be constant or declining as the prey’s abundance increases. However, empirical studies have instead shown that predators make greater numbers of attacks on aposematic prey when those prey are more common. I show that this failure of theory to predict behaviour likely arises from limitations of the learning models in question. Rather than mechanistic models of conditioned learning, I use signal detection theory to provide a functional characterization of the response uncertainty encountered by inexperienced predators. This characterization explains otherwise puzzling data on aposeme predation and can offer insight on the selective pressures driving the evolution of aposematism and mimicry.
Lynn, S.K. 2005. Learning to avoid aposematic prey. Animal Behaviour 70(5):1221-1226.