ABSTRACT
In the last chapter we saw that supernormal effects are widespread in nature and that many animal signals exploit a preference for extremes. In this chapter I argue that the power of exaggeration may reflect an intrinsic feature of recognition systems. Whenever stimuli or categories must be distinguished, exemplars that exaggerate the differences between those categories often elicit enhanced responses compared with the learned exemplars. I turn now to the evidence that such a preference for extremes is widespread, occurring in humans, other animals, and even computer recognition systems.
