Michael Kalish Department of Cognitive Science. University of California San Diego. La Jolla, CA. U.S.A.
The primary function of perceptual learning is to alert an animal to new and changing affordances in its niche. Distinguishing a situation's affordances is sometimes a question of detecting a deterministic specific change in the ambient array; however, other times objects lie about their affordances, making the detectable changes in the array only probabilistically related to the affordance. Since this is the case, animals must be flexible enough to learn in the presence of uncertainty, as when jacamars learn to distinguish palatable from unpalatable butterflies (Marden & Chai, 1991). Faced with a probabilistic discrimination between the two choices, the birds become sensitive to a higher-order array variable which is maximally informative of the butterfly's palatability. Many decisions about environmental events call for categorical judgements, but many other cases require continuous judgements, for example judgements about age (Pittenger, Shaw, & Mark, 1979). Learning to make such decisions under cue ecologies of varying informativeness is a fundamental and general task for all but the simplest animals.