The central focus of this article is a derivation of stochastic utility functions from the basic postulates of stimulus sampling learning theory. Generalizations of the results to multiple choice situations are discussed, as well as relations to the entropy of a set of alternatives and to Luce's choice axiom.