A review of recent evidence indicates that contingent reinforcement is neither a necessary nor a sufficient condition for operant learning. This dilemma reopens the old question of "what is learned." It is proposed that what laboratory Ss characteristically learn is not a response to a stimulus, but rather 2 kinds of expectancies. 1 kind of expectancy corresponds rather accurately with environmental stimulus-outcome contingencies, while the 2nd is a less faithful representation of response-outcome contingencies. The reinforcement procedure merely permits both expectancies to be learned. (36 ref.) (PsycINFO Database Record (c) 2016 APA, all rights reserved)