different outcomes of equal incentive value are employed.
Except for the use of the cognitive-sounding term expectancy, there is, in fact, very little that is necessarily cognitive about the model presented in this paper. Indeed, the formulation depicted in Figure 8.1 is pure S-R associationism, and is essentially identical to the mediational theories of 30 or more years ago (e.g., Hull, 1930; Hull, 1931; Osgood, 1953). Whether the expectancy is regarded as cognitive or noncognitive depends primarily upon where one believes it resides, i.e., whether one believes it is a central or peripheral event. This is, of course, a classic issue in animal learning, and one on which there has been little progress in the past half century. Peripheral responses identified as likely candidates for the position of expectancy have seldom been found to behave the way mediation theory says they should (cf. Rescorla & Solomon, 1967; Trapold & Overmier, 1972). Moreover, even if a situation were found in which a certain peripheral response correlated perfectly with the predictions of the theory, the possibility would remain that the response was merely the surface manifestation of some more primary central event. It should be noted, however, that the expectancy model does not require that the location of the expectancy be known. The only requirement is that the response to a stimulus is found to be predictable from a knowledge of 1) the particular outcome with which that stimulus has been associated and 2) the response the subject has most recently learned to make in the presence of stimuli associated with that outcome. This has been amply demonstrated to be the case. Thus, the abstract concept of the expectancy is useful independently of whether the "real" expectancy is ever identified.
As noted above, expectancy mediation of behavior is not a new idea. The Differential Outcomes procedure, however, is relatively new. Why did the development of a laboratory technique for studying expectancy mediation lag so far behind its conceptual development? In the main, I think this is attributable to a particular inhibitory effect which traditional reinforcement theory, especially the orthodox operant paradigm, has had on laboratory procedural variation. The idea of differentiation of behavior via differential consequences is, of course, the fundamental concept of operant conditioning. However, according to the traditional doctrine, the mechanism by which differential consequences control behavior is quite different from the expectancy mechanism. The emphasis in discussions of response differentiation within the operant paradigm has always been on differential reinforcement in the sense of reinforcing one class of behavior and not reinforcing another. No mention is made of the role of anticipations or expectancies of reinforcement or nonreinforcement in this process. Moreover, as far as the process of reinforcement is concerned, an implicit assumption of the traditional account seems to be that, to paraphrase Gertrude Stein, a reinforcer is a reinforcer is a reinforcer. On this view, qualitatively different reinforcing substances may differ quantitatively in their effectiveness as the reinforcer, but otherwise one