No abstract

Prior work has shown that when the separate correct responses of a conditional discrimination are followed by different reinforcing outcomes, performance is enhanced relative to that obtained under the conventional, single-reinforcer procedure. Four experiments with pigeons yielded the analogous finding when the different outcomes were reinforcement and explicit nonreinforcement. Controls indicated that the results could not be attributed to the effects of intermittent reinforcement, to possible differences in cue duration, or to a variety of potential sources of conditioned reinforcement. An interpretation in terms of expectancy learning is proposed. Trapold (1970) originally demonstrated what we will refer to in this paper as the differential outcomes effect (DOE). He found that rats learn a conditional discrimination (i.e., if stimulus S.. then response R 1 is correct, but if S2, then R 2 is correct) faster when correct responses to S, are consistently followed by reinforcing outcome 0 1 (e.g., sucrose solution) and correct responses to S2 are consistently followed by a different outcome O 2 (e.g., food pellets) than if correct responses to S, and S2 are both reinforced by the same outcome. Trapold hypothesized that the DOE occurs because (1) when S, and S2 are each followed by a unique outcome event, subjects learn an expectancy of 0 1 to S, and an expectancy of O 2 to S2' and (2) these different expectancies provide an additional source of differential stimulation that can assume stimulus control over the choice behavior. In contrast, subjects which receive the same outcome for correct responding to S, and S2 learn the same expectancy to Sl and S2' and hence do not have available this additional source of potential differential stimulus control over R1 and R2.Trapold provided additional support for this theo- retical analysis by showing that subjects who are pretrained so as to preestablish the expectancies involved in a subsequent differential-outcomes conditional discrimination problem learn that problem faster than subjects pretrained to expect the wrong outcome. Wielkiewicz (1972, 1976) extended Trapold's findings by showing a DOE for food reward outcomes of different delays and different magnitudes in rats. Overmier, Bull, and Trapold (1971) demonstrated a DOE in dogs when the differential outcomes were avoidance of electric shocks to different loci on the body. Brodigan and Peterson (1976) and Peterson, Wheeler, and Armstrong (1978) found a DOE with pigeons using food and water as the different outcomes. These latter studies also provided considerable additional support for the assumption that the DOE is mediated by differential outcome expectancies which function as part of the discriminative stimulus complex to which the correct response becomes learned. Specifically, they showed that the DOE is much more profound when a delay is inserted between the offset of the conditional cue (S, or S2) and the opportunity to choose between R 1 and R 2. With a standard procedure, insertion of such a delay of ev...