Partial reward, the negative contrast effect, and incentive averaging

Abstract

A total of 169 rats distributed across six experiments, received training in a straight runwa Y d with a 5-min intertrial interval. A variety of shifts in reward schedules, to~nd from partial rewar , were employed to assess the effects of partial reward on the successi~e ne~atIve contrast effect. The results were seen as supportive of an incentive averaging approach to partial remforcement.

Repeated successive contrast in consummatory behavior with repeated shifts in sucrose concentration

“…The reinforcement-level theory (Capaldi, 1974) states that contrast is due to a discrepancy from an average, or expected, level of reinforcement. There are substantial data which show that animals average different levels of reward received in the past and that degree of contrast is a function of degree of discrepancy from this average level of reward (Flaherty, Becker, & Osborne, in press;McHose & Peters, 1975). The failure of negative contrast to diminish substantially with repeated shifts in any of these experiments indicates that the expected reward level is quickly "reset" to the maximum value when animals are returned to the preshift reward.…”

An Adaptive Response to Uncertainty Generates Positive and Negative Contrast Effects

“…Successive contrast effects pose a serious challenge to accounts of rational choice in which past states or alternative current states are irrelevant (6). So far, all theories to explain the existence of contrast effects have been based on descriptive models of psychological or physiological processes (4,7,(15)(16)(17)(18), largely ignoring their evolutionary basis [though see (10)]. Here we address this gap, by developing a simple and general optimality model to show that both positive and negative successive contrast effects can arise from an adaptive response to uncertainty about how conditions change over time.…”

Expectancy, salience, and habit: A noncontextual interpretation of the effects of changes in the conditions of reinforcement on simple instrumental responses.

“…Previous studies (McHose & Peters, 1975;Mikulka et al, 1967) have clearly shown that partial reward in the preshift period attenuates the negative contrast effect. Since partial as compared to continuous reward should affect the intensity of K through the laws of classical conditioning, it follows that subjects trained on large partial reward and subsequently shifted to small continuous reward should display less contrast than subjects shifted from large to small continuous reward.…”

“…For present purposes, this simply means that HA and HK differences between groups trained on partial as compared to continuous reward are no more nor less complicated than such differences for groups trained on different amounts of continuous reward (cf. McHose & Peters, 1975, for a detailed account of the various effects of partial and varied reward on negative contrast that are described by this approach). That the current model also provides a nice fit to the data bearing upon the effects of partial reward on positive contrast may be seen by reference to a study by Lehr (1974), in which positive contrast was obtained following a shift from small to large partial reward (as compared to a large partial-reward control condition).…”

“…It must be emphasized that such a paradigm is satisfactory only to the extent that the M condition equals the incentive average of L and S values used, a condition that can be approximated (cf. McHose & Peters, 1975) using continuous reward in all conditions or that can be accomplished by, for example, administering L -* L/S, S -> L/S, and L/S conditions, where L/S represents the appropriate 50% varied reward-amount condition. We find no data that match this paradigm.…”