Cue effects of drive and reward as a function of discrimination difficulty: Evidence against the Yerkes–Dodson law.

Abstract

To investigate the possible cue (stimulus) as opposed to motivational functions of drive and reward, the present study assessed the effects of problem difficulty (easy, moderate, and difficult), reward magnitude (1, 2, and 4 pellets), and drive level (moderate and high) on the performance of rats in visual discrimination tasks entailing both choice and speed measures. Errors and trials to criterion were greater with increasing problem difficulty and smaller reward magnitude, with the effect of reward magnitude being amplified by increasing problem difficulty. High drive produced fewer errors and trials to criterion than moderate drive, with this difference being amplified by both smaller reward magnitude and greater problem difficulty. Choice and arm speeds were positively related to reward magnitude, with high drive producing faster speeds than moderate drive at all reward levels. Collectively, the data support a distinctive-cue interpretation of the effects of reward on discrimination learning and provide evidence contrary to the Yerkes-Dodson law.The present study was prompted by recent findings showing that response-contingent neutral and aversive stimuli (e.g., white noise and electric shock) will facilitate the learning of a difficult, but not an easy, discrimination when administered for either a food-rewarded or nonrewarded response (e.g., Fago & Fowler, 1972; Fowler, 1971b;Fowler & Wischner, 1969). The fact that the facilitation is observed only in relatively difficult discriminations, where the discriminative stimuli and/or other characteristics of the alternatives (e.g., physical features, response feedback) are highly similar, argues that the effect is not due simply to the acquired reinforcing or nonreinforcing properties of the events, since such properties

Transfer from Pavlovian appetitive to instrumental appetitive conditioning: Signaling versus discrepancy interpretations.

“…In reverse fashion, a four-pellet ApCS-should retard discrimination learning because, as a signal for the absence of a large magnitude of food (i.e., as a strong conditioned frustrative stimulus) occurring in the correct alternative, it can promote equivalence with the incorrect alternative where no reinforcement (and hence frustration) occurs, and thus it can increase between-alternative generalization effects. These discriminability effects are especially likely for strong ApCSs because they have been shown independently to operate for the primary reinforcer itself: Larger magnitudes of food reinforcement will facilitate learning in a difficult discrimination merely on the basis of their distinctive stimulus features and associated consummatory feedback (•Hochhauser & Fowler, 1975). Thus, it is not unlikely that the expectancies generated by strong ApCSs will have a similar function.…”

“…Second, because the size of the Pavlovian reinforcer for a subject would later be matched in discrimination training, the upper limit of reinforcement magnitude was restricted to that value for which one could be reasonably confident of the absence of a floor effect, that is, a physical limit, in the rate of discrimination learning [cf. Hochhauser & Fowler, 1975].) Of the four Pavlovian training conditions, three were designed to establish an ApCS+, CSo, or CS-; the fourth entailed only US presentations (US alone, i.e., USa) and was designed to provide a novel-CS control.…”

“…If discriminability effects of the fourpellet ApCSs confounded the evaluation of signaling and/or discrepancy interpretations of transfer in the prior difficult discrimination, then the two interpretations might be better evaluated in the context of an easy discrimination, even though reinforcement magnitude effects are constrained in such a problem (e.g., Hochhauser & Fowler, 1975;Reynolds, 1949). The rationale for this judgment stems from prior research (e.g., Fowler & Wischner, 1969;Hochhauser & Fowler, 1975) showing that the distinctive cue effects of response-contingent stimuli, such as food and shock, are virtually eliminated in an easy visual discrimination where generalization (and hence discriminability) effects are largely precluded by high dissimilar discriminative stimuli. On the other hand, if the four-pellet findings of the prior difficult discrimination reflect conditioned reinforcing and/or higher-order conditioning effects of the ApCSs, they should not be restricted to a difficult discrimination but should reappear in an easy discrimination.…”