Reward pretraining and discrimination learning set

Abstract

Pigtailed monkeys were trained on discrimination problems involving objects that had been previously paired with reward, previously paired with nonreward, or objects not associated with any pretraining. In two separate studies, no differences on these various problem formats were observed, but performance improved considerably with practice. These results are contrary to the theories for learning set formation of Spence and Medin.When monkeys are given a series of object discriminations, their performance on new problems gets better and better. Early in the series, they make many errors but, after extensive practice, learning takes place after just a few errors. A major theoretical problem is to specify the source of improvement in discrimination performance.The present study tests one implication of some theories for learning set formation. The basic procedure involves pairing objects with reward or nonreward during pretraining, followed by discrimination training where two previously rewarded objects are paired and two previously nonrewarded objects are paired. The variable of main interest is the relative rate of learning these two kinds of problems. First, it is in order to review the theories making a clear prediction concerning this paradigm.Although it is certainly not a well-publicized fact, Spence's (1936) original theory of discrimination learning predicts that learning to learn will occur. The most extensive discussion of the application of Spence's theory to learning set is given by Reese (1964). Space does not permit a detailed review of Spence's theory, but it is important to bring out the facets of the theory which underlie improvement on object discriminations. These main factors are: (l) the initial strengths associated with the two objects of a discrimination problem, (2) the difference in these initial strengths, which provides a basis for stimulus preferences, and (3) the difference in strengths of irrelevant cues which must be overcome before a probelm is solved. We shall consider these in turn.Initial strengths alter performance through their influence on the effectiveness of reward and nonreward. The effectiveness of a reward first increases with strength to a maximum and then, for high strengths, decreases once more. The change in strength produced by a nonreward increases linearly with strength. Overall then, changes in strength are greatest when strengths are *This research was supported by United States Public Health Service Grants MH25134 and MH16100.in the middle to high range. When the preferred stimulus is incorrect, stimulus preferences can serve to retard learning. As a consequence, decreases in initial stimulus preferences facilitate learning. Spence's original paper illustrates the influence of irrelevant cues, such as spatial or position habits, on learning. Decreases in these strength differences are also associated with faster learning.To assess predictions of Spence's theory concerning the effects of pretraining with rewards, computer simulations of these pretraining condit...

Preference effects on acquisition and retention of concurrent discriminations by rhesus monkeys

“…In overview, the results from our series of tests are in accord with reviews (Bessemer & Stollnitz, 1971;Medin & Davis, 1974) that have characterized monkey memory as proficient. In the concurrent situation, this retention proficiency does not seem sternly related to the choice behaviors that a monkey has displayed on its first exposure to discriminanda.…”

“…Medin (1977) reviewed evidence on the relationship between retention and the extent of prior discrimination training. Although there was little generality of results, he did note instances in which some retention measures were better after learning a relatively small number of problems than they were after the animals had undergone extensive training.…”

Bilateral Orbital Prefrontal Cortex Lesions in Rhesus Monkeys Disrupt Choices Guided by Both Reward Value and Reward Contingency

“…It is well established that the presence or absence of food carries informational value independent of its hedonic value (Medin, 1977;Gaffan, 1979). Although a change from the presence to the absence of food reward, as occurs in object reversal learning, could be construed as an extreme devaluation of the reinforcer, we contend that it is a manipulation in which the information about reward contingencies is conveyed by the visual processing system (by the presence or absence of food reward) rather than the affective processing system.…”

Conditional Discrimination in Mentally Retarded Subjects: Programming Acquisition and Learning Set

“…The following account makes a parallel assumption for conditional discrimination. In addition, it incorporates elements of earlier accounts of simple discrimination learning set, such as Harlow's (1959) error factor theory and the "win-stay, lose-shift" of hypothesis theory (for reviews, see Medin, 1977;Schrier, 1984).…”

“…These errors usually produce a rapid change in comparison selection, an outcome presumably made more probable by prior comparison discrimination reversal training. Information-processing and hypothesis-theory accounts of simple discrimination learning set (see Medin, 1977;Schrier, 1984) Riopelle & Copelan, 1954).…”