Within-Subjects Procedural Variations in Two-Choice Probability Learning

Abstract

Rats were given extended training (more than 1000 trials) in a two-choice probabilistic food reinforcement situation with three different procedures: (a) noncorrection, (b) correction, (c) a forced-choice procedure that ensured rewarded experience with the less frequently rewarded alternative. The noncorrection and forced-choice procedures resulted in near maximizing (the more frequently rewarded alternative chosen on most of the trials), while the correction procedure resulted in near matching (response probabilities approximate reinforcement probabilities). There were large individual differences in response to the variations in procedures. A second experiment with water reinforcement and correction procedures showed similar results.