Patterns of Errors in the Reversal Shift and the Overlearning-Reversal Effect

Abstract

In two experiments adults ( ns = 90, and 40) learned a two-choke discrimination to eight stimuli. After either initial acquisition training or overtraining a reversal shift was presented. In a subsequent phase post-reversal-shift subjects were instructed to respond, without feedback, to the stimuli in terms of either the acquisition phase or reversal-shift phase patterns of responding. In Exp. I four stimuli, presented without feedback, were interspersed during acquisition and the reversal shift to differentiate between subjects using a conceptual mode of solution based only on values of the relevant dimension and a nonconceptual, rote learning, solution mode. The overlearning-reversal effect and associated error pattern were found only among nonconceptual subjects. Post-reversal-shift test performance indicated that overtraining differentially affected response strategies used in the reversal shift as a function of the mode of solution subjects attained. In Exp. II stimuli were arranged so that a conceptual solution was not possible. The data replicated Exp. I. A response-mediation interpretation of the effects of overtraining on reversal-shift behavior and reversal-shift error pattern is proposed.