Abstract

Two experiments with rat subjects separated learning about the discriminative stimulus versus the operant response in the extinction of discriminated operant learning. Each was designed to separate 2 forms of error that could generate extinction learning from an error-correction perspective: Stimulus error, where the discriminative stimulus overpredicts the reinforcer in extinction, and response error, where the response is higher than what the current reinforcer supports. Stimulus error would cause correction of the Pavlovian stimulus-reinforcer association, whereas response error could cause correction of the instrumental response through adjustment of the response-reinforcer association or direct inhibition of the response. Previous research has supported a role for prediction error in instrumental extinction but has confounded these 2 potential sources of error. Using new variations of the concurrent excitor paradigm (Experiment 1) and the overexpectation paradigm (Experiment 2), the present experiments manipulated response error while controlling stimulus error. Both demonstrated that response error plays a role in instrumental extinction. When a discriminated operant response is not reinforced, response error correction may cause the animal to learn to suppress that specific response.