The present study suggests that relapse can be defined in behavioural terms; relapse is when a subject reverts back to a previously-reinforced behaviour following the disruption of a recently-reinforced behaviour. It is argued that two research areas in the field of the Experimental Analysis of Behaviour are relevant to the study of relapse behaviour. Firstly, Behavioural Momentum Theory addresses the question of what factors determine how disruptable the recently-reinforced behaviour is. Secondly, Resurgence addresses the question of what determines which behaviours increases as the recently-reinforced behaviour decreases following disruption.

Eight experimentally naive pigeons were trained in a matching-to-sample task. Subjects were reinforced for making one response in Condition 1, and a different response in Condition 2. Condition 2 responding was disrupted by three single extinction sessions which occurred at successive points during the condition. Resistance to change was measured as the persistence of responding, consistent with Condition 2 contingencies, during extinction. Resurgence was measured as a shift back to responding which was consistent with Condition 1 contingencies, when reinforcement was withheld in Condition 2.

In Group 1 (Extinction Group), responding consistent with Condition 1 contingencies was specifically extinguished in the process ofreinforcing responding in Condition 2. In Group 2 (Response Prevention Group), responding consistent with Condition 1 contingencies was prevented from being reinforced through never presenting choice stimuli in Condition 2 that were consistent with Condition I contingencies.

Results indicated that an increase in the duration of reinforcement contingencies did not increase resistance to change. However, subjects in the Extinction Group maintained a higher proportion of correct response during extinction, relative to baseline, compared to subjects in the Response Prevention Group. Difficulties in the measurement of resistance to change, caused by the experimental procedure used, were discussed.

Original analysis of errors made during extinction showed resurgence of responding consistent with Condition I contingencies in the Response Prevention Group but not in the Extinction Group. This finding held when a lack of stimulus control was accounted for in the analysis of errors. In the Response Prevention Group, resurgence decreased as the duration of reinforcement contingencies in Condition 2 increased. These findings were shown to be consistent with the response prevention hypothesis of resurgence. The results of the present study were discussed in light of their application to relapse behaviour in humans.