The Effects of Reinforcer Distribution During Response Elimination on Resurgence of an Instrumental Response

Resurgence is the relapse of an extinguished instrumental behavior that can occur when an alternative behavior that was introduced to replace it is itself extinguished. In a typical resurgence experiment, rats are trained to make a response (R1) for food reinforcers. In a second phase, responses on R1 are no longer reinforced, but a new response (R2) is introduced and responses on it are reinforced. During a third phase, reinforcement for R2 is removed and behavior on R1 often returns (or resurges ) despite remaining on extinction. The current experiments were designed to examine the effects of the temporal distribution of reinforcers delivered during Phase 2 on later resurgence. The role of these alternative reinforcers is central to theories that have been proposed to account for resurgence. The experiments provided a special opportunity to contrast predictions made by the Shahan-Sweeney Model (Shahan & Sweeney, 2011) and a contextual account of resurgence (Winterbauer & Bouton, 2010). Experiments 1 and 2 examined resurgence when alternative reinforcement during Phase 2 was delivered according to the same set of daily reinforcement schedules presented in different orders. That is, one group received rich reinforcement rates that were gradually thinned to leaner ones (Group Thinning) and another group received lean rates that were gradually increased to richer ones (Group Reverse Thinning). Both procedures weakened resurgence compared to that in a group that received the richest rate (a variable interval, or VI 10-s schedule that arranged for a reinforcer to be available for a response every 10s on average) during all of the Phase 2 sessions. However, the forward thinning procedure was more effective than the reverse thinning procedure at eliminating the resurgence effect. Experiment 3 examined resurgence when alternative reinforcement was only available for R2 during every other session. The results indicated that daily alternations of a VI 10-s schedule with an extinction schedule for R2 weakened resurgence compared to groups that either received the same average rate over the entire phase (VI 17.5-s) or that received the same terminal rate (VI 10-s) in every session. The Shahan-Sweeney model cannot account for several of the current results. Instead, the results are most consistent with a contextual account of resurgence. That is, resurgence can be conceptualized as an ABC renewal effect in which extinguished R1 behavior returns when an animal is removed from an extinction context provided by R2 reinforcement. Lean reinforcement rates at any time during Phase 2 allow the animal to learn to inhibit R1 under conditions that generalize to the extinction conditions that prevail during the resurgence test. The results also suggest that experience with alternating extinction sessions or lean reinforcement rates close to the final resurgence test are especially effective at eliminating the resurgence effect