In resurgence, a target behavior (R1) is acquired in an initial phase and extinguished in a second phase while an alternative behavior (R2) is reinforced. When reinforcement for the second response is removed, however, R1 behavior returns or “resurges.” The resurgence paradigm may have implications for understanding relapse after behavioral interventions in humans such as contingency management, or CM, in which (for example) drug users can earn vouchers contingent upon drug abstinence. The present experiments examined the effectiveness of a putative retrieval cue for treatment in attenuating the resurgence effects and determined the likely mechanism by which this cue functions. Experiment 1 established that a 2-second cue associated with delivery of the alternative reinforcer in Phase 2 can attenuate R1 resurgence and promote R2 behavior during testing. Experiment 2 demonstrated that this effect occurs regardless of whether the cue is delivered contingently or noncontingently on responding during the resurgence test, and Experiment 3 demonstrated that for the cue to be effective in reducing resurgence, it must be paired with alternative reinforcement during Phase 2. This might mean that pairing the cue with reinforcement serves to maintain attention to the cue. Experiment 4 suggested that a cue paired with alternative reinforcement did not serve as a conditioned reinforcer in that making it contingent on a new behavior did not increase the likelihood of that behavior. Experiment 5 demonstrated that the cue must be experienced in sessions that also include the extinction of R1. Experiment 6 found that a cue produced by R1 during the second phase of a resurgence paradigm (analogous to a conditioned inhibitor) does not attenuate resurgence of an extinguished instrumental response. Together, the results suggest that a neutral cue can serve as an effective cue that attenuates resurgence if it is first paired with alternative reinforcement and presented in sessions in which R1 is extinguished. One way to view the results is that creating greater generalization between the extinction context and the testing context results in less resurgence.