Behavior change processes studied in the learning laboratory, such as extinction and counterconditioning, do not involve destruction of the original learning. Instead, they often result in new behavior that is strongly dependent on the current context, whether provided by external cues, internal state, recent events, or time. Lapse and relapse effects may therefore occur after various manipulations of the context. Theory and preliminary evidence suggests that long-term maintenance of changed behavior may be promoted by a number of factors, including situating the new learning in the most relevant contexts, providing retrieval cues after the new learning is complete, and varying the contexts in which the new learning takes place. Furthermore, because original learning is often more context free than the learning that replaces it, the most efficient way to reduce risk behavior in the general population may be to find ways to ensure that healthy behaviors and attitudes are learned first.