Following an initial smoking program, 33 subjects were randomly assigned to one of four maintenance strategy conditions developed to investigate the Marlatt and Gordon cognitive-behavioral model of relapse. Two out of three components of the model were supported. Abstinent subjects given problem solving training did not show significant relapse while those given an attention placebo control did relapse significantly. Self-efficacy expectations predicted smoking rate up to 5-month follow-up. Contrary to expectations based on the Marlatt and Gordon model, subjects taught to cope with the guilt and sense of personal failure associated with a relapse episode showed considerable relapse.