A multicomponent smoking relapse prevention treatment based on Marlatt and Gordon's (1980) model of the relapse process was developed and evaluated. Behavior-analytic methods were used to develop assessment instruments, training situations, and coping responses. The prevention components were presented in the context of a basic broad-spectrum stop-smoking program, and were compared with the basic program plus discussion control, and the basic program alone. Smoking-related dependent variables generally did not differ between groups at any time from pre-treatment to 12 month follow-up. Only the subjects in the relapse prevention condition improved problem-solving and social skills needed to cope with high-risk situations. These subjects also tended to take longer to relapse and smoke fewer cigarettes at the time of relapse. Subjects above the median level of competence on measures of social skill at post-treatment remained abstinent significantly longer. Maintenance of non-smoking was found to be related to the degree of competence with which individuals deal with high-risk situations. Results are discussed in relation to models of compliance with therapeutic regimens.