The present investigation initiated a dismantling strategy in which subjects were assigned to treatment stages (preparation, aversion, maintenance) or to combinations of stages in a factorial design. It was hypothesized that multistage conditions would be superior to single stage conditions and that maintenance would retard relapse. Forty men and 33 women were randomly assigned to one of seven conditions. Results supported both hypotheses although significant effects were no longer evident at 12-month follow-up. Considerable relapse occurred in all conditions. Specific treatment components appeared to be of limited importance as indicated by generally very similar results for preparation and aversion. Booster sessions appeared to be ineffective. Interpretation of the findings is limited, however, by a relatively small subject enrollment in each condition. It was concluded that attempts to isolate extremely precise treatment elements are unlikely to be successful. Suggestions for further research included adoption of more structured maintenance strategies emphasizing coping skills and a more systematic focus upon potentially important process variables, notably group cohesiveness.