INTRODUCTION: The association between smoking and mental disorders has been confirmed by several studies using cross-sectional and retrospective designs. The present study illustrates the need for differentiating subthreshold psychiatric disorders in the analysis. METHODS: The analysis is based on cumulative ("lifetime") prevalences of mental disorders and smoking in the Zurich study. This is a longitudinal community study with a stratified sample of 591 participants and six interviews from 1979 to 1999. RESULTS: The percentage of lifetime smokers in the Zurich study was higher both in persons with a lifetime psychiatric diagnosis (72%) and in persons with subthreshold disorders (60%) than in those without any diagnosis (40%). DISCUSSION: The association between smoking and mental disorders turned out to be clearly stronger if subthreshold mental disorders were appropriately considered in the analyses. Constructing appropriate reference groups is as crucial for the analysis of mental disorders and their outcomes as constructing adequate diagnostic groups.