AIMS: To assess the validity of the Fagerström test for nicotine dependence (FTND, six items) and of a short-form of this questionnaire, the Heaviness of Smoking Index (HSI, two items), in a population of relatively light smokers. DESIGN: Comparison of item content with published definitions of addiction. Test-retest reliability and multiple tests of construct validity, based on a secondary analysis of a cohort study conducted between November 1995 and June 1996. SETTING: University of Geneva, Switzerland. PARTICIPANTS: Students (82%), academic (12%) and administrative staff (6%): 643 smokers at baseline and 482 smokers at follow-up. MEASUREMENTS: French-language versions of the FTND and HSI, smoking status, saliva cotinine level, self-efficacy for quitting smoking and other variables related to addiction with cigarettes. FINDINGS: A literature review showed that both composite scales fail to assess several recognised aspects of tobacco dependence. In this population of relatively light smokers (average: 12 cigarettes per day), both tests had important floor effects with, respectively, 55% and 63% of participants with scores equal to 0 or 1 on these scales. In addition, two of the FTND items (Difficult-to-refrain and Hate-most-to-give-up) had poor psychometric properties. Even though FTND and HSI correlated about as expected with criterion variables, the number of cigarettes smoked per day performed better than either composite scale on most validation criteria. CONCLUSION: In a population of relatively light smokers, FTND and HSI seem to measure little more than the number of cigarettes per day. Designing a new and more broadly applicable test of addiction to cigarettes is a research priority.