To evaluate possible misclassification of smokers and nonsmokers, we compared self-reported cigarette consumption and serum cotinine levels in a sample of 743 Mexican American participants in the Hispanic Health and Nutrition Examination Survey (HHANES). The study sample was stratified by sex and self-reported cigarettes consumed per day (0, 1 to 9, 10 to 19, and greater than or equal to 20) and selected from those with available serum. We defined biochemical smokers as persons with serum cotinine levels greater than or equal to 0.084 microM/L (14 ng/ml). Misclassification was defined as a discrepancy between self-reported smoking and the serum cotinine level used to define a biochemical smoker. Of 189 self-reported nonsmokers, 12 (6.3%) were defined as biochemical smokers and possibly misclassified by self-report. Among 124 never smokers only 5 (4%) were biochemical smokers compared with 7 of 65 (10.8%) self-reported former smokers. Only 1 of the 12 misclassified nonsmokers reported living with a current smoker. In 9 of the 12 misclassified nonsmokers, serum cotinine levels were consistent with light smoking. Among the 547 self-reported smokers, 66 (12.1%) were found to have serum cotinine levels less than or equal to 0.084 microM/L (14 ng/ml) and possibly misclassified by self-report. Of these, one person reported 20 or more cigarettes per day. We conclude that self-reported cigarette consumption may be an insufficient measure of the risks associated with tobacco use and measurement of serum cotinine may be important to assess the magnitude of misclassification of smoking status in epidemiologic studies.