Smoking prevalence in southern Germany was studied in 1984-1985 using a representative cohort of 4022 subjects aged 25 to 64 years, with 3753 reinterviewed in 1987-1988. Data were available for analysis from interviews on self-reported smoking behavior and from serum cotinine measurements in both investigations. More men than women reported current smoking, and particularly heavy smoking. Serum cotinine levels increased steadily with the daily number and nicotine yield of cigarettes smoked. Mean cotinine levels in ex-smokers were higher than those in never smokers, suggesting that a higher percentage of current smokers are misclassified as ex-smokers than never smokers. Using cotinine rather than self-reported smoking data increased the proportion of true smokers in the subgroup of self-reported smokers by about 3% in males and by about 1% in females. Data from the reinterviews revealed that reported smoking status confirmed by cotinine measurement in 1987-1988 conflicted in a number of cases with the data obtained in 1984-1985 using the same procedure. For example, 0.1% of those who stated they were current regular smokers, 4.3% of those who stated they were current occasional smokers, and 17.6% of those who stated they were ex-smokers in 1984-1985 claimed in 1987-1988 to have never smoked. This misclassification of ex-smokers was higher in women. Altogether the true proportion of ex-smokers among self-reported never smokers was about 9.7% (17.8% in men and 6.7% in women). The widely variable uptake of tobacco smoke by smokers, as well as the misclassification of true smokers and ex-smokers as never smokers, needs to be considered in epidemiological studies evaluating the health risks from both active and passive smoking.