To develop a procedure for maximizing the discrimination of smoking status, the authors analysed parallel samples of thiocyanate and cotinine in serum, and carbon monoxide (CO) in expired air in a cohort of 145 male subjects aged 45-65 years. The sensitivity and specificity were 93% and 82%, 97% and 83%, and 98% and 100% for thiocyanate, cotinine, and CO respectively. The results were not significantly improved when combining two or three methods as compared with CO separately. Also, cotinine in urine was analysed in a subgroup of 21 subjects. The correlation coefficient between cotinine in serum and urine was 0.92. In a subgroup of 44 subjects with extensive information on smoking habits, CO was the only indicator significantly related to the quantity of tobacco smoked. We conclude that CO seems sufficient for validating smoking status, but as atypical smokers who are vulnerable to misclassification may be overrepresented in smoking cessation programmes, combining two methods could still be useful. Validating the amount of tobacco smoked is of limited use with the current methods.