Exposure to environmental tobacco smoke (ETS) is most often estimated using questionnaires, but they are unreliable. Biomarkers can provide valid information on ETS exposure, the preferred biomarker being cotinine. However, no reference range of hair cotinine exists to distinguish among active, passive, and unexposed nonsmokers. This study identifies cutoffs to validate cotinine as a marker for exposure to ETS. Data were obtained from six databases (four US, one Canada, one France). Active smoking and exposure to ETS were measured in the hair of women of reproductive age, pregnant women, their children, and neonates. Subjects were classified into active smokers, passively exposed to ETS, and unexposed nonsmokers. A total of 1746 cases were available for analysis. For active smokers, mean hair cotinine concentrations (95% confidence interval) were 2.3 to 3.1 ng/mg for nonpregnant women and 1.5 to 1.9 ng/mg for pregnant women. In the group of passive smokers, mean hair cotinine concentrations were 0.5 to 0.7 ng/mg for nonpregnant women, 0.04 to 0.09 ng/mg for pregnant women, 0.9 to 1.1 for children, and 1.2 to 1.7 for neonates. Among unexposed nonsmokers, mean hair cotinine was 0.2 to 0.4 ng/mg in nonpregnant women, 0.06 to 0.09 ng/mg in pregnant women, and 0.3 to 0.4 ng/mg in children. Cutoff values for hair cotinine were established to distinguish active smokers from passive or unexposed (0.8 ng/mg for nonpregnant women and 0.2 ng/mg for pregnant women). A cutoff value of 0.2 ng/mg was accurate in discriminating between exposed children and unexposed. These new values should facilitate clinical diagnosis of active and passive exposure to tobacco smoke. Such diagnosis is critical in pregnancy and in a large number of tobacco-induced medical conditions.