The discrepancy between cigarette smoking status reported during an interview and measured level of serum cotinine, a nicotine biomarker, was investigated in a representative sample of the US population aged >/=17 years (N = 15,357). Data were collected from participants in the Third National Health and Nutrition Examination Survey (1988-1994). Among self-reported smokers, 7.5% (95% confidence interval: 6.3, 8.7) had a serum cotinine level less than or equal to 15.0 ng/ml, the selected cutoff point for identifying nonsmokers. Age (p < 0.01), race/ethnicity (p < 0.01), and average number of cigarettes smoked per day (p < 0.01) were associated with these discrepant findings. Among self-reported nonsmokers, 1.4% (95% confidence interval: 1.1, 1.7) had a serum cotinine level greater than 15.0 ng/ml, the selected cutoff point for identifying smokers. Race/ethnicity (p < 0.01), education (p < 0.01), number of household members who smoked in the home (p = 0.03), and self-reported smoking status from an earlier home interview (p < 0.01) were associated with these discrepant findings. Differences in smoking patterns, including the extent of nicotine dosing, may explain most of the discrepancy observed among self-reported smokers, whereas deception regarding smoking status may explain most of the discrepancy among self-reported nonsmokers. This study provides evidence that self-reported smoking status among adult respondents to a population-based survey conducted in a private medical setting is accurate.