To describe serum cotinine levels in a rural Italian population and to examine its usefulness as an epidemiologic biomarker of nicotine exposure, cross-sectional data collected in 1993 for the MATISS Project (2098 men and 1352 women, aged 20-79 years) were used. The study population consisted of 977 current smokers, 882 nonsmokers reporting exposure to environmental tobacco smoke (ETS) and 1520 nonsmokers reporting no ETS exposure. Mean values of serum cotinine measured by radioimmunoassay for never smokers, ex-smokers and current smokers (including four categories of cigarette consumption), and for categories of ETS exposure in all nonsmokers were calculated. In univariate analysis, there was a positive association between self-reported nicotine exposure and serum cotinine levels in all groups. Using self-reported status as truth, sensitivity and specificity for various cotinine cutoff points were estimated to distinguish nonsmokers from smokers. The value of 15 ng/mL represented the best combined levels of sensitivity (95%) and specificity (96%). Using this cutoff point, the overall misclassification rate for self-reported nonsmokers was 2.1% and about two times greater for the more vs. the less educated. In multivariate analysis, reported ETS exposure among nonsmokers was significantly associated with serum cotinine even after adjusting for age, socio-demographic and behavioural factors, though the strength of the association was not strong. In conclusion, serum cotinine represents a reliable epidemiological marker of nicotine intake and may be helpful when studying ETS exposure. Improved information collection is needed to reduce misclassification among nonsmokers and enhance our understanding of the relationship between ETS and cotinine measures.