BACKGROUND Low to moderate agreement between self-reported exposure to environmental tobacco smoke (ETS) and serum cotinine levels in non-smokers questions the accuracy of the measurement of ETS exposure. We examined the relationship of serum cotinine to different self-reported ETS questionnaires in a large community-based study. METHODS Subjects were derived from four Scottish MONICA surveys. Agreement between self-reported ETS (yes/no) and serum cotinine levels (> 0, 0) in non-smokers was tested by K, and the difference in cotinine levels among self-reported ETS exposure by ANOVA and the relationship by linear regression. RESULTS None of the values for K was > 0.24 for any ETS questionnaire. In non-smokers with serum cotinine > 0, cotinine levels increased with increasing ETS exposures. In the first and second surveys with the questionnaire of ETS exposure in the last 3 days, standardised coefficients were 0.28-0.39, while in the third and fourth surveys with the questionnaire of a total exposure to ETS at work, at home and in other places the standardised coefficients were 0.19-0.36, with the questionnaire of ETS daily exposure hours, 0.23-0.36. The relationship between self-reported ETS and cotinine levels varied with the questionnaires, and with the time of day of the blood sample collection. In current smokers, cotinine levels were significantly related to both the number of cigarettes smoked daily (the coefficients were 0.13-0.41) and time elapsed since the last cigarette (-0.24 to -0.40). CONCLUSIONS The findings raise the question of whether it is ideal to take only serum cotinine as an index of ETS exposure in adults, because of time delays between ETS exposure and blood collection, and suggest the combined use of appropriately worded self-reported questionnaires and cotinine levels to estimate ETS exposure.