Abstract A validation study was carried out on self-reported smoking for 1177 people in Sydney and Melbourne in 1983. Because of its long half life and the fact that smoking is its only source in body fluids, saliva cotinine was chosen as the validation measure. Cotinine levels above 250 nmol/1 were used to classify people as smokers. The sensitivity of self-reported smoking was 92.6% and the specificity was 93.4%. There was some evidence that people in the process of changing their smoking status might be slow in updating their self-classification. The smoking prevalence estimate based on cotinine levels was found to be 1.7% lower than that for self-reported smoking status. The small proportion of false negatives and false positives suggests that commercially collected data banks can be valid sources of prevalence data. Correlation between cotinine level and reported cigarette consumption was not affected by sample volume, and was similar to that achieved for carbon monoxide and thiocyanate at a low 0.34. Regression analysis using self-reported cigarette consumption filter/non-filter cigarettes, and time since last cigarette as predictors, explained 13.6% of the variance in cotinine level.