There is no direct evidence that workplace environmental tobacco smoke (ETS) increases lung cancer risk. Demands for regulation of workplace smoking are based on studies reporting increased risk in non-smoking women whose husbands smoke. Although denying smoking can artificially elevate risk estimates, and although many studies reporting an increase have been conducted in Asia, no previous study of smoking habit misclassification has been conducted there. In this study 400 married Japanese women answered questions on smoking and ETS exposure and supplied urine for cotinine analysis. Of 106 with a cotinine/creatinine ratio (CCR) indicating current smoking (> 100 ng/mg), 22 reported never smoking. These misclassified smokers had a median CCR (1408 ng/mg) similar to the 78 self-reported current smokers (1483 ng/mg). Of current smokers, 89.7% had a currently smoking husband, while this was true of 51.0% of non-smokers. Among 264 confirmed non-smokers (with CCR < 100 ng/mg), CCR was non-significantly lower if the husband smoked (11.51 vs 17.98 ng/mg) and was unrelated to various indices of smoking by the husband. Japanese epidemiological studies using "marriage to a smoker" to index ETS exposure may therefore have compared groups with similar ETS exposure, suggesting that associations reported between lung cancer and this index in some of these studies may result from bias. While other biases, including confounding, may also be important, bias resulting from smoking misclassification combined with husband/wife smoking concordance is shown to be of major concern. The high misclassification rates in Japan, much higher than in Western populations, undermine conclusions from epidemiological studies conducted there.