suggest that socio-demographic, lifestyle and mental health characteristics are independently associated with current smoking. These factors should be considered in smoking cessation policies as program components. (J Occup Health 2007; 49: 443-452) Key words: Smoking behavior, Socio-demographic factors, Education, Lifestyle, Affect balance scale, Japanese civil servants Tobacco smoking is of great public health concern world-wide [1][2][3] . Although the prevalence of smoking has steadily decreased, from 82.3% (in 1965) to 46.9% (in 2004) in men and from 15.7% (in 1965) to 13.2% (in 2004) in women 4) , the prevalence of smoking among Japanese men is still higher than in other developed countries 5) . Although the prevalence of smoking among Japanese women is lower than in Western countries, it has gradually increased especially among the 20-30 year age group 6, 7) . Recently, in European and American populations, many previous studies have reported that smoking is associated with socio-demographic factors [8][9][10] , low education level [10][11][12][13] , work stress 14,15) and mental health 16,17) . In Japan, some studies have demonstrated a significant correlation between smoking behavior and socio-demographic status 18,19) , low education level 20) , but only a limited range of variables related to smoking behavior have been analyzed in such studies, and many of them did not investigate for possible cofounders. Although some occidental studies have found a positive association between smoking and mental health in males 16,17) , no association was observed in the Japanese surveys 21,22) . The relationship between smoking behavior and possible cofounder factors such as socio-demographic factors, lifestyle, mental health, and work characteristics and so on has not been investigated fully in the Japanese population. The questionnaire included items on socio-demographic factors, education level, grade of employment, lifestyle, affect balance scale, and work characteristics. Smoking status was divided into current smoker, ex-smoker and never smoked. Multiple logistic regression analysis was performed to evaluate the relationship between smoking and the other items. Men presented a higher smoking prevalence rate than women (53.1% vs. 4.9%). In men, a strong relationship between current smoker and advanced age (40 yr or older), low education level, less physical activity, irregular breakfast and negative affect balance was found. Among men with a low education, the prevalence of smoking cessation was significantly lower in comparison to men with a high education. In women, being young (20-29 yr), unmarried (single or other), having a hobby, and irregular breakfast were associated with smoking behavior. Furthermore, smoking cessation was significantly associated with having a hobby and negative affect balance. The above results