To update the evidence on the association between smoking and mortality, we analyzed data from a population-based prospective study in Japan. In total, 19950 men and 21534 women aged 40 - 59 who reported their smoking history and had no serious disease at baseline survey were followed. During 1990 - 1999, 1014 men and 500 women died. Smokers were associated with an unhealthy lifestyle. Relative risks (RRs) for selected cause of death due to smoking were slightly attenuated by adjusting for possible confounding factors. Age- and area-adjusted RRs of male current smokers compared with never smokers were 1.66 (95% confidence intervals (CI): 1.40, 1.95) for all causes, 1.69 (1.31, 2.18) for all cancers, 1.67 (1.20, 2.34) for all circulatory system disease, and 1.63 (1.24, 2.15) for other causes, while those of females were 2.03 (1.52, 2.73), 2.06 (1.35, 3.15), 2.99 (1.75, 5.11), 1.31 (0.69, 2.51), respectively. After adjusting for multivariate variables, the corresponding RRs of male smokers were 1.55 (1.29, 1.86), 1.61 (1.20, 2.15), 1.41 (0.97, 2.03), and 1.61 (1.17, 2.19), against 1.89 (1.36, 2.62), 1.83 (1.14, 2.95), 2.72 (1.45, 5.07), and 1.39 (0.71, 2.73) for females. Twenty-two percent of death from all causes, 25% of all cancer, and 17% of all circulatory system disease deaths, could be attributed to cigarette smoking in males, and 5%, 4%, and 11% in females, respectively. Cumulative dose as indicated by pack-years was clearly associated with cancer death. These findings provided information as to the quantitative risk for premature death due to smoking among middle-aged Japanese men and women, and showed that the elevated risk was not explained by the unhealthy lifestyle of smokers.