To clarify the relationship between lifestyles and total mortality, a 6-year cohort study was conducted for three communities in Wakayama Prefecture. During 1988-1990, information on lifestyles were obtained from a self-administered questionnaire utilizing 125 items. The survey protocol included a concentrated effort to train all observers for standardization and quality control. For the baseline survey, a total of 4,238 residents (1,947 men and 2,291 women) aged 40 to 79 years old were recruited from resident registration lists. Of these, 3,048 participants (1,308 men and 1,740 women) formed the cohort (response rate: 72%). Death certificates and change of residence registrations were analyzed annually utilizing local public health center and town records. Relative risk (RR) for each lifestyle component against total mortality was calculated applying Mantel-Haenszel methods to adjust for age, past history, walking, smoking (male) and BMI (female). During 6 years, 171 deaths (106 men and 65 women) and 63 resident changes (20 men and 43 women) were observed. Statistical significant RRs were observed for walking (RR: 1.7; p < 0.01) in men, walking (RR: 1.9; p < 0.01) in women, BMI (RR: 2.1; p < 0.001) in women, when calculated with all deaths during 6 years. As of smoking for men, RR was 1,4 (not significant), but by considering the possibility of effects of a latent period, and excluding deaths occurring in 1989 and 1990, RR became 2.1 (p < 0.01). These results suggest that non-smoking, walking daily and maintaining a moderate BMI may lower total mortality in this population.