To investigate secular trends in prevalence and incidence of atrial fibrillation and associated factors, data from population studies of a northeast rural Japanese community, were explored. Cross-sectional studies for men and women aged 40-69 year were conducted in 1963-1966, 1972-1975 and 1984-1987. Age-adjusted prevalence rates of atrial fibrillation show no significant change in both men and women during these three periods. Over 80% of atrial fibrillation in each period showed no clinical evidence of rheumatic valvular disease, myocardial infarction, idiopathic cardiomyopathy, congenital heart disease and hyperthyroidism. Cohorts of men and women aged 40-69 year without atrial fibrillation at baseline were constructed in 1963-1966 (1,920 persons) and in 1972-1975 (2,325 persons) and followed until 1974 and 1987, respectively. In both cohorts, incidence of atrial fibrillation without these organic diseases was positively associated with hypertension related funduscopic abnormality and urine protein. Age-adjusted incidence rates of atrial fibrillation declined in both men and women between the two cohorts. This decline which was greater in hypertensives than in normotensives may in part reflect improvements in hypertension control in this community.