We examined longitudinal changes in smoking behavior among older adults in three community cohorts of the Established Populations for Epidemiologic Studies of the Elderly. Smoking prevalence declined from 15% at baseline to 9% during 6 years of follow-up. Annual smoking cessation and relapse rates were 10% and less than 1%, respectively. Interval diagnosis of myocardial infarction, stroke, or cancer increased subsequent smoking cessation but not relapse. Although smoking cessation around diagnosis is increased, primary prevention could yield greater benefits.