This study tested the predictive value of self-rated health on 3-year mortality, with attention focused on healthy, high-functioning elders. Data from the MacArthur Field Study of Successful Aging were utilized. Subjects were 70–79-year-old (at baseline) residents of the communities of East Boston, MA, New Haven, CT, and a five-county area in and around Durham County, NC (N = 1192) which comprise three sites of the Established Populations for Epidemiologic Studies of the Elderly (EPESE). In-home interviews were conducted in 1988 and 1991. Logistic regression was performed to create odds ratios adjusted for age, sex, race, marital status, education, alcohol consumption, cigarette smoking, chronic diseases, past hospitalizations, and cognitive function. The sample was then divided into healthy and less healthy cohorts based on number of chronic diseases, and the analyses were repeated. The adjusted odds ratios for self-rated health (poor/bad ratings compared to excellent ratings) in relation to mortality were 19.56 in the general sample, 93.51 in the healthy cohort, and 2.75 in the less healthy cohort. Self-rated health is predictive of mortality in controlled analyses, with the greatest impact seen in healthy individuals. Health care professionals should be sensitive to the significance of poor self-rated health in apparently healthy patients.