Community mortality surveillance for 1984-1988 was conducted by researchers of the Strong Heart Study, which examined the incidence, prevalence, and risk factors of cardiovascular disease in three American Indian populations, aged 45-74 years, in Arizona, Oklahoma, and South/North Dakota. All-cause and cardiovascular disease mortality rates were determined through the use of death certificate data. Cardiovascular disease deaths were confirmed by independent systematic review of medical records. In all three populations, men had higher all-cause and cardiovascular disease mortality rates than did women. Oklahoma exhibited slightly lower 5-year, age-adjusted, all-cause mortality (96/1,000) than did Arizona (107/1,000) and South/North Dakota (114/1,000). The leading cause of death among both sexes in Oklahoma and in South/North Dakota was cardiovascular disease. Diabetes mellitus led among Arizona women. The other major causes of death were cancer, liver disease including cirrhosis, and injury. When compared with the rates in each state, average annual all-cause mortality rates were higher for the American Indian populations in almost every age group. The all-cause annual mortality rates in the three Indian populations were close to rates in the US black population and higher than the rates of the entire US population and of US whites. This trend was amplified in the 45- to 64-year age group. Only in the 65- to 74-year age group did mortality rates in the Indian population approach those of the US population. Cardiovascular disease mortality rates were close to the US averages in Arizona and Oklahoma, but they were more than two times higher in South/North Dakota among those between 45 and 64 years of age. Thus, American Indians in Arizona, Oklahoma, and South/North Dakota exhibit high all-cause mortality rates. In particular, the South/North Dakota population cardiovascular disease death rate appears to present a potential target for community-based programs to intervene on known risk factors to promote healthy lifestyles.