To assess the long-term relationship of midlife blood pressure levels to late-life cognitive function. The 4678 surviving cohort members of the prospective Honolulu Heart Program (baseline, 1965-1968) were examined a fourth time in 1991 through 1993 and given a cognitive test. The subjects were 3735 Japanese-American men living in Hawaii in the community or in institutions, with an average age of 78 years at the fourth examination. Cognitive function, measured by the 100-point Cognitive Abilities Screening Instrument (CASI), was categorized into good (reference: a CASI score of 92 to 100), intermediate (< 92 to 82), and poor (< 82). Midlife systolic blood pressure (SBP) and diastolic blood pressure (DBP) values were measured in 1965, 1968, and 1971. A respondent was classified into the following categories if two of three measurements fell into the following groups: for SBP, < 110, 110 to 139, 140 to 159, and > or = 160 mm Hg; and for DBP, < 80, 80 to 89, 90 to 94, and > or = 95 mm Hg. When we controlled for age and education, the risk for intermediate and poor cognitive function increased progressively with increasing level of midlife SBP category (P for trend < .03 and < .001, respectively). For every 10-mm Hg increase in SBP there was an increase in risk for intermediate cognitive function of 7% (95% confidence interval [CI], 3% to 11%) and for poor cognitive function of 9% (95% CI, 3% to 16%). Adjustment for prevalent stroke, coronary heart disease, and subclinical atherosclerosis reduced the strength of the relationship between midlife SBP and poor cognitive function to 5% (95% CI, 0% to 12%). The level of cognitive function was not associated with midlife DBP. Midlife SBP is a significant predictor of reduced cognitive function in later life. Early control of SBP levels may reduce the risk for cognitive impairment in old age.