The relation of longitudinally measured blood pressure to cognitive performance in the absence of clinically diagnosed cerebrovascular disease was investigated in the Framingham Study. In 1976-1978, neuropsychologic testing was administered to 1993 participants aged 55-89 years. Performance on an education-adjusted composite of these tests was examined in relation to measures of chronicity of hypertension as well as the average systolic and average diastolic blood pressure. All analyses were stratified by antihypertensive medication use during the 2 years prior to cognitive testing and adjusted for age, sex, occupation, alcohol consumption, and participation rate in prior examination cycles. Among subjects on drug therapy for hypertension, there was no association between cognitive performance and longitudinally measured blood pressure. The proportion of cycles in which hypertension was present and average systolic and diastolic blood pressure had a significant inverse relation with cognitive performance only in the group not on antihypertensive drug therapy. However, among subjects on antihypertensive medication at earlier cycles, there was a highly significant graded relation between cognitive impairment and the probability of being off medication at the time of testing. These results suggest that hypertension-related subclinical vascular disease is not an important cause of cognitive impairment in the elderly. Cognitive impairment may, however, be associated with a reduced adherence to drug treatment regimens.