Few studies have assessed the accuracy of patient recall of doctor-diagnosed heart attack and stroke on postal questionnaire, yet such data are widely used in epidemiologic studies. In the national prospective British Regional Heart Study of 7,735 men aged 40-59 years, based in general practice and followed up for a mean 13.8 years, a mailed questionnaire was sent to all available survivors in 1992. Patient recall of doctor-diagnosed heart attack and stroke was compared with the 316 new general practice-reported heart attacks and 102 new general practice-reported strokes from the medical record reviews. Both study and general practice records were checked for all discordant findings, and corrections were made to the study database. Patients tended to overrecall major cardiovascular events more than they underrecalled them, 33% versus 6% for heart attacks and 25% versus 11% for strokes. Among overrecalled heart attacks, other circulatory problems were present in 78% of the subjects; transient ischemic attacks accounted for 57% of overrecalled strokes. In contrast, the general practice record review system tended to underreport events rather than to overreport them, 3% versus 0.3% for heart attacks and 23% versus 5% for strokes. Patient recall of doctor-diagnosed heart attack and stroke provides a useful method for estimating prevalence rates and resource needs, but the tendency to overestimation needs to be recognized. In etiologic studies when strict diagnostic case criteria are essential, patient recall should be used to complement rather than to supplant medical record data.