The data reported here document levels of accuracy in reports of blood pressure and identify correlates of inaccurate reporting. The data come from a long-term follow-up of a cohort of African-American women who registered for antepartum care between September, 1967 and June, 1969. At the follow-up interview, these women were asked whether they had ever received a diagnosis of hypertension from a physician. The self-reports of hypertension were compared with information contained in the medical records of these women. Twenty-five percent reported having high blood pressure but 53% of these reports were unconfirmed by their medical records (overall misreporting rates was 15.9% with 2.5% underreporting and 13.4% overreporting). The factors related to misreporting included a psychiatric diagnosis (based on the Diagnostic Interview Schedule) of major depressive disorder or drug and/or alcohol abuse and a small social network. The conjunction of these three variables significantly affected accuracy of reporting (100% misreporting with all three variables). These results suggest that, using currently standard methodology, there is an unreliable subpopulation of respondents in health surveys that may require the collection of data on health status from a second source to confirm data from self-reported health measures.