A random sample of ten community-based Vita-Stat automatic blood pressure recorders (ABPR) was evaluated for accuracy and repeatability. Each of 518 adult volunteers had two randomly ordered systolic and diastolic Vita-Stat blood pressure measurements compared with two corresponding measurements obtained by a trained observer using a Random-Zero sphygmomanometer. Eight of the ten Vita-Stat recorders underestimated systolic blood pressure. In contrast, diastolic blood pressure was overestimated by nine of the ten machines. While the overall differences in average blood pressure were small (mean +/- SEM = -2.4 +/- 0.6 mm Hg for systolic blood pressure and +2.3 +/- 0.4 mm Hg for diastolic blood pressure), the magnitude of the average discrepancy varied considerably by machine (+4.7 to -13.8 mm Hg for systolic and +5.0 to -2.0 for diastolic blood pressure). At every level of systolic and diastolic blood pressure the Vita-Stat ABPR provided a less accurate method of classifying blood pressure among individuals than the human observer. Additional analyses exploring the ability of the Vita-Stat machine to measure an individual's blood pressure within 2, 4, or 6 mm Hg of the corresponding Random-Zero value again suggested that the Vita-Stat ABPR was less accurate than the human observer. Duplicate Vita-Stat readings were less repeatable than corresponding Random-Zero measurements. Based on the findings of this study, the Vita-Stat ABPR appears to be an unsatisfactory tool for community-based self-measurement of blood pressure.