The aim of this study was to investigate if four duplicate blood pressure (BP) remeasurements are sufficient for diagnosing hypertension in potentially hypertensive subjects. The subjects were 99 outpatients who were included on the basis of elevated diastolic (95 < or = DBP < or = 115 mm Hg) or systolic (160 < or = SBP < or = 200 mm Hg) BP. After inclusion all patients underwent nine subsequent duplicate BP measurements over a period of 7 months. None of the patients received hypotensive drug treatment during the study. Between the first (initial) and second measurements there were significant reductions in systolic (161.0-152.5 mm Hg) and diastolic (101.5-97.1 mm Hg) BPs (P < 0.01). Differences between the subsequent measurements were not statistically significant. A linear regression analysis proved that the "conceptual average BP' (the average of the last five visits) which was chosen as the reference value was stable. The decline of standard deviations of differences between two, three and four duplicate remeasurements on one hand, and the reference value on the other was found to be strikingly small. After four duplicate remeasurements, there was misclassification in 56% (systolic) and 38% (diastolic). We conclude that the numbers of two, three or four BP measurements recommended by international guidelines for diagnosing hypertension are too low. Even after four duplicate remeasurements a considerable amount of misclassification remains.