OBJECTIVE To delineate more precisely an operational threshold for making clinical decisions based on ambulatory blood pressure (ABP) measurement by studying the ABP in subjects who were diagnosed as either normotensive or hypertensive by conventional blood pressure (CBP) measurement. METHODS Twenty-four research groups recruited 7069 subjects. Of these, 4577 were normotensive (CBP < or = 140/90 mmHg), 719 were borderline hypertensive (systolic CBP 141-159 mmHg or diastolic CBP 91-94 mmHg) and 1773 were definitely hypertensive. Of the subjects in the last of these categories, 1324 had systolic hypertension (systolic CBP > or = 160 mmHg) and 1310 had diastolic hypertension (diastolic CBP > or = 95 mmHg). Combined systolic and diastolic hypertension was present in 861 subjects. Hypertension had been diagnosed from the mean of two to nine (median two) CBP measurements obtained at one to three (median two) visits. RESULTS The 95th centiles of the ABP distributions in the normotensive subjects were (systolic and diastolic, respectively) 133 and 82 mmHg for 24-h ABP, 140 and 88 mmHg for daytime ABP and 125 and 76 mmHg for night-time ABP, respectively. Of the subjects with systolic hypertension, 24% had 24-h systolic ABP < 133 mmHg. Similarly, 30% of those with diastolic hypertension had 24-h diastolic ABP < 82 mmHg. The probability that hypertensive subjects had 24-h ABP below these thresholds tended to increase with age and was two- to fourfold greater if the CBP of the subject had been measured at only one visit and if fewer than three CBP measurements had been averaged for establishing the diagnosis of hypertension. By contrast, for each 10-mmHg increment in systolic CBP, this probability decreased by 54% for 24-h systolic ABP and by 26% for 24-h diastolic ABP, and for each 5-mmHg increment in diastolic CBP it decreased by 6 and 9%, respectively. In comparison with 24-h ABP, the overlap in the daytime and night-time ABP between normotensive and hypertensive subjects was of similar magnitude and was influenced by the same factors. CONCLUSIONS The ABP distributions of the normotensive subjects included in the present international database were not materially different from those in previous reports in the literature. One-fifth to more than one-third of hypertensive subjects had an ABP which was below the 95th centile of the ABP of normotensive subjects, but this proportion decreased if the hypertensive subjects had shown a higher CBP upon repeated measurement. The prognostic implications of elevated CBP in the presence of normal ABP remain to be determined.