OBJECTIVE To compare the short-term reproducibility of four diagnostic tests: resting blood pressure, exercise blood pressure, non-invasive daytime ambulatory blood pressure and echocardiographic left ventricular mass. METHODS Blinded, prospective test-retest (reliability) study. METHODS Hypertension research units in two teaching hospitals. METHODS Six normal volunteers and 22 patients with untreated borderline to mild hypertension, mean age 44 years. METHODS The intraclass correlation coefficient (RI) and standard deviation of the difference (SDD) between visits. RESULTS The mean blood pressures and left ventricular mass did not differ between visits. Concordance between visits reached RI = 0.86 systolic/0.66 diastolic for ambulatory blood pressure and RI = 0.85 systolic/0.64 diastolic for nurse-measured random-zero sphygmomanometer resting blood pressure. The respective variabilities were SDD = 9/8 and 8/8 mmHg. Submaximal exercise systolic blood pressure (SBP) and echo left ventricular mass showed excellent reliability. Echo left ventricular mass and resting SBP or ambulatory SBP were significantly more reproducible than resting diastolic blood pressure (DBP) or ambulatory DBP. CONCLUSIONS Despite averaging many readings within each day, clinically important between-visit variations in ambulatory blood pressure remained. The between-visit variability of daytime ambulatory blood pressure was similar to that of resting blood pressure when carefully measured by a research nurse. The echo left ventricular mass appears to be more reproducible over the short term than the current diagnostic standard for hypertension, the resting DBP.