Thirty commercially available noninvasive devices for self-measurement of blood pressure were tested for accuracy and reproducibility against intraarterial measurement in 15 normotensive and 15 hypertensive subjects. For reference to outpatient measurements, a standard sphygmomanometer was included. Analysis of the data for any device was performed according to a statistical model in which two types of errors, representing accuracy and reproducibility, were distinguished: offset (mean difference between intraarterial measurement and device) and standard deviation (comprising intraindividual and interindividual variability). Linearity of the devices was analyzed by linear regression with intraarterial measurement as independent variable. The average offset of all tested devices amounted to 11.7 mm Hg (ranging from -0.2 to 21.4 between instruments) for systolic blood pressure and -1.6 mm Hg (range -11.4 to 8.1) for diastolic blood pressure. The standard deviation between instruments ranged from 7.0 to 15.4 mm Hg for systolic and from 3.3 to 16.0 mm Hg for diastolic blood pressure. Except for the devices based on the oscillometric principle, most (semi-) automated devices demonstrated a variability similar to the mercury sphygmomanometer (offset = 14.3 for systolic and 0.1 mm Hg for diastolic pressure; standard deviation = 7.5 for systolic and 5.6 mm Hg for diastolic blood pressure). Devices based on the oscillometric principle had a significantly larger variability. It is concluded that new devices should be tested for accuracy and reproducibility before application for clinical use. Blood pressure measurements from instruments with larger offset are only comparable to values from other instruments after rescaling in accordance with their offset.