Radial artery tonometry provides continuous measurement of non-invasive arterial pressure (CNAP) by a sensor positioned above the radial artery. An inflatable upper arm cuff enables intermittent oscillometric calibration. CNAP was compared with invasive radial artery pressure recordings from the opposite wrist in 22 high-risk surgical patients with an inter-arm oscillometric mean arterial pressure difference < or = 10 mm Hg. Oscillometric, tonometric and invasive digital pressure values, and invasive and CNAP waveforms were obtained by the same instrument (Colin BP-508). Correlation coefficients (r) of invasive vs oscillometric values (n = 481 pairs) were 0.83, 0.90 and 0.92, and mean absolute errors of oscillometry were 7.6, 4.7, and 2.6 mm Hg for systolic, diastolic and mean arterial pressures, respectively. Correlation was poor for systolic (r = 0.80), diastolic (r = 0.77) and mean (r = 0.84) invasive vs CNAP values (n = 1375). Compared with oscillometry, mean absolute errors of 15.2, 10.9 and 9.4 mm Hg for systolic, diastolic and mean CNAP, respectively, were significantly (P < 0.001) higher. Mean prediction errors of CNAP, compared with invasive values, were -5.8 (SD 14.2) mm Hg for systolic, +7.2 (8.3) mm Hg for diastolic and +3.9 (8.8) mm Hg for mean arterial pressure. Individual patient accuracy of CNAP was assessed as good (individual prediction error < or = 5 (8) mm Hg and individual absolute error < or = 10 mm Hg) in seven patients, as acceptable (< or = 10 (12) and < or = 15 mm Hg) in 11 patients, and as inadequate in four of 22 patients. Individual accuracy of oscillometry was good or acceptable in all 22 patients. The trend in CNAP changes (difference between consecutive measurements) was sufficiently accurate during induction of anaesthesia, as only 47 (7.6%), 14 (2.3%) and 27 (4.4%) of 616 systolic, diastolic and mean CNAP values differed by more than 10 mm Hg of invasive pressure trends. We conclude that: intermittent oscillometry provides accurate arterial pressure monitoring; CNAP measurements offer a reliable trend indicator of pressure changes during induction of anaesthesia and may be considered an alternative to invasive pressure measurements, should arterial cannulation be difficult in an awake patient; and accuracy of absolute CNAP values is only moderate and unpredictable, thus radial artery tonometry should not replace invasive monitoring in high-risk patients during major surgical procedures.