Recently, a method has been proposed for the quantitative measurement of regional myocardial blood flow (MBF) using oxygen-15-labelled water and positron emission tomography (PET). A multicentre project was organized with the intention of evaluating the accuracy of this method, particularly as a multicentre clinical investigative tool. Each of seven institutions performed PET studies on more than five normal volunteers following a specified protocol. The PET study included a transmission scan, a 15O-carbon monoxide static scan and a 15O-water dynamic scan, thereby yielding MBF values which should have been independent of the spatial resolution of the PET scanner employed. Fifty-three subjects (aged 20-63 years, mean+/-SD 36+/-12 years) were studied at rest, and 31 of these subjects were also studied after dipyridamole in five institutions. Inter-institution consistency and intra-subject variation in MBF values were then evaluated. MBF averaged for all subjects was 0.93+/-0.34 ml min(-1) g(-1) at rest and 3.40+/-1.73 ml min(-1) g(-1) after the administration of dipyridamole, and the flow reserve (defined as the ratio of the two MBF values) was 3.82+/-2.12; these values are consistent with previous reports. Resting MBF values were significantly correlated with the heart rate-blood pressure product (RPP) (y=0.31+6.56E-5x, P<0.010), and RPP was in resting MBF observed in all institutions was well explained by the age-dependent RPP. No significant difference was observed in resting MBF among the institutions. Except in one institution, no significant difference was seen in dipyridamole MBF or myocardial flow reserve. No significant difference was found among the myocardial segments. Regional variation was reasonably small in five institutions, but was not acceptable in two institutions, which was attributed to the scanner performance. These observations suggest that the 15O-water PET technique is useful for a multicentre clinical study if the PET scanner can provide time-activity data with good count statistics.