The performance of computer planning systems has been tested by comparing calculations using the local beam data and computer facilities with measurements at the local installations. Relative absorbed dose distributions have been determined in a water phantom, irradiated with megavoltage photon beams with qualities from 60Co up to 25 MV. Three clinically relevant situations were studied: oblique incidence, tangential beams and wedged fields. For 45 degrees oblique incidence the mean deviation between the calculated and measured relative absorbed dose was less than 1%. Individual deviations, however, ranged from -2% up to +7%. A systematic difference, due to a straightforward application of the modified effective SSD method, was observed. For tangential irradiation the planning systems which do not consider the lack of scattering material showed deviations up to 8% between calculated and measured relative absorbed dose. For wedged beams, especially when they impinge obliquely on the phantom surface, differences were found up to 10%. Differences up to 20% were found for a point in the build-up region of an obliquely impinging wedged beam. From this study it can be concluded that planning systems may produce clinically unacceptable errors.