From August 1990 to February 1991, a dosimetry intercomparison of breast treatment was performed at all 21 radiotherapy centres in The Netherlands. The absorbed dose was measured in three planes in a breast phantom during tangential breast irradiation, according to a prescribed technique. The beam energy could be chosen by the radiotherapy centre as normally applied for this type of "patient", and varied between 60Co and 8 MV X rays. The dose measured by the visiting team in 22 points inside the phantom was compared with the dose calculated by the institution using their local treatment planning system. In the institutions the mean ratio (the mean value of the ratios of the absolute calculated dose and the measured absolute dose in the 22 points) varied between 0.92 and 1.08 with an overall mean ratio of 1.04. There was no significant difference in this ratio between the three planes in a particular institution. In the isocentre the mean ratio of calculated and measured dose was 1.021 with a SD of 0.028, i.e. the algorithms in the six different commercial treatment planning systems calculate the dose generally somewhat too high. In order to explain the results, a measurement of the output under reference conditions was performed at each treatment unit. The mean ratio of the dose stated by the institution and the dose measured by the visiting team was 1.011 with a SD of 0.015 with a maximum deviation of 0.040. This small deviation explains therefore only part of the variation in the ratio of calculated and measured dose for tangential breast irradiation. In several institutions large deviations between the actual beam data and the beam data implemented in the planning system were found, which was the main reason for the observed discrepancies.