OBJECTIVE The objectives of this study are: (1) to investigate the dosimetric differences of the different treatment planning systems (TPS) in breast irradiation with tangential fields, and (2) to study the effect of beam characteristics on dose distributions in tangential breast irradiation with 6 MV linear accelerators from different manufacturers. METHODS Nine commercial and two university-based TPS are evaluated in this study. The computed tomographic scan of three representative patients, labeled as "small", "medium" and "large" based on their respective chest wall separations in the central axis plane (CAX) were used. For each patient, the tangential fields were set up in each TPS. The CAX distribution was optimized separately with lung correction, for each TPS based on the same set of optimization conditions. The isodose distributions in two other off-axis planes, one 6 cm cephalic and the other 6 cm caudal to the CAX plane were also computed. To investigate the effect of beam characteristics on dose distributions, a three-dimensional TPS was used to calculate the isodose distributions for three different linear accelerators, the Varian Clinac 6/100, the Siemens MD2 and the Philips SL/7 for the three patients. In addition, dose distributions obtained with 6 MV X-rays from two different accelerators, the Varian Clinac 6/100 and the Varian 2100C, were compared. RESULTS For all TPS, the dose distributions in all three planes agreed qualitatively to within +/- 5% for the "small" and the "medium" patients. For the "large" patient, all TPS agreed to within +/- 4% on the CAX plane. The isodose distributions in the caudal plane differed by +/- 5% among all TPS. In the cephalic plane in which the patient separation is much larger than that in the CAX plane, six TPS correctly calculated the dose distribution showing a cold spot in the center of the breast contour. The other five TPS showed that the center of the breast received adequate dose. Isodose distributions for 6 MV X-rays from three different accelerators differed by about +/- 3% for the "small" patient and more than +/- 5% for the "large" patient. For two different 6 MV machines of the same manufacturer, the isodose distribution agreed to within +/- 2% for all three planes for the "large" patient. CONCLUSIONS The differences observed among the various TPS in this study were within +/- 5% for both the "small" and the "medium" patients while doses at the hot spot exhibit a larger variation. The large discrepancy observed in the off-axis plane for the "large" patient is largely due to the inability of most TPS to incorporate the collimator angles in the dose calculation. Only six systems involved agreed to within +/- 5% for all three patients in all calculation planes. The difference in dose distributions obtained with three accelerators from different manufacturers is probably due to the difference in beam profiles. On the other hand, the 6 MV X-rays from two different models of linear accelerators from the same manufacturer have similar beam characteristics and the dose distributions are within +/- 2% of each other throughout the breast volume. In general, multi-institutional breast treatment data can be compared within a +/- 5% accuracy.