The accuracy of the photon convolution/superposition dose algorithm employed in a commercial radiation treatment planning system was evaluated for conditions simulating tangential breast treatment. A breast phantom was fabricated from machineable wax and placed on the chest wall of an anthropomorphic phantom. Radiographic film was used to measure the dose distribution at the axial midplane of the breast phantom. Subsequently, thermoluminescent dosimeters (TLDs) were used to measure the dose at four points within the midplane to validate the accuracy of the film dosimetry. Film measurements were compared with calculations performed using the treatment planning system for four types of treatment: optimized wedged beams at 6 and 18 MV and two-dimensional compensated beams at 6 and 18 MV. Both the film- and TLD-measured doses had a precision of approximately 0.6%. The film-measured doses were approximately 1.5% lower than the TLD-measured doses, ranging from 0-3% at 6 MV and 0.5-1% at 18 MV. Such results placed a high level of confidence in the accuracy and precision of the film data. The measured and calculated doses agreed to within +/-3% for both the film and TLD measurements throughout the midplane exclusive of areas not having charged particle equilibrium. Good agreement was not expected within these regions due to the limitations in both film dosimetry and the dose-calculation algorithm. These results indicated that the treatment planning system calculates doses at the midplane with clinically acceptable accuracy in conditions simulating tangential breast treatment.