The accuracy of Monte Carlo codes in dose calculation systems relies on the correctness of the input data. Monte Carlo calculations are performed to generate phase-space descriptions of the Varian 2100C accelerator at 6 and 18 MeV. Before these data can be reliably used as the input for dose calculations in patients, they must be properly validated. This validation consists of three different stages: validation of the coding of the geometry, validation of the user code for the Monte Carlo code, and validation of calculated results. Geometric validation is performed by isolating and testing treatment head components independently. The user code is checked by testing for energy conservation and the variance reduction schemes incorporated into the user code are checked by comparison of results calculated with and without their employment. Validation of the phase-space description is performed by calculation of depth dose curves and lateral profiles for dose deposition in phantom, with difference plots used to illustrate any discrepancies. Calculated and experimental in-phantom output is also determined. After complete validation, the calculated data can then be reliably used as the input for dose calculations.