A dosimetric study of anterior electron beam irradiation for treatment of retinoblastoma was performed to evaluate the influence of tissue heterogeneities on the dose distribution within the eye and the accuracy of the dose calculated by a pencil beam algorithm. Film measurements were made in a variety of polystyrene phantoms and in a removable polystyrene eye incorporated into a tissue substitute phantom constructed from a human skull. Measurements in polystyrene phantoms were used to demonstrate the algorithm's ability to predict the effect of a lens block placed in the beam, as well as the eye's irregular surface shape. The eye phantom was used to measure dose distributions within the eye in both the sagittal and transverse planes in order to test the algorithm's ability to predict the dose distribution when bony heterogeneities are present. Results show (1) that previous treatment planning conclusions based on flat, uniform phantoms for central-axis depth dose are adequate; (2) that a three-dimensional heterogeneity correction is required for accurate dose calculations; and (3) that if only a two-dimensional heterogeneity correction is used in calculating the dose, it is more accurate for the sagittal than the transverse plane.