Forty-four eyes in 38 children were treated between 1963 and 1991 by external radiotherapy for retinoblastoma. Treatment modalities varied widely during this period; in addition to radiotherapy there was chemotherapy (16/44), photocoagulation (14/44), and laser therapy or cryotherapy (14/44). Treatment technique and dose fractionation also varied widely; lateral beam technique (39/44) versus anterior or anterior/lateral beam; doses per fraction ranged from 1 to 4.5 Gy, total doses from 30 to 61.5 Gy, and overall times from 22 to 49 days. Patients were followed at 3-month intervals, and actuarial survival at 10 years was 88%, with 62% local control. Ten eyes showed clinical evidence of retinopathy. A multivariate analysis of factors associated with increased risk of retinopathy was carried out using the Cox proportional hazards model and the mixture model of Farewell. The estimated latent time was 17 months (95% confidence interval, 14-20 months). The only factors found to be significantly associated with retinopathy were total dose multiplied by dose per fraction, or total dose normalized to the equivalent total dose in 2-Gy fractions as estimated from the LQ model, and these gave equivalent descriptions. There were trends (not significant) for increased risk of retinopathy when treatments included chemotherapy or photocoagulation, and for decreased risk (also not significant) when cryotherapy was used in conjunction with radiotherapy. No significance could be attached to any of the following: number of sites per eye, Reese-Ellsworth stage, and family history. We conclude that hypofractionation carries a significant risk for retinopathy in the treatment of retinoblastoma.