The risk of developing intracranial tumors was studied in a Swedish cohort of 11,805 infants treated with 226Ra for hemangioma of the skin between 1930 and 1965. The cohort was followed up in the Swedish Cancer Register during the years 1958 through 1992, giving 402,958 person-years at risk. Forty-seven intracranial tumors developed in 46 individuals. The standardized incidence ratio (SIR) was 1.80 (95% CI 1.32-2.39). An excess number of cases was found for many histopathological subgroups, although the excess was statistically significant only for gliomas (n = 23, SIR 1.89, 95% CI 1.20-2.83) and meningiomas (n = 8, SIR 2.42, 95% CI 1.06-4.77). Absorbed doses to the brain have been calculated. The mean dose to the brain was only 7.2 cGy. The dose response was evaluated by Poisson regression methods. There was an excess of intracranial tumors in all dose categories, but no clear dose-response relationship could be seen for the data. This indicates that explanations other than irradiation may also be involved. The selection of the reference population has been studied and confirmed to be representative. For individuals treated before 7 months of age, however, the data could be fitted to both the linear and the linear-quadratic model. These results may indicate that the youngest infants may be more sensitive to ionizing radiation and that even very low doses to the brain may increase the risk for intracranial tumors.