Concern about the adverse affects of brain irradiation used in treatment for childhood leukaemia on children's learning have been put forward since the 1960s. Early work based on assessment of IQ suggested considerable problems associated with CNS irradiation of 2400cGy, and that children who were younger on diagnosis (below 5 years) were particularly at risk. Consequently, new protocols were introduced in which the amount of irradiation was reduced, or children were treated by chemotherapy alone. There is some evidence that reduction from 2400 to 1800cGy irradiation is beneficial, especially for younger children, as is treatment by chemotherapy alone. Methodological problems in conducting this work and limitations in relying on IQ tests are discussed. Where there are no indications of differences in survival, there would seem to be implications for reducing as far as possible the use of CNS directed therapy among young children treated for leukaemia.