Serial cranial computed tomograms were carried out in 136 children with acute lymphoblastic leukaemia who were receiving 24 Gy or 18 Gy of cranial irradiation and continuing treatment with doses of methotrexate given weekly orally or intramuscularly. The findings were correlated with treatment variables, the development of fits, and the intelligence quotient (IQ). Reversible brain shrinkage, attributed to treatment with steroids, was found on 87 of 114 initial scans (76%); 14 showed changes in white matter during treatment (10%), and calcification was found in 13 either during or after treatment (10%). Eight children (6%) had fits, and in six of the eight there were changes in white matter or calcification on the scans. Comparison of the two radiotherapy dosages showed no difference in the incidence of abnormalities seen on computed tomography, fits, or serial IQ measurements, but children receiving intramuscular methotrexate had a higher incidence of calcification and a lower mean IQ at one year than those who received the drug orally, although this difference was not apparent later. Younger children were more likely to develop changes on computed tomograms and fits, and to have low IQs on completion of treatment, with changes most apparent in those less than 2 years of age. There were highly significant correlations between abnormalities on computed tomography, fits, and IQ. These findings confirm the neurological vulnerability of younger children with acute lymphoblastic leukaemia, show an association between abnormalities on computed tomography and intellectual deficit, and suggest that methotrexate is more toxic when given intramuscularly than orally. They provide no evidence that 18 Gy of cranial irradiation is less toxic than 24 Gy, and indicate the need for alternative treatment regimens.