Children with acute lymphocytic leukemia, who were in remission after induction with prednisone and vincristine and consolidation with intravenous methotrexate, were randomized into three groups receiving (1) no further therapy, (2) BCG, and (3) chemotherapy with biweekly methotrexate and monthly prednisone and vincristine. Children continuing in remission after 8 mo on chemotherapy in group 3 were rerandomized into three similar groups, i.e., no therapy, BCG, and chemotherapy. In the primary randomization, the median duration of remission was identical in the groups receiving no therapy or BCG, (4 and 4.3 mo respectively), and both were significantly less than the median duration of remission on chemotherapy which had not been reached prior to secondary randomization at 8 mo. Results of secondary randomization were similar to those of primary randomization. As used in this study, BCG was ineffective in prolonging drug-induced remissions either early in remission or when the leukemic cell population might have been further reduced after 8 mo of maintenance chemotherapy.