Developing improved therapy for the one-third or more of patients who can be expected to relapse after initial treatment for acute lymphoblastic leukemia would be less difficult if one could identify potential failures unequivocally at diagnosis. Subgroups of patients who should be considered candidates for highly experimental therapy include infants (less than 1 year of age), patients with the Philadelphia chromosome and perhaps patients with B-cell leukemia. The most important factor that determines the success of therapy after relapse is the length of the patient's initial remission. We recommend bone marrow transplantation for children whose first remission did not exceed 18 months. For all others, it appears that intensive chemotherapy affords as great a potential for cure as one could expect from transplantation. We favor intensive chemotherapy over transplantation in cases of late bone marrow relapse (greater than 18 months), because of the currently high peritransplantation mortality rate. It is not clear whether either modality will be adequate for patients relapsing on contemporary treatment programs.