One hundred and twenty-six infants with acute lymphoblastic leukaemia (ALL) were treated on two consecutive protocols, Infant 87 (n = 40) and Infant 92 (n = 86), in an attempt to improve the poor prognosis of this disease. Both included intensive induction and consolidation with intrathecal and high-dose systemic therapy for central nervous system (CNS) protection. Intensification therapy was modified and high-dose chemotherapy with bone marrow transplantation in first remission was permitted in Infant 92. Four-year event-free survival was superior in Infant 92 (33%; 95% CI 23-44%) compared with Infant 87 (22.5%; 95% CI 12-37%) (P = 0.04) and survival at 4 years was also superior, 46% (95% CI 35-57%) c.f. 32.5% (95% CI 20-48%) (P = 0.01), largely as a result of a significant reduction in remission deaths. Twelve patients in Infant 92 underwent bone marrow transplantation (BMT) in first remission, but their survival was no better than that of patients receiving chemotherapy. Multivariate analysis of prognostic factors showed the adverse influence of younger age, CNS involvement at diagnosis and a high initial leucocyte count, but not of CD10 expression. Cytogenetic analysis, available in 93% of patients in Infant 92, showed that 67% had chromosomal rearrangements involving 11q23 of which 39% had the translocation t(4;11) (q21;q23). There was no significant difference in event-free survival between cytogenetic subgroups, although no children under 6 months of age with 11q23 abnormalities, other than the t(4;11), survived. In conclusion, infants with lymphoblastic leukaemia remain a high-risk group, but it is unclear whether their adverse prognosis can be attributed to unfavourable cytogenetics alone. The role of high-dose therapy and BMT in first remission remains uncertain.