The importance of banded chromosome analyses in predicting long-term outcome in acute lymphoblastic leukemia (ALL) was evaluated in this follow-up study of 329 patients from the Third International Workshop on Chromosomes in Leukemia. Patients were divided into ten groups according to pretreatment karyotype: no abnormalities, one of the following structural abnormalities [the Philadelphia chromosome, translocations involving 8q24,t(4;11), 14q+, 6q-] or, in the remaining cases, modal number [less than 46, 46, 47 to 50, greater than 50]. Achievement and duration of complete remission (CR) and survival differed among chromosome groups (P less than .0001). Karyotype was an independent prognostic factor for duration of first CR and survival, even when age, initial leukocyte count (WBC), French-American-British (FAB) type, and immunologic phenotype were considered. Among adults, prolonged remission and survival were uncommon in all chromosome groups. Only in the normal karyotype group was median survival even two years. Among children, striking differences in long-term remission and survival were seen depending upon karyotype. Children in the greater than 50 group did best, with 70% remaining in first CR for a median duration in excess of five years. Children in the 47–50, 6q-, and normal karyotype groups also had prolonged survivals. In contrast, certain translocations [t(9;22)(q34;q11), t(4;11)(q21;q14–23), t(8;14)(q24;q32)] identified children who had short survivals, even in the presence of favorable prognostic factors including a low WBC, L1 morphology, and non-T, non-B immunologic phenotype. We conclude that chromosome analysis is required at diagnosis in patients with ALL, and that children with these specific translocations should be managed as having high-risk ALL.