Analysis of leukemic cell phenotypes using cell surface antigens and various enzymes indicates that acute lymphoblastic leukemia (ALL) is a biologically heterogeneous disease consisting of four major subclasses with additional subsets existing within these subclasses. These different types of ALL appear to reflect sequential stages of early lymphocyte ontogeny. There is a strong association between cell phenotype and first remission duration in ALL (p trend less than 0.0001) and an equally strong correlation between remission duration and white blood cell count at presentation. If common ALL and thymic ALL (T-ALL) are compared after adjustment of white blood cell counts, then the prognostic differences between these two major subclasses almost disappear (p = 0.38). It is suggested, therefore, that an immunological (and enzymatic) phenotype of ALL subclasses may not be an independent correlate of prognosis but nevertheless is linked to other differentiation-linked features, especially growth rate and sites of clonal expansion (e.g., marrow versus thymus), which critically influence the size of the clonogenic leukemic population and its associated evolutionary status with respect to drug resistant mutants at the time of diagnosis and introduction of therapy. An extensive library of monoclonal antibodies has been used to further define the phenotypic heterogeneity of T-and non-T All. Several of the antigenic structures identified by these monoclonal antibodies have been isolated and characterized. T-ALL can be dissected into several subsets corresponding to stages of intrathymic differentiation. Non-T ALL (null-ALL, common ALL, and B-ALL) all have a phenotype indicative of B-lineage affiliation indicating that "non-T, non-B" ALL may originate from B-cell precursors in bone marrow. A cell type is identified in normal bone marrow which has the same identical monoclonal antibody-defined phenotype as common ALL and may provide the target cell for this disease.