Lymphoblast colony-culture of adult acute lymphoblastic leukemia (ALL) was studied to explore its clinical implication. Among 13 marrow cultures from ALL patients with full-blown disease, 11 developed leukemic colonies. A type of colony, very similar to a lymphoblastic colony and possibly T-cell in origin, could be found in four cultures of the six control marrows. To minimize the difficulty in differentiating a leukemic blast colony and a normal lymphocyte colony, based solely on morphology, a quantitative approach was used. Since both the mean of blast colony count and the mean of blast percentage of leukemic marrow were significantly higher than those of the control group, mean value plus two standard deviations of the control group were defined arbitrarily as upper normal limits. The defined normal range was then used to examine the relationship between results of the cultures and clinical outcome for the ALL patients. Early relapse or incomplete remission following chemotherapy could be predicted in four patients by these quantitative colony-culture assays 0.5-2 months before full-blown disease. The low colony count and low blast percentage in the colony-culture assay of the fifth patient is compatible with the clinical observation of continuous remission. One culture, growing clusters only, had an increased blast percentage; this correlated well with cytogenetic relapse two months later. In summary, the quantitative colony-culture assay could detect morphologically unidentifiable leukemic cells in ALL patients with early relapse or incomplete remission. This quantitative colony-culture system, though not ultrasensitive in the detection of minimal residual leukemic cells, was of potential value as a prognostic assay.