Techniques based on the polymerase chain reaction (PCR) to detect rearrangement of the immunoglobulin or T-cell receptor genes can detect residual disease in leukemia and hence have the potential to improve prognosis and treatment. Such techniques may involve either detection of monoclonality, which is simple and quick but has limited sensitivity, or specific detection of the leukaemic clone, which is complex and time-consuming but has high sensitivity. The PCR was used to detect monoclonal rearrangements of the immunoglobulin heavy chain and/or T-cell receptor gamma chain genes in archival marrow specimens from 185 children with acute lymphoblastic leukemia who achieved remission during two consecutive Australasian trials of treatment. A monoclonal rearrangement was detected at diagnosis in 152 (84%) patients and in these patients detection of the same rearrangement in the remission marrow at the end of induction therapy was highly significantly correlated with outcome. There were nine patients in whom polymerase chain reaction showed only the monoclonal rearrangement and eight (89%) relapsed; there were 26 patients in whom PCR showed the leukemic monoclonal rearrangement as well as polyclonal rearrangements from normal lymphocytes and 12 (46%) relapsed; and there were 117 patients in whom only polyclonal rearrangements could be detected and only 29 (25%) relapsed. In patients who relapsed, remissions were shorter in those patients in whom the leukemic rearrangements had been detected in the remission marrow. Treatment in the later trial was more intensive than in the earlier trial, the results were better and the PCR detected the leukemic rearrangement in the remission marrow in significantly fewer patients. We conclude that detection by PCR of the monoclonal gene rearrangement of the leukemic clone in remission marrow indicates that numerous leukemic cells have survived induction therapy and is a good predictor of relapse. However, due to limited sensitivity of the test, failure to detect the leukemic clone by PCR is not a sufficiently good predictor of ultimate cure.