We have analysed the behaviour of minimal residual disease (MRD) after allogeneic bone marrow transplantation (allo-BMT) in 71 children with acute lymphoblastic leukaemia (ALL). The method relied on PCR of IgH, TCRdelta and/or TCRgamma gene rearrangements followed by electrophoretic size resolution and allele-specific oligoprobing. Patients were similarly conditioned; 55 received marrow from unrelated donors and 16 from related donors. MRD was assessed at various time-points up to 24 months after BMT. Three children were not evaluable due to transplant-related mortality. MRD was detected in 28/32 patients (88%) who relapsed post-BMT; 16 were positive at all times and 12 were initially negative but became positive at a median of 3 months (range 1.5-11) prior to relapse. In contrast, only eight of 36 (22%) patients who remained in continuing complete remission (CCR) (median follow-up 43 months, range 20-94) showed MRD at any time after BMT (P<0.0001). In these eight patients MRD was found up to 9 months after transplant and at low levels (0.01-0.001%). All eight (median follow-up 39 months, range 24-87) had at least two MRD-negative samples tested subsequently and five of the eight had evidence of grade I-II acute graft-versus-host disease (GvHD), raising the possibility of a graft-versus-leukaemia effect. In general, any evidence of MRD after allo-BMT is a poor prognostic sign. However, if immunotherapy were to be targeted towards patients with evidence of persisting MRD after BMT, the method described would expose only a small proportion of patients to unnecessary additional toxicity.