The recent introduction of a variety of techniques for removing T cells from bone marrow grafts has reduced the incidence of graft-versus-host disease (GVHD)*-associated morbidity and mortality (1–4). Whether this advance will be translated into improved patient survival is unclear at present, mainly because these procedures increase the risk of graft failure (2, 5,6).Since 1983 we have transplanted 25 consecutive leukemia patients with HLA-identical sibling grafts purged of T cells by a single incubation with the monoclonal antibody Campath-1 and donor complement (7). This approach was successful in reducing T cell contamination of the graft and preventing acute and chronic GVHD. In this group of patients two suffered irreversible graft failure and one developed reversible graft failure. In a similarly sized group of patients previously transplanted with unpurged marrow according to the Seattle protocol, no episodes of graft failure occurred (8). Since other causes of graft failure, such as drug toxicity or viral infections, could be largely excluded, this suggested that the graft failures were specifically related to the purging process.In haploidentical bone marrow transplantation (BMT) O'Reilly has identified residual host-versus-graft activity (HVG) as a cause of graft failure (9). The causes and mechanisms of graft failure in T-depleted HLA-identical sibling transplants have not been extensively investigated to date. In the three graft failures observed by us, the loss of the graft was preceded by the appearance of a population of activated lymphocytes. We have determined the phenotype and origin of this population and investigated its interactions with donor hemopoietic tissue in vitro.