Early clinical trials using T lymphocyte-depleted human marrow for transplantation have reported that such grafts reduce, to varying degrees, both the incidence and the severity of graft-v-host disease (GVHD). However, to date, no clear estimates have been made as to what degree of T cell depletion is necessary to prevent GVHD in every case. To address this problem, we used a limiting dilution assay (LDA) to quantitate residual clonable T lymphocytes in human T cell-depleted bone marrow in 31 HLA-identical transplants for leukemia. The number of phytohemagglutinin -interleukin 2-responsive T lymphocytes determined by LDA and expressed as T cell per kilogram recipient weight was found to correlate with the subsequent development of GVHD: no patients who received less than 1 X 10(5) T cell per kilogram developed GVHD (N = 24). Of the seven patients who received 1 X 10(5) to 4.4 X 10(5) T cell per kilogram, four patients developed grade I or II skin GVHD. This study thus provides a quantitative estimate of the number of T lymphocytes necessary to initiate clinically detectable GVHD in an HLA- identical host.