A systemic graft-vs.-host (GVH) reaction was initiated by the intravenous injection of parental strain thoracic duct lymphocytes (TDL) into irradiated F1 hybrid recipients with in-dwelling thoracic duct cannulae. The migration of the donor lymphocytes was followed by labeling them in vitro with either [3H] or [14C]uridine and measuring radioactivity by scintillation counting of the spleen and lymph nodes of the recipients removed 24 h after injection and in TDL collected throughout this period. The localization of labeled cells was always compared to that of a reference population of nonreactive lymphocytes, e.g. F1 hybrid, labeled with the alternative isotope (Fig. 1). A consistent surplus of the reactive label was found in the spleen which was balanced by a deficit of the reactive label in TDL; lymph nodes gave intermediate values. The same distribution pattern was noted when the reference population was a specifically unresponsive population of the parental strain. This differential distribution depends on recognition of the recipient's Ag-B antigens because when normal lymphocytes were injected together with specifically unresponsive lymphocytes into a "third party" F1 hybrid (against which both populations were reactive) there was no surplus of the normal cells in the spleen and no deficit in the lymph. Moreover in an Ag-B identical strain combination there was no detectable difference in the distribution of reactive and nonreactive populations. The distribution of a labeled reaction population can be accounted for if a substantial minority of cells are immobilized in the spleen and lymph nodes as a consequence of antigen recognition (Fig. 3). When the donor cells in the spleen were assayed 24 h after injection there was paradoxically a slight reduction in their specific GVH activity, which is at least partly because they are under-represented in a single cell suspension. The size of the splenic surplus (23%) and the thoracic duct deficit (12%) suggested that the minority of nonimmune lymphocytes which recognize each Ag-B complex carry 12% of the radioactive label in the original population. It is argued that this provides a near estimate of the frequency of T lymphocytes which can recognize each Ag-B antigenic complex.