The cellular basis of immunological memory, particularly with respect to T cells is not understood. In humans, monoclonal antibodies to CD45 have been used to identify memory (CD45R0) and naive (CD45RA) T cells. However, this identification has been called into question by various studies which suggest that high molecular weight CD45 isoforms may be re-expressed by previously activated cells. In the present study, using cultures which supported responses of naive T cells, we examined the responses of purified CD45R0brightRA- or CD45R0(-)-RAbright T cell subsets. The former subset was found to respond preferentially to recall antigens with minimal responses apparent to neo-(or non-recall)-antigens. The inverse pattern was found for CD45R0-RAbright T cells, which converted to CD45R0brightRA- after stimulation with a neo-antigen. Moreover, the two populations of T cells exhibited distinct response kinetics with a faster response evident from the CD45R0brightRA- T cells compared to the CD45R0-RAbright subset. The poor responses of CD45R0-RAbright T cells to recall antigens compared to neo-antigens suggests that this putative naive population is specifically depleted of reactive T cells following an encounter with antigen. We propose that T cell priming results in the stimulation of many CD45R0-RAbright T cells with various T cell receptor specificities from which memory T cells are selected for survival. If re-expression of higher molecular weight isoforms does occur in humans in vivo, our results suggest that R0 expression would be retained (CD45R0+RA+). Alternatively, if primed CD45R0-RAbright T cells exist, they are not prevalent in peripheral blood and thus may be sequestered within lymphoid tissues. Our data support the view that in human peripheral blood, CD45R0bright and CD45RAbright expression identify memory and naive CD4+ T cells, respectively.