Background. Whether alloreactive T cells in a naive host derive from naive or memory T cells remains unclear. It is also unclear whether graft rejection alters the phenotype of these T cells. Proliferation assays and cytokine enzyme-linked immunosorbent assays performed on culture supernatants do not differentiate primary T-cell alloreactivity from recall responses in allograft-primed mice, suggesting that these methods are inadequate measures of the alloreactive immune repertoire. Methods. To better characterize alloreactivity in native and skin allograft-primed mice, we used a modified, high-resolution cytokine enzyme-linked immunosorbent spot assay capable of detecting cytokine production over short time periods. Results. Twenty-four- hour analysis of alloreactivity in mice that rejected fully MHC-disparate skin allografts revealed a high frequency of interferon (IFN)-γ- and interleukin (IL)-4-producing, L-selectin-negative T cells, consistent with a memory phenotype. In contrast, 24-hr allostimulation of T cells from naive mice resulted in IL-2 production with minimal secretion of IFN-γ or IL-4. The frequency of IL-2 producers was lo and their phenotype was L-selectin positive, suggesting that they were naive and not memory T cells. When maintained in culture of 48 hr, however, the T cells from the primary mixed lymphocyte reaction began producing IFN-γ, consistent with in vitro priming. Conclusions. The primary alloresponse does not seem to involve clones that have been preprimed by environmental antigens, but instead behaves similarly to self-MHC-restricted immunity directed toward prototypic protein antigens: T cells with a naive phenotype are specifically induced to differentiate into high-frequency memory populations. These findings may have important implications for therapeutic induction of allograft tolerance.