Cluster of differentiation (CD)8(+) T cells exist as naive, central memory, and effector memory subsets, and any of these populations can be genetically engineered into tumor-reactive effector cells for adoptive immunotherapy. However, the optimal subset from which to derive effector CD8(+) T cells for patient treatments is controversial and understudied. We investigated human CD8(+) T cells and found that naive cells were not only the most abundant subset but also the population most capable of in vitro expansion and T-cell receptor transgene expression. Despite increased expansion, naive-derived cells displayed minimal effector differentiation, a quality associated with greater efficacy after cell infusion. Similarly, the markers of terminal differentiation, killer cell lectin-like receptor G1 and CD57, were expressed at lower levels in cells of naive origin. Finally, naive-derived effector cells expressed higher CD27 and retained longer telomeres, characteristics that suggest greater proliferative potential and that have been linked to greater efficacy in clinical trials. Thus, these data suggest that naive cells resist terminal differentiation, or "exhaustion," maintain high replicative potential, and therefore may be the superior subset for use in adoptive immunotherapy.