The conventional approach to cytotoxic T-lymphocyte (CTL) induction uses maximal antigen concentration with the intent of eliciting more CTL. However, the efficacy of this approach has not been systematically explored with regard to the quality of the CTLs elicited or their in vivo functionality. Here, we show that a diametrically opposite approach elicits CTLs that are much more effective at clearing virus. CTLs specific for a defined peptide epitope were selectively expanded with various concentrations of peptide antigen. CTLs generated with exceedingly low-dose peptide lysed targets sensitized with > 100-fold less peptide than CTLs generated with high-dose peptide. Differences in expression of T-cell antigen receptors or a number of other accessory molecules did not account for the functional differences. Further, high-avidity CTLs adoptively transferred into severe combined immunodeficient mice were 100- to 1000-fold more effective at viral clearance than the low-avidity CTLs, despite the fact that all CTL lines lysed virus-infected targets in vitro. Thus, the quality of CTLs is as important as the quantity of CTLs for adoptive immunotherapy, and the ability to kill virally infected targets in vitro is not predictive of in vivo efficacy, whereas the determinant density requirement described here is predictive. Application of these principles may be critical in developing effective adoptive cellular immunotherapy for viral infections and cancer.