Induction of a functional CD8(+) T-cell response is the important criterion for cancer vaccines, and it is unclear whether acute or chronic live vectors are better suited for cancer antigen delivery. We have evaluated the tumor protective ability of two recombinant vectors, Listeria monocytogenes (LM) and Salmonella typhimurium (ST), both expressing ovalbumin (OVA). Although both vectors induced a similar OVA-specific CD8(+) T-cell response in the long term, LM-OVA induced mainly central-phenotype (T(CM), CD44(high)CD62L(high)), whereas ST-OVA induced mainly effector-phenotype (T(EM), CD44(high)CD62L(low)) cells. Both vectors induced functional OVA-specific CD8(+) T cells that expressed IFN-gamma and killed targets specifically in vivo. However, only LM-OVA-vaccinated mice were protected against B16-OVA tumors. This correlated to the ability of CD8(+) T cells generated against LM-OVA, but not against ST-OVA, to produce interleukin 2 and exhibit profound homeostatic and antigen-induced proliferation in vivo. Furthermore, adoptive transfer of memory CD8(+) T cells generated against LM-OVA (but not against ST-OVA) into recipient mice resulted in their trafficking to tumor-draining lymph nodes conferring protection. Although cytotoxicity and IFN-gamma production are considered to be the principal functions of memory CD8(+) T cells, the vaccine delivery strategy may also influence memory CD8(+) T-cell quality, and ability to proliferate and traffic to tumors. Thus, for efficacy, cancer vaccines should be selected for their ability to induce self-renewing memory CD8(+) T cells (CD44(high)IL-7Ralpha(high)CD62L(high)) besides their effector functions.