Adoptive cell therapy (ACT) destroys tumors with infused cytotoxic T lymphocytes (CTLs). Although successful in some settings, ACT is compromised due to impaired survival or functional inactivation of the CTL. To better understand the mechanisms involved, we have exploited a mouse model of leukemia expressing ovalbumin as a tumor neoantigen to address these questions: (i) Is CTL impairment during ACT antigen specific? (ii) If yes, which are the antigen-presenting cells responsible? (iii) Can this information assist the development of complementary therapies to improve ACT? Our results indicate that the target (tumor) cells, not cross-presenting cells, are the main culprits of antigen-specific CTL inactivation. We find that the affinity/avidity of the CTL-tumor cell interaction has little influence on ACT outcomes, while tumor density is a major determinant. Reduction of tumor burden with mild non-lymphoablative and non-inflammatory chemotherapy can dramatically improve the efficacy of ACT and may minimize side-effects. We propose a general mechanism for the inactivation of anti-self CTL in the same tissues where the activity of anti-foreign CTL is preserved, based on the density of target cells. This mechanism, which we tentatively call stunning, may have evolved to protect infected sites from self-destruction and is exploited by tumors to inactivate CTL.