We showed previously that NKT cell–deficient TCR Jα18^−/− mice are more susceptible to methylcholanthrene (MCA)-induced sarcomas, and that normal tumor surveillance can be restored by adoptive transfer of WT liver-derived NKT cells. Liver-derived NKT cells were used in these studies because of their relative abundance in this organ, and it was assumed that they were representative of NKT cells from other sites. We compared NKT cells from liver, thymus, and spleen for their ability to mediate rejection of the sarcoma cell line (MCA-1) in vivo, and found that this was a specialized function of liver-derived NKT cells. Furthermore, when CD4⁺ and CD4⁻ liver-derived NKT cells were administered separately, MCA-1 rejection was mediated primarily by the CD4⁻ fraction. Very similar results were achieved using the B16F10 melanoma metastasis model, which requires NKT cell stimulation with α-galactosylceramide. The impaired ability of thymus-derived NKT cells was due, in part, to their production of IL-4, because tumor immunity was clearly enhanced after transfer of IL-4–deficient thymus-derived NKT cells. This is the first study to demonstrate the existence of functionally distinct NKT cell subsets in vivo and may shed light on the long-appreciated paradox that NKT cells function as immunosuppressive cells in some disease models, whereas they promote cell-mediated immunity in others.