The recognition by effector T lymphocytes of novel antigenic targets on tumor cells is the premise of specific targeted immunotherapy of cancer. With the molecular characterization of peptide epitopes from melanoma antigens, and more recently broadly expressed tumor antigens, there has been considerable enthusiasm for clinical evaluation of peptide tumor vaccines. Immunologic monitoring of vaccinated patients has demonstrated expansion of CD8+ T cells that react with the relevant peptide and, more importantly, with native tumor. In most instances however, vaccine-induced CD8+ T cell responses alone have not been sufficiently robust or sustained to translate into a high percentage of durable clinical responses. Vaccine strategies have also utilized dendritic cells (DCs) that have been modified to present tumor antigens. The superior antigen processing capacity and co-stimulatory function of DC convey a powerful stimulatory signal to both CD4+ and CD8+ T cells. Several strategies are attempting to broaden the immune response beyond single antigens by introducing the entire complement of tumor antigens into DCs. Adoptive immunotherapy is a promising strategy to recover tumor-reactive precursor T cells from patients, stimulate them to induce numerical expansion, and then re-infuse them. Ex vivo manipulation of the tumor-reactive T cells also permits cytotoxic therapy to be administered to the patient without damaging the effector cells. Recently, host lymphodepletion prior to adoptive transfer of effector T cells has resulted in an extremely high and sustained frequency of effectors that has achieved therapeutic efficacy against bulky metastatic disease in a substantial fraction of treated patients.