Immunotherapy of virus induced tumours by adoptive transfer of virus specific cytotoxic T cells (CTL) is now feasible in experimental murine systems. These CTL recognize viral peptide sequences of defined length presented in the groove of MHC class I molecules. Effective eradication of large tumour masses requires coadministration of IL-2. In essence, T cell immunity against virus induced tumours does not differ from anti-viral T cell immunity in general. Tumour escape strategies are numerous but, in various instances, can be counteracted by defined measures. Initiation of CTL responses against poorly immunogenic non-virus induced tumours (the majority of human cancer) requires novel strategies to overcome T cell inertia. Rather than waiting to see whether tumour specific CTL (against unknown antigens) can be cultured from TIL, we propose an alternative strategy in which CTL are raised against target molecules of choice, including differentiation antigens of restricted tissue distribution (autoantigens) or mutated/overexpressed oncogene products. The various steps proposed include: (a) identification of target molecules of choice; (b) identification in these target molecules of MHC allele specific peptide motifs involved in peptide binding to MHC molecules; (c) evaluation of actual binding of such peptides to specific MHC class I molecules; (d) in vitro CTL response induction by such peptides, presented either by highly efficient antigen presenting cells (such as processing defective cells, which carry empty MHC class I molecules) loaded with a single peptide or by dendritic cells, both cell types being capable of primary CTL response induction in vitro and (e) adoptive transfer of tumour specific CTL generated in vivo or, more conveniently, vaccination with immunodominant peptides. The latter possibility seems to be feasible because peptide vaccination with a single immunodominant viral peptide can install CTL memory and confer protection against lethal virus infection.