Both clinical and experimental evidence indicate that T lymphocytes can mediate antileukemic effects in acute myelogenous leukemia (AML). These antileukemic effects can be either nonspecific cytotoxicity (killer cell activity) or reactivity against leukemia-specific antigenic peptides presented by self-HLA molecules. The antigen-specific T cell activation requires recognition of specific peptides together with costimulatory signalling. For most patients the AML blasts express both HLA class I and class II molecules for antigenic presentation, but patients are heterogeneous with regard to: (1) expression of costimulatory binding molecules; (2) expression of receptors/counterreceptors involved in induction of apoptosis; (3) constitutive release of immunomodulatory soluble mediators. This heterogeneity suggests that the ability of AML blasts to initiate an antileukemic T cell response will differ between individual patients. Thus, clinical approaches for immunotherapy in AML have to overcome three major problems. First, the therapy should reduce the patient heterogeneity so that therapeutic effects become more predictable; or alternatively one should define patient subsets which are likely to benefit from immunotherapy. Second, immunotherapy should enhance antileukemic T cell reactivity or blast susceptibility to immune attacks. Third, the therapeutic procedures must be safe and suitable for routine use. All three problems probably have to be solved before immunotherapy can become a routine treatment.