A differentiation block and an accumulation of immature myeloid cells characterize acute myelogenous leukemia (AML). However, native AML cells usually show some morphological signs of differentiation that allow a classification into different subsets, and further differentiation may be induced by exposure to various soluble mediators, for example, all-trans retinoic acid (ATRA) and several cytokines. Combination therapy with ATRA and chemotherapy should now be regarded as the standard treatment of the acute promyelocytic leukemia (APL) variant of AML. Although several agents can also induce leukemic cell differentiation for other AML subgroups, in vitro studies as well as clinical data have demonstrated that these agents often have heterogeneous effects on the leukemic progenitors. This makes the clinical impact of differentiation induction therapy for individual patients difficult to predict. However, differentiation induction should be regarded as a promising therapeutic approach, especially as a part of immunotherapy or in combination with intensive chemotherapy to increase the susceptibility of AML blasts to drug-induced apoptosis. Although the morphology-based French-American-British classification was used to identify APL as an AML subset that required a special treatment, it seems unlikely that this classification alone can be used to identify new subsets of AML patients with special therapeutic requirements. Future studies on differentiation induction in AML should therefore focus on A) the identification of therapeutic agents with more predictable effects; B) the use of clinical and laboratory parameters to define new subsets of AML patients in which differentiation induction has a predictable and beneficial effect, and C) the characterization of how AML blast sensitivity to drug-induced apoptosis is altered by differentiation induction.