The chromosomal translocation t(8;21), generating the AML1-ETO fusion protein, is frequently associated with French-American-British (FAB) type M2 acute myeloid leukemia (AML). t(8;21) fuses the runt domain from the hematopoietic transcription factor RUNX1 with almost the entire transcriptional repressor ETO. AML1-ETO inhibits normal definitive hematopoiesis and blocks erythroid differentiation. Several mechanistic models for the role of AML1-ETO in leukemia development have emerged over the last decade. Most of these models have emphasized the capacity of the fusion protein to redirect repressive cofactors, such as histone deacetylases (HDACs) and DNA methyltransferases (DNMTs), to RUNX target genes, thereby reversing the hematopoietic transcriptional program activated by wild-type RUNX1a phenomenon referred to collectively in this review as the "classical" corepressor model. Because erythropoiesis occurs in a RUNX-independent manner, this dominant-negative "classical" model cannot explain the prominent repression of red-cell development by AML1-ETO. This review will consider the clinical and mechanistic significance of erythroid inhibition by AML1-ETO. Additional models to account for this mysterious oncogenic function are proposed.