An understanding of the mechanism that controls growth and differentiation in normal cells would seem to be an essential requirement to elucidate the origin and reversibility of malignancy. For this approach I have mainly used normal and leukemic blood cells, and in most studies have used myeloid blood cells as a model system. Our development of systems for the in vitro cloning and clonal differentiation of normal blood cells made it possible to study the controls that regulate growth (multiplication) and differentiation of these normal cells and the changes in these controls in leukemia. Experiments with normal blood cell precursors have shown that normal cells require different proteins to induce growth and differentiation. We have also shown that in normal myeloid precursors, growth-inducing protein induces both growth and production of differentiation-inducing protein so this ensures the coupling between growth and differentiation that occurs in normal development. The origin of malignancy involves uncoupling of growth and differentiation. This can be produced by changes from inducible to constitutive expression of specific genes that result in asynchrony to the coordination required for the normal developmental program. Normal myeloid precursors require an external source of growth-inducing protein for growth, and we have identified different types of leukemic cells. Some no longer require and other constitutively produce their own growth-inducing protein. But addition of the normal differentiation-inducing protein to these malignant cells still induces their normal differentiation, and the mature cells are then no longer malignant. Genetic changes that produce blocks in the ability to be induced to differentiate by the normal inducer occur in the evolution of leukemia. But even these cells can be induced to differentiate by other compounds, including low doses of compounds now being used in cancer therapy, that induce the differentiation program by other pathways. This differentiation of leukemic cells has been obtained in vitro and in vivo, and our in vivo results indicate that this may be a useful approach to therapy. In some tumours, such as sarcomas, reversion from a malignant to a non-malignant phenotype can be a result of chromosome changes that suppress malignancy. But in myeloid leukemia, the stopping of growth in mature cells by induction of differentiation bypasses the genetic changes that produce the malignant phenotype. These conclusions can also be applied to other types of normal and malignant cells.