The rationale for exchange transfusion (ET) in severe Plasmodium falciparum malaria is to lower the parasite burden rapidly, to replenish unparasitized cells and to correct severe anaemia. In addition, parasite antigens or 'toxins' may be removed. Despite reports of successful ET in the treatment of malaria since 1974, it remains controversial and its role and technique have been poorly defined. We present a mathematical model of ET which relates volume of exchange to reduction in parasitaemia and change in haemoglobin concentration. This model fits published and unpublished clinical data well, and should facilitate standardization of ET in future.