Malaria parasites cause much morbidity and mortality to their human hosts. From our evolutionary perspective, this is because virulence is positively associated with parasite transmission rate. Natural selection therefore drives virulence upwards, but only to the point where the cost to transmission caused by host death begins to outweigh the transmission benefits. In this review, we summarize data from the laboratory rodent malaria model, Plasmodium chabaudi, and field data on the human malaria parasite, P. falciparum, in relation to this virulence trade-off hypothesis. The data from both species show strong positive correlations between asexual multiplication, transmission rate, infection length, morbidity and mortality, and therefore support the underlying assumptions of the hypothesis. Moreover, the P. falciparum data show that expected total lifetime transmission of the parasite is maximized in young children in whom the fitness cost of host mortality balances the fitness benefits of higher transmission rates and slower clearance rates, thus exhibiting the hypothesized virulence trade-off. This evolutionary explanation of virulence appears to accord well with the clinical and molecular explanations of pathogenesis that involve cytoadherence, red cell invasion and immune evasion, although direct evidence of the fitness advantages of these mechanisms is scarce. One implication of this evolutionary view of virulence is that parasite populations are expected to evolve new levels of virulence in response to medical interventions such as vaccines and drugs.