To characterize red cell susceptibility to invasion in malaria, a selectivity index (SI) was calculated as the ratio of observed number of multiple-infected red cells to that expected from a random process (Poisson distribution). In patients with falciparum malaria (n = 100) SI decreased with increasing parasitaemia (P < 0.001), and correlated inversely with plasma lactate concentrations, chosen prospectively as a measure of disease severity (r = -0.36, P < 0.001). For parasitaemias < 5%, the SI was lower in patients with severe malaria (geometric mean 1.35; 95% confidence interval 1.01-1.80) than in uncomplicated malaria (2.31; 1.89-2.81; P = 0.003), despite similar parasite counts. The geometric mean (range) SI in vivax malaria (n = 20), 7.69 (1.67, 29.75), was significantly greater than that in falciparum malaria at comparable parasitaemias (< or = 2%), 2.44 (0.45, 14.05), P < 0.001, suggesting that about 13% of circulating erythrocytes were susceptible to invasion by Plasmodium vivax. This translates into susceptibility for about 2 weeks after emergence from the bone marrow, if age is the sole determinant of this process. In falciparum malaria selectivity was inversely proportional to severity; lack of selectivity could reflect either a 'favourable' host red cell phenotype, or an indiscriminate parasite population. Both are dangerous for the host.