Summary1. In Africa, as in the United States, the great majority of cases of sickle‐cell anaemia are homozygous for the sickle‐cell gene.2. Carriers of the sickle‐cell trait, i.e. individuals heterozygous for the sickle‐cell gene, occur in frequencies up to 40 % in Africa, 30 % in India and 17 % in Greece. It is difficult to explain the distribution of the sickle‐cell trait in terms of ancestral origins alone.3. The incidence of sickle‐cell anaemia is not accurately known, but there is indirect evidence that in Africa, as in the United States, the majority of homozygotes for the sickle‐cell gene are affected. A minority of these homozygotes reaches adult life and reproduces. It is estimated that the fitness of the sickle‐cell homozygote is not more than a quarter that of the heterozygote.4. The elimination of sickle‐cell genes in cases dying of sickle‐cell anaemia is so rapid that it is very unlikely + hat these genes could be replaced by reourrent mutation. It appears that those who are heterozygous for the sickle‐cell gene are naturally resistant to subtertian malaria, so that they have a selective advantage over the normal homozygotes in regions where malaria is hyperendemic. The sickle‐cell polymorphism would then be the result of a balance determined mainly by the severity of subtertian malaria, which would tend to increase the frequency of the gene, and the elimination of genes in cases of sickle‐cell anaemia. The sickle‐cell trait attains high frequencies, in fact, only in regions where malaria is severe.5. It is estimated that the fitness of the sickle‐cell heterozygote in malarious environments is 1–26 X the fitness of the normal homozygote.6. The expected rate of change of the sickle‐cell carrier frequency in populations under different conditions has been calculated. In two instances such an estimate corresponds approximately with the observed data. It seems that the incidence of the sickle‐cell trait may be greater in adults than in young children in populations exposed to malaria, as expected. And the United States Negroes have a lower frequency of the sickle‐cell trait than would be expected on the basis of a one‐third admixture with Caucasian and Indian blood. The fall in frequency would correspond to an elimination of three‐quarters of the homozygotes without heterosis for