GLUCOSES-PHOSPHATE dehydrogeiiase (G6PD) deficiency, the most common enzymopathy of humans, affects over 400 million people¹. The geographical correlation of its distribution with the historical endemicity of malaria suggests that this disorder has risen in frequency through natural selection by malaria^2,3. However, attempts to confirm that G6PD deficiency is protective in case-control studies of malaria have yielded conflicting results^4‡-8. Hence, for this X-linked disorder, it is unclear whether both male hemizygotes and female heterozygotes are protected or, as frequently suggested, only females^1,5‡-11. Furthermore, how much protection may be afforded is unknown. Here we report that, in two large case-control studies of over 2,000 African children, the common African form of G6PD deficiency (G6PD A‡-) is associated with a 46‡-58% reduction in risk of severe malaria for both female heterozygotes and male hemizygotes. A mathematical model incorporating the measured selective advantage against malaria suggests that a counterbalancing selective disadvantage, associated with this enzyme deficiency, has retarded its rise in frequency in malaria-endemic regions. Although G6PD deficiency is now regarded as a generally benign disorder, in earlier environmental conditions it could have been significantly disadvantageous.