Background Malaria remains a major cause of mortality and morbidity in Africa. Many approaches to malaria control involve reducing the chances of infection but little is known of the relations between parasite exposure and the development of effective clinical immunity so the long-term effect of such approaches to control on the pattern and frequency of malaria cannot be predicted. Methods We have prospectively recorded paediatric admissions with severe malaria over three to five years from five discrete communities in The Gambia and Kenya. Demographic analysis of the communities exposed to disease risk allowed the estimation of age-specific rates for severe malaria. Within each community the exposure to Plasmodium falciparum infection was determined through repeated parasitological and serological surveys among children and infants. We used acute respiratory-tract infections (ARI) as a comparison. Findings 3556 malaria admissions were recorded for the five sites. Marked differences were observed in age, clinical spectrum and rates of severe malaria between the five sites. Paradoxically, the risks of severe disease in childhood were lowest among populations with the highest transmission intensities, and the highest disease risks were observed among populations exposed to low-to-moderate intensities of transmission. For severe malaria, for example, admission rates (per 1000 per year) for children up to their 10th birthday were estimated as 3·9, 25·8, 25·9, 16·7, and 18·0 in the five communities; the forces of infection estimated for those communities (new infections per infant per month) were 0·001, 0·034, 0·050, 0·093, and 0·176, respectively. Similar trends were noted for cerebral malaria and for severe malaria anaemia but not for ARI. Mean age of disease decreased with increasing transmission intensity. Interpretation We propose that a critical determinant of life-time disease risk is the ability to develop clinical immunity early in life during a period when other protective mechanisms may operate. In highly endemic areas measures which reduce parasite transmission, and thus immunity, may lead to a change in both the clinical spectrum of severe disease and the overall burden of severe malaria morbidity.