We investigated local-scale variation in malaria transmission and infection in children within a continuous landscape by retrospective spatial analysis of entomological and clinical data collected during 1988 and 1989 in The Gambia, West Africa. Parasite prevalence was negatively correlated with vector abundance and exposure to malaria parasites in 10 villages where entomological surveillance had been carried out. Variation in bednet use did not explain this finding. Mosquito-breeding habitat was retrospectively mapped using 20-m spatial resolution multispectral SPOT satellite imagery from 1988. From these data we estimated by linear regression the risk of exposure to malaria parasites in 26 villages where clinical surveys of children had been made. As exposure increased, so did parasite prevalence; but at higher levels of exposure, parasite prevalence declined. Our findings demonstrate marked differences in exposure to malaria in villages over distances of less than 2 km from mosquito breeding sites and suggest that there are also large differences in immunity between neighbouring settlements.