Between 1995 and 2000 there were marked changes in the epidemiology of malaria in Ifakara, southern Tanzania. We documented these changes using parasitological and clinical data from a series of community- and hospital-based studies involving children up to the age of 5 years. There was a right shift and lowering in the age-specific parasite prevalence in the community-based cohort studies. The incidence of clinical malaria in placebo-receiving infants in additional study cohorts dropped from 0.8 in 1995 to 0.43 episodes per infant per year in 2000, an incidence rate ratio of 0.53 (95% confidence interval: 0.404, 0.70, P<0.0001). At the same time, there was an increase in the total number of malaria admissions and a marked right shift in the age pattern of these admissions (median age in 1995 1.55 years vs. 2.33 in 2000, P<0.0001). However, the burden of malaria deaths remained in infants. We discuss how these dramatic changes in the epidemiology of malaria may have arisen from the use of currently available malaria control tools. Caution is required in the interpretation of hospital-based data as it is likely to underestimate the impact of anaemia on mortality in the community, where most paediatric deaths occur. Even in low/moderate malaria transmission settings, where older children suffer most malaria episodes, targeting effective malaria control at infants may produce important reductions in infant mortality caused by malaria.