BACKGROUND: It is controversial to what degree α(+)-thalassaemia protects against episodes of uncomplicated malaria and febrile disease due to infections other than Plasmodium. METHODS: In Tanzania, in children aged 6-60 months and height-for-age z-score < -1.5 SD (n = 612), rates of fevers due to malaria and other causes were compared between those with heterozygous or homozygotes α(+)-thalassaemia and those with a normal genotype, using Cox regression models that accounted for multiple events per child. RESULTS: The overall incidence of malaria was 3.0/child-year (1, 572/526 child-years); no differences were found in malaria rates between genotypes (hazard ratios, 95% CI: 0.93, 0.82-1.06 and 0.91, 0.73-1.14 for heterozygotes and homozygotes respectively, adjusted for baseline factors that were predictive for outcome). However, this association strongly depended on age: among children aged 6-17 months, those with α(+)-thalassaemia experienced episodes more frequently than those with a normal genotype (1.30, 1.02-1.65 and 1.15, 0.80-1.65 for heterozygotes and homozygotes respectively), whereas among their peers aged 18-60 months, α(+)-thalassaemia protected against malaria (0.80, 0.68-0.95 and 0.78, 0.60-1.03; p-value for interaction 0.001 and 0.10 for hetero- and homozygotes respectively). No effect was observed on non-malarial febrile episodes. CONCLUSIONS: In this population, the association between α(+)-thalassaemia and malaria depends on age. Our data suggest that protection by α(+)-thalassaemia is conferred by more efficient acquisition of malaria-specific immunity.