The feasibility of a malaria vaccine is supported by the fact that children in endemic areas develop naturally acquired immunity to disease. Development of disease immunity is characterized by a decrease in the frequency and severity of disease episodes over several years despite almost continuous infection1, suggesting that immunity may develop through the acquisition of a repertoire of specific, protective antibodies directed against polymorphic target antigens1–3. Plasmodium falciparum erythro-cyte membrane protein 1 (PfEMPI) is a potentially important family of target antigens, because these proteins are inserted into the red cell surface and are prominently exposed4–6 and because they are highly polymorphic and undergo clonal antigenic variation7,8,18, a mechanism of immune evasion maintained by a large family of var genes9–11. In a large prospective study of Kenyan children, we have used the fact that anti-PfEMP1 antibodies agglutinate infected erythrocytes in a variant-specific manner10,12–16, to show that the PfEMPI variants expressed during episodes of clinical malaria were less likely to be recognized by the corresponding child's own preexisting antibody response than by that of children of the same age from the same community. In contrast, a heterologous parasite isolate was just as likely to be recognized. The apparent selective pressure exerted by established anti-PfEMPl antibodies on infecting parasites supports the idea that such responses provide variant-specific protection against disease.