A prospective study was conducted to measure the selective effect of pyrimethamine prophylaxis on point mutations in Plasmodium falciparum dihydrofolate reductase (DHFR). A total of 109 Malian children were given pyrimethamine weekly for 5 weeks. P. falciparum infections were analyzed by polymerase chain reaction for DHFR mutations, which were dramatically more frequent among prophylaxis-breakthrough infections than at baseline: the prevalence of Asn-108 rose from 13% to 100%, Ile-51 from 4% to 50%, and Arg-59 from 11% to 90%. Eight persistent infections lacking detectable DHFR mutations at baseline developed multiple mutations within 1 week of the patients' starting pyrimethamine prophylaxis. Microsatellite analysis found no evidence of clonal identity among baseline and breakthrough infections. Analysis of these data demonstrates that under prophylaxis conditions, pyrimethamine is strongly selective for DHFR mutations, which arise extremely rapidly under drug pressure, even when undetectable in the initial infection. These findings have implications for prophylaxis regimens with other antifolate drugs.