The genome sequence of Plasmodium falciparum, the causative agent of the most severe form of malaria in humans, rapidly approaches completion, but our ability to genetically manipulate this organism remains limited. Chromosomal integration has only been achieved following the prolonged maintenance of circularised episomal plasmids which selects for single crossover recombinants. It has not been possible to construct genetic deletions via double crossover recombination, presumably due to the low frequency of this event. We have used the Herpes simplex virus thymidine kinase gene and the Escherichia coli cytosine deaminase gene for negative selection of P. falciparum. Parasites were transformed with plasmids expressing the thymidine kinase and cytosine deaminase genes by positive selection for the human dihydrofolate reductase gene. Parasites expressing thymidine kinase are susceptible to the pro-drug ganciclovir while those expressing cytosine deaminase are sensitive to 5-fluorocytosine. Parental parasites were inherently resistant to these drugs. A significant 'bystander effect' was evident in cultures with either ganciclovir or 5-fluorocytosine. Positive and negative selection of the thymidine kinase transformants with both ganciclovir and WR99210 resulted in the selection of parasites containing a genetic deletion of the Pfrh3 gene, the first targeted double crossover deletions in P. falciparum. The use of negative selection for gene disruptions via double crossover recombination will dramatically improve our ability to analyse protein function and opens the possibility of using this strategy for a variety of gene deletion and modification experiments in the analysis of this important infectious agent.