Ethnobotanical survey in 2 communities in western Kenya revealed that the most commonly known repellent plants were Ocimum americanum L. (64.1%), Lantana camara L. (17.9%), Tagetes minuta L. (11.3%) and Azadirachta indica A. Juss (8.7%) on Rusinga Island, and Hyptis suaveolens Poit. (49.2%), L. camara (30.9%) and O. basilicum L. (30.4%) in Rambira. Direct burning of plants is the most common method of application for O. americanum (68.8%), L. camara (100%) and O. basilicum (58.8%). Placing branches or whole plants inside houses is most common for H. suaveolens (33.3 and 57.8% for the respective locations), A. indica (66.7 and 100%), and T. minuta (54.8 and 56.0%). The repellency of plants suggested by the ethnobotanical survey and other empirical information was evaluated against the malaria vector Anopheles gambiae s.s. Giles in experimental huts within a screenwalled greenhouse. Thermal expulsion and direct burning were tested as alternative application methods for the selected plants O. americanum, O. kilimandscharicum Guerke, O. suave Willd., L. camara, A. indica, H. suaveolens, Lippia uckambensis Spreng and Corymbia citriodora Hook. When thermally expelled, only H. suaveolens failed to repel mosquitoes, whereas the leaves of C. citriodora (74.5%, P < 0.0001), leaves and seeds of O. suave (53.1%, P < 0.0001) and O. kilimandscharicum (52.0%, P < 0.0001) were the most effective. Leaves of C. citriodora also exhibited the highest repellency (51.3%, P < 0.0001) by direct burning, followed by leaves of L. uckambensis (33.4%, P = 0.0004) and leaves and seeds of O. suave (28.0%, P = 0.0255). The combination of O. kilimandscharicum with L. uckambensis repelled 54.8% of mosquitoes (P < 0.0001) by thermal expulsion. No combination of plants increased repellency by either method. The semi-field system described appears a promising alternative to full-field trials for screening large numbers of candidate repellents without risk of malaria exposure.