Mosquito nets treated with permethrin, deltamethrin or alpha-cypermethrin at 25 mg/m(2) were evaluated in experimental huts in an area of rice irrigation near Moshi, in northern Tanzania. The nets were deliberately holed to resemble worn nets. The nets treated with permethrin offered the highest personal protection against Anopheles arabiensis (61.6% reduction in fed mosquitoes) and Culex quinquefasciatus (25.0%). Deltamethrin and alpha-cypermethrin provided lower personal protection against An. arabiensis (46.4% and 45.6%, respectively) and no such protection against Cx. quinquefasciatus. Permethrin performed poorly in terms of mosquito mortality, however, killing only 15.2% of the An. arabiensis and 9.2% of the Cx. quinquefasciatus exposed to the nets treated with this pyrethroid (after correcting for control mortality). The alpha-cypermethrin and deltamethrin performed marginally better, with respective mortalities of 32.8% and 33.0% for An. arabiensis and 19.4% and 18.9% for Cx quinquefasciatus. The poor killing effect of permethrin was confirmed in a second trial where a commercial, long-lasting insecticidal net based on this pyrethroid (Olyset) produced low mortalities in both An. arabiensis (11.8%) and Cx. quinquefasciatus (3.6%). Anopheles arabiensis survivors collected from the verandahs of the experimental huts and tested on 0.75%-permethrin and 0.05%-deltamethrin papers, in World Health Organization susceptibility kits, showed mortalities of 96% and 100%, respectively. The continued use of permethrin-treated nets is recommended for personal protection against An. arabiensis. In control programmes that aim to interrupt transmission of pathogens by mosquitoes and/or manage pyrethroid resistance in such vectors, a combination of a pyrethroid and another insecticide with greater killing effect should be considered.