Effects of bednet coverage (C) on prevalence of malaria were analysed using data from 1990-92 from 9 Papua New Guinean villages. Effects of coverage varied by age, resulting in a shift in age of peak prevalence from 4.7 (C = 0%) to 11.6 (C = 100%) years for Plasmodium falciparum, from 3.4 to 4.9 years for P. vivax and from 11.0 to 16.8 years for P. malariae. In small areas with no bednets the age distribution of P. falciparum parasitaemia was like that of a holoendemic area. Where coverage was complete the pattern corresponded to mesoendemicity. Thus, protracted use of bednets can result in profound changes in the endemicity of malaria even when coverage is incomplete and without insecticide treatment. Average entomological inoculation rates (EIRs) estimated from indoor landing rates on individuals without bednets were 35, 12 and 10 infectious bites per person per annum for P. falciparum, P. vivax and P. malariae, respectively. Logistic regression analyses indicated that the EIR estimate for P. falciparum was related to prevalence of this species independently of effects of bednet coverage. However, the recent EIR still accounted for much less variation than did the bednets. A similar pattern was seen for P. malariae, while there were no significant relationships between the recent EIR and the parasite positivity for P. vivax. It is concluded that short-term variations in inoculation rate are not important determinants of parasite prevalence in this population.