Despite its key role in determining the stability and intensity of malaria transmission, the infectiousness of human populations to mosquitoes has rarely been estimated. Field-based analyses of malaria transmission have frequently relied on the prevalence of asexual parasites or gametocytes as proxies for infectiousness. We now summarize empirical data on human infectiousness from Africa and Papua New Guinea. Over a wide range of transmission intensities there is little relationship between the infectiousness of human populations to vector mosquitoes and mosquito-to-human transmission intensity. We compare these data with the predictions of a stochastic simulation model of Plasmodium falciparum epidemiology. This model predicted little variation in the infectiousness of the human population for entomologic inoculation rates (EIRs) greater than approximately 10 infectious bites per year, demonstrating that the lack of relationship between the EIR and the infectious reservoir can be explained without invoking any effects of acquired transmission-blocking immunity. The near absence of field data from areas with an EIR < 10 per year precluded validation of the model predictions for low EIR values. These results suggest that interventions reducing mosquito-to-human transmission will have little or no effect on human infectiousness at the levels of transmission found in most rural areas of sub-Saharan Africa. Unless very large reductions in transmission can be achieved, measures to prevent mosquito-to-human transmission need to be complemented with interventions that reduce the density or infectiousness of blood stage parasites.