A synoptic view is given of the data amassed by WHO, with technical assistance from the Imperial College of Science and Technology, on the origins of blood-meals in Anopheles samples collected from 1971 to 1978. Attention is focused on the proportion of each sample found to contain human blood and on the problems of interpreting from this the human blood index or degree of biting-contact with man exhibited by vector populations. The difficulties of overcoming bias in sampling, which are formidable in unsprayed areas, are further compounded where the dwellings are treated with a slow-acting residual insecticide which knocks down many engorged mosquitoes before they can be collected from their daytime resting places. There is evidence to suggest that the host-selection patterns of those vectors which are ‘opportunistic’ feeders may be heavily influenced, even from village to village or from month to month, by the changing availability of alternative hosts, particularly cattle. This suggests in turn that the possibilities of manipulating the degree of mosquito-man contact by encouraging deflection to animals (zooprophylaxis) or by measures to afford a degree of personal protection should not be under-estimated by malaria strategists. It may sometimes be found less difficult to reduce the vector’s human blood index than it is to measure it, but in view of the epidemiological importance of this parameter, suggestions are put forward for improving entomological field practice in this area. They include a quantitative survey of the biotopes available to the mosquito population as diurnal shelters, a longitudinal survey of the densities of blood-fed females per biotope, and a survey of the numbers and the respective distribution of people and domestic animals available as hosts. The work-load entailed by such a thorough form of investigation, to be repeated where necessary at different seasons of the year, underlines the necessity to concentrate efforts on a small number of localities, carefully chosen for the malaria situations they represent and the vector populations they support. A large-scale blood-meal sampling programme, confined to these selected localities, is most likely in our estimation to yield information of value for controlling malaria vectors.