Summary Insecticide-treated nets have proved successful in the prevention of malaria as a result of both the personal protection with which they provide the sleeper and also the 'mass effect' on the local mosquito population when they are used on a community-wide basis. Personal protection estimates are normally based on comparisons of the numbers of bloodfed mosquitoes found in rooms with and without nets, however it seemed possible that a number of those mosquitoes may not have fed on the occupants of the rooms in which they were found but had entered after feeding elsewhere. To address this possible source of error, we used an 8-locus microsatellite system to identify the source of bloodmeals of Anopheles gambiaes.l. and A. funestus mosquitoes collected in rooms and window traps in Tanzanian villages with and without nets treated with alphacypermethrin. DNA fingerprints were produced from blood samples taken from people who had slept in these rooms and were matched to fingerprints obtained from the mosquito bloodmeals. We were able to type successfully over 90% of the bloodmeals collected and found that proportions of bloodfed mosquitoes that had fed on occupants of the rooms in which they were found were high and only slightly greater in villages without treated nets than those with them (95% and 88%, respectively). When these percentages were used to adjust estimates of personal protection, it was found that the error due to mosquitoes not feeding in the rooms in which they were collected is negligible.