Anopheline mosquitoes were sampled inside houses, where residents slept under untreated bednets, by CDC light traps and human-biting catches to evaluate light traps as a means for determining human exposure to malaria vectors in Kilifi District, Kenya. Mosquitoes were sampled during 2 all-night collections by light traps and one all-night biting catch in a series of 262 houses. Collections yielded 1,721 Anopheles gambiae s.l. and 46 An. funestus, and 60.3% of the houses were negative for anophelines. There was a significant correlation in numbers of An. gambiae s.l. captured by light traps and human-biting collections (r = 0.64), but light traps were biased and underestimated An. gambiae s.l. abundance. This bias increased with increasing mosquito abundance. In addition, the proportion of An. gambiae s.l. infected by Plasmodium falciparum was 2.3-fold higher in light traps than in human-biting collections. Along the coastal zone of Kenya where vector abundance is low, light traps do not provide an adequate estimate of man-vector contact when such information is required at the household level in epidemiological studies of malaria parasite transmission.