Mosquitoes were also collected while landing on two persons. The same persons were employed in site B and C, whereas one of the two persons was replaced in site A. One dog- and one man-bait trap were employed simultaneously in the three sites for 17 sampling nights, i.e., for six nights in site A (August 5-September 15, 1997), six nights in site B (July 29-September 23, 1998), and five nights in site C (June 23-September 15, 1999). The traps consisted of a cylindrical structure of wood (2.3 m of diameter × 2.0 m in height for dog and 2.0 m × 1.5 m for humans) covered by a net, five cm above the ground to allow mosquitoes to enter the trap [12,13] (Figure 1). The risk of mosquitoes escaping from the trap was considered negligible, due to both the insect host-seeking behaviour (in which mosquitoes rarely fly downwards when leaving an enclosed space), and to the fact that engorged females do not move after the blood meal, resting on the net [14]. In each site, the human and dog baits were located at a distance of at least ten meters to avoid interference between mosquito attractants.