A 4-year entomological study was carried out in Southern Madagascar to identify malaria vectors, evaluate the transmission and compare the influence of irrigation in the sub-arid and adjacent humid regions. Three villages were involved in this entomological survey: Androvasoa (located in the natural sub-arid ecosystem), Pépiniére (sited at the centre of an irrigated rice scheme in the sub-arid region) and Esana (bordered with rice fields in the humid region). Mosquitoes were collected inside and outside dwellings when landing on human beings, with light traps and with knockdown indoor sprays. Anopheles arabiensis, Anopheles merus and Anopheles funestus were found in every village while Anopheles gambiae s.s. was only found in the village bordering the rice-fields (Pépiniére) and Anopheles mascarensis, a mosquito native to Madagascar, was only found in the humid region (Esana). In Pépiniére, the annual entomological inoculation rate (EIR) was low (EIR=0.4 infective bites/man/year (IBM)). In the irrigated scheme of the sub-arid region, malaria transmission was 150 times higher (mean EIR=63 IBM) than in the natural ecosystem and A. funestus was the main vector, responsible for 90% of infective bites. In Esana, the level of malaria transmission was high (EIR=41 IBM) and 2/3 of the infective bites were due to A. mascarensis, despite the presence of A. gambiae s.s. and A. funestus. These results are discussed with reference to the impact of irrigation on malaria in Africa (“the paddies paradox”).