Habitat preferences
The distribution of all the captures of the seven main species of the community in the eight room types is given in Table 2. Small mammal species appear not to be randomly distributed in the room categories defined (khi-2 = 401.34; 42df; p = 3.62 10-60). The distribution of M. musculus appears as the most divergent from random expectations, the species being clearly over-represented in kitchens (and to a lesser extent in stores) and under-represented in outdoors, granaries, and stock rooms. The distribution of the room types and their modalities (nature of floor, walls and ceiling) across the localities sampled did not show any particular trend, as evidenced by the results of fCA (Fig. S3). An overall opposition between more urbanized (Petite Côte, Tambacounda) vs . more rural (Boundou) areas appears however, associated with a dominance of distinct room type (workshops and non-food stores vs . granaries) or construction materials (concrete and metal vs . adobe and straw). From there, a k+1 analysis was performed between the cPCA of small mammals and the fCA of room types / modalities grouped by geographic areas. This K+1 C1-C2 factorial map illustrates graphically (Fig. 2) the relationships between species, geographic areas and room types and characteristics. The most visible associations are, on axis 1, between M. musculus , kitchens and concrete walls, mainly on the Petite Côte, the Kaolack-Tambacounda axis, and in Tambacounda (positive side of C1). On the opposite (negative) side of C1 are all the other small mammals,M. natalensis excepted, stock rooms and adobe walls (and to a lesser extent presence of large food stocks and straw ceiling), in Boundou, Basse Casamance and the Tambacouda – Kidira axis, to a lesser extent. C2 mainly contrasts M. natalensis associated with average levels of food and concrete floor in the area of Kédougou vs .R. rattus and M. erythroleucus related to non-food stores and outdoors, absence of stocks, concrete floor and ceiling, and metal sheet walls.