A significant and stable difference in the residual night-biting collection (NBC, observed/expected) of A. gambiae s.l. was observed among different villages of the West African savannah when sprayed for 2 years with propoxur. The residual mosquito density of a given village was positively associated with some of its pre-spraying characteristics: the NBC/PSC (pyrethrum spray collections) ratio, the median biting hour, and possibly the proportion of species B. It was not significantly associated with several other pre-spraying characteristics (absolute mosquito density, the ratios between NBC indoors and NBC outdoors, between ETC (exit-trap collections) and PSC, between fed and gravid in the PSC or ETC, between males and females in the PSC), and was not associated with variations in recorded coverage, with latitude, or with distance from unsprayed villages.In a comparison between A. gambiae s.l., A. funestus, and A. pharoensis, the residual NBC (observed/expected) was positively associated with the pre-spraying NBC/PSC ratio, and was negatively associated with the median biting hour.The pre-spraying NBC/PSC ratio thus appears to be a predictor of the variation, between villages or species, in the reduction of the NBC by residual spraying, and may assist in forecasting the effect of a residual insecticide. The limitations of any absolute prediction must, however, be kept in mind.