We used a Simrad EK60 echosounder equipped with two split-beam transducers to develop a sampling strategy for assessing fish resources in Tunisian man-made lakes. Day and night surveys, using vertical and horizontal beaming, were carried out between December 2006 and February 2007, a period when fish catchability is high. Four reservoirs with differing surface areas and bathymetries were selected. Echogram analysis revealed that fish communities were mainly composed of individual targets. A few schools were detected near the surface during daylight, but these schools dispersed slightly at night. In these multispecies reservoirs, considerable day and night differences in density existed, but with no clear trend. Target strength (TS) distribution mode values detected at night were always lower or equal to daytime values. Biomass estimates were significantly higher during daytime in three reservoirs, corresponding with higher TS modal values. In the other reservoir, the biomass estimate was significantly higher during nighttime corresponding with higher mean density during this period. Using only a vertically aimed transducer in our study reservoirs would have led to an underestimate of density and biomass by 5-100% and 20-100%, respectively, depending on the man-made lake. We conclude that acoustic sampling in our reservoirs must be done during day and night and that both vertical and horizontal beaming must be used to obtain the best possible picture of the fish stocks.