Fish screens can help prevent the entrainment or injury of fish at irrigation diversions, but only when designed appropriately. Design criteria cannot simply be transferred between sites or pump systems and need to be developed using an evidence-based approach with the needs of local species in mind. Laboratory testing is typically used to quantify fish responses at intake screens, but often limits the number of species that can studied and creates artificial conditions not directly applicable to screens in the wild. In this study a field-based approach was used to assess the appropriateness of different screen design attributes for the protection of a lowland river fish assemblage at an experimental irrigation pump. Direct netting of entrained fish was used along with sonar technology to quantify the probability of screen contact for a Murray-Darling Basin (Australia) fish species. Two approach velocities (0.1 and 0.5 m.sec(-1)) and different sizes of woven mesh (5, 10 and 20 mm) were evaluated. Smaller fish (