A revised model for the interaction of water flow, bed topography, and the rate and mean grain size of bed load in river bends is presented and compared with a large range of observational data. The model represents a modification of the approaches used by F. Engelund, J. S. Bridge, G. Parker, S. Ikeda, and coworkers. Model predictions generally agree with data from the River South Esk, Muddy Creek, River Dommel, and Hooke's laboratory channel. This simple model apparently performs at least as well as more complicated flow models. To make further improvements it is necessary to more accurately specify the interaction between local bed shear stress, sediment transport, and the local bed configuration.