The SHETRAN physically based, spatially distributed model is used to investigate the scaling relationship linking specific sediment yield to river basin area, for two contrasting topographies of upland and more homogeneous terrain and as a function of sediment source, land use and rainfall distribution. Modelling enables the effects of the controls to be examined on a systematic basis, while avoiding the difficulties associated with the use of field data (which include limited data, lack of measurements for nested basins and inability to isolate the effects of individual controls). Conventionally sediment yield is held to decrease as basin area increases, as the river network becomes more remote from the headwater sediment sources (an inverse relationship). However, recent studies have reported the opposite variation, depending on the river basin characteristics. The simulation results are consistent with these studies. If the sediment is supplied solely from hillslope erosion (no channel bank erosion) then, with uniform land use, sediment yield either decreases or is constant as area increases. The downstream decrease is accentuated if rainfall (and thence erosion) is higher in the headwaters than at lower elevations. Introducing a non-uniform land use (e.g. forest at higher elevations, wheat at lower elevations) can reverse the trend, so that sediment yield increases downstream. If the sediment is supplied solely from bank erosion (no hillslope erosion), the sediment yield increases downstream for all conditions. The sediment yield/basin area relationship can thus be inverse or direct, depending on basin characteristics. There still remains, therefore, considerable scope for defining a universal scaling law for sediment yield.