Three main diffusion-based models are currently used to study grain-size distributions. In this paper, two of these approaches - perfect sorting and imperfect sorting - are compared in a parameter study. First, the numerical solution of the imperfect-sorting model is extensively discussed, and numerical tests are performed. Then, the two sedimentation models are compared for a basin under varying conditions. For some of the imposed variations, predictions of both models differ markedly due to the different approach. The position of the gravel front in the perfect sorting model depends on gravel input and proximal accommodation space. The position of grain-size boundaries in the imperfect-sorting model is strongly controlled by gravel input, the position of the basin axis and the difference in diffusivities. As a result, those two models may predict gravel progradation for different situations. Both models suggest that gravel progradation should always be coupled with sedimentation rates in order to suggest an explanation of gravel progradation observed in the geological record. Simulations with the imperfect sorting model show that this criterion may also fail, showing that a unique interpretation of gravel progradation may be impossible. (C) 2000 Elsevier Science B.V. All rights reserved.