The vertical structure is an important feature of mixed swards, as it influences the inter-species competition for light, as well as the patterns of grazing. Although the experimental measurement of the vertical sward structure is relatively straightforward, the mechanistic and dynamic modelling of the leaf density and lamina density profiles is complicated by the unpredictability of leaf angles. Therefore only static, descriptive profiles have been incorporated in grassland simulation models to date. Two mechanistic models are presented which predict the leaf and lamina density profiles of perennial ryegrass and white clover, based on variables which can readily be measured or produced by simulation models. The `basic model' for grass requires only three input variables: the average leaf length, the average sheath length, and either the total herbage mass or the Leaf Area Index. For clover, only the average petiole length, and either the total weights or areas of petiole and lamina material are required. Use of this basic model is restricted to the simulation of homogeneous swards in which all herbage material shows a random spatial orientation. The `extended model' is more flexible and capable of simulating a wide range of different sward types. However, it requires variables which have not been produced by simulation so far such as the distributions of leaf top heights, sheath top heights, and petiole top heights. Both models were evaluated against a measured profile of a grass–clover sward. The leaf density and lamina density profiles of both perennial ryegrass and white clover, as predicted by the extended model, closely matched the observed profiles. However, the predictions of the basic model were significantly inaccurate for both ryegrass and clover. The extended model can be incorporated as a sub-model into dynamic grassland simulation models in order to enhance the accuracy of simulation of both competition for light and of grazing.