A whole crop computer simulation model of winter wheat has been written in FORTRAN and used to simulate the growth of September- and October-sown crops of Hustler wheat at Rothamsted for the years 1978–9, 1979–80 and 1980–1. Results of the simulations, which are for crops with adequate water and nutrients, are compared with observations from experiments at Rothamsted. The model uses daily maximum and minimum temperatures and daylength to calculate the dates of emergence, double ridge, anthesis and maturity of the crops and the growth and senescence of tillers and leaves. In the simulations, the canopy intercepts daily radiation and produces dry matter that is partitioned between roots, shoots, leaves, ears and grain. Partial simulations, using observed LAI values, produced dry matter in close agreement with observations of late-sown crops, but consistently overestimated the total dry-matter production of the early-sown crops. Full simulation described satisfactorily the average difference in dry-matter production to be expected with changes in time of sowing, but did not give as close correspondence for individual crops. A grain growth submodel, that linked maximum grain weight to average temperatures during the grain growth period, correctly simulated the observed growth of individual grains in the 1981 crop. The benefits to be obtained by combining whole crop modelling with detailed crop observations are discussed.