Abstract. The effects of grazing upon the composition of a grassland were examined in a replicated grazing experiment with sheep over eight years. Grazing was controlled to produce two levels of grazing in each of winter, spring and summer, combined in a 2×2×2 factorial experimental design. Particular species were either increased or decreased by winter grazing while, in general, dicotyledonous species were favoured by grazing in both winter and spring. A model fitting the main effects of grazing in three seasons showed the importance of grazing in winter and spring to achieve sward diversification.