Competition between the annual legume Stylosanthes humilis and the annual grass weed Digitaria ciliaris, between the perennial legume Stylosanthes hamata and Digitaria ciliaris, and between Stylosanthes hamata and the perennial pasture grass Urochloa mosambicensis was studied at Katherine, N.T., by the replacement method. All species were mutually exclusive in mixtures. In pure stands S. hamata yielded significantly more than the three other species, and its competition with Digitaria was significantly stronger than that for S. humilis in competition with Digitaria (as measured by the relative crowding coefficient, k). Changes in species proportion in mixed swards were measured for two seasons in three phases of the annual life cycle: (1) germination–establishment, (2) main growth period, and (3) seed production, and were expressed by a 'phase parameter' analogous to the relative reproductive rate commonly used in competition experiments. When the legume constituted 75% of the viable seed sown in the first year, the legume proportion in terms of dry matter at the end of the second season was 0.02 for S. humilis–Digitaria, 0.35 for S. hamata–Urochloa and 0.80 for S. hamata–Digitaria. These results are discussed with reference to the phase parameters. Simulation, with variation of the phase parameters assumed log normal and independent within the observed ranges, indicated that under climatic conditions resulting in grass dominance in undefoliated S. humilis–Digitaria mixtures, legume dominance could be expected in undefoliated, high fertility S. hamata–Digitaria mixtures.