Seeds of two grass (Arrhenatherum elatius and Festuca rubra), two herb (Plantago lanceolata and Rumex acetosa) and two legume (Lotus corniculatus and Trifolium repens) species were sown in summer 1995 at four densities (no seed, 1000, 10 000 and 50 000 seeds m−2) into an established rabbit-grazed grassland given factorial combinations of rabbit fencing (with and without fences) and soil disturbance (with and without cultivation). On plots where no seeds were sown, only the species with persistent seed banks (P. lanceolata, L. corniculatus and T. repens) showed enhanced seedling emergence in response to disturbance. In disturbed soil, seedling densities of all species increased with increasing density of sown seeds, the effects of which were still evident for plant cover 2 years after seed sowing. In undisturbed vegetation, A. elatius, F. rubra, P. lanceolata and R. acetosa showed increased seedling densities following seed sowing; but in each case, there was an upper asymptote to seedling recruitment, presumably due to microsite limitation. Rabbit grazing reduced seedling densities, with this reduction being more pronounced with disturbance than without. However, the effect of rabbit grazing did not persist for some species; seedling mortality of R. acetosa, P. lanceolata, L. corniculatus and T. repens was higher on plots without rabbit grazing, so that plant densities of these species in summer 1996 and plant cover in summer 1997 were greater on grazed plots. The results indicate interactions between soil disturbance, propagule availability and herbivory, rather than disturbance alone, will play an important role in controlling seedling recruitment and species habitat distributions in grasslands.