Two experiments conducted in spring and autumn 1992 examined the effect of mollusc grazing on seedling regeneration from natural grassland seedbanks by creating artificial gaps in plots in a grassland sward. Molluscs were excluded from half the gaps by application of molluscicide. Mollusc grazing in both the spring and autumn experiment significantly reduced seedling recruitment, though the intensity of grazing was greatest in autumn. Recruitment of five species was markedly influenced by molluscicide application. In spring, plots from which molluscs were excluded contained significantly more seedlings of Chenopodium polyspermum and Ranunculus acris. In the autumn, exclusion of molluscs resulted in increased populations of R. acris, Stellaria graminea and Rumex acetosa. Cerastium holosteoides populations were greatest in autumn grazed plots. Other species, notably the grasses Holcus lanatus and Agrostis capillaris and the legume Trifolium repens were unaffected by molluscicide application. Species diversity was significantly decreased by molluscicide application in the autumn. Gap size significantly affected the recruitment of two species. Ranunculus acris populations were significantly higher in small gaps in both spring and summer, while Chenopodium recruitment in the spring was greater in small gaps. Gap size also significantly influenced the risk of mollusc attack on Ranunculus as molluscs appeared to show an aggregative feeding response in the high seedling density small gaps. Selective grazing of vulnerable seedlings by molluscs may influence the eventual relative proportions of the species present and so provide a potent mechanism in shaping community composition in grasslands.