1. Contrasting patterns of change in the seed bank of natural grasslands are frequently found in response to grazing by domestic herbivores. Here, we studied the hypotheses that a) patterns of change in seed bank density and composition in response to grazing depend on spatial variation in resource availability and productivity, and b) that variation among species in patterns of seed bank response to grazing is linked to differences in species size traits (i.e. size of plant, dispersal unit and seed).

2. Effects of sheep grazing exclusion on the seed bank were followed during five years in a semiarid Mediterranean annual plant community in Israel. Seed bank density and composition were measured in autumn, before the rainy season, inside and outside fenced exclosures in four neighboring topographic sites differing in vegetation characteristics, soil resources and primary productivity: Wadi (dry stream terraces, high productive site), Hilltop, South- and North-facing slopes (less productive sites).

3. Topographic sites differed in seed density (range ca 2500–18000 seed m−2) and in seed bank response to grazing exclusion. Fencing increased seed density by 78, 51 and 18% in the Wadi, South- and North-facing slopes, respectively, but had no effect in the Hilltop. At the species level, grazing exclusion interacted with site conditions in determining species seed bank density, with larger or opposite changes in the high productive Wadi compared to the other less productive sites.

4. Changes in seed bank structure after grazing exclusion were strongly related to species size traits. Grazing exclusion favored species with large size traits in all sites, while seed density of tiny species decreased strongly in the high productive Wadi. Species with medium and small size traits showed lesser or no responses.

5. The size of plants, dispersal units and seeds were strongly correlated to each other, thus confounding the evaluation of the relative importance of each trait in the response of species to grazing and site conditions. We propose that the relative importance of plant size vs seed size in the response to grazing changes with productivity level.