Background and Aims In Horqin sandy land, vegetation degradation usually occurs in improper management regimes under fragile conditions, e.g. clearing and grazing. However, few information is available about the effects of grazing and enclosing on the structure and performance of soil seed banks in the semi-arid desert environment. Therefore, a field experiment was conducted in grazed and enclosed grasslands to examine the structure and performance of soil seed banks. Methods In late March 2003, three parallel 200-m line transects (20 m apart) were established in two experimental sites, grazed and enclosed grasslands. For each transect, 20 sampling points were set up at 10-m intervals, and a soil sample of 5 cm in depth and 20 cm × 20 cm in dimension was collected from each sampling point. All soil samples were transported to the laboratory in open plastic bags and chopped and sieved (mess width is 0.2 mm) soon. The sieved soil samples were placed in plastic germination pots (33 cm in diameter and 12 cm in depth) and spread evenly to form an approximately 1 cm thick layer in individual pots. Pots were placed in an unheated greenhouse for seed germination. Pots were watered daily with a very fine nozzle in the afternoon. Emergent seedlings were identified to species and carefully removed from the pots. Seed germination and seedling identification were carried out continuely in the following three months. However, some unidentifiable seedlings remained longer in the pots until they were identified. The density of the existing seed bank was expressed as the number of viable seeds per square meter. At each site, the frequency of species in the soil seed bank was determined in terms of the 60 sampling points or the 60 quadrats. Key Results The study was conducted in grazed and enclosed grasslands to examine the relationship between density and composition of soil seed bank and vegetation in Horqin sandy land, Inner Mongolia. The results showed that there were 22 and 30 plant species in seed banks in the grazed and enclosed grasslands, respectively. In the grazed and enclosed grasslands, the seed bank density were 20 657±3 342 and 16 149±1 900 viable seeds·m -2 , respectively, with annual herbaceous plants dominating in both sites (occupied 99% and 98%, respectively) Shannon-Wiener index and richness of the grazed grassland were 0.836 3 and 4.954 9, respectively, which were distinctly lower than those in the enclosed grassland (0.968 2 and 7.226 0), suggesting negative impacts of grazing on seed bank diversity. Seed bank density strongly related to the standing vegetation density in both grazed and enclosed grasslands (p0.001), indicating the standing vegetation density increased with the increase of seed bank density. Differences in seed bank density accounted for 78% and 58% of the variance of standing vegetation density in grazed and enclosed communities, respectively. Conclusions Our results showed that adoption of enclosure management practice significantly increased density and species diversity of the seed bank, compared with the grazed grassland. Therefore, reseeding, enclosure and other management steps should be used to speed up the restoration process of the degraded grassland.