Four experiments were conducted to facilitate the improvement of degraded grasslands in north Syria. The first examined the survival of legume seeds fed to penned Awassi wethers. The second and third observed the germination and establishment of legumes from faecal pellets in the field and in a glasshouse respectively. The fourth explored the possibility of transporting seeds from legume-rich (source) to legume-poor (target) grasslands using commercial flocks on communally-owned land. Seeds of small-seeded clovers passed through sheep in greater numbers (58–72%) than did seeds of larger-seeded species (10–40%). Of the clover seeds Trifolium campestre (seed size 0.45 mg) disintegrated least (72% passage) and T. haussknechtii (seed size 2.68 mg) disintegrated most (10% passage). Recoveries of Medicago noeana and Scorpiurus muricatus seeds were larger than expected on the basis of their seed sizes. Maximum rate of recovery was at 36 h after the seed meal for all species, all seeds were recovered by 120 h and 90% of the recovered seeds were passed in 72 h. Ingestion had little effect on the hardness and viability of most seeds. Experiments 2 and 3 suggested that seeds in pellets germinated and established as successfully as naked seeds. Burial increased establishment, whether in the field or under conditions of low moisture stress in the glasshouse. The small-seeded clovers, T. tomentosum and T. campestre, established most successfully whether from pellets or from naked seeds. About 500 seeds m−2 were successfully transported from the source to the target grassland. About half were Trigonella monspeliaca, a naturally-occurring legume found on both the source and target grasslands. Of the species sown on the source grassland, 115 seeds m−2 of Trifolium campestre (33% of the source seed bank), 62 seeds m−2 of T. tomentosum (27%) and 78 seeds m−2 of T. purpureum (21%) were detected in the target. Few medics were transported (for example, <2% of M. noeana). Apart from Trigonella monspeliaca, Trifolium tomentosum (16 plants m−2) was the most frequent species found in the target grassland in the following April. The results suggest that using sheep is a cheap and practical way of dispersing the seeds of pasture legumes and thereby improving the degraded grasslands of north Syria. They clearly demonstrate the role of small-seeded legumes (<1 mg) in grassland improvement, especially the small-seeded clovers.