The consumption and preferences of polyphagous ground beetles (Coleoptera: Carabidae) for the seeds of herbaceous plants was determined. The seeds were stuck into plasticine in small tin trays and exposed to beetle predation on surface of the ground. In the laboratory the effect of carabid (species, satiation) and seed (species, size) on the intensity of seed predation was investigated. The consumption of the generally preferred Cirsium arvense seed by 23 species of common carabids increased with body size. Seed of Capsella bursa-pastoris was preferred by small carabids and their consumption rates were not related to their size. The average daily consumption of all the carabid species tested (0.33 mg seeds . mg body mass-1 . day-1) was essentially the same for both kinds of seed. Because of satiation the consumption of seed of C. arvense provided ad libitum to Pseudoophonus rufipes decreased over a period of 9 days to 1/3-1/4 of the initial consumption rate. Preferences of P. rufipes (body mass 29.6 mg) and Harpalus affinis (13.4 mg) for the seeds of 64 species of herbaceous plants were determined. The small H. affinis preferred smaller seed than the large P. rufipes. Predation of seed present on the ground in the field was studied in 1999-2000, at Praha-Ruzyne (50°06´ N 14°16´E). Seeds were placed in stands of different crops as in the laboratory experiments and vertebrate predation was excluded by wire mesh cages. Pitfall traps placed near the cages revealed that carabids were the only seed predators active in the area. Rates of removal of seed of 6 weed species varied with crop, season, seed and site. Average rate of removal in June-August was 2.5 seeds.day-1.tray-1 and was smaller before and after this period. The rates of removal increased with increasing activity density of the carabids and paucity of seed from naturally occurring weeds, which may have satiated the carabids. In stands of winter wheat, millet and soybeans there were significant differences in the rates of removal of the seed of 43 herbaceous species. The field preferences were correlated with those established in the laboratory. Predation of seed on the ground in arable fields can be as high as 1000 seeds.m-2.day-1 and may selectively influence the quantity of seed of particular herb species that enters the soil seed bank. Seed predation thus may be an effective component of weed control on arable land, particularly at low weed densities.