Weed seed consumption experiments involving comparison of rates of seed loss by seed feeders were conducted over a five-week period in low-input (no insecticide, low herbicide usage) conventional- and no-tillage soybean agroecosystems. Seeds of four broadleaf weed species (ragweed [Ambrosia artemisiifolia L.], pigweed [Amaranthus retroflexus L.], sicklepod [Cassia obtusifolia L.], and jimsonweed [Datura stramonium L.]) and one grain crop species (wheat [Triticum aestivum L.]) were provided in a free choice design at densities of 10, 25, and 50 seeds/24 cm3. Approximately 2.3 times more seeds overall, and 1.4 times more large seeds as a group were consumed in notillage systems than in conventional-tillage systems. In our experimental low-input, notillage treatments, large ground beetles (Carabidae: Coleoptera) (15–25 mm) and mice preferentially fed on the larger seed species, while small carabids (< 15 mm), ants and crickets, fed almost exclusively on the smaller seed species. Carabid beetles were responsible for more than half of all seeds consumed. Laboratory and field studies indicated that ground beetles selectively consumed specific seed species. In conventional-tillage, ants were one of the dominant consumers of seeds, suggesting different patterns of resource partitioning in each tillage system. We suggest that selective feeding by arthropod seed feeders, in combination with their high number, could affect the species composition and possibly the abundance of weeds in low-input, no-tillage agroecosystems.