Two field experiments were conducted during 1993/4 and 1994/5 in Cristalina-GO, Brazil, to test the effect of seed size on the performance of soybean. The following cultivars were compared: FT-Cristalina, IAC-8, Cariri and Doko-RC. The lots were classified according to three seed sizes, with manual sieves. The experimental design was a split plot with four replicates. The cultivars were considered main plots while the seed sizes were tested in the subplots (in the first year a control without sieving was included, which was not done in the second year because it was similar to the average of the three sizes). Seeding, harvesting and treshing were carried out manually so as to avoid mechanical seed injuries. The harvested seeds were classified by size and weighted. The small seeds always represented the lowest fraction of the seed lots (less than 20%). None of the agronomic parameters evaluated: plant population 30 days after seeding, height and number of pods per plant, yield, size and weight of the harvested seeds, were affected by seed size in soybean.