Most prairie restorations fail to produce the diversity of species found in unplowed remnants. This lack of restored diversity is hypothesized to be partly due to the inhibition of forb species by high seeding densities of dominant grasses and partly due to the low seeding densities of forbs used in many restorations. We tested this hypothesis by sowing various densities of forb and warm-season grass seeds into a restoration begun on bare soil. This is the first replicated restoration experiment we are aware of that varies grass seeding densities to examine the effects on forbs. Four years after seeding, we found that higher densities of grass seeds decreased forb cover, biomass, and richness, and higher densities of forb seeds increased forb richness. These results suggest that dominant grasses compete strongly with native forb species and that many forb species thrive when they are spatially separated from dominant grasses. The results also suggest that seed availability limits the establishment of some forbs. Forb diversity can therefore be increased by decreasing grass seeding density, by increasing forb seeding density, or both. However, forb seeds are generally expensive, and increasing forb seeding density across the entire area of a restoration may be prohibitively expensive. We therefore recommend a low seeding density of dominant grasses, and we recommend spatially separating forbs from dominant grasses by adding most forb seeds to areas with little to no dominant grasses and by adding the rest of the forb seeds to areas with a low density of dominant grasses.