At the Appleton-Whittell Research Ranch, an ungrazed grassland preserve in southeastern Arizona, soil seed banks were sampled in June, August, and October 2002 and June 2003. Wildfire had previously burned 90% of the research ranch in May 2002. Seed density and species richness in burned native grassland (2 plots) were compared to those in burned exotic grassland (2 plots). Averaged over 4 sample dates, seed densities were as follows: burned native grassland, 591 ±243.1 seeds ·m⁻² and 784 ±334.9 seeds ·m⁻²; burned exotic grassland, 501 ±198.9 seeds ·m⁻² and 196 ±123.8 seeds ·m⁻². Species richness in the seed bank, also averaged over 4 sample dates, was as follows: burned native grassland, 16.3 ±1.7 species ·m⁻² and 19.5 ±1.0 species ·m⁻²; burned exotic grassland, 12.0 ±3.4 species ·m⁻² and 11.06 ±2.5 species ·m⁻². The seed bank of burned exotic grassland contained significantly fewer seeds and species than that of burned native grassland. In addition, the seed bank in burned exotic grassland comprised mainly exotic grasses, whereas annual and perennial herbs, most of them native, dominated the seed bank of burned native grassland. Of the 50 species detected in soil samples, only 20 had a persistent seed bank, and only 1 of these was a native perennial bunchgrass. The preponderance of transient species means that eradication of exotic grasses must be followed by reseeding of native grasses and herbs, perhaps repeatedly, if native grassland is to replace exotic grassland.