Within a region and community type, site-to-site differences in plant diversity are explained primarily by differences in disturbance regimes, resource availability, and propagule supply. Less attention has been given to the effect of the physical soil environment, despite correlative evidence that diversity differs between deep and shallow soil grasslands. In correlative studies, soil depth effects can be confounded with differences in soil texture, moisture, and fertility. Because root architecture and foraging strategies differ among species, we hypothesized that soil depth alone could affect plant diversity. We evaluated soil-depth affects on plant species composition and richness (S) by seeding 36 species into replicated 0.2-m² plots that had 20-cm, 42-cm, or unrestricted rooting depths. To our knowledge this is the first direct experimental test of soil depth on plant species richness. We hypothesized that S would decline with decreasing soil depth, with a concomitant change in community composition. After three years, S significantly increased with soil depth (P<0.0001), but there was no change in species composition (MRPP, P>0.2). Species presence was strongly nested, with medium-depth species reflecting a subset of deep-soil species, and shallow-soil species reflecting a subset of medium-depth species. All depth treatments contained the same dominant grasses, thus differences in S resulted from the nested loss of forbs. Conversely, increasing soil depth added sets of new species, but the specific identity of the species present appeared interchangeable among plots of a given depth. We conclude that soil depth can contribute to among-site differences in S independent of species composition within tallgrass prairie.