A confirmatory structural equation model was built in order to test the generality of Grace and Pugesek's model of species richness. A main feature of their model was that light reaching the soil surface had the strongest effects on species richness, and that disturbance and biomass effects were largely indirect via effects on light. Their model was not confirmed for the understory vegetation of floodplain oak savannas and a new model had numerous fundamental differences. Disturbance history had the strongest direct effects on richness and these were independent of biomass effects. Richness was maximal at intermediate disturbance and biomass. Bivariate relationships between soil quality and species density were very weak because soil quality simultaneously had negative direct effects and positive indirect effects (through biomass), such that the total effect of soil was negligible. This provides an example of how structural modeling can provide insights that are not possible with other numerical methods. The complex effects of soils support recent findings that some soil components tend to increase richness via a species pool effect while other components tend to reduce richness via biotic interactions. The effects of light were not significant, but canopy trees had weak, positive effects, and this contradicts other structural models which have generally shown that shading reduces species richness. Here, species richness increases with shade presumably because of species pool effects, whereby the species pool increases by including prairie, savanna, and some woodland species and indirectly by reducing dominance by warm-season grasses. The results have implications for management because of the overall importance of disturbance history, however the majority of the variation in richness was left unexplained and this suggests other factors such as dispersal limitation, soil fungi, and historical effects may be of overriding importance in these oak savannas.