This paper evaluates both the ideas and empirical evidence pertaining to the control of species density in herbaceous plant communities. While most theoretical discussions of species density have emphasized the importance of habitat productivity and disturbance regimes, many other factors (e.g. species pools, plant litter accumulation, plant morphology) have been proposed to be important. A review of literature presenting observations on the density of species in small plots (in the vicinity of a few square meters or less), as well as experimental studies, suggests several generalizations: (1) Available data are consistent with an underlying unimodal relationship between species density and total community biomass. While variance in species density is often poorly explained by predictor variables, there is strong evidence that high levels of community biomass are antagonistic to high species density. (2) Community biomass is just one of several factors affecting variations in species density. Multivariate analyses typically explain more than twice as much variance in species density as can be explained by community biomass alone. (3) Disturbance has important and sometimes complex effects on species density. In general, the evidence is consistent with the intermediate disturbance hypothesis but exceptions exist and effects can be complex. (4) Gradients in the species pool can have important influences on patterns of species density. Evidence is mounting that a considerable amount of the observed variability in species density within a landscape or region may result from environmental effects on the species pool. (5) Several additional factors deserve greater consideration, including time lags, species composition, plant morphology, plant density and soil microbial effects. Based on the available evidence, a conceptual model of the primary factors controlling species density is presented here. This model suggests that species density is controlled by the effects of disturbance, total community biomass, colonization, the species pool and spatial heterogeneity. The structure of the model leads to two main expectations: (1) while community biomass is important, multivariate approaches will be required to understand patterns of variation in species density, and (2) species density will be more highly correlated with light penetration to the soil surface, than with above-ground biomass, and even less well correlated with plant growth rates (productivity) or habitat fertility. At present, data are insufficient to evaluate the relative importance of the processes controlling species density. Much more work is needed if we are to adequately predict the effects of environmental changes on plant communities and species diversity.