Although many indices estimate diversity, species richness recently has been used as a surrogate for diversity in many studies in ecology, biogeography, and conservation. Underlying assumptions of this approach are that all diversity indices, including those that weight species importance by their relative abundance (e.g., evenness), are correlated positively, and that richness accounts for a large proportion of the variance in diversity. We addressed these assumptions with data from six grassland sites using univariate and multivariate analyses of a variety of indices (species evenness, richness, rarity, dominance, and Simpson's diversity index). Univariate correlations between plant species evenness and richness were weak and negative at each site. Principal-component analyses consistently revealed two significant components of variation in diversity. Richness and evenness were largely orthogonal, with Simpson's diversity loading between them. Thus, measures of species diversity based on relative abundance, as well as richness, may be necessary to capture the full complexity of diversity in conservation studies and in experiments of biodiversity and ecosystem functioning. At these and perhaps other sites, species richness was an incomplete surrogate for diversity.