The database comprises 393 non-experimental datasets describing geographic gradients richness gradients at any spatial scale. We searched journals, computerized literature databases and databases generated by other NCEAS working groups. The response variable was taxonomic richness (usually of species, but occasionally of genera or families) of any plant or animal group. A study was included if we could extract coefficients of determination of independent variables from any standard linear or non-linear statistical technique. Because a wide range of independent variables are measured in diversity studies, we classified them into six basic types (Climate/Productivity, Heterogeneity/Disturbance, Edaphics/Nutrients, Area, Biotic interactions and History, each of which represents a hypothesis type). We also included a few studies where pairs of independent variables were combined by authors, as long as they belonged to the same class of variable. A study was excluded if it considered only a single hypothesis type. In cases where raw data could be extracted from the original paper, we recalculated the models to verify the published results, transforming variables if appropriate and computing coefficients of determination if the original authors did not. Because some data were reanalyzed, the coefficients of determination sometimes differ from those originally reported. To facilitate comparative analyses, for each case five categorical grouping variables are provided: taxon, insularity, habitat medium, grain and extent. The final two are facets of scale, which was the main focus of the working groups analysis.