The relationship between the number of species and the area sampled is one of the oldest and best-documented patterns in community ecology. Several theoretical models and field data from a wide range of plant and animal taxa suggest that the slope, z, of a graph of the logarithm of species richness against the logarithm of area is roughly constant, with z approximately 0.25. We collected replicated and randomized plant data at 11 spatial scales from 0.01 to 10(8) square meters in Great Britain which show that the slope of the log-log plot is not constant, but varies systematically with spatial scale, and from habitat to habitat at the same spatial scale. Values of z were low (0.1 to 0.2) at small scales (<100 square meters), high (0.4 to 0.5) at intermediate scales (1 hectare to 10 square kilometers), and low again (0.1 to 0.2) for the largest scale transitions (e.g., East Berks to all of Berkshire). Instead of one process determining changes in species richness across a wide range of scales, different processes might determine plant biodiversity at different spatial scales.