In order to investigate continental-scale patterns of plant species richness and rarity, distribution maps of 3661 plant species were digitized into a one degree grid of sub-Saharan Africa using the WORLDMAP computer programme. Cells with high species richness were also likely to be those containing the greatest number of species of restricted range, but areas such as the South African Cape and the Eastern Arc mountains were found to have more restricted-range species than predicted from their richness scores. The two environmental predictors which had the strongest individual relationships with both species richness and range-size rarity were absolute maximum annual temperature and mean monthly potential evapotranspiration. However, correlative predictive powers of these variables were low, with R=−0.58 and R=−0.54, respectively (P < 0.01). Multiple regression also failed to produce a strong explanatory model for observed continental-scale patterns of diversity. Spatial variability analysis showed that this was likely to be because different environmental parameters predicted different centres of richness and rarity. West African species richness was better predicted by absolute maximum annual temperature, whereas East African species richness was better predicted by mean monthly potential evapotranspiration. © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 142, 187–197.