Abstract
Summary Quaternary pollen records may contribute uniquely to the understanding of present plant diversity. Pollen assemblages can reflect diversity at community and landscape scales but the time resolution of most studies does not match that of modern ecological studies. Because of the complicating effects of differential pollen productivity and dispersal, pollen records do not directly reflect equitability aspects of vegetation diversity. Vegetation diversity indices other than S (the total number of taxa) are therefore not appropriate for pollen assemblages. As a measure of the species richness palynological richness is biased by the lack of taxonomic precision, by a possible interference on pollen dispersal from vegetation structure and by pollen representation. The nonlinear relationship between species richness and pollen‐taxa richness may be used in attempts to estimate past floristic richness from fossil pollen assemblages. Using a hypothetical example the strong effect of cover shifts in the vegetation affecting taxa with different representation (R rel ) values on observed palynological richness is demonstrated. It is suggested that estimates of relative pollen productivity should be used to guide the pollen sum on which pollen‐type richness is estimated by rarefaction techniques and this approach is illustrated using a paired site study of late Holocene diversity dynamics. The need for a modern training set relating pollen‐type richness to species richness, pollen productivity and vegetation structure is emphasized.