Erwin's much debated estimate of 30 million species of arthropods is revised. The original estimate is based on the evaluation of host specificity of guilds in beetle samples, and subsequent hierarchical ratio extrapolations. The growing number of studies including mass sampling of arthropods have provided several data sets suitable for obtaining an empirical basis of this estimate. The structure in this modified version is somewhat changed compared to the original estimate in order to make each hierarchical step more easily testable. Plant species are separated into different growth forms, and host specificity measures are based only on phytophagous species. Effective specialization is applied as a measure of host specificity to correct for the fauna shared between plant species. A between community correction factor is applied to correct for differences in host specificity at different spatial scales. There are still great uncertainties attended with such estimates. The largest problems refer to the between community correction factor and the proportion of canopy species to total species. Further work on host specificity and the least known hyperdiverse groups are also needed. The revised version of the estimate does not support hyperestimates of 30 100 million species. Rather, it compares nicely with estimates derived from other estimation methods, indicating a global arthropod species richness of 5–10 million species.