Taxonomic structure of the fossil record is shaped by sampling bias

Understanding biases that affect how species are partitioned into higher taxa is critical for much of paleobiology, as higher taxa are commonly used to estimate species diversity through time. We test the validity of using higher taxa as a proxy for species diversity for the first time by examining one of the best fossil records we have, that of deep-sea microfossils. Using a new, taxonomically standardized, data set of coccolithophorid species and genera recorded from 143 deep-sea drilling sites in the North Atlantic, Caribbean, and Mediterranean, we show that there is a two-stepped change in the ratio of species to genera over the last 150 myr. This change is highly unexpected and correlates strongly with changes in both the number of deep-sea sites yielding coccolithophorids that have been studied and with the number of taxonomists who have published on those sections. The same pattern is present in both structurally complex heterococcoliths and the simpler nannoliths, suggesting that increasing complexity is not the driving factor. As a stepped species-to-genus ratio exists even after subsampling to standardize either the numbers of sites or numbers of papers, both factors must be contributing substantially to the observed pattern. Although some limited biological signature from major extinction events can be recognized from changes in the species-to-genus ratio, the numbers of sites and the numbers of taxonomists combined explain some 82% of the observed variation over long periods of geological time. Such a strong correlation argues against using raw species-to-genus ratios to infer biological processes without taking sampling into account and suggests that higher taxa cannot be taken as unbiased proxies for species diversity.