SKELETONS IN THE CLOSET: ESTIMATING THE NUMBER OF INVALID SPECIES NAMES

Large-scale analyses of biodiversity patterns depend on accurate taxonomy. Many databases attempt to track and exclude known synonyms and nomina dubia, but these efforts are compromised by invalid names that have not yet been unrecognized. Because new names are proposed at a steady rate, this problem always will persist. A new, simple method allows estimating the true proportion of valid names given the past history of invalidation and revalidation: if the per-taxon, per-historical year invalidation rate is x and the revalidation rate y, the equilibrium valid name proportion or "flux ratio" is y/(x+y). The method is applied to a complete database of some 5200 proposed North American Cretaceous and Cenozoic fossil mammal species names. About 69% of names currently are accepted as valid, but no more than 58% actually are. Thus, > 16% of accepted names are invalid and diversity is overestimated by > 19%. Similar results are obtained using an older method of Solow's that differs because it ignores revalidation and assumes a fixed pool of truly valid names. Flux ratios for different subsets of names show additional patterns. (1) The quality of names varies through historical time: for example, names from 1870 - 1900 are highly unreliable. However, names improve only irregularly through the 20th century, and the lack of time to vet new names is a more important factor. (2) Validity rates for names proposed by productive taxonomists range broadly from 26% (Hay) to 82% (Matthew), and vetting rates also vary considerably. (3) Compared to ungulates, insectivoran and rodent names are on average some 20 - 30 years younger and less than half as likely to be invalid, but twice as many of these invalid names are currently in use. (4) There are many invalid names in the late Eocene and late Oligocene, and the proportion of hidden invalid names is generally high in the Paleogene. Although the figures cannot be used directly to correct sampling-standardized diversity curves, they indicate that significant revisions need to be made. For example, taxonomic error may conceal a diversity drop during the mid-Eocene, coincident with known shifts in body mass distributions and taxonomic composition.