COMPARISON OF MUSEUM AND PUBLISHED RELATIVE ABUNDANCES REVEALS A CONSISTENT PUBLICATION BIAS

Recent developments in paleobiology indicate that analysis of abundance data will be essential for continued progress in studies of paleodiversity through time. Many large-scale studies of paleodiversity are based on data from published literature, which at best reports abundance data inconsistently at worst does not report it at all. To assess the effects of such biases, the abundance data from published literature for seven mammal faunas from the Miocene of the Great Basin, western USA, are compared to the museum collections from which they were drawn. Analyses of rarefied richness and Hurlbert's evenness indicate that there is a consistent bias towards excessive evenness in the published data. This means that rarified richness is often overestimated, but to different extents for different assemblages; in general, faunas with a higher percentage of collections published tend to have a more accurate rarified richness, but monographic effects and unpublished taxa can affect this generality. Evenness data are more consistent, but examination of more faunas will be necessary to discern how robust this metric is to publication bias. Additionally, none of the rarified museum data matched a predicted slope of 0.5 in log-log space, calling into question one of the assumptions of the occurrences-squared weighted method of faunal resampling. These results indicate that future paleodiversity studies will not only require an analysis of abundance data but will also require a careful census of specimens in museum collections.