Abstract
Body size is a key organismal trait with profound implications ranging from individual physiology to large-scale macroecological or macroevolutionary phenomena. Among extant terrestrial vertebrates, peak diversity commonly occurs at small body size. Similarities between the body size distributions of fossil and extant mammals have been used to argue that fossil record signals are robust, yet preservation and collector biases disproportionately favour the sampling of large taxa and probably under-represent small-sized diversity. Here, we quantify the effects of these biases on the body size distributions of North American mammals through the Cenozoic. We assess how these distributions have changed with new palaeontological discoveries and evaluate sampling standardization as a potential correction for body size bias. Our results show bias in the mammal record to be persistent and severe. Sampling standardization has no consistent effect on recovered distribution shape probably because sample coverage estimators cannot account for changes in the scope of the sampling universe driven by a combination of historical worker interest and the preservational characteristics of a small pool of formations. Short of a novel standardization method that can account for publication biases, deriving non-artefactual fossil body size signals may ultimately depend on targeted, systematic sampling of exceptional deposits.