Standardizing fossil disparity metrics using sample coverage

Estimating past biodiversity using the fossil record is a central goal of palaeobiology. Because raw estimates of biodiversity are biased by variation in sampling intensity across time, space, environments and taxonomic groups, sampling standardization is routinely applied when estimating taxonomic diversity (e.g. species richness). However, sampling standardization is less commonly used when estimating alternative currencies of biological diversity, such as morphological disparity. Here, we show the effects of standardizing fossil time series of morphological disparity to equal sample completeness, or 'coverage', of the underlying taxon-frequency distribution. We apply coverage-based standardization to three published datasets of discrete morphological characters (echinoderms, ichthyosaurs and ornithischian dinosaurs), and quantify disparity using two metrics: weighted mean pairwise dissimilarity (WMPD) and the sum of variance (SOV). We also compare the effects of coverage-based and sample-size-based standardization. Our results show that coverage standardization can yield estimates of disparity through time that dramatically deviate from raw estimates, both in magnitude and direction of changes. These findings demonstrate that future studies of morphological disparity should control for variation in sampling intensity to enable more reliable inferences.