Sampling biases obscure the early diversification of the largest living vertebrate group.

Extant ray-finned fishes (Actinopterygii) dominate marine and freshwater environments, yet spatio-temporal diversity dynamics following their origin in the Palaeozoic are poorly understood. Previous studies investigate face-value patterns of richness, with only qualitative assessment of biases acting on the Palaeozoic actinopterygian fossil record. Here, we investigate palaeogeographic trends, reconstruct local richness and apply richness estimation techniques to a recently assembled occurrence database for Palaeozoic ray-finned fishes. We identify substantial fossil record biases, such as geographical bias in sampling centred around Europe and North America. Similarly, estimates of diversity are skewed by extreme unevenness in the occurrence distributions, reflecting historical biases in sampling and taxonomic practices, to the extent that evenness has an overriding effect on diversity estimates. Other than a genuine rise in diversity in the Tournaisian following the end-Devonian mass extinction, diversity estimates for Palaeozoic actinopterygians appear to lack biological signal, are heavily biased and are highly dependent on sampling. Increased sampling of poorly represented regions and expanding sampling beyond the literature to include museum collection data will be critical in obtaining accurate estimates of Palaeozoic actinopterygian diversity. In conjunction, applying diversity estimation techniques to well-sampled regional subsets of the 'global' dataset may identify accurate local diversity trends.