The fossil record is our only direct means for evaluating shifts in biodiversity through Earth's history. However, analyses of fossil marine invertebrates have demonstrated that geological megabiases profoundly influence fossil preservation and discovery, obscuring true diversity signals. Comparable studies of vertebrate palaeodiversity patterns remain in their infancy. A new species-level dataset of Mesozoic marine tetrapod occurrences was compared with a proxy for temporal variation in the volume and facies diversity of fossiliferous rock (number of marine fossiliferous formations: FMF). A strong correlation between taxic diversity and FMF is present during the Cretaceous. Weak or no correlation of Jurassic data suggests a qualitatively different sampling regime resulting from five apparent peaks in Triassic-Jurassic diversity. These correspond to a small number of European formations that have been the subject of intensive collecting, and represent 'Lagerstätten effects'. Consideration of sampling biases allows re-evaluation of proposed mass extinction events. Marine tetrapod diversity declined during the Carnian or Norian. However, the proposed end-Triassic extinction event cannot be recognized with confidence. Some evidence supports an extinction event near the Jurassic/Cretaceous boundary, but the proposed end-Cenomanian extinction is probably an artefact of poor sampling. Marine tetrapod diversity underwent a long-term decline prior to the Cretaceous-Palaeogene extinction.