In recent years several authors have questioned the reality of a widely accepted and apparently large increase in marine biodiversity through the Cenozoic. Here we use collection-level occurrence data from the rich and uniquely well documented New Zealand (NZ) shelfal marine mollusc fauna to test this question at a regional scale. Because the NZ data were generated by a small number of workers and have been databased over many decades, we have been able to either avoid or quantify many of the biases inherent in analyses of past biodiversity. In particular, our major conclusions are robust to several potential taphonomic and systematic biases and methodological uncertainties, namely non-uniform loss of aragonitic faunas, biostratigraphic range errors, taxonomic errors, choice of time bins, choice of analytical protocols, and taxonomic rank of analysis.The number of taxa sampled increases through the Cenozoic. Once diversity estimates are standardized for sampling biases, however, we see no evidence for an increase in marine mollusc diversity in the NZ region through the middle and late Cenozoic. Instead, diversity has been approximately constant for much of the past 40 Myr and, at the species and genus levels, has declined over the past ∼5 Myr. Assuming that the result for NZ shelfal molluscs is representative of other taxonomic groups and other temperate faunal provinces, then this suggests that the postulated global increase in diversity is either an artifact of sampling bias or analytical methods, resulted from increasing provinciality, or was driven by large increases in diversity in tropical regions. We see no evidence for a species-area effect on diversity. Likewise, we are unable to demonstrate a relationship between marine temperature and diversity, although this question should be re-examined once refined shallow marine temperature estimates become available.