Size bias in the fossil record limits its use for interpreting patterns of past biodiversity and ecological change. Using comparative size frequency distributions of exceptionally good regional records of New Zealand Holocene and Cenozoic Mollusca in museum archive collections, we derive first-order estimates of the magnitude of the bias against small body size and the effect of this bias on completeness of the fossil record. Our database of 3907 fossil species represents an original living pool of 9086 species, from which ∼36% have been removed by size culling, 27% from the smallest size class (<5 mm). In contrast, non-size-related losses compose only 21% of the total. In soft rocks, the loss of small taxa can be reduced by nearly 50% through the employment of exhaustive collection and preparation techniques.