Quantifying the history of biodiversity requires counts of fossil specimens to estimate the number of individuals per taxon in a sample. Counting fragments of individuals may introduce biases that are not present when examining only whole body fossils or other non-repeating skeletal elements that represent a known fraction of an individual (e.g., one valve of a bivalve representing half an individual). This study reports on biases revealed by quantitative comparisons of counts of non-repeating skeletal elements of Pleistocene bivalves (hinge fragments and whole valves) with counts of other shell fragments (non-hinge fragments). Compared to the hinge fragments and whole valves, non-hinge fragments yielded lower measured diversity with specific biases toward large-sized species and species whose shell microstructure was foliated or nacreous. The faunal composition preserved in non-hinge fragments was distinct from that preserved in whole valves and hinge fragments according to multivariate analysis, even though all specimens were drawn from the same sediment bulk samples. These biases should be considered when agglomerating data from multiple studies to address questions about large-scale patterns in the history of life.