Much of the contribution of bivalves to paleoecology is based on the fraction preserved whole. The use of fragments has been limited by the inability to reconstruct the whole shell. For this study, shells of selected species were broken, metrics obtained for shell reconstruction, and a stepwise model developed and tested to evaluate the potential of reconstructing whole shells from fragments. The model is based on the facts that shell thickness is well correlated with shell length, width, and weight, that fragments can be assigned to shell areas, and that each shell area can be matched to a potential donor shell by combinations of length, width, and weight. Simulations using the model show that shell reconstructions are feasible and that the size frequency can be reconstructed more accurately than the number of shells. If all fragments are preserved, shell number tends to be modestly overestimated. Assuming fragment loss, the shell number estimated will be a conservative estimate of original abundance. Reconstructed size frequencies did not differ significantly from the original size frequencies of the whole shells even if 40% of the fragments were not preserved. Biovolume, consequently, may also be estimable. Standardly, bivalve fragments are enumerated by counting intact umbos. If all fragments are preserved, this approach is preferable, as shell model reconstructions overestimate abundance under that circumstance. However, the expectation that fragments with intact umbos are uniquely well preserved is unrealistic and if fragment loss has occurred, relying on all fragments may provide an improved estimate of the original abundance and, in addition, an estimate of the size frequency and the latter proves to be robust even with 40% of the fragments lost.