An analysis of the changes in mortality estimates that occur with changes in the population representations used for damage assessment. Representations for the U.S. population are created by laying a square mesh grid of a specified dimension over a map of the United States and adding up the population contained within each cell. The populations represented within grid sizes of 5, 10, 20, and 50 km are considered. Two hypothetical attacks are employed: a low-collocation attack and an indiscriminate or unrestrained targeting attack. For both attacks, mortality estimates tend to decrease as the grid size of the population representation increases.