Falling weight deflectometer (FWD) testing is now considered routine for the evaluation of pavement structures. Averaging of FWD load and deflection data collected at the same location and similar drop loads is common before further analysis. The stated reason for this averaging is to decrease the effect of the random error inherent in FWD deflection sensors. Implicit in this methodology is the assumption that there is no significant trend in deflection results with successive load applications. FWD data from the Long-Term Pavement Performance program were analyzed to determine whether a significant trend exists between deflection and drop number. In a majority of data sets, a statistically significant trend does exist for at least one deflection sensor. The trends are most common for the highest load level and the center deflection sensor. These trends are less often practically significant but are common enough that averaging deflection data can be expected to increase total error in a large minority of cases.