An assimilated sea ice motion product is used to track ice parcels in several regions of the Arctic over time periods of one day to several weeks during 1992-1993. Motions simulated using a two-dimensional, dynamic-thermodynamic sea ice model are combined with motions derived from daily 85 GHz special sensor microwave/imager (SSM/I) imagery using an optimal interpolation method that minimizes error covariance. Assimilation attenuates the tracking error over the stand-alone model in comparison to buoy trajectories with the same starting location and time. The average 14-day assimilated trajectory's displacement error is as much as 34% lower than the model trajectory, while the RMS direction error is decreased by up to 10 degrees (24%). Assimilation can also yield an estimate of dispersion, which is not retrievable by point buoy observations. An assimilation approach improves estimates of ice drift and has the potential to further the understanding of ice mass flux, freshwater flux, and pollutant transport in the polar regions.