Characteristics of the cross-frontal water exchange over Georges Bank are investigated through a sequence of numerical model experiments featuring realistic three-dimensional (3-D) bathmetry, bi-monthly averaged climatological stratification, tidal forcing, and mean and observed wind-forcing. The model used in this study is the ECOM-si version of the Blumberg and Mellor <1987> primitive equation model with Mellor-Yamada <1982> level 2.5 turbulence closure. The model domain includes Georges Bank and the Gulf of Maine, and is forced at the open ocean boundary by the semidurnal M2 tide. Fluid particles were tracked in the 3-D, time-dependent Eulerian flow field to examine the Lagrangian flow field, and passive tracer experiments were conducted to investigate the relative roles of advection and turbulent diffusion on cross-frontal exchange. Two distinct paths for the on-bank water movement were detected on Georges Bank: one is over the northwestern flank of the bank where the bottom topography changes sharply in both along- and cross-bank directions, and the other is near the bottom around the bank where the tidal mixing front is located. Over the northern flank, the cross-bank component of the Lagrangian residual current is generally opposite in direction to that of the Eulerian residual current, resulting in an on-bank, cross-frontal water transport near the bottom. Over the southern flank, the near-bottom water tends to converge toward the tidal mixing and shelf-break fronts, so that the near-bottom flow over the shelf between these two fronts is divergent. The response to wind-forcing varied with ambient stratification and water depth. In winter, strong winds can drive a significant off-bank water transport, tending to wash out the bank. In summer, winds are generally too weak to alter the general pattern of tidal-driven particle motion within the mixed region on the crest of the bank and the surrounding tidal mixing front. Some wind-driven off-bank transport occurs near the surface in the stratified region on the outer southern flank, but this has little influence on water movement near the bottom. Passive tracer experiments reveal that the net cross-frontal water flux near the bottom is caused primarily by advection and horizontal diffusion. Tidal-induced vertical diffusion tends to make the tracer mix rapidly upward, thus reducing the percent of the cross-frontal flux due to advection. Our new model results are consistent with previous model studies of Georges Bank, and agree in general with drifter and dye patches observed in the U.S. GLOBEC NW Atlantic/Georges Bank program.