Wind-driven flow in the upper 90 meters during autumn–winter in the midlatitude North Pacific is investigated using satellite-traced drifting buoys (i.e., drifters) deployed nearly simultaneously but drogued at different depths. The difference in the relationship between drifter velocity and wind stress as a function of drogue depth is observed to change when its drogue entered the deepening mixed layer. This change is characterized by a sudden increase in the amplitude of the near-inertial motions observed by the drifters and by the onset of a persistent net displacement whose downwind component is approximately 3 times as large as its crosswind component. Attempts to model this downwind flow as a windage result in a large unexplained residual downwind velocity component. On the other hand, 80–90% of the observed crosswind displacement is explained by an Ekman slab model (i.e., with flow uniform over the mixed layer). This large residual downwind velocity component combined with the Ekman driven... Abstract Wind-driven flow in the upper 90 meters during autumn–winter in the midlatitude North Pacific is investigated using satellite-traced drifting buoys (i.e., drifters) deployed nearly simultaneously but drogued at different depths. The difference in the relationship between drifter velocity and wind stress as a function of drogue depth is observed to change when its drogue entered the deepening mixed layer. This change is characterized by a sudden increase in the amplitude of the near-inertial motions observed by the drifters and by the onset of a persistent net displacement whose downwind component is approximately 3 times as large as its crosswind component. Attempts to model this downwind flow as a windage result in a large unexplained residual downwind velocity component. On the other hand, 80–90% of the observed crosswind displacement is explained by an Ekman slab model (i.e., with flow uniform over the mixed layer). This large residual downwind velocity component combined with the Ekman driven...