An array including seven moorings was in place between 10 August and 25 September 1982 on southern Georges Bank. Ten of the records from the Vector-Averaging Current Meters (VACMs) were complete. Short-term drifters with window-shade drogues at 10m seaward of the shelfbreak (200m) documented flow of shelf water across bathymetry. Winds were available for this period from NOAA buoy 44003 at the southwestern corner of Georges Bank near the array.The major structure in the subtidal flow is an alongshelf maximum toward the west (mean speed ~27cms - 1 with bursts often >30cms - 1 ) at the shelfbreak front (~100m): the shelfbreak jet. The first two modes in empirical orthogonal analysis of the cross-bank structure contain 82% of the variance. Mode 1 is associated with the large-scale alongshelf current and is partially correlated with local winds. Mode 2, generally not correlated with wind, may be generated by features initiated by the Gulf Stream. The current record at the northern periphery of the warm eddy or remnant of a warm-core ring has a mean velocity of 5cms - 1 toward the east for the period, opposite the general circulation over the shelf. Elsewhere the circulation over the shelf seemed minimally affected even with the proximity of the offshore feature. A mean estimate for volume transport of shelf water during the period is 0.83+/-0.2Sv with periods of greater than 1Sv related to NE winds leading by 6h. While a bias is possible due to undersampling of the velocity field, this mean value and variability during late summer are higher than previously reported estimates for transport of shelf water. This is primarily due to the additional contribution of the shelfbreak jet near the 100m isobath. Confidence limits for the transport were calculated by two independent methods, geometric weighting and block estimation by universal kriging.