In this thesis, production of dense water that feeds the dense overflows across the Greenland-Scotland Ridge has been considered A new circulation scheme is developed which is consistent with the water masses, currents and air-sea fluxes in the region, and with the important observation that the dense overflows show little or no seasonal or interannual variability. An inverse box model has been constructed that shows that the new circulation scheme is consistent with conservation statements for mass, heat and salt as well. According to the new circulation scheme the major buoyancy is lost in the North Atlantic Current, which enters the Norwegian Sea between Iceland and Scotland, and flows northward towards the Arctic Ocean and the Barents Sea. The transformation is due to a large net annual heat loss over the North Atlantic Current, combined with a long residence time (2-3 years) and a large surface area. After subduction, one branch of the North Atlantic Current enters the Arctic Ocean, is modified in hydrographic properties into those associated with the Denmark Strait Overflow Waters in the western North Atlantic, exits the Arctic Ocean in the western Fram Strait and flows with the East Greenland Current towards the Denmark Strait Another branch of the North Atlantic Current recirculates directly in the Fram Strait and flows towards the Denmark Strait with the East Greenland Current This branch will not sink to the bottom of the North Atlantic as it is less compressible than the Arctic branch. The third branch of the North Atlantic Current enters the Barents Sea, continues to lose buoyancy, and enters the Arctic Ocean at intermediate depth. This branch exits the Arctic Ocean in the western Fram Strait, circulates around the Greenland Sea, enters the Norwegian Sea, and flows towards the Faeroe-Shetland Channel. The traditional view holds that the major sources of the dense overflows are the Iceland and Greenland gyres, west of the North Atlantic Current. Aside from the finding that the new circulation scheme is more likely in terms of water mass properties, currents etc., one fundamental problem with the old scheme lies with supplying a substantial overflow. There are indications that the production of dense water in the gyres is sensitive to the highly variable surface conditions and that indeed the production tends to shut on and off. The reservoirs in the gyres are so small that they would be drained within a few years if they were to supply the overflows during a shutdown period. Production of dense water within the North Atlantic Current is less sensitive to surface conditions. The density in the gyres is gained at a temperature around freezing, whereas in the North Atlantic Current the density is gained well above freezing. Therefore a freshwater anomaly in the two domains will have different consequences for vertical · overturning: within the North Atlantic Current the freshening can be overcome by further cooling, whereas in the gyres freezing will occur and the vertical overturning will cease. The observed lack of a significant seasonal signal associated with the dense overflows is consistent with the new circulations scheme. The net annual cooling dominates the seasonal oscillation in the atmospheric heat loss for time scales comparable with the residence time of the Atlantic Water within the domain. Thus winter formation of dense water within the North Atlantic Current does not induce a seasonal signal in the transport field of the dense water.