Over 32 years of wind data, which are based on surface pressure maps, from the Hindcast data base that has been developed by the Norwegian Meteorological Institute have been used to study the wind stress curl field, over the Nordic Seas. The mean wind stress curl pattern is characterized by very large values over most of the area, especially over the Greenland Sea. Within the framework of Sverdrup dynamics this gives rise to a cyclonic circulation in the area, with a maximum transport in the western boundary current of about 35 Sv. Only one gyre is present, and its center is situated between Jan Mayen and Greenland. The seasonal variation of the wind stress curl is very large, with almost negligible values during summer, from May through August. During September the wind stress curl starts to build up and has reached its full winter strength by November. This maximum is maintained until April, when it begins to decline to typical summer values. In order to study the spatial and temporal scales of the wind stress curl, an empirical orthogonal function analysis was performed on the wind stress curl after filtering it with a half-power period of 50 days and then resampling the wind stress curl every 30 days. A Monte Carlo method is used to estimate the statistical significance of the various modes. The first 11 modes are found to be significant, and they represent 83.6% of the total variance of the data set. The first mode clearly shows the seasonal fluctuations of the mean wind stress curl pattern, representing the seasonal variation in the intensity of the Icelandic low. The next two modes show mostly interannual variability, with a decadal time scale. Most of these variations are happening in the northern part of the Greenland Sea. The other significant modes are mainly describing variability that is limited either in space or in time. A large part of this anomalous variability occurs in the 1980s.