The seasonal variability of the wind stress curl over the North Atlantic is investigated by means of empirical orthogonal function (EOF) analysis. The curl field is calculated from 1 year of First Global GARP Experiment wind data. It was found that 44% of the variability is contained in four significant eigenvectors. Their spatial patterns are characterized by basin-sized oscillations with larger amplitude to the north of 40¿N. Their associated time series coefficients have the highest amplitude during the winter and show tendency toward a white frequency spectrum which nevertheless exhibits noticeable peaks or gaps at certain frequencies. Physically, the first EOF is seen as the seasonal fluctuations of the mean wind stress curl pattern. Five other eigenvectors are also found to be above the noise level, but they account for only a smaller percentage of variability (19%). They are characterized by smaller spatial scales than the basis size. Their time series of coefficients show a whiter frequency spectrum.