The subinertial frequency-wavenumber structure and spatial coherence of wind stress curl, and their spatial and temporal variability, are described with the emphasis on characteristics which affect the ocean's response to the wind stress curl forcing function. Wind stress curl over the North Pacific between 25° and 60°N was computed for 3 years (1985–1988) from the Fleet Numerical Oceanography Center wind product. The range from winter peak to summer trough for the mean and variance are typically a factor of 3–4 and ≈10 respectively, with superimposed interannual changes of up to a factor of 2. Power spectra vary seasonally and interannually in an essentially frequency-independent manner that is consistent with the curl variance. However, the spectrum over the period band 5–100 days often fails a statistical test for whiteness at the 95% level. Zonal and meridional wavenumber spectra were estimated for 13 sites distributed around the central-eastern North Pacific using the maximum likelihood method. Directional trends with frequency are comparable to those from earlier studies if 3-year-long data segments are analyzed, with approximate wavenumber symmetry except for the zonal term at mid-latitudes for periods shorter than 10 days, where eastward propagation is dominant. However, spectra for shorter data sections sometimes display eastward and westward excursions which are only weakly similar over distances of ≈1000 km and essentially dissimilar over longer separations. The spatial correlation structure of wind stress curl is shown typically to have a main lobe of ≈1000 km and multiple intercorrelation lobes separated by ≥ 2000 km. The strength and location of the intercorrelation peaks vary slowly with time. These results suggest that curl behavior is more complex than was previously believed, that the use of long-term averages of or simple parameterizations for the frequency-wavenumber spectrum of wind stress curl in model studies may be unrealistic, and that more attention to actual curl characteristics at the time oceanic measurements are collected will be required to reconcile models with observations.