The well-known along-track resolution loss in synthetic aperture radar (SAR) ocean wave image spectra is investigated comparing simulations based on Hasselmann's nonlinear integral transform and measurements from the Norwegian Continental Shelf Experiment 1988 experiment. In the literature the resolution loss has often been modeled as a low-pass filter process, described by the rms azimuth shift width &sgr;x, acting within a quasi-liner SAR transformation. Estimates from real data of &sgr;x, as a function the range-to-platform velocity (R/V) and incidence angle, are compared to Hasselmann's new nonlinear spectral transformation and the widely used quasi-linear model. Simulations correlated with real wave data show that the quasi-linear model, with contributions to &sgr;x from the entire spectrum, overestimates &sgr;x by roughly 30--40%. The conformity between Hasselmann's model and real data is excellent, however. The numerics also indicate that the degree of nonlinearity in Hasselmann's transform is explicitly related to the surface truth parameters' significant wave height and peak wavelength. Furthermore, the spectral bandwidth, including the subresolution part of the ocean wave spectrum, seems to be of minor importance. It is also shown that the resultant smearing (due to imaging nonlinearities) cannot explicitly be separated from the coherent linear velocity bunching part of the transform. This point is especially discussed since it has led previously to some dissension regarding which ocean spectral components are most essential in the smearing process. ¿ American Geophysical Union 1993