Two theoretical formulations of the sea state bias (SSB) by Srokosz <1986> (hereafter S86) and Elfouhaily et al. <2000> (hereafter E00) are applied to directional ocean wave spectra from WAM and from NDBC moored buoys collocated with Topex altimeter data. S86 SSB computations obtained for WAM with idealized ocean current fields reveal a quasi-linear relationship between the SSB coefficient and the r.m.s. ocean slope, which remains valid for a wide range of wind/wave/current conditions. The SSB coefficient is shown to be better parameterized in terms of slope, instead of the significant wave height and wind speed presently used in empirical models. The theoretical SSB for the buoys' 2-D wave spectra are compared with estimates from five recent empirical SSB models calculated from the collocated Topex data. The magnitude of the E00 SSB shows a strong sensitivity to the choice of high frequency tail model used to extend the 2-D wave spectra. We find that the E00 theory can adequately model the radar frequency dependence of the SSB given a suitable choice of high frequency tail. However, the E00 theory is found to systematically underestimate the magnitude of the SSB coefficient with respect to empirical SSB models. In contrast, the long-waves-only S86 theory displays good agreement with the empirical SSB estimates. Thus our calculations suggest that the sea state bias may be primarily governed by the slope of long gravity waves, presumably through its well-documented influence on the modulation of short waves, as also suggested by recent empirical findings.