The TOPEX/POSEIDON satellite carries the first dual-frequency radar altimeter. Monofrequency (Ku-band) algorithms are presently used to retrieve surface wind speed from the altimeter's radar cross-section measurement (&sgr;Ku0). These algorithms work reasonably well, but it is also known that altimeter wind estimates can be contaminated by residual effects, such as sea state, embedded in the &sgr;Ku0 measurement. Investigating the potential benefit of using two frequencies for wind retrieval, it is shown that a simple evaluation of TOPEX data yields previously unavailable information, particularly for high and low wind speeds. As the wind speed increases, the dual-frequency data provides a measurement more directly linked to the short-scale surface roughness, which in turn is associated with the local surface wind stress. Using a global TOPEX &sgr;0 data set and TOPEX's significant wave height (Hs) estimate as a surrogate for the sea state's degree of development, it is also shown that differences between the two TOPEX &sgr;0 measurements strongly evidence nonlocal sea state signature. A composite scattering theory is used to show how the dual-frequency data can provide an improved friction velocity model, especially for winds above 7 m/s. A wind speed conversion is included using a sea state dependent drag coefficient fed with TOPEX Hs data. Two colocated TOPEX-buoy data sets (from the National Data Buoy Center (NDBC) and the Structure des Echanges Mer-Atmosphere, Proprietes des Heterogeneites Oceaniques: Recherche Experimentale (SEMAPHORE) campaign) are employed to test the new wind speed algorithm. A measurable improvement in wind speed estimation is obtained when compared to the monofrequency Witter and Chelton <1991> model. ¿ 1998 American Geophysical Union