On the SEASAT-A satellite, a microwave scatterometer was used to determine the vector wind over the world's oceans. The technique is based on the sensitivity of microwave radar backscatter to the centimeter length ocean waves created by the action of the surface wind. This paper describes the algorithm used to convert the scatterometer's measurements of ocean normalized radar cross section, &sgr;0, to the neutral stability vector wind at 19.5 m height and the comparison of these winds with high quality surface observations. The wind vector algorithm used an empirical &sgr;0 model function to describe the dependence of the ocean &sgr;0 on the 19.5-m neutral stability wind vector. Two model functions, developed from a limited base of aircraft and satellite &sgr;0 measurements, were evaluated by using an independent set of in situ surface wind observations from the Joint Air Sea Interaction Experiment (JASIN). Although these model functions were found to have some weaknesses, the results of these comparisons produced better results than the SEASAT specifications of wind speed accuracy of ¿2 m/s and wind direction accuracy of ¿20¿ over the 0--16 m/s range of winds observed during JASIN. An improved model function was later developed by 'tuning' to these JASIN data so that the remaining biases between the observed surface wind and the scatterometer-derived winds were minimized. Results are presented for this model function compared against other surface wind observations from the Gulf of Alaska SEASAT Experiment and the SEASAT Storms (Hurricane) Experiment.