A statistical technique is developed for determining fully empirical model functions relating altimeter radar backscatter (σ0) measurements to near-surface neutral stability wind speed. By assuming that σ0 varies monotonically and uniquely with wind speed, the method requires knowledge only of the separate, rather than joint, distribution functions of σ0 and wind speed. Analytic simplifications result from using a Weibull distribution to approximate the global ocean wind speed distribution; several different wind data sets are used to demonstrate the validity of the Weibull approximation. The technique has been applied to 1 year of Geosat data. Validation of the new and historical model functions using an independent buoy data set demonstrates that the present model function not only has small overall bias and RMS errors, but yields smaller systematic error trends with wind speed and pseudowave age than previously published models. The present analysis suggests that generally accurate altimeter model functions can be derived without the use of collocated measurements, nor is additional significant wave height information measured by the altimeter necessary.