The overall accuracies of vector wind measurements from the NASA scatterometer (NSCAT) are quantified by comparisons with collocated data from operational U.S. National Data Buoy Center ocean buoys. A vector correlation statistic is used to examine the geographical distribution of full-mission NSCAT-buoy differences. The dependences of the vector correlation on collocation radius and scatterometer errors such as wind speed offset and random component errors are quantified. For purposes of NSCAT validation, 30 ocean moored buoys having high full-mission sample correlations are considered. When applied to this set of ocean buoys, a nonlinear analysis shows that the 25-km resolution NSCAT wind speeds have unity gain, an offset of -0.3 m s-1, an rms error of 1.3 m s-1, and component standard deviations of ~1.3 m s-1. For wind speeds greater than 6 m s-1, fewer than 3% of all collocated solutions have significant ambiguity removal errors. The rms directional difference for the remaining vectors is less than 17¿. At lower wind speeds the frequency of ambiguity removal errors increases and the directional accuracy of the correct vectors decreases with decreasing wind speed, consistent with the assumptions of the random error model underlying the nonlinear wind speed regression approach. While the primary emphasis of this paper is on determining the quantitative accuracy of the NSCAT wind velocity measurements, the analysis techniques and interpretations have more general application to validation studies involving vector quantities. ¿ 1999 American Geophysical Union