The ocean surface vector wind can be measured from space by scatterometers. For a set of measurements observed from several viewing directions and collocated in space and time, there will usually exist two, three, or four consistent wind vectors. These multiple wind solutions are known as ambiguities. Ambiguity removal procedures select one ambiguity at each location. We compare results of two different ambiguity removal algorithms, the operational median filter (MF) used by the Jet Propulsion Laboratory (JPL) and a two-dimensional variational analysis method (2d-VAR). We applied 2d-VAR to the entire NASA Scatterometer (NSCAT) mission, orbit by orbit, using European Centre for Medium-Range Weather Forecasts (ECMWF) 10-m wind analyses as background fields. We also applied 2d-VAR to a 51-day subset of the NSCAT mission using National Centers for Environmental Prediction (NCEP) 1000-hPa wind analyses as background fields. This second data set uses the same background fields as the MF data set. When both methods use the same NCEP background fields as a starting point for ambiguity removal, agreement is very good: Approximately only 3% of the wind vector cells (WVCs) have different ambiguity selections; however, most of the WVCs with changes occur in coherent patches. Since at least one of the selections is in error, this implies that errors due to ambiguity selection are not isolated, but are horizontally correlated. When we examine ambiguity selection differences at synoptic scales, we often find that the 2d-VAR selections are more meteorologically reasonable and more consistent with cloud imagery.