Three kinds of satellite altimeter data have been combined, along with 5'¿5' bathymetric data, to calculate a 0¿.125 ocean wide gridded set of 2.3¿106 free-air gravity anomalies. The procedure used was least squares collocation that yields the predicted anomaly and standard deviation. The value of including the bathymetric data was shown in a test around the Dowd Seamount where the root mean square (rms) difference between ship gravity measurements decreased from ¿40 mgal to ¿20 mgal when the bathymetry was included. Comparisons between the predicted anomalies and ship gravity data is described in three cases. In the Banda Sea the rms differences were ¿20 mgal for two lines. In the South Atlantic rms differences over lines of 2000 km in length were ¿7 mgal. For cruise data in the Antarctica region the discrepancies were ¿12 mgal. Comparisons of anomalies derived from the Geosat geodetic mission data by Marks and McAdoo [1992> with ship data gave differences of ¿6 mgal showing the value of the much denser Geosat geodetic mission altimeter data. ¿ American Geophysical Union 1992