By means of 1¿¿1¿ gravimetric geoids in the North Atlantic, Northwest Pacific, and Indian oceans, comparisons have been made with Geos 3 radar altimeter estimates of geoid height. Most commonly, there are constant offsets and long-wavelength discrepancies between the two data sets; there are many probable causes, including radial orbit error, scale errors in the geoid, or bias errors in altitude determination. Across the Aleutian trench the 1¿¿1¿ gravimetric geoids do not measure the entire depth of the geoid anomaly. This is due to averaging over 1¿ squares and subsequent aliasing of the data. Especially above trench systems in the Pacific, there can be considerable energy in the geoid for frequencies higher than are resolvable by 1¿¿1¿ geoids; such a signal, though, is detected by the Geos 3 altimeter. After adjustment of Geos 3 data to eliminate long-wavelength discrepancies, agreement between the altimeter geoid and gravimetric geoid is between 1.7 and 2.7 m in rms error. In a hypothetical experiment a comparison can be made of an altimeter determining the geoid and a gravimeter measuring gravity. At longer wavelengths an altimeter will have a higher signal to noise ratio, but at shorter wavelengths a gravimeter will perform better.