Numerous techniques for empirical correction of the dominant once per revolution orbit error present in satellite altimeter data have been suggested along with various tests of their validity. These validations typically involve comparison of data from other sources, for example tide gauges, with altimetry, comparison of different orbit error models with which to approximate the radial error, or comparison of variabilities from collinear analyses with those from crossover differences. A new means of validating empirical orbit error correction is here presented through use of crossover difference differences (CDDs) to independently eliminate the dominant orbit error terms. A CDD, Ci,j, is defined to be Ci,j=Xi+1,j+1-Xi,j, where Xi,j is the crossover difference derived from the intersection of ascending arc number i with descending arc number j; the arcs i and i+1, for example, are separated in time by one orbital period. This provides a measure of the change in ocean height difference between two geographical locations on the same circle of latitude, separated by an angle &khgr;, equal to the angle through which the Earth has rotated during the orbital period. Removal of residual gravity field modeling error and systematic solar radiation pressure orbit errors produce a result that is purely ocean signal differences plus residual noise contamination. CDDs have been applied with Geosat crossover difference data from the Southern Ocean to validate an empirical correction technique obtained by combining the methods suggested by Milbert et al. (1988) and Tai (1988). The comparison has shown that a maximum, nonsteric annual signal of less than 1.5 cm was absorbed by empirical correction. In addition, it has been shown that relative height time series can be generated &khgr; due east or west of a known time series without any need for further orbit error reduction.