A study of the power spectra of auroral radar echoes shows that they are sometimes asymmetrical. This is probably due to two waves (at different heights) having different line-of-sight velocities. The velocity splitting occurs over a limited region of space (~300 km in geographic extent) that moves toward the equator. The velocity splitting seems to be associated with divergence of the STARE drift velocities. This has been calculated and found to have significant time and space variations-much too large to be explained by magnetic field changes. The divergence is greatest at times when the power spectrum is asymmetrical and more than one velocity may be present. Several possible corrections (including a few unlikely ones) have been tried, but none reduce the magnitude of the divergences appreciably. We must conclude that STARE does not measure the electric field exactly and that a correction velocity of the order of 25 m s-1 and divergence of the order of 10-4 s-1 is required. We have not been able to find what it depends on, except its association with velocity splitting.