The U.S. Navy's Geosat altimeter mission between 1985 and 1989 provided the first long-term measurements of sea surface topography with sufficient accuracy to observe large-scale changes in the ocean circulation. However, the overall accuracy of these measurements has been difficult to assess because of the lack of coincident measurements of the radar range corrections. In particular, Geosat, unlike Seasat, did not make simultaneous measurements of the range delay caused by refraction from atmospheric water vapour. This delay varies between 5 and 50 cm. In order to correct for this delay, estimates of columnar water vapour in the altimeter path are modelled using estimates from numerical weather forecast models or non-coincident satellite measurements. Comparisons of these estimates with direct radiosonde measurements have been made by others and clearly indicate a preference for the satellite measurement over the forecast model. NASA's Very Long Baseline Interferometry programme routinely estimates the wet tropospheric range delay at radio frequencies for all antenna locations at 15 minute intervals with a precision of a few cm rms. We have used these measurements to make a direct assessment of the Geosat wet troposphere range correction. The VLBI measurements show strong diurnal variations in columnar water vapour at several sites. These variations are a source of error (order 3 cm root-mean-square) in any non-coincident measurement of the wet troposphere range delay. These errors can be aliased to lower frequencies because of the satellite sampling scheme and have an effect on studies of annual and interannual changes in sea level with Geosat data.