The variability of sea level and surface geostrophic currents in the Southern Ocean is investigated from the first 26 months of unclassified Geosat altimeter data (November 1986 to December 1988). Because of problems unique to Geosat, it has been necessary to develop new techniques for analyzing the height data. These techniques are presented here, and the processed Geosat data are used to examine the relation between mesoscale variability and the mean circulation (as determined from historical hydrographic data). The two are shown to be significantly correlated, implicating the importance of hydrodynamic instabilities in the Antarctic Circumpolar Current. The geographical patterns of both the mean flow and the mesoscale variability are shown to be controlled by the bathymetry. An efficient objective analysis algorithm is introduced for generating smoothed fields from observations randomly distributed in time and two space dimensions. The algorithm is applied to the 26 months of Geosat data, and the smoothed fields are used to investigate the large-scale, low-frequency variability of sea level and surface geostrophic velocity in the Southern Ocean. Approximately 33% of the variance is accounted for by the first three empirical orthogonal functions (EOFs) of sea level variability. These three modes describe variability over seasonal time scales and separate into an annual cycle (mode 1), a semiannual cycle (mode 2) and a mode which describes year-to-year variability in the seasonal cylces for 1987 and 1988 (mode 3). The complexity of the spatial patterns of the second- and higher-order modes and small percentages of variance accounted for by the first three modes (15%, 10% and 8%, respectively) reflect the generally regional, as opposed to coherent circumpolar, nature of sea level variability in the Southern Ocean. The inherent weak zonal coherence of the variability is further emphasized by EOF analysis separately within each basin of the Southern Ocean. ¿ American Geophysical Union 1990