A global statistical analysis of the first 10 months of TOPEX/POSEIDON merged geophysical data records is presented. The global crossover analysis using the Cartwright and Ray (1990) (CR) tide model and Gaspar et al. (this issue) electromagnetic bias parameterization yields a sea level RMS crossover difference of 10.05 cm, 10.15 cm, and 10.15 cm for TOPEX-TOPEX, POSEIDON-POSEIDON, and TOPEX-POSEIDON crossovers, respectively, All geophysical corrections give reductions in the crossover differences, the most significant being with respect to ocean tides, solid earth tide, and inverse barometer effect. Based on TOPEX-POSEIDON crossovers and repeat-track differences, we estimate the relative bias between TOPEX and POSEIDON at about -15.5 +/- 1 cm. This value is dependent, however, to the electromagnetic bias corrections used. An orbit error reduction method based on global minimization of crossover differences over one cycle yields an orbit error of about 3 cm RMS. This is probably an upper estimate of the orbit error since the estimation absorbs other altimetric signals. The RMS crossover difference is reduced to 8.8 cm after adjustment. A repeat-track analysis is then performed using the CR tide model. In regions of high mesoscale variability, the RMS sea level variability agrees well with the Geosat results. Tidal errors are also clearly evidenced. A recent tide model (Ma et al., this issue) determined from TOPEX/POSEIDON data considerably improves the RMS sea level variability. The reduction of sea level variance is (4 cm)(2) on average but can reach (8 cm)(2) in the southeast Pacific, southeast Atlantic, and Indian Oceans. The RMS sea level variability thus decreases from 6 cm to only 4 cm in quiet ocean regions. The large-scale sea level variations over these first 10 months most likely show for the first time the global annual cycle of sea level. Finally, we analyze the TOPEX and POSEIDON sea level anomaly wavenumber spectral characteristics. TOPEX and POSEIDON have identical spectral characteristics at low wavenumbers. For wavelengths shorter than 100 km, however, POSEIDON spectra are more energetic. This is probably related to the TOPEX tracker characteristics and to the way the acceleration correction is made in the geophysical data records. POSEIDON repeat-track noise level is estimated at about 3 cm for a 1-s average. The TOPEX repeat-track noise level is about 1.8 cm RMS but this probably corresponds to averages over several seconds.