The relation between sea level and barometric pressure and, specially, the validity of the inverted barometer (IB) approximation is examined over the global oceans, using nearly 2 years of TOPEX-POSEIDON altimeter measurements. Both crossover differences and collinear differences between consecutive cycles are utilized in this study. Linear regressions between barometric pressure and sea level time series yield coefficients between 0.8 and 1 cm/mbar poleward of 20¿ and as low as 0.5 cm/mbar in the equatorial regions. Such deviations from the IB value of 1 cm/mbar can be due to the presence of data errors or to correlations between pressure and adjusted sea level (i.e., sea level corrected for IB effect). A simple error model for the pressure fields and a number of sensitivity tests are used to evaluate the changes in the regression coefficient possibly induced by data errors (pressure errors, altimeter measurements errors, and radial orbit errors). The combined (root-mean-square) effect of the different errors amounts to 0.8 mm/mbar poleward of 20¿ and 1.8 mm/mbar within 20¿ of the equator, in general smaller than the observed deviations from the IB value. Regression coefficients thus imply a correlation between adjusted sea level and pressure. Results from a shallow-water, global ocean model forced by realistic wind and pressure fields corroborate this finding. The model is able to explain the observed coefficients, within measurement errors, with wind-driven effects being most important in accounting for differences from the simple IB model. Pressure-forced dynamical signals cause maximum deviations of only 1 mm/mbar. The analyses point to the general validity of the IB approximation over the deep oceans but also highlight the complex relation between sea level and barometric pressure resulting from correlations between various sea level signals.¿ 1997 American Geophysical Union