Despite its fundamental importance in radiative transfer, atmospheric dynamics, and the hydrological cycle, atmospheric water is inadequately characterized particularly at a global scale. Occultation measurements from the Global Positioning System (GPS) should improve upon this situation. Individual occultations yield profiles of specific humidity accurate to 0.2 to 0.5 g/kg providing sensitive measurements of lower and middle tropospheric water vapor with global coverage in a unique, all-weather, limb-viewing geometry with several hundred meters to a kilometer vertical resolution. We have derived water vapor profiles from June 21 to July 4, 1995, using GPS occultation data combined with global temperature analyses from the European Center for Medium-Range Weather Forecasts (ECMWF) and reanalyses from the National Centers for Environmental Prediction (NCEP). The zonal mean structure of the profiles exhibits basic climatological features of tropospheric moisture. Specific humidity biases between the GPS results and the ECMWF global humidity analyses in the middle to upper troposphere are ∼0.1 g/kg or less. Occultation results below 6 km altitude are generally drier than those of ECMWF with the bias generally increasing toward warmer temperatures. Near the height of the trade wind inversion, the ECMWF analyses are significantly moister than the occultation results due to vertical smoothing and overextension of the boundary layer top in the analyses. Overall, the occultation results are drier than the NCEP reanalyses with a marked exception near the Intertropical Convergence Zone (ITCZ) where occultation results are wetter by more than 10%. The occultation results are significantly wetter near the ITCZ and drier in the subtropics than the classical moisture climatology of Peixoto and Oort. Similarities between the NCEP and the Peixoto and Oort near-ITCZ differences suggest that a common analysis/model problem may be responsible. The generally wetter Peixoto and Oort results in the subtropics are due in part to moist radiosonde biases. Discrepancies between these data sets are significant and limit our ability to resolve uncertainties in moisture control and feedbacks in a changing climate.