The physically based total column water vapor retrieval algorithm of Wentz (1992) is tested for sensitivity to realistic vertical distributions of temperature and water vapor. The spatial pattern of systematic retrieval error of total column water vapor due to this sensitivity is simulated using the European Center for Medium Range Weather Forecasts (ECMWF) monthly averaged temperature and humidity fields. Effects of liquid water and ice are not considered. The estimated systematic error is within 0.1 g cm-2 over about 70% of the global ocean area; systematic errors greater than 0.3 g cm-2 are expected to exist only over a few well-defined regions, about 3% of the global oceans, assuming that the global mean value is unbiased. Over subtropical and middle latitudes the largest biases occur in regions with strong temperature inversions, such as the northwest Pacific during summer. The warm temperature structure is significantly different from the globally averaged temperature structure used implicitly in the retrieval algorithm. The largest biases in the tropics appear to be confined to regions with horizontal wind convergence and large amounts of water vapor in the lower troposphere. The large amounts of water vapor cause saturation of the 22-GHz channel by water vapor, which increases the sensitivity of the retrieval algorithm to discrepancies introduced by simplifications in radiative transfer equation that forms the basis of the Wentz algorithm. ¿ American Geophysical Union 1993