Monthly averaged evaporation over the global oceans is estimated from 1 year of special sensor microwave imager (SSM/I) data using the method proposed by Liu (1984). Intercomparisons involving SSM/I and in situ data are made over a wide range of oceanic conditions during August 1987 and February 1988 to determine the source of errors in the evaporation estimates. The most significant spatially coherent evaporation errors are found to come from estimates of near-surface specific humidity q. Systematic discrepancies of over 2 g kg-1 are found in the tropics, as well as in the middle and high latitudes. The q errors are partitioned into contributions from (1) the parameterization of q in terms of the columnar water vapor W (i.e., the Liu q:W relationship) and (2) the retrieval algorithm for W. It is found that the spatial pattern of discrepancies due to the q:W relationship can be interpreted readily in terms of dynamical and physical processes that maintain the vertical profile of atmospheric water vapor content. The effects of W retrieval errors are found to be smaller over most of the global oceans and due primarily to the implicitly assumed vertical structures of temperature and specific humidity on which the physically based SSM/I retrievals of W are based. To improve estimates of q, it will thus be necessary to have supplementary information on both the temperature and water vapor profiles that is able to distingish the most important dynamical and physical regimes over the global oceans. ¿ American Geophysical Union 1993