This study examines a mix of seven statistical and physical Special Sensor Microwave Imager (SSM/I) passive microwave algorithms that were designed for retrieval of over-ocean precipitable water (PW). The aim is to understand and explain why the algorithms exhibit a range of discrepancies with respect to measured PWs and with respect to each other, particularly systematic regional discrepancies that would produce substantive uncertainties in water vapor transports and radiative cooling in the context of climate dynamics. Data analysis is used to explore the nature of the algorithm differences, while radiative transfer analysis is used to explore the influence of several environmental variables (referred to as tangential environmental factors) that affect the PW retrievals. These are sea surface temperature (SST), surface wind speed (Us), cloud liquid water path (LWP), and vertical profile structure of water vapor [q(z)]. The main datasets include the Wentz matched radiosonde-SSM/I point database consisting of 42 months of globally distributed oceanic radiosonde profiles paired with coincident SSM/I brightness temperatures, and globally compiled instantaneous orbit-swath maps of SSM/I brightness temperatures for January and July 1990.Results demonstrate that the seemingly good agreement found in past studies and herein, within the conventional framework of scatter diagram analysis that ignores regional classification, gives way to poor agreement in the framework of monthly and zonally averaged differences. It is shown how much of the disagreement inherent to statistical algorithms is due to disjoint training datasets used in deriving algorithm regression coefficients. The investigation also explores how tangential environmental factors composed of variations in SST, Us, cloud LWP, and q(z) structure impart dissimilar errors to retrieved PWs, according to the design of the retrieval algorithms. A discussion on implications of the discrepancies vis-à-vis the Global Energy and Water Cycle Experiment program is given, with suggestions on mitigating discrepancies in algorithm designs.