Cloud liquid water path (LWP) observed over the oceans by microwave sensors is one of only a handful of physical cloud parameters measured from space. These products have been used in a limited way in climate and cloud feedback studies, and may also have future application in data assimilation. However, we believe one reason these products have not been used more extensively is because little is known about their error characteristics. This study utilizes previously developed satellite techniques to help interpret cloud LWP products from the TMI (TRMM Microwave Imager) under different cloud conditions and provide estimates of their accuracy over the global oceans. These methods include cloud clearing TMI LWP products using VIRS (Visible and Infrared Scanner) data and comparing these products to similar products derived from VIRS for cloudy scenes. Results for July and January 1998 show that the TMI products contain significant clear-sky biases (mainly positive) that vary from region to region and season to season. Under conditions of overcast nonprecipitating warm clouds, both TMI and VIRS products are highly correlated but the TMI products are biased too high (about 40-50% on a zonal average basis). Using monthly mean clear-sky fields to adjust the TMI products brings these products into excellent agreement with the VIRS products, suggesting the TMI product biases are mainly attributed to biases in water vapor/oxygen absorption and/or biases in the surface characterization in the LWP retrievals. Under broken cloud scenes, the TMI products show a negative bias that clearly increases with decreasing cloud amount (i.e., the beam-filling effect), demonstrating that in these situations LWP products derived from coarse-resolution microwave measurements are not a true indicator of cloud LWP.