Water vapor adsorption isotherms a t five temperatures between 0 and 40°C were obtained on Li and Na kaolinite. By a careful experimental technique, the error in the determination of weight of water adsorbed was reduced to +O.Ol mg/g and that of equilibrium vapor pressure was reduced to t0.0002 mm Hg. The data were used to calculate the integral and differential entropy of the adsorbed water phase. The differential entropy of adsorption (the partial molar entropy of water in the sorbate phase minus the molar entropy of normal liquid water) has large positive values a t low surface coverage. As more water is adsorbed, the differential entropy drops to a negative value approximately equal to that of a two-dimensional liquid which is about the same as the entropy for hypothetical ice a t this temperature and pressure. The water can hardly be treated as " ice " because the large negative entropy persists over a very smallp/po range and occurs a t a p/pO such that the surface is only about 70 percent covered. The integral entropy of adsorption (the molar entropy of water in the sorbate phase minus the molar entropy of normal liquid water) has positive values on both Li and Na kaolinite throughout the p/po range investigated; plpo = 0 to plpo = 0.5. These entropy data indicate that the water molecules in the adsorbed phase on Li or Na kaolinite possess greater randomness than the water molecules in normal liquid water from the dry clay up to at least p/po = 0.5. The difference between integralanddifferential entropies is explained as a " configurational entropy " that changes with the amount of the adsorbed phase. A hypothesis to explain the entropy data is proposed in which one builds inward from normal liquid water to the solid surface rather than the more conventional mechanisms which build outward from the solid surface. I n this way the entropy data can be explained in terms of what happens to the structure of normal liquid water when different sized ions or nonpolar molecules, or both, are added. Literature data for the low density and dielectric coefiicient of the sorbed phase plus the large amounts of-unfrozen water still present a t-5'C are frequently used as evidence for the quasi-solid structure of the sorbed phase. A consideration of the water molecules in the sorbed phase as more random than normal liquid water is a n equally plausible explanation. Nuclear magnetic resonance and magnetic susceptibility data which cannot be explained by an increase in water structure are in complete agreement with the entropy data and the proposed new working hypothesis. It is concluded that not only the present entropy data but also all literature data (known to the author) concerning the sorbed phase can be satisfactorily explained if one considers the water molecules in the sorbed phase as being more random, less structured, than normal liquid water.