The heat capacities of very small adsorbed amounts of poly(methyl methacrylate) on high-surface-area silica (Cab-O-Sil) were measured using temperature-modulated differential scanning calorimetry (TMDSC) using a quasi-isothermal method and interpreted via different models. The composition-dependent heat capacities of the adsorbed samples were measurably less than those predicted with a simple mixture model. A two-state model, composed of tightly and loosely bound polymer, fits the data better with heat capacities of the tightly bound polymer found to be 70-80% (glassy region) and 70-94% (rubbery region) of that of the bulk polymer at the same temperatures. The amount of tightly bound polymer was estimated to be about 1.2 mg/m(2) (about 1 nm thickness) in both the glassy and rubbery regions, consistent with heat flow measurements. The data sets were also extensive enough to model them with a more detailed layered gradient model, including a nonzero heat capacity for the polymer at zero adsorbed amount, which increased based on an exponential growth function to bulk polymer value of the heat capacity away from the surface. More importantly, this gradient model mimicked the experimental dependence on adsorbed amounts in the tightly bound adsorbed amount region (approximately 1 mg/m(2)). This model provided, for the first time, an experimental estimate of the heat capacity of the polymer adsorbed closest to the surface. The fractional heat capacity of the adsorbed polymer closest to the silica surface, relative to bulk polymer, increased with temperature from 0.3 (well below) to 0.8 (well above the bulk T-g). It was also possible to estimate the exponential growth parameter of the development from the initial heat capacities to the bulk heat capacity as 0.4 to 0.6 mg/m(2), identifying a distance scale (0.3 to 0.5 nm) consistent with the notion of a transition from tightly bound to loosely bound polymer.