We have developed a new molecular thermodynamic model to predict the glass transition temperature (T-g) of nanocomposites consisting of polymer (1) and nanoparticle (2) based on the configurational entropy model and the Flory-Huggins theory. Four configurational entropies have been taken into account in this study: the disorientation entropy of the polymer (Sdis-1), the confinement entropy of the nanoparticle (Scon-2), the mixing (Smix-12), and the specific interaction entropies (Sspe-12) of the polymer/nanoparticle. Quantitative descriptions according to the proposed model are consistent with experimental T-g data of polymer nanocomposites. Our results demonstrate that T-g values of polymer nanocomposites can be enhanced or depressed relative to neat polymer depending on the intensity of specific interaction between polymer and nanoparticles, which is quantitatively identified by a physical parameter (gamma(spe)) in this model. (c) 2007 Wiley Periodicals, Inc.