This work presents the main results obtained within a project on mechanical properties of polymer based nanocomposites. The specific point was how to analyze and model the filler–filler interactions in the description of the viscoelastic behavior of these materials. This paper aims at presenting the general strategy used by the different partners to address this question, together with original experimental results and micro-mechanical modeling. Different nanocomposite materials were fabricated using the latex route, leading to random dispersions of rigid submicronic particles (PS = polystyrene, silica) in a flexible polybutylacrylate matrix at various volume fractions. In addition, encapsulated silica particles in a styrene–acrylate copolymer were produced, leading, after film formation, to a limited number of contacts between silica fillers. The processing route of these encapsulated particles was optimized and the resulting morphology was analyzed by TEM experiments. In the case of random mixtures, a strong effect of reinforcement appears in the rubbery field of the soft phase when the filler content is above a critical fraction (percolation threshold). The reinforcement in the rubbery plateau can be still exacerbated in the case of the PS particles if the material undergoes a heat treatment above the main relaxation of the PS phase. These experimental results illustrate the difference between geometrical percolation (when particles are just in contact) and mechanical percolation (with strong interactions between the fillers). The comparison of the results for PS and silica fillers shows once more that the strength of the interactions plays an important role. To account for the whole set of experimental data, two ways of modeling were explored: (i) homogenization methods based on generalized self-consistent schemes and (ii) a discrete model of spheres assembly which explicitly describes the ability of the contacts to transmit efforts.