It has recently been shown that the linear elastic steady-state compliance J(e)(0) reacts very sensitively on the addition of nanoparticles to a polymer melt. This effect can be attributed to an interaction between the particles and the matrix molecules. Creep-recovery experiments have evolved as a very suitable tool to measure J(e)(0), because the time window can be extended wide enough to detect the processes underlying the interactions. Whereas the effect of the particle geometry on J(e)(0) has already been investigated to some extent, the influence of the molecular structure of the matrix is still an open question. Therefore, in this study investigations of two polystyrenes with different molar masses and distributions and composites with 1 vol.% silica nanoparticles each are reported. One polystyrene is an anionic product (aPS) with a narrow molar mass distribution, the other a radically polymerized sample (PS 158K) with a broader molar mass distribution. Due to their molecular structures the unfilled polymers already differ significantly in their rheological properties. The linear steady-state elastic compliance is found to be J(e)(0) = 1.9 x 10(-5) Pa-1 for the aPS and J(e)(0) = 2.5 x 10(-4) Pa-1 for the PS 158K, which is in agreement with the literature. Investigations on nanocomposites with a poly (methyl methacrylate) of M-w/M-n = 1.5 as the matrix have shown that the elasticity, measured by J(e)(0) in the creep-recovery experiment, strongly increases with the specific surface area of the nanoparticles added. Also for the PS composites an increase of J(e)(0) was found by adding 1 vol.% of silica nanoparticles. However, the relative increase strongly depends on the elasticity of the unfilled matrix. Whereas for the PS 158K an increase of J(e)(0) of only 70% is found, it is much larger, namely 470%, in the case of the anionic PS. (C) 2011 Elsevier Ltd. All rights reserved.