The reinforcement of elastomers by inorganic fillers, a concept of very high technological importance, is commonly understood to result from the presence of a mechanical network of partially aggregated filler particles. The non-linear mechanical properties, in particular the decrease of the modulus at high strain (Payne effect), are further interpreted to be a consequence of the breakdown of this filler network. There are, however, many open questions concerning the actual nature of the interparticle connections, where a modified polymer layer forming "glassy bridges" constitutes one possibility. In this work, we address this issue with a suitable silica-filled model elastomer, where we characterize the silica dispersion by SANS in combination with reverse Monte-Carlo modeling, and the mobility modification of the polymer by low-field proton NMR spectroscopy. In our samples, we identify a glassy layer as well as a region of intermediate mobility (possibly modified Rouse modes). Based on the structural information from SANS, we are able to quantify the amount of interparticle connections, and correlate it with the magnitude of the Payne effect taken from shear rheology. This works only if we assume that these connections comprise both, the glassy layer as well as the region of intermediate mobility. The amount of glassy immobilized polymer only does not suffice to explain the mechanical properties.