Sound reduction index measurements were performed inside a 1:10 scale model, reproducing the two adjoining rooms of an acoustic laboratory, with the aim of investigating the influence of the boundary conditions on the acoustic performance of lightweight double walls. The model was entirely built in polymethyl methacrylate (PMMA). The specimen double walls were built using the same acrylic material. At first, measurements were performed on a single homogeneous panel, in order to check the reliability of the whole procedure and the repeatability of the measurements. A good correspondence was found between theoretically calculated values and measured values of the sound reduction index above the critical frequency. A second set of measurements was carried out on model double walls constituted by two leaves, connected each other only by a border frame running along the whole wall perimeter. It has been found that the effect of the border frame is a reduction of the performance of the double wall under that of a single plate with the same total surface mass. A third set of measurements was carried out inserting vertical studs to connect the two leaves. It has been found that the presence of the studs causes a general decrease of the sound reduction index values and a decomposition of the critical frequency. A fourth set of measurements was carried out on the same double wall using border frames made of different materials. The experimental results show that damped materials (wood, rubber) result in a better performance than rigid materials (acrylic, aluminium).