Existing unreinforced masonry buildings frequently suffer out-of-plane local collapse mechanisms when undergoing earthquake ground motion. The energy damping that occurs during the motion, due to impacts of a wall against the foundation or against other walls, is a relevant parameter on the response. An experimental investigation has been carried out to estimate the dissipation of kinetic energy that takes place during free oscillations. Restraint conditions allow for two-sided rocking (wall resting on a foundation) and one-sided rocking (wall resting on a foundation adjacent to transverse walls). Five specimens have been tested, modelling walls acted out-of-plane (fa double dagger ades). When one-sided rocking is under consideration, different depths of the contact surface between fa double dagger ade and transverse walls are considered. In the case of two-sided rocking, the experimental coefficient of restitution is slightly lower than the analytic coefficient. In the case of one-sided rocking, an analytic formulation is proposed and this is compared against experimental data. Although the coefficient of restitution of one-sided rocking is less than half that of two-sided rocking, it is not equal to zero. Thus, it cannot induce a sudden stop of the motion. Hence, nonlinear time history analyses performed under this assumption may prove unsafe. Moreover, a comparison has been carried out between overturning maps, induced by twenty natural accelerograms, computed for the analytic coefficient of restitution and those computed for the experimental coefficient of restitution. The increased energy dissipation reduces the frequency of overturning and causes a more regular behaviour