Since recent literature has quantified the persistence of changes in the real exchange rate (RER) using trade-weighted data, in this paper we ask whether the trade-weighted RER is mean reverting. We focus on post-Bretton Woods data for the G7 countries and, after revising the strong correlation between the RER and the nominal exchange rate over that period, we follow a fractional integration approach. We consider different assumptions for the residuals and allow for breaks at unknown dates. We conclude that the nonstationary behaviour of the RER is mean reverting (i.e., it is integrated of order d∈0.5,1) for about half of the G7 countries and that allowing for structural breaks affects the test results obtained in absence of breaks but do not invalidate them.