Charge inversion is the phenomenon in which an electric double layer contains more counterions than needed to compensate the surface charge. For colloidal particles this has the consequence that the apparent surface charge, as inferred from electrophoresis or interaction studies, has a sign opposite to that of the actual surface charge, obtainable by titration. This phenomenon has been known for over a century. According to the traditional interpretation, the inversion is caused by (chemical) specific adsorption of ions. However, beginning in the early 1980s it has been predicted by a large number of workers that charge inversion should occur as a consequence of many-body ion−ion correlations. For surfaces of sufficiently high surface charge density in the presence of divalent or multivalent counterions, charge inversion is expected to be ubiquitous even in the absence of specific adsorption. Testing this prediction has proved difficult because chemical specific adsorption is a very common phenomenon and can outweigh the effects of ion correlations. So far, no experimental systems have been thoroughly investigated where strong specific adsorption could be unambiguously ruled out under conditions where charge inversion due to ion−ion correlations is predicted. Here, we solve this problem by studying the mercury/aqueous MgSO4 interface. This system has the advantage that highly accurate double layer data are available for a variety of conditions, including some where chemical specific adsorption is known to be absent (or at least very weak). From precise data for this system [Harrison, J. A.; Randles, J. E. B.; Schiffrin, D. J. J. Electroanal. Chem. 1970, 25, 197] one can establish the ionic components of charge and surface charge density. To extract quantitative theoretical predictions about the consequences of ion−ion correlations, we use the highly accurate anisotropic hypernetted chain (AHNC) method, where ion−ion correlations in the double layer are taken into account in a fully self-consistent fashion. It is found that for moderate to large negative surface charge densities and not too high concentration, the variation in the ionic components of charge with the surface charge density can to a large extent be quantitatively explained by enrichment of ions close to the surface due to ion−ion correlations. That chemical specific adsorption of Mg2+ is negligible is supported by considering the properties of the double layer close to the electrocapillary maximum. In view of the large body of evidence indicating that the counterions tend to specifically adsorb on the mercury surface for positive polarization but not for negative, the agreement between theory and experiment for negative surface charge constitutes strong evidence for ion−ion correlations as the origin of charge inversion.