A series of recent papers have indicated that reconstructed bulk compositions of abyssal peridotites define chemical correlations, namely increasing FeO tot and decreasing SiO 2 with increasing MgO, which cannot be produced by simple extraction of partial melts. However, no general consensus exists on the reliability of these trends, because they could be artifacts of the adopted calculations, and on the origin of abyssal peridotites. Specifically, it has been inferred that abyssal peridotite compositions are consistent with combined histories of partial melting and subsequent melt migration which caused either olivine addition, or dissolution and precipitation reaction, by equilibrium porous flow [Niu et al., Earth Planet. Sci. Lett. 152 (1997) 251–265; Asimow, Earth Planet. Sci. Lett. 169 (1999) 303–319]. We report combined bulk-rock and mineral chemical data for ophiolitic peridotites from the Erro-Tobbio (ET) Unit (Voltri Massif, Ligurian Alps), which represent lithosphere remnants of the Jurassic Ligurian Tethys embryonic ocean. These peridotites include sp-lherzolites and sp-harzburgites, and display overall depleted geochemical signature. However, comparison between bulk rock and corresponding mineral compositions reveals that these rocks cannot be residues of simple (equilibrium or fractional) melt extraction. Mineral compositions are similar in all the samples. By contrast, the bulk rock compositions define striking correlations, i.e. increasing FeO tot, Ni, Co, and decreasing Al 2O 3, SiO 2, CaO, Sc, Cr, Yb N, with increasing MgO: the MgO–FeO tot and MgO–SiO 2 correlations are similar to those recognized in abyssal peridotites. Thus, the ET peridotites provide evidence that the above trends are indeed consistent with similar variations in on-land peridotites; also, these trends cannot simply result from progressive melt depletion, because constituent minerals in the different ET samples have rather uniform composition. Calculated bulk modes indicate that the observed chemical variations are coupled to systematic modal changes, namely decrease in cpx and opx, and increase in olivine, at increasing bulk MgO. The ET peridotites also display decrease of the cpx/opx ratio at increasing bulk MgO, and this argues against a process of simple olivine addition. By contrast, some peculiar bulk-mineral compositional features – e.g. (i) nearly constant olivine Mg* [=Mg/(Mg+Fe tot)] values, at increasing modal olivine, (ii) opposite bulk Cr and Ni correlations, (iii) bulk Cr and Yb decrease, and parallel Ni increase, strikingly correlated with progressive modal cpx decrease and concomitant modal olivine increase – are consistent with the expected chemical and modal effects, during a process of interaction between depleted peridotites and melts migrating by equilibrium porous flow (involving pyroxene dissolution and olivine precipitation reactions). Bulk-rock and mineral chemistry data in the ET peridotites thus indicate that the major element correlations inferred by Niu et al. for the abyssal peridotites are reliable, and most likely result from a combined history of partial melting and melt interaction by reactive porous flow.