F, Cl and Br contents of tholeiitic volcanic glasses dredged along the Mid-Atlantic Ridge from 53° to 28° N, including the transect over the Azores Plateau, are reported. The halogen variations parallel those of 87 Sr/ 86 Sr, La/Sm or other incompatible elements of varying volatility. The latitudinal halogen variation pattern is not obliterated if only Mg-rich lavas are considered. Variations in extent of low-pressure fractional crystallization or partial melting conditions do not appear to be the primary cause of the halogen variations. Instead, mantle-derived heterogeneities in halogens, with major enrichments in the mantle beneath the Azores, are suggested. The Azores platform is not only a ‘hotspot’ but also a ‘wetspot’, which may explain the unusually intense Azores volcanic activity. The magnitude of the halogen and incompatible element enrichments beneath the Azores appear strongly dependent on the size of these anions and cations, but independent of relative volatility at low pressure. The large anions Cl and Br behave similarly to large cations Rb, Cs and Ba, and the smaller anion F similarly to Sr and P. Processes involving crystal and liquid (fluid and/or melt), CO 2 rather than H 2 O dominated, seem to have produced these largescale mantle heterogeneities. Geochemical ‘anomalies’ beneath the Azores are no longer apparent for coherent element pair ratios of similar ionic size. Values of such ‘unfractionated’ coherent trace element ratios provide an indication of the mantle composition and its nature before fractionation event (s) which produced the inferred isotopic and trace element heterogeneities apparently present beneath the North Atlantic. The relative trace element composition of this precursor mantle does not resemble that of carbonaceous chondrites except for refractory trace element pairs of similar ionic size. It is strongly depleted in halogens, and to a lesser extent in large alkali ions Rb and Cs relative to refractory Ba. These relative depletions are comparable within a factor of 5 to Ganapathy & Anders’s estimates for the bulk Earth, with the exception of Cs. There is also evidence for removal of phosphorus into the iron core during its formation. With the exception of San Miguel, alkali basalts from the Azores Islands appear to have been derived from the same mantle source as tholeiitic basalts from the ridge transect over the Azores Platform but by half as much degree of partial melting. The Azores subaerial basalts seem to have been partly degassed in Cl, Br and F, in decreasing order of intensity. A working model involving metasomatism from release of fluids at phase transformation during convective mantle overturns is proposed to explain the formation of mantle plumes or diapirs enriched in larger relative to smaller halogen and other incompatible trace elements. The model is ad hoc and needs testing. However, any other dynamical model accounting for the 400 -1000 km long gradients in incompatible trace elements, halogens and radiogenic isotopes along the Mid-Atlantic Ridge should, at some stage, require either (1) some variable extent of mixing or (2) differential migration of liquid relative to crystals followed by re-equilibration (or both), as a diffusion controlled mechanism over such large distances is clearly ruled out, given the age of the Earth.