SUMMARY A controlled-source electromagnetic sounding survey centred on an axial volcanic ridge (AVR) segment of the Reykjanes Ridge at 57°45aeN was performed as part of the RAMESSES experiment. Low-frequency (0.35‐11 Hz) electromagnetic signals were transmitted through the crust to an array of horizontal electric field recorders at the seafloor to ranges of 15 km from the source, which was a 100 m long horizontal electric dipole towed at heights of 50‐80 m from the seafloor. Coincident seismic and magnetotelluric studies were conducted during the rest of the RAMESSES experiment. Data were interpreted using a combination of 1-D forward modelling and inversion, and iterative forward modelling in two dimensions. On the axis of the AVR, the resistivity at the seafloor is 1 V m. There is a steep resistivity gradient in the upper few hundred metres of the crust, with the resistivity reaching approximately 10 V ma t a depth of 500 m. In order to explain the low resistivities, the upper layer of the crust must be heavily fractured and saturated with sea water. The resistivity increases with distance from the axis as the porosity decreases with increasing crustal age. The most intriguing feature in the data is the large diVerence in amplitude between fields transmitted along and across the AVR axis. A significant zone of low-resistivity material is required at approximately 2 km depth beneath the ridge crest in order to explain this diVerence. It is coincident with the low-velocity zone required by the seismic data, and has a total electrical conductance in excellent agreement with the results of the magnetotelluric study. The low-resistivity zone can be explained by the presence of a body of partially molten basalt in the crust. Taken together, these results provide the first clear evidence for a crustal magma chamber at a slow spreading mid-ocean ridge. The data constrain the melt fraction within the body to be at least 20 per cent, with a melt volume suYcient to feed crustal accretion at this segment of the ridge for of the order of 20 000 years. Since this body would freeze in the order of 1500 years, this finding lends support to the hypothesis that, at slow spreading rates, crustal accretion is a cyclic process, accompanying periodic influxes of melt from the mantle to a crustal melt reservoir.