Melt generated by mantle upwelling is fundamental to the production of new oceanic crust at mid-ocean ridges, yet the forces controlling this process are debated1,2. Passive-flow models predict symmetric upwelling due to viscous drag from the diverging tectonic plates, but have been challenged by geophysical observations of asymmetric upwelling3,4,5that suggest anomalous mantle pressure and temperature gradients2,6,7, and by observations of concentrated upwelling centres8consistent with active models where buoyancy forces give rise to focused convective flow2. Here we use sea-floor magnetotelluric soundings at the fast-spreading northern East Pacific Rise to image mantle electrical structure to a depth of about 160 kilometres. Our data reveal a symmetric, high-conductivity region at depths of 20–90 kilometres that is consistent with partial melting of passively upwelling mantle9,10,11. The triangular region of conductive partial melt matches passive-flow predictions, suggesting that melt focusing to the ridge occurs in the porous melting region rather than along the shallower base of the thermal lithosphere. A deeper conductor observed east of the ridge at a depth of more than 100 kilometres is explained by asymmetric upwelling due to viscous coupling across two nearby transform faults. Significant electrical anisotropy occurs only in the shallowest mantle east of the ridge axis, where high vertical conductivity at depths of 10–20 kilometres indicates localized porous conduits. This suggests that a coincident seismic-velocity anomaly12is evidence of shallow magma transport channels13,14rather than deeper off-axis upwelling. We interpret the mantle electrical structure as evidence that plate-driven passive upwelling dominates this ridge segment, with dynamic forces being negligible.