The seismic activity in the western part of the Corinth Rift (Greece) over the period 2000– 2007, monitored by a dense network of three-component stations, is analysed in terms of multiplets and high precision relocation using double difference techniques. This detailed analysis provides new insights into the geometry of faults at depth, the nature and the structure of the active zone at 6–8 km depth previously interpreted as a possible detachment, and more generally into the rifting process. The seismicity exhibits a complex structure, strongly varying along the rift axis. The detailed picture of the seismic zone below the rift indicates that its shallower part (at depths of 6–8 km) is 1–1.5 km thick with a complex microstructure, and that its deeper part (at depths of 9–12 km) gently dipping to the north (10–20°) is 0.1–0.3 km thick with a microstructure consistent with the general slope of the structure. Although the nature of this seismic zone remains an open question, the presence of seismicity beneath the main active area, the strong variability of the structure along the rift over short distances and the complex microstructure of the shallower part revealed by the multiplet analysis are arguments against the hypothesis of a mature detachment under the rift: this active zone more likely represents a layer of diffuse deformation. The geometry of the mapped active faults is not well defined at depth, as no seismicity is observed between 0 and 4 km, except for the Aigion Fault rooting in the seismic layer at 6 km depth with a dip of 60°. A distinct cloud of seismicity may be associated with the antithetic Kalithea Fault, on which the 1909 Fokis earthquake (M s = 6.3) may have occurred. The link between the 1995 rupture (M s = 6.2) and the faults known at the surface has been better constrained, as the relocated seismicity favours a rupture on an offshore, blind fault dipping at 30° , rather than on the deeper part of the East Helike Fault. Consequently, the 1995 event is expected to have decreased the Coulomb stress on the East Helike Fault. To explain these seismic observations along with the geodetic observations, a new mechanical model for the rifting process in this region is proposed, involving non-elastic, mostly aseismic uniform NS opening below the rift axis, coupled with the downward and northward growth of a yet immature detachment: the reported GPS rates would mainly result from this deep, silent source, and the seismicity would reveal the detachment position, not yet connected to the ductile lower crust. In such a model, the strong fluctuations of microseismicity would result from small strain instabilities, undetected by continuous GPS and possibly related to pore pressure transients.