Data used for the present study were recorded at the small aperture cross-linear array station which was installed at Kaptagat (in NW Kenya) by the University of Durham. The seismic array data from local earthquakes have been analysed by velocity/azimuth filtering technique. Apparent velocities and azimuths for first and later arrival phases were measured for local rift events from the immediate east, for local events from the south west and for more distant rift events to the north and south of Kaptagat. Data from local rift events originating from the immediate east of Kaptagat were used in the present analysis to study the structure of the lithosphere beneath the Gregory rift at about 0.5 N latitude. The first arrival data (apparent velocities and azimuths) were determined to a high degree of accuracy. The first and later arrival data have been interpreted in terms of a simple two layer model with a horizontal refracting interface at a depth of 13 + 5 km and having upper and lower layer uniform velocities of 5.8 + 0.2 km/ s and7.2 + 0,2 km/s respectively. The minimum lateral extent of the top surface of this refractor is estimated at about 30 km. A maximum dip of about 6 on the interface is allowed by the data. In the preferred three layer model, a 10 km thick top horizontal layer of velocity 5.8 km/s overlies a 10 km thick intermediate layer in which velocity increases uniformly from6.0 km/s at 10 km depth to 7.5 km/s at a depth of 20 km. The intermediate layer, in turn, overlies a 7.6 km/s refractor. The models derived from the present data are consistent with the theory that upward perturbation of the lithospher asthenosphere boundary giving rise to domal uplift, lithospheric tension and magmatic activity, is the primary causeof rifting.