Abstract In February 1994, the Kenya Rift International Seismic Project carried out two wide-angle reflection and refraction seismic profiles between Lake Victoria and Mombasa across southern Kenya. Our investigation of the data has revealed evidence for the presence of two upper mantle reflectors beneath southwestern Kenya, sometimes at short range, from seven shotpoints. Two-dimensional forward modelling of these reflectors using a pre-existing two-dimensional velocity-depth model for the crust [Birt, C.S., Maguire, P.H.K., Khan, M.A., Thybo, H., Keller, G.R., Patel, J., 1997. The influence of pre-existing structures on the evolution of the Southern Kenya Rift Valley — evidence from seismic and gravity studies. Tectonophysics 278, 211–242], has shown them to lie at depths of approximately 51 and 63 km. The upper reflector, denoted d 1, shallows by about 5–10 km in the area beneath Lake Magadi, situated in the rift itself. Correlations for the deeper reflector, denoted d 2, are sparse and more difficult to determine, so it was not possible to define any shallowing corresponding to the surface expression of the rift. Only limited control exists over the upper mantle velocities used in the modelling. Immediately beneath the Moho we use a value of P n calculated from the crustal model, and constraints from previous refraction, teleseismic and gravity studies, to determine the velocity at depth. At the d 1 reflector a reasonable velocity contrast was introduced to produce a reflector for modelling purposes. Beneath the d 1 reflector the velocity decreases to the average value over 3 km. Beneath the rift the velocity also rises across d 1 and again, decreases to the average value over the next 3 km. At the d 2 reflector a similar model is used. This model accounts for the presence of the mantle reflectors seen in the data by using layers of thin higher velocity in a lower background velocity. Due to the uncertainty in the velocities the absolute position of both d 1 and d 2 could vary, but the relative upwelling beneath the rift is reasonably well constrained and data from four different shotpoints which indicate the shallowing show good agreement. A significant result of this study is that the continuity of the d 1 reflector indicates that the sub-Moho lithosphere has not been substantially disrupted by mantle upwelling, even though probably thinned and stretched.