The Cenozoic Kenya rift valley has developed in an area of great lithospheric complexity, partly related to late Proterozoic orogenic events, which has strongly influenced the regional patterns of faulting, subsidence, uplift and magmatism. The mantle sources of the magmas are poorly understood, but seem to be heterogeneous as regards trace elements and isotopes, and to have undergone one or more depletion-enrichment events. Mafic magmas erupted in the rift zone may have both asthenospheric and lithospheric components, but the evidence is equivocal. Uncertainties about geothermal gradients and the abundances and speciation of volatiles in the melting zone make it difficult to constrain the depths and degrees of partial melting. Partial fusion may, however, have taken place over the depth range 100+ to 23 km. The largest volumes of basalt last equilibrated at pressures <l5 kb and have trace element characteristics consistent with generation in the spinel-garnet peridotite transition zone. The thickness and internal structure of the lithospheric mantle and attempts to interpret rift evolution in terms of lithosphere-asthenosphere interactions are poorly constrained. Evidence in favour of active (plume-driven) rifting includes the long wavelength gravity and topographic anomalies, the volume of eruptive rocks, the relative timing of magmatism and rifting, the distribution of geophysical anomalies, and the pattern of the tectonomagmatic evolution of the rift. A passive role for the asthenosphere is suggested by the continent-scale correlation between episodes of rifting and magmatism and by fault geometry. Active and passive mechanisms may have acted together in rift development.