A generic, three-dimensional (3-D) model has been developed which explains the three-dimensionality exhibited by magnetovariational (MV) and magnetotelluric (MT) data from southern Kenya. In this model, observed variations in electromagnetic strike with period and location, impedance phase splitting, and peaks in tipper magnitude are all understood in terms of two regionally two-dimensional (2-D) structures striking NW-SE and N-S, respectively. The observed period and location dependence of the electromagnetic strike may arise as an indirect consequence of a rotating stress field, with regional-scale structures formed at different stages of the stress-strain history of Kenya being preserved as conductive lineaments. These conductive structures are not all confined to the upper crust. Thus, whereas stress data provide constraints on rifting at the upper crustal scale, the MT impedance tensor data provide constraints at lithospheric scales. The constraints and resolution provided by the MV and MT data have been rigorously investigated using 3-D forward modeling. Decoupling of the period and site dependence of electromagnetic strike aids resolution and constraint of conductors, rendering attempts to fix an average strike in space and frequency inexpedient. An anomalous apparent strike at the center of the Rift Valley reflects neither the N-S strike of the rift, nor the NW-SE striking shear fabric, but is shown to be a virtual strike arising as a result of coupling between the respective strike directions. The NW-SE trending conductivity anomaly straddles the rift and both its flanks, extends to at least middle to lower crustal depths, and appears substantially more electrically anomalous than the rift itself. A hypothesis that melt exists in the mantle directly below the rift at latitude 1.8 ¿S is not supported by the MT data. ¿ 2000 American Geophysical Union