The southward propagation of the East Africa rift presents an opportunity to study plate boundary formation. We tabulate orientation data which confirm the province of NW-SE directed most compressive horizontal principal stress ("Wegener stress anomaly") earlier tentatively attributed to ridge push. We also collect information on stress "regime," described by the associated Andersonian fault type(s). We use thin shell finite element models with realistic rheology to test three causes of stress: (1) lateral variations in density moment, (2) resistance of unbroken lithosphere to relative plate rotation, and (3) stress concentration ahead of a crack tip. Models with stress due primarily to variations in density moment are unsuccessful in their predictions (59--73% incorrect regimes; 32--40¿ azimuth errors). Models in which Africa-Somalia spreading is regulated at realistic rates by remote boundary conditions are more accurate (18--41% incorrect regimes; 25--35¿ azimuth errors). Treating the East Africa rift as a frictionless crack degrades the fit in either case. Apparently, the Wegener stress anomaly is caused primarily by resistance to the relative rotation between the Somalia and Africa plates. The East Africa rift north of 21¿S may be weakened by strain but has residual friction ≥0.1. Greater strength of oceanic lithosphere is likely to cause stress increases, reorientations, and regime changes offshore. The predicted strain rate map has high rates along the rift, curving at 12¿S into a western arc through Angola-Namibia-South Africa. Seismic hazard in Namibia may be greater than the instrumental catalog suggests. However, a number of unfit data indicate that these models represent only a first step.