Seismological observations suggest the inner core has been rotating faster
than the mantle during the past 30-50 years. Differential rotation can prod
uce large gravitational torques on the mantle which, if not balanced by oth
er torques, would be detectable as changes in the rotation;rate of the mant
le (Length of Day or LOD). We compare seismological estimates of inner core
rotation with estimates derived by inverting LOD observations using models
of gravitational coupling. Model predictions depend on several parameters,
but for the expected range of parameter values, the seismological estimate
s of relative rotation are at least an order of magnitude larger than those
predicted by the LOD data. Furthermore, the LOD data predict oscillation i
n the angular alignment of the inner core and mantle during the past 30-50
years, while the seismological data seem to require a component of steady r
otation. We show that steady rotation of the inner core implies a steady to
rque on the mantle. This torque is not accelerating the rotation rate of th
e mantle, so it must be opposed by another torque. The required torque is c
onsistent with a frictional electromagnetic stress on the base of the mantl
e due to a westward flow at the top of the core.