Precessional motion of Earth's rotation axis relative to its orbit is a well-known source of long-period climatic variation. It is less well appreciated that growth and decay of polar ice sheets perturb the symmetry of the global mass distribution enough that the geographic location of the rotation axis will change by at least 15 km and possibly as much as 100 km during a single glacial cycle. This motion of the pole will change the seasonal and latitudinal pattern of temperatures. We present calculations, based on a diurnal average energy balance, which compare the summer and winter temperature anomalies due to a 1¿ decrease in obliquity with those due to a 1¿ motion of the rotation pole toward Hudson Bay. Both effects result in peak temperature perturbations of about 1¿ Celsius. The obliquity change primarily influences the amplitude of the seasonal cycle, while the polar motion primarily changes the annual mean temperatures. The polar motion induced temperature anomaly is such that it will act as a powerful negative feedback on ice sheet growth. We also explore the evolution of the coupled system composed of ice sheet mass and pole position. Oscillatory solutions result from the conflicting constraints of rotational and thermal stability. A positive mass anomaly on an otherwise featureless Earth is in rotational equilibrium only at the poles or the equator. The two polar equilibria are rotationally unstable, and the equatorial equilibrium, though rotationally stable, is thermally unstable. We find that with a plausible choice for the strength of coupling between the thermal and rotational systems, relatively modest external forcing can produce significant response at periods of 104--106 years, but it strongly attenuates polar motion at longer periods. We suggest that these coupled oscillations may contribute to the observed dominance of 100 kyr glacial cycles since the mid-Pleistocene and will tend to stabilize geographic patterns that are suitable to glaciations. ¿ 1999 American Geophysical Union