In the past, the influences of large earthquakes on Earth rotation and on changes of the gravity field have been the subject of extensive investigations. Though it is now recognized that earthquakes cannot excite the Chandler wobble and that they only drive a very modest secular drift of the Earth's spin axis, their rotational signatures present some intriguing peculiarities. In fact, they preferentially drive the rotation axis toward a well-defined direction (i.e., ~140 ¿E) and systematically induce negative variations of the Earth's oblateness. In the first part of this work the static dislocation theory is employed to ascertain the origin of these two remarkable trends. This task is accomplished by considering the effects of seismicity on Earth rotation during the last two decades. We find that the above tendencies are operating even on a yearly time scale, and we explain their nature by means of simple geometrical and physical arguments. We also find that the trend of the pole toward 140 ¿E is essentially due to the seismicity which characterizes the western Pacific subduction zones. The seismic excitation function takes for shallow thrust earthquakes a direction roughly opposite to the excitation function associated with mass injection along active subduction zones; this point highlights the potential effects of aseismic mass rearrangement at subduction zones on the secular drift of the pole. Our findings, based on a simple model of steady state subduction, show that in the last two decades this process substantially acts to counterbalance the polar drift due to global seismic activity. We verify that this balance between earthquake- and subduction-induced rotational signatures also characterizes the time variations of the dynamic oblateness, which is intrinsically independent of the inertia changes that drive polar motion. From these results, which may potentially have some impact on the interpretation of long-term polar motion, we conclude that the pole of rotation cannot be significantly perturbed by tectonic motions on a decade time scale. ¿ 1998 American Geophysical Union