We study the spatial clustering of shallow aftershock hypocenters with respect to focal mechanisms of mainshocks. We use the Harvard centroid moment tensor (CMT) global catalog, the Preliminary Determination of Epicenters (PDE) earthquake list, the California Institute of Technology/U.S. Geological Survey catalog of earthquakes in southern California, and a catalog of focal mechanisms for all earthquakes since 1850 in southern California with magnitude larger than 6. We need to account for possible systematic bias in hypocenter distribution due to the geometry of seismogenic zones, especially that of subduction zones. We also select only strike-slip earthquakes from the catalogs to investigate aftershock clustering in circumstances more favorable for direct observation. We compare the spatial distribution of hypocenters before each strong earthquake with the distribution during the first 250 days after the earthquake and for the time interval extending beyond 250 days. If the friction coefficient in the Coulomb criterion is positive one expects that after a strong earthquake, aftershocks and other earthquakes would concentrate in the direction of the P axis (dilatational quadrant) rather than in the direction of the T axis (compression quadrant). Such correlations have been pointed out previously for selected earthquakes sequences, but is such correlation a general feature of earthquake occurrence? We study spatial earthquake distributions before and after each event for several choices of focal sphere partition, cutoff magnitude, focal mechanisms of large events, time periods, distance from a mainshock, etc. Although some earthquake distributions agree with a nonzero friction coefficient, others produce the opposite pattern, suggesting that the concentration of events along the P and T axes is due to random effects. This result implies that the friction coefficient in the Coulomb law is close to zero. ¿ 1998 American Geophysical Union