Shear-wave splitting with 20 to 120 msec time separation is observed on 3-component seismograms from earthquakes occurring at depths of 5 to 18 km in the crystalline basement beneath the Los Angeles basin. Shallow events that occur in the basement at the base of the 5-km-thick sedimentary section exhibit little splitting, while deeper events show progressively greater splitting with depth. The polarization direction of the fast shear wave is N-S, independent of the azimuth of ray path between the event and station. We interpret that the shear-wave splitting occurs mainly in the crystalline basement and is the result of vertical crustal microcracks aligned in N-S direction. Under the assumption that the alignment of microcracks is due to the subsurface stress regime, the results from the shear-wave splitting data yield a maximum principal stress direction of N-S ± 15° at depth beneath the northern Los Angeles basin, consistent with results from geological mapping and fault-plane solutions. Ray trace modeling indicates that the observations of shear-wave splitting can be explained in terms of an anisotropic basement containing vertical microcracks aligned in the N-S direction and with a crack density of ∼0.04, overlain by a weakly anisotropic sedimentary section with a crack density less than 0.02. During the last 4 yr, three major earthquakes occurred in the Los Angeles basin area. They are the M 4.6 and M 4.4 Montebello double earthquakes of 1989, the M 5.5 Upland earthquake of 1990, and the M 5.8 Sierra Madre earthquake of 1991, with epicenters of 10, 55, and 40 km away from the recorder site SCS located in the northern Los Angeles basin, respectively. We did not observe significant variations of shear-wave splitting at SCS before and after the Upland and Sierra Madre earthquakes. However, we found fairly rapid variations of shear-wave splitting following the Montebello events. The time separation between the split shear waves sharply decreased from 70 msec to less than 20 msec after the mainshock and then returned to the average level gradually. This rapid decrease in shear-wave splitting may be a temporal variation or a result of location variations of the Montebello events that were located approximately where the line singularity is expected.