Changes in shear wave splitting are observed at KNW before and after the M = 6 North Palm Springs earthquake of July 8, 1986. KNW is a station of the Anza seismic network monitoring the Anza seismic gap on the San Jacinto fault, southern California. The gradual increase in the delays between the split shear waves over 3 years at KNW, reported by Peacock et al. (1988), ended in June 1986. The further 2 years of observations analyzed here show that the behavior of the delays changed abruptly near the time of the North Palm Springs earthquake, 33 km north of KNW. Peacock et al. demonstrated that the increase in delays could be simulated by increasing the aspect ratio of stress-aligned fluid-filled inclusions, and speculated that this increase might be the result of a build up of stress before an impending earthquake. The new data appear to confirm this speculation, but the temporal variations require a more complex interpretation, although they still suggest that the changes in shear wave splitting are caused by earthquake-induced stress changes to the fluid-filled inclusions throughout the rockmass. Central to our interpretation of temporal changes in shear wave splitting is the well established existence, throughout at least the uppermost 10 to 20 km of the crust, of small fluid-filled cracks, microcracks, and pores. The existence of such inclusions introduces a compliant quality to otherwise stiff crustal rock. We term these distributions of inclusions extensive dilatancy anisotropy or EDA, and the individual inclusions EDA cracks because, although they may include a wide range of shapes, many of the seismic properties can be simulated by distributions of thin parallel cracks. We present a further 2 years of data that splitting) than seismograms displayed as timeindicate that the shear wave splitting recorded series, and all the measurements of shear wave at KN• appears to change before and after the splitting in this paper are taken from polariza-M = 6 North Palm Springs (NPS) earthquake of tion diagrams. Figure 2 shows rotated seismo-July 8, 1986. Figure 1 shows a map of the Anza grams and horizontal polarization diagrams of a seismic network and location of the NPS earth-number of events. The positions of the arrowquake. The temporal changes display different heads marking the arrival times of the split behavior along different ray paths that place shear waves are those read from polarization constraints on the possible cause of the changes. diagrams. Many of the polarization diagrams in We speculate on how .these variations may be Figure 2 and throughout the data set display interpreted in terms of stress-induced modific-classic patterns of shear wave splitting familiar ations to the cracked rockmass. from synthetic modelling [Crampin. 1978; Crampin and Booth, 1985], so that the analysis is Crampin, S., Seismic wave propagation through a cracked EH9 3LA, Scotland. solid: polarization as a possible dilatancy diag-J.B. Fletcher, U.S. Geological Survey, 345 Middlefield nostic, Geophys. •. R. /ts tr. Soc., 53, 46...