Using a repeatable airgun source we have monitored the travel times of several seismic phases along 13- and 18-km baselines from Bouquet Reservoir to boreholes on the San Andreas fault near Palmdale, California. The first arriving phases refract at a depth of 2-3 km, while one later arrival is possibly a reflection from 10 km. The majority of some 200 travel time determinations for the most prominent phases were made between August 1979 and March 1980; isolated measurements were made in November 1978 and March 1981. There exists ample geodetic and other evidence for significant strain changes during this period. Allowing for source variation due to fluctuating reservoir water levels, we observed no travel time changes to the limit of our precision of ¿1-2 ms for the first arrivals and ¿3-5 ms for secondary arrivals, or about 0.1% of the total travel time. An earlier study, which monitored rays penetrating only 500 m into the crust, found that travel times varied by up to 0.3% over a period of several days. The variations showed some correlation with strain changes due to local seismic activity and were not permanent. To understand this different behavior of crustal rocks between 500 and 2000 m, we investigate the depth dependence of their velocity-strain sensitivity. Using measured P and S velocity profiles to gauge the rate at which confining pressure increases velocity, we find that 1 bar of overpressure at 500 m can vary the P velocity by 0.1%, while 1 bar at 2000 m will vary the velocity by only 0.01%. Applying the O'Connell-Budiansky physical model for cracked rock to the velocity depth profiles, we can further characterize the local crust as having a high crack density, a mobile fluid component, and an average crack aspect ratio of 2.5¿10-3. The combination of a mobile fluid content and high crack density suggests that fluids play a role in the transient velocity anomalies observed at a depth of 500 m an argues against significant undersaturation phenomena at seismogenic depths on the San Andreas fault.