Studies of P-wave travel times at a station within 0.5 km of the ground rupture associated with the Parkfield, California, earthquake of 1966 show no systematic variations for a time period of 7 months before the mainshock to at least 13 months after the event. Sources used include quarry blasts, regional earthquakes, explosions at the Nevada Test Site (NTS), and teleseismic earthquakes. The data from the quarry blasts and regional earthquakes have a scatter of less than ± 0.15 sec. With suitable source corrections, the scatter in the NTS data can be reduced to about ± 0.25 sec (making the catalog of nuclear explosions potentially useful for monitoring large travel-time changes). The data from teleseismic P waves have much more scatter than do the data from the more local sources.
The regional earthquake data (expressed as the time differences between the station near the ground rupture and one farther to the north) show temporal variations, but these variations appear to be due to systematic changes in the hypocentral locations of the sources rather than changes in the seismic velocity near the recording stations. The quarry blasts are not as subject to this bias and consequently are more reliable for the monitoring of seismic velocity changes.
The negative results of our study do not rule out the possibility that a velocity anomaly was associated with the Parkfield earthquake; they do, however, require that any velocity change as large as 15 per cent be confined to a volume that is either less than about 5 km deep by several kilometers wide or that does not coincide spatially with the rupture zone.