In 1971 researchers hoping to predict earthquakes in the short term by identifying precursory phenomena (those that occur a few days before large quakes but not otherwise) turned their attention to changes in seismic waves that had been detected prior to earthquakes. An explanation for such changes was offered by "dilatancy theory," based on a well-known
phenomenon observed in rocks in the laboratory: as stress builds, microfractures in rock close, decreasing the rock`s volume. But as stress continues to increase, the rock begins to crack and expand in volume, allowing groundwater to seep in, weakening the rock. According to this theory, such effects could lead to several precursory phenomena in the field, including a change in the velocity of seismic waves, and an increase in small, nearby tremors. Researchers initially reported success in identifying these possible precursors, but subsequent analyses of their data proved disheartening. Seismic waves with unusual velocities were recorded before some earthquakes, but while the historical record confirms that most large earthquakes are preceded by minor tremors, these foreshocks indicate nothing about the magnitude of an impending quake and are indistinguishable from other minor tremors that occur without large earthquakes. In the 1980s, some researchers turned their efforts from short-term to long-term prediction. Noting that earthquakes tend to occur repeatedly in certain regions, Lindh and Baker attempted to identify patterns of recurrence, or earthquake cycles, on which to base predictions. In a study of earthquake-prone sites along the San Andreas Fault, they determined that quakes occurred at intervals of approximately 22 years near one site and concluded that there was a 95 percent probability of an earthquake in that area by 1992. The earthquake did not occur within the time frame predicted, however.Evidence against the kind of regular earthquake cycles that Lindh and Baker tried to establish has come from a relatively new ♦field, paleoseismology. Paleoseismologists have unearthed and dated geological features such as fault scarps that were caused by earthquakes thousands of years ago. They have determined that the average interval between ten earthquakes that took place at one site along the San Andreas Fault in the past two millennia was 132 years, but individual intervals ranged greatly, from 44 to 332 years.