In the sparsely-populated land along the San Andreas fault in California, researchers have been trying to see if they can find the holy grail of seismology -- the ability to predict an earthquake before it does any harm.
Today a group of scientists report they've taken a step in the right direction. In a two-month experiment with instruments lowered into the ground near Parkfield, Calif., they were able to detect minute changes in the Earth's crust -- up to 10 hours before a small earthquake took place.
It's a hopeful sign, but the scientists who did the research caution that it's too early to say they are actually on the way to predicting tremors.
"We're very encouraged, but we want to do a lot more monitoring to confirm what we saw," said Paul G. Silver of the Carnegie Institution of Washington, one of the researchers. He, along with Fenglin Niu of Rice University and three colleagues, published their findings in this week's edition of the journal Nature.
The study earned a place in one of the world's leading scientific journals, mainly because research into earthquake prediction has essentially gone nowhere up until now.
The scientists lowered instruments more than half a mile down two holes that had been drilled into the ground near the San Andreas fault at Parkfield. Using those instruments, they were able to measure movement of the ground between the two holes, the amount of stress in the rocks, and the spread of vibrations in one direction or another.
Twice -- one time two hours in advance of a small earthquake, the other 10 hours -- they thought they saw shifts suggesting the quakes were coming. But they said they could not be sure the changes were directly connected to the tremors that followed.
"Those of us who have been working at this for a while have already been burned once or twice, seeing something we think is connected to the earthquakes that doesn't pan out," said Lucy Jones of the U.S. Geological Survey in an e-mail to ABC News. Jones was not involved in the current experiment.
"We have learned the hard way to be very careful to state the problem clearly," she said, "and in fact look for the ways in which we could be fooling ourselves into seeing a signal that isn't really there."
Part of the problem is that earthquakes are so common, and so complicated, that it's hard to say if a shift in the ground is related to a specific tremor. In California alone, there are some 30,000 earthquakes every year, most of them too small for people to notice.
But scientists have tried for decades to figure out a way to predict major quakes. Even a few seconds' notice, they say, would give people time to protect themselves from harm.
Silver said he and Niu will be back in Parkfield in September, planning to redeploy their instruments for a longer experiment.