Before the quake: detecting the slow groan.

Earth typically gives some clue before unleashing its fury. People threatened by hurricanes, volcanic eruptions, and other natural disasters receive anywhere from a few minutes to several days of warning, often enabling them to seek safety. Yet the planet extends no such courtesy in the case of earthquakes, which catch victims completely unawares and have thus far eluded scientific attempts at prediction.

A pair of seismologists now reports finding evidence that some large earthquakes start off slowly, building up for several minutes before letting loose violent vibrations. While the discovery does not provide a direct means of forecasting tremors, it yields new and puzzling information about the birth of earthquakes -- a topic scientists must address before they can even hope to predict damaging shocks.

"We're looking at something new in terms of physics. With the data we have, we're facing some new type of earthquake dynamics that is not understood," says Pierre F. Ihmle of the Institut de Physique du Globe in Paris. Ihmle and Thomas H. Jordan of the Massachusetts Institute of Technology described their earthquake search in the Dec. 2 SCIENCE.

Ihmle and Jordan studied the low-frequency vibrations recorded from 107 shallow-depth earthquakes around the world. In 20 cases, they found indirect

On a map of Earth's tectonic plates, shaded diamonds show 19 oceanic earthquakes that started slowly. A 1990 jolt in southern Sudan (shaded triangle) was the only continental earthquake that displayed similar precursory features. evidence of large energy releases -- called slow earthquakes -- starting hundreds of seconds before the jerky, high-frequency vibrations that traditionally define the start of an earthquake. Scientists have identified isolated examples of slow precursors in the past, but this is the first systematic search, Jordan says.

If seismologists could catch slow precursors in action, they could -- in theory -- alert endangered populations minutes before the start of damaging quakes. But the feasibility of such a scenario remains unclear.

Because all but 1 of the 20 slow earthquakes occurred on oceanic faults, Ihmle and Jordan cannot tell how their findings relate to continental earthquakes, which pose the most risk to people. If the slow precursors arise from unusual properties of oceanic rock, they may not typically precede tremors on land. In fact, scientists have found no evidence of such slow deformation prior to well-monitored earthquakes in California.

Ihmle and Jordan call their conclusions tentative because they have not been able to see the slow precursors directly on seismograms of the 20 earthquakes they studied. In fact, they delayed reporting their results for several years in order to rule out alternative interpretations of the data.

William L. Ellsworth of the U.S. Geological Survey in Menlo Park, Calif., describes these findings as tantalizing. "It reinforces the possibility that there are short-term processes that are occurring immediately before the earthquake propagates down the fault," says Ellsworth. But he adds that "seismologists are always going to be cautious until they can see things in the actual seismograms."