Early in May, NASA earth scientists monitoring infrared images of the earth noticed unusual patterns in southwestern China. One sent an email to colleagues, noting: Something is happening in Sichuan province.

For Friedemann Freund, a chemist-turned-NASA geophysics researcher, it was more support for his simple, though hotly contested theory: Earthquakes are the culmination of drawn-out physical processes that can be tracked sometimes more than a week ahead of the main event.

The main idea: Rocks put under enough pressure -- for example, when tectonic plates shift -- turn into batteries. The resulting electrical currents can travel miles into the earth, Dr. Freund says. The infrared images observed by NASA, for example, were concentrated several hundred miles from the epicenter of the roughly 8.0 magnitude earthquake that struck on May 12, killing at least 34,000 people.

Dr. Freund describes his discovery as simple, made at 2 p.m. on a Friday afternoon in early 2005 just before he and his graduate students finished packing up a temporary laboratory they had been using. For experiment No. 167, one for the road, they decided to use a copper contact to test whether a squeezed rock emitted a current. It did.

"This is something that should have been discovered 50 years ago," he said.

Certainly, people have tried. For more than a century, researchers have debated the pursuit of the "holy grail" of earthquake prediction. There is still no widespread support for linking electromagnetic signals, infrared emissions or atmospheric changes to an approaching quake.

Dr. Freund's conclusions are questioned by many in the field, who say it isn't clear if the rocks in his laboratory are a reliable model of the properties beneath the earth's surface. Critics also charge that he hasn't done enough to prove that currents, if they do exist, travel through wet rock or for long distances.

"At this point that is largely in the realm of speculation," said Mike Blanpied, the associate coordinator of the U.S. Geological Survey's Earthquake Hazards Program. The USGS funds around 100 earthquake research projects but not Dr. Freund's work.

Dr. Blanpied adds, however, that he hasn't entirely ruled out the possibility of some kind of early-warning system for earthquakes. "Even if the chances are relatively small, it's so important that it's worth pursuing," he said.

Stephen Park, who led a team monitoring the Parkfield section of the San Andreas fault when a 6.0 magnitude earthquake hit in 2004, said what he saw there was "nothing." Some signals occur regularly and are tempting to attribute to earthquakes in retrospect, said Dr. Park, a research geophysicist at the University of California at Riverside.

But here is a mind-bending problem: The inability to track earthquakes means it is hard to get evidence that would help create an earthquake tracker. Early-warning signals could be more apparent at big earthquakes, but scientists would have to get themselves and their on-the-ground equipment to the right place at the right time. "If we could predict that, we wouldn't need to run the experiment," said Dr. Park.

That doesn't discourage Dr. Freund, who is 75 years old. He has invested $1 million of savings held by him and his wife, the novelist Hisako Matsubara, in his research, he said. Their son, Minoru Freund, is the director of the Center for Advanced Aerospace Materials and Devices at the NASA Ames Center in nearby Mountain View, Calif., and has worked with his father on his theories. The pair hope that satellites could help them take their theories to the next step.

Dimitar Ouzounov, the NASA scientist who sent the Sichuan email, was studying thermal patterns as part of his work on atmospheric changes ahead of earthquakes. His research competes with that of Dr. Freund but is predicated on the same concept: that it could be possible to warn that quakes are coming.

Existing earthquake-warning systems rely on seismic tracking of shaking that has already begun, giving, at best, hours of warning. The only formal predictions of earthquakes are probabilities based on historic and geographic data.

Other, more anecdotal evidence of a quake's approach, such as unexplained lights in the sky, migraines, strange animal behavior and water changes might be explained by Dr. Freund's electronic-charge theories.

"I believe he's onto something," said Nevin Bryant, head of cartographic applications at NASA's Jet Propulsion Laboratory in Pasadena, Calif. "In all earthquakes where the weather has been kind to us, we see anomalous numbers" in satellite images of the earth, he said.

The earth-science division at NASA Ames Center, where Dr. Freund also works on his projects, agrees. "There's a sense that at least there's a strong plausibility in what Friedemann's proposing," said Steve Hipskind, director of the division. Dr. Freund acknowledges his theory needs refining and says it needs more scientific and financial support to get there. He hopes to raise as much as $200 million to launch more orbiting satellites to track changes at the earth's surface.

There are other hurdles to turning Dr. Freund's findings into a full-fledged early-warning system, such as: How do you distinguish between a small rumbling and a large earthquake?

"There are hundreds of problems that need to be addressed vigorously," he acknowledges. But he is confident they will be resolved. "I know and, well, I have bet almost everything I own to carry this through."