Carl Johnson, 35, a geophysicist, wakes at 4 a.m. at least once a week to worry about earthquakes.

Sometimes he just gives up on sleep and goes off in the pitch dark to his office at the California Institute of Technology.

"He's thinking there may be something he can do to predict the first big earthquake," says his wife, Nancy.

Chuck Koesterer, 32, an electronics technician, has found his life moving with the erratic rhythms of the Pacific plate, the huge piece of the earth's crust sliding ponderously up the coast of his native California. It's his job to get government sensors to the spot of a major quake in the mountains or deserts, so every large temblor sets off his electronic beeper.

"It always seems to happen in the middle of the night," said his wife, Sheryl.

Johnson and Koesterer, with hundreds of other geologists, mathematicians, technicians, graduate students and psychics up and down this coast, are engaged in one of the great scientific races in American history, to find a way to predict earthquakes before the next great quake devastates a major city in California.

The thrill of the chase, which intrigues nearly everyone living here, is enough to make many decline much better paying jobs in the oil industry.

The first successful earthquake prediction in the United States, in the view of some scientists, occurred eight years ago in the Adirondack Mountains of New York. A team from Columbia's Lamont-Doherty Geological Observatory predicted a magnitude three quake with fairly precise time and location after measuring wave velocities from a series of tiny foreshocks.

Unfortunately, the same technique later proved "an abysmal failure" in California, Johnson said.

Chinese scientists claim to have predicted large tremors and saved many lives, but such success has eluded American scientists. The last death from an earthquake in the United States occurred here 10 years ago, but the numbers of small quakes in southern California have increased recently and there are enough other signs of unusual underground activity to put earthquake watchers under some pressure.

Johnson, head of the U.S. Geological Survey field office at Cal Tech, is trying to computerize the earthquake data in a way that will provide new clues.

Earthquake scientists come from many fields, such as Cal Tech researcher Kate Hutton, an astronomer lured by the excitement of the field and its intriguing mathematical puzzles.

And some are not scientists at all, but earthquake buffs who volunteer services to the cause.

Robert Parsons, 45, a telephone company transmission technician, said he has invented a seismic device that emits a high-pitched tone within 24 hours of a major quake. In the northern California town of Carmel, Clarisa Bernhardt said she has visions of a photo, or a calendar with the date circled, and the word earthquake stamped across it, and has successfully foreseen earthquakes this way.

In recent years, scientific research on earthquake prediction has focused on history and happenstance. Scientists have spent considerable time recording all quakes in recent history, just to see if there is a pattern that might yield clues to their causes.

Johnson and Hutton, looking at the pattern of recent quakes and the way they come sometimes in complex "swarms," reported to the American Geophysical Union in San Francisco recently that a system of faults near the Salton Sea thought to have been dormant has shown the potential for damaging quakes.

Cal Tech assistant professor Kerry Sieh has unearthed old fissures that indicate the last great southern California earthquake in 1857 was part of a series recurring every 123 to 225 years.

According to one U.S. government estimate based on Sieh's work, that quake has a better than 50 percent chance of recurring in the next 30 years, and could kill more than 13,000 southern Californians, leave 100,000 homeless and cause $15 billion in damage.

The fascination with earthquakes among researchers here tends to extend to their family and social lives. "I would prefer to talk about something else occasionally," said Nancy Johnson, who teaches the physically handicapped in East Los Angeles.

On one occasion, local reporters panicked when an earthquake occurred and no one could be found at the Cal Tech center to give its location and magnitude. Hutton says all the scientists had been at a party that had gotten so lively they had failed to feel the quake.

The Cal Tech field office has since added its beeper system so Hutton, Johnson and Koesterer are immediately alerted to anything over magnitude three, a mere rattling of teacups. A buzzer alarm sounds at the second-floor Cal Tech field office if one of the sensors scattered about southern California picks up such a quake.

Several universities and the U.S. Geological Survey have scattered mechanical earthquake detectives all over the state--seismometers, strainmeters, radon meters, tiltmeters, magnometers and creep meters.

Some scientists think the measurements of underground radon gas, on a recent upswing, may signal approaching quakes. But none of the sensors have yielded consistent clues. With so few large quakes, they can only be put to a test sporadically.

Earthquake scientists have already proved the value of seismometers, which measure vibrations in the earth, by their accurate prediction of eruptions at Mount St. Helens. Volcanoes, however, are far easier to predict because, unlike earthquakes, their precise location is usually known in advance and seismometers can be placed in the best positions.

Ned North, 64, and Robert Parsons, 45, have ignored the drawbacks and formed their own company, Earthquake Sentry. They say they can predict quakes over over magnitude 5 anywhere in southern California within 24 hours.

Since there has been no quake of that magnitude on land in the area in the two years they have been forecasting, North said, "Our prediction has been, 'No earthquake.' " He said they have predicted quakes outside the area, however, and feel the government should be paying more attention to their device.

Parsons said he is reluctant to show his earthquake-predicting device to a Cal Tech expert, or patent it, for fear his secret to improving ordinary seismological sensors will be stolen. North, a former Los Angeles fire commission member working out of a tiny office in Huntington Beach, has written dozens of companies, offering the earthquake forecast service to the first 10 major corporate subscribers at $42,000 each.

He advises companies who buy the service that "once they are made aware that a damaging earthquake will occur, they should not issue any public warning. This is strictly and privately for their own use."

Parsons and North have gotten no offers. Company executives seem convinced by scientific evidence that major quakes are so far very difficult to predict. But that does not stop the friends of California earthquake researchers from hoping for inside information.

"My friends always ask me, 'Let me know if something is going to happen,' " Koesterer said. His wife said friends at the basement ophthalmology clinic where she works tell each other, "When Sheryl starts running upstairs, we'll know."

The Koesterers have told their children how to crawl under the table in case of a quake, and have told Sheryl's grandmother, who stays with them, how to turn off the gas and water. They have candles, flashlights, bottled water and some canned goods ready for use.

Carl and Nancy Johnson, on the other hand, admit that, like most Californians, they have made hardly any preparations.

"Basically, we consider ourselves lucky to get the laundry done," says Nancy Johnson.