LOS ANGELES -- When his friend Steven Kilston brought him a new theory on earthquakes and lunar-solar movements six years ago, UCLA geophysicist Leon Knopoff reacted with amused scorn.

Kilston, an astronomer now working for Lockheed Aircraft Corp., thought he had detected a correlation between the northernmost drift of the moon's orbit and several major California earthquakes. Since Kilston was a friend, Knopoff agreed to put the theory to the test. After several tries, he could not find a significant flaw in Kilston's statistics. Although bothered by the lack of any physical explanation for the link between moon and quakes, Knopoff, after some hesitation, agreed to be listed as co-author of a 1983 article in the British journal Nature describing the connection.

In passing, the article suggested the next likely occasion for an earthquake stronger than magnitude 6 on the Richter scale would be four years on either side of November 1987. Knopoff spent some time squelching journalists who tried to say he was making an earthquake prediction, then went back to his real work, constructing computer models to show the distribution of stress over different fault systems.

On Monday and Tuesday it all came back to haunt him. Two earthquakes, one tentatively magnitude 6.0, the other 6.3, struck within 12 hours of each other in California's Imperial Valley -- in the middle of November 1987.

As the 1983 Kilston-Knopoff article predicted, the two faults apparently affected ran in a northwesterly direction, and occurred at sunset and sunrise, when the sun's position might have the greatest influence on surface stresses.

The game of earthquake prediction has had so many spectacular failures, and drawn so many charlatans, that the few scientists trying to inch their way toward a real science of earthquake forecasting shy away from even an apparent success. Kilston could not be reached for comment, but Knopoff said, "This is just one more statistical count on the side of a weak statistic."

He said he still can see no physical explanation for strong earthquakes occurring during the northernmost track of the moon's 18.6-year cycle of changing orbit positions. But he is still looking at it, he said, in the small amount of time he spends on the problem, and thinks the sunset-sunrise correlation may also have some promise.

And he has a lot of other work to keep him busy before the next major stress point in the theory comes around, sometime about May 2006.