SANTA BARBARA, Calif. (AP) _ The southern San Andreas Fault is far more complex than once believed, muddying researchers' efforts to predict when a disastrous earthquake will rock Southern California, studies show.

Scientists often say a great quake measuring 8 on the Richter scale is at least 50 percent likely within 30 to 50 years on the southern San Andreas.

Research presented last week during the Seismological Society of America's annual meeting doesn't necessarily change that timing, but is ''raising the uncertainty of our knowledge,'' said Gordon Jacoby, of Lamont-Doherty Geological Observatory in Palisades, N.Y.

It also increases uncertainty about whether Southern California's next big temblor will be a ''great'' quake, measuring 8 or above on the Richter scale, or one or more magnitude 7 ''major'' quakes, said Jacoby, California Institute of Technology geologist Kerry Sieh and U.S. Geological Survey seismologist Lucile Jones.

Federal officials estimate a great quake would kill up to 14,000 people, seriously injure up to 55,000 and cause $17 billion in damage.

The San Andreas Fault snapped in great quakes before, both prehistorically and in 1857 in Southern California and 1906 in San Francisco.

Jones said it inevitably would rupture violently again as the gigantic Pacific plate of the Earth's crust drags the Pacific ocean floor and much of Southern California northwest past the North American plate at 3 inches a year.

Stress builds until a fault breaks in a quake, then rebuilds until another quake. But the 650-mile fault, which stretches from northwest of San Francisco to southeast of Los Angeles, doesn't break at once.

The fault is divided into segments. Stress accumulates and is released during quakes at different times on different segments, although adjacent segments can rupture in a single larger quake.

Long-term predictions of when a major or great quake will happen are based on the theory that certain segments produce big quakes at characteristic time intervals.

Studies by Jacoby and Long Beach City College geologist Howard Shifflett suggest two segments of the fault overlap, producing big quakes less predictably than was thought, and that one segment consists of smaller subsegments, which would produce smaller quakes if they broke separately instead of together.

''It throws a monkey wrench into our canned explanations of how faults work,'' Sieh said.

''We're no longer looking for the quick, simple answers,'' Jones added. ''Earth is just (a lot) more complicated than that.''

Geophysicists believe stresses are near the breaking point on the two southermost parts of the fault: the Mojave segment, which starts near Tejon Pass north of Los Angeles and runs southeast to Cajon Pass, north of San Bernardino, and the Indio segment, which starts at Cajon Pass and stretches southeast to the Salton Sea.

Sieh's pioneering research in recent years found the Mojave segment generates a big quake roughly every 145 years, the last time in 1857, when several segments to the northwest also snapped. The Indio segment hasn't produced a big quake in California's recorded history, but Sieh found prehistoric big quakes occurred every 200 or 250 years.

So the Mojave segment is considered ripe for a big quake, and the Indio segment is considered overdue.

Jacoby studied annual growth rings from trees directly on what was thought to be the Mojave segment of the fault at Wrightwood, near Cajon Pass. He found trees were stunted by a big quake in December 1812, only 44 years before the great January 1857 quake.

Because a 44-year interval beween big quakes is far less than Sieh's repeat times of 145 years for the Mojave segment and 200 to 250 years for the Indio segment, Jacoby's study suggests both segments overlap at Wrightwood, so a quake on either segment shakes the area. But it also is possible Wrightwood is on a separate segment.

Either possibility casts doubt on the predictability of future big quakes by questioning the previous concept that the Mojave and Indio segments are distinct and produce quakes at predictable intervals.

Shifflett studied a 7-foot-diameter circle of stones which Indians built on the Indio segment near the Salton Sea about 1,000 years ago. As quakes made one side of the fault slip past the other, the circle was split into two semicircles.

Yet two of four big prehistoric quakes Sieh detected farther northwest on the Indio segment didn't change the stone circle, meaning those two quakes didn't rupture the whole Indio segment. That suggests the segment consists of subsegments that might break separately in major quakes, instead of together in a great quake.

The Richter scale is a measure of ground motion as recorded on seismographs. Every increase of one number on the scale means a tenfold increase in magnitude. Thus a great quake with a reading of 8 reflects an earthquake 10 times stronger than a major quake with a 7 reading.