A mysterious flash of light detected by a space satellite over the South Atlantic last Sept. 22 was "probably not" caused by a nuclear explosion as initially believed, a panel of scientists appointed by the White House reported yesterday.

In a conclusion that still left many questions unanswered and was at odds with the findings of other in the government, the scientists said it is more likely that the flash detected by the satellite was light reflected from debris from the satellite after it was struck by a small meteoroid.

The scientists said they cannot be sure exactly what caused the signal detected by the Vela satellite. But whatever the cause, they said that their eight-month investigation found "sufficient internal inconsistency to cast serious doubt whether that signal originated from a nuclear explosion or in fact from any light source not in the proximity of the Vela satellite."

The conclusion of the panel, which was headed by Dr. Jack Ruina of the Massachusetts Institute of Technology, is not likely to end the internal government division over whether a nuclear device was or was not exploded over the South Atlantic last Sept. 22.

On Monday, Pentagon sources said that the Defense Intelligence Agency had reached precisely the opposite conclusion as yesterday's report from the White House panel -- that the flash probably was from a clandestine nuclear explosion.

A White House official who made the panel's report public yesterday said the dispute within the government will probably continue. "A lot of people remain convinced that this was a nuclear explosion," he said.

The initial belief of most government experts was that the flash picked up by the satellite was a nuclear explosion. There was speculation that South Africa, the closest country to the site of the activity, had conducted a nuclear test, and later there were stories that Israel had exploded an atomic bomb in the region.

South Africa and Israel, however, denied conducting any nuclear tests and government officials were puzzled by the absence of other evidence of a nuclear explosion -- the detection of fallout, for example.

As a result, the White House convened the panel of nine scientists to investigate.

The scientists said their conclusion that the signal detected by the satellite was probably not caused by a nuclear explosion was based on three major findings:

The signal strongly resembled signals detected by satellites from nuclear explosions, but also differed from those other signals "in a very significant way." That difference suggested that the origin of the signal was close to the satellite, and not near the surface of the earth where nuclear tests are conducted.

The satellite's detection devices, known as "bhangmeters," have recorded hundreds of thousands of signals of nonnuclear origins. In a few cases, signals triggered by unknown causes known to scientists as "zoo events" have had some of the characteristics of signals set off by nuclear explosions.

A search for other evidence of a nuclear explosion has turned up only "ambiguous" data. "At this date, there is no persuasive evidence to corroborate the occurrence of a nuclear explosion on Sept. 22," the panel said.

The scientists said they examined a number of alternative explanations for the flash detected by the satellite and discarded most of them. They said the possibility that the satellite was struck by a meteoroid, and event they said could be expected to occur about once in a decade, "appears to be the best candidate for a nonnuclear origin of the signal."

White House officials said investigations of the signal will continue and that if the source was a nuclear explosion it may be possible to pinpoint its location near the coast of Antarctica rather than in the large expanse of the South Atlantic studied so far.