On Sept. 13, 1981, in a major speech in West Berlin, Secretary of State Alexander M. Haig Jr. announced to the world that the United States had firm evidence that deadly chemical weapons were being used in Asia.

Only 13 days before his declaration, U.S. analysts had warned the State Department not to make such an announcement. The evidence, they said, could not yet support such a charge.

With that dramatic discrepancy the U.S. government began its campaign to show that "yellow rain," a terrible new poison weapon, was being used by the Soviets and their allies in Southeast Asia.

It was called "yellow rain" because many of the reports of "attacks" referred to droplets of a yellow substance that sounded like rain as it fell on leaves and rooftops. The toxins, called trichothecenes, are made naturally by one variety of fungus. The United States said the Soviets grew the fungus and harvested the poison, which was then given to communist troops in Southeast Asia and used against resistance fighters and Hmong tribesmen in Laos and Kampuchea (Cambodia).

Documents obtained by The Washington Post through the Freedom of Information Act show that while U.S. officials -- including President Reagan on numerous occasions -- actively pursued the charge for two years after Haig's speech, the government was privately accumulating and withholding evidence that suggested the accusation was probably not true.

David Lambert, the State Department's spokesman on the issue, said last week that official U.S. policy remains that the Soviets sponsored lethal chemical warfare in Southeast Asia.

But interviews with U.S. officials who were involved in the controversy suggest that the yellow rain accusation was a case of high-level policy makers, including Haig and Richard R. Burt, then head of the State Department's Bureau of Politico-Military Affairs, taking a shred of early, ambiguous evidence and pressing it into service as a major public relations device too early and too forcefully, making it difficult to retreat from the position when the evidence went against it.

Burt emerges as a pivotal figure in the case because he heard the preliminary evidence and decided to make it a major issue.

In a telephone interview last week from Bonn, where he now serves as ambassador to West Germany, Burt said he has not recently examined evidence for the charges, but believed at the time that the preliminary reports were good. He did not want to wait for confirmation, he said, because "I wanted to get the story out and stop the killing. There are people in government who always urged you to wait a while."

Those asking for the government to wait "were Carter administration holdovers who didn't want to say anything bad about the Soviets," Burt said.

The yellow rain accusations have been widely cited by conservatives as an example of duplicity by the Soviets, who, they say, not only use abhorrent weapons but casually break treaties banning their use. The Wall Street Journal has pressed the issue relentlessly on its editorial pages.

On the other side through the past six years has been Matthew Meselson, a prominent Harvard biochemist and an expert on chemical warfare, who relentlessly pursued the opposite view.

"The facts now speak for themselves," he said. "The early claims of toxins were not validated by the much larger studies done later in U.S. Army and British laboratories. The {refugee} interviews were cast into doubt by the careful work of the State-Defense interviewers. And yellow rain itself turned out to be entirely made of pollen."

In an article to be published Sept. 6 in Foreign Policy magazine, Jeanne Guillemin, Julian Perry Robinson and Meselson write, "In large measure it was the systematic efforts of American government investigators that undermined the administration's case."

Documents independently obtained by The Post and by Guillemin, a sociology professor at Boston College, show that:The most highly potent sample of the "weapon" offered as evidence by the State Department after further testing was found to have no poisons. The government did not reveal that fact. U.S. teams in 1983 and 1984 investigating attack sites in Southeast Asia and interviewing those supposedly suffering from symptoms of chemical attack found no verifiable chemical attacks, no victims actually suffering the effects of chemical agents, and stories from refugees that were highly contradictory and unreliable. A captured pilot who said he had personally fired chemical rockets against the Hmong in Laos, and whose statements the administration used as important evidence for chemical attacks, was "highly unlikely" to have dropped anything but conventional weapons, according to a government interviewer.

Former ambassador James Leonard, who was the U.S. negotiator on the 1972 biological and toxin weapons convention, said, "There couldn't be a worse way of handling an issue like this."

"Credibility is one of the factors people use in evaluating American intentions, and you become less credible when you make a pronouncement and refuse to back off it," he said.

Reports of chemical warfare in Southeast Asia first began to trickle in to American intelligence in 1976. U.S. investigators over the next three years found evidence that tear gas and other riot agents were used. The CIA reported that of the suspected chemical attacks, 90 percent were believed to be with these agents, which are not covered by chemical warfare treaties.

But refugees began to tell American investigators of many deaths which they connected with yellow powders and yellow spots that came from the sky. The symptoms shown by the sick and dying varied so greatly that they could not be matched with any of the common stock of chemical agents held by the Soviets or the United States.

Then in December 1980, Sharon Watson, an Army toxicologist and a member of an interagency working group studying chemical warfare, suggested there might be a new weapon in use in Southeast Asia, a weapon made from the natural poison made by fungi. Because the most deadly and most famous outbreak of trichothecene poisoning took place in the Soviet Union 50 years ago, she pointed out, the Soviets had studied these poisons extensively.

Watson decided to test one of the many samples brought back from Southeast Asia, a moldy looking leaf and stem. She had the samples sent to a laboratory at the University of Minnesota, which in summer 1981 found three trichothecene toxins.

At the time, Haig faced a serious problem in Europe. U.S. plans for the defense of Europe were under attack, and a debate raged over whether U.S. missiles such as the Cruise and Pershing should be planted on European soil.

Burt, one of Haig's top aides, saw possibilities in the situation. It looked as if the Soviets might be caught in a major blunder, caught using weapons that had for decades been universally renounced.

He wanted to make the story public, and to press it hard. But Watson resisted. She and others said they could not assure officials that the toxin wasn't from a natural outbreak.

In a memo dated August 31, 1981, less than two weeks before Haig's West Berlin speech, she told the State Department that going public with any accusations was "ill-advised for several reasons. These results are based on only one sample . . . . Before the possibility of natural occurrence can be entirely ruled out, results of analyses of normal flora samples from the region need to be screened." She said "this type of background evidence is crucial."

Haig went ahead with his speech, provoking a worldwide outcry. The Soviets immediately denounced the charges.

Two months after the first sample results were announced, the State Department said more trichothecene toxin was found in samples of vegetation and water from Southeast Asia. Blood and urine from refugees was analyzed and again trichothecene toxins were found.

Army intelligence used the Minnesota laboratory to test samples from Southeast Asia. The lab found five samples of leaves, water or powder containing toxin. Another 18 to 20 people who said they were victims of chemical attacks were found to have the toxin in their blood or urine. About 150 to 200 samples tested in Minnesota were negative.

Only one sample tested in Minnesota was also tested in other independent labs, according to documents. It was a sample of a yellow powder scraped off a rock in Laos, and it was divided into parts which were shipped to different labs.

What was then called the Army Chemical Systems Laboratory in Aberdeen, Md., analyzed it and found nothing. The British military analyzed it at their Porton Downs laboratory and also found nothing.

"It now appears probable that this sample contains no mycotoxins despite the originally reported finding," the Defense Department concluded in a memo. "Any other results that have not been verified by analysis in more than one laboratory should also be used cautiously."

Another memo was even more cautionary: "We believe that single analyses of complex samples should not be taken as absolute proof but only as indicative of what might be: they should be the justification for a more rigorous testing founded in scientific methodology. Unless this approach is taken any single analysis can be refuted by another single analysis . . . .

Still, the government continued to look at samples in a single lab, and continued to report the powder from Laos as a positive result. Other labs did come into the picture, however, since laboratories in several countries acquired their own samples from the area and tested them.

Aberdeen tested more than 80 samples in total, and found no positives. The British tested more than 100 samples from about 35 "attack" sites, and found no positives. Sweden collected another 20 or so samples and found none positive. A Canadian report said two labs tested more than two dozen samples. One sample was found positive.

Meanwhile, the government was marshaling what it considered to be additional evidence. In March 1982, Haig devoted a substantial part of a special report on yellow rain to the story of a captured Lao pilot. But what the Haig report did not mention was the result of an interview carried out by an American pilot to determine whether the Lao pilot's missions actually fit what would have happened on a flight dropping chemical munitions.

"It seems very unlikely that the rockets' fire contained BW-CW {biological warfare-chemical warfare} agents," the interviewer said. The Lao pilot did not even have a rudimentary knowledge of wind effects on smoke, crucial information in dropping chemical munitions.

Hundreds of refugee reports that lethal chemicals were being used against them were also used to buttress the government's case. The stories the refugees told were reported back, unconfirmed by the American teams that had gathered them, to Washington.

"We got many different stories and the reliability was not always known," said Dr. Frederick Sidel, who was an Army member of one those teams, adding: "I don't know whether there was any mycotoxin poisoning out there or not. Possibly not. Something was happening to these people, but we were never sure what."

Finally, at the end of 1983, a joint Pentagon-State Department team was sent for the first time to try to confirm the refugee reports. Documents obtained by Guillemin detail the results of these checks.

The hottest area of reported chemical attacks cited by the government was in the Phu Bia region in Laos, where 200 attacks had been reported. After interviewing a former resistance fighter who had led 40 men in that area since 1975, the interviewers cabled that "he denied ever having experienced a CBW attack nor ever having seen evidence of CBW use. When questioned why many other Hmong refugees related accounts of CBW attacks and he as a resident of Phu Bia for eight years had seen nothing, {he} stated that he was an educated man who related what he saw and not what he felt. He added that other Hmong are different in that they relate what they hear and feel."

Medical checks of those said to be victims turned up no clear cases of exposure to chemical attack, though a dozen attacks were reported to the team. The victims typically turned out to be suffering from other conditions.

The State Department has maintained that refugee reports represented a clear constellation of symptoms that were a "perfect fit" for the symptoms of trichothecene poisoning and no other condition. The key symptoms mentioned repeatedly were hemorrhaging, bloody vomiting, skin irritation and diarrhea.

Guillemin reported that an analysis she did of 217 reports over five years showed about 85 percent of the refugees interviewed did speak of a yellow substance. Only eight of the 217 interviewed said they had bloody vomiting, 10 percent bloody diarrhea, and 21 percent rashes or blisters. Only 3 percent reported all three symptoms.

By 1982, the government had also learned something about the samples of leaf and other materials from "attacks" that would lead to serious difficulties with their public campaign. Early that year, the United States learned what the yellow powder in yellow rain actually was -- pollen.

Asked how the Soviets would have been able to concoct a weapon by putting together poison from a mold with pollen, John H. Hawes, deputy assistant secretary of state for politico-military affairs, said, "I have no idea how the Soviets produce this stuff. We've not been in their factory."

Meselson and Yale bee expert Thomas Seeley, finding it hard to accept the idea of a pollen-and-poison weapon, sought other explanations of the confusing situation.

Seeley suggested that the yellow spots cited in so many reports might be the droppings of honey bees, which eat large amounts of pollen. In his work on Southeast Asian bees, Seeley knew that showers of yellow spots occurred commonly, when the bees went on their occasional "cleansing flights" to let go of their droppings while in the air rather than inside their hives.

This theory gained strength when another report of yellow rain was discovered in China. Years earlier, a shower of yellow rain was reported in a Chinese village. People feared the yellow spots were toxic and there was fear of widespread illness. But a close count showed that despite reports from the farmers, the local clinic had no more reports of illness than usual. Eventually, Chinese researchers discovered that the showers were from honey bee flights, and even called the droppings a "yellow rain."

An uncanny repeat of the incident occurred in Thailand during the height of the government's action on yellow rain. In Ban Sa Tong, a Thai village a few miles from the Kampuchean border, a shower of yellow rain occurred and villagers associated the rain of spots with an airplane that flew over. Reports of the "chemical attack" spread quickly, but clinics reported no significant increase of illness or death.

The yellow spots were analyzed and found to be like bee droppings in appearance, and to consist almost entirely of pollen. Canadian tests showed most of the samples had no trichothecenes, but at least one showed a level like those occurring in natural trichothecene outbreaks. The U.S. government told reporters that the incident must have been staged to cover the tracks of those perpetrating chemical war.

Ultimately the explanation of the full array of bizarre incidents and fragmentary evidence may never be clear. But by last year, the United States had stopped sending teams to Southeast Asia to investigate yellow rain. Officially, the government says no more attacks have been confirmed since 1983.

Within the government there is now a split between the believers and the skeptics, and widely differing beliefs about what the government should do. For now, the charges stand, and are repeated in public when it is convenient.

Meselson suggests several possibilities for what happened: First, that bee showers or at least the appearance of yellow spots were mistaken for the agents of chemical war. Second, that because there are so few positive results and they come mostly from one lab, simple lab error can explain the appearance of trichothecenes. Third, some natural occurrence of trichothecenes in Southeast Asia may explain some of the other positive results.

The explanation still has numerous loose ends, Meselson says.

"In retrospect," Meselson concluded in the Foreign Policy article, "it is easy to see how the administration came to be in its present predicament. The U.S. intelligence community departed from established procedures for verifying laboratory and field information and instead supported a conclusion that should have been regarded only as a hypothesis."