In early 1979, a government scientist in a laboratory in Southwest Washington started giving caffine to more than 300 pregnant rats. His purpose: to confirm that the drug caused birth defects in the rats to pinpoint how small a dose could produce the defects.

Some 22 months later, three government inspectors made a routine check of that laboratory and found it dirty and disorganized, a bad place to do a good study.

Those two events have led to a bureaucratic collision within the Food and Drug Administration, a collision confusing the issues the agency wanted to clarify: Does caffine, a dietary staple of millions of Americans through their consumption of coffee, tea, chocolate and cola, represent a health hazard? To whom? And what should the FDA do about the nearly 2 million pounds of caffine added to America's soft drinks annually?

The story of the caffine study is a textbook example of the pitfalls facing policymakers who try to translate scientific research, with all its uncertanties, into the certanties of public policy.

In 1978 a group of scientists under contract to the FDA recommended that caffine no longer automatically be considered safe in sodas. They were particularily concerned about caffine's effect on the behavior of children.

Dr. Jere E. Goyan, then the FDA commissioner, took the scientists' advice and ordered a review of the scientific studies of caffine. But while the review was under way, attention shifted from the drug's possible effects on behavior to indications that high doses of caffine caused birth defects in animals -- something that, in scientific terms, was easier to prove.

Some 20 years' worth of laboratory studies pointed to the conclusion on birth defects.But many of the studies were done when the science of teratology, the study of chemicals that harm the fetus, was new.Other studies used small staples.

So Goyan took an unusual step. Instead of depending on industry scientists or academics to do more research, he authorized an FDA scientist to do, in essence, a product test of his own, rather than merely reviewing the work of others. The aim was to confirm the earlier work and to determine the highest level of caffine that could be consumed safely by pregnant animals.

In the spring of 1980, Dr. Thomas F. X. Collins gave his superiors his preliminary findings. The offspring of rats fed the equivalent of 12 to 24 strong cups of coffee daily had an unusually high proportion of missing or deformed toes, he reported. Also, some delay in skeletal development was observed in the offspring of rats fed the equivalent of two cups of coffe a day.

Last September, the agency issued a general statement cautioning pregnant women against consuming caffine. Goyan used the Collins study as Exhibit A to support the warning. And just six weeks after the results were made public, the FDA issued its long-debated proposals to tighten government regulation of caffine in sodas. Those proposals, according to FDA spokesman Wayne Pines, were "triggered by the Colins study and supported by a large scientific history."

The proposed regulations on caffine, published in the Federal Register last Oct. 21, satisfied few people involved in the debate. Officials of the soft drink industry wre upset because caffine was to be taken off the list of substances "generally regarded as asfe" and given a twilight "interim" status until soft drink manufacturers did extensive new tests. The FDA acted prematurley, industry scientists said, because there was no clear evience caffine would be hazardous to humans.

Consumer groups, like the Center for Science in the Public Interest, wre unhappy that the FDA didn't go further in issuing public warnings or limiting caffine's availabilty on supermarket shelves. It was a case of too little, too late, the activists said.

The Collins study gave agency officials a handy answer for both. The study, they said, was more extensive than earlier efforts and reinforced their doubts. But there was still need for further testing before a decision could be made on more drastic action against the additive.

For a time, it seemed the FDA had successfully finessed a tricky issue. Then abruptly the regulators were blindsided from an unexpected quarter.

Just a week after the agency cautioned pregnant women about the changes of caffine, three inspectors from the Baltimore regional office of the FDA made a scheduled reinspection of the Bureau of Foods labs at 200 C St. SW, where a second phase of the Collins study was being conducted. They found the lab poorly organized and unsanitary.

The inspectors then reviewed the work of Collins and his collaborators and found it wanting. Samples appeared to have been mixed up, the inspectors said. They were leery of the sampling process. They raised so many questions that the FDA officials convened a panel of non-FDA government scientists to review Collins' work, evaluate the inspectors' objections and make their own report. Meanwhile, late last year, the proposed regulations were postponed.

The outside scientists' review, released last month, did little to clear up the confusion. It agreed that Collins' data were solid, but it questioned his interpretation of those data. " . . . High levels of caffine were associated were teratogenicity [birth defects]," their report said. "However, the study did not confirm if caffine was a direct teratogen."

Now, FDA officials find themselves involved in a rear-guard action, defending the Collins study, yet playing down its importance in the rule-making process. "Collins [wasn't] the basis for regtulatory action but a tool to raise the questions," Bureau of Foods Director Sanford Miller said recently.

"The problem is that now . . . everyone and his uncle is taking potshots at the study," he added. "There is no scientific study that has ever been done that won't have some components people are going to criticize. There's a feeling that if somebody can find something wrong with Collins the whole [rule-making] process is going to go away. And that's not so."