Scientists still do not fully understand exactly where or how SARS emerged 18 months ago. But it is now clear that the most threatening source of the deadly virus today may be places they know intimately -- their own laboratories.

The recent announcement of nine cases of severe acute respiratory syndrome linked to China's National Institute of Virology brings to three the number of lab outbreaks of the disease in the past eight months. The three events -- including one in Singapore in September and another in Taiwan in December -- account for all but four of the known SARS cases since last year's epidemic was brought under control.

The Beijing incident, unlike the others, led to person-to-person transmission of the virus outside the lab. It caused one death and required quarantining about 200 people in two provinces to stop the virus from spreading. It was an epidemic "near-miss" and has led to calls for greater international monitoring of labs working on a virus that caused more than 8,000 illnesses and 774 deaths last year.

Together, the three SARS outbreaks have highlighted the unique hazards to public health that arise from accidental laboratory releases of germs that no longer exist -- or barely exist -- in the wild.

Such an event happened 26 years ago when the last cases of smallpox -- the only human disease ever eradicated -- occurred after a laboratory accident. Another one may have happened in 1977, when an influenza virus not seen for 27 years inexplicably reappeared and circulated worldwide. Leaders of the campaign to wipe out polio are working to ensure that such a thing never happens with that disease. They are already inventorying and urging destruction of global stocks of polio virus.

Laboratory workers can be infected in myriad ways, including needle sticks, animal bites, splashes in the mouth or eyes, and undetected inhalation of infected droplets. When a person recalls no definite exposure, in most cases the microbe somehow got into the air, usually because of poor lab technique and occasionally because of faulty equipment.

The number of fatalities in the United States from lab accidents is unknown, as there is no requirement to report lab accidents or cases of illness caused by them to government authorities. Thirty years ago, a University of Texas microbiologist attempted to count all known laboratory-acquired infections worldwide. He found 3,921 -- 4.2 percent of them fatal -- with most occurring before the 1960s.

Improvements in lab equipment and technique since, as well as development of vaccines against some of the more dangerous microbes, have greatly reduced lab hazards. But infections still occur, and not just from SARS. Russian health officials recently reported that a scientist working on Ebola virus at the Vector State Research Center of Virology and Biotechnology in Siberia died after sticking herself on May 5 with a contaminated needle. A team of experts from the World Health Organization is investigating China's lab-associated SARS cases. It has not announced its findings or any recommendations. But the problem goes far beyond what happened in the Beijing lab, some experts say.

"Does the WHO know how many laboratories in the world have this organism?" Robert Webster, a virologist at St. Jude Children's Research Hospital in Memphis, said of SARS. "It would seem to be time to collect this information. It really is time that the whole world, not just China, rounded up these things and put them away."

Webster has helped research numerous new strains of influenza, including the H5N1 strain of avian flu that killed millions of birds and 19 people last winter. He thinks lab stocks of dangerous influenza strains, as well as SARS virus samples, are a major but largely unrecognized threat to public health.

In the wake of the SARS epidemic, WHO recommended that the virus be handled only in laboratories rated "biosafety level 3," or "BSL-3." Such labs limit access, and their workers must handle microbes in sealed or vented cabinets and wear protective clothing. Only BSL-4 labs, where technicians and scientists must wear spacesuits that have their own air supply, are more restrictive.

BSL-4 labs, however, are rare and expensive to operate. The United States has four. SARS was designated a BSL-3 pathogen in part for practical reasons.

"We put the virus at a level appropriate to avoid its accidental release, but also at a level at which enough work could be done on it. If we'd put it on Level 4, it would restrict it to just a handful of laboratories worldwide," said John Mackenzie, a scientist at Curtin University of Technology in Australia, who helped formulate the WHO guidelines.

The three SARS lab outbreaks appear to have had distinct causes.

In Singapore in September, a sample of West Nile virus contaminated with SARS virus infected a 27-year-old lab worker at the Environmental Health Institute. The lab was not known to have stocks of SARS, but it had been pressed into service during the epidemic, which is presumably when the contamination occurred. Although the lab claimed a BSL-3 rating, a WHO inspection team found it did not meet those standards.

The Taiwan case happened in a BSL-4 lab when a 44-year-old military scientist failed to follow procedures in cleaning up a spill of SARS-containing fluid. It was judged to be a case of individual carelessness or a failure of training.

The Beijing case is the most mysterious and troubling. There, a 26-year-old graduate student developed SARS in late March, just two weeks after she started working at the virology institute. In mid-April, a 31-year-old man in the same lab also came down with the disease. Neither had been working with the SARS virus.

The graduate student went home to Anhui province, where she infected her mother, who died. The student then became ill enough to be hospitalized and infected a nurse. The nurse, in turn, infected five others -- three relatives, a patient and a relative of that patient -- in a "third generation" of infection. When the outbreak became known, Taiwan and Australia instituted health alerts, screening visitors from China or ordering special surveillance of recent travelers who became ill.

What the latest accident "has done beyond a shadow of a doubt is show that we do need some international agreement as to what a Biocontainment Level 3 lab is," Mackenzie said. "Unless everyone has the same standards, we may be talking about chalk and cheese." He said he and several others are calling for an international system to "accredit the laboratories and accredit the training of the people working in them."

Larry Anderson, chief of respiratory pathogens at the Centers for Disease Control and Prevention, said safe handling of SARS requires the right lab design, personal equipment, adequate training of workers, proper technique and medical surveillance of people at risk of exposure. Since the virus emerged, the CDC has distributed samples of it to 56 laboratories at universities, government departments and companies. Fourteen are overseas. Anderson would not name any of them and said he does not think the large number is inherently unsafe.

"I think one lab that is working with it inappropriately is too many. Fifty-six working with it appropriately is not a problem," he said.

The Beijing incident is reminiscent of a notorious smallpox release in Birmingham, England, in August 1978 -- 10 months after the last wild infection occurred in Somalia.

Henry S. Bedson, head of the microbiology department at a medical school, was rushing to finish his experiments before the deadline to turn in or destroy his stocks of smallpox. The lab's containment had been judged unsatisfactory by WHO inspectors, but they did not have the power to close it.

The smallpox virus apparently became aerosolized in Bedson's lab and traveled up one floor through air ducts to the school's photographic studio and darkroom. A 40-year-old photographer became infected and died, even though she had been vaccinated 12 years earlier. She transmitted the virus to her mother, who also became ill but survived. Her father did not become infected but had a fatal heart attack.

Bedson, despondent, slashed his throat in his potting shed, leaving a note in which he said, "I am sorry to have misplaced the trust which so many of my friends have placed in me and my work."

The leaders of the effort to eradicate polio, who hope to finish the task this year or next, have been working since 1999 to ensure that no such tragedy mars that historic achievement. They have asked nearly 200,000 labs around the world whether they hold polio virus. To date, 833 have said they do, either in pure form or in fecal samples, Christopher Wolff, a WHO scientist, said recently. About 50 have since destroyed their stocks, and many more expect to do so once the disease disappears.

The biggest disease outbreak that may have arisen from a laboratory was the mini-pandemic of "Russian flu" in 1977 and 1978.

Despite its name, that strain of influenza virus appeared in Tientsin, China, in May 1977. It spread around the world, causing mild infection that almost exclusively hit people younger than 20. Millions of people became ill, although overall flu mortality did not increase.

What is curious is that this virus had a genetic fingerprint virtually identical to a strain that had last circulated in 1950. Flu viruses evolve at a fairly predictable rate "and it is extremely difficult to explain why the . . . strains . . . are so strikingly familiar," a team of scientists wrote in 1978.

There are two possible explanations. The first is that the 1950 virus was somehow "genetically frozen" in nature -- possibly in ice or perhaps in some human or animal carrier that has never been discovered. The second is that it escaped from a laboratory in China.

Many scientists think the second is the more probable.

Chinese workers were quarantined after colleagues contracted SARS.Bottles are taken from a Chinese lab where investigators suspect two cases of SARS originated this spring. Neither infected worker had handled the virus.