Four years ago, while U.S. troops were toppling the Taliban in Afghanistan, the Environmental Protection Agency lobbed a different sort of bombshell at the Defense Department. EPA scientists recommended strictly regulating a chemical that is a key component of munitions, but that has seeped into drinking-water supplies.

The EPA said it had determined that the chemical, called perchlorate, endangers babies' brain development when present even at trace levels. As a prelude to possible formal regulation, it proposed declaring that a safe level of the chemical in drinking water would be just one part per billion. That's an amount so minute it wouldn't even have been detectable a few years ago.

Pentagon officials were aghast. Defense suppliers had discharged massive quantities of the chemical into soil and streams during the Cold War, and they still need it for weaponry. Such a strict limit could mean the Pentagon and defense contractors would have to clean up scores of water sources in 35 states and even the mighty Colorado River, with its water flow of 67,000 gallons a second at the Hoover Dam.

Fearing both costs and possible curbs on arms production, the Pentagon took its case to the White House, which told the EPA to stand down while an outside scientific panel looked at the issues. The panel then issued a middle-ground report that has left some senior EPA scientists deeply unhappy and the Pentagon still pressing for the minimum possible cleanup.

The standoff, involving two high-profile federal agencies, shows how the burgeoning science of low-dose chemical exposure is raising both the stakes and the stratagems in today's pollution fights. There's no question perchlorate interferes with the body's ability to make thyroid hormone, a substance that everyone needs but babies especially so. The question is how much exposure it takes to do harm. The controversy has intensified with science's growing ability to detect and test chemicals at extraordinarily low exposure levels.

The appeal to the White House was just one of the several moves by defense interests in a long struggle with the EPA over whether and how to regulate perchlorate. Among other tactics: Perchlorate users financed a study of the chemical's health effects -- then undermined their own study when results went against them.

Perchlorate, used chiefly in solid rocket fuel, first polluted groundwater decades ago at a munitions plant outside Sacramento, Calif., triggering years of resistance by the plant's operator to state regulatory efforts. Then in 1997, after technical breakthroughs allowed detection of the chemical at far lower levels than before, it began to be found in water supplies in Southern California.

EPA scientists traced one plume up the Colorado River aqueduct to Las Vegas. There they found the source in an old plant that once manufactured the missile propellant. The soil beneath was tainted and the chemical was seeping into the river.

In the human body, perchlorate blocks the thyroid gland from absorbing iodide, which the gland needs to make thyroid hormone. The Pentagon and defense industry say such interference isn't dangerous, at least so long as it's only partial, because most adults produce plenty of the hormone.

The EPA, however, focused on fetuses and infants. They need thyroid hormone every day, because it is critical during brain development. And unlike adults, they can't store a supply. Because risk levels weren't well understood, the EPA and the Pentagon agreed in the late 1990s to cooperate to find answers. Several defense contractors, linked in what was called the Perchlorate Study Group, agreed to pay for new research.

The centerpiece was a $3 million experiment involving 3,000 mother, infant and fetal rats. Pregnant rats and pups were fed varying levels of perchlorate for several months. Scientists then dissected the rats' thyroid glands and brains. Researchers started with the rats that got the largest dose of perchlorate, intending to work downward until they found a dose so small that it had no effect.

They never found such a dose. Even at the lowest dose tested -- 0.01 milligrams per kilogram of rat weight per day -- the scientists saw a pattern of altered growth in several regions of the baby rats' brains. They also saw effects on their thyroid cells and hormone output.

Chemicals don't necessarily affect rats and humans the same way. Still, the test results would be considered "adverse effects" under EPA policy, the agency's team leader, Ann Jarabek, warned the defense interests. She told them the results would tend to reduce the level of perchlorate exposure the EPA ultimately would deem safe.

Sponsors of the study then did something unusual. Instead of submitting the final results of the study to the EPA, the defense companies that paid for the study commissioned a critique of their own research. They hired a consulting firm, which asked five academic scientists to study the study.

A few months later, in May 2001, the defense contractors delivered to the EPA a 200-page critique of their own study. It found fault with the study's design, with the handling of rat pups, with what the pups were fed and with the way rat brains were sliced and preserved. Conclusion: They said the multimillion-dollar study they financed was highly flawed.

The agency's chief of neurotoxicology, William Boyes, says he had never seen sponsors of a study attack their own work. "Usually," he says, they either "stand behind their data or they go back and do another study."

Also puzzling: The head of the consulting firm the defense industry hired to critique the original study had been that study's science adviser.

This consultant is Michael Dourson, who leads a nonprofit science consulting firm called Toxicology Excellence for Risk Assessment, or TERA. Dr. Dourson says the critique wasn't an attempt to discredit the rat study, but simply to explain its "biological significance."

The laboratory that had done the rat study says it stood ready to do it over if necessary to correct any flaws identified. But the defense industry didn't ask the lab, Argus Research Laboratories in Horsham, Pa., to do it over. Asked why not, an executive of one major user of perchlorate, the Aerojet missile unit of GenCorp Inc., said it was because EPA guidelines regarded animal studies as inferior to human ones anyway. So, he said, the industry had by this time decided to focus on human research.

In early 2002, the EPA, equipped with the rat study's final results and also the critique of it, issued a draft risk assessment for perchlorate, proposing a safe limit for the chemical in drinking-water supplies. This would constitute the first step toward possible regulation, which can occur only after further study, including a cost-benefit analysis. The EPA's proposed safe limit was quite strict: a mere one part per billion.

Pentagon officials felt sandbagged. The defense industry paid for the rat study in the expectation that they would hear privately from the EPA about any problems it presented. Instead, they learned at the same time as the public of the strict safe limit the EPA now wanted.

"All of a sudden, up on the screen popped this one parts per billion standard -- where did that come from?" says Raymond DuBois, a former deputy U.S. undersecretary of defense who's now acting under secretary of the army. This limit, he says, "had no consistent scientific confirmation."

EPA officials, asked why they didn't warn the industry the strict proposal was impending, said that while they cooperate with industry on research, the job of setting safe exposure levels is theirs alone.

"Perchlorate is now among the better understood compounds," says Paul Gilman, the EPA's former chief scientist. "At some point, the agency had to step inside itself as a regulatory body and determine the weight of the evidence."

The furor the EPA had stirred was soon evident at a gathering known as a peer-review workshop, where a panel of scientists discussed the proposal. The workshop took place in early 2002 in Sacramento, near the site of decades of groundwater perchlorate pollution from an Aerojet missile factory.

The session was tumultuous, featuring environmentalists, regulators, consultants and lobbyists. Among the speakers was La Donna White, president of an African-American doctors' group, who said the EPA proposal would divert funds from "real health issues" affecting blacks and "scare the public." She later repeated her points in an op-ed essay in a local newspaper -- and in a news release put out by a lobbying group for perchlorate users, the Council on Water Quality.

Dr. White, a family physician, says she had learned about the issues from a guest at one of her medical-society meetings, Eric Newman. He is a lobbyist for a Sacramento firm that has lobbied on perchlorate matters for defense contractors. Dr. White says she didn't know he was a lobbyist when he asked her to speak to the EPA. She didn't reply to an email asking whether anyone had helped her draft her perchlorate commentaries -- two of which misspelled her first name. Mr. Newman didn't return messages left for him.

Perchlorate users and the Pentagon said the chemical was safe in drinking water at 200 times the safe limit the EPA wanted, that is, at up to 200 parts per billion. The Pentagon's Mr. DuBois appealed in early 2003 to the White House Office of Management and Budget, which referees inter-agency disputes. Given the strict limit the EPA was pushing, he says, "I said, 'Time out!' "

The White House told the EPA to halt further action on the chemical, and arranged for the EPA and three other agencies to sponsor further review by the National Research Council, a federally funded group that vets issues for the government and others. The council, in turn, named a panel of scientists, who did a wide-ranging assessment that included public hearings in 2003 and 2004.

At the hearings, the EPA came in for harsh criticism from perchlorate users and consultants working for them. An Air Force colonel, Daniel Rogers, termed the EPA's work "biased, unrealistic and scientifically imbalanced." Col. Rogers also said perchlorate is critical to U.S. security because while highly explosive, it is stable during handling and storage. Besides missiles, it is used in various battlefield weapons and flares and in munitions for training.

In January 2005, the National Research Council panel announced its conclusions. It called the rat research inconclusive and said perchlorate's key effect of blocking iodide from entering the thyroid gland, and thereby interfering with production of thyroid hormone, was not in itself dangerous. Still, it said, exposure to perchlorate should be restricted because of the high stakes for babies.

The panel recommended a maximum safe exposure level of 0.0007 milligrams per kilogram of body weight per day, based on a small study of human volunteers. For an adult drinking a normal amount of water, that would permit about 24 parts per billion of perchlorate in drinking water -- assuming people ingested no perchlorate from any source except water.

In fact, however, the EPA's working assumption in such cases is that drinking water accounts for only 20 percent of people's exposure to a waterborne contaminant. Recent studies indicate that small amounts of the chemical are in a wide variety of fruits and vegetables, possibly from irrigation water, as well as in some dairy products and breast milk.

Some EPA staffers assumed their agency would reduce the safe level in drinking water well below 24 ppb to adjust for several factors, including exposure through food. Instead, the EPA quickly adopted the panel's assessment as its own, eschewing the internal and external peer reviews that normally precede a formal EPA listing of a safe level for a chemical.

An EPA spokeswoman said no additional reviews were needed before adopting the 24 ppb safe limit because of extensive internal and external scrutiny of the chemical done several years ago. She also said it was natural to use the National Research Council's conclusion as the EPA's own because the EPA was among those who sponsored the review.

Some state agencies criticized both the National Research Council assessment and the EPA for quickly adopting it. Massachusetts complained to the EPA that the research-council panel had based its analysis on a study of just seven adults, rather than on babies. Massachusetts reaffirmed its own health advisory that is as strict as the safe limit the EPA envisioned in 2002: one part per billion in water. Meanwhile, two regulators from Connecticut and Maine wrote a science-journal commentary accusing the EPA of superseding its own scientific judgment with a flawed review by an outside body.

Today, Pentagon and White House officials are drafting new guidance for toxic-site cleanup officials. Intended to go out under the EPA's name, the guidance under consideration would effectively fix the cleanup standard for federal pollution sites at 24 ppb. The result is that many water bodies with less perchlorate than that would escape cleanup.

Several senior EPA staffers believe the agency would be better off with no perchlorate cleanup policy than with this one, emails reviewed by The Wall Street Journal show. "We got a very ugly set of comments from Office of Management and Budget last week that eviscerated the guidance" to be given to cleanup officials in the field, one senior EPA staffer emailed a colleague this fall. "Doing nothing was better than accommodating those comments." EPA spokeswoman Eryn Witcher said the policy is still undergoing internal deliberation.

All the skirmishing thus far still doesn't determine whether the federal government ever will actually regulate perchlorate with a mandatory water standard. To help decide that, the EPA plans to test drinking-water supplies nationwide over the next several years. It is also monitoring blood and urine screenings and tests of food, to measure Americans' exposure from sources other than drinking water.

The arms industry thinks even the safe limit of 24 parts per billion is far too strict. It notes that the National Research Council said the effect on the thyroid wasn't itself adverse to health, but merely could possibly lead to ill effects, in a chain of events. Says Dr. Dourson, the defense-industry consultant: "The committee chose a precursor to a precursor to a precursor to an adverse effect in the development of its safe dose."