As the United States nears a decision on whether to go to war, the Bush administration is carefully ducking one central question of the public debate: What would it cost, in American lives, to expel Iraq from Kuwait?

Estimates outside the Pentagon extend from hundreds of U.S. troops killed in action to tens of thousands, a range so broad that it discourages meaningful argument about the matter. Even inside the military, according to sources, casualty projections vary widely.

Some projections are based on historical precedent, using loss ratios in previous wars to estimate how many of the 430,000 U.S. troops who will be in the Persian Gulf area by early February might die. Other projections are built around computer models of the lethality of modern weapons and the expected intensity and duration of combat. Still others result from negotiations between military bureaucracies, such as those concerned with medical supplies and troop replacements, whose plans depend on the results.

"Sometimes they'll say, 'That's impossible, I just don't believe it,' and they'll just raise or lower the number," said one source with extensive experience in the process.

None of the Defense Department's projections has been made public partly because, according to civilian and military sources, the administration fears potential negative repercussions in public support for a confrontation with Iraq.

"You have two ways to go" on public discussion of casualties, said a Pentagon official familiar with the thinking at senior levels. "One is to say, 'It's really worth it' without being specific, which I think it's fair to say is the present course. The other is to say, 'Here's what the cost is going to be,' and cause a sharp intake of breath and second thoughts."

The Pentagon has chosen the former course, the official said, "so the natural tendency in any democracy, which is to debate and eventually decide it ain't worth the price, can't work against us."

No Defense Department official, military or civilian, agreed to be interviewed by name on the subject.

In the military's central command, where senior officers must anticipate casualties to plan operations and logistics, projections of troops killed and wounded are classified. So are the numbers of hospital beds, body bags and graves registration units in the theater of operations.

In a recent interview, Lt. Gen. John J. Yeosock, the commander of Army forces in the gulf, got through several minutes of conversation without once using the words "wounded," "dead" or "casualties." He spoke only of "outcomes" and "scenarios" and said he did not know what they would be.

"It could be one {casualty} and it could be something else," said Army Lt. Gen. Calvin A.H. Waller, deputy commander of U.S. forces in the gulf. "I mean, who knows what's going to happen in war? I have no idea what it's going to be."

Few subjects are handled more skittishly by the Pentagon's public relations apparatus. Spokesmen for individual services have been instructed to refer all questions on casualties to the Defense Department's main office of public affairs. That office, in turn, declines to answer them.

In eastern Saudi Arabia, where reporters may not visit troops or bases without permission, dozens have placed their names on sign-up sheets to visit the mortuary units trained to recover, identify and evacuate soldiers killed in battle. Officers at the U.S.-run Joint Information Bureau -- who gladly put reporters inside state-of-the-art tanks and airplanes and in secret forward positions with combat troops -- have not granted any of the requests.

As they begin to anticipate combat, U.S. military authorities also have proposed ground rules for reporters that would sharply restrain television and photographic coverage of the casualties of battle. Supplemental rules issued Thursday, and challenged by many media organizations, would ban images of "patients suffering from severe disfigurement" and "personnel in agony or severe shock."

Some Pentagon officials, citing the British example, say the administration is wrong to avoid talk of the possible cost of combat. British commanders, such as Brigadier Patrick Cordingley of the 7th Armoured Brigade (Desert Rats), are urging their countrymen to be ready "for a particularly unpleasant war," with "a lot of casualties."

"I applaud that," one Pentagon official said. "I think that's a good way to handle it. People need to know what the cost is."

At the same time, experts inside and outside the Pentagon insist that no one really knows the answer.

"Please don't say that Epstein predicts anything," said Joshua Epstein, a Brookings Institution analyst who briefed members of the Pentagon's Joint Staff about his computer model of a conflict in the Persian Gulf. "All I can say is that these numbers appear to be in a very large ballpark."

Epstein's model, designed with associate Alf Hutter, follows the tradition of Frederick W. Lanchester, a British meteorologist who first used mathematical equations to describe ground warfare after World War I. The modern method, based on so-called Armored Division Equivalents, or ADEs, measures the "combat value" of opposing forces against the standard of a U.S. armored division. Combat value is computed using such factors as the number of soldiers, the quantity and quality of their equipment and their mobility on battlefields.

By early February, when the last U.S. reinforcements arrive in the gulf, Epstein calculates that the international coalition arrayed against Iraq will have 17.6 ADEs in the Kuwaiti theater of operations, 11.5 of them American. Iraqi forces, he calculates, will have 17 ADEs in Kuwait and southern Iraq. Because the defense is widely dispersed, however, Epstein calculates the Iraqis will have only 2.3 ADEs in the main sector of the likeliest U.S. attack.

A typical ADE lost in combat, according to Epstein's methodology, means the loss of 290 tanks, for example, and 72 crew-served artillery pieces. The corresponding number of casualties is 3,851, about two-thirds of them wounded and one-third dead.

Epstein plotted U.S. casualties according to two broad outcomes. The optimistic case, based on 15 days' intense combat, left 3,344 casualties, 1,049 of them dead. The pessimistic case, based on 21 days' intense combat, showed 16,059 casualties, 4,136 of them dead.

Like all models, Epstein's depends on assumptions about the way U.S. forces will attack. Since only a handful of top officials know the actual war plan, outside analysts like Epstein can only guess. Epstein assumed a two-phase plan, beginning with an air attack against strategic targets in Iraq and Kuwait and followed by a ground assault involving a breakthrough against Iraqi defenses in the eastern part of southern Kuwait, a siege of Kuwait City and a flanking attack on the main reserve of Iraqi forces in northern Kuwait.

As Epstein freely acknowledges, there are other plausible war plans. Retired Col. Trevor N. Dupuy, a military historian and author of a book titled "Attrition," has sketched six possible war plans and calculated six likely outcomes in U.S. casualties. Dupuy, a supporter of strong action against Iraq, projected a low of 300 American dead and 1,700 injured and a high of 3,000 dead and 15,000 injured, depending on which plan was used.

The Center for Defense Information, which is generally opposed to a war in the gulf, published a projection based on an overland drive all the way to Baghdad that showed 10,000 Americans killed and 35,000 wounded.

At bottom, however, according to the center's David Isenberg, "anybody who pretends to give you ultra-precise estimates is someone who's either hugely arrogant or hugely ignorant."