For the past five years, as it purchased almost 3,000 Bradley Fighting Vehicles at $1.5 million apiece, the Army fended off accusations that the Bradley could not protect the troops it was designed to carry.

Then last month, under pressure from critics in Congress and the Pentagon, the Army for the first time subjected a Bradley, loaded with fuel and ammunition, to antitank fire. The Army discovered that had live soldiers instead of dummies been riding the test range at Aberdeen, Md., there would have been many casualties.

So, fiery results in hand, the Army is preparing a package of "survivability improvements" to install in existing Bradleys and build into future ones at a possible extra cost of $75,000 per vehicle, according to Lt. Col. Craig A. MacNab, a Pentagon spokesman.

Army officials said the live-fire tests did not teach them anything they didn't know about the Bradley, which was never designed to withstand rocket-propelled grenades and antitank missiles.

"It does just what it's supposed to do, and it is in fact more resistant to damage than we thought it would be," MacNab said. "But we can make it even more resistant."

But Army brass, who sold Congress on the Bradley as a vehicle that would fight "side by side" with the M1 tank, now talk about holding it behind battle lines, leading some critics to ask whether it is worth more than 10 times the armored personnel carrier it is replacing.

"What they're counting on is, we're so stupid, we'll forget they said it was going to fight alongside the tanks," said an aide to Rep. Denny Smith (R-Ore.). "Do we have to pay $2 million for a vehicle that's not going to see any action?"

Smith also criticized the Army for delaying survivability tests until more than a third of its projected force of almost 7,000 had been purchased. "They're a couple of billion dollars late, and it's going to cost a chunk of money to straighten it out," Smith said.

The Army began working in 1963 on a successor to its M113 troop carrier, settling on the Bradley to not only carry troops, but fight with a 25-mm cannon and antitank TOW missiles as well. Since its birth, the Bradley has been criticized as -- variously -- too big, too high-profiled on the battlefield, too expensive, too unwieldy to carry on a C141 cargo plane without being dismantled, not heavily armored enough and flawed in its basic concept of storing explosive ammunition in a troop-mover.

The Army and FMC Corp., which makes the Bradley in San Jose, Calif., have shepherded their $13 billion program past all those accusations. Army generals say they have a vehicle that is superior to the Soviet BMP, fast enough to keep up with the U.S. M1 Abrams tank, armed to take part in battle -- and better protected than any other personnel carrier in the world.

Many of those generals worry, however, that Pentagon critics have decided to concentrate their fire on the Bradley now that they have shot down the Sgt. York Divad antiaircraft gun, a $6 million weapon that the Army defended vociferously until Defense Secretary Caspar W. Weinberger concluded it would not work.

An internal Army memo prepared for a briefing of the vice chief of staff in October began this way: "Army likes the Bradley. Serious threat to the program. Improved survivability is the priority. Cut in 3 quick fixes."

The "quick fixes," according to Army officials and others, include moving ammunition to safer locations, changing flow of diesel fuel so less of it is exposed to fire, shielding electrical cables and adding more or better armor. In addition, FMC has prepared mockups of what it calls a "high-survivability" model of the Bradley. Army generals viewed the proposed new models at a private exhibition in San Jose earlier this year.

Army and FMC officials said the proposed modifications do not indicate past laxity, but improving technology, the need to meet growing threats from improved antitank weapons and normal improvements common to any weapon program. They said adding much more armor would decrease mobility and add weight, tradeoffs that were carefully considered long ago.

Brig. Gen. Donald P. Whalen, director of weapons systems in the Army research office, told Congress earlier this year that live-fire tests had not been conducted until Congress and Weinberger's office demanded them because the Army knew a large artillery round would "do terrible damage" and "there is not a great virtue at the early stages of confirming what we already know."

But in the early stages, before Congress had funded the Bradley, the Army did little to advertise its vulnerabilities. In fact, Brig. Gen. Stan Sheridan told Congress in 1978 that the Bradley would "fight side by side" with the M1 tank, offering "sufficiently increased protection to allow infantry and cavalry to fight from within the vehicle."

When a congressional aide asked whether the Bradley would protect the M1 by "distracting the enemy tank and giving him another target to shoot at," Sheridan said that was "correct."

"It is a matter of the enemy knowing he has to shoot at the IFV Infantry Fighting Vehicle , because the IFV has the ability to kill tanks with its missiles," Sheridan said.

Today, Army generals say no one would be foolish enough to offer the Bradley as a target. In an interview with Armed Forces Journal in October, Maj. Gen. John W. Foss, Army infantry chief, said "the best place for the Bradley is to stand back."

"You don't want to make them vulnerable, you don't want to run it up the hill to attack tanks," Foss said. "It would be a pretty dumb commander who would do that, who would have his Bradley right behind his tanks."

Meanwhile, the Pentagon is supposed to report the results of its live-fire tests to Congress today. They are classified. While Smith said they revealed serious problems in the vehicle, MacNab said only one of the four test vehicles was destroyed and no more than three of the five dummies riding in each vehicle were "killed" by any shot.

"We're thrilled with the results," he said. "The notion that if you kick the tires of a Bradley the whole thing's going to go up in a glowing fireball just didn't pan out."