The Army gave Congress an unrealistically rosy view of how well the Bradley Fighting Vehicle will stand up in combat after conducting vulnerability tests last fall, the General Accounting Office has concluded.

The Army told Congress that live-fire tests proved the $1.7 million troop carrier to be more resistant to enemy fire, and better able to protect soldiers, than the Army had expected. But the GAO, the investigative arm of Congress, said in a report not yet released that the Army did not conduct realistic tests and did not give Congress a fair analysis of the tests it did conduct.

"Results do not provide a realistic picture of the vehicle's vulnerability or of the number of casualties likely in combat," the report, requested by Sen. William V. Roth Jr. (R-Del.), found.

The GAO said the Army fired less powerful weapons against the Bradley than it would meet in combat; purposely fired them away from the most explosive spots; used the vehicle configuration that carries fewest troops, so casualty figures were understated, and did not fire as many shots as planned. In addition, the report said, the Army updated computer predictions of casualties and "catastrophic kills," but did not provide that information to Congress.

Army officials said they could not comment because they have not seen the GAO report. But they have said before that they avoided firing weapons at the Bradley which they knew would blow up the vehicle because little would be learned from such tests.

The Bradley has emerged as one of the Army's most contentious requests this year. Army officials, who say they need the tracked vehicle to complement the M1 Abrams tank, have bought almost 3,000 Bradleys and want to buy about 4,000 more for $13 billion.

Because the vehicle is supposed to fight as well as carry infantry, it is equipped with guns and antitank missiles. Critics charge that the dual mission forces soldiers to ride into battle alongside stowed explosives, making the Bradley a dangerous proposition for the GI. The Army conducted its first live-fire tests last fall, and will run a second series this spring. After the tests, the Army designed a $75,000-per-vehicle improvement package that will be fitted onto existing Bradleys to make them less vulnerable.

"Critics will argue that the fact that we plan modifications proves that the Bradley is unsatisfactory as is," an Army report said. "In fact, the test proved that the Bradley is pretty damned good just as it is."

But Rep. Denny Smith (R-Ore.), a Bradley critic, said the GAO report suggests the Bradley may meet the same fate as the Divad (Sgt. York) antiaircraft gun, which Defense Secretary Caspar W. Weinberger canceled after test results were disappointing.

"I thought the Pentagon had learned a lesson on the Sgt. York," Smith said. "If the spring tests aren't run and reported honestly, the Bradley could become the next casualty of Gramm-Rudman."