The Pershing II missiles that will be scrapped under terms of the U.S.-Soviet Intermediate-Range Nuclear Forces Treaty might not be able to hit their targets anyway, according to critics who charge that the weapon has not been sufficiently flight-tested.

The Army claims the Pershing II can fly more than 1,000 miles and deliver its nuclear warhead with near-pinpoint accuracy -- with at least half the missiles hitting within 120 feet of their targets.

The Pershing II's test record, however, reveals there is no ground-to-ground operational flight test at maximum range to substantiate the Army's claims. Army spokesmen vigorously deny that the Pershing is inaccurate and say it has been extensively flight-tested.

The Pershing is a two-stage missile with a unique radar guidance system to maneuver the nuclear warhead to its target. The radar, which turns on during the last 12 seconds of flight, is key to the missile's accuracy. An on-board computer compares the image seen by the radar to a stored digital map of the target area and generates guidance corrections to aim the plummeting warhead directly at the target.

In an early developmental flight test conducted at White Sands, N.M., in 1978, the Pershing "came so close it was amazing," said David Harris of the Army's Missile Command, the testing agency. "We publicly released photos of other successful tests in 1983," he said, "and nobody paid any attention."

Testing experts say there is a vast difference between the kind of developmental tests the Army has publicized and operational tests, which are far more demanding.

There are three big differences, according to retired vice admiral Robert Monroe. Operational testing, he said, is performed in the field, not the laboratory. It is performed with the troops who will use the weapon, "not technicians or engineers." Most important, the testing is done "against a simulated enemy who is being as unpleasant as he possibly can, using countermeasures and every dirty trick in the book."

The Pershings were rushed into operational service in West Germany two years before the first maximum-range operational test-firing in December 1985. At that time, a two-stage Pershing was fired from Cape Canaveral to a spot in the Atlantic Ocean. Since then, Harris said, the Army has conducted 31 operational flight tests. However, 23 of those flights were conducted from Canaveral, where the radar guidance system did not fly against real targets, or against countermeasures, but instead flew preprogrammed maneuvers before hitting the water.

"Those 23 flights were much closer to developmental tests than operational tests," a testing expert said. The other operational flight tests, eight in all, were conducted at the White Sands test range, with single-stage Pershing IIs, where the missile's radar seeker was required to find and home in on a real target. Harris says White Sands isn't big enough to fire the two-stage Pershing.

"Any weapon with artificial intelligence develops a personality," said one expert. "Things creep into the hardware and software, and you may not even know it. And with single-stage tests, the altitude, temperature and vibration effects are not the same."

"You must test the complete missile with the second stage," said another expert. "You have to test for things like voltage surges at first-stage separation, the higher dynamic pressures, things like that."