The crash of an Air Force radar jet in rural Maryland last month that killed 21 people was caused by an undetermined malfunction in the aircraft's flight control system, the Air Force said yesterday.

Brig. Gen. Peter W. Odgers said Air Force investigators had concluded that a malfunction in the system that controls the jet's stabilizers pitched the four-engine plane into a nose dive that the pilot was unable to stop.

"How that happened is unknown," Odgers said at a press conference at Andrews Air Force Base. "I would have to say [the cause] was mechanical."

Odgers added that a factor in the crash was confusion in the cockpit immediately after the jet's nose dropped abruptly. The plane had been cruising smoothly at 29,000 feet on automatic pilot when the pitch-over created sudden weightlessness on the flight deck. This cut off electrical power and probably also hurled debris and unstrapped-in passengers against the jet's ceiling, Odgers said.

For a minute and a half after its nose tipped, the jet plunged toward the ground at a speed of more than 330 feet per second until it exploded above the ground and slammed into rye and barley fields north of Walkersville in Frederick County, Md.

Odgers said the pilot's wife was in the cockpit when the mechanical failure occurred and that another crewman's wife may have been there also. Those in the cockpit included the pilot, Capt. Joseph C. Emilio, and his wife Peggy, who was strapped beside him in the copilot's seat. Capt. Donald V. Fonke was strapped in the navigator's seat behind the pilot and Fonke's wife Linda may have been standing nearby, Odgers said.

The women were aboard as authorized observers as part of an Air Force program to give spouses a chance to see what wives or husbands do. The program was suspended June 4 and is being reviewed by the Air Force.

When the nose pitched, the plane began to accelerate at a 30-degree angle, a condition known as "runaway nose-down trim," Odgers said. After 8 to 10 seconds, he said, the plane cannot recover from this position because the wing flaps that could pull the plane out of a dive simply cannot overcome the speed of the plummeting jet.

At 5,000 feet there was a fire or an explosion, Odgers said. The jet banked left and at 1,300 feet a "major explosion" occurred when jet fuel in the main tanks ignited. To eyewitnesses on the ground, the plane appeared as "a red and orange ball of fire hanging in the fog" before it crashed.

"The most difficult part is what happened during the pitchover," Odgers said. "One of the things we don't understand is the confusion in the cockpit. What inhibited him [the pilot] from taking more aggressive action to raise the nose? Eight to 10 seconds isn't long if you have debris and people floating around. If the pilot caught it immediately, if he said, 'This is what's going on,' he probably could have done it."

But, Odgers added, investigators had duplicated the events leading up to the crash in simulators and some of the pilots didn't recover. He said Captain Emilio had 3,000 hours flight experience.

"The outcome would have been no different had the copilot's seat been empty," Odgers said referring to the presence of Emilio's wife in the copilot's chair. Asked whether the pilot would have had a better chance to stop the jet's descent if a copilot had been beside him to help, Odgers declined to speculate because investigators do not know what happened on the flight deck.

The jet, an EC135N missile tracking plane, was on a routine training mission May 6 out of Wright-Patterson Air Force base in Dayton, Ohio. It disappeared off ground-based radar screens at 10:51 a.m. There had been no reports of problems.

During the five-week inquiry into the crash, the Air Force interviewed more than 20 witnesses and, inasmuch as possible, reconstructed the shattered plane in a cavernous concrete hangar at Andrews Air Force Base.

The findings have been compiled in two reports that will be sent to the Inspector General of the Air Force.

Odgers declined to discuss the reports' recommendations. Officially, he said, no procedural changes have been adopted for the 749 other similar jets operated by the Air Force.

The horizontal stabilizers on the jet's tail are adjusted by a motor that moves a screw up and down. Searchers recovered the motor, which was displayed at yesterday's press conference, as well as cockpit indicators that showed that the stabilizers had been driven to the maximum "nose down" position.

The flight control system was being guided by the auto pilot, which keeps the plane on its course and at its altitude. "We think the auto-pilot drove the trim system to a more nose down position," Odgers said.

Investigators found that a switch in the cockpit that cuts off power to the trim system and thus prevents it from further changing the pitch of the plane had been thrown, but too late -- the stabilizers had already reached the pitch-over angle and, 10 seconds later, sealed the fate of flight.