A simple plumbing leak foiled last month's test of an experimental interceptor that is central to the Pentagon's effort to develop a national missile defense system at a cost of $12.7 billion over the next six years, investigators have concluded.

The conclusion suggests that no fundamental scientific flaw caused the interceptor to miss its target, a dummy warhead streaking over the Pacific. But the failed intercept attempt has forced a delay of at least several weeks in the next test of the anti-missile system--now scheduled tentatively for the second half of May--as the Pentagon and defense contractors make certain they understand what went wrong and take steps to avoid a recurrence, Pentagon officials said.

The delay, in turn, threatens to upset President Clinton's plan to decide this summer whether to build the system, beginning with construction next year of silos and other facilities in Alaska for the first 100 interceptor missiles.

The system is intended to shoot down a small number of ballistic missiles of the type being developed by North Korea and Iran. Whether it is ready--and reliable enough to justify its high cost and disruptive effect on arms control agreements--is being debated in Washington and on the presidential campaign trail.

Advocates of the system can take heart that the Jan. 18 test failure was caused by a mechanical glitch, not something that calls into question the basic physics and design of the proposed anti-missile shield.

On the other hand, the crippling of the interceptor by something as mundane as punctured tubing may be seen as evidence that the system is just too complicated, and its development too rushed, to be reliable.

White House officials said yesterday there has been no change in Clinton's self-imposed deadline for a decision in mid-summer on whether to proceed from development to deployment. But after a direct hit last October in the first intercept attempt, followed by last month's failure, administration officials had been counting on a third test this spring to help gauge the system's feasibility.

The Pentagon is due to deliver its own assessment in June, ahead of Clinton's decision. That schedule now appears to leave little time to pore over results of the next test.

Aside from the plumbing leak, almost everything else went right during the January test, including first-time use of satellites and ground-based tracking radar to guide the interceptor, officials said.

The leak occurred in a small metal tube that conveys nitrogen gas to refrigerate a pair of infrared sensor panels in the interceptor. The sensors serve as the interceptor's eyes, spotting enemy warheads by detecting heat signatures and other features.

With the sensors not working, the interceptor or "kill vehicle," launched from the Marshall Islands, ended up essentially blind in the final seconds of its flight and shot past its target.

It was a narrow miss, with the kill vehicle passing within 50 to 150 yards of the target. This left an impression that even in its damaged condition, the kill vehicle had used one remaining non-infrared, or "visible," sensor to distinguish the mock warhead from two other nearby objects--a decoy balloon and a container that had carried the dummy warhead and balloon into space.

Actually, though, happenstance played a large part in how close the interceptor got, according to defense officials.

The visible sensor was not programmed to pick out the mock warhead, but simply to center on whatever objects appeared in its field of vision. Initially, it saw all three objects, and the imitation warhead happened to be positioned in the middle.

As the kill vehicle approached, the two outside objects fell out of the picture. The sensor then focused on the mock warhead.

"If something else had been in the center--very honestly, it just as likely would have gone to that object as it would have gone to another object," said a senior official involved in developing the weapon.

In the last few seconds of flight, the kill vehicle went into a final homing mode that depends on infrared sensor data to compute the target's exact location. Not getting any infrared readings, the kill vehicle shut down its final guidance program and "went into a coast mode," the senior official said.

One of the main technical challenges raised by the test is how to avoid what engineers call a "single point failure"--that is, the loss of one component crippling an entire system.

With two infrared sensors on board the kill vehicle, if one fails, the other still can guide the spacecraft. But both sensors draw their gas coolants from a single network.

The Pentagon and Raytheon Co., which produces the kill vehicle, have begun considering ways to strengthen the plumbing and possibly even to add a second cooling network. "For the long term, for the reliability of the system, we have tasked Raytheon to look at how we can modify the design to eliminate a single point failure in the flow of gas to the sensors," the senior official said.

For now, though, the plan is to recheck all the plumbing in the next kill vehicle to fly.

Nothing of the last kill vehicle remains; it burned up reentering Earth's atmosphere. So investigators have had to deduce what went wrong by combing through extensive telemetry data, information beamed from the kill vehicle and collected by monitoring equipment.

The data revealed what one official described as a "slight rotation of the vehicle that shouldn't have been there" as it sped through space. Based on the nature of the rotation and other information, investigators concluded that a tiny jet of leaking gas in the vicinity of the coolant tubing was to blame.

Even before last month's setback, some members of Congress from both political parties had begun asking why a presidential decision had to be made this summer. European allies, concerned about stiff Russian and Chinese opposition to the U.S. project, also have questioned the urgency.

Defense officials say a decision is necessary this year if the system is to be operational by 2005--in time to address what U.S. intelligence analysts predict will be the eventual possession of intercontinental missiles by North Korea and Iran.

This Pentagon timetable allows three years for building the interceptor site in Alaska and a fourth year for moving the interceptors into place and training military crews to operate the system. In order for ground to be broken next spring, defense officials say, contracts must be awarded later this year.

And before construction begins, officials say, there must be agreement with Russia to amend the 1972 Anti-Ballistic Missile Treaty, or the president must decide to abrogate the treaty and give six months' notice of U.S. intent to withdraw.

Since 1983, the United States has spent more than $60 billion on missile defense research.