LATE ONE SEPTEMBER night in 1983, as shells landed near the U.S. ambassador's residence outside Beirut, Robert C. McFarlane had a chance meeting with his host, ambassador Robert S. Dillon, and showed him a cable he had drafted recommending the use of U.S. air strikes and naval gunfire in "self-defense" of the embassy and the U.S. marines in Lebanon.

The surprised ambassador said he'd like to think about the implications of this proposed new use of American force. But McFarlane, then deputy national security adviser, was in a hurry. He rushed to the radio room at the rear of the residence to send what later became known to administration officials as the "sky-is-falling cable."

McFarlane's cable was a small step on the road toward the NSC's cult of secrecy -- and the eventual Iran-contra debacle. The naval gunfire was approved by President Reagan, despite Pentagon reluctance, ostensibly to check the influence of Syria, a perceived Soviet proxy. The first volleys from ships off the coast made the United States a combatant in Lebanon; soon afterward 241 U.S. marines were killed in their Beirut barracks.

The episode at the embassy compound illustrates many of the points made by internal critics of President Reagan's management of foreign policy through the National Security Council, which was established in 1947 to integrate domestic, foreign and military policies of the executive branch. Long before the Iran-contra affair,the critics say, the Reagan NSC process was overly secretive. Senior NSC staff officials often overrode or refused to consult their own experts inside the government and often took action outside conventional channels. These secret maneuvers, frequently in the name of combatting communism, were the flaws that flowered into the scandal now being probed by Congress.

"The NSC really began to go operational right then, when your peace negotiator became an artillery spotter," said one of McFarlane's NSC colleagues from the Lebanon period.

When McFarlane became national security adviser himself in October of 1983, he and John M. Poindexter, his chief aide and eventual successor, turned increasingly inward, officials said. They used a small clique of NSC staffers, including Oliver L. North, to accomplish covertly what they could not do through regular policy-making channels.

The "management style" of McFarlane and Poindexter has received far less attention than that of Ronald Reagan. But it's no less important, because the two aides were, in many ways, the twin architects of the Iran-contra disaster. Several officials cited their compulsion for secrecy, encouraged by such senior actors as CIA Director William J. Casey and White House Chief of Staff Donald Regan, as a critical weakness.

Another weakness, according to these officials, was an overreliance on a few like-minded staff members housed in a political-military section of the NSC that was created in 1983 and initially headed by Donald Fortier, whom McFarlane had brought over from the State Department. It included North and later Howard R. Teicher and, as a consultant, Michael A. Ledeen.

This small group, plus Poindexter, were McFarlane's "soulmates," according to a U.S. official who watched the interaction at close range. They had special access to him and to the most secret of his clandestine plans. For these officials, it was such a high-pressure existence that, "after three years you either burn out or think you're the president," one NSC colleague close to that small group recalled.

McFarlane, Fortier, Teicher and Ledeen had worked together at State under Secretary of State Alexander M. Haig Jr. and shared a Haigian world view that stressed closeness to and confidence in Israel, passionate concern about international terrorism, a determination to eradicate the post-Vietnam reluctance to use U.S. military force and deep frustration with the U.S. government bureaucracy, according to colleagues. To one degree or another, these tendencies were evident in the U.S. policies forged by this group in Lebanon, Libya and Iran.

"At the State Department, {this group} could come up with 10 ideas for every one that was given a serious hearing, but now suddenly at NSC there was no filter between them and the ability to act," said an official who served with them. Another former NSC colleague said the "soulmates" operation at the White House in 1983-86 turned out to be "Haig's revenge" on the administration that forced him out in mid-1982 but left his proteges in power.

As national security advisers to the president in earlier administrations, Henry A. Kissinger and Zbigniew Brzezinski had wielded great power and authority and were often considered more important than the Cabinet secretaries. But when the Reagan administration came into office in 1981, it initially downgraded the position in an effort to avoid bureaucratic conflict and strengthen "cabinet government." Reagan's first national security adviser, Richard V. Allen, all but disappeared from public view and was forced to report to the president through White House Counsel Edwin Meese III.

Haig as secretary of state was running his own initiatives with little White House input or clearance in the early months, and McFarlane was his secret emissary. Haig had learned to operate in secret outside normal channels in his days as a trusted assistant to Kissinger. Haig in turn imparted these traits to McFarlane, who initially operated from the post of State Department counselor. In this daisy chain of secret operations, McFarlane later would look to his subordinates, his "soulmates" at the NSC, to carry on the back-channel maneuverings.

A prominent foreign ambassador posted in Washington said Haig informed him shortly after becoming secretary of state that McFarlane would be the channel for secret messages and dealings too important to be conducted through normal channels.

"McFarlane used to take lots of secret trips for Haig without the White House knowing," said one ex-official. For example, he said, early in 1981 McFarlane traveled to Pakistan and Saudi Arabia for Haig, without Allen or other White House officials being informed.

Haig would then hand a report of the trip to Meese, who would file it in his briefcase and eventually -- sometimes weeks later -- hand it on to Allen rather than the president. Allen in turn would hand it on to NSC aides, who would call their counterparts in State Department geographic bureaus, who often never had heard of the McFarlane secret missions.

In January, 1982, Allen was succeeded by William P. Clark, a longtime aide and friend of Reagan who had been Haig's chief deputy at State. "Clark changed at least one thing. Now you had a national security adviser who not only had access to the president but who could speak for him," recalled retired general Brent Scowcroft, who was President Ford's national security adviser and a member of the Tower Commission, which studied the Iran-contra affair.

Clark brought McFarlane with him from State as his deputy and also inherited Poindexter, who had joined the NSC staff as Allen's military assistant. Poindexter had a PhD in nuclear physics, but had never been assigned to a policymaking job requiring political skills.

One of Poindexter's initial tasks for Clark was an attempt to stop leaks with an order sharply limiting contact between officials and reporters. The Poindexter plan would have required advance approval for nearly all press contacts throughout the executive branch involving "NSC information" and a "memorandum of conversation" for higher-ups after the interview.

Clark informally presented Poindexter's handiwork, in the admiral's presence, to several senior diplomatic correspondents, who denounced it as unrealistic and unworkable.

The press-control aspect of the plan was hastily dropped but not before embarrassing Poindexter, complicating his future press relations and illustrating his inexperience. (His reputation with reporters suffered again when he sought to mislead them deliberately about the invasion of Grenada and when he approved a plan of proposed "disinformation" aimed at destabilizing the Libyan regime of Moammar Gadhafi.)

McFarlane took over as NSC advisor in October of 1983 after Clark was nudged out of the NSC because of differences with other top White House aides. McFarlane, like Poindexter after him, was named to the job after feuding by more senior officials cut down more prestigious rival candidates. Neither McFarlane nor Poindexter felt he had the clout within the administration to resolve the often-deep differences between the Cabinet "heavies," Secretary of State George P. Shultz and Secretary of Defense Caspar W. Weinberger.

McFarlane told the Iran-contra committees that he didn't "have the guts" to stand up and tell the president his secret Central America policy wouldn't work because he feared Casey, Weinberger or U.N. Ambassador Jeane Kirkpatrick "would have said I was some kind of commie."

Having a strong NSC chief was important because Reagan refused to choose between opposing views of his Cabinet officers or to enforce his decisions if he was able to make them. Unable to establish and maintain clear lines of policy in disputed areas, McFarlane and Poindexter took some key issues underground, cutting out the rest of the government, including many officials of the NSC itself.

The watchword at the NSC in McFarlane's day was "compartmentalization." This meant, for example, that secret dealings with Iran were kept from James R. Stark, Shirin Tahir-Kheli and Dennis Ross, Mideast experts who shared responsibilities and in some cases offices with North and Teicher. Teicher told Ross in the spring of 1986 that he was "not authorized" to discuss one highly secret project which he felt Ross, as chief of the Middle East section of the NSC staff, really should know about -- the arms sales to Iran.

Stark and Ross were present at an interview on Aug. 20, 1986 when Teicher flatly denied to a Washington Post reporter that the United States was covertly helping to supply arms to Iran, either directly or through Israel, or that the NSC was dealing with Iran on a private track through North. By then, Teicher and North had been to Tehran with McFarlane on an arms-trading mission, something Stark and Ross said later they did not know.

On paper, the NSC decision-making system included 25 senior interagency groups of high officials (SIG's) for discussion and resolution of issues, 55 interagency groups (IG's) and more than 100 other task forces, working groups and coordinating committees.

But by the time Poindexter settled in as NSC advisor in 1986, most of these committees were hollow bureaucratic shells. They met infrequently and rarely agreed on important lines of policy. Congress heard testimony that in the contra-support program, most of the discussion took place in a restricted interagency group (RIG) which at key points was limited to only three officials.

Poindexter could be even more secretive than McFarlane and was less inclined than his predecessor to deal with Congress or reporters. McFarlane, who was deeply interested in arms-control negotiations with the Soviet Union, held "weekly, interminable, Cecil B. DeMille meetings" of interdepartmental groups on this subject, said a participant, but Poindexter "would call in 10 participants one by one" rather than hold a meeting.

"Poindexter was paranoid about leaks," the same official said. "He'd call me in and spend part of the meeting telling me to be sure and not tell something to somebody else -- 'Don't mention this to so and so.'" A computer enthusiast, Poindexter tended to do business with individual staff members by electronic communications. This reduced the chances of leaks, but also reduced chances for critical consideration and advice from those who were not already committed to the enterprise being discussed.

If Poindexter was suspicious of meetings and open discussion, he was enthusiastic about titles. At the time he was forced out of the NSC, Poindexter was being assisted by four deputy national security advisers, one special counsel, 10 special assistants to the president, 14 senior directors of the NSC staff, 40 directors and nine deputy directors.

Zbigniew Brzezinski, who operated with an NSC professional staff of about 40 (with less exalted titles), said the Reagan staff reminded him of "a bank in which every teller held the rank of vice-president."

In part to prevent leaks, Poindexter orchestrated a new "top-down" approach to arms control policy-making last summer. He and the top Cabinet officials agreed on an initiative to Gorbachev that was then handed down to working-level officials. Last fall, when the Soviet leader surprised the administration with his proposal for a quick summit at Reykjavik, Poindexter was responsible for the preparations, which proved to be inadequate to the task, according to many sources.

"Reykjavik happened in part because we had no philosophy for arms control and what part it was supposed to play in U.S.-Soviet relations or what our overall strategic policy is. That was impossible to develop because you had a secretary of state and a secretary of defense with fundamentally different perspectives and a president who was not willing to decide between the two," said Scowcroft.

After the resignation of Poindexter last Nov. 25, Reagan turned to a Washington veteran, Frank C. Carlucci, to impose order on the shell-shocked NSC staff as his fifth national security adviser. Carlucci took over the job on Jan. 2 amid high expectations that he could make the institution work. But the current controversy over U.S. military operations in the Persian Gulf has raised doubts about whether Reagan has yet found the key to coherent foreign and defense policymaking in his administration.

Charles Babcock and Don Oberdorfer are staff reporters of The Washington Post.