"What best emphasizes the complexity of Dense Pack," says presidential science adviser George A. Keyworth about the latest scheme for basing the MX missile, "is the fact that people are high on it on Monday and down on it on Wednesday."

However, Keyworth said he is now pretty high on the controversial plan, which calls for placing 100 MX missiles in new underground silos crowded together with extra concrete and steel protection against the blast of attacking enemy missiles.

The theory is that the Soviets could not effectively attack the silos simultaneously because the blast and debris from the first missiles to arrive would destroy the following ones or blow them off course. Therefore, a number of MX missiles would survive and could quickly fire back at the Soviets.

"I feel very confident about it right now," said Keyworth, 42, a former atomic weapons scientist whose views may be very important in the final decision on whether to go ahead with Dense Pack. "But my skepticism has not been reduced to zero."

After several weeks of intensive study, he said, "I think people are just beginning to understand the myriad complexities" about Dense Pack and about whether the Soviets could figure out how to attack the system successfully.

In another interview a month ago, Keyworth said, "I really believe that there is no one . . . and I emphasize no one . . . in this country today who totally understands" those complexities.

Another senior official with extensive military background also attested to the ups and downs resulting from the sheer complexity of the Dense Pack concept. When he first heard the plan discussed some months ago, he said, "I thought it was a crazy idea." Since then, he said, his confidence has gone up and then dropped back a little.

The Reagan administration is now in the final phase of analyzing the $23 billion scheme, which is known formally as "closely spaced basing." A skeptical Congress has rejected numerous basing plans because they did not provide confidence that the missile could survive a surprise Soviet attack and did not close the so-called "window of vulnerability."

President Reagan campaigned hard two years ago on a pledge to close that window, an allusion to the possibility that the 1,000 U.S. land-based Minuteman missiles could be wiped out in a first strike by increasingly accurate Soviet missiles.

The president has promised Congress an answer on the future of MX and Dense Pack by Dec. 1. The decision will be heavily influenced by the recommendations of Secretary of Defense Caspar W. Weinberger, White House national security adviser William P. Clark and a panel of outside experts set up by Weinberger.

A decision to bet on Dense Pack, however, involves probably the most technically complicated problem ever to surround a major new weapon. The solution depends on the effect of one nuclear weapon, and the debris it kicks up, upon another -- an effect impossible to assess realistically in underground nuclear testing. The plan has many critics outside the administration and some skeptics within, reportedly including Weinberger.

Within the administration, Keyworth -- a nuclear physicist who spent 13 years at the government's Los Alamos, N.M., weapons laboratory -- probably is the civilian most qualified to judge the technical merits of the plan.

The White House is eager to get MX deployed to bolster its claims of strengthening the U.S. strike force, to balance the first-strike threat it sees Soviet missiles posing, and to gain bargaining strength in negotiations with Moscow on reducing atomic missile forces.

Politically, Keyworth is a true "Reaganaut," saying that his admiration for both Reagan and presidential counselor Edwin Meese III, through whom Keyworth reports to the president, "goes to a rather extreme degree."

So what would happen, he was asked, if it turns out that he has serious technical reservations about a system that he senses the White House, perhaps even more than the Pentagon, really wants?

"It's not so difficult," he said, "and the reason is because I am not in the decision process. If my opinion were to diverge from someone else's, I believe that would be interpreted as the best opinion I can provide from the perspective I have . . . as a scientist . . . and that is not the whole perspective. I try to stick to the issues that I understand and do not sound off as an expert on aspects of issues where I am not an expert.

"What we are looking for right now is the Achilles heel," some fatal technical flaw that could shoot down the Dense Pack concept. "But it doesn't appear to be there," he said. "Unless we find an Achilles heel, I think Dense Pack will be a very effective and relatively inexpensive mode of achieving substantial progress toward closing the window of vulnerability."

In a way, he said, the "very complexity" of Dense Pack will increase the uncertainty in Moscow about how well it would work, thereby decreasing the likelihood that anyone would gamble on attacking it.

The uncertainty works both ways, however. Even though the Air Force has claimed it can "super-harden" the Dense Pack silos to withstand very high pressures from an atomic attack, Keyworth said "there is not likely to be a positive yes-or-no answer to that. . . . There will always be uncertainties."

What is certain, he said, is that Soviet missiles "possess a very real threat" to the survival of land-based missiles in fixed locations "and there is no way in modern technology to completely nullify that threat."

Keyworth said, "The real issue about Dense Pack right now is, first, does it possess more deterrence and is it less vulnerable, better protected, against Soviet attack than conventional silo basing like Minuteman. Secondly, if that question is answered satisfactorily, and I believe it will be, then you ask how far does it go" in doing so.

If Dense Pack works, Keyworth said, it would force the Soviets to prolong their assault, thus giving an American president the option of launching MX missiles while they are actually under attack. A policy based on "launch under attack," however, could be very controversial because it conveys the image of a quickly escalating all-out nuclear war.

The United States, as Pentagon officials point out, has always tried to design its forces, including MX, so they could survive an attack and then fire back. There has always been private discussion about the usefulness of the launch-under-attack option in deterring the Soviets from attacking in the first place. It has never become declaratory U.S. policy, although it has always remained an option that Moscow could not be sure about.

As Keyworth sees it, the president is not apt to launch a massive retaliatory strike if there are indications that one Soviet missile may have been launched. But neither is he apt to wait until after there have been 1,000 nuclear detonations in this country. "Obviously, the president is not going to sit back and watch our nation be annihilated before he retaliates, and the question is when you retaliate," he said.

"In that whole course of that first half hour or hour . . . , action taken by the president would be launch under attack," he said, adding that "I think it is very unrealistic to say we simply will not do that. . . . To tell the Soviet Union that the United States will not launch under attack is neither candid nor wise."

The final decision on how MX is based, he said, "will obviously determine whether launch under attack is an important part of the deterrence or an unimportant part. In the case of Dense Pack, I think the option of launch under attack may represent a substantial part of its overall deterrence."