Before attacking Iraq in March 2003, the United States told the world that Saddam Hussein had reconstituted his nuclear weapons program in defiance of the United Nations. That claim, used to justify the war, was based on assessments provided by the United States intelligence community. But as everyone now knows, those assessments were wrong. So Jeffrey T. Richelson's history of American nuclear intelligence, including our attempts to learn about Iraq's nuclear program, could hardly be more timely.

In "Spying on the Bomb," Richelson, the author of several books on American intelligence, has brought together a huge amount of information about Washington's efforts to track the nuclear weapons projects of other countries. He examines the nuclear projects of Nazi Germany, the Soviet Union, China, France, Israel, India, South Africa, Taiwan, Libya, Pakistan, Iran and North Korea, as well as Iraq. Through interviews and declassified documents as well as secondary works, he sets out briefly what we currently know about those projects and compares that with assessments of the time.

This may sound like heavy going, but Richelson writes with admirable clarity. And along the way he has fascinating stories to tell: about plans to assassinate the German physicist Werner Heisenberg during World War II; about discussions in the Kennedy and Johnson administrations on the possibility of attacking Chinese nuclear installations; about Indian measures to evade the gaze of American reconnaissance satellites; and about the bureaucratic infighting over the estimates on Iraq.

The United States has put an enormous effort into gathering information about the nuclear projects of other countries. After World War II it equipped aircraft with special filters to pick up radioactive debris from nuclear tests for isotopic analysis. It created a network of stations around the world to register the seismic effects of nuclear explosions. Most important, in 1960 it began to launch reconnaissance satellites that could take detailed photographs of nuclear sites in the Soviet Union and China. Richelson occasionally speculates about the role of communications intercepts and of spies, but these appear from his account to have been much less important than the other methods of collecting information.

Through these means the United States has gathered a vast quantity of data, sometimes to surprising effect. Intelligence played a crucial role in the cold war, for instance, by reducing uncertainty about Soviet nuclear forces. Alongside such successes, however, there have been failures. One notable example concerned the first Soviet test, which took place in August 1949, much sooner than the C.I.A. had predicted. Another was the failure to detect Indian preparations for tests in May 1998, even though at an earlier time the United States, with the help of satellite intelligence, had managed to learn about preparations the Indians were making and to head off their tests.

But the most serious failure of all was in Iraq in 2003, because in no other case did the intelligence assessments serve as justification for the use of military force. The information needed for avoiding political surprise is one thing. That needed for preventive war is quite another, if only because of the consequences of making a mistake.

Beyond making the uncontroversial recommendation that "aggressive and inventive intelligence collection and analysis" should continue, Richelson draws no general conclusions. That is a pity, because his rich material points to issues that cry out for further analysis. He suggests in one or two cases that failures sprang from the mind-set of the intelligence community, but he does not elaborate on this point. He has little to say about relations between policy makers and the intelligence community, even though the quality of intelligence and the use made of it depend heavily on that relationship.

His focus is no less narrow in his discussion of foreign nuclear projects. He concentrates on the programs themselves, paying very little attention to their political context. Does that reflect a technological bias in nuclear intelligence? Would, for example, the prewar assessment of Iraqi nuclear capabilities have been more accurate if it had paid more attention to the broader political and economic circumstances of Hussein's regime?

The task of intelligence has become more complex than it was during the cold war. A single dominant nuclear opponent has now been replaced by a number of nuclear states, along with states and stateless terrorists that are aiming to get their hands on nuclear weapons. Meanwhile, the technology needed for producing nuclear weapons has become easier to acquire.

Many critics believe the recent performance of the intelligence community shows it has not responded adequately to this new situation. Richelson does not have much to say on this question; nor does he discuss the likely impact of the current reforms, initiated in response to the Iraq war, on the quality of intelligence. His reticence may imply that he does not think reform is necessary. Still, it is disappointing that he does not draw on his historical survey to discuss whether new approaches are needed for dealing with nuclear threats, and, if so, what those new approaches might be.

You burst into the room. Sitting on a chair, blindfolded, his hands tied behind his back, is your prisoner. The room is dark, except for a lonely naked bulb hanging from the ceiling. He is sweating. He is afraid.

"Tell me where it is!" you scream. "Now!" You know there is little time left. Somewhere in your city, a time bomb is ticking. Whether it spits serin into the air, uranium into the water or atomic fire into the heavens, you do not know.

He does. But he is not talking. Involuntarily, you raise your hand as if to strike. What you are about to do violates the law and your conscience. And yet...

In peacetime, torture ranks next to murder as a primal sin. But during war, the debate begins over whether this evil can ever be justified to combat the seemingly greater evil of the enemy. Harvard law Professor Alan Dershowitz has said torture should be legalized.

In early October, the U.S. Senate voted 90-9 to ban it. Although Secretary of State Condoleezza Rice and President Bush have both recently asserted that "We do not torture," five U.S. Army Rangers were charged in November for punching and kicking detainees in Iraq, secret U.S. prisons have caused anxiety in Europe, and Vice President Dick Cheney has battled to win the CIA an exemption from the torture ban. As late as December, the U.S. House of Representatives stood poised to defeat the White House.

Few of us will ever be asked to torture. But, indirectly, all of us have to make a choice: to support, as citizens, those politicians who back torture, or those who seek its prohibition.

The decision of an individual to support, or reject, torture seems at first to be a purely moral question. But what would be the long-term consequences to society if we were to make this radical break with the past?

One cannot do experiments with societies, or predict the future, but, it turns out, one can attempt to address this issue using the cold, hard tools of mathematics and logic. This story begins in 1963.

The United States and the Soviet Union are on the perpetual brink of war, balanced like two sides of an equation. On the American side are "game theorists" like Thomas Schelling, recently awarded the Nobel Prize for his work on the strategy of conflict. On the Soviet side, there is the solitary mathematical psychologist Vladimir Lefebvre.

Just as mathematics could be used to describe logical reasoning, Lefebvre saw that mathematics could be used to describe ethical reasoning. If something was good -- for example, "church," "democracy," "prosperity," "kindness" -- it had value "1."

If something was evil -- "earthquake," "famine," "military defeat," "murder" -- it had value "0." But rarely were ethical situations so simple. For instance, "killing" is bad (0) but protecting one's country is good (1) -- so is war 1 or 0?

Lefebvre saw that, at the crudest level, there were essentially two types of ethical systems. Those that held that employing evil means to attain just ends was good, and those that saw that employing evil means to attain good ends was wrong.

There were also, crudely put, two types of relations between individuals: those entailing compromise (or cooperation) and those entailing confrontation.

Of course, evil people rarely see themselves as evil. So Lefebvre had to incorporate in his model of human nature the capacity of human beings to judge -- correctly or incorrectly -- the goodness or evil of their own acts, and to reflect upon their own judgments, and others'. "Reflexive Theory" was born.

It quickly became a paradigm within the Soviet defense establishment, with the publication of books such as "Mathematics and Armed Conflict." Nothing like it was known in the West.

With very simple assumptions -- for instance, that an individual who correctly sees his actions to be good when they are good, and evil if they are evil, is more highly regarded by society than an individual who incorrectly sees himself -- Lefebvre showed that in a society that accepted the compromise of good with evil, individuals would more often seek the path of confrontation with each other.

Lefebvre's insights were called upon by the State Department during negotiations with Mikhail Gorbachev in Reykjavik, Iceland. (And perhaps Lefebvre's model could be re-enlisted to help U.S. officials understand and negotiate with Arab and Muslim heads of state, who must also negotiate with their people.)

In support of Lefebvre's revolutionary new theory, a survey of Soviet émigrés and Americans was conducted in the 1970s. They were asked questions like, "Should a doctor conceal from a patient that he has cancer in order to diminish his suffering?" Overwhelmingly, the Americans would say no, and overwhelmingly, the Soviets yes. The Soviets accepted the compromise of good with evil; the Americans rejected it.

What does this mean? If Americans begin to accept the use of torture, American society might turn into a society of individuals in conflict.

Not uniformly, thanks to something called free will, but generally, with harmful consequences for society: Imagine two roads, with a stream of cars moving along each one. Each driver wants to reach his destination as quickly as possible; on occasion, drivers will impede each other.

On the first road, drivers rise in their own, and in other drivers', estimation if they yield. Drivers on the second road lose face when they yield. It is clear that traffic will move faster on the first road than on the second.

It can be argued that repressive states like Saudi Arabia, which bred most of the Sept. 11 hijackers, are on the second road. If the United States moved to accept torture, it could veer toward the second road, too -- the road of the Soviet Union.

And we know where that road ends. The Soviet Union no longer exists.