The distinguished international relations theorist Kenneth Waltz (still) thinks that the emergence of new nuclear powers is not necessarily to be regretted. In a world of many nuclear powers, in Waltz's view, a major war would be practically impossible. He therefore believes that "more may be better"-that what has come to be called "nuclear proliferation" might actually be a good thing. Scott Sagan, a professor of political science at Stanford University and a leading authority in the field of strategic studies, disagrees; in Sagan's view, "more will be worse." In 1995, Waltz and Sagan published The Spread of Nuclear Weapons: A Debate, in which each laid out his argument and then responded to the other's argument. The discussion was spirited and often quite interesting, and the book deservedly received a good deal of attention. A revised edition, with new passages on terrorism and missile defense and a new chapter on India and Pakistan, has just been published. Its appearance provides us with a timely opportunity to confront the proliferation question yet again-a chance to get to the bottom of what is perhaps the most important international issue the world will face in the years to come.

Why does Waltz think that the spread of nuclear weapons is not necessarily to be avoided? His fundamental claim is that nuclear forces have a very powerful deterrent effect, and that nuclear states would therefore be extremely reluctant to tangle with each other. If one such state does take aggressive action-if a nuclear Iraq invaded Kuwait, for example-other powers would find it too dangerous to use military force against the aggressor state, and would only be able to take non-military measures. If the United States at the time of the Iraqi invasion of Kuwait in 1990 "had thought that Iraq might have had a few bombs, it would have had to manage the Iraq-Kuwait crisis differently, say by employing only an embargo." If, however, by some chance a limited war between two nuclear states did break out, that conflict would be very unlikely to escalate. Escalation that threatened an adversary's core interests, especially an invasion of the adversary's homeland with conventional forces, would be "too risky to contemplate." But if a country's core interests actually were threatened, then even that situation would scarcely lead to disaster. There would be no massive escalation. Instead nuclear weapons would be used in a very limited way. A "few judiciously delivered warheads" would in such circumstances be a very effective wake-up call, and would probably "bring rapid deescalation."

A nuclear world, according to Waltz, is thus a world of safety nets beneath safety nets. In the unlikely event that serious problems between nuclear states developed at one level of conflict, the dangers would very probably be contained before matters escalated to the next level. The mere possession of a nuclear capability, Waltz says, induces "caution in any state" and "especially in weak ones." A nuclear Iraq, he now writes (Libya was the state mentioned in the corresponding passage in the 1995 text), would actually be more cautious than a non-nuclear Iraq. The possession of nuclear weapons by an adversary, moreover, would have an extremely powerful deterrent effect. In Waltz's view, "not much is required to deter." "A low probability of carrying a highly destructive attack home", he believes, "is sufficient for deterrence." A large force is not necessary for this purpose; a relatively small number of bombs would do the trick. That small force, of course, would have to be able to survive an enemy attack, but since bombs can be small and light, they are "easy to hide and to move" and delivering them, even after a surprise attack, would not be hard to do. "Bombs can be driven in by trucks from neighboring countries", he writes. "Ports can be torpedoed by small boats lying offshore." Even "weak and poor states", he writes, can "easily" build small, survivable nuclear forces, and when they do, they will deter even the strongest nuclear powers. With nuclear weapons, he says, "any state will be deterred by another state's second-strike forces." Relative strength then no longer matters: "if no state can launch a disarming attack with high confidence, force comparisons are irrelevant." "A minimal deterrent", he writes, "deters as well as a maximal one."

The atomic bomb, Waltz believes, is thus the great equalizer in international politics. It is not just that a small nuclear force is as good as a big one. He now argues (in two passages that did not appear in the 1995 text) that the presence of nuclear forces also makes disparities in conventional military power meaningless. "Nuclear weapons", he says, "negate the advantages of conventional superiority because escalation in the use of conventional force risks receiving a nuclear strike. With nuclear weapons, not only is a small second-strike force equivalent to a large second-strike force but also small conventional forces are equivalent to large conventional forces because large forces cannot be used against a nuclear power."

In such a world, where neither the nuclear nor the conventional balance matters, the weak are as strong as the strong. Everyone is deterred-and deterred equally. No one will dare start a war, because the risks are so immense. There is thus practically no risk of war, or at any rate of a war in which nuclear weapons are used in a major way. In a nuclear world, a general war is a virtual "impossibility"; in such a world, "only limited wars can be fought" and "the probability of major war among states having nuclear weapons approaches zero." Nuclear weapons are thus the great bulwark of international peace; and it is because they have such a powerfully stabilizing effect that in Waltz's view the "gradual spread of nuclear weapons is more to be welcomed than feared."

Sagan does not think a nuclear world would be nearly as stable as Waltz makes out. Indeed, looking at South Asia, he thinks that deterrence will eventually break down-that nuclear weapons will someday be used in a conflict between India and Pakistan. For him, the root problem with Waltz's argument is the assumption that states are under very strong pressure to behave rationally, and that one can therefore assume that states in fact will act rationally. To Sagan, nuclear weapons are controlled by "imperfect humans inside imperfect organizations", and, given the way large organizations actually work, such a high degree of rationality simply cannot be expected. Nuclear weapons are in the hands of professional military organizations, and those organizations behave in only "boundedly rational" ways that "are likely to lead to deterrence failures and deliberate or accidental nuclear war."

In the 1995 version of his essay, Sagan focused on three dangers resulting from the "wide-spread biases and imperfections in military organizations." First was the problem of preventive war: when a country begins to build a nuclear force, one or more of its rivals might decide to launch an attack on that country's nuclear facilities before it is too late. Military officers, Sagan argues, are particularly attracted to this kind of thinking, and this, he says, can be a real source of danger, especially in countries where the armed forces are not under firm civilian control. The second problem has to do with the survivability of a nuclear force. In a crisis, vulnerable forces might invite enemy attack, and for various organizational reasons, professional militaries, if left to themselves, might not actually build invulnerable forces. The third problem has to do with the possibility of a nuclear accident; this includes the problem of unauthorized use, since such use is "accidental" from the standpoint of the central authorities. And there are limits, Sagan insists, to how accident-proof large organizations are likely to be. (Sagan is the author of an important 1993 book called The Limits of Safety, which lays out this argument in some detail.) The empirical record, he point outs here, shows that accidents are possible. An Iraqi bomb, the UN inspectors discovered in the early 1990s, would have been "highly unstable" and might have gone off "if it fell off the edge" of a desk. Such an accident, especially if it were to take place near the front lines in a war, might well in his view lead to a major nuclear exchange.

In the new edition, Sagan also talks about a fourth problem, although it can also be viewed as an extreme form of the problem of unauthorized use. This is the problem of terrorism. If "unstable states, especially unstable Islamic states", acquire nuclear weapons, they might fall into the hands of terrorist groups like Al-Qaeda. Since the bombs would have "no return address", the terrorists could use those weapons "without fear of retaliation." Deterrence would therefore not work. The best way to handle this problem is to prevent such states from developing nuclear capabilities in the first place. But this point simply underscores the seriousness of the problem and strengthens Sagan's basic conclusion that "more will be worse."

It is hard to quarrel with most of what Sagan has to say, but there is still something odd about his position. It is as though, in his view, proliferation as such is not a problem. Sagan, it seems, would have no objection to it if it were managed properly-if forces were made invulnerable to enemy attack and secure against unauthorized use, if the weapons themselves would not go off by accident, and so on. But to take that view is in effect to concede the heart of Waltz's argument: if the weapons were designed and deployed the right way, then a nuclear world really would be better than a non-nuclear world. The Waltz-Sagan debate seems to melt away: Nuclear deterrence, in that case, really could serve as the basis of a very stable international order.

But is that really so? How would a world in which a large number of countries have invulnerable nuclear forces actually work? The real issue here has to do with how political conflicts would run their course in such a world. Is there any reason to assume that that process would be more peaceful in a nuclear environment than in a non-nuclear one?

Waltz does not approach the problem this way. For him, wars are started by one side or another. There is an attacker and a defender; with nuclear weapons, the attacker is deterred and war is avoided. "Where nuclear weapons threaten to make the cost of wars immense", he asks, "who will dare to start them?" The Soviet Union would have been deterred by any state that might have been able to deliver one or two simple fission bombs on Moscow. Indeed, he argues, "with nuclear weapons, any state will be deterred by another state's second-strike forces." "A nation", he says, "will be deterred from attacking even if it believes there is only a possibility that its adversary will retaliate." There is no doubt in Waltz's mind about this; for him, the deterrent effect is absolute: no one will start a war, and wars-at least major wars, wars in which nuclear weapons will be used-will simply not happen.

In the real world, however, wars are often not simply "started" by one side, and the distinction between defender and attacker can be very problematic. In 1914, for example, who "started" the First World War? Germany, by invading Belgium and attacking France? Or Russia, by ordering general mobilization a few days earlier, knowing full well that such action made war virtually inevitable? Who was the "defender"? Austria, supported by Germany, for trying to prevent Serbia from serving as a base for terrorist activities directed against the Habsburg Monarchy? Or Russia, supported by the Western powers, for trying to defend Serbian sovereignty and maintain its own political position in the Balkans? And if all the major powers had been armed with nuclear weapons at the time, is it clear who exactly would have been deterred? Or take the case of the coming of World War II in 1939. If both Britain and Germany had been nuclear powers at the time, again, is it clear who would have been deterred? Waltz thinks that Germany would have backed off: Hitler would not have "started" a war that would destroy the Third Reich. But Hitler did not intend to "start" a war with Britain at that point; his aim was to get Britain to back down in the confrontation over Poland. Nor did Britain intend to start a war with Germany. War broke out not because either side wanted war in late 1939, but rather because neither was willing to give way-and because each was hoping that the other would.

Once we get away from the idea that wars are simply "started" by one side and that the "attacker" can be readily identified, the whole problem appears in an entirely different light. If war is seen as the outcome of a process in which two sides interact, it makes no sense to focus simply on the calculations of just one side. Instead, the calculations of both sides, and especially their calculations about each other, have to be taken into account. Each side may be trying to deter the other-to get its way without war if it can. Each side might be afraid of escalation, but those fears are balanced by the knowledge that one's adversary is also afraid, and his fears can be exploited. In the case of a conflict between two nuclear powers, if either side believed that Waltz's analysis was correct-if either side believed that its adversary would give way rather than run any risk of nuclear attack, as long as his vital interests were not threatened-there would be no reason for that country not to take advantage of that situation. That side could threaten its adversary with nuclear attack if its demands were not met in the firm belief that its opponent was bound to give way, and that it would therefore not be running any risk itself. That belief might turn out to be correct, but if it were not-if its rival was unwilling to allow it to score such an easy victory-there could be very serious trouble indeed. And if both sides were convinced by Waltz's arguments, and both adopted strong deterrent strategies, the situation would be particularly dangerous. Each side would dig in its heels, convinced that when confronted with the risk of nuclear war, the other side would ultimately back down. Such a situation could quickly get out of hand. As Dean Rusk pointed out in 1961, "one of the quickest ways to have a nuclear war is to have the two sides persuaded that neither will fight."

This is an extreme case, but it illustrates the problem. In the real world, states will not be so sure that their opponent "will be deterred" by the prospect of nuclear war and that they can therefore go as far as they like in a political dispute-say, in the Cold War case, in a dispute over Cuba or Berlin. Nor will they themselves, in all probability, be absolutely deterred by the threat of nuclear war. They would be under a certain competitive pressure to play the same game as their rivals; their rivals could not be allowed to profit so easily from a simple threat-making strategy while themselves running no real risk at all. Each side would be afraid of escalation, but each side would in the final analysis also be willing to run a certain risk. Each side would know that its adversary was also worried about what would happen if things got out of hand, and that an unwillingness to run any risk at all would remove that element of restraint and give the adversary too free a hand. Each side would know that its adversary was probably also willing to run a certain risk for the same reason, which is why each side could not be sure that its opponent would be deterred in a confrontation.

In such situations, it is impossible to say how all these calculations would sort themselves out. Deterrence cuts in more than one way, and it for this reason that in a nuclear world, no one can know how far things will go before a conflict is resolved, or whether it even can be resolved before nuclear weapons are actually used. Each side may calculate that if it is just a bit tougher, its opponent may back down. Having gone so far, wouldn't it make sense to go further still? And there is no natural end-point to that process. For Waltz, if deterrence fails, "a few judiciously delivered warheads are likely to produce sobriety in the leaders of all of the countries involved and thus bring rapid deescalation." But it is just as likely that if a few bombs are exploded, the country that had been targeted would choose to retaliate in kind. It might even choose to escalate the conflict. A political dispute can thus become a gigantic poker game, with each side raising the stakes in the hope that its opponent, frightened by the prospect of nuclear war, will fold before things go too far.

Conflicts in such a world, as Thomas Schelling argued years ago, would become "contests in risk-taking." The side with the greater resolve, the side more willing to run the risk of nuclear war, has the upper hand and will prevail in a showdown. In the pre-nuclear world, more or less objective factors-above all, the balance of military power-played a key role in determining how political conflicts ran their course. The weak tended to give way to the strong; in an admittedly rough and imperfect way, the military balance gave some indication as to how a dispute would be worked out. But in a world of invulnerable nuclear forces, as Waltz points out, the military balance counts for little. Subjective factors, like will and resolve, would play the key role in determining how political conflicts are worked out.

The result is that in such a world there would be a great premium on resolve, on risk-taking, and perhaps ultimately on recklessness. In international politics, as in other areas of life, what you reward is what you get. Resolve would tend to harden, and the parties involved would tend to dig in their heels. A reputation for toughness would be of fundamental importance, since one has to worry not just about the present but about the future, and this would provide further incentive to take a tough stand. And as each side hardens its position, its rival is also led by competitive pressure to do the same. Why would anyone think that a world of that sort, where political outcomes are up for grabs and victory goes to the side with the strongest nerves, would be particularly stable?

There is one great objection to this whole line of worried argument, and it has to do with the historical record. Doesn't history show that nuclear weapons are a force for stability? Isn't that what the history of the Cold War tells us? Waltz certainly thinks so. "Never in modern history", he says, "have great powers enjoyed a longer period of peace than we have known since the Second World War. One can scarcely believe that the presence of nuclear weapons does not greatly help to explain this happy condition."

Waltz makes a number of specific claims that can be tested against the historical record. He says that in a nuclearized system, leaders of major powers are extremely reluctant to run risks; they are easily deterred, and will avoid taking any action that might lead to even a few bombs falling on their homelands. It was for that reason, he thinks, that an attack on Cuba during the missile crisis had to be ruled out: even if an air strike could wipe out 90 percent of the Soviet missiles in Cuba, "retaliation by the remaining 10 percent was unacceptable to us." The fear of retaliation from the surviving Soviet missiles in Cuba was certainly a factor making for restraint in 1962, but it is important to note that the deterrent effect was far from absolute. The Americans accepted a showdown with the Soviets over the missiles even though they knew at the outset, as Rusk put it at the time, that a crisis "could well lead to general war." And if a settlement had not been reached, the United States probably would have invaded Cuba at the end of October, even though the U.S. government believed that some of the Soviet missiles on the island would at that time be operational.

Waltz also argues that the mere possession of nuclear weapons makes for a cautious policy; as noted above, he thinks in particular that a nuclear Iraq would be more cautious than a non-nuclear Iraq. Again, the general claim here is testable, and the most important counter-example has to do with Soviet policy in the period after the USSR first began to develop a nuclear force. After the Soviets broke the American nuclear monopoly in late 1949, their foreign policy quickly hardened, and the winter of 1950-51 was one of the darkest periods of the Cold War. There is no doubt that these military and political developments were connected. In the nuclear age as in the pre-nuclear past, military developments cast a deep political shadow. Countries tend to avoid war when the military balance is unfavorable, but when their strategic position improves, they often feel free to take a tougher line.

A final point has to do with what Waltz says about the risk of nuclear escalation. He does not take all the talk during the Cold War about going nuclear in a crisis-about "massive retaliation" and so on-very seriously. "Did our own big words", he asks, "or the Soviet Union's prattling about nuclear war-fighting, ever mean anything?" He thinks not. If the Soviets had attacked Western Europe, the NATO response would have been limited: "The threat, if it failed to deter, would have been followed not by spasms of violence but by punishment administered in ways that conveyed threats of more to come." No evidence is given to support this view; it seems to be based essentially on Waltz's intuition about how states in a heavily nuclearized environment would have had to behave.

My own view is very different. If a war had broken out in the NATO area during the 1950s, or even as late as 1962, there was a good chance, I think, that it would have escalated to the level of a full-scale nuclear exchange. President Eisenhower's whole philosophy was that in the event of war in Europe, the United States would have to open up with everything it had. The U.S. government, he said, would just be fooling itself if it thought it "could fight such a war without recourse to nuclear weapons." And once a nuclear exchange began, the United States "could not stop" until it had "finished off the enemy." There might be no winners in such a war, he said, but "we just don't want to lose worse than we have to." In such circumstances, restraint was simply out of the question: "The United States would be applying a force so terrible that one simply could not be meticulous as to the methods by which the force was brought to bear."

Kennedy also seemed to take it for granted that if a war broke out in Europe, America would "have no choice but to use nuclear weapons." For him, as for Eisenhower, nuclear escalation was by no means ruled out. As long as the United States had the upper hand in strategic terms-as long as America, if it struck fast enough and hard enough, would not suffer really massive destruction from a Soviet retaliatory strike-a major nuclear attack on the USSR, in the event, for example, of a showdown over Berlin, might be the least bad U.S. option.

Those Kennedy and Eisenhower remarks, made in private meetings with top officials, are not to be dismissed as mere talk or as speaking for effect. Serious people really felt at the time that war was not entirely out of the question, since the United States, if push came to shove, was not going to capitulate on the core political issues despite the threat of nuclear war. And if a war did break out, it was widely assumed in high political and military circles that the country might have to do what it could to limit damage to itself-or, as Eisenhower's military chiefs put it in a very important document, to "hold damage to nationally manageable proportions." For Waltz, it is almost inconceivable that responsible leaders would ever be willing to accept a degree of nuclear devastation. But no matter what we think today, the attitude reflected in those Eisenhower and Kennedy comments illustrate what a responsible state might actually do if its leaders felt that fundamental political interests were at stake.

If all this is true, however, how were we able to avoid a nuclear exchange during the Cold War period? The Soviets were able to build a small force soon after they broke the American nuclear monopoly in 1949. If that small force had been able to balance America's much larger force even during the Eisenhower period, then political disputes at that time would indeed have been "contests in risk-taking"; the side with the stronger will, the side willing to accept the greatest level of risk, would have prevailed in the end. Since there is no way to know how far such a contest would go before one side or the other gave way, a world of that sort, with each side under a certain pressure to dig in its heels, would have been a very dangerous place. But under Eisenhower, and for most of the Kennedy period as well, U.S. leaders felt they still had the upper hand in strategic terms: if war broke out, the United States might well not suffer really heavy damage, while the Soviet Union would be utterly destroyed as a functioning society. In such circumstances, the Americans felt they could in the final analysis take a strong line in a crisis, and that the Soviets would be under great pressure to give way in the end. American leaders believed that the strategic balance played a central role in determining how political conflicts would run their course. The USSR could not engage in a "contest in risk-taking" with America on equal terms. The U.S. blockade of Cuba during the missile crisis, for example, could not be answered by a Soviet counter-blockade of Berlin, and so the United States was essentially able to get its way.

By late 1963, as Kennedy was told in September of that year, clear American nuclear superiority finally had become a thing of the past. The post-1963 system of great power political relations turned out to be relatively stable, but not because the USSR had become militarily stronger and could now stand up to the United States as an equal. The loss of clear American nuclear superiority, all else equal, could be expected to have led to a tougher Soviet policy, and thus to a more unstable international system. If that system turned out to be stable, it was because all other things were not equal. It was because a set of political understandings had been struck by late 1963 that allowed a relatively stable system to take shape. If those understandings had not been worked out, the world might well have been in for serious trouble. A Berlin crisis in a period of nuclear parity could have been a very dangerous affair.

It is a mistake to think of nuclearization as a kind of magic wand bringing stability to what would otherwise be an unstable international system. That view is not supported by a close analysis of the Cold War historical record, which suggests that nuclearization might well be a source of instability in international affairs. This is important, because the way we think about these issues can have a real political effect. The United States, of course, is going to oppose the spread of nuclear weapons in any event, for the simple reason (as Rusk pointed out in 1963) that "it was almost in the nature of nuclear weapons that if someone had them, he did not want others to have them." Nuclearization does have a certain equalizing effect and, as other countries build nuclear forces, the U.S. power advantage is necessarily diluted. But the seriousness and intensity with which the U.S. government pursues a non-proliferation policy depends on how the problem is understood. It makes a difference whether America's policy is rooted simply in its own narrow national interests, or whether it is based as well on a concern for the interests of the international system as a whole.

The same point applies to states considering whether to build a nuclear force. A nuclear capability would certainly be a source of strength for most; it would enable them, to a certain extent, to stand up to even the greatest powers on the planet. In regional terms, the weaker of two rival powers-Pakistan, for example, in its conflict with India-is bound to favor nuclearization as a way of leveling the playing field. Once one such power goes nuclear, moreover, the rival power has good reason to follow suit. Perhaps the spread of nuclear weapons is inevitable for those reasons. But if a world of many nuclear powers is to be avoided, it will only be avoided because those incentives are offset by other factors-by the cost of building a secure and survivable nuclear force, and even more by a sense for how dangerous a nuclear world would be.

Perhaps a world of many nuclear powers will turn out to be unavoidable. But there is no point to making it more unavoidable than it would otherwise be by assuming that a fully nuclearized world would have a near rock-solid stability, and that nuclear proliferation is the path to peace. It won't, and it isn't.