To the editors:

In his review, Jean-Pierre Dupuy praises Daniel Ellsberg’s memoir The Doomsday Machine. Dupuy approves and develops the ideas presented by the author about the risks associated with nuclear weapons and nuclear deterrence.

Ellsberg is the famed whistle-blower who leaked the Pentagon Papers while working at the RAND Corporation in Santa Monica, California. He is also a seasoned nuclear analyst. Dupuy lives in California, is well-known for his scholarship, and is one of the few Western writers who has delved into the philosophical intricacies and moral dilemmas of nuclear deterrence. Dupuy and Ellsberg have both made invaluable contributions to the field.

I have never met Ellsberg, but during a one-year stay at RAND in the 1990s, I regularly dined at Chez Jay, the famed restaurant where Ellsberg passed the Pentagon Papers to a New York Times reporter. At that time, some of the towering figures of the US nuclear strategic community, such as Albert Wohlstetter, still had offices at RAND. But my stay convinced me that, by that time, RAND’s experts often had an incomplete and distorted view of what nuclear strategy had become.

My argument is that, for all of Dupuy’s ethical insights and Ellsberg’s historical and technical knowledge, both men mischaracterize past and current nuclear policy, and end up overstating the risks of nuclear war. In short, their writings reflect an exceedingly theoretical, partly obsolete, overly pessimistic, and ultimately incomplete view of nuclear deterrence.

An Exceedingly Theoretical View

Dupuy and Ellsberg make statements that reveal confusion on five points: the role of irrationality in nuclear strategy, the concept of mutually assured destruction (MAD) and its implementation, the role of preemptive nuclear strikes in US strategy, the notion of a first strike, and the existence of so-called doomsday machines. Their views are by and large divorced from theory, history, and reality.

Irrationality Is Not Integral to Deterrence

Dupuy claims that deterrence requires irrationality: in order to deter effectively, one must appear crazy in the eyes of the adversary. As Dupuy writes, “This disregard for rationality is … fundamental to the internal logic of deterrence theory. … The madman theory is … integral to the MAD doctrine.”

Dupuy claims that this point was made early on by one of the main theorists of deterrence, Thomas Schelling. But this is not what Schelling thought—nor what the United States embraced. The idea of “embracing non-rationality and simply giving the impression that U.S. leadership was crazy” was indeed discussed by Schelling and analysts at RAND, among other places.¹ But, as noted by Austin Long, a historian of RAND, Schelling, “along with most other analysts, found this approach unpromising.”² Schelling did note that “it does not always help to be, or to be believed to be, fully rational, cool-headed, and in control of one’s country.”³ But what he prescribed was ambiguity, not irrationality. He believed that deterrence is enhanced by not being entirely clear in declaratory threats, emphasizing the unpredictable nature of crisis management and escalation, and “leaving something to chance.”⁴

Ellsberg and Dupuy see US Presidents Richard Nixon and Donald Trump as applying the irrationality principle. Nixon wanted to appear crazy to terrify US adversaries, in particular the North Vietnamese.⁵ But this was a personal and commonsensical choice, not the implementation of a fundamental principle of deterrence. The same can be said about Trump.

MAD Is Only One Version of Deterrence

Dupuy states that “the most basic form of deterrence theory is known as MAD.” He argues that to implement MAD, “it is essential that neither protagonist tries to protect against itself attacks from their opponent.”

But MAD is actually an elaborate and demanding, rather than basic, version of deterrence. MAD suggests that deterrence is stable only when both opponents possess an assured destruction capability—that is, an ability to inflict unacceptable damage on the other side, even after a first strike.

There are, in fact, several versions of MAD. As discussed in the 1960s, MAD involved a retaliation-only posture, counter-cities targeting, and the acceptance of territorial vulnerability. But neither the US nor the Soviet Union implemented MAD in this form. Both sides retained the possibility of first use of nuclear weapons on the battlefield and the option of a counterforce strike, and they both believed that limited missile defenses could be useful. According to Charles Fairbanks: “Through the 1960s, the Defense Department and successive presidential administrations allowed mutually assured destruction (MAD) to be perceived as strategic doctrine. However, MAD never became, in practice, America’s strategic doctrine.”⁶ Post–Cold War interviews revealed that the Soviet Union, for its part, never examined the question of mutually assured destruction as a condition that they should accept, much less pursue. The Soviet Union never embraced vulnerability as desirable. The Soviets also believed that, given the military uncertainties, mutually assured destruction was only a theoretical conclusion.⁷

In official US doctrine, the concept of “assured destruction” was utilized at the Department of Defense in 1963 and 1964 merely to size its forces. Sizing the force to be able to completely disarm the adversary turned out to be too expensive in budgetary terms and potentially too costly in human terms.⁸ The US plans ended up targeting a mix of industrial, military, and nuclear installations.

The reality of MAD was a degraded version of the theoretical construct of the early 1960s. By the 1970s, both the United States and the Soviet Union had, among other options, the capability to inflict unacceptable damage to the other in most probable circumstances. Ellsberg is correct to claim that MAD has never been a US objective. He argues that the only possible alternative to fully embracing MAD was preemption. But, in reality, modern strategic nuclear planning is not limited to a single course of action. Instead, there is an array of options to be considered: leadership must decide if they will strike first or second, respond in kind or escalate, and target adverse forces, traditional military targets, command centers, or war-supporting industries, among others.⁹

Preemption Is Not about Disarming

Due to the traumas of the Second World War, US and Soviet strategic thinking during the Cold War was focused on finding the best way to respond to a surprise attack in the case that deterrence failed. Broadly speaking, there were three ways to manage this scenario: 1) protecting civilian and military targets with air and missile defense and shelters; 2) guaranteeing the ability to retaliate by dispersing forces, putting bombers in the air, hardening silos, or deploying submarines; and 3) seeking to actively reduce the threat by targeting enemy forces in wartime.

Ellsberg and Dupuy suggest that preemption—striking first as soon as it becomes clear that the enemy is about to attack—was about disarming the adversary and was a preferred strategy of the US.¹⁰ But once the Soviet Union possessed nuclear forces in significant numbers, preemption in the US was aimed not to fully disarm through force, but to reduce threat. It was more appropriately labeled “damage limiting counterforce.”¹¹

Most importantly, there is no evidence that the US favored a strategy based on preemption after the 1960s. The inherent risks of preemption given the impossibility of ensuring the complete destruction of adverse nuclear forces were recognized as early as the Cuban Missile Crisis. In 1969, three out of five strategic nuclear strike options were preemptive.¹² But the situation changed afterward, as made clear by the declassified record.

Under the Nuclear Weapons Employment Policy that Nixon signed in 1974, the ability to “destroy or neutralize … the nuclear offensive capabilities of the enemy … in order to assist in limiting damage and to reduce the enemy’s forces for nuclear coercion” ranked only third on a list of a desired capabilities.¹³ Forces were to be sized and designed for two different scenarios. The first, and presumably most important, regarded the case of having “alert forces with damage”—that is, striking second. In this case, Soviet forces were the last targeting priority. The second scenario was “generated forces without damage”—that is, striking first. In this case, Soviet forces were only the second targeting priority.¹⁴ Under Presidential Directive 18 signed by Jimmy Carter in 1977, controlled escalation was a priority as “one of few means for limiting damage to [the] US,” given that “US counterforce capability [was] waning” and the US “lack[ed] improved civil defense measures.”¹⁵ Thus it was directed that only “if control of escalation fails, seek to limit damage to US and allies.”¹⁶

In later documents, preemption ranked even lower in US preferences. Carter’s Presidential Directive 59 of 1980 stated that while it will remain our policy not to rely on launching nuclear weapons on warning that an attack has begun, appropriate pre-planning, especially for ICBMs that are vulnerable to a preemptive attack, will be undertaken to provide the option of so launching.¹⁷

In 1981, Ronald Reagan’s National Security Decision Directive 13 directed that “we must have the capability to attack the widest range of targets … even when retaliating to a massive strike received without strategic warning.”¹⁸ It also directed that, although “we must be prepared to launch our recallable bombers upon warning that a Soviet nuclear attack has been initiated,” “it will remain our policy not to rely on launching our nuclear forces in an irrevocable manner upon warning that a Soviet missile attack has begun.”¹⁹ The directive clearly emphasized that the bombers should be recallable.

Likewise, by the 1970s the Soviet Union had also moved away from a preemption strategy to one based on launch under attack and retaliation. The Soviet General Staff continued to maintain all options: preemption, launch on warning, and retaliation. But according to post–Cold War interviews of high-ranking officials, “The Soviet political leadership never discussed the possibility of launching a first strike.”²⁰ In fact, by the mid-1980s, first strike options had disappeared from Moscow’s strategic planning.²¹

A First Use Is Not Necessarily a First Strike

There is a difference between first use, which is the initial use of nuclear weapons by one of the parties during a war, and first strike, which is the use of nuclear weapons against adverse strategic forces. Ellsberg and Dupuy are aware of this difference, but they sometimes conflate the two terms. Dupuy is correct to affirm with Ellsberg that “there has not been a single US president … who has pledged not to use nuclear weapons first.”²² But there are many ways to use nuclear weapons first; a first strike would be an extreme option among others, the most likely being deployment in the European theater and not necessarily against strategic forces. Dupuy is wrong when he claims that current events vindicate Ellsberg’s thesis. When John Bolton argued for striking North Korea first, he referred to conventional, not nuclear, preemption. And Trump’s refusal to “take any cards off the table” in nuclear policy—an expression also used by his predecessors—is hardly confirmation of a preference for nuclear preemption.

True Doomsday Machines Were Never Built

Ellsberg and Dupuy muse about Dr. Strangelove’s concept of a doomsday machine, a system that would ensure reprisals even if a nation’s leadership had been killed. They claim that such structures were actually designed and built, and that, like in the Stanley Kubrick film, neither Washington nor Moscow informed its opponent of their existence.

It is important here to distinguish four concepts that Ellsberg and Dupuy seem to fuse. The first is devolution, the transfer of political authority from the president, or its foreign equivalent, to his or her successor. In the United States, the procedure was public and well known.²³

The second is delegation, authorization given in advance to military commanders to use nuclear weapons in specific circumstances. From 1957 to 1968, delegation to US commanders in chief (CINCs)—and possibly to the next level, as revealed by Ellsberg—existed for when time and circumstances did not allow for a presidential decision or for when communication from the president to the military was impossible. From 1965 until the end of the Cold War, delegation to the CINC of the North American Air Defense Command existed within the continental United States, albeit “under severe restrictions and specific conditions of attack.” The CINC had to repeatedly try to contact civilian authorities and could only use low-yield defensive weapons on US and Canadian territory or waters.²⁴

The third concept is a broader form of last-resort delegation combined with emergency communication systems, sometimes confused with so-called dead hand mechanisms. The Soviet Union deployed Perimeter, a communications system that could be activated if detonations occurred on Russian territory and all contact with political and military authorities had been lost. Only if those two conditions were met, officers could give the launch order via emergency communication rockets. The US, for its part, deployed an Emergency Rocket Communications System from 1963 to 1991, but it could only be activated by the Strategic Air Command.

The fourth concept is a fully automated reprisals system, a true doomsday machine. Ellsberg acknowledges that such a mechanism was not set up in the Soviet Union.²⁵ Nor was it available in the US. “Strangelove and RAND analysts alike,” Long has remarked, “conclude that this method is impractical.”²⁶

A Partly Obsolete View

Ellsberg’s first-hand nuclear policy experience is from the 1960s, a decade during which he was seconded from RAND to the US Department of Defense. Dupuy, for his part, seems sometimes unfamiliar with the evolution of nuclear planning and strategy during the last few decades. There is no evidence that delegation exists in the US today. There is also no evidence that preemption remains the preferred choice, or that “preemptive, launch-on-warning” is still “at the heart of our strategic alert,” as Ellsberg claims.²⁷ In fact, by the 1990s, launch-on-warning had been explicitly demoted to the least favored approach. Robert Bell, director for defense policy under Bill Clinton, explained,

There is no change in this PDD [Presidential Decision Directive] with respect to U.S. policy on launch on warning … We direct our military forces to continue to posture themselves in such a way as to not rely on launch on warning—to be able to absorb a nuclear strike … Our policy is to confirm that we are under nuclear attack with actual detonations before retaliating.²⁸

Prompt launch has remained an option since the 1970s to maximize options given to the US president. The same is probably true in Russia, although clear public evidence is lacking. But historical precedents suggest that it would not be a favored option.

There is no evidence either that the US military now views nuclear weapons as war-fighting tools. “Ellsberg believes,” Dupuy remarks, “that the US military has never regarded the use of nuclear weapons as a taboo … Nuclear weapons are far from the weapons of nonuse suggested by the theory of deterrence.” Making such a bold claim when all available official documents state that US nuclear weapons are for deterrence—though they could be used if deterrence fails—requires extraordinary evidence, of which there is none to be found. The Clinton administration said it removed from presidential guidance

all previous references to being able to wage a nuclear war successfully or to prevail in a nuclear war … The emphasis in this PDD is therefore on deterring nuclear wars or the use of nuclear weapons at any level, not fighting [with] them.²⁹

Under George W. Bush, conventional preemption was emphasized, but nuclear was not. Nothing in Trump’s 2018 Nuclear Posture Review suggests otherwise.

There is no evidence that decapitation, the targeting of political and military leaders, is currently considered by the US or Russia. In their discussion of doomsday machines, Ellsberg and Dupuy refer to the possibility of a decapitation strike: “The best way to ensure the success of a nuclear first strike,” writes Dupuy, “one might assume, is to target the opposing president or leader.” But US leaders have always been very cautious about implementing such a concept, for fear that a nuclear war would become unmanageable. The Nixon Nuclear Weapons Employment Policy explicitly stated that targeting enemy command and control should be avoided.³⁰ In fact, the only time when Washington reportedly embraced such a concept was under the Carter administration. Carter’s now declassified nuclear guidance states only that “pre-planned options, capable of relatively prolonged withhold or of prompt execution, should be provided for attacks on the political control system…”³¹ Decapitation was one option among others. The succeeding directive under Reagan ordered that the plan to attack the Soviet leadership was “designed to be withheld for a protracted period … so that they have a strong incentive to conflict termination.”³² No publicly available document or testimony suggests that decapitation remains an important US or Russian planning goal.

An Exceedingly Pessimistic View

To make his case about the dangers posed by nuclear weapons, Ellsberg expands the definition of a doomsday machine to a system that could be “triggered on a false alarm, a terrorist action, unauthorized launch, or a desperate decision to escalate.”³³ He describes such a machine as being able to “with unknowable but possibly high probability bring about the global destruction of civilization and of nearly all human life on earth.”³⁴

Dupuy nods approvingly: “It is not inconceivable that a false alarm could trigger a nuclear conflict.” But to claim that false alarms have brought the world close to catastrophe is debatable. As I have written elsewhere, a close look at the historical record suggests that the system has worked. Technical, procedural, and human barriers to an accidental or unauthorized use have functioned properly. And luck does not necessarily explain the absence of nuclear use.³⁵ Dupuy mentions the false alert sent to the Hawaiian population in 2018 announcing a North Korean missile launch, but such local warning systems have nothing to do with defense nuclear attack warnings. Of course, it is “not inconceivable” that the doomsday machine, as any human construct, could backfire. The question is whether the probability is small enough that it is worth the potential cost.

Deliberate use of nuclear weapons would be another matter. But here the record speaks for itself. Many US presidents and a few other foreign leaders have toyed with the idea of using nuclear weapons, but none of them has actually come close to pushing the button.³⁶ The only time we may have come close to nuclear war was during the Cuban Missile Crisis, in particular when an exhausted Soviet submarine commander considered firing a nuclear-tipped torpedo against a US ship. This would have been a colossal event, but why would it have meant an inevitable escalation toward the extinction of all life on earth?

The pioneer of game theory John von Neumann believed, along with many other theorists of the Cold War, that deterrence could not exist or last for long and that nuclear war was certain. It has not happened—a fact that Schelling marveled about in his 2005 Nobel Prize acceptance speech. But the best explanation is that the tens of thousands of persons who have had charge, at one level or another, of nuclear weapons since 1945 “must have taken much greater care than is taken in any other situation involving human agents and complex mechanical systems.”³⁷

Drawing on the debates among the Manhattan Project scientists about whether a nuclear reaction could accidentally set fire to the entire atmosphere and end life on earth, Ellsberg and Dupuy suggest that even a vanishingly small probability of such a scenario is excessive. At first glance, this seems commonsensical. But is it really true? Anyone who drives to work every day knows that there is a non-trivial probability that they might be killed in a car accident. Commuters are nevertheless ready to take a total risk—death for themselves and possibly others—over time for a significant benefit on a day-to-day basis. Why would the same kind of calculation applied to nuclear deterrence be wholly unreasonable? This is a legitimate debate, to which philosophers such as Dupuy can make invaluable contributions.

An Incomplete View

I concur with Dupuy’s criticism that Ellsberg’s book entirely disregards any positive role for the policy of deterrence. I also agree that one of the book’s shortcomings is “the absence of a theoretical perspective.” But theory, whether based on philosophical principles or on mathematical devices, only goes so far when we try to predict how humans might behave in a crisis involving the life and death of nations. Especially now that we realize we understood the Soviets much less than we thought we did.

Nuclear deterrence is a risky business and calculating its costs and benefits requires a sobering, facts-based, and dispassionate analysis. By taking a well-meaning and thoughtful, but partly uninformed view, as Dupuy has done, or an experienced, but partly obsolete and militant one, as Ellsberg has, the philosopher and the practitioner are missing the view of the policymaker. A dispassionate review of nuclear history, doctrines, and strategies of the past seven decades does not support their exceedingly pessimistic vision of nuclear risks, which is based on outdated information, conceptual approximations, and unconvincing analysis.

Bruno Tertrais

Jean-Pierre Dupuy replies to Bruno Tertrais, David Omand, M. V. Ramana, Thomas Shea, and Jacek Kugler:

Let me start by saying how grateful I am to my five commentators. Even when I disagree with them, I acknowledge that they bring interesting and at times essential elements to a never-ending discussion.

They had to comment on not a sui generis paper but the review of a book. These are two very different exercises. About a review one can ask questions such as, is the reviewer faithful to the work reviewed, has he understood it correctly, are his criticisms fair and well argued? One can also choose to criticize the work reviewed, independently of what the reviewer said about it, or one can focus on the reviewer’s own take on the problems at hand. The majority, if not all, of my commentators, implicitly or explicitly, have not raised those questions. One of them, Bruno Tertrais, repeats the phrase “Ellsberg and Dupuy” as if we are like two peas in a pod. Only Jacek Kluger has taken issue with me on the fact that I minimized what he calls the “most fundamental flaw” of the book, namely “Ellsberg’s claim about the role nuclear weapons play in preserving peace.”

My commentators should be forgiven since they, aside from Tertrais, do not know me and I did not explain my methodological choices. All of them are specialists of the field broadly defined. I am not. I am a philosopher who has specialized these last 30 years in applied analytical philosophy and metaphysics for moral and political issues. The domains of application have been varied, from the foundations of rational choice and game theory to the ethics of nanobiotechnology. One issue has been the rationality and ethics of nuclear deterrence. I have read the relevant literature, but that does not make me a practitioner or a policymaker.

I am an activist of sorts. Through a thinktank, I militate for the abolition of the French force de frappe, or nuclear strike force, in the same way that Tertrais is militating for its preservation. These differing stances do not prevent us from grounding our actions on arguments and reasons as finely honed as possible.

In discussing Ellsberg’s book, I made a methodological decision. I would not question the facts. I would not even discuss the major arguments. I would simply report them. My added value, if any, would consist solely in putting them in the service of a philosophical inquiry. I am not the first philosopher to do that: David K. Lewis, Robert Stalnaker, Gregory Kavka, Jean Hampton, David Gauthier, Edward McClennen, Michael Bratman, and even Alvin Plantinga,³⁸ to name just a few, are no more practitioners than I am. They became interested in the topic of nuclear deterrence and chose to analyze it because the topic lends itself to a reconsideration of the axioms of the philosophy of time. Or they chose to analyze it because the topic lends itself to a questioning of the foundations of rational choice theory, including game theory, on the basis of its paradoxes. I have inscribed my inquiry in that tradition.

This aim is more apparent in a book that was published in France last February under the title La Guerre qui ne peut pas avoir lieu: Essai de métaphysique nucléaire, or “The War That Cannot Take Place: An Essay in Nuclear Metaphysics.”³⁹ The paper published by Inference reflects some of the ideas developed in that book.

Am I in agreement with Ellsberg? Most of my commentators seem to have assumed so. Tertrais is the most explicit: “Dupuy praises Daniel Ellsberg’s memoir … [He] approves and develops the ideas presented by the author …” This is inaccurate.⁴⁰ Ellsberg is not only the whistleblower we all know; he is above all a subtle critic of rational choice theory. Every student of economics has to become familiar with the paradox he published in 1961, a work that administered a near fatal blow to the axioms put forward by Leonard Savage to formalize rational choice in a situation of radical uncertainty.⁴¹ In The Doomsday Machine, this dimension of the author’s talent is less apparent, although it is not difficult to unearth, as I have endeavored to do. My conclusions are different from his. I am led to take nuclear deterrence more seriously than he does, and, above all, I am much more pessimistic and negative than he is. I have shown that it is indeed possible to preserve theoretically the rationality of nuclear deterrence based on fact-based truth—no nuclear apocalypse has occurred since 1945. Yet one has to pay an exorbitant metaphysical price for that peace—a price so high that the simple possession of a nuclear arsenal appears to be a moral abomination. The rationality of nuclear deterrence and its ethical appraisal are two inseparable issues. I have referred to my method as enlightened doomsaying, a phrase that has been taken up by a number of scholars, in France as well as in the US.⁴²

In the paper that my commentators had to read, almost none of this opinion was apparent.⁴³ My task was simply that of a reviewer. I did not judge it appropriate to profit from the occasion to expound my own theories on the subject matter. This I did in the book already cited. I may then be the only guilty party in the errors of interpretation committed by my critics.

In this light, let me comment briefly on my readers’ reactions.

David Omand’s paper is a remarkable illustration of the mistake in misinterpreting the statement “X said that Y said that P” as “X said that P.” Omand asserts that “Dupuy is mistaken in asserting that a nation has only one goal in seeking to acquire nuclear weapons, namely, to prevent others from using them.” What I wrote at the very beginning of my paper was

Nuclear weapons are widely believed to be weapons not meant for use. In seeking to acquire them, a nation has only one goal: to prevent others from using them. This is the notion at the heart of deterrence theory.

I hope that no one except Omand understands the second sentence as being an expression of what I believe. Both the grammar and the rest of the paper,⁴⁴ which Oman appears not to have read,⁴⁵ exclude such an interpretation. The text says that the many who believe in deterrence theory are thereby led to believe that a nation has only one goal. But it does not say that the author of the review, nor the author of the book, share that belief. It should be obvious that we do not. Ellsberg hardly takes deterrence theory seriously and I have shown that it is a flawed and dangerous concept. We do believe that in acquiring nuclear weapons, a nation pursues goals other than ensuring nuclear peace, as M. V. Ramana has perfectly understood in his cogent contribution.

One finds in Ramana’s paper an outstanding quotation that summarizes beautifully what Ellsberg and I want to say about nuclear deterrence. It is by an eminent practitioner, General Lee Butler, a former commander-in-chief of the US Strategic Command:

The goal—the wish, really—might be to prevent nuclear war, but the operational plan had to be to wage war. … Wish and plan collided at every point—psychologically, intellectually, but, above all, operationally.

Preventing nuclear war thanks to nuclear weapons is the very paradox that lies at the heart of nuclear deterrence. It is a very old paradox if we bring it together with a passage from the Gospel according to Matthew (12:26): “Satan casts out Satan.” Deterrence theory has been unable to unravel it. A French president once remarked that a nuclear arsenal is not maintained for the purpose of liquidating 60 million innocent lives; it is there to ensure that the conditions that would lead us to liquidate 60 million innocent lives will never occur. The intention to make good on one’s promise to retaliate if attacked is a self-stultifying intention, as Gregory Kavka interpreted it—giving his name to the enigma for want of solving it.

I have little to say about the interesting paper by Thomas Shea as it does not address the main concerns that motivated my review. I have already mentioned Kugler’s critique addressed to Ellsberg and incidentally to me: our lack of solution for preserving peace. What I believe I have shown through an a priori approach is this: a necessary condition for deterrence to work is that the decision to retaliate remain indeterminate, in a very specific sense. Retaliating and caving in must not be two alternatives but a superposition of states.

Last but not least comes my fellow French citizen’s contribution. I have already mentioned that Tertrais and I stand at opposite ends of the spectrum of opinions regarding nuclear weapons in general and the French force de frappe in particular. He concludes that

a dispassionate review of nuclear history, doctrines, and strategies of the past seven decades does not support [Ellsberg and Dupuy’s] exceedingly pessimistic vision of nuclear risks, which is based on outdated information, conceptual approximations, and unconvincing analysis.

I find his unforgiving judgment insulting for a great number of eminent policymakers and outstanding scholars. I am thinking principally of my Stanford colleagues gathered at the Center for International Security and Cooperation. Among them are William Perry, former Secretary of Defense for President Bill Clinton, Scott Sagan, David Holloway, Lynn Eden, Barton Bernstein, and Martin Hellman. These people are no less pessimistic than Ellsberg or myself. In his last book, Perry goes so far as to write, “Today, the danger of some sort of a nuclear catastrophe is greater than it was during the Cold War and most people are blissfully unaware of this danger.”⁴⁶ By contrast, Tertrais’s blind optimism appears exceedingly naive or rather partisan. His unexamined trust in the efficiency of nuclear deterrence could lead, if it is shared by many, to a major disaster.

There should be no room for wars of words in these supremely important matters. Only a dispassionate examination of the arguments exchanged can bring some light to the issues at hand. In the present case, there is a problem. The letter by Tertrais is longer than my original review. A point by point response would require an even longer paper. I will focus here solely on two of his major critiques: they pertain to the role of irrationality in nuclear strategy and the notion of a first strike. These are the easiest critiques to respond to, but I believe I could argue likewise about almost all of the others.

The first point in contention is my assertion that irrationality is an integral part of deterrence theory. Tertrais mixes up two different ideas. On the one hand, there is madman theory, whose source is generally ascribed to Nixon, but is claimed by Ellsberg himself.⁴⁷ Tertrais is right to assert that this so-called theory was not “the implementation of a fundamental principle of deterrence” and that it has almost never really been taken seriously. On the other hand, and this is the gist of my assertion, there is the idea that even if someone is convinced that deterrence is a rational plan in a particular situation, its implementation is, as they say in rational choice theory, time-inconsistent. The plan is to threaten the enemy in order to deter them and, if the plan fails, to make good on one’s threat. It would be highly irrational to execute the second part of the plan since it would risk triggering what Carl von Clausewitz called escalation to the extreme—that is, in the nuclear case, mutual annihilation. In the madman theory, the madness is an act. In deterrence theory, it is real. Thomas Schelling was well aware of this distinction. I thank Tertrais for noting that Schelling prescribed ambiguity rather than irrationality.

As for the distinction between first strike and first use, Tertrais takes exception with me about the way I illustrate Ellsberg’s thesis in relation to current events. In particular, John Bolton’s declaration that “it is perfectly legitimate for the United States to respond to the current ‘necessity’ posed by North Korea’s nuclear weapons by striking first.”⁴⁸ I note that this course of action would be prevention and not preemption, contrary to what Bolton claimed, and therefore illegal. Tertrais rejoins: “[Bolton] referred to conventional, not nuclear, preemption.” How does Tertrais know that? Bolton was clever enough to preserve the ambiguity. If he had been interrogated, he would have probably responded, like Trump pushed by Chris Matthews, “I’m not going to use nuclear, but I’m not taking any cards off the table.”

I was writing this when I received a paper by Scott Sagan and his colleagues, published in the Bulletin of the Atomic Scientists. The paper is about the attitude of Americans about the prospect of attacking North Korea first, whether with conventional or nuclear weapons. Here is Sagan’s conclusion, summarized by John Mecklin, editor-in-chief of the Bulletin:

Most Americans do not want the United States to launch a preventive war against North Korea. But a large, hawkish minority—about a third of respondents—approves of a US preventive strike across scenarios—even when US use of nuclear weapons could be expected to kill 1 million North Korean civilians. … Although it is good news that most Americans do not support preemptive war against North Korea, far too many don’t hesitate to approve, even if nuclear weapons are used in the attack [emphases added]. And far too many Americans are mistakenly convinced that such an attack would not result in harm to US citizens.⁴⁹

Bolton was certainly no less hawkish than the “large, hawkish minority” of US citizens who approve of a nuclear first strike against North Korea.

I would like to conclude these notes with an immodest hurrah. Former Secretary of Defense Perry has just publicly endorsed the method I have been advocating to tackle these nuclear issues. Asked by Mecklin why he believes that it is important to publicly disseminate “realistic (and therefore horrifying) scenarios of nuclear war,” he replied that “what we’re really trying to do is find ways of averting doom.” Or as phrased in the interview’s title, “We must describe doomsday to keep it from happening.”⁵⁰ This is the exact formula of my enlightened doomsaying.