The Doomsday Machine

by Daniel Ellsberg

  Why do all options provide for total expenditure of the force? Why is no provision made for a strategic reserve?

  What is the ability of the JCS to accept the surrender of the enemy forces during the execution of the SIOP? What preparations have been made for this event? Have acceptable surrender terms been drafted? How reliable is the capability to STOP remaining attacks after Execute orders have been transmitted? Have preparations been made to monitor conformity to surrender terms?

  Some of my questions couldn’t possibly have occurred to anyone simply from reading the briefing. I threw them in to warn the recipients that someone working for Gilpatric was already familiar with the problems of the operational planning:

  Does planned coordination assume that all offensive vehicles will get the Execute message simultaneously? If so, what is the estimated validity of this assumption? What will effects on coordination be of estimated lags in receipt of message? Or of different wind direction and strength affecting different parts of the attacking forces, in planning to avert interference?

  Since these questions were supposedly coming from Gilpatric, who hadn’t been given the JSCP, I had to find a way to draft them so that they would purport to be based only on the briefing paper that the JCS had given to him. But anyone who knew the actual plans would know that the person writing those questions was not Gilpatric. It had to be someone who was intimately familiar with the JSCP itself and all the controversies that lay behind it, who probably had a copy of it sitting in front of him. In other words, the Joint Staff and their bosses, the JCS, would know immediately that a copy of the JSCP had finally found its way to the Office of the Secretary of Defense. Moreover, they would know that someone who was either a military planner himself (a mole) or who had been very well educated by such planners was advising the deputy secretary.

  The questions were the message. They were intended to leak into the JCS the news that their processes, their conflicts, compromises, and maneuvers, had become transparent to the Office of the Secretary of Defense. I hoped they would figure the game was up; they had to come clean and give straight answers. They had to fear that any efforts to lie or evade would be quickly spotted by whoever wrote these questions. (He might actually know, through some direct channel, their inner discussions of how to deal with the problem of responding to Gilpatric.)

  Each question, still more the whole set, was designed to convey that someone working for Gilpatric knew, as military staffers used to say, “where all the bodies were buried.” It would be clear not only that “he knows the JSCP, and he’s got a copy,” but also that he knew, somehow, why it was written the way it was, where the controversies were, how they had been papered over, and all the other things that it would be hard or impossible for the JCS to explain or justify.

  I don’t have the final memo that I submitted to Gilpatric, in which the questions were more elegantly phrased than above. To keep up the pretense (however transparent to the recipients) that the questions were based on the briefing to Gilpatric rather than on the JSCP itself, most of my thirty or so queries started with a reference to a statement in the briefing paper and then presented a list of sub-questions purporting to relate to it. I happen to recall verbatim the wording of the first question:

  You say, on page 1, that each operational plan is submitted for review and approval to the next higher level of command.

  a) Was the JSCP submitted to Secretary of Defense Gates for his review and approval?

  b) When in the annual planning cycle, is it customary to submit the JSCP to the Secretary of Defense for his review and approval?

  True answers would have been (a) No; and (b) Never. It was obvious that the drafter of the questions knew that. It was not obvious what a satisfactory explanation of those answers would be. Or those for the rest of the questions, which tended to get tougher from there.

  When I handed the list to Gilpatric, he glanced through it, nodded his head, and said appreciatively, “These are very … penetrating questions.” He read it over more carefully, shook his head several times, thanked me warmly, and sent it off to the Joint Staff with a cover letter and without any changes.

  There was no good way for the JCS or its staff to respond to these questions. If they lied or evaded, it was clear they would be found out. But if they answered truthfully, it would have seemed appropriate to send at the same time their letters of resignation. Bob Komer, McGeorge Bundy’s deputy at the NSC, put that more strongly. After he read the draft I showed him in his office next door to the White House, he said to me, “If these were Japanese generals, they would have to commit suicide after reading these questions.”

  The generals and admirals who got the questions were not Japanese. They did not commit suicide, but they did get the message. Within hours of the questions being sent, the director of the Joint Staff was on the phone to Harry Rowen. As Harry told me, he asked very intensely: “Do you know anything about a set of questions we just received from Gilpatric?”

  Harry said, “I might.”

  There was a long pause. Then a curt “Who wrote them?”

  Harry declined to say. End of conversation.

  In a season when military staffs were working night and day to meet without fail or delay the secretary’s short deadlines on numerous studies, this was the one set of questions that was simply never answered at all. As the first deadline approached, the director asked for an extension, and when time ran out on that, he asked for another, then a third. When I asked Gilpatric about it, in a later meeting, they still had not made a formal reply. They never did.

  “That’s perfect,” Gilpatric told me. “We’ll just leave them hanging there. Then if they fight us on the new plans, we’ll just say, ‘Well, then, let’s go back to a discussion of your old plans.’ And we’ll start with those questions again.”

  Meanwhile, my revised guidelines on basic national security policy were signed by the secretary of defense, sent to the JCS as Secretary of Defense Guidance on War Planning, and eventually became the new policy. (President Kennedy had decided not to issue a new BNSP in his own name.)

  * * *

  As it turned out, one of the questions I had drafted for Gilpatric got a different treatment. As part of the list he sent to the Joint Staff, it got no more response than the rest. But it was picked out of my list by Bob Komer at the White House and sent to the JCS as a presidential query. And this question, to my surprise, got a quick, specific, and apparently accurate answer.

  As recounted in the prologue, this question was the following: “If existing general war plans were carried out as planned, how many people would be killed in the Soviet Union and China alone?”

  In posing that question originally, my tentative understanding from Lukeman and his Air Staff colleagues was that the JCS had never calculated an up-to-date answer to it for the current operational plans, which called for the quick and reliable destruction of a target system that included every major city in the Soviet Union and China. That might seem a peculiar supposition, but I had a basis for it. Despite my knowledge of the war-planning process and the plans themselves, which was extensive and virtually unique for a civilian, I had never seen such an estimate. Colonels Lukeman, Cragg, and others had told me they had never seen one either, and they believed it did not exist. And it was easy for someone familiar with the military bureaucracy to imagine bureaucratic considerations that would have blocked it from ever being investigated, having to do with a fear of leaks to the public, but also with the use that internal military critics of the plans could make of realistically horrific figures.

  So I thought that the JCS would probably have to admit that they didn’t know. Or they would have to ask for more time to calculate an answer. Either response would put them off balance in defending their current plans against our proposed alternatives. “What, you don’t even know the consequences of your own plans for human fatalities?” It was to make that as embarrassing as possible that I drafted the question to cover the Soviet Union and Chi
na alone, so that they couldn’t pretend they needed extra time merely to calculate answers for fatalities in Albania.

  I thought it was also possible that they would turn out a hasty answer, which could probably be shown to be absurdly low. The only estimates I had ever seen in war plans had that character. They were from the early 1950s and ranged from about one million dead in the Soviet Union at the beginning of the decade to up to ten million or fifteen million dead in plans a few years later. I had read those as ridiculously low even for the era of A-bombs (which were already, by that time, much larger than the Hiroshima and Nagasaki weapons). If they came back with any estimate at all, I expected that it would be comparably unrealistic in the era of thermonuclear weapons, H-bombs. New underestimates would serve the same purposes in the inner bureaucratic bargaining over the plans as no estimates at all. The possibility that the JCS would come up quickly with a realistic estimate was one I barely considered.

  I was mistaken. So were the usually knowledgeable colonels I had consulted. Not only did some section of the Joint Staff have a plausible computer model for calculating such effects, but they supplied the White House with an answer within a day or two. As I’ve described earlier, it was classified “Top Secret—For the President’s Eyes Only,” but since I had drafted the question, Komer called me over to the NSC offices to look at it.

  The answer was in the form of the graph depicted in the prologue (here) that showed 275 million would die in the first few hours of our attacks and 325 million would be dead within six months. (I had only asked for fatalities, not for casualties, which would have included wounded and sick.) While this was for the Soviet Union and China alone, the speed of their response suggested that they had an existing computer model and probably had estimates on hand for other areas as well. So it proved. I drafted for Komer a follow-up question covering areas contiguous to the Sino-Soviet bloc, and the Joint Staff provided comprehensive estimates with equal dispatch. These were listed in a table.

  Another hundred million or so would die in the Eastern European satellite countries from the attacks contemplated in our war plans, many of which were on air defenses and military installations in those countries, most of them near cities (even though Eastern Europe cities were not targeted as such). To open “air corridors” for subsequent bombers advancing toward the Soviet Union through Warsaw Pact territories, the first wave of bombers would “bomb as they go,” dropping megaton weapons on radar stations, antiaircraft installations, and surface-to-air missile sites as they came to them in Eastern Europe. (Recall General Power’s remark, at the SIOP briefing attended by John Rubel, on the unhappy fate of Albania, site of a radar station en route to Russia.) Although population destruction was not regarded as a “bonus” in the “captive nations”—as it was in the Soviet Union and China, where it was deliberately maximized—most warheads in Eastern Europe, as elsewhere, were ground-burst, maximizing fallout.

  Fallout from our surface explosions in the Soviet Union, its satellites, and China would decimate the populations in the Sino-Soviet bloc as well as in all the neutral nations bordering these countries—Finland, Sweden, Austria, and Afghanistan, for example—as well as Japan and Pakistan. Given prevailing wind patterns, the Finns would be virtually exterminated by the fallout from surface bursts on Soviet submarine pens near their borders. These fatalities from U.S. attacks, up to another hundred million, would occur without a single U.S. warhead landing on the territories of these countries outside the NATO and Warsaw Pacts.

  Fallout fatalities inside our Western European NATO allies from U.S. attacks against the Warsaw Pact would depend on climate and wind conditions. As a general testifying before Congress put it, these could be up to a hundred million European allied deaths from our attacks, “depending on which way the wind blows.”

  As I had intended, the JCS had clearly interpreted the phrase “if your plans were implemented as planned,” to mean “if U.S. strategic forces struck first, and executed their planned missions without disruption from a Soviet preemptive strike.” These figures clearly presumed that all or most U.S. forces had gotten off the ground with their weapons without having been attacked first. That is, it was implicit in these calculations—as in the greater part of our planning—that the United States would be initiating all-out nuclear war: either as escalation of a limited regional conflict that had come to involve Soviet troops or in preemption of a Soviet nuclear attack of which we had tactical warning. (The warning, it was understood, could be a false alarm. Or if it were not, the Soviet attack under way might be in response to a Soviet false alarm of a U.S. attack.)

  The phrase “implemented as planned” referred to the assumption on which nearly all our planning was based: that in the whole range of circumstances in which nuclear war was likely to occur, we would “take the initiative.” Before enemy warheads had arrived or, perhaps, had been directed to launch, we would be striking first.

  Thus, the fatalities the JCS were reporting to the White House were the estimated results of a U.S. first strike. The total death count from our own attacks, in the estimates supplied by the Joint Staff, was in the neighborhood of six hundred million dead, almost entirely civilians. The greater part inflicted in a day or two, the rest over six months.

  And these were solely the effects of U.S. warheads, not including any effects from Soviet retaliatory attacks on the United States or U.S. and Allied forces in Europe or elsewhere. The CIA intelligence estimate in June 1961 credited the Soviets with well over a hundred ICBMs at that time, of which they claimed to be able to locate, and thus to target, only a small fraction. Estimates of U.S. fatalities from Soviet retaliation consistent with those estimates would have added scores of millions of U.S. dead to the total, even after a very effective U.S. first strike.

  Army and Navy estimates of Soviet ICBMs threatening America were “a few.” But by all estimates, several hundred intermediate and medium-range missiles were aimed at Western Europe, Germany in particular, along with hundreds of medium-range bombers. Even after the most successful U.S. and NATO first strike, Soviet retaliation against Europe could have added a hundred million to the Western European death count from blast, fire, and immediate exposure to radiation even before the fallout from our own attacks had arrived from the east, borne on the wind.

  Holding the graph in my hand—the answer to my initial query, covering only fatalities from the Soviet Union and China—looking at it in an office of the White House Annex on a spring day in 1961, I realized, “So they knew.” As I said in the prologue, the graph seemed to me the depiction of pure evil. It should not exist; there should be nothing real on earth that it referred to.

  But what it dealt with was all too real. I had seen some of the smaller bombs myself, H-bombs with an explosive yield of 1.1 megatons each—equivalent to 1.1 million tons of high explosive, each bomb half the total explosive power of all the bombs of World War II combined. I saw them slung under single-pilot F-100 fighter-bombers on alert at Kadena Air Base on Okinawa, ready to take off on ten minutes’ notice. On one occasion I had laid my hand on one of these, not yet loaded on a plane. On a cool day, the smooth metallic surface of the bomb was warm from the radiation within: a bodylike warmth.

  Three thousand warheads like these—most of them much larger in yield, up to twenty times as great—would be delivered on the Soviet bloc and China in the first stage of execution of the SIOP. Most of them, I knew, would be ground-burst, with fallout that would annihilate the population not only of the Sino-Soviet bloc but of its neighbors, including allies and neutrals.

  It was not only the size of the numerical estimate of deaths that threw me into a state of shock, though I was not at all used to seeing estimates like that in classified studies. At RAND I’d read almost exclusively estimates of the population damage the United States could reliably threaten to produce in a retaliatory second strike, for purposes of deterring a Soviet first strike. In the context of Albert Wohlstetter’s and RAND’s concern that a well-designed Soviet
Pearl Harbor–like attack might totally eliminate SAC’s currently deployed retaliatory forces, the focus in those studies was how to assure that Soviet casualties from U.S. retaliation could be as high as, say, the Soviets had suffered in World War II: in the neighborhood of twenty million. Nothing less, Wohlstetter argued, and his acolytes (like me) accepted, would reliably deter cold-blooded Bolshevik leaders in an intense crisis. I’m not sure that I’d ever seen in a RAND study an estimate of the human consequences of a U.S. first strike by an undamaged force, a possibility that was scarcely considered at RAND except by Herman Kahn.

  But since I’d seen actual war plans in the Pacific and the Pentagon, I’d become aware that they focused preeminently on just such a case: a U.S. first strike, either preemptively or in escalation from a regional conflict. And not, in either case, what RAND analysts would have considered a well-designed first strike—that exclusively focused on Soviet military targets—but one that explicitly attacked all Soviet (and Chinese) cities in the early waves. So I’d long been aware that the destruction wrought by executing such a plan (unimagined by RAND, for the United States) would be “huge,” “horrible,” beyond the RAND calculations I’d seen, but I’d never formulated a specific measure of it in my own mind, and I’d never seen one. This one seemed realistic.

  To see it in print was startling, despite the fact that I had long privately thought, while reading war plans during the previous two years, that I was looking at the way the civilized world might end. These were plans for destroying the world of cities, plans that someday might be carried out. But I had thought that none of the others reading or writing them had faced up to that.

  The shock, for me, was to realize that the Joint Chiefs were, after all, aware of it. Their planning process was not so mindless of overall consequences as I had come to suppose. It was worse. What was beyond surprising—it was unfathomable—was that they felt they could afford to be so candid about this particular answer, so prompt, responsive, realistic, while they stalled on all the others.