No Launch on Warning

by Alan F. Phillips, M.D.

(see also Note on Terminology)

This paper argues for abandoning “Launch on Warning” (L-o-W), as a simple and quick method of greatly reducing the risk of nuclear war, pending the permanent elimination of nuclear weapons which has been promised by all the Nuclear Weapon States that have signed the Non-Proliferation Treaty. Elimination seems likely to take a considerable time, and meanwhile the world is at serious risk from a possible nuclear war, and particularly from a purely accidental war. It is this last risk which could be reduced almost to zero by abandoning L-o-W.

The discussion here is based on the simplifying assumption of a one-against-one antagonism – U.S.A. vs. Russia – with stability based on nuclear deterrence. The assumption is appropriate because L-o-W is only relevant between adversaries with arsenals adequate for them to consider a “disarming first strike”, rather than, say, a surprise attack on cities. Also, the war that it is essential for humanity to prevent at the present time is one between those two countries. Their arsenals are so large that if they should go to war the result would be complete destruction of civilization in the northern hemisphere, nuclear winter, and other sequelae that might combine to exterminate the human species.

Regarding the several other states that have, or may have, nuclear weapons deployed, their arsenals are so much smaller that only the civilizations of the warring states would be destroyed by a nuclear war between them. Collateral damage in other countries could be extensive, but not completely destructive. With their present arsenals, nuclear winter would be threatened only if practically all the weapons were detonated. These states with smaller arsenals are not believed to have the policy of “Launch on Warning”, so that in any case the discussion in this paper does not apply to them.

The concept that stability can be achieved by nuclear deterrence is an assumption that is seriously questioned. However, it is a concept that is assumed in arms control discussions between, and agreements made by, the governments and military establishments of U.S.A. and Russia. Nuclear deterrence is accepted for the present discussion because, without implying its validity, I am arguing for a change of just one feature in the two States’ military posture, for their own safety and the safety of the people of the world. The change is a logical necessity; it is urgently needed, and does not require any change in the assumptions upon which policy is based, whether these are valid or not. It is financially neutral, not requiring substantial extra expense nor yielding significant savings.

1. Definition of L-o-W, and distinction from “Launch Under Attack”

The term “Launch on Warning” is used here in reference to retaliation with rocket-mounted nuclear weapons to a perceived nuclear attack. It is the policy that, when there is a warning (by radar or satellite sensors) of attacking missiles on the way, requires a decision to launch or not to launch retaliatory weapons to be made so promptly that the launch can be done before any incoming warhead has arrived and detonated.

Another term, “Launch under Attack”, has been used less precisely by U.S. Strategic Command and in Congress, possibly sometimes with the intention of causing confusion. It is commonly supposed to mean a prompt launch of retaliation as soon as one or more incoming nuclear weapons have detonated. However, in the late 1970’s it was included in the dictionary of military terms by the Joint Chiefs of Staff, explained as “execution by National Command Authorities of Single Integrated Operational Plan Forces subsequent to tactical warning of strategic nuclear attack against the United States and prior to first impact.” This is identical to “L-o-W”. But at times military spokespeople have said their policy is not L-o-W, but “launch under attack”, implying that there is a difference, and that retaliation would be launched only after impact or detonation.

An alternative distinction has sometimes been implied: that L-o-W means to launch on a warning from one system (radar or satellite) alone, and “launch under attack” means launching retaliation before detonation, but only if the warning is confirmed by a second system. This is too uncertain a distinction to rely on. If communication with one system were temporarily disrupted there would be great pressure to act on an indication from only one. In fact it is believed that, because the Russian satellite fleet is incomplete, there are periods when segments of their periphery are not doubly monitored. Also some of the radar complexes installed under the Soviet system are now in independent States, and there is said to be a corridor along which missiles could approach with no warning early enough for evaluation of the situation before impact.

Both Russia and U.S.A. are believed to have the policy of L-o-W, as defined above, at the present time. If this is true of Russia, they must be relying on warning from only one system for a large fraction of the time. Their satellite fleet is incomplete and there are periods when segments of their periphery are not doubly monitored. Some of the radar complexes installed under the Soviet system are now in independent States. There is said to be a corridor along which missiles could approach giving no warning early enough for evaluation of the situation before impact. We have no way of knowing whether, for that direction of attack, their retaliation would be purely reflex or would wait for impact.

2. History of L-o-W, and the reasons for L-o-W

The avowed function of nuclear ballistic missiles is “deterrence”. Deterrence is in theory achieved when a potential attacker is convinced that an attack will be unavoidably followed by retaliation so devastating that it would be irrational to try it.

When the extreme destructiveness of nuclear weapons was realized, it became clear that a massive first salvo directed at command and weapon locations and at communications could diminish or eliminate a response. Knowledge of that by the enemy would weaken deterrence and invite a “disarming first strike”. To avoid this weakening of deterrence, the possibility was explored of launching retaliation before the first impact and detonation, L-o-W. It was probably put into effect as soon as such a quick launch became possible. The development of solid fuel as rocket propellant (around 1960) was a decisive factor.

During atmospheric testing of nuclear weapons in the early 1950’s the electrical phenomenon called “Electro-Magnetic Pulse” (EMP) was discovered. The Electro-Magnetic Pulse (EMP) is an extremely sharp and energetic electromagnetic impulse that is emitted by electrons traveling at nearly the speed of light from a nuclear explosion. It is maximal when the detonation is at very high altitude and the electrons interact with the earth’s magnetic field above the atmosphere. It disrupts unshielded electrical and electronic equipment over a wide area. Around1960 the U.S. conducted a series of high-altitude nuclear explosions to investigate it, incidentally causing significant disruption of radio communications each time. The purpose was presumably two-fold: to explore the possibility of using the phenomenon to enable a disarming first strike, and to study methods of protecting their own electronic equipment so that deterrence would be maintained even if the enemy was planning to use EMP. The possibility that electrical disruptions might prevent retaliation made a second reason to adopt L-o-W.

As early as 1960 the propriety and morality of adopting L-o-W was being discussed , because of the recognized danger of launching on a false warning and so starting an unintended nuclear war. In that year the Planning Board wrote that it was “essential” to avoid the possibility of launching unrecallable missiles based on a false warning. They stressed the importance of a “reliable bomb alarm system to provide early positive information of actual missile hits.” In 1962, Robert McNamara said that as long as he was Secretary of Defense and Jack Kennedy was President, the U.S. would never launch on warning. But the same year, the Secretary of the Air Force must have been thinking of L-o-W when he informed Kennedy that once the Minuteman missiles had been deployed in the first complex, in their “normal alert status”, all “twenty missiles will be able to be launched in thirty seconds.”

A discussion in 1969 is on record as showing that some who were opposing “Ballistic Missile Defense” favoured L-o-W, but The White House is stated to have opposed it “on the grounds that 50% of warnings from Over-the-Horizon Radar were false”. (The validity of this position seems dubious, because no true warning of a nuclear ballistic missile attack has ever been received. Presumably the other 50% were true observations of test rocket launches.) Lawrence Lynn, of U.S. National Security Council Staff, responded that the new satellite early warning system was estimated to produce only one false alarm per year, which he evidently regarded as acceptable. Georgy Arbatov, a Soviet deterrence specialist who had joined the National Security Council, assured Council members that “neither side would wait if it received warning of an attack but instead … would simply empty its silos by launching a counter-strike at once.” That removes concern about failure of deterrence against a surprise first strike, but underlines the danger from a false warning.

It seems probable that by 1969 L-o-W had been the military policy on both sides, for a number of years, notwithstanding the record that in 1973 Secretary of Defense Melvin Laird expressed the hope that “that kind of strategy would never be adopted by any Administration or by any Congress.” The recollections of former officers and enlisted men of Strategic Air Command (SAC) from the early 1970’s confirm that L-o-W was in effect then.

The policy of L-o-W is apparently still in effect both in U.S.A. and Russia, even though the Cold War is regarded as over. This seems inexcusably dangerous.

3. The danger of inadvertent N_war from false warnings or accidental coincidences.

Launch on Warning has kept the world exposed, for at least 30 years, to the danger of a nuclear war caused by nothing but a coincidence of radar, sensor, or computer glitches, and a temporary failure of human alertness to appreciate that an unexpected message of attack from the warning system is false, the enemy having done nothing. There is at most 20 minutes for the human operators and commanders to call and conduct a “threat conference”; and for the chief of SAC is put in touch with the President to advise him, and the President decides whether to order retaliation. The disaster of an accidental nuclear war has not happened yet, in spite of a large number of false warnings of which at least a few have had very dangerous features. This is a credit to the care and alertness of the military in both Russia and U.S.A. It should not be taken as reassurance. A single instance of launch of nuclear weapons on a false warning would result in nuclear war, and the end of civilization, just as surely as a nuclear war started by an actual attack. There would be no chance to review the system to make it safer after one failure of that kind.

The threat conferences require, and so far have achieved, the extraordinary standard of perfect accuracy. They have been called many times. Probably most of them have been routine and it was easy to exclude a real attack. To get some idea of how the laws of chance apply to the situation, suppose we make a very conservative assumption, that just one conference a year had a risk of error as high 1% (and that the rest had a much lower risk). It is simple mathematics to calculate that one 1% risk of disaster per year for 30 years does not result in 30%, but in a 26% probability of one actual disaster in that time. On that assumption, then, we had approximately 3 to 1 odds in favour of surviving the period 1970 – 2000, and we did survive. But so, from the risk of accidental war alone, we had a one in four chance of not surviving, which is not very safe. A single trial of Russian roulette is safer: it gives a one in six chance of death, or 5 to 1 odds on surviving. (This is not an attempt to calculate an actual probability. It is merely an example to illustrate the cumulative effect of any low-probability risk that is taken repeatedly, or accepted continuously, over a period of time.)

During the Cold War, many mishaps within the nuclear retaliation system on the U.S. side are known, including false warnings. There must have also been many such on the Russian side. One has been reported in which an officer decided on his own initiative not to report an apparently grave warning on his computer screen, on the correct belief that it was a false warning. He may have saved the world, but was disgraced for failing to follow his orders; his career was ruined, and he suffered a mental breakdown.

Scott Sagan described a large number of errors and accidents within the U.S. nuclear deterrence system, in a study of rival theories of accident probabilities. He concluded that the risk of nuclear war from accidents had not been excessive. I came to the opposite conclusion from his data. I have collected from that source and others, 20 mishaps that, with less alertness among military officers, or some coincidental problem, might possibly have started a war.

One example of a situation that was difficult to assess correctly at the Command Center, was this: On the night of 24 November, 1961, all communication links between SAC HQ and NORAD went dead, and so cut SAC HQ off from the three Ballistic Missile Early Warning Sites (BMEWS), at Thule (Greenland), Clear (Alaska), and Fylingdales (England) . For General Power at SAC HQ, there were two possible explanations: either enemy action, or the coincidental failure of all the communication systems, which had multiple ostensibly independent routes, including commercial telephone circuits. The SAC bases in U.S.A. were therefore alerted and B-52 nuclear bomber crews started their engines, with instructions not to take off without further orders. In the hope of clarifying the situation, radio contact was made with an orbiting B-52 on airborne alert which was near Thule (5,000 kilometers away) at the time. Its crew contacted the BMEWS station and could report that no attack had taken place, so the alert was cancelled. The reason for the “coincidental” failure was that the “independent” routes for telephone and telegraph between NORAD and SAC HQ all ran through one relay station in Colorado (which seems to show a remarkably stupid error in system design). At that relay station a small fire had interrupted all the lines.

That particular false warning was followed by a coincidental mishap that could have been disastrous. It seems there was an error in transmitting the alert code to 380th Bomb Wing at Plattsburg, New York. A former aircraft maintenance technician, who was serving at that B-52 bomber base, recently told me his recollection of the incident (which is vivid) is that the coded order first received by the bomber crews instructed them to take off and proceed directly to their pre-assigned targets, and bomb. The order was down-graded to the code meaning “wait with engines running”, before any plane had taken off. If the corrected code had not been received in time it could have been very difficult to stop the bombers.

The episode just described took place before L-o-W was instituted for the ICBMs that were in service. By 1979 the policy of L-o-W was in effect, and on the morning of 9 November a war games tape was running in a reserve computer when failure of the operational computer automatically switched in the reserve to take its place. The Threat Conference saw the picture of a massive attack in a realistic trajectory from Russian launch sites. On that occasion, preparation to retaliate got as far as launch of the president’s National Emergency Airborne Command Post (though without the president), before the error was discovered.

The most recent example known to the public was on 25 January 1995 when, as described in a report of the Standing Committee for Foreign Affairs and International Trade, “the Russian missile early warning system detected a scientific rocket launched off the coast of Norway. This area is frequented by U.S. submarines, whose ballistic missiles could scatter eight nuclear warheads over Moscow within fifteen minutes. Norway had informed the Russian Foreign Ministry about the upcoming launch, but this information had not been transmitted to the military. Over the next several minutes President Yeltsin was informed of the possible American attack, and, for the first time ever, his ‘nuclear briefcase’ was switched into alert mode for emergency use, allowing him to order a full Russian nuclear response. Tension mounted as the rocket separated into several stages, but the crisis ended after about eight minutes (just a few minutes before the procedural deadline to respond to an impending nuclear attack) when it became clear that the rocket was headed out to sea and would not pose a threat to Russia.”

4. Recommendations to U.S. Government on the need for de-alerting.

Several reports to governments have indicated the importance of abandoning this hair-trigger stance with weapons of such terrible destructive power. The first two steps in the recommendations of the Canberra Commission were: “• Taking nuclear forces off alert • Removal of warheads from delivery vehicles.”

5. A working definition of “de-alerting”, and how it might be done.

“De-alerting” is a term commonly used in suggestions and recommendations that nuclear weapons should be taken off “hair-trigger alert”. A definition that covers the current usage is that physical changes are made so that there is an unavoidable delay between a decision to launch, and the irrevocable step by which the rocket is actually launched. It would be incompatible with L-o-W, and would provide a period in which to review the fateful decision and take counsel.

A wide variety of methods have been suggested to introduce this delay. A very radical one would be to have all warheads removed from all delivery vehicles, and stored at a distance from them. Others include:

• making a heap of earth and rocks on silo lids that requires heavy machinery to remove it;
• removing hydraulic fluid from the machines that raise silo lids;
• inactivating the mechanism that rolls back garage roofs (Russia);
• pinning open a switch in a place that takes time to reach, or within a casing that takes time to open; and
• removing batteries, gyroscopes, or guidance mechanisms from rockets or re-entry vehicles

Some methods would be applicable to some types of warhead, and some to others. For each one a suitable method would have to be chosen and the details adjusted to achieve the required delay.

De-alerting would make sure that nuclear weapons could not be brought into use hastily. It would tend to reduce reliance on them in crisis situations, and thus be a step towards their eventual elimination from national arsenals. De-alerting would also make unauthorized launch of a nuclear weapon far more difficult to do, and would remove entirely the risk of accidental war due to a false warning. It would make more improbable the already unlikely event of a serious dispute between Russia and U.S.A. pushing either of the two into intentionally starting a war, by giving more time for diplomatic exchanges between the hostile governments and for conciliatory efforts by third parties.

However, there seems to be an obstacle that is difficult to overcome. Until elimination of the weapons is complete and assured by treaty, it has to be assumed that governments will go on regarding possession of nuclear weapons as essential deterrence against use of them by their enemy.

The two governments will require approximate equality of arsenals, and the military will insist on close symmetry of the de-alerting delays enforced. That would require rather complex arrangements, because of the differences in warhead types and in delivery vehicles. In the event that one side was actually intending to start a war, they would as rapidly as possible start to reverse the de-alerting process and ready their first salvo to fire. The other side would follow suit immediately, assuming that a verification system was set up so that they knew immediately. During the re-alerting process the weapon system would be vulnerable to attack, so whatever different systems were in use, the enforced delay times would have to be equal.

The governments will also demand intrusive verification to ensure the completeness of the de-alerting measures actually carried out, and that they cannot be secretly reversed. This may require representatives from neutral countries, and perhaps from the adversary, in the vicinity of each one’s launch sites. The military on both sides will then be concerned about maintaining the secrecy of features of their systems. To include verification acceptably for submarine-launched missiles would be extremely difficult.

It would take a prolonged technical study and discussion to set up these two systems — the de-alerting itself and the verification — in a way that would satisfy the two adversaries. When the experts and the military establishments were satisfied (which might prove impossible) a formal written agreement would be needed. This could require an actual treaty which would need ratification by the parliament on each side, and this raises another possibility of disappointing failure after years of work.

The only other way of getting the two governments and military establishments to accept de-alerting, is to persuade them that deterrence with nuclear weapons is no longer necessary. It may be even more difficult to do that.

6. NO L-o-W would be much easier to achieve.

A very simple change can be made which would remove entirely the risk of a launch of nuclear weapons in response to a false alarm. It is to abandon the policy of Launch on Warning. Since a false warning is immediately revealed as such when the predicted time has passed for the first rockets to arrive and no detonation has been detected, simply delaying retaliation until there has been a nuclear detonation guarantees that a war will not be started accidentally from that cause.

Incidents from which a purely accidental war might have been started seem to have outnumbered the actual geopolitical crises when nuclear war was intentionally threatened. Most of these threats, though dangerous, have been regarded as threatening gestures rather than an actual thought of going to war.

Since the Berlin Wall came down, the most serious threat of a nuclear war between Russia and U.S.A. known to the public has been the “Norwegian Rocket event” of January 1995, described above. Without L-o-W, the Russian alert and the anxious few minutes would still have occurred, but there would have been absolutely no danger of nuclear war because the rocket was unarmed and there could not have been a nuclear explosion.

To change from L-o-W to NO L-o-W does not require any change of alert status of the retaliatory system. It only requires a change of standing orders and standard operating procedure, such that no launch may take place until a nuclear detonation is reported.

A possible procedure might be as follows: As soon as a warning is received that the threat conference deems might be real, the order to prepare for launch is given, and arrangements are made to put the Chief of Strategic Command in touch with the President (or the equivalent steps on the Russian side, if it is their warning system that has come into action). As the situation develops over the next few minutes, the President decides whether, in the event of it proving to be a real attack, he will launch immediate retaliation. If he decides to do so, he will give a conditional order to retaliate using the appropriate SIOP option. Perhaps half of the launch code will be sent to the silos at once. The remainder will sent with the order to launch as soon as a detonation is reported to the military command. If the actual launch order is not received at the silos within a predetermined time, the preparatory order is regarded as cancelled and the preparatory steps are reversed without further orders.

Bomb alarms were installed many years ago near all military installations and all big cities in U.S.A., and presumably in Russia, which automatically and instantaneously indicate at the Strategic Command Centers any nuclear explosion and its location. If, and only if, indication of a nuclear explosion was received at the predicted arrival time of the attack, the final order to launch would be sent immediately to the silos. No delay to obtain presidential authorization would be needed at that point. The actual retaliatory launch could probably take place within a minute of the first detonation.

With a procedure like that in place, there would be no possibility of launching nuclear-armed rockets unless an actual attack had taken place. It would probably eliminate 90% of the current risk of nuclear war between U.S.A. and Russia.

A secondary benefit would be the reduced stress on the President during those vital minutes. He would know that he was not in danger of starting a war on a false warning. Under L-o-W, that worry might impair his concentration on the main issues.

It is worth noting that the change to NO L-o-W can start with a unilateral decision. Neither side wants an accidental war. They know that if either side mistakenly launches nuclear weapons both countries are going to be destroyed: it makes no difference who started it. If one side changes to NO L-o-W the risk of a purely accidental war from a false warning is approximately halved, immediately. It does not depend on the other side knowing that the change has been made.

7. Any loss of deterrence is minimal.

There can be few grounds for objection, by the military or by the governments, to this very necessary safety measure. One objection has to be taken seriously: that “NO L-o-W” might impair deterrence and tempt the enemy to try a “disarming first strike”. There are good reasons why this objection should not be allowed to prevent the change.

• First, for a first strike to be a rational option, the attacking side would have to be sure that the first salvo could effectively disarm the enemy within one minute of the first detonation. Their degree of alertness has not been reduced, and retaliation for a real attack would still be launched within a minute, as explained above. Synchronisation of detonation times on widely separated launch sites and command posts could not be assured to that accuracy, even if the many other uncertainties could be overcome.

The other possible method to prevent retaliation would be a first salvo engineered to maximize EMP and disable enemy electronics instantly. It is hardly credible that the attacking side could feel sure that their EMP would disrupt communication and launch mechanisms sufficiently, since they know that military electronics will have been shielded. Furthermore, they will know that submarine-launched missiles will not be disabled, because the sea-water shields the submarines and their contents.

• Second, the side planning an attack would have to be sure that their adversary had in fact changed to and remained under a policy of NO L-o-W. They cannot be sure without verification. A verification agreement is therefore not only unnecessary but actually undesirable.

• Lastly, the original reason for L-o-W does not now apply. During the Cold War it was assumed by both sides that the other would make a pre-emptive first strike in the event that it believed war to be inevitable, or would contemplate an unprovoked attack if that were possible without unacceptable damage to itself. At the present time it seems most unlikely that either U.S.A. or Russia is contemplating an attempt to destroy the other by a nuclear war. Nor are there any disagreements on the horizon that are likely to make war “inevitable”.

If the situation should change so that either side believed a first strike against it to be once again a real threat, then the NO L-o-W policy can be changed back instantly to L-o-W by an executive order, and the small changes that were made in operating procedure could be overridden. In the ordinary course of events there would be no reason to change back: NO L-o-W on either side is an advantage to both, simply because it halves the risk of a purely accidental nuclear war.

If, despite these arguments, the military establishment on either side is not convinced, the head of state must balance the elimination of the very definite risk of accidental war due to a false warning, against a hypothetical possibility of weakened deterrence. The results of a nuclear war would be the same, whether started by accident or by intention.

8. Conclusion.

Compared with real de-alerting that introduces a time delay, the change to NO L-o-W is quick and simple. It does not need symmetry, verification, agreement, nor even trust, between the adversaries. If adopted unilaterally by one, it is of immediate benefit to both. It does not impair deterrence. Unilateral operation of NO L-o-W by one country for a time might well be sufficient for the other to understand the benefit and to realize that the change did not in fact invite a first strike.

Putting NO L-o-W into effect requires only an executive order, followed by a change in Standing Orders to the effect that no rocket is launched until a nuclear explosion is reported to Strategic Command. There is no reduction in alert status. There would be minor changes in the launch sequence to suit whatever safeguards would be made to ensure that no launch could occur while the crews in the silos were waiting for the final order, and that they would be ready for instant launch if that order came through.

All the world’s people would be safer for the change. Therefore all governments have a duty to their people to urge the U.S. and Russian governments to make it at once. NGOs should consider making this their primary objective until it is achieved. Thereafter, they will continue to urge the Nuclear Weapon States to fulfil their undertaking to eliminate nuclear weapons from their arsenals.


Ballistic Missile Early Warning System
ElectroMagnetic Pulse
Inter-Continental Ballistic Missile
Launch on Warning
Non-Governmental Organization
North American Aerospace Defense Command
Strategic Air Command (later changed to “Strategic Command”)
Single Integrated Operational Plan
Submarine-Launched Ballistic Missile
Union of Soviet Socialist Republics