reciprocal measures for arms stabilization

Reciprocal Measures for Arms StabilizationThe Harvard community has made this

article openly available. Please share howthis access benefits you. Your story matters

Citation Schelling, Thomas C. 2005. Reciprocal measures for armsstabilization. Daedalus 134(4): 101-117.

Published Version http://dx.doi.org/10.1162/001152605774431455

Citable link http://nrs.harvard.edu/urn-3:HUL.InstRepos:3425941

Terms of Use This article was downloaded from Harvard University’s DASHrepository, and is made available under the terms and conditionsapplicable to Other Posted Material, as set forth at http://nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of-use#LAA

http://osc.hul.harvard.edu/dash/open-access-feedback?handle=&title=Reciprocal%20Measures%20for%20Arms%20Stabilization&community=1/1&collection=1/2&owningCollection1/2&harvardAuthors=88ea4bef064aa909a5bb5fa5d65180ed&departmentEconomics

http://nrs.harvard.edu/urn-3:HUL.InstRepos:3425941

http://nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of-use#LAA



There has been a widespread change inthe thinking on arms control in the lastyear or so. Much of it is due to the focusof attention on “measures to safeguardagainst surprise attack” (to use the of-½cial terminology). Although this sub-ject is still listed anachronistically un-der “disarmament,” it is differently ori-ented. It assumes deterrence as the key-stone of our security policy, and tries toimprove it. It accepts a retaliatory capa-bility as something to be enhanced, notdegraded–something to be made moresecure, less accident-prone, less in needof striking quickly to avoid its own de-struction, less capable of gaining advan-tage from a sudden attack of its own. An anomaly of this approach to armscontrol is that it does not necessarilyinvolve “disarmament” in the literalsense.

Another anomaly, which rather shakesthe disarmament tradition, is that weap-ons may be more stabilizing and less ag-gressive if they are capable of civilianreprisal rather than of military engage-ment. A standoff between two retalia-tory forces is in some ways equivalent to an exchange of hostages; and “inhu-mane” weapons, capable of inflictingdamage but not able to go after the ene-my’s strategic forces, acquire virtue be-cause of their clearly deterrent functionand the lack of temptation they give ei-ther side to strike ½rst.

More important, though, is the factthat schemes to avert surprise attack are manifestly compatible with a nation-al military policy, not a renunciation ofit. They emphasize the possibility thatone can simultaneously think seriouslyand sympathetically about our militaryposture and about collaborating withour enemies to improve it. To propose,as does the notion of “measures to safe-guard against surprise attack,” that mil-itary cooperation with potential ene-mies may offer opportunities to improveour military posture, opens a new ½eldfor imaginative scienti½c and militarythinking, and may eventually enlist thesupport of the military services them-selves.

Dædalus Fall 2005 101

Thomas C. Schelling

Reciprocal measures for arms stabilization

Thomas C. Schelling, a Fellow of the AmericanAcademy since 1967, is Distinguished UniversityProfessor at the School of Public Policy at theUniversity of Maryland. At the time of this es-say’s publication in the Fall 1960 issue of “Dæ-dalus,” Schelling was professor of economics andassociate of the Center for International Affairs at Harvard University.

© 2005 by the American Academy of Arts & Sciences

Most of this progress is still ahead ofus; the revolution in thinking aboutarms control is barely started. Of½cial-ly we have taken only the most hesitantsteps in de½ning arms control in a waythat does not contradict our national se-curity policies. We still talk of½cially asthough “disarmament” can only savemoney, without noticing that under thenew philosophy it could cost more. Westill work of½cially with an image of dis-armament that makes it solely a peace-time (cold-wartime) process of negoti-ating explicit detailed agreements in amultinational context for the reductionor elimination of weapons, without ad-equately recognizing that, as in limit-ing war, limiting the arms race can be amore tacit and less formal process thanthe “treaty” idea implies. More impor-tant, the prevalent image of disarma-ment is still one that gives the process auniquely de½ned end point–the point ofno arms at all, or virtually none except inthe hands of some international authori-ty or synthetic state that would have thepower to police the world against inter-national violence but against nothingelse.

The cautious and the skeptical, thepessimists and the realists, have doubtsabout how rapidly that end point can be approached, whether it will be ap-proached at all, and whether the pro-cess once started may not be reversed.But the ultimate goal is rarely challengedexcept by those who have no interest inarms control. And by far the most fre-quent argument raised in favor of par-ticular limited measures of arms con-trol, perhaps the most widely persua-sive, is that these limited measures are at least “steps toward” the goal of ulti-mate disarmament. We have not facedup to the implications of the anomalythat “measures to safeguard against sur-prise attack” are designed to preserve a

nuclear striking power, and are not easilyconstrued as just another “step toward”ultimate disarmament.1

We still talk about “levels” of arma-ment or disarmament, as though therewere only two directions in which to go,up and down, the arms race going in onedirection and arms control in the other.We have not yet admitted that, even inthe framework of arms control, it couldbe an open question whether we oughtto be negotiating with our enemies formore arms, less arms, different kinds ofarms, or arrangements superimposed onexisting armaments. We have given littlethought even to the weapon system thatwould be required by that ultimate inter-national authority that might police theworld against armed violence, and towhether it, too, would be embarrassedby a “massive retaliation” doctrine thatwould lack credibility; whether it, too,might be subject to surprise attack;whether it, too, would lack resolution(as some think nato might lack resolu-tion) to reach an awful collective deci-sion in response to nibbling aggressionor bland violation.

The point of this paper is that there is a vast new area to be explored once we break out of the traditional con½ne-ment of “disarmament”–the entire areaof military collaboration with potentialenemies to reduce the likelihood of waror to reduce its scope and violence. It is an area worth exploring because ourpresent military policies and prospects,however we feel about the adequacy of current programs, cannot promisesecurity from a major thermonuclear

102 Dædalus Fall 2005

Thomas C.Schelling

1 See T. C. Schelling, “Surprise Attack and Dis-armament,” in Klaus Knorr, ed., nato andAmerican Security (Princeton, N.J.: PrincetonUniversity Press, 1959), or the shorter versionin T. C. Schelling, The Strategy of Conflict (Cam-bridge, Mass.: Harvard University Press, 1960),chap. 10.

war; and even modest improvementsachieved through cooperation with theSoviets should be welcome.

It is not true that in the modern worlda gain for the Russians is necessarily aloss for us, and vice versa. We can bothsuffer losses, and this fact provides scopefor cooperation. We both have–unlessthe Russians have already determined to launch an attack and are preparing for it–a common interest in reducingthe advantage of striking ½rst, simply be-cause that very advantage, even if com-mon to both sides, increases the likeli-hood of war. If at the expense of somecapability for launching surprise attackone can deny that capability to the other,it may be a good bargain. We both have acommon interest in avoiding the kind offalse alarm, panic, misunderstanding, orloss of control, that may lead to an un-premeditated war, in a situation aggra-vated by the recognition on both sidesthat it is better to go ½rst than to go sec-ond. We have a common interest in notgetting drawn or provoked or panickedinto war by the actions of a third party(whether that party intends the result or not). And we may have an interest in saving some money by not doing onboth sides the things that, if we both dothem, tend to cancel out.

This common interest does not de-pend on trust and good faith. In fact itseems likely that unless thoroughgoingdistrust can be acknowledged on bothsides, it may be hard to reach any realunderstanding on the subject. The intel-lectual clarity required to recognize thenature of the common interest may beincompatible with the pretense that wetrust each other, or that there is any se-quence of activities in the short run bywhich either side could demonstrate itsgood faith to the other.

Ancient despotisms may have under-stood better than we do how to tranquil-

ize relations between them while hatingand distrusting. They exchanged hos-tages, drank wine from the same glass,met in public to inhibit the massacre ofone by the other, and even deliberatelyexchanged spies to facilitate transmittalof authentic information. And perhaps,having exchanged a son for a daughter in the cold-blooded interest of contractenforcement, they may have reducedtension suf½ciently to permit a little af-fection to grow up in later generations.

The premise underlying my point ofview is that a main determinant of thelikelihood of war is the nature of pres-ent military technology. We and the Rus-sians are trapped by our military tech-nology. Weapon developments of thelast ½fteen years, especially of the lastseven or eight, have themselves been re-sponsible for the most alarming aspectsof the present strategic situation. Theyhave enhanced the advantage, in theevent war should come, of being the one to start it. They have inhumanlycompressed the time available to makethe most terrible decisions. They havealmost eliminated any belief that a reallybig war either could be or should be lim-ited in scope or brought to a close by anyprocess other than the sheer exhaustionof weapons. They have greatly reducedthe con½dence of either side that it canpredict the weapons its enemy has orwill have in the future. In these and oth-er ways the evolution of military tech-nology has exacerbated whatever pro-pensities toward war are inherent in thepolitical conflict between us and our en-emies. It might be naïve to say that thisis an unmixed evil for both us and theSoviets, since it powerfully affects thebilateral contest between us; neverthe-less, it is hard to escape the judgmentthat nature might have been kinder inthe way she let our military technology


Reciprocalmeasures for armsstabilization

unfold itself over the last decade and ahalf.

It is interesting–more than that, it is useful–to ask what technologicalachievements (available both to us andto our enemies) we wish had never oc-curred, and what technological failureswe wish had turned out otherwise. Dowe wish the hydrogen bomb had nevercome along to make intercontinentalmissiles economical? Do we wish thatnuclear-powered aircraft had made air-borne alert so cheap that retaliatory aircraft could stay aloft rather than bevulnerable on the ground to a missileattack? Do we hope that no one ever dis-covers an economical means of nullify-ing ballistic-missile submarines, so thatneither side can hope to preclude retalia-tion by sudden attack? Do we wish thatwarning systems were so nearly perfectthat “false alarm” were virtually impos-sible, or so poor that we could never betempted to rely on them? Do we wishthat missiles had never become so accu-rate that they could be used to destroy an enemy’s missiles in an effort to ne-gate an enemy’s retaliatory threat? Dowe wish that radioactive fallout couldnot occur, or do we welcome it as a pe-culiarly retaliatory (and hence deter-rent) weapon effect that is of little use in a preemptive attack? Do we wish thatsecrecy about weapons and weapon pro-duction were much more dif½cult tomaintain than it is, or welcome certainkinds of secrecy as a form of mutuallyappreciated security against surpriseattack?

The reason why it is productive tospeculate on these questions, rather than merely fanciful, is that arms con-trol can usefully be thought of as a wayof changing some of the answers. In ad-dition to what we can do unilaterally to improve our warning, to maintainclose control over our forces, to make

our forces more secure against attack, to avoid the need for precipitate deci-sions, and to avoid accidents or the mis-taken decisions that they might cause,there may be opportunities to exchangefacilities or understandings with our en-emies, or to design and deploy our forcesdifferently by agreement with our ene-mies who do likewise, in a way that en-hances those aspects of technology welike and that helps to nullify those thatwe do not.

If we wish that radar were better andcheaper and less limited by the Earth’scurvature, we might make it so by ex-changing real estate with the Russiansfor the construction by each of us ofobservation posts on each other’s soil. If we hope that no one can ever predictwith con½dence how his own missileswould do, in a surprise attack, againstthe hardened missile sites of his oppo-nent, we might deny each other the nec-essary knowledge by banning tests oflarge weapons in the era in which any-one actually has a missile in a hard un-derground site that he could use in aweapon-effects test. If instead we wishthat each side might preserve the priva-cy of its railroad lines for mobile mis-siles, we might jointly eschew certainsurveillance techniques; and if wethought that antimissile defenses ofmissile sites might be more feasible, and retaliatory forces correspondinglyless vulnerable, with the further testingof nuclear weapons and their effects, wemight look with more favor on contin-ued weapon testing. These considera-tions are by no means the whole story in arms control, but they do remind usthat we and our enemies can both joint-ly welcome, or jointly deplore, certaintechnological developments (like theimproved accuracy of long-range mis-siles) and may possibly ½nd ways, joint-ly, to enhance them or to offset them,


Thomas C.Schelling

over and above the things that we can do unilaterally.

These examples suggest some of thecriteria that can be applied to limitedarms-control schemes, and some of thedif½culties in implementing them. As to criteria, the ½rst thing to emphasize is that it takes a good deal of strategicanalysis to decide whether a particularlimitation or augmentation of weaponsor facilities is a good one or a bad one.Viewing limited measures on their indi-vidual merits, and not as steps in a com-prehensive program that can be justi½edonly by a long sequence of steps to fol-low, one has to ask whether the techno-logical and economic consequences ofa particular scheme are or are not con-ducive to military stability; and the an-swer is very unlikely to be closely cor-related with whether more weapons or fewer weapons are involved, biggerweapons or smaller ones, or even wheth-er notions of “more” and “less,” “big-ger” and “smaller,” can be applied.Whether we would like to see reconnais-sance satellites banned or encouragedmay depend, for example, on whetherwe think they will mainly provide target-ing information to the initiator of war or mainly provide warning to a potentialdefender so that a potential attacker isthe more deterred. Whether we like bigmissiles or not may depend on whetherwe believe, as so many believed a fewyears ago, that missiles would be simpleand sturdy and hard to destroy in theirunderground sites or believe as so manyfear now that increased accuracies andyields make the present generation ofmissiles better for a ½rst strike than for a second strike. Whether we wish mis-sile technology to be advanced or retard-ed may depend on whether or not we be-lieve, as many do, that the next genera-tion of missiles will be easier to protect,

easier to hide, or easier to keep moving,and therefore less insecure. Whether one welcomes nuclear-powered ballistic-missile submarines on both sides or de-plores them depends on whether theyseem to be peculiarly good at survivingand retaliating, and hence “deterrent,”or peculiarly good at getting up close for a no-warning strike on an enemy’sretaliatory power. And if it were some-how possible to enforce a ban on “dir-ty” bombs, there would still be a gen-uine strategic question of whether or not we wish deterrent capabilities to beenhanced by the greater punitive powerof dirty bombs, recognizing that com-paratively slow-acting fallout may be ofmuch less utility to a potential attacker,whose main interest is to minimize re-taliation on himself.

The fact that developments such asthese require strategic analysis before it can be decided whether they are goodor bad is, aside from being true, discour-aging. It means that even among the ex-perts there will be disagreement aboutthe consequences of any particular pro-hibition or exchange of military facili-ties; it may be next to impossible to getwidespread understanding of the rele-vant arguments, even within govern-ments. And if fairly detailed analysis isrequired, and careful distinctions haveto be made, prohibitions might have to be speci½ed in equally careful detailand with equally ½ne distinctions. Thisis certainly an obstacle to negotiation.Furthermore, any analysis–and any pro-hibition or agreement or exchange offacilities that is justi½ed on the basis ofsuch analysis–is subject to rapid obso-lescence. The friendly warning satelliteappears, a year later, as a vicious target-ing aid to the surprise attacker; the net-work of warning systems originally de-signed for mutual reassurance proves in



operation to have too high a false-alarmrate; the missile-guidance systems thatwe deplored because of their extremeaccuracy and the advantage they wouldgive the attacker may prove, after weoutlaw them, to have been the mainhope for mobile missile systems desiredfor their invulnerability and hence fortheir stability. By the time we reachagreement on precisely what to allow inour satellites, where to place our radar,or what missiles to ban, new evidence ornew analysis comes along to suggest thatthe justi½cation of the particular schemewe are about to subscribe to is all wrong.

Finally, by the time we look at individ-ual schemes in suf½cient detail to judgewhether their strategic implications are“good” for both us and our enemies, wemay have narrowed them down to thepoint where they are intolerably biased.It is probably a mathematically soundprinciple that the more measures we putin a package, the more their bilateral bi-ases will cancel out, and hence the great-er will be the joint gain relative to thecompetitive advantage. This may meanthat once a potential arms-control sys-tem is dissected into suf½ciently smallpieces to apply the right kind of analy-sis, we shall have more individual bar-gaining counters too small and too bi-ased for the negotiating process.

The recent negotiations on weapontests may prove to be typical. First, therehas been almost no public discussion ofwhether the further testing of weaponsand weapon effects would really be con-ducive to the development of greater bi-lateral military stability or instabilityover the coming years.2 Even if the pub-

lic could be got interested in this crucialquestion, it would be unlikely to havethe information it would need to judgethe answer. (There has been a good dealof public discussion of the merits andpossible demerits of preventing the fur-ther spread of nuclear weapons to smallcountries, but remarkably little discus-sion of just how a test ban would ob-struct the spread.) Second, while it mayseem a mischievous stroke of fortunethat somebody discovered, between the two conferences, facts or ideas thatmade the policing of a test ban appearmore dif½cult than it had appeared theyear before, this may be exactly what wehave to expect in every case. If today wehad “completely solved” the new techni-cal problems introduced by the “decou-pling” technique, we should still have tobe prepared for somebody’s discoveringnext year a new possibility that had beenoverlooked, one that contemporary de-tection technology could not yet copewith.

The test-ban discussions also illustratethat, when an issue has been narroweddown, the bias in the advantages mayseem to outweigh the joint advantages.There is more controversy, and under-standably so, over whether a prohibitionon small-weapon tests is in the Ameri-can interest, than on whether a prohibi-tion covering the whole spectrum is.

But of all the characteristics of thepresent test-ban negotiations, the most


Thomas C.Schelling

2 That is, whether further testing would main-ly facilitate the development of more secure re-taliatory weapon systems with better commu-nication and control, less subject to accidentand false alarm, or instead would mainly en-

hance the potency of weapons for preemptiveattack and aggravate the urge, when in doubt,to strike quickly and without restraint. The an-swer is by no means obvious for the period im-mediately ahead. It should be noted that testsinvolve not only new-weapon performance butweapon effects on previously untested targets,and the latter may be especially relevant tosuch things as anti-icbm defense, civil defense,and the vulnerability of ½xed or mobile weap-ons, warning systems, and communication andcontrol systems.

signi½cant may be that we have had amoratorium for some time without aformal agreement. (We do not, ofcourse, have rights of inspection; so wecannot be sure that the moratorium hasbeen kept; but it likely has been, exceptpossibly for the most easily disguisedtests.) And this moratorium resultedfrom no detailed negotiations, no care-ful speci½cations, and no written docu-ments to be initialed and rati½ed. I donot think this result can be wholly ex-plained by the pressure of public opin-ion. Part of the motivation must be that,whatever one side is sacri½cing in im-proved technology, the other side is alsoforegoing tests, and each would proba-bly resume them if the other did. Thusthe main sanction of an arms-controlagreement–the expectation that eachwill abstain only if the other does–isprobably present in this case. It is there-fore a genuine instance of “arms con-trol.” If it suffers from being tentative,temporary, quali½ed, and conditional, somight any arms-control agreement, evenif duly negotiated and signed; further-more, who can say yet that the present“agreement,” if such we may call it, willnot be of some duration?

Here, I think, we have an importantclue to a process by which arms controlmay be reached, and the kinds of armscontrol that can be reached by that pro-cess. Maybe arms control is destined tobe something more informal than is sug-gested by the great diplomatic deploy-ments in Geneva. Maybe limited meas-ures of arms control can be arrived at byquite indirect and incomplete communi-cation; maybe they will take the form ofa proposal embodied in unilateral action(or abstention from action), which con-tinues if matched by corresponding ac-tion on the other side and only for solong as it is. Maybe instead of arguing

about what we should do, we will sim-ply do it and dare the other side to dolikewise, or do it and quietly suggest thatwe would like to keep it up, but only ifthey ½nd it in their interest to do some-thing comparable.

But if arms control is to be arrived atby a more tacit and informal process,and if we are going to call “arms con-trol” any of the military things that weand the Russians abstain from becauseof an awareness that as long as each ab-stains the other probably will too, weshould look around and see whether wedo not already have a good deal of armscontrol. If we have, we should look at itclosely to see what lessons we can draw.

Offhand, it appears (but a more imagi-native examination might prove other-wise) that the tacit understandings wehave with the Russians concern what wedo with our weapons more than what wepossess.3 We seem to have some under-standings about traf½c rules for patrol-ling bombers; there are apparently cer-tain lines we stay on this side of, linesthe Russians presumably can recognize,the crossing of which they can probab-ly monitor to some extent. This is cer-tainly a restraint that we unilaterallyobserve in the interest of reducing mis-understandings and alarms. As far as Iknow, the traf½c rules are communicat-ed, not explicitly, but simply by behav-



3 A possible exception is civil defense. Theextraordinary aversion to civil defense in theU.S. government must be complex in its ex-planation; but an element is very likely a be-lief that a genuine civil defense program mightopen up a new dimension of the arms race,leading either to a “civil-defense race” with the ussr or just to an aggravation of the armscompetition. The same may be true in theussr. An interesting question is how much“clandestine” civil defense the Russians areundertaking, and their reasons for keeping itprivate. (In pointing this out, the author is nottrying to justify the aversion to civil defense.)

ing in accordance with them (perhapsconspicuously in accordance with them)and possibly by having chosen the divid-ing lines in such a way that their signi½-cance is recognizable. We both abstainfrom harassing actions on each other’sstrategic forces; we do not jam eachother’s military communications, scareeach other with fallout from weaponstests, or wage surreptitious peacetimeundersea wars of attrition.4 We may yet develop tacit understandings aboutzones and traf½c rules for submarines,and may (or may not) develop a tradi-tion for leaving each other’s reconnais-sance satellites alone. We both very ob-viously abstain from assassination. TheRussians recently “negotiated” (by aprocess of nudging) a sharper under-standing about sharing the Paci½c fortarget practice. It remains to be seenwhether the U-2 incident causes certaintacit or latent understandings to comeunstuck.5

In all likelihood we may abstain fromthe use of nuclear weapons in some lim-ited war, though both sides often seemto denounce of½cially the notion that aserious limited war should be, or couldbe, fought without nuclear weapons.Here is an interesting case of an armslimitation that may be tacitly recognizedby both sides, and recognized only be-cause each thinks the other may observe

it too, yet one that is not only not for-mally agreed on but even denounced anddenied by both sides. It seems doubtfulwhether this tacit understanding couldbe made much stronger by a writtendocument.6 A restraint on the use ofnuclear weapons may be more persua-sive if it seems to rest on the enemy’sown self-interest–on his understandingthat if he abstains we may too, but onlyif he does–than if it pretends to rest onthe power of a written agreement or on a½ction of “good faith.”

In fact, all of the tacitly agreed limitsthat do apply, or may apply, in limitedwar can be construed as a kind of infor-mal arms control tacitly arrived at. Myimpression is that we and the Russianswill go to some length to avoid havingAmerican and Russian troops directlyengage each other in a limited war, sim-ply because such an engagement mightcreate extremely unstable expectationsabout whether the war could remainlimited. We and the Russians both rec-ognize many legalistic limitations inwar, such as the distinction betweenNorth Koreans and Chinese, betweenvolunteers and regulars, between theprovision of materials to an ally and theprovision of manpower, between doingan ally’s reconnaissance for him and do-ing his bombing, perhaps even the dis-tinction between local air½elds that arefair game because they are on the groundwithin a disputed country and the decksof carriers offshore that might for somereason be construed as “sanctuary.”


Thomas C.Schelling

4 Not yet, that is, or not very much. Preserv-ing some of the mutual restraints we now en-joy may be as important an “arms-control”objective as creating more.

5 It seems a correct interpretation that there is still some element of implicit understandingabout not transferring nuclear weapons to oth-er countries. Its status is presently a great dealmore ambiguous than the author expected acouple of years ago; nevertheless there must bea general awareness on both sides that the re-straint of either will be weakened or dissolvedby promiscuousness on the other’s part.

6 It could be made much stronger by variousunilateral actions. One would be to increaseour capability to get along without nuclears inlimited war. Another would be to add symbol-ic support to the understanding; the test-bannegotiations–especially if a formal agreementis reached–almost certainly do this, whetherthey are intended to or not.

Most of these limits are arbitrary, con-ventional, and casuistic–purely mattersof tradition and precedent. For that rea-son they are uncertain and insecure; no-body is even nominally committed tohonor them. But they demonstrate thatit is possible for potential enemies to ar-rive tacitly, or by indirect communica-tion, at a meeting of minds about somerules, and about how to interpret inten-tions through the way one operates anddeploys his resources. Most important,the limits that can be observed in limit-ed war are a powerful demonstrationthat sheer self-interest–the recognitionof a need to collaborate with an enemyin wartime, to reach understandings that transcend the formalities of explic-it communication; the recognition of amutual interest in avoiding accidents,incidents, misunderstandings and un-necessary alarms, and in holding to anyconstraints that can be found–can pro-vide potent sanctions that need not reston explicit negotiation and formal agree-ments.

We may, then, increase our under-standing of the nature of arms control,what it rests on and how it may comeabout, by recognizing limited war as akind of arms control in itself. And per-haps it differs from peacetime (i.e., cold-war) arms control less than we custom-arily think. Perhaps the psychology andthe sanctions and the mode of com-munication, the kinds of reasoning in-volved, the lack of formal agreement oreven acknowledgment, that typify lim-ited war, represent a more central andtypical process of international negoti-ation than we usually give it credit for.

There is another aspect of limited warthat deserves emphasis in this connec-tion. The limits in limited war are ar-rived at not by verbal bargaining, but bymaneuver, by actions, and by statementsand declarations that are not direct com-

munication to the enemy. Each sidetends to act in some kind of recogniz-able pattern, so that any limits that it is actually observing can be appreciatedby the enemy; and each tries to perceivewhat restraints the other is observing.For that reason the limits themselvesmust be clear-cut, must be of an “obvi-ous” character, must be based on quali-tative distinctions rather than matters of degree. They must not be too selec-tive, too gerrymandered in discriminat-ing between what is inside and what isoutside the limit. They must attachthemselves to benchmarks, demarca-tion lines, and distinctions that comenaturally. They must have simplicity.They must take advantage of conven-tions and traditions and precedents thatexist, even if the precedents and tradi-tions are biased between the two sides or a nuisance to both sides. Often theymust involve all-or-none distinctions, or across-the-board distinctions like thatbetween land and water, between mate-rial and manpower, between two sides of a border, or even some arbitrary butpotent and highly suggestive feature likea parallel of latitude.7

This is certainly true in the case ofthe use of nuclear weapons in limitedwar. It is enormously more likely that a limit against any use of nuclear weap-ons could be recognized, sensed, andadhered to by both sides on conditionthat each other observe it, than that anyparticular quantitative limitation, tar-get limitation, ½ssion vs. fusion limita-tion, or limitation based on who is the“aggressor,” could be jointly and tacitlyconverged on by the participants.



7 For an extensive analysis of tacit bargaining,with special reference to limited war, see chap-ters 3 and 4 and Appendix A of T. C. Schelling,The Strategy of Conflict (Cambridge, Mass.: Har-vard University Press, 1960).

But the same is certainly true of a testsuspension. A tacitly reached moratori-um on testing nuclear weapons–mutualand reciprocal but essentially unilateralon both sides–is much more likely to bestable and durable, much less likely to beeroded by ambiguous behavior, than aselective moratorium. If we and the Rus-sians are very selective in our unilateralrestraints, each choosing the particularyields, altitudes, ½ssion-fusion combina-tions, and localities for tests, it seemsunlikely either that both sides will hit onthe same limitations and maintain themwith con½dence, or that both will hit on“equivalent” though different restraints.

To some extent, then, the gains andlosses of a particular agreement, i.e., theway any particular understanding that isreached may discriminate between thetwo parties (or among more than twoparties), are likely to be dictated some-what by the elements of the problem,and not altogether by the detailed pref-erences of the parties to the understand-ing or their bargaining skill. An absoluteban on weapon tests, for example, or anyother across-the-board prohibition, issomewhat arbitrary in the way it distrib-utes the advantages; but perhaps someof its appeal is precisely in the fact that it is somewhat arbitrary, somewhat de-termined by chance or by the very struc-ture of the problem, dictated by circum-stances rather than by either side to theother.

If an important part of our arms con-trol–or let us call it “mutual arms ac-commodation”–with our enemies isgoing to be tacit and informal, a matterof reciprocated unilateral actions andabstentions, we need to take seriouslythe problem of communicating with ourenemies about what we are doing, and ofreaching understandings with them. Insome respects informal communication

is easier, in some ways harder; the pro-cess is different from that of formal, ex-plicit, detailed negotiation, and imposesdifferent requirements. Informal com-munication is usually ambiguous; a gov-ernment speaks by hint as well as byovert statement and proposal, it speaksindirectly through the medium of pressconferences, leaks of information, andremarks to third parties. It speaks withmany voices, in the executive branch, in the congress, and even in private arti-cles and news stories that are “inspired”or are inferred to be so. And it speaksthrough the actions it takes.8

The differences should not be exagger-ated; even when large teams of profes-sional diplomats and technical expertsare assembled in Geneva, much of thecommunication takes these other forms.Nevertheless, the strategy of communi-cation is different, particularly becauseof the greater need in informal negoti-ations to reach a real understanding. Informal and explicit negotiation, whateventually matters is to a large extentwhat gets written down and agreed to;even if there was not a meeting ofminds, there may have been a meeting of words that provides a record of theexpectations of both sides and the obli-gations perceived. In informal negotia-tion the ultimate sanction depends lesson a piece of paper than on the clarity of the understanding reached. If one be-haves in a particular way, in anticipationof the other’s reciprocation, there is aneed to make clear precisely how one isbehaving, with what mutual purpose inmind, so that the other can read the pro-posal in it, infer what would constitutereciprocation, and design its own behav-ior accordingly.


Thomas C.Schelling

8 In a sense, the abortive summit conference ofMay 1960 did not involve less “negotiation” justbecause the meeting never took place.

There is furthermore a greater need tobe persuasive. In explicit negotiation, itmay be possible to reach an agreementwhose terms are reasonably well under-stood without agreement on principlesor any reciprocal understanding of eachother’s motives. If the letter of the agree-ment is clear, the spirit can remainsomewhat in doubt. In informal negotia-tion, the spirit bears most of the burden;and if the idea behind what we think weare doing is not perceived by our partner(enemy), what we expect of him–orwhat we may reasonably be expected toexpect of him–may be too dimly per-ceived to be the basis for genuine recip-rocation.

Suppose we decide to put more em-phasis on ballistic-missile submarines,for example, in the belief that they arepeculiarly “stable” weapons because oftheir lesser susceptibility to destructionin case of a surprise attack and becausethey are not so much under obligation to strike quickly in the event of an am-biguous warning (or war itself ), or elsebecause their smaller warheads, withpossibly a lesser degree of accuracy ascompared with ground-based missiles,makes them less of a threat to the ene-my’s retaliatory forces and more of agenuine deterrent. Suppose we decidethat we could afford to do this only ifthe enemy himself oriented his ownstrategic program toward similarly “sta-ble” weapon systems. It might not be at all clear to the Russians what our mo-tives are, or what the conditions were forour going through with the program. Orsuppose we have a crash program for thedevelopment of a more secure ground-based missile force, this program to be½nanced by a sharp increase in the de-fense budget, with a good deal of expen-diture on command, control, and com-munication arrangements so as to re-duce both the vulnerability of our weap-

ons and their sensitivity to accident orfalse alarm. In particular, suppose thatour budget rises because of increasedoutlays associated with our desire for a slow reacting force, rather than one that must react rapidly. In such circum-stances, our actions may be stabilizingor destabilizing, depending on whetherthe enemy can perceive that we are mak-ing the world safer for him rather thanincreasing his need (and ours) to jumpthe gun in a crisis. If we institute an air-borne alert, it may be important to do so in a way that enhances the apparentas well as the real security and stabili-ty of our retaliatory weapon systems.This might mean that we would have tochoose deliberately, say, flight patternsthat manifestly enhance the security ofour forces rather than the speed withwhich they could initiate a surpriseattack of their own.

By far the most important prerequisiteis that we understand our own motiveswell enough to take actions that are con-sistent with a deterrent philosophy, andwell enough so that we can articulate itto ourselves. If we have such a philoso-phy, and if our actions are consistentwith it, and if for our own purposes wearticulate that philosophy in explainingour budget decisions here at home, weare probably well on the way to convey-ing that philosophy persuasively to ourenemy, if he is at all receptive. A specialproblem here is that our overt positionon disarmament must not be too incon-sistent with the philosophy that we aretrying to display and get across to ourenemy. If, for example, we really be-lieved in a policy of collaborating withthe Russians to develop a stable situa-tion of mutual deterrence, and if we de-termined to make important changes, to this end, in the con½guration of ourweapons but these changes were not inthe direction of general disarmament,



we would put a double burden on ourcommunication if the front we present-ed on arms-control questions bore norelation to that philosophy. This doesnot necessarily mean that we have tospeak in our formal disarmament di-plomacy in a manner that is sincere andconsistent with what we are fundamen-tally trying to get across to the Russians.It may just mean that our insincerityshould be as manifest as the inconsis-tency, so that when we do contradictourselves the Russians know that this is for show and that they should look for the real message elsewhere. Still, itwould help if we could ½nd the diplo-matic courage to shift even the formaldiscussions of arms control more intoaccord with our basic military policy, at the same time as we try to adapt thatmilitary policy in directions that theRussians can appreciate and reciprocate,so that disarmament negotiations canhelp a little, or at least hinder as little aspossible, the development of a genuineunderstanding.

Even so, it is still an unanswered ques-tion whether the Russians are at all dis-posed to participate in any “mutual armsaccommodation” with us, beyond whatwe already do in a tacit way. And it is adif½cult technical question whether,even if they are disposed to cooperatewith us and appreciate the principle ofstable retaliatory systems with mini-mum proclivity toward false alarm andminimum temptation toward surpriseattack, there are any promising actionsto be undertaken. Weapon systems canrarely be classi½ed indisputably as ½rst-strike or second-strike weapons, as “ac-cident-prone” or “accident-proof”; agood deal of technical analysis has to liebehind a judgment, many of the techni-cal judgments may not be made equallyby us and our enemies, the judgment hasto be made in the context of an evolving

weapon system for which facts are reallyonly forecasts, and what is known todaymay no longer be true tomorrow. It is,furthermore, too much to expect themassive bureaucracy of our defense es-tablishment and our foreign service, and the partisan conflicts in Congress, to produce and maintain a coherent phi-losophy and transmit it with high ½deli-ty to a suspicious enemy whose receptiv-ity and reasoning processes we can onlypoorly evaluate. But it is worth trying.

One possibility, already adverted to, is to design our military forces conspic-uously and deliberately in the directionof deterrence, stability, and slow reac-tion. That is, to articulate as a policy thedesign of a strategic force that is pecu-liarly good at waiting out crises, at sur-viving a surprise attack, and at punish-ing an attacker ex post facto, and not par-ticularly good at initiating a preventiveattack, not in need of responding rapidlyto warning.

This may not be a bad policy to followunilaterally; but the advantage of pursu-ing it is greater if the enemy pursues ittoo. The more each side perceives theother as designing his force for a suddenpreemptive attack in a crisis, or for a pre-meditated surprise attack, the more oneis tempted himself to develop a quick-reacting system, one that is peculiarlysuited to catching the enemy’s militaryforces before they have left the ground.Thus to some extent such a policy is aconditional policy; the motive is great-er if the principle is reciprocated by theenemy.

It would be extraordinarily dif½cult,perhaps impossible, to negotiate a de-tailed understanding of precisely whatkinds of weapons in what con½gura-tions, and how deployed, would meetthe “stability” criterion. For that rea-son the idea may not be one that lends


Thomas C.Schelling

itself to explicit detailed negotiatedagreements. But that does not rule outthe possibility that both sides may per-ceive value in pursuing such policies in a general way, and may recognize thattheir own behavior not only helps theother side pursue a similar policy buthelps to induce it by the tacit promise of reciprocation. As mentioned above,we already do this in such matters as the traf½c rules we both unilaterally ob-serve and reciprocate; there may be agood deal of room for gradually extend-ing this kind of reciprocal unilateral ac-tion, even though the subject may neverappear on the agenda of a diplomatic ne-gotiation.

Compared with a peaceful world dis-armed, schemes to stabilize mutual de-terrence are a poor second best; judgedagainst the prospect of war, measures tomake it less likely may be attractive. Thispoint of view will not appeal to any whobelieve that war results from the sheerexistence of arms and the temptation touse them, or from the influence of mili-tarists in modern society whose prestigeincreases in proportion to the arms bud-get, and who believe that distrust is onlyaggravated by people’s acting as thoughdistrust exists. History shows, it is said,that man cannot live in a world witharms without using them. History rare-ly shows anything quite that universal;but even granting it, the question is notwhether it is asking much of man tolearn to live in a world with arms andnot to use them excessively. The ques-tion is whether it takes more skill andwisdom for man to learn to live in aworld with arms and not to use themthan it does for man to disarm himselfso totally that he can’t have war even ifhe wants it (or can’t want it any longer).If modern social institutions are capa-ble of achieving disarmament in the ½rstplace, and of avoiding arms races in per-

petuity thereafter, perhaps they are ca-pable of supporting a world with armswithout war. Those who argue thatpeace with arms is impossible but act as though peace and disarmament arenot, may be using a double standard.

And it must be remembered that to-tal disarmament, even if achieved, doesnot by itself preclude subsequent armsraces; nor does a good start toward totaldisarmament preclude a violent reversal.To the extent that an arms advantage ismore easily obtained when the level ofarmaments on both sides is low–to theextent that the consequences of cheatingare greater in a world with few arms–arms races might become more violent,the lower the level of armament fromwhich they start. Particularly in a worldin which the pace of scienti½c progress is rapid but jerky, uneven as betweencountries, and full of opportunities anduncertainties for weapons development,it is not at all clear that the world wouldbe less uneasy about arms advantages ifeach side continually thought of itselfas nearly naked. What can explain thecomplacency of the American responseto the ½rst Soviet sputnik except a feel-ing (superbly rationalized) that the ex-isting level of arms provided so muchsecurity that no single new achievement,or even a revision of the comparativetime schedules by a year or two, couldquite upset the balance.

Another area of possible cooperationis in damping the arms race through theexchange of information. I am not muchimpressed with the budgetary fury ofour participation in the arms race, but it is not hard to imagine that the budget-ary arms race might get into much high-er gear. If it does, part of the motivation(at least in this country) may be due touncertainty about the level of armamenton the other side. The “missile gap” that



one estimates, or feels obliged to assumeto exist in the absence of information,may exceed the actual missile gap, caus-ing a more frantic increase in armamentsthan would be undertaken with betterinformation. And it may induce recipro-cal action on the other side, which alsowishes to avoid an intolerably unfavor-able imbalance.

To illustrate: suppose that either sidefelt reasonably secure against suddenattack as long as its enemy’s numericalsuperiority in missiles never reached,say, 2 to 1. In this case, just knowing whateach other possesses and is producingcould make possible a stable equilibriumat a modest level of strategic armaments,while ignorance of the enemy’s strengthmight seem to require an unlimited ef-fort to avoid falling too far behind. Withactual weapons such simple calculationsare of course impossible; but the princi-ple is valid.

An important dif½culty of applying it,though, is that the ways by which onecan get authentic information about theother’s present and projected strengthmay provide more strategic informationthan the other side can tolerate.9 A spe-cial dif½culty is that the Soviets may al-ready know most of what they need toknow for this purpose; it is mainly wewho do not.

But it is interesting that they mightpossibly prefer that we know the truth. If in fact we are on the verge of a crashprogram based on an exaggerated esti-mate of what they have already done, it could cost them money (and perhapsan increase in the risk of war) to keep

up with us. It is also interesting that thetruth is probably not something thatthey could readily reveal on their own.They have to ½nd some way of giving us evidence for believing the truth (or aless exaggerated estimate of the truth)and give it in a way that does not yieldtargeting and other information thatthey would ½nd intolerable. The fact that this intelligence gap is mainly onour side does not preclude Soviet inter-est in some means of conveying the in-formation to us, and it does not obviatethe need for cooperative techniques forreceiving it.

Measures to prevent “accidental war,”war by misunderstanding, war by falsealarm, are another possibility. One as-pect of this has been mentioned: thereciprocal development of the kinds offorces and modes of behavior that mini-mize accidents or their consequences,minimize alarms and misunderstand-ings, minimize the need to react quicklyin the face of ambiguous evidence. Butthere is another type of joint or recipro-cal activity that could help. It would beto arrange in advance, even if crudelyand informally, communication proce-dures, exchange of information, andinspection facilities, for use in the eventof an accident, alarm, or misunderstand-ing that created a crisis. Part of this isjust procedural–making sure that weand the Russians have the same ideaabout who gets in touch with whomwhen communication or bargaining issuddenly required. Part of it is intellec-tual–thinking ahead of time about howone would go about reassuring the Rus-sians in the event they had a false alarm,and what we could demand of them forour own reassurance if we ever got am-biguous evidence. Part of it is physical–making sure that, if we should need in-spectors on a particular scene within a


Thomas C.Schelling

9 Also, one side yields a bluf½ng or bargain-ing advantage if it reveals that its weaponry isless impressive than may have been thought. It loses, too, the possibility of surreptitiouslyachieving a dominant superiority. But losses of this kind are the price of arms control in the ½rst place.

few hours to verify that something wasan accident, or to verify that the Rus-sians were calm, or to verify that theRussians were not taking actions wethought they were taking, the neces-sary inspectors and equipment would be available within a few hours’ traveltime from where we would need them.Just having some Russians available atstrategic points around the UnitedStates, able to see things with their owneyes if we suddenly wanted them to andable to report home instantly throughauthentic channels, might be usefulsomeday. And if we ever want them, wemay want them in a hurry; there maynot be time to identify them, brief them,ship them over here, and train them fortheir job, once the accident occurs or thecrisis is on or the misinformation ½ltersthrough the Russian warning system.10

There is a more ambitious possibility.Neither we nor the Russians at the pres-ent time take arms control terribly seri-ously; we do not view it as an alternativeto a war that is imminent. But it is notimpossible to imagine crises in whichthe likelihood of immediate war wouldbecome a grave preoccupation. Once thethreat of imminent war rises above somethreshold, the mere consciousness thateach side is preoccupied with it–andwith the importance of being the one to start it, if it should come–will aggra-vate the propensities that already exist.It is perfectly conceivable that in a realcrisis there would be a sudden and dras-tic change in the attitudes of both sidestoward arms control. “Preventive armscontrol” might begin to look like a riskybut attractive alternative to a possibly

inevitable preemptive war. Sudden anddrastic “measures to safeguard againstsurprise attack” might have to be nego-tiated on an acutely demanding timeschedule.

If so, success may depend on wheth-er one or both sides is intellectually pre-pared for the contingency, whether someunderstandings have been reached inadvance, and whether certain facilitiescan be improvised to monitor whateverarrangements might be forthcoming.One of the important “limited” arms-control measures that we might take inadvance of such a crisis, either by our-selves or with our enemies, either infor-mally or explicitly, is a development ofunderstandings, procedures, personnel,and equipment, of an imaginative andadaptable sort, capable of going intoaction at such time as we and the Rus-sians both decide that now is the timefor arms control and we can’t wait.

A ½nal possibility, a pessimistic but a serious one and one suggested by theanalogy between arms control and limit-ed war, is the role of arms control in gen-eral war if general war occurs. We usual-ly think of arms control or deterrence as having failed if war breaks out; and so it has, but it can fail worse if we giveup at that point. It is not entirely clearthat a general war–a war between theusa and the ussr, involving their stra-tegic forces on a large scale–would nec-essarily be unlimited either in the way itwould be fought or in the way it wouldbe concluded. Particularly as we come to think about an inadvertent war–onethat results by some kind of accident ormisunderstanding, or one that is reluc-tantly initiated by the Russians or by usin the belief that it is urgent to preemptat once–it is worthwhile to considerwhether fury is the only guide we needin conducting the war, and whether the



10 A more extensive discussion of this pointwill appear in T. C. Schelling, “Arms Control:Proposal for a Special Surveillance Force,”World Politics (October 1960).

exhaustion of weapons on both sides isthe only condition for terminating it.

It is commonly taken for granted thatif the Russians initiate a general war itwould be in a vicious effort to extermi-nate us both as a nation and as a people,and that they would be so impatient todo this as to spend valuable weapons tocreate civil damage at the outset. But it is not obvious that a coldly calculatingenemy would afford himself the luxuryof going after cities and people whenthere are more urgent targets that he has to destroy in order to reduce thescale of our retaliation. Nor is it obvi-ous that an impetuous attacker, onewhose motivation is partly the fear thatif he does not strike ½rst he will be sec-ond, would be immune to the thoughtthat he might want to surrender if thething went badly, to accept our surren-der if it went well, or to negotiate a trucebetween those extremes. If there is noimmediate strategic need to kill our peo-ple, it may occur to him that they areworth more alive than dead; the threatof killing them gives him something tobargain with in the course of the war orat its termination. Similarly for us: if thewar was a mistake we might be more in-terested in minimizing the consequencesof the error, whosever error it was, andin maintaining the possibility of a nego-tiated outcome that limited damage onboth sides. For this bargaining purpose,live Russians and our unspent weaponsare assets, and about the only ones we’dhave.

The subject is a complicated one andcannot be decided here. It has to be ac-knowledged that there are dangers insuggesting to the Russians that we areeven aware of the possibility that anattack on us might not be cataclysmic for us both. But the possibility is so uni-versally unmentioned and so terriblyimportant that it deserves to be brought

into the open for study. Its relation toarms control is that the mere possibili-ty of limiting a general war between usand our principal enemy may depend onsome understanding, tacit and informalas it may be, that we share ahead of time.There may be little national advantage in abstaining from certain targets in theevent of war, or in attempting to com-municate, unless the enemy can be alertto what is going on.

Terminating a war through anythingother than the sheer exhaustion of weap-ons on both sides would require someform of arms control. It is a noteworthycharacteristic of a possible World WarIII that even unconditional surrendermay be physically impossible. How dothe Russians persuade us that they havedestroyed (or are prepared to destroy or deliver us) some or all of their signi½-cant weapons and are prepared to sub-mit to our political demands? We can-not even trust them not to test weaponsunder a test-suspension agreement; incircumstances in½nitely more desper-ate, when a one-hour pause in the warmay be of strategic bene½t to somebody,if they send us an urgent message ac-knowledging their guilt in the war andproposing that we preserve our world by letting them surrender to us, are welikely to be able to do anything? If theyare fooling, and if we are fooled, the costwill be tremendous; if they are not fool-ing and we choose to ignore them, thecost will be tremendous. Can we think of what they might do to prove that theymean it? Have we got the facilities tomonitor them and to police them? Havewe incorporated in our strategic forces,and in the operating doctrine of thoseforces, recognition of their potential role in policing the disarmament bywhich the war might be brought to aclose?


Thomas C.Schelling

Actually “surrender” is a poor wordhere. Anywhere between the twoextremes of unconditional surrender by one side or the other, the truce orunderstanding or scheme for bringingthe war to a close might better be de-scribed as “disarmament” or “arms control.” Historically one might haveallowed an enemy, when he “condition-ally” surrendered, to keep some purelydefensive weapons as a hedge against the victor’s violating his promise. This is a kind of asymmetrical disarmamentscheme. In the future, at the close of ageneral war, one might have to allow theconditionally surrendering enemy to re-tain some retaliatory weapons, these be-ing the only kind that two major powerscan use to enforce promises from eachother. In effect, “measures to safeguardagainst surprise attack,” possibly one-sided, possibly bilateral, and certainlymore drastic than any that have yet beenconsidered, might be the minimum re-quirement of a conditionally surrender-ing enemy.

Thus anywhere between the twoextremes of total surrender, the out-come should be viewed as a disarma-ment process, with the asymmetry pre-sumably reflecting the degree of victoryor defeat. But as remarked above, eventhe extremes of unconditional surren-der require much the same kind of pro-cedure for mutual relaxation, cessationof hostilities, inspection, enforcement,and so forth. Any general war that is ter-minated by a bilateral understanding, byanything other than the independent ex-haustion of weapons on both sides, re-quires something in the nature of anenormous, complex and dynamicscheme for arms control.

If this possibility is to be left open, weneed to anticipate it in the design of ourstrategic forces and in our plans for theiruse. It may require special facilities and

equipment to bring a war to a close, ofa kind not necessarily provided for in aplan that considers only the contingencyof an all-out war to the ½nish. But it alsorequires some mutual awareness aheadof time, on the part of both our enemyand ourselves, and perhaps some crudeand tacit, if not careful and explicit, un-derstanding about the modes and tech-niques of negotiation in the event ofwar.



This article has been cited by:

1. Avinash Dixit. 2006. Thomas Schelling's Contributions to Game Theory. Scandinavian Journal of Economics 108:2, 213-229.[CrossRef]

http://dx.doi.org/10.1111/j.1467-9442.2006.00447.x

reciprocal measures for arms stabilization

Documents