Trustees of Princeton University

Cooperation Under the Dilemma Author(s): Source: World Politics, Vol. 30, No. 2 (Jan., 1978), pp. 167-214 Published by: Cambridge University Press Stable URL: http://www.jstor.org/stable/2009958 . Accessed: 20/09/2011 21:08

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Cambridge University Press and Trustees of Princeton University are collaborating with JSTOR to digitize, preserve and extend access to World Politics.

http://www.jstor.org UNDER THE

By ROBERT JERVIS*

I. AND THE SECURITY DILEMMA T HE lackof an internationalsovereign not only permits to occur,but also makesit difficultfor states that are satisfiedwith the statusquo to arriveat goals thatthey recognize as beingin theircom- mon interest.Because there are no institutionsor authoritiesthat can make and enforceinternational laws, the policiesof cooperationthat will bringmutual rewardsif otherscooperate may bringdisaster if theydo not. Because statesare aware of this,anarchy encourages be- haviorthat leaves all concernedworse off than they could be, even in theextreme case in whichall stateswould like to freezethe status quo. This is trueof themen in Rousseau's"Stag Hunt." If theycooperate to trapthe stag, they will all eat well. But if one persondefects to chasea rabbit-whichhe likesless thanstag-none of the otherswill get any- thing.Thus, all actorshave the same preferenceorder, and thereis a solutionthat gives each his firstchoice: (i) cooperateand trapthe stag (the internationalanalogue being cooperation and disarmament);(2) chasea rabbitwhile othersremain at theirposts (maintain a highlevel of armswhile others are disarmed); (3) all chaserabbits (arms compe- titionand highrisk of ); and (4) stayat theoriginal position while anotherchases a rabbit (being disarmedwhile othersare armed).1

* I am gratefulto Robert Art, Bernard Brodie, and for comments,and to the Committee on Research of the UCLA Academic Senate for financial support. An earlier version of this essay appeared as Working Paper No. 5, UCLA Program in and . 1 This kind of rank-orderingis not entirelyan analyst's invention, as is shown by the following section of a Britisharmy memo of I903 dealing with Britishand Russian railroad constructionnear the Persia-Afghanistanborder: The conditions of the problem may . . . be brieflysummarized as follows: a) If we make a railway to Seistan while Russia remains inactive, we gain a considerable defensive advantage at considerable financial cost; b) If Russia makes a railway to Seistan, while we remain inactive, she gains a considerable offensiveadvantage at considerable financial cost; c) If both we and Russia make railways to Seistan, the defensive and advantages may be held to neutralize each other; in other words, we shall have spent a good deal of money and be no betteroff than we are at present. On the

0043-887I/78/3002-0 I 67$02-40/I ? I978 Princeton UniversityPress For copying information,see Contributorpage 168 WORLD POLITICS Unlesseach personthinks that the others will cooperate,he himself willnot. And why might he fearthat any other person would do some- thingthat would sacrifice his own firstchoice? The othermight not understandthe situation, or mightnot be ableto controlhis impulses ifhe sawa rabbit,or might fear that some other member of the group is unreliable.If theperson voices any of thesesuspicions, others are morelikely to fearthat he willdefect, thus making them more likely to defect,thus making it morerational for him to defect.Of coursein thissimple case-and in manythat are morerealistic-there are a numberof arrangementsthat could permit cooperation. But the main pointremains: although actors may know that they seek a common goal,they may not be ableto reachit. Evenwhen there is a solutionthat is everyone'sfirst choice, the in- ternationalcase is characterizedby three difficulties not present in the StagHunt. First, to theincentives to defectgiven above must be added thepotent fear that even if the other now supports the staus quo, itmay become dissatisfied later. No matterhow much decision makers arecommitted to thestatus quo, they cannot bind themselves and their successorsto thesame path. Minds can be changed,new leaderscan cometo power,values can shift,new opportunitiesand dangerscan arise. The secondproblem arises from a possiblesolution. In orderto pro- tecttheir possessions, states often seek to controlresources or land outsidetheir own territory.Countries that are notself-sufficient must tryto assurethat the necessary supplies will continueto flowin war- time.This was part of the explanation for Japan's drive into China and SoutheastAsia beforeWorld War II. If therewere an international authoritythat could guarantee access, this motive for control would disappear.But sincethere is not,even a statethat would prefer the statusquo toincreasing its area of control may pursue the latter policy. Whenthere are believed to be tightlinkages between domestic and foreignpolicy or betweenthe domestic politics of twostates, the quest forsecurity may drive states to interferepreemptively in thedomestic politicsof others in orderto provide an ideologicalbuffer zone. Thus,

other hand, we shall be no worse off,whereas under alternative (b) we shall be much worse off. Consequently, the theoreticalbalance of advantage lies with the proposed railway extension from Quetta to Seistan. W. G. Nicholson, "Memorandum on Seistan and Other Points Raised in the Discussion on the Defence of India," (Committee of Imperial Defence, March 20, I903). It should be noted that the possibilityof neitherside building railways was not mentioned, thus strongly biasing the analysis. SECURITY DILEMMA 169 Metternich'sjustification for supervising the politics of theItalian states has been summarizedas follows: Everystate is absolutelysovereign in itsinternal affairs. But this implies thatevery state must do nothingto interferein theinternal affairs of anyother. However, any false or perniciousstep taken by anystate in its internalaffairs may disturbthe reposeof anotherstate, and this consequentdisturbance of anotherstate's repose constitutes an inter- ferencein thatstate's internal affairs. Therefore, every state-or rather, everysovereign of a greatpower-has the duty,in the nameof the sacredright of independenceof everystate, to supervisethe govern- mentsof smallerstates and to preventthem from taking false and per- nicioussteps in theirinternal affairs.2 Morefrequently, the concern is withdirect attack. In orderto protect themselves,states seek to control,or at leastto neutralize,areas on their borders.But attemptsto establishbuffer zones can alarm otherswho have stakesthere, who fearthat undesirable precedents will be set,or who believethat their own vulnerabilitywill be increased.When buff- ers are are soughtin areas emptyof greatpowers, expansion tends to feedon itselfin orderto protectwhat is acquired,as was oftennoted by thosewho opposedcolonial expansion. Balfour's complaint was typical: "Everytime I come to a discussion-atintervals of, say,five years-I findthere is a new spherewhich we have got to guard,which is sup- posed to protectthe gatewaysof India. Those gatewaysare getting furtherand furtheraway from India, and I do notknow how farwest theyare goingto be broughtby theGeneral ."3 Thoughthis process is mostclearly visible when it involvesterritorial expansion,it oftenoperates with the increaseof less tangiblepower and influence.The expansionof powerusually brings with it an expan- sion of responsibilitiesand commitments;to meet them,still greater poweris required.The statewill takemany positions that are subjectto challenge.It will be involvedwith a wide rangeof controversialissues unrelatedto itscore values. And retreatsthat would be seen as normal if made by a smallpower would be takenas an indexof weaknessin- vitingpredation if made by a largeone. The thirdproblem present in internationalpolitics but not in the Stag Hunt is thesecurity dilemma: many of themeans by which a state triesto increaseits security decrease the securityof others.In domestic

2 Paul Schroeder,Metternich's Diplomacy at Its Zenith, i82o-i823 (Westport, Conn.: Greenwood Press I969), I26. 3 Quoted in Michael Howard, The Continental Commitment (Harmondsworth, Eng- land: Penguin I974), 67. 170 WORLD POLITICS society,there are several ways to increasethe safety of one's person and propertywithout endangering others. One can moveto a saferneigh- borhood,put barson the windows,avoid dark streets, and keep a distancefrom suspicious-looking characters. Of coursethese measures arenot convenient, cheap, or certainof success.But no one savecrim- inalsneed be alarmedif a persontakes them. In internationalpolitics, however,one state'sgain in securityoften inadvertently threatens others.In explainingBritish policy on navaldisarmament in the inter- war periodto the Japanese,Ramsey MacDonald said that"Nobody wantedJapan to be insecure."4But the problem was notwith British desires,but with the consequences of her policy. In earlierperiods, too, Britainhad neededa navylarge enough to keepthe shipping lanes open.But such a navycould not avoid being a menaceto any other state witha coastthat could be raided,trade that could be interdicted,or coloniesthat could be isolated.When Germanystarted building a powerfulnavy before I, Britainobjected that it couldonly be an offensiveweapon aimed at her. As SirEdward Grey, the Foreign Secretary,put it to King EdwardVII: "If theGerman Fleet ever be- comessuperior to ours,the German Army can conquerthis country. Thereis no correspondingrisk of this kind to Germany;for however superiorour Fleet was, no navalvictory could bring us anynearer to Berlin."The Englishposition was halfcorrect: Germany's navy was an anti-Britishinstrument. But the British often overlooked what the Germansknew full well: "in everyquarrel with England, German coloniesand tradewere . . . hostagesfor England to take."Thus, whethershe intendedit or not,the British Navy constituted an im- portantinstrument of coercion.5

II. WHAT MAKES COOPERATION MORE LIKELY? Giventhis gloomy picture, the obvious question is, why are we notall dead?Or, to put it lessstarkly, what kinds of variables ameliorate the impactof anarchy and the security dilemma? The workingsof several

4Quoted in Gerald Wheeler, Prelude to Pearl Harbor (Columbia: University of Missouri Press i963), i67. 5Quoted in Leonard Wainstein, "The Dreadnought Gap," in Robert Art and Ken- neth Waltz, eds., The Use of Force (Boston: Little, Brown I971), I55; Raymond Sontag, European Diplomatic History, I871-1932 (New York: Appleton-Century-Croits1933), 147. The French had made a similar argument 50 years earlier; see James Phinney Baxter III, The Introductionof the Ironclad Warship (Cambridge: Press 1933), I49. For a more detailed discussion of the securitydilemma, see Jervis, Perceptionand Misperceptionin InternationalPolitics (Princeton: Princeton University Press I976), 62-76. SECURITY DILEMMA 171 can be seenin termsof theStag Hunt or repeatedplays of thePrisoner's Dilemma.The Prisoner'sDilemma differsfrom the Stag Hunt in that thereis no solutionthat is in the bestinterests of all the participants; thereare offensiveas well as defensiveincentives to defectfrom the coalitionwith the others;and, if the game is to be playedonly once, the onlyrational response is to defect.But if the game is repeatedin- definitely,the latter characteristic no longerholds and we can analyze thegame in termssimilar to thoseapplied to the Stag Hunt. It would be in theinterest of each actorto have othersdeprived of thepower to defect;each would be willing to sacrificethis abilityif otherswere similarlyrestrained. But if the othersare not,then it is in the actor's interestto retainthe power to defect.'The game theorymatrices for thesetwo situationsare given below, with the numbersin the boxes beingthe order of theactors' preferences.

STAG HUNT PRISONER'S DILEMMA

COOPERATE DEFECT COOPERATE DEFECT A A COOPERATE COOPERATE

B1 43 4BEE34 DEFECT DEFECT43

We can see the logical possibilitiesby rephrasingour question: "Giveneither of theabove situations, what makes it moreor less likely thatthe playerswill cooperateand arriveat CC?" The chancesof achievingthis will be increasedby: (i) anythingthat increases incentivesto cooperateby increasingthe gains of mutualcooperation (CC) and/ordecreasing the coststhe actorwill pay if he cooperates and theother does not (CD); (2) anythingthat decreases the incentives fordefecting by decreasingthe gains of takingadvantage of the other (DC) and/orincreasing the costsof mutual noncooperation(DD); (3) anythingthat increases each side's expectationthat the otherwill cooperate!7

6 Experimentalevidence for this proposition is summarizedin JamesTedeschi, Barry Schlenker,and ThomasBonoma, Conflict, Power, and Games (Chicago: Aldine r973), '35-4'. 7The resultsof Prisoner'sDilemma games played in thelaboratory support this argu- ment.See Anatol Rapoportand AlbertChammah, Prisoner's Dilemma (Ann Arbor: Universityof MichiganPress i965), 33-50. Also see RobertAxelrod, Conflict of Interest (Chicago:Markham I970), 60-70. 172 WORLD POLITICS

THE COSTS OF BEING EXPLOITED (CD) The fearof beingexploited (that is, the costof CD) moststrongly drivesthe security dilemma; one of themain reasons why international lifeis notmore nasty, brutish, and shortis thatstates are not as vulner- able as menare in a stateof nature. People are easyto kill,but as Adam Smithreplied to a friendwho fearedthat the Napoleonic Wars would ruinEngland, "Sir, there is a greatdeal ofruin in a nation."8The easier it is to destroya state,the greater the reason for it eitherto join a larger and more secureunit, or else to be especiallysuspicious of others,to requirea large army,and, if conditionsare favorable,to attackat the slightestprovocation rather than wait to be attacked.If the failureto eat thatday-be itvenison or rabbit-meansthat he will starve,a person is likelyto defectin the Stag Hunt even if he reallylikes venison and has a highlevel of trustin his colleagues.(Defection is especiallylikely if theothers are also starvingor if theyknow thathe is.) By contrast, ifthe costs of CD are lower,if people are well-fedor statesare resilient, theycan affordto takea morerelaxed view of threats. A relativelylow costof CD has the effectof transformingthe game fromone in whichboth players make theirchoices simultaneously to one in whichan actorcan make his choiceafter the otherhas moved. He will nothave to defectout of fearthat the other will, but can wait to see what theother will do. Statesthat can affordto be cheatedin a bargainor thatcannot be destroyedby a surpriseattack can moreeasily trustothers and need not act at the first,and ambiguous,sign of menace.Because they have a marginof timeand error,they need not match,or morethan match, any others'arms in peacetime.They can mobilizein theprewar period or evenat thestart of thewar itself,and stillsurvive. For example,those who opposeda crashprogram to de- velopthe H-bomb felt that the U.S. marginof safetywas largeenough so thateven if Russia managed to gain a lead in the race, America would not be endangered.The program'sadvocates disagreed: "If we letthe Russians get the super first, catastrophe becomes all butcertain."' When the costsof CD are tolerable,not only is securityeasier to attainbut, what is evenmore imporant here, the relatively low level of armsand relativelypassive foreign policy that a status-quopower will be able to adopt are less likelyto threatenothers. Thus it is easierfor

8 Quoted in Bernard Brodie, Strategyin the Missile Age (Princeton: Princeton Uni- versityPress 1959), 6. 9 Herbert York, The Advisors. Oppenh/7einer, Teller, and the Superboinb (San Fran- cisco: Freeman I976), 56-60. SECURITY DILEMMA 173 status-quostates to act on theircommon interests if theyare hardto conquer.All otherthings being equal, a worldof small states will feel theeffects of anarchy much more than a worldof large ones. Defensi- ble borders,large size, and protectionagainst sudden attack not only aid thestate, but facilitate cooperation that can benefitall states. Of course,if one stategains invulnerability bybeing more powerful thanmost others, the problem will remain because its security provides a basefrom which it can exploit others. When the price a statewill pay forDD is low,it leaves others with few hostages for its good behavior. Otherswho are morevulnerable will growapprehensive, which will leadthem to acquire more arms and will reduce the chances of cooper- ation.The bestsituation is one in whicha statewill not suffer greatly ifothers exploit it, for example, by cheating on an armscontrol agree- ment(that is, the costs of CD arelow); butit will pay a highlong-run priceif cooperationwith the othersbreaks down-for example, if agreementscease functioning orif there is a longwar (thatis, the costs ofDD arehigh). The state'sinvulnerability is then mostly passive; it providessome protection, but it cannotbe usedto menaceothers. As we will discussbelow, this situation is approximatedwhen it is easier forstates to defendthemselves than to attackothers, or whenmutual deterrenceobtains because neither side can protectitself. The differencesbetween highly vulnerable and lessvulnerable states are illustratedby the contrasting policies of Britainand Austriaafter theNapoleonic Wars. Britain's geographic isolation and political stabil- ityallowed her to takea fairlyrelaxed view of disturbanceson the Continent.Minor wars and small changes in territoryorin thedistribu- tionof power did notaffect her vital interests. An adversarywho was outto overthrowthe system could be stoppedafter he had madehis intentionsclear. And revolutionswithin other states were no menace, sincethey would not set off unrest within England. Austria, surrounded by strongpowers, was not so fortunate;her policyhad to be more closelyattuned to all conflicts.By the time an aggressor-statehadclearly shownits colors,Austria would be gravelythreatened. And foreign revolutions,be theydemocratic or nationalistic,would encourage groupsin Austriato upset the existing order. So itis notsurprising that Metternichpropounded the doctrinesummarized earlier, which de- fendedAustria's right to interferein theinternal affairs of others, and thatBritish leaders rejected this view. Similarly, Austria wanted the Congresssystem to be a relativelytight one, regulating most disputes. The Britishfavored a lesscentralized system. In otherwords, in order 174 WORLD POLITICS to protectherself, Austria had eitherto threatenor to harmothers, whereasBritain did not.For Austriaand herneighbors the security dilemmawas acute;for Britain it was not. The ultimatecost of CD is ofcourse loss of . This cost can varyfrom situation to situation.The lowerit is (forinstance, because thetwo states have compatible ideologies, are similarethnically, have a commonculture, or becausethe citizens of thelosing state expect economicbenefits), the less the impact of thesecurity dilemma; the greaterthe costs, the greater the impact of the dilemma. Here is another reasonwhy extreme differences in valuesand ideologiesexacerbate internationalconflict. It is throughthe lowering of the costsof CD thatthe proposed Rhodesian"safety net"-guaranteeing that whites who leave the coun- trywill receive fair payment for their property-would have the para- doxicaleffect of making it more likely that the whites will stay. This is lesspuzzling when we seethat the whites are in a multi-personPrison- er'sDilemma with each other. Assume that all whitesare willingto stayif mostof theothers stay; but, in the absenceof guarantees,if thereis goingto be a massexodus, all wantto be amongthe first to leave(because late-leavers will get less for their property and will have moretrouble finding a countryto takethem in). Thenthe problem is toavoid a self-fulfillingprophecy in whicheach person rushes to defect becausehe fearsothers are going to. In narrowingthe gap betweenthe payofffor leaving first (DC) and leavinglast (CD) by reducingthe costof the latter, the guarantees make it easierfor the whites to coop- erateamong themselves and stay. SubjectiveSecurity Demands. Decision makers act in termsof the vulnerabilitythey feel, which can differfrom the actual situation; we musttherefore examine the decisionmakers' subjective security re- quirements.10Two dimensionsare involved.First, even if theyagree aboutthe objective situation, people can differabout how much secur- itythey desire-or, to putit moreprecisely, about the price they are willingto pay to gain incrementsof security.The morestates value theirsecurity above all else(that is, see a prohibitivelyhigh cost in CD), themore they are likelyto be sensitiveto evenminimal threats, and to demandhigh levels of arms.And if armsare positively valued be- 10 For thedevelopment of theconcept of subjectivesecurity, see ArnoldWolfers, Dis- cordand Collaboration(Baltimore: Johns Hopkins Press i962), chap. io. In thepresent sectionwe assumethat the state believes that its security can be bestserved by increasing itsarms; later we will discusssome of theconditions under which this assumption does nothold. SECURITY DILEMMA 175 causeof pressuresfrom a -industrialcomplex, it will be espe- ciallyhard for status-quo powers to cooperate.By contrast, the security dilemmawill not operate as stronglywhen pressing domestic concerns increasethe opportunity costs of armaments.In thiscase, the net ad- vantageof exploiting the other (DC) willbe less,and the costs of arms races(that is, one aspectof DD) will be greater;therefore the state willbehave as thoughit were relatively invulnerable. The secondaspect of subjectivesecurity is theperception of threat (thatis, theestimate of whetherthe other will cooperate)."A state thatis predisposedto seeeither a specificother state as an adversary,or othersin generalas a menace,will reactmore strongly and more quicklythan a statethat sees its environment as benign. Indeed, when a statebelieves that another not only is notlikely to be an adversary, buthas sufficient interests in commonwith it to be an ally,then it will actuallywelcome an increasein theother's power. Britishand Frenchforeign policies in theinterwar years illustrate thesepoints. After the rise of Hitler,Britain and Francefelt that in- creasesin eachother's arms increased rather than decreased their own security.The differingpolicies that these states followed toward Ger- manycan be explainedby their differences on both dimensions of the variableof subjectivesecurity.12 Throughout the period, France per- ceivedGermany as moreof a threatthan England did. The British weremore optimistic and argued that conciliation could turn Germany intoa supporterof thestatus quo. Furthermore,in the years immedi- atelyfollowing World War I, Francehad beenmore willing to forego othervalues in orderto increase her security and had therefore followed a morebelligerent policy than England, maintaining a larger army and movingquickly to counterGerman assertiveness. As thisexample shows,one cannot easily say how much subjective security a state should seek.High securityrequirements make it verydifficult to capitalize on a commoninterest and runthe danger of settingoff spirals of arms racesand hostility. The Frenchmay have paid this price in theI920'S. Low securityrequirements avoid this trap, but run the risk of having toofew arms and oftrying to conciliatean aggressor. One aspectof subjective security related to thepredisposition to per-

11The questionof whenan actorwill see anotheras a threatis importantand under- studied.For a valuable treatment(although one marredby seriousmethodological flaws),see RaymondCohen, "Threat Perceptionin InternationalRelations," Ph.D. diss. (Hebrew University1974). Amongthe importantfactors, touched on below,are the lessonsfrom the previouswar. 12 Stillthe best treatment is ArnoldWolfers, Britain and France Between Two Wars (New York: Harcourt,Brace I940). 176 WORLD POLITICS ceivethreat is thestate's view of how many enemies it must be prepared to fight.A statecan be relaxedabout increases in another'sarms if it believesthat there is a functioningcollective security system. The chancesof peace are increased in a worldin whichthe prevailing inter- nationalsystem is valuedin itsown right, not only because most states restraintheir ambitions and thosewho do notare deterred(these are theusual claims for a Concertsystem), but also because of the decreased chancesthat the status-quo states will engagein unnecessaryconflict out of thequest for security. Indeed, if therewere complete faith in collectivesecurity, no statewould want an army.By contrast,the securitydilemma is insolublewhen each state fears that many others, farfrom coming to itsaid, are likelyto join in anyattack. Winston Churchill,as FirstLord of the Admiralty, was settinga highsecurity requirementwhen he noted: Besidesthe Great Powers, there are many small states who are buying or buildinggreat ships of war and whose vessels may by purchase, by somediplomatic combination, or by duress, be broughtinto the line againstus. None of these powers need, like us, navies to defendtheir actualsafety of independence. They build them so as toplay a partin worldaffairs. It is sportto them. It is deathto us.13 It takesgreat effort for any one stateto be ableto protectitself alone againstan attackby several neighbors. More importantly, it is nextto impossiblefor all statesin thesystem to havethis capability. Thus, a state'sexpectation that allies will be availableand that only a fewothers willbe ableto join against it is almosta necessarycondition for security requirementsto be compatible.

GAINS FROM COOPERATION AND COSTS OF A BREAKDOWN (CC AND DD) The maincosts of a policyof reactingquickly and severelyto in- creasesin theother's arms are notthe price of one'sown arms,but ratherthe sacrifice of thepotential gains from cooperation (CC) and theincrease in thedangers of needless arms races and wars (DD). The greaterthese costs, the greater the incentives totry cooperation and wait forfairly unambiguous evidence before assuming that the other must be checkedby force. Wars would be muchmore frequent-even if the firstchoice of all stateswas thestatus quo-if theywere less risky and costly,and if peaceful intercourse did not provide rich benefits. Ethiopia recentlyasked for guarantees that the Territoryof Afarsand Issas wouldnot join a hostilealliance against it whenit gainedindepend-

13 Quoted in Peter Gretton,Former Naval Person (London: Cassell i968), I5I. SECURITY DILEMMA 177 ence.A spokesmanfor the Territory replied that this was notnecessary: Ethiopia"already had the bestpossible guarantee in the railroad"that links the two countriesand providesindispensable revenue for the Territory."4 The basicpoints are well knownand so we can moveto elaboration. First,most statesmen know thatto entera war is to set offa chain of unpredictableand uncontrollableevents. Even if everythingthey see pointsto a quick victory,they are likelyto hesitatebefore all theuncer- tainties.And if the battlefieldoften produces startling results, so do the councilchambers. The statemay be desertedby allies or attacked by neutrals.Or the postwaralignment may rob it of the fruitsof victory,as happenedto Japanin i895. Second,the domesticcosts of wars mustbe weighed.Even strongstates can be underminedby dis- satisfactionwith the way thewar is run and by thenecessary mobiliza- tionof men and ideas. Memoriesof such disruptionswere one of the mainreasons for the era of relativepeace thatfollowed the Napoleonic Wars.Liberal statesmen feared that large armies would lead to despot- ism; conservativeleaders feared that wars would lead to revolution. (The otherside of this coin is that when thereare domesticconse- quences of foreignconflict that are positivelyvalued, the net cost of conflictis loweredand cooperationbecomes more difficult.)Third- turningto the advantagesof cooperation-forstates with large and diverseeconomies the gains from economic exchange are rarelyif ever sufficientto preventwar. Norman Angell was wrong about World War I being impossiblebecause of economicties among the powers; and beforeWorld War II, theU.S. was Japan'smost important trading partner.Fourth, the gains from cooperation can be increased,not only if each side getsmore of thetraditional values such as wealth,but also if each comesto value theother's well-being positively. Mutual cooper- ationwill thenhave a double payoff:in additionto the directgains, therewill be the satisfactionof seeingthe otherprosper.15 While high costsof war and gainsfrom cooperation will ameliorate the impactof the securitydilemma, they can createa differentprob- lem.If thecosts are highenough so thatDD is thelast choice for both sides,the game will shiftto "Chicken."This game differsfrom the Stag Hunt in thateach actorseeks to exploitthe other;it differsfrom Prisoner'sDilemma in thatboth actorsshare an interestin avoiding

14 Michael Kaufman, "Tension Increases in French Colony," New York Times, July II, I976. 15 Experimental support for this argument is summarized in Morton Deutsch, The Resolutionof Conflict(New Haven: Yale University Press I973), 8-95. 178 WORLD POLITICS mutualnon-cooperation. In Chicken, if you thinkthe otherside is goingto defect, you have to cooperate because, although being exploited (CD) is bad, it is not as bad as a totalbreakdown (DD). As the familiarlogic of deterrence shows, the actor must then try to convince hisadversary that he is goingto standfirm (defect) and thatthe only waythe other can avoiddisaster is toback down (cooperate). Commit- ment,the rationality ofirrationality, manipulating the communications system,and pretending not to understandthe situation, are among the tacticsused to reach this goal. The samelogic applies when both sides areenjoying great benefits from cooperation. The sidethat can credibly threatento disruptthe relationshipunless its demandsare metcan exploitthe other. This situation may not be stable,since the frequent use ofthreats may be incompatiblewith the maintenance of a cooperative relationship.Still, de Gaulle'ssuccessful threats to breakup theCom- monMarket unless his partners acceded to hiswishes remind us that theshared benefits of cooperation as well as theshared costs of defection can providethe basisfor exploitation. Similarly, one reasonfor the collapseof the Franco-British entente more than a hundredyears earlier was thatdecision makers on bothsides felt confident that their own countrycould safely pursue a policythat was againstthe other's inter- est becausethe othercould not affordto destroythe highlyvalued relationship."Because statesmen realize that the growthof positive interdependencecan provideothers with new leversof influenceover them,they may resist such developments more than would be expected fromthe theories that stress the advantages of cooperation.

GAINS FROM EXPLOITATION (DC) Defectingnot only avoids the danger that a statewill be exploited (CD), butbrings positive advantages by exploitingthe other (DC) The lowerthese possible gains, the greater the chances of cooperation. Evena relativelysatisfied state can be temptedto expandby the hope ofgaining major values. The temptationwill be lesswhen the state sees otherways of reachingits goals, and/or places a low valueon what exploitationcould bring.The gainsmay be low eitherbecause the immediateadvantage provided by DC (forexample, having more arms thanthe other side) cannot be translatedinto a politicaladvantage (for example,gains in territory),or becausethe political advantage itself

16 Roger Bullen, Palmerston, Guizot, and the Collapse of the Entente Cordiale (London: Athlone Press I974), 8i, 88, 93, 2I2. For a differentview of this case, see Stanley Mellon, "Entente, Diplomacy, and Fantasy," Reviews in European History, ii (September 1976), 376-80. SECURITY DILEMMA 179 is not highlyvalued. For instance,a statemay not seek to annex addi- tionalterritory because the latterlacks raw materials,is inhabitedby peopleof a differentethnic group, would be costlyto garrison,or would be hard to assimilatewithout disturbing domestic politics and values. A statecan reducethe incentivesthat another state has to attackit, by notbeing a threatto thelatter and byproviding goods and servicesthat would be lostwere the otherto attemptexploitation. Even wherethe directadvantages of DC are great,other considera- tionscan reducethe net gain. Victoryas well as defeatcan set offun- desireddomestic changes within the state.Exploitation has at times beenfrowned upon by theinternational community, thus reducing the prestigeof a statethat engages in it. Or othersmight in the futurebe quickerto see thestate as a menaceto them,making them more likely to arm,and to opposeit later.Thus, Bismarck'sattempts to get other powersto cooperatewith him in maintainingthe statusquo afteri87I weremade moredifficult by thewidely-held mistrust of him thatgrew out of his earlieraggressions.17

THE PROBABILITY THAT THE OTHER WILL COOPERATE The variablesdiscussed so far influencethe payoffsfor each of the fourpossible outcomes. To decidewhat to do, thestate has to go further and calculatethe expectedvalue of cooperatingor defecting.Because such calculationsinvolve estimating the probability that the otherwill cooperate,the statewill have to judge how the variablesdiscussed so faract on the other.To encouragethe otherto cooperate,a statemay tryto manipulatethese variables. It can lowerthe other'sincentives to defectby decreasingwhat it could gain by exploitingthe state (DC)- the detailswould be similarto thosediscussed in the previouspara- graph-and it can raise the costsof (DD). But if the state cannotmake DD theworst outcome for the other,coercion is likelyto be ineffectivein theshort run because the other can respondby refusing to cooperate,and dangerousin thelong run becausethe otheris likely to becomeconvinced that the stateis aggressive.So the statewill have to concentrateon makingcooperation more attractive. One way to do thisis to decreasethe coststhe otherwill pay if it cooperatesand the statedefects (CD). Thus, the statecould tryto make the otherless vulnerable.It was forthis reason that in thelate 1950's and earlyi960's

17Similarly, a French diplomat has argued that "the worst result of Louis XIV's abandonment of our traditionalpolicy was the distrustit aroused towards us abroad." Jules Cambon, "The Permanent Bases of French Foreign Policy," Foreign Aflairs, viii (JanuaryI930), I79. 180 WORLD POLITICS someAmerican defense analysts argued that it would be good forboth sidesif the Russians developed hardened missiles. Of course,decreasing theother's vulnerability also decreasesthe state's ability to coerceit, and opens the possibilitythat the otherwill use thisprotection as a shield behindwhich to engagein actionsinimical to thestate. But by sacrific- ing someability to harmthe other, the state can increasethe chances of mutuallybeneficial cooperation. The statecan also tryto increasethe gains thatwill accrueto the otherfrom mutual cooperation (CC). Althoughthe state will of course gain if it receivesa shareof any new benefits,even an incrementthat accruesentirely to the otherwill aid the stateby increasingthe likeli- hood thatthe other will cooperate.18 This line of argumentcan be continuedthrough the infiniteregres- sions that game theoryhas made familiar.If the otheris ready to cooperatewhen it thinksthe state will, the state can increasethe chances of CC byshowing that it is planningto cooperate.Thus thestate should understatethe gainsit would make if it exploitedthe other(DC) and the costsit would pay if the otherexploited it (CD), and stressor exaggeratethe gains it would make under mutualcooperation (CC) and the costsit would pay if thereis deadlock (DD). The statewill also want to convincethe otherthat it thinksthat the otheris likely to cooperate.If the otherbelieves these things, it will see thatthe state has strongincentives to cooperate,and so it will cooperatein turn.One pointshould be emphasized.Because the other,like the state,may be drivento defectby the fearthat it will be exploitedif it does not,the stateshould tryto reassureit thatthis will not happen.Thus, when Khrushchevindicated his willingnessto withdrawhis missilesfrom Cuba, he simultaneouslystressed to Kennedythat "we are of sound mind and understandperfectly well" thatRussia could not launch a successfulattack against the U.S., and thereforethat there was no rea- son for the U.S. to contemplatea defensive,pre-emptive strike of its own.19 There is, however,a danger.If the otherthinks that the statehas littlechoice but to cooperate,it can crediblythreaten to defectunless the state providesit with additionalbenefits. Great advantagesof mutualcooperation, like high costsof war, providea leverfor com-

18 This assumes, however, that these benefitsto the other will not so improve the other's power position that it will be more able to menace the state in the future. 19 Walter LaFeber, ed., The Dynamics of World Power; A Documentary History of United States Foreign Policy I945-I973, II: Eastern Europe and the Soviet Union (New York: Chelsea House in association with McGraw-Hill I973), 700. SECURITY DILEMMA 181 petitivebargaining. Furthermore, for a stateto stresshow muchit gains fromcooperation may be to implythat it is gainingmuch morethan the otherand to suggestthat the benefitsshould be distributedmore equitably. When each sideis readyto cooperateif it expectsthe other to, inspec- tion devicescan amelioratethe securitydilemma. Of course,even a perfectinspection system cannot guarantee that the other will not later developaggressive intentions and the militarymeans to act on them. But by relievingimmediate worries and providingwarning of com- ing dangers,inspection can meeta significantpart of the feltneed to protectoneself against future threats, and so make currentcooperation more feasible.Similar functions are servedby breakingup one large transactioninto a seriesof smallerones.20 At each transactioneach can see whetherthe other has cooperated;and itslosses, if the other defects, will be small.And sincewhat eitherside would gain by one defection is slightcompared to the benefitsof continuedcooperation, the pros- pects of cooperationare high. Conflictsand wars among status-quo powerswould be much more commonwere it not for the factthat internationalpolitics is usuallya seriesof small transactions. How a statesmaninterprets the other'spast behaviorand how he projectsit into the futureis influencedby his understandingof the securitydilemma and his abilityto place himselfin the other'sshoes. The dilemmawill operatemuch more stronglyif statesmendo not understandit, and do notsee thattheir arms-sought only to securethe statusquo-may alarm othersand thatothers may arm,not because theyare contemplatingaggression, but because they fear attack from the firststate. These two failuresof empathyare linked.A statewhich thinksthat the other knows that it wants only to preservethe status quo and thatits arms are meant only for self-preservationwill conclude thatthe other side will reactto itsarms by increasingits own capability onlyif it is aggressiveitself. Since the otherside is not menaced,there is no legitimatereason for it to objectto thefirst state's arms; therefore, objectionproves that the otheris aggressive.Thus, the followingex- changebetween Senator Tom Connallyand Secretaryof StateAcheson concerningthe ratificationof the NATO treaty: SecretaryAcheson: [The treaty]is aimedsolely at armedaggression. SenatorConnally: In otherwords, unless a nation. . . contemplates, meditates,or makesplans looking toward aggression or armedattack on anothernation, it has no causeto fearthis treaty. 20 ThomasSchelling, The Strategyof Conflict(New York: OxfordUniversity Press i963), I34-35- 182 WORLD POLITICS SecretaryAcheson: That is correct,Senator Connally, and it seemsto me thatany nation which claims that this treaty is directedagainst it shouldbe remindedof the Biblicaladmonition that 'The guiltyflee whenno manpursueth.' SenatorConnally: That is a veryapt illustration. WhatI had in mindwas, when a Stateor Nationpasses a criminal act,for instance, against burglary, nobody but thosewho are burglars or gettingready to be burglarsneed have any fear of theBurglary Act. Is thatnot true? SecretaryAcheson: The onlyeffect [the law] would have [on an innocentperson] would be forhis protection,perhaps, by deterring someoneelse. He wouldn'tworry about the imposition of thepenalties on himself.21 The otherside of thiscoin is thatpart of theexplanation for detente is thatmost American decision makers now realizethat it is at leastpos- siblethat Russia may fearAmerican aggression; many think that this fearaccounts for a rangeof Sovietactions previously seen as indicating Russian aggressiveness.Indeed, even 36 percentof militaryofficers considerthe SovietUnion's motivations to be primarilydefensive. Less thantwenty years earlier, officers had beendivided over whether Russia soughtworld or onlyexpansion.22 Statesmenwho do not understandthe securitydilemma will think thatthe moneyspent is the onlycost of buildingup theirarms. This beliefremoves one importantrestraint on armsspending. Furthermore, it is also likelyto lead statesto settheir security requirements too high. Since theydo not understandthat trying to increaseone's securitycan actuallydecrease it, theywill overestimatethe amountof securitythat is attainable;they will thinkthat when in doubtthey can "playit safe" by increasingtheir arms. Thus it is verylikely that two stateswhich supportthe status quo butdo notunderstand the security dilemma will end up, if notin a war,then at leastin a relationshipof higherconflict thanis requiredby the objectivesituation. The beliefthat an increasein militarystrength always leads to an increasein securityis oftenlinked to the beliefthat the only routeto securityis throughmilitary strength. As a consequence,a whole range of meliorativepolicies will be downgraded.Decision makerswho do not believe that adoptinga more conciliatoryposture, meeting the

21 U.S. Congress, Senate, Committee on Foreign Relations, Hearings, North Atlantic Treaty,8ist Cong., ist sess. (1949), I7. 22 Bruce Russett and Elizabeth Hanson, Interest and Ideology (San Francisco: Free- man I975), 260; Morris Janowitz, The Prolessional Soldier (New York: Free Press i960), chap. I3- SECURITY DILEMMA 183 other'slegitimate grievances, or developingmutual gains from cooper- ationcan increasetheir state's security, will not devote much attention oreffort to these possibilities. On theother hand, a heightenedsensitivity to the security dilemma makesit morelikely that the state will treat an aggressoras thoughit werean insecuredefender of thestatus quo. Partlybecause of their viewsabout the causes of World War I, theBritish were predisposed to believethat Hitler sought only the rectification of legitimate and lim- itedgrievances and that security could best be gainedby constructing an equitableinternational system. As a resultthey pursued a policywhich, althoughwell designedto avoidthe dangerof creatingunnecessary conflictwith a status-quoGermany, helped destroy Europe.

GEOGRAPHY, COMMITMENTS, BELIEFS, AND SECURITY THROUGH EXPANSION A finalconsideration does not easily fit in thematrix we havebeen using,although it can be seenas an aspectof vulnerability and of the costsof CD. Situationsvary in theease or difficultywith which all statescan simultaneouslyachieve a highdegree of security.The influ- enceof militarytechnology on thisvariable is thesubject of thenext section.Here we wantto treatthe impact of beliefs,geography, and commitments(many of whichcan be consideredto be modifications ofgeography, since they bind states to defend areas outside their home- lands).In thecrowded continent of Europe, security requirements were hardto mesh.Being surrounded by powerful states, Germany's prob- lem-or theproblem created by Germany-was always great and was evenworse when her relations with both France and Russiawere bad, suchas beforeWorld War I. In thatcase, even a status-quoGermany, ifshe could not change the political situation, would almost have been forcedto adoptsomething like the Schlieffen Plan. Becauseshe could nothold off both of her enemies, she had to be preparedto defeatone quicklyand thendeal withthe other in a moreleisurely fashion. If Franceor Russiastayed out of a warbetween the other state and Ger- many,they would allow Germany to dominatethe Continent (even if thatwas notGermany's aim). Theytherefore had to denyGermany thisability, thus making Germany less secure. Although Germany's arrogantand erratic behavior, coupled with the desire for an unreason- ablyhigh level of security (which amounted to the desire to escape from hergeographic plight), compounded the problem, even wise German statesmenwould have been hard put to gaina highdegree of security withoutalarming their neighbors. 184 WORLD POLITICS A similarsituation arose for France afterWorld War I. She was committedto protectingher allies in EasternEurope, a commitment she couldmeet only by takingthe offensive against Germany. But since therewas no way to guaranteethat France might not laterseek expan- sion,a Francethat could successfullylaunch an attackin responseto a Germanmove into Eastern Europe would constitutea potentialdanger to German values. Similarly,a United States crediblyable to threatenretaliation with strategic nuclear if the SovietUnion attacksWestern Europe also constitutesa menace, albeit a reducedone, to the Sovietability to maintainthe status quo. The incompatibilityof thesesecurity requirements is not complete.Herman Kahn is correct in arguingthat the UnitedStates could have Type II deterrence(the abilityto detera majorSoviet provocation) without gaining first-strike capabilitybecause the expected Soviet retaliation following an American strikecould be greatenough to deterthe U.S. fromattacking unless the U.S. believedit would sufferenormous deprivation (for instance, theloss of Europe) if it did not strike.23Similarly, the Franco-German militarybalance could have been such thatFrance could successfully attackGermany if the latter's armies were embroiled in EasternEurope, butcould notdefeat a Germanythat was freeto devoteall herresources to defendingherself. But thisdelicate balance is veryhard to achieve, especiallybecause statesusually calculate conservatively.Therefore, sucha solutionis notlikely to be available. For the United States,the problemposed by the need to protect Europe is an exception.Throughout most of its history,this country has been in a much more favorableposition: relatively self-sufficient and securefrom , it has not onlybeen able to get securityrela- tivelycheaply, but by doing so, did not menaceothers.24 But ambitions and commitmentshave changed this situation.After the American conquestof thePhilippines, "neither the United States nor Japan could assureprotection for their territories by military and naval meanswith- out compromisingthe defensesof the other.This problemwould plague Americanand Japanesestatesmen down to I94I."25 Further- more,to the extentthat Japancould protectherself, she could resist Americanthreats to go to war if Japandid not respectChina's inde-

23 Kahn, On Thermonuclear War (Princeton: Princeton UniversityPress i960), 138- 6o0 It should be noted that the French example is largely hypotheticalbecause France had no intentionof fulfillingher obligations once Germany became strong. 24Wolfers (fn. 9), chap. I5; C. Vann Woodward, "The Age of Reinterpretation," American Historical Review, Vol. 67 (October i960), i-i9. 25 William Braisted, The United States Navy in the Pacific, 1897-1909 (Austin: Uni- versityof Texas Press1958), 240. SECURITY DILEMMA 185 pendence.These complicationswere minor compared to thosethat followedWorld War II. A worldpower cannot help but have the abil- ityto harmmany others that is outof proportion to theothers' ability toharm it. Britainhad beenable to gainsecurity without menacing others to a greaterdegree than the Continental powers, though to a lesserone than theUnited States. But the acquisition of colonies and a dependenceon foreigntrade sacrificed her relative invulnerability of being an island. Once shetook India, she had to considerRussia as a neighbor;the latter was expandingin CentralAsia, thusmaking it muchmore difficult for bothcountries to feelsecure. The need to maintainreliable sea lanes to India meantthat no statecould be allowedto menaceSouth Africa and, later,Egypt. But theneed to protectthese two areasbrought new fears, new obligations,and new securityrequirements that conflictedwith thoseof otherEuropean nations. Furthermore, once Britainneeded a flowof importsduring both peace and wartime,she requireda navy thatcould preventa .A navysufficient for that task could not help but be a threatto any otherstate that had valuabletrade. A relatedproblem is raisedby thefact that defending the statusquo oftenmeans protectingmore than territory.Nonterritorial interests, norms,and the structureof the internationalsystem must be main- tained.If all status-quopowers agree on thesevalues and interpretthem in compatibleways, problems will be minimized.But thepotential for conflictis great,and the policiesfollowed are likelyto exacerbatethe securitydilemma. The greaterthe range of intereststhat have to be protected,the more likelyit is that nationalefforts to maintainthe statusquo will clash.As a Frenchspokesman put it in I930: "Security! The termsignifies more indeed than the maintenanceof a people's homeland,or evenof theirterritories beyond the seas. It also meansthe maintenanceof theworld's respect for them, the maintenanceof their economicinterests, everything in a word,which goes to make up the grandeur,the lifeitself, of the nation."" When securityis thoughtof in thissense, it almostautomatically has a competitiveconnotation. It involvesasserting one state'swill over others,showing a high degree ofleadership if not dominance, and displayinga pricklydemeanor. The resultingbehavior will almostsurely clash with that of otherswho definetheir security in the same way. The problemwill be almostinsoluble if statesmenbelieve that their securityrequires the threateningor attackingof others."That which

26 Cambon (fn. I7), I85. 186 WORLD POLITICS stops growingbegins to rot," declareda ministerto Catherinethe Great.27More commonis thebelief that if theother is secure,it will be emboldenedto act againstone's own state'sinterests, and thebelief that in a war it will not be enoughfor the stateto protectitself: it mustbe able to takethe war to theother's homeland. These convictionsmake it verydifficult for status-quo states to developcompatible security poli- cies,for theylead the stateto concludethat its securityrequires that othersbe renderedinsecure. In othercases, "A countryengaged in a war of defensemight be obligedfor strategic reasons to assumethe offensive," as a Frenchdele- gate to an interwardisarmament conference put it.28That was the case forFrance in I799: The Directory'spolitical objectives were essentially defensive, for the Frenchwanted only to protectthe Republic from invasion and preserve thesecurity and territoryof thesatellite regimes in Holland,Switzer- land,and Italy.French leaders sought no new ;they wanted onlyto preservethe earlier gains of theRevolution. The Directorybe- lieved,however, that only a militaryoffensive could enable the nation to achieveits defensive political objective. By inflictingrapid and decisive defeatsupon one or moremembers of the coalition, the directors hoped to ruptureallied unity and forceindividual powers to seeka separate peace.29 It did notmatter to thesurrounding states that France was not attack- ing becauseshe was greedy,but because she wantedto be leftin peace. Unlessthere was some way her neighborscould provideFrance with an alternateroute to her goal, Francehad to go to war.

III. OFFENSE, DEFENSE, AND THE SECURITYDILEMMA Anotherapproach starts with the centralpoint of the securitydi- lemma-that an increasein one state'ssecurity decreases the security of others-and examinesthe conditionsunder which thisproposition holds.Two crucialvariables are involved:whether defensive weapons and policiescan be distinguishedfrom offensive ones, and whetherthe

27 Quoted in Adam Ulam, Expansion and Co-Existence (New York: Praeger i968), 5. In I920 the U.S. Navy's General Board similarly declared "A nation must advance or retrocedein world position." Quoted in William Braisted, The United States Navy in the Pacific, 1909-1922 (Austin: Universityof Texas Press I971), 488. 28 Quoted in Marion Boggs, Attemptsto Define and Limit "Aggressive" Armament in Diplomacy and (Columbia: University of Missouri Studies, xvi, No. i, I94'), 4I. 29 Steven Ross, European Diplomatic History,1789-i815 (Garden City,N.Y.: Double- day i969), I94. SECURITY DILEMMA 187 defenseor the offense has the advantage. The definitionsare not always clear,and many cases are difficult tojudge, but these two variables shed a greatdeal of light on thequestion of whether status-quo powers will adoptcompatible security policies. All thevariables discussed so far leavethe heart of the problem untouched. But when defensive weapons differfrom offensive ones, it is possiblefor a stateto makeitself more securewithout making others less secure. And whenthe defense has theadvantage over the offense, a large increase in one state'ssecurity onlyslightly decreases the security of the others, and status-quo powers can all enjoya highlevel of security and largelyescape from the state ofnature.

OFFENSE-DEFENSE BALANCE Whenwe saythat the offense has the advantage, we simplymean that it is easierto destroythe other's army and takeits territory than it is to defendone's own. When the defense has theadvantage, it is easier toprotect and tohold than it is to moveforward, destroy, and take.If effectivedefenses can be erectedquickly, an attackermay be able to keepterritory he has takenin an initialvictory. Thus, the dominance ofthe defense made it veryhard for Britain and Franceto pushGer- manyout of France in WorldWar I. Butwhen superior defenses are difficultfor an aggressorto improviseon thebattlefield and mustbe constructedduring peacetime, they provide no directassistance to him. The securitydilemma is at its mostvicious when commitments, strategy,or technologydictate that the onlyroute to securitylies throughexpansion. Status-quo powers must then act like aggressors; thefact that they would gladly agree to foregothe opportunityfor expansionin returnfor guarantees for their security has no implica- tionsfor their behavior. Even if expansionis notsought as a goal in itself,there will be quickand drasticchanges in thedistribution of territoryand influence.Conversely, when the defense has theadvan- tage,status-quo states can make themselves more secure without gravely endangeringothers.30 Indeed, if thedefense has enoughof an advan- tageand if the states are of roughly equal size, not only will the security dilemmacease to inhibit status-quo states from cooperating, but aggres- sionwill be nextto impossible,thus rendering international anarchy relativelyunimportant. If states cannot conquer each other, then the

30 Thus, when Wolfers(fn. IO), I26, arguesthat a status-quostate that settles for roughequality of powerwith its adversary,rather than seeking preponderance, may be able to convincethe other to reciprocateby showingthat it wantsonly to protectitself, not menacethe other,he assumesthat the defensehas an advantage. 188 WORLD POLITICS lack of sovereignty,although it presentsproblems of collectivegoods in a numberof areas,no longerforces states to devotetheir primary attentionto self-preservation.Although, if force were not usable, there would be fewerrestraints on the use of nonmilitaryinstruments, theseare rarelypowerful enough to threatenthe vital interests of a majorstate. Two questionsof the offense-defense balance can be separated.First, doesthe state have to spendmore or lessthan one dollaron defensive forcesto offseteach dollar spent by the other side on forcesthat could be usedto attack? If thestate has one dollar to spendon increasingits security,should it put it into offensive ordefensive forces? Second, with a giveninventory of forces,is it betterto attackor to defend? Is there an incentiveto strikefirst or to absorbthe other's blow? These two aspectsare often linked: if each dollar spent on offensecan overcome eachdollar spent on defense,and if bothsides have the same defense budgets,then both are likely to build offensive forces and find it attrac- tiveto attackrather than to wait for the adversary to strike. Theseaspects affect the security dilemma in differentways. The first hasits greatest impact on armsraces. If thedefense has theadvantage, andif the status-quo powers have reasonable subjective security require- ments,they can probably avoid an armsrace. Although an increasein one side'sarms and securitywill stilldecrease the other's security, the former'sincrease will be largerthan the latter's decrease. So ifone side increasesits arms, the other can bring its security back up toits previous levelby addinga smalleramount to itsforces. And if thefirst side reactsto this change, its increase will also be smallerthan the stimulus thatproduced it. Thus a stableequilibrium will be reached.Shifting fromdynamics to statics,each side can be quitesecure with forces roughlyequal to those of the other. Indeed, if the defense is muchmore potentthan the offense, each side can be willingto haveforces much smallerthan the other's, and can be indifferentto a wide rangeof theother's defense policies. The secondaspect-whether it is betterto attackor to defend- influencesshort-run stability. When the offense has theadvantage, a state'sreaction to internationaltension will increasethe chancesof war.The incentivesfor pre-emption and the"reciprocal fear of sur- priseattack" in thissituation have been made clear by analyses of the dangersthat exist when two countrieshave first-strikecapabilities.31 Thereis no wayfor the state to increaseits security without menacing,

31Schelling (fn. 20), chap.9. SECURITY DILEMMA 189 oreven attacking, the other. Even Bismarck, who once called committingsuicide from fear of death,"said that"no govern- ment,if it regardswar as inevitableeven if it doesnot want it, would be so foolishas to leaveto theenemy the choice of timeand occasion andto wait for the moment which is mostconvenient for the enemy."32 In anotherarena, the same dilemma applies to the policeman in a dark alleyconfronting a suspected criminal who appearsto be holdinga .Though racism may indeed be present,the security dilemma can accountfor many of the tragic shootings of innocent people in the ghettos. Beliefsabout the course of a warin whichthe offense has the advan- tagefurther deepen the security dilemma. When there are incentives to strikefirst, a successful attack will usually so weakenthe other side thatvictory will be relativelyquick, bloodless, and decisive.It is in theseperiods when conquest is possibleand attractivethat states con- solidatepower internally-for instance, by destroying the feudal barons -and expandexternally. There are several consequences that decrease thechance of cooperationamong status-quo states. First, war will be profitablefor the winner. The costswill be low and thebenefits high. Of course,losers will suffer;the fear of losingcould induce states to tryto form stable cooperative arrangements, but the temptation of vic- torywill make thisparticularly difficult. Second, because wars are expectedto be bothfrequent and short, there will be incentivesfor high levelsof arms,and quickand strongreaction to theother's increases in arms.The statecannot afford to waituntil there is unambiguous evidencethat the other is buildingnew weapons. Even large states that havefaith in theireconomic strength cannot wait, because the war will be overbefore their products can reachthe army. Third, when wars are quick,states will haveto recruitallies in advance.33Without the opportunityfor bargainingand re-alignmentsduring the opening stagesof hostilities, peacetime diplomacy loses a degreeof thefluidity thatfacilitates balance-of-power policies. Because must be securedduring peacetime, the international system is morelikely to becomebipolar. It is hardto saywhether war therefore becomes more or less likely,but thisbipolarity increases tension between the two campsand makesit harderfor status-quo states to gainthe benefits of cooperation.Fourth, if wars are frequent, statesmen's perceptual thresh- oldswill be adjustedaccordingly and theywill be quickto perceive 32 Quoted in Fritz Fischer,War of Illusions (New York: NortonI975), 377, 46i. 33 George Quester, Oflense and Defense in the International System (New York: JohnWiley I977), I05-06; Sontag(fn. 5), 4-5. 190 WORLD POLITICS ambiguousevidence as indicatingthat others are aggressive. Thus, there willbe morecases of status-quopowers arming against each other in theincorrect belief that the other is hostile. Whenthe defense has theadvantage, all theforegoing is reversed. The statethat fears attack does not pre-empt-since that would be a wastefuluse of its military resources-but rather prepares to receivean attack.Doing so doesnot decrease the security of others,and several statescan do it simultaneously;the situation will thereforebe stable, and status-quopowers will be ableto cooperate.When Herman Kahn arguesthat ultimatums "are vastly too dangerous to givebecause . . . theyare quite likely to touchoff a pre-emptivestrike,"34 he incorrectly assumesthat it is alwaysadvantageous to strikefirst. Moreis involvedthan short-run dynamics. When the defense is dom- inant,wars are likelyto becomestalemates and can be won onlyat enormouscost. Relatively small and weak states can hold off larger and strongerones, or can deter attack by raising the costs of conquest to an unacceptablelevel. States then approach equality in what they can do to eachother. Like the.45-caliber pistol in theAmerican West, fortifica- tionswere the "great equalizer" in someperiods. Changes in thestatus quo areless frequent and cooperationis morecommon wherever the securitydilemma is therebyreduced. Manyof thesearguments can be illustratedby themajor powers' policiesin theperiods preceding the two world wars. Bismarck's wars surprisedstatesmen by showing that the offense had the advantage, and bybeing quick, relatively cheap, and quite decisive. Falling into a com- monerror, observers projected this pattern into the future.35 The result- ingexpectations had several effects. First, states sought semi-permanent allies.In theearly stages of the Franco-Prussian War, Napoleon III had thoughtthat there would be plentyof timeto recruitAustria to his side.Now, others were not going to repeat this mistake. Second, defense budgetswere high and reacted quite sharply to increaseson theother side.It is notsurprising that Richardson's theory of arms races fits this periodwell. Third, most decision makers thought that the next Euro-

34Kahn (fn. 23), 2II (also see I44). 35 For a generaldiscussion of suchmistaken learning from the past, see Jervis(fn. 5), chap. 6. The importantand still not completelyunderstood question of why this beliefformed and was maintainedthroughout the war is examinedin BernardBrodie, War and Politics (New York: MacmillanI973), 262-70; Brodie,"Technological Change, StrategicDoctrine, and PoliticalOutcomes," in Klaus Knorr,ed., Historical Dimensions of Problems (Lawrence: UniversityPress of Kansas i976), 290-92; and Douglas Porch,"The FrenchArmy and the Spiritof the Offensive,1900-14," in BrianBond and Ian Roy,eds., War and Society (New York: Holmes & Meier i975), II7-43. SECURITY DILEMMA 191 pean war would notcost much blood and treasure.36That is one reason why war was generallyseen as inevitableand why mass opinionwas so bellicose.Fourth, once war seemedlikely, there were strong pressures to pre-empt.Both sides believed that whoever moved first could pene- tratethe otherdeep enoughto disruptmobilization and thusgain an insurmountableadvantage. (There was no suchbelief about the use of navalforces. Although Churchill made an ill-advisedspeech saying that if Germanships "do notcome out and fightin timeof war theywill be dug out like ratsin a hole,"37everyone knew thatsubmarines, mines, and coastalfortifications made this impossible.So at the startof the war each navy preparedto defenditself rather than attack,and the short-rundestabilizing forces that launched the armiestoward each otherdid not operate.)38Furthermore, each side knew thatthe other saw the situationthe same way,thus increasing the perceiveddanger thatthe otherwould attack,and givingeach added reasonsto precipi- tatea war ifconditions seemed favorable. In thelong and theshort run, therewere thusboth offensive and defensiveincentives to strike.This situationcasts light on thecommon question about German motives in I9I4: "Did Germanyunleash the war deliberatelyto becomea world poweror did she supportAustria merely to defenda weakeningally," therebyprotecting her own position?3 To some extent,this question is misleading.Because of the perceivedadvantage of the offense,war was seen as the bestroute both to gainingexpansion and to avoiding drasticloss of influence.There seemed to be no way for Germany merelyto retainand safeguardher existingposition. Of coursethe war showedthese beliefs to have been wrongon all points.Trenches and machineguns gave thedefense an overwhelming advantage.The fightingbecame deadlocked and producedhorrendous casualties.It made no sensefor the combatantsto bleed themselvesto death.If theyhad knownthe power of the defensebeforehand, they would have rushedfor their own trenchesrather than for the enemy's territory.Each side could have done thiswithout increasing the other's

36 Some were not so optimistic.Gray's remark is well-known: "The lamps are going out all over Europe; we shall not see them lit again in our life-time."The German Prime Minister,Bethmann Hollweg, also feared the consequences of the war. But the controllingview was that it would certainlypay for the winner. 37 Quoted in Martin Gilbert, Winston S. Churchill, III, The Challenge of War, I9I4- I9I6 (Boston: Houghton MifflinI971), 84. 38 Quester (fn. 33), 98-99. Robert Art, The Influence of Foreign Policy on Seapower, II (Beverly Hills: Sage Professional Papers in International Studies Series, I973), 14-i8, 26-28. 39 Konrad Jarausch,"The Illusion of : Chancellor Bethmann Hollweg's Calculated Risk, July I914," Central European History, ii (March i969), 50. 192 WORLD POLITICS incentivesto strike.War mighthave broken out anyway, just as DD is a possibleoutcome of Chicken, but at leastthe pressures of time and thefear of allowing the other to get the first blow would not have con- tributedto thisend. And, had bothsides known the costs of thewar, theywould have negotiated much more seriously. The obviousquestion is whythe states did notseek a negotiatedsettlement as soonas the shapeof thewar becameclear. Schlieffen had said thatif his plan failed,peace should be sought.40The answeris complex,uncertain, and largelyoutside of the scope of our concerns. But part of the reason was thehope and sometimesthe expectation that breakthroughs could be madeand the dominance of the offensive restored. Without that hope, thepolitical and psychologicalpressures to fightto a decisivevictory mighthave been overcome. The politicsof the interwar period were shaped by the memories of theprevious conflict and the belief that any future war would resemble it.Political and militarylessons reinforced each other in ameliorating the securitydilemma. Because it was believedthat the First World Warhad beena mistakethat could have been avoided by skillful con- ciliation,both Britain and, to a lesserextent, France were highly sensi- tiveto thepossibility that interwar Germany was nota realthreat to peace,and alert to the danger that reacting quickly and stronglyto her armscould create unnecessary conflict. And because Britain and France expectedthe defense to continueto dominate,they concluded that it wassafe to adopt a morerelaxed and nonthreatening military posture.41 Britainalso felt less need to maintaintight bonds. The Allies' militaryposture then constituted only a slightdanger to Germany; had thelatter been content with the status quo, it wouldhave been easy forboth sides to have felt secure behind their lines of fortifications. Of coursethe Germans were not content, so it is notsurprising that they devotedtheir money and attentionto findingways out of a defense- dominatedstalemate. tactics were necessary if theywere to useforce to changethe status quo. The initialstages of thewar on theWestern Front also contrasted withthe First World War. Only with the new air arm were there any

40 Brodie (fn. 8), 58. 41 PresidentRoosevelt and the Americandelegates to the League of NationsDis- armamentConference maintained that the tank and mobile heavyartillery had re- establishedthe dominanceof the offensive,thus makingdisarmament more urgent (Boggs,fn. 28, pp. 31, io8), but thiswas a minorityposition and may not even have been believedby the Americans.The reducedprestige and influenceof the military, and the high pressuresto cut governmentspending throughout this period also con- tributedto the loweringof defensebudgets. SECURITY DILEMMA 193 incentivesto strikefirst, and these forces were too weak to carry out the grandioseplans that had beenboth dreamed and feared.The armies, stillthe main instrument, rushed to defensivepositions. Perhaps the alliescould have successfully attacked while the Germans were occu- piedin Poland.42But belief in thedefense was so greatthat this was neverseriously contemplated. Three months after the start of the war, theFrench Prime Minister summed up theview held by almost every- onebut Hitler: on theWestern Front there is "deadlock.Two Forcesof equalstrength and the one that attacks seeing such enormous casualties thatit cannotmove without endangering the continuation of thewar orof the aftermath."43 The Allieswere caught in a dilemmathey never fullyrecognized, let alone solved. On theone hand, they had very high waraims; although unconditional had notyet been adopted, theBritish had decidedfrom the start that the removal of Hitlerwas a necessarycondition for peace.44 On theother hand, there were no realisticplans or instrumentsfor allowing the Allies to imposetheir willon theother side. The BritishChief of the Imperial General Staff noted,"The Frenchhave no intentionof carrying out an offensivefor years,if at all"; theBritish were only slightly bolder.45 So theAllies lookedto a longwar that would wear the Germans down, cause civilian sufferingthrough shortages, and eventuallyundermine Hitler. There waslittle analysis to support this view-and indeed it probably was not supportable-butas long as thedefense was dominant and the numbers on each siderelatively equal, what else could the Allies do? To summarize,the security dilemma was muchless powerful after WorldWar I thanit had been before. In thelater period, the expected powerof thedefense allowed status-quo states to pursuecompatible securitypolicies and avoidarms races. Furthermore, high tension and fearof war did not setoff short-run dynamics by whicheach state, tryingto increaseits security, inadvertently acted to makewar more likely.The expectedhigh costs of war, however, led the Allies to believe

42 Jon Kimche, The Unjought (New York: Stein i968); Nicholas William Bethell, The War Hitler Won: The Fall of Poland, September 1939 (New York: Holt I972); Alan Alexandroffand Richard Rosecrance, "Deterrence in I939," World Politics, XXIX (April I977), 404-24- 43 Roderick Macleod and Denis Kelly, eds., Time Unguarded: The Ironside Diaries, I937-I940 (New York: McKay i962), I73. 4 For a short time, as France was falling, the BritishCabinet did discuss reaching a negotiated peace with Hitler. The officialhistory ignores this, but it is covered in P.M.H. Bell, A Certain Eventuality (Farnborough, England: Saxon House I974), 40-48. 45 Macleod and Kelly (fn. 43), I74. In flatcontradiction to common sense and almost everythingthey believed about , the Allies planned an expedition to Scandinavia to cut the supply of iron ore to Germany and to aid Finland against the Russians. But the dominant mood was the one described above. 194 WORLD POLITICS thatno saneGerman leader would run the risks entailed in an attempt to dominatethe Continent, and discouragedthem from risking war themselves.

Technologyand Geography.Technology and geography are the two mainfactors that determine whether the offense or thedefense has the advantage.As Brodienotes, "On thetactical level, as a rule,few physical factorsfavor the attacker but many favor the defender. The defender usuallyhas the advantage of cover. He characteristicallyfires from be- hindsome form of shelterwhile his opponent crosses open ground."46 Anythingthat increases the amount of ground the attacker has to cross, or impedeshis progress across it, or makeshim more vulnerable while crossing,increases the advantage accruing to thedefense. When states areseparated by barriers that produce these effects, the security dilemma is eased,since both can haveforces adequate for defense without being able to attack.Impenetrable barriers would actually prevent war; in reality,decision makers have to settlefor a gooddeal less. Buffer zones slowthe attacker's progress; they thereby give the defender time to prepare,increase problems of logistics,and reducethe number of sol- diersavailable for the final assault. At theend of the i9thcentury, ArthurBalfour noted Afghanistan's "non-conducting" qualities. "So longas itpossesses few roads, and no railroads,itwill be impossiblefor Russiato makeeffective use of hergreat numerical superiority at any pointimmediately vital to the Empire." The Russiansvalued buffers for thesame reasons; it is not surprising that when Persia was being divided intoRussian and Britishspheres of influencesome yearslater, the Russianssought assurances that the British would refrain from building potentiallymenacing railroads in theirsphere. Indeed, since railroad constructionradically altered the abilities of countries to defendthem- selvesand to attackothers, many diplomatic notes and muchintelli- genceactivity in thelate i9thcentury centered on thissubject.47 Oceans,large rivers, and mountain ranges serve the same function as bufferzones. Being hard to cross,they allow defense against superior numbers.The defenderhas merelyto stayon his sideof thebarrier andso canutilize all themen he canbring up toit. The attacker'smen, however,can crossonly a fewat a time,and theyare very vulnerable

46 Brodie (fn. 8), I79. 47 ArthurBalfour, "Memorandum," Committee on ImperialDefence, April 30, I903, pp. 2-3; see the telegramsby Sir ArthurNicolson, in G. P. Gooch and Harold Tem- perley,eds., British Documents on the Origins of the War, Vol. 4 (London: H.M.S.O. 1929), 429, 524. These barriersdo not preventthe passageof long-rangeaircraft; but even in the air, distanceusually aids the defender. SECURITY DILEMMA 195 whendoing so. If all stateswere self-sufficient islands, anarchy would be muchless of a problem.A smallinvestment in shore defenses and a smallarmy would be sufficienttorepel invasion. Only very weak states wouldbe vulnerable,and onlyvery large ones could menace others. As notedabove, the United States, and to a lesserextent Great Britain, havepartly been able to escapefrom the state of naturebecause their geographicalpositions approximated this ideal. Althoughgeography cannot be changedto conformto borders,bor- derscan and do changeto conform to geography. Borders across which an attackis easytend to be unstable.States living within them are likely to expandor be absorbed.Frequent wars are almostinevitable since attackingwill oftenseem the best way to protectwhat one has.This processwill stop, or at least slow down, when the state's borders reach- byexpansion or contraction-aline of natural obstacles. Security with- out attackwill thenbe possible.Furthermore, these lines constitute salientsolutions to bargainingproblems and, to theextent that they are barriersto migration,are likelyto divideethnic groups, thereby raisingthe costs and loweringthe incentives for conquest. Attachmentto one's state and itsland reinforce one quasi-geograph- ical aid to thedefense. Conquest usually becomes more difficult the deeperthe attacker pushes into the other's territory. Nationalism spurs thedefenders to fightharder; advancing not only lengthens the attack- er'ssupply lines, but takes him through unfamiliar and often devastated landsthat require troops for garrison duty. These stabilizing dynamics will notoperate, however, if thedefender's war materielis situated nearits borders, or ifthe people do notcare about their state, but only aboutbeing on thewinning side. In suchcases, positive feedback will be at workand initialdefeats will be insurmountable.48 Imitatinggeography, men have tried to createbarriers. Treaties may providefor demilitarized zones on bothsides of theborder, although suchzones will rarely be deepenough to provide more than warning. Even thiswas not possiblein Europe,but the Russiansadopted a gaugefor their railroads that was broader than that of the neighboring states,thereby complicating the logisticsproblems of any attacker- includingRussia. Perhapsthe most ambitious and at leasttemporarily successful at- temptsto constructa system that would aid thedefenses of bothsides werethe interwar naval treaties, as theyaffected Japanese-American 48 See, for example, the discussion of warfare among Chinese warlords in Hsi-Sheng Chi, "The Chinese Warlord System as an International System," in Morton Kaplan, ed., New Approaches to InternationalRelations (New York: St. Martin's i968), 405-25. 196 WORLD POLITICS relations.As mentionedearlier, the problem was thatthe United States could not defendthe Philippines without denying Japan the abilityto protecther home islands.49(In I94I this dilemmabecame insoluble whenJapan sought to extendher control to Malaya and theDutch East Indies.If the Philippineshad been invulnerable,they could have pro- videda securebase fromwhich the U.S. could interdictJapanese ship- ping betweenthe homeland and theareas she was tryingto conquer.) In theI920's and earlyi930's each sidewould have been willing to grant theother security for its possessions in returnfor a reciprocalgrant, and the WashingtonNaval Conferenceagreements were designedto ap- proachthis goal. As a Japanesediplomat later put it, theircountry's "fundamentalprinciple" was to have "a strengthinsufficient for attack and adequate for defense."50Thus, Japanagreed in I922 to accepta navyonly three-fifths as large as thatof theUnited States, and theU.S. agreednot to fortifyits Pacific islands."' (Japan had earlierbeen forced to agree not to fortifythe islands she had taken fromGermany in World War I.) Japan'snavy would not be large enough to defeat America'sanywhere other than close to thehome islands. Although the Japanesecould stilltake the Philippines, not only would theybe unable to move farther,but theymight be weakenedenough by theirefforts to be vulnerableto counterattack.Japan, however, gained security.An Americanattack was renderedmore difficultbecause the American bases were unprotectedand because,until i930, Japan was allowed unlimitednumbers of cruisers,destroyers, and submarinesthat could weakenthe American fleet as it made itsway acrossthe ocean.52 The othermajor determinantof the offense-defensebalance is tech- nology.When weaponsare highlyvulnerable, they must be employed beforethey are attacked.Others can remainquite invulnerable in their bases.The formercharacteristics are embodiedin unprotectedmissiles and manykinds of bombers.(It shouldbe notedthat it is not vulner- abilityper se thatis crucial,but the location of thevulnerability. Bomb- ersand missilesthat are easyto destroyonly after having been launched towardtheir targets do not createdestabilizing dynamics.) Incentives to strikefirst are usuallyabsent for naval forcesthat are threatenedby Some American decision makers, including militaryofficers, thought that the best way out of the dilemma was to abandon the Philippines. 50 Quoted in Elting Morrison,Turmoil and Tradition: A Study of the Life and Times of Henry L. Stinson (Boston: Houghton Mifflini960), 326. 51 The U.S. "refused to consider limitationson Hawaiian defenses,since these works posed no threat to Japan." Braisted (fn. 27), 6I2. 52 That is part of the reason why the Japanese admirals stronglyobjected when the civilian leaders decided to accept a seven-to-tenratio in lighter craft in I930. Stephen Pelz, Race to Pearl Harbor (Cambridge: Harvard UniversityPress I974), 3. SECURITY DILEMMA 197 a naval attack.Like missilesin hardenedsilos, they are usuallywell protectedwhen in theirbases. Both sides can then simultaneouslybe preparedto defendthemselves successfully. In groundwarfare under some conditions, forts, trenches, and small groupsof men in preparedpositions can hold offlarge numbersof attackers.Less frequently,a few attackerscan stormthe defenses.By and large,it is a contestbetween fortifications and supportinglight weaponson theone hand,and mobilityand heavierweapons that clear theway for the attack on theother. As theerroneous views held before the two worldwars show,there is no simpleway to determinewhich is dominant."[T]hese oscillationsare not smoothand predictablelike thoseof a swingingpendulum. They are unevenin both extentand time.Some occurin thecourse of a singlebattle or campaign,others in the courseof a war, stillothers during a seriesof wars."Longer-term oscillationscan also be detected: The earlyGothic age, from the twelfth to thelate thirteenthcentury, withits wonderful cathedrals and fortifiedplaces, was a periodduring whichthe attackersin Europegenerally met seriousand increasing difficulties,because the improvement in the strengthof fortressesout- ran the advancein the powerof destruction.Later, with the spread of firearmsat theend of thefifteenth century, old fortresseslost their powerto resist.An age ensuedduring which the offensepossessed, apartfrom short-term setbacks, new advantages.Then, duringthe seventeenthcentury, especially after about i66o, and untilat leastthe outbreakof the War of theAustrian Succession in 1740, the defense regainedmuch of the groundit had lost since the greatmedieval fortresseshad provedunable to meetthe bombardmentof the new and morenumerous .53 Anotherscholar has continuedthe argument:"The offensivegained an advantagewith new formsof heavymobile artilleryin the nine- teenthcentury, but the stalemateof World War I createdthe impres- sion thatthe defenseagain had an advantage;the German invasion in World War II, however,indicated the offensivesuperiority of highlymechanized armies in thefield."'54 The situationtoday with respectto conventionalweapons is un-

53John Nef, War and Human Progress (New York: Norton i963), I85. Also see ibid., 237, 242-43, and 323; C. W. Oman, in the Middle Ages (Ithaca, N.Y.: Cornell UniversityPress I953), 70-72; John Beeler, Warfare in Feudal Europe, 730-1200 (Ithaca, N.Y.: Cornell UniversityPress I971), 2I2-I4; Michael Howard, War in European History (London: Oxford UniversityPress I976), 33-37. 54Quincy Wright, A Study of War (abridged ed.; Chicago: Universityof Chicago Press i964), I42. Also see 63-70, 74-75. There are important exceptions to these gen- eralizations-the American , for instance, falls in the middle of the period Wright says is dominated by the offense. 198 WORLD POLITICS clear.Until recentlyit was believedthat tanks and tacticalair power gave the attackeran advantage.The initialanalyses of the 1973 Arab- Israeli war indicatedthat new anti-tankand anti-aircraftweapons have restoredthe primacyof the defense.These weapons are cheap, easyto use,and can destroya highproportion of theattacking vehicles and planesthat are sighted.It thenwould make sensefor a status-quo powerto buylots of $20,000 missilesrather than buy a fewhalf-million dollartanks and multi-milliondollar fighter-bombers.Defense would be possibleeven againsta large and well-equippedforce; statesthat careprimarily about self-protection would not need to engagein arms races. But furtherexaminations of the new technologiesand the his- toryof the OctoberWar cast doubt on these optimisticconclusions and leave us unable to renderany firmjudgment.55 Concerningnuclear weapons, it is generallyagreed that defenseis impossible-a triumphnot of the offense,but of deterrence.Attack makes no sense,not becauseit can be beatenoff, but becausethe at- tackerwill be destroyedin turn.In termsof the questionsunder con- siderationhere, the resultis the equivalentof the primacyof the de- fense.First, security is relativelycheap. Less than one percentof the G.N.P. is devotedto deterringa directattack on the United States; most of it is spenton acquiringredundant systems to providea lot of insuranceagainst the worst conceivable contingencies. Second, both sides can simultaneouslygain securityin the form of second-strike capability.Third, and relatedto the foregoing,second-strike capability can be maintainedin the face of wide variationsin the otherside's militaryposture. There is no purelymilitary reason why each side has to react quickly and stronglyto the other'sincreases in arms. Any spendingthat the otherdevotes to tryingto achieve first-strike capabilitycan be neutralizedby the state'sspending much smaller sums on protectingits second-strikecapability. Fourth, there are no incentivesto strikefirst in a crisis. Importantproblems remain, of course. Both sides have interests thatgo well beyonddefense of the homeland.The protectionof these interestscreates conflicts even if neitherside desiresexpansion. Fur- thermore,the shiftfrom defense to deterrencehas greatlyincreased the importanceand perceptionsof resolve.Security now restson each side's beliefthat the otherwould preferto run high risks of total destructionrather than sacrificeits vital interests.Aspects of the se-

55 GeoffreyKemp, Robert Pfaltzgraff, and Uri Ra'anan,eds., The OtherArms Race (Lexington,Mass.: D. C. Heath I975); JamesFoster, "The Future of Conventional Arms Control,"Policy Sciences, No. 8 (Spring I977), I-I9. SECURITY DILEMMA 199 curitydilemma thus appear in a new form.Are weapons procure- mentsused as an index of resolve?Must theybe so used? If one side failsto respondto theother's buildup, will it appearweak and thereby invitepredation? Can both sides simultaneouslyhave imagesof high resolve or is there a zero-sumelement involved? Although these problemsare real, theyare not as severeas those in the prenuclear era: thereare manyindices of resolve,and statesdo not so muchjudge imagesof resolvein the abstractas ask how likelyit is thatthe other will stand firmin a particulardispute. Since statesare most likely to standfirm on matterswhich concern them most, it is quite possible for both to demonstratetheir resolve to protecttheir own security simultaneously.

OFFENSE-DEFENSE DIFFERENTIATION The other major variable that affectshow stronglythe security dilemma operatesis whetherweapons and policies that protectthe statealso providethe capabilityfor attack.If theydo not, the basic postulateof the securitydilemma no longer applies. A statecan in- crease its own securitywithout decreasing that of others.The ad- vantageof the defensecan only amelioratethe securitydilemma. A differentiationbetween offensive and defensivestances comes close to abolishingit. Such differentiationdoes not mean, however,that all securityproblems will be abolished.If the offensehas the advantage, conquestand aggressionwill still be possible.And if the offense's advantageis greatenough, status-quo powers may findit too expen- sive to protectthemselves by defensiveforces and decide to procure offensiveweapons even thoughthis will menaceothers. Furthermore, stateswill stillhave to worrythat even if the other'smilitary posture shows that it is peacefulnow, it may develop aggressiveintentions in the future. Assumingthat the defenseis at least as potentas the offense,the differentiationbetween them allows status-quostates to behave in ways thatare clearlydifferent from those of aggressors.Three bene- ficialconsequences follow. First, status-quo powers can identifyeach other,thus laying the foundationsfor cooperation.Conflicts growing out of the mistakenbelief that the otherside is expansionistwill be less frequent.Second, status-quostates will obtain advance warning when othersplan aggression.Before a statecan attack,it has to de- velop and deployoffensive weapons. If procurementof theseweapons cannotbe disguisedand takesa fairamount of time,as it almostal- ways does, a status-quostate will have the timeto take countermeas- 200 WORLD POLITICS ures.It need not maintaina high level of defensivearms as long as itspotential adversaries are adoptinga peacefulposture. (Although beingso armedshould not, with the one important exception noted be- low,alarm other status-quo powers.) States do, in fact, pay special atten- tionto actionsthat they believe would not be takenby a status-quo statebecause they feel that states exhibiting such behaviorare ag- gressive.Thus theseizure or developmentof transportationfacilities will alarmothers more if thesefacilities have no commercialvalue, and thereforecan onlybe wantedfor military reasons. In i906, the Britishrejected a Russianprotest about their activities in a district of Persiaby claimingthat this area was "onlyof [strategic]impor- tance[to the Russians] if they wished to attackthe Indian frontier, or toput pressure upon us bymaking us thinkthat they intend to attack * "156 it. The sameinferences are drawn when a stateacquires more weapons thanobservers feel are needed for defense. Thus, the Japanese spokes- man at the 1930 Londonnaval conference said thathis countrywas alarmedby the American refusal to giveJapan a 70 percentratio (in placeof a 6o percentratio) in heavy cruisers: "As longas Americaheld thatten percent advantage, it was possiblefor her to attack.So when Americainsisted on sixtypercent instead of seventypercent, the idea wouldexist that they were trying to keep thatpossibility, and the Japanesepeople could not acceptthat."57 Similarly, when Mussolini told Chamberlainin January1939 thatHitler's arms program was motivatedby defensive considerations, the Prime Minister replied that "Germanmilitary forces were now so strongas to makeit impossible forany Power or combinationof Powersto attackher successfully. She couldnot wantany further armaments for defensive purposes; whatthen did shewant them for?""5 Ofcourse these inferences can be wrong-asthey are especially likely to be becausestates underestimate the degree to whichthey menace others.59And when they are wrong, the security dilemma is deepened. Becausethe state thinks it has receivednotice that the other is aggres- sive,its own armsbuilding will be lessrestrained and thechances of

Richard Challener, Admirals, Generals, and American Foreign Policy, i898-i9I4 (Princeton: Princeton University Press 1973), 273; Grey to Nicolson, in Gooch and Temperley (fn. 47), 414- 57Quoted in James Crowley, Japan's Quest for Autonomy (Princeton: Princeton University Press I966), 49. American naval officersagreed with the Japanese that a ten-to-sixratio would endanger Japan's supremacy in her home waters. 5 E. L. Woodward and R. Butler, eds., Documents on British Foreign Policy, i9i9- I939, Third series, III (London: H.M.S.O. 1950), 526. 59 Jervis (fn. 5), 69-72, 352-55. SECURITY DILEMMA 201 cooperationwill be decreased.But the dangersof incorrectinferences shouldnot obscurethe main point:when offensiveand defensivepos- turesare different,much of theuncertainty about the other's intentions thatcontributes to the securitydilemma is removed. The thirdbeneficial consequence of a differencebetween offensive and defensiveweapons is thatif all statessupport the statusquo, an obviousarms control agreement is a ban on weaponsthat are usefulfor attacking.As PresidentRoosevelt put it in his messageto the Geneva DisarmamentConference in 1933: "If all nationswill agree wholly to eliminatefrom possession and use the weapons which make pos- siblea successfulattack, defenses automatically will becomeimpregna- ble, and the frontiersand independenceof everynation will become secure."60The factthat such treatieshave been rare-the Washington naval agreementsdiscussed above and theanti-ABM treaty can be cited as examples-showseither that states are not alwayswilling to guaran- tee the securityof others,or that it is hard to distinguishoffensive fromdefensive weapons. Is such a distinctionpossible? Salvador de Madariaga,the Spanish statesmanactive in thedisarmament negotiations of theinterwar years, thoughtnot: "A weapon is eitheroffensive or defensiveaccording to which end of it you are looking at." The French Foreign Minister agreed (although French policy did not always follow this view): "Everyarm can be employedoffensively or defensivelyin turn.... The only way to discoverwhether arms are intendedfor purelyde- fensivepurposes or are held in a spiritof aggressionis in all cases to enquireinto the intentionsof the countryconcerned." Some evidence forthe validity of thisargument is providedby the fact that much time in theseunsuccessful negotiations was devotedto separatingoffensive fromdefensive weapons. Indeed, no simpleand unambiguousdefini- tionis possibleand in manycases no judgmentcan be reached.Before the Americanentry into World War I, wantedto arm merchantmenonly with guns in the back of the ship so they could not initiatea fight,but thisexpedient cannot be appliedto more commonforms of armaments." There are severalproblems. Even when a differentiationis possible, a status-quopower will want offensivearms under any of threecondi- tions.(i) If the offensehas a greatadvantage over the defense,pro- tectionthrough defensive forces will be too expensive.(2) Status-quo

60 Quoted in Merze Tate, The UnitedStates and Armaments(Cambridge: Harvard UniversityPress I948), io8. 61 Boggs (fn. 28), I5, 40. 202 WORLD POLITICS statesmay need offensive weapons to regain territory lost in theopening stagesof a war.It mightbe possible,however, for a stateto waitto procurethese weapons until war seems likely, and they might be needed onlyin relativelysmall numbers, unless the aggressor was ableto con- structstrong defenses quickly in theoccupied areas. (3) The statemay feelthat it mustbe preparedto takethe offensive either because the otherside will make peace only if it losesterritory or because the state has commitmentsto attack if theother makes war on a thirdparty. As notedabove, status-quo states with extensive commitments are often forcedto behavelike aggressors.Even when theylack such commit- ments,status-quo states must worry about the possibilitythat if they are able to hold offan attack,they will stillnot be able to end the war unlessthey move intothe other'sterritory to damage its military forcesand inflictpain. Many Americannaval officersafter the Civil War, forexample, believed that "only by destroyingthe commerceof theopponent could theUnited States bring him to terms."62 A furthercomplication is introducedby the factthat aggressors as well as status-quopowers require defensive forces as a preludeto ac- quiringoffensive ones, to protectone frontierwhile attackinganother, or forinsurance in case thewar goes badly.Criminals as well as - men can use bulletproofvests. Hitler as well as Maginotbuilt a line of forts.Indeed, Churchillreports that in 1936 the German Foreign Ministersaid: "As soon as our fortificationsare constructed[on our westernborders] and the countriesin Central Europe realize that France cannotenter German territory,all thesecountries will begin to feel verydifferently about theirforeign policies, and a new con- stellationwill develop."63So a statemay not necessarilybe reassuredif its neighborconstructs strong defenses. Morecentral difficulties are created by the fact that whether a weapon is offensiveor defensiveoften depends on the particularsituation- forinstance, the geographical setting and theway in whichthe weapon is used. "Tanks . . . spearheaded the fatefulGerman thrustthrough the Ardennesin 1940, but if the Frenchhad disposedof a properly concentratedarmored reserve, it would have providedthe bestmeans fortheir cutting off the penetration and turninginto a disasterfor the Germanswhat became insteadan overwhelmingvictory."" Anti-air- craftweapons seem obviouslydefensive-to be used, theymust wait

62 Kenneth Hagan, American Gunboat Diplomacy and the Old Navy, i877-i889 (Westport, Conn.: Greenwood Press I973), 20. 63 Winston Churchill, The Gathering Storm (Boston: Houghton I948), 206. 64 Brodie, War and Politics (fn. 35), 325. SECURITY DILEMMA 203 forthe otherside to come to them.But the Egyptianattack on Israel in 1973 would have been impossiblewithout effective air defensesthat coveredthe battlefield.Nevertheless, some distinctionsare possible. Sir JohnSimon, then the BritishForeign Secretary, in responseto the views cited earlier,stated that just because a fine line could not be drawn,"that was no reasonfor saying thatthere were not stretchesof territoryon eitherside which all practicalmen and women knew to be well on thisor thatside of the line." Althoughthere are almost no weaponsand strategiesthat are usefulonly for attacking, there are somethat are almostexclusively defensive. Aggressors could want them forprotection, but a statethat relied mostly on themcould not menace others.More frequently,we cannot"determine the absolutecharacter of a weapon, but [we can] make a comparison. . . [and] discover whetheror not the offensivepotentialities predominate, whether a weapon is moreuseful in attackor in defense."65 The essenceof defenseis keepingthe other side out of yourterritory. A purelydefensive weapon is one thatcan do thiswithout being able to penetratethe enemy'sland. Thus a committeeof militaryexperts in an interwardisarmament conference declared that armaments "in- capableof mobility by means of self-containedpower," or movableonly afterlong delay,were "only capable of being used forthe defenseof a State'sterritory."" The mostobvious examples are fortifications.They can shelterattacking forces, especially when theyare builtright along the frontier,67but theycannot occupy enemy territory. A statewith onlya strongline of forts,fixed guns, and a small armyto man them would not be much of a menace.Anything else thatcan serveonly as a barrieragainst attacking troops is similarlydefensive. In this cate- goryare systemsthat provide warning of an attack,the Russian's adop- tionof a differentrailroad gauge, and nuclearland minesthat can seal offinvasion routes. If total immobilityclearly defines a systemthat is defensiveonly, limitedmobility is unfortunatelyambiguous. As noted above, short- rangefighter aircraft and anti-aircraftmissiles can be used to coveran attack.And, unlikeforts, they can advancewith the troops.Still, their inabilityto reach deep into enemyterritory does make them more

65 Boggs (fn. 28), 42, 83. For a good argumentabout the possibledifferentiation betweenoffensive and defensiveweapons in the I930's, see Basil Liddell Hart,"Aggres- sion and the Problemof Weapons,"English Review, Vol. 55 (JulyI932), 7I-78. 68 Quoted in Boggs (fn. 28), 39. 67 On thesegrounds, the Germansclaimed in I932 that the Frenchforts were of- fensive(ibid., 49). Similarly,fortified forward naval basescan be necessaryfor launch- ing an attack;see Braisted(fn. 27), 643. 204 WORLD POLITICS usefulfor the defense than for the offense. Thus, the United States and Israelwould have been more alarmed in theearly I970's had theRus- siansprovided the Egyptians with long-range instead of short-range aircraft.Naval forces are particularly difficult to classify in theseterms, butthose that are veryshort-legged can be usedonly for coastal de- fense. Anyforces that for various reasons fight well only when on theirown soilin effectlack mobilityand thereforeare defensive.The mostex- tremeexample would be passiveresistance. Noncooperation can thwart an aggressor,but it is veryhard for large numbers of peopleto cross theborder and stagea sit-inon another'sterritory. Morocco's recent marchon the SpanishSahara approached this tactic, but its success dependedon specialcircumstances. Similarly, is defensiveto the extentto whichit requirescivilian support that is likelyto be forthcomingonly in oppositionto a foreigninvasion. In- deed,if guerrilla warfare were easily exportable and if it tookten de- fendersto destroyeach guerrilla, then this weapon would not only be one whichcould be usedas easilyto attackthe other's territory as to defendone's own, but one in whichthe offense had theadvantage: so thesecurity dilemma would operate especially strongly. Ifguerrillas are unable to fighton foreignsoil, other kinds of armies maybe unwillingto do so. An armyimbued with the idea thatonly defensivewars were just would fight less effectively, ifat all,if the goal wereconquest. Citizen militias may lack both the ability and thewill foraggression. The weaponsemployed, the short term of service,the timerequired for mobilization, and thespirit of repellingattacks on thehomeland, all lendthemselves much more to defensethan to at- tackson foreign territory.68 Lessidealistic motives can produce the same result. A leadingstudent of medievalwarfare has describedthe armiesof thatperiod as fol- lows:"Assembled with difficulty, insubordinate, unable to maneuver, readyto melt away from its standard the moment that its short period of servicewas over,a feudalforce presented an assemblageof un- soldierlikequalities such as haveseldom been known to coexist. Prima- rilyintended to defend its own borders from the Magyar, the Northman, or theSaracen . . . , theinstitution was utterlyunadapted to takethe offensive."69Some political groupings can be similarlydescribed. In-

68 The Frenchmade this argumentin the interwarperiod; see RichardChallener, The FrenchTheory of the Nation in Arms (New York: ColumbiaUniversity Press 1955), i8i-82. The Germansdisagreed; see Boggs (fn. 28), 44-45. 69Oman (fn. 53), 57-58. SECURITY DILEMMA 205 ternationalcoalitions are more readilyheld togetherby fearthan by hope of gain. Thus Castlereaghwas not being entirelyself-serving when in i8i6 he arguedthat the QuadrupleAlliance "could onlyhave owed itsorigin to a senseof commondanger; in itsvery nature it must be conservative;it cannotthreaten either the securityor the liberties of otherStates."70 It is no accidentthat most of the major campaigns of expansionhave been waged by one dominantnation (for example, Napoleon'sFrance and Hitler'sGermany), and thatcoalitions among relativeequals are usuallyfound defending the statusquo. Most gains fromconquest are too uncertainand raisetoo manyquestions of future squabblesamong the victorsto hold an alliancetogether for long. Al- thoughdefensive coalitions are by no means easy to maintain-con- flictingnational objectives and the free-riderproblem partly explain whythree of themdissolved before Napoleon was defeated-thecom- mon interestof seeingthat no statedominates provides a strongin- centivefor solidarity. Weaponsthat are particularlyeffective in reducingfortifications and barriersare of greatvalue to the offense.This is not to deny thata defensivepower will want some of thoseweapons if the otherside has them: Brodieis certainlycorrect to argue thatwhile theirtanks allowed the Germansto conquerFrance, properly used Frenchtanks could have haltedthe attack.But Francewould not have neededthese weaponsif Germanyhad not acquiredthem, whereas even if France had no tanks,Germany could not have foregonethem since they pro- vided the only chance of breakingthrough the Frenchlines. Mobile heavyartillery is, similarly, especially useful in destroyingfortifications. The defender,while needingartillery to fightoff attacking troops or to counterattack,can usuallyuse lighterguns sincethey do not need to penetratesuch massiveobstacles. So it is not surprisingthat one of the few thingsthat most nationsat the interwardisarmament con- ferenceswere able to agreeon was thatheavy tanks and mobileheavy guns were particularlyvaluable to a stateplanning an attack.7' Weaponsand strategiesthat depend for their effectiveness on surprise are almostalways offensive. That factwas recognizedby some of the delegatesto theinterwar disarmament conferences and is theprinciple behindthe commonnational ban on concealedweapons. An earlier representativeof thiswidespread view was themid-Igth-century Phila- delphianewspaper that argued: "As a measureof defense, knives, dirks,

70Quoted in Charles Webster, The Foreign Policy of Castlereagh, II, i815-i822 (London: G. Bell and Sons i963), 5I0. 71 Boggs (fn. 28), I4-I5, 47-48, 6o. 206 WORLD POLITICS and swordcanes are entirely useless. They are fitonly for attack, and all suchattacks are of murderouscharacter. Whoever carries such a weaponhas prepared himself for homicide."72 It is, of course,not always possibleto distinguishbetween forces thatare mosteffective for holding territory and forcesoptimally de- signedfor taking it. Such a distinctioncould not have been made for thestrategies and weaponsin Europeduring most of theperiod be- tweenthe Franco-Prussian War and World War I. Neithernaval forces nortactical air forcescan be readilyclassified in theseterms. But the pointhere is thatwhen such a distinctionis possible, the central char- acteristicof the security dilemma no longerholds, and one of the most troublesomeconsequences of anarchyis removed. Ogense-DefenseDifierentiation and StrategicNuclear Weapons. In theinterwar period, most statesmen held the reasonable position that weaponsthat threatened civilians were offensive.73 But whenneither sidecan protect its civilians, a counter-city posture is defensivebecause thestate can crediblythreaten to retaliateonly in responseto an at- tackon itselfor itsclosest allies. The costsof thisstrike are so high thatthe state could not threaten to use it forthe less-than-vital interest ofcompelling the other to abandonan establishedposition. In thecontext of deterrence, offensive weapons are those that provide defense.In thenow familiar reversal of commonsense, the state that couldtake its population out of hostage, either by active or passivede- fenseor by destroyingthe other's strategic weapons on the ground, wouldbe ableto alter the status quo. The desireto prevent such a situa- tionwas one of the rationales for the anti-ABM agreements; it explains whysome arms controllers opposed building ABM's to protectcities, but favoredsites that covered ICBM fields.Similarly, many analysts wantto limitwarhead accuracy and favormultiple re-entry vehicles (MRV's), butoppose multiple independently targetable re-entry vehi- cles (MIRV's). The formerare more useful than single warheads for penetratingcity defenses, and ensurethat the state has a second-strike capability.MIRV's enhancecounterforce capabilities. Some arms con- trollersargue that this is alsotrue of cruisemissiles, and thereforedo notwant them to be deployedeither. There is someevidence that the Russiansare not satisfiedwith deterrence and are seekingto regain thecapability for defense. Such an effort,even if notinspired by ag- gressivedesigns, would create a severesecurity dilemma. 72 Quotedin PhilipJordan, Frontier Law and Order (Lincoln:University of Nebraska Press I970), 7; also see i6-I7. 73Boggs (fn. 28), 20, 28. SECURITY DILEMMA 207 What is most importantfor the argumenthere is that land-based ICBM's are both offensiveand defensive,but when both sides rely on Polaris-typesystems (SLBM's), offenseand defenseuse different weapons.ICBM's can be used eitherto destroythe other'scities in retaliationor to initiatehostilities by attackingthe other'sstrategic missiles.Some measures-forinstance, hardening of missilesites and warningsystems-are purely defensive, since they do not make a first strikeeasier. Others are predominantlyoffensive-for instance, passive or active city defenses,and highlyaccurate warheads. But ICBM's themselvesare usefulfor both purposes.And because statesseek a high level of insurance,the desirefor protection as well as the con- templationof a counterforcestrike can explainthe acquisitionof ex- tremelylarge numbersof missiles.So it is verydifficult to inferthe other'sintentions from its militaryposture. Each side's effortsto in- creaseits own securityby procuringmore missiles decreases, to an ex- tentdetermined by the relativeefficacy of the offenseand the defense, the otherside's security.That is not the case when both sides use SLBM's. The point is not that sea-basedsystems are less vulnerable thanland-based ones (thisbears on the offense-defenseratio) but that SLBM's are defensive,retaliatory weapons. First, they are probablynot accurateenough to destroymany military targets.74 Second, and more important,SLBM's are notthe main instrumentof attackagainst other SLBM's. The hardestproblem confronting a statethat wants to take its citiesout of hostageis to locate the other'sSLBM's, a job thatre- quires not SLBM's but anti-submarineweapons. A statemight use SLBM's to attackthe other'ssubmarines (although other weapons would probablybe moreefficient), but without anti- (ASW) capabilitythe task cannotbe performed.A status-quostate thatwanted to foregooffensive capability could simplyforego ASW researchand procurement. There are two difficultieswith thisargument, however. First, since thestate's SLBM's are potentiallythreatened by theother's ASW capa- bilities,the statemay want to pursueASW researchin orderto know what the othermight be able to do and to designdefenses. Unless it does this,it cannotbe confidentthat its submarinesare safe.Second, becausesome submarines are designed to attacksurface ships, not launch missiles,ASW forceshave missionsother than taking citiesout of hostage.Some U.S. officialsplan fora long war in Europewhich would requirekeeping the sea lanes open againstRussian submarines.De-

74 See, however,Desmond Ball, "The CounterforcePotential of AmericanSLBM Systems,"Journal of Peace Research, xiv (No. I, I977), 23-40. 208 WORLD POLITICS signingan ASW forceand strategythat would meetthis threat with- out endangeringSoviet SLBM's would be difficultbut not impossible, sincethe two missionsare somewhatdifferent.75 Furthermore, the Rus- siansdo not need ASW forcesto combatsubmarines carrying out con- ventionalmissions; it mightbe in America'sinterest to sacrificethe abilityto meeta threatthat is not likelyto materializein orderto re- assurethe Russiansthat we are not menacingtheir retaliatory capabil- ity. When both sidesrely on ICBM's, one side's missilescan attackthe other's,and so the statecannot be indifferentto the other'sbuilding program.But because one side's SLBM's do not menace the other's, each sidecan buildas manyas it wantsand theother need notrespond. Each side'sdecision on the size of itsforce depends on technicalques- tions,its judgmentabout how much destructionis enough to deter, and the amountof insuranceit is willingto pay for-and thesecon- siderationsare independentof the size of the other'sstrategic force. Thus the crucialnexus in the armsrace is severed. Here two objectionsnot onlycan be raisedbut have been,by those who feelthat even if American second-strike capability is in no danger, the UnitedStates must respond to a Sovietbuildup. First, the relative numbersof missilesand warheadsmay be used as an index of each side's power and will. Even if thereis no militaryneed to increase Americanarms as theRussians increase theirs, a failureto respondmay lead thirdparties to thinkthat the U.S. has abandonedthe competition withthe U.S.S.R. and is no longerwilling to pay the price of world leadership.Furthermore, ifeither side believes that nuclear "superiority" matters,then, through the bargaininglogic, it will matter.The side with"superiority" will be morelikely to standfirm in a confrontation ifit thinks its "stronger" military position helps it, or ifit thinksthat the otherthinks its own "weaker"military position is a handicap.To allow the otherside to have moreSLBM's-even if one's own second-strike capabilityis unimpaired-willgive the'otheran advantagethat can be translatedinto political gains. The secondobjection is thatsuperiority does matter,and not only becauseof mistakenbeliefs. If nuclearweapons are used in an all-or- none fashion,then all that is needed is second-strikecapability. But limited,gradual, and controlledstrikes are possible.If the otherside has superiority,it can reducethe state'sforces by a slow-motionwar 75 Richard Garwin, "Anti-Submarine Warfare and National Security," Scientific American, Vol. 227 (JulyI972), I4-25. SECURITY DILEMMA 209 of attrition.For the stateto strikeat the other'scities would invite retaliation;for it to replywith a limitedcounterforce attack would furtherdeplete its supplyof missiles.Alternatively, the othercould employdemonstration attacks-such as takingout an isolatedmilitary base or explodinga warheadhigh over a city-in orderto demonstrate its resolve.In eitherof thesescenarios, the statewill sufferunless it matchesthe other's arms posture.76 These two objections,if valid,mean thateven with SLBM's one can- not distinguishoffensive from defensivestrategic nuclear weapons. Compellencemay be more difficultthan deterrence,77but if decision makersbelieve that numbers of missilesor of warheadsinfluence out- comes,or if theseweapons can be used in limitedmanner, then the postureand policythat would be needed forself-protection is similar to thatuseful for aggression. If thesecond objection has merit,security wouldrequire the ability to hitselected targets on theother side, enough ammunitionto wage a controlledcounterforce war, and the willing- ness to absorblimited countervalue strikes. Secretary Schlesinger was correctin arguingthat this capability would not constitutea first-strike capability.But becausethe "SchlesingerDoctrine" could be used not only to cope with a parallel Russian policy,but also to supportan Americanattempt to changethe statusquo, the new Americanstance would decreaseRussian security. Even if the U.S.S.R. were reassured thatthe presentU.S. Governmentlacked the desireor courageto do this,there could be no guaranteethat future governments would not use thenew instrumentsfor expansion. Once we move away fromthe simpleidea thatnuclear weapons can onlybe used forall-out strikes, halfthe advantage of havingboth sides rely on a sea-basedforce would disappearbecause of the lack of an offensive-defensivedifferentiation. To theextent that affects political relations, it would be harderfor the United Statesand the SovietUnion to cooperateeven ifboth supported the status quo. Althougha full explorationof thesequestions is beyondthe scope of thispaper, it should be noted thatthe objectionsrest on decision makers'beliefs-beliefs, furthermore, that can be stronglyinfluenced byAmerican policy and Americanstatements. The perceptionsof third 76The latter scenario, however, does not require that the state closely match the number of missiles the other deploys. 77 , Arms and Influence (New Haven: Yale UniversityPress i966), 69-78. Schelling's arguments are not entirelyconvincing, however. For furtherdiscus- sion, see Jervis," Re-Visited," Working Paper No. I4, UCLA Pro- gram in Arms Control and International Security. 210 WORLD POLITICS nationsof whetherthe detailsof the nuclearbalance affectpolitical conflicts-and,to a lesserextent, Russian beliefs about whether superi- orityis meaningful-arelargely derived from the Americanstrategic debate.If most Americanspokesmen were to take the positionthat a securesecond-strike capability was sufficientand thatincrements over that (short of a first-strikecapability) would only be a waste of money,it is doubtfulwhether America's allies or the neutralswould judge the superpowers'useful military might or politicalwill by the size of theirstockpiles. Although the Russiansstress war-fighting abil- ity,they have not contendedthat marginal increases in strategicforces bring politicalgains; any attemptto do so could be renderedless effectiveby an Americanassertion that this is nonsense.The bargaining advantagesof possessingnuclear "superiority" work best when both sidesacknowledge them. If the"weaker" side convincesthe otherthat it does notbelieve there is anymeaningful difference in strength,then the "stronger"side cannot safelystand firmbecause thereis no in- creasedchance that the otherwill back down. This kind of argumentapplies at leastas stronglyto the secondob- jection.Neither side can employlimited nuclear options unless it is quite confidentthat the otheraccepts the rules of thegame. For if the otherbelieves that nuclear war cannot be controlled,it will either refrainfrom responding-whichwould be fine-or launch all-out retaliation.Although a statemight be readyto engagein limitednuclear war withoutacknowledging this possibility-and indeed, that would be a reasonablepolicy for the United States-it is not likelythat the otherwould have sufficientfaith in thatprospect to initiatelimited strikesunless the statehad openlyavowed its willingnessto fightthis kind of war. So the United States,by patientlyand consistentlyex- plainingthat it considerssuch ideas to be mad and thatany nuclear wars will inevitablyget out of control,could gain a large measureof protectionagainst the danger that the Soviet Union mightseek to employa "SchlesingerDoctrine" against an Americathat lacked the militaryability or politicalwill to respondin kind. Such a position is made more convincingby the inherentimplausibility of the argu- mentsfor the possibilityof a limitednuclear war. In summary,as long as statesbelieve that all thatis neededis second- strikecapability, then the differentiationbetween offensive and defen- siveforces that is providedby relianceon SLBM's allows each side to increaseits securitywithout menacing the other,permits some infer- encesabout intentions to be drawnfrom military posture, and removes themain incentivefor status-quo powers to engagein armsraces. SECURITY DILEMMA 211

IV. FOUR WORLDS The two variableswe have been discussing-whetherthe offense or the defensehas the advantage,and whetheroffensive postures can be distinguishedfrom defensive ones-can be combinedto yieldfour possibleworlds.

OFFENSE HAS DEFENSE HAS THE ADVANTAGE THE ADVANTAGE

2 OFFENSIVE POSTURE Security dilemma, but NOT DISTINGUISHABLE Doubly dangerous security requirements FROM DEFENSIVE ONE maybe compatible.

4 Nlo security dilemma, but OFFENSIVE POSTURE aggressionpossible. DISTINGUISHABLE Status-quo states con follow Doubly stable FROM DEFENSIVE ONE differentpolicy than aggressors. Warninggiven.

The firstworld is the worstfor status-quostates. There is no way to get securitywithout menacing others, and securitythrough defense is terriblydifficult to obtain.Because offensiveand defensivepostures are the same,status-quo states acquire the same kind of armsthat are soughtby aggressors.And becausethe offensehas the advantageover the defense,attacking is the best routeto protectingwhat you have; status-quostates will thereforebehave like aggressors.The situation will be unstable.Arms races are likely.Incentives to strikefirst will turncrises into wars. Decisive victories and conquestswill be common. Stateswill grow and shrinkrapidly, and it will be hard forany state to maintainits size and influencewithout trying to increasethem. Cooperationamong status-quopowers will be extremelyhard to achieve. There are no casesthat totally fit this picture, but it bearsmore than a passingresemblance to Europebefore World War I. Britainand Ger- many,although in manyrespects natural allies, ended up as enemies.Of coursemuch of the explanation lies in Germany'sill-chosen policy. And fromthe perspective of our theory,the powers' ability to avoidwar in a seriesof earliercrises cannot be easilyexplained. Nevertheless, much of the behaviorin thisperiod was the productof technologyand beliefs 212 WORLD POLITICS thatmagnified the security dilemma. Decision makers thought that the offensehad a big advantageand saw littledifference between offensive and defensivemilitary postures. The era was characterizedby arms races.And oncewar seemedlikely, mobilization races created powerful incentivesto strikefirst. In the nuclearera, the firstworld would be one in whicheach side relied on vulnerableweapons thatwere aimed at similarforces and each side understoodthe situation.In thiscase, the incentivesto strike firstwould be veryhigh-so high thatstatus-quo powers as well as aggressorswould be sorelytempted to pre-empt.And sincethe forces could be used to changethe statusquo as well as to preserveit, there wouldbe no wayfor both sides to increasetheir security simultaneously. Now the familiarlogic of deterrenceleads both sides to see the dan- gersin thisworld. Indeed, the new understandingof thissituation was one reasonwhy vulnerablebombers and missileswere replaced.Ironi- cally,the i950's would have been morehazardous if thedecision mak- ers had been aware of the dangersof theirposture and had therefore feltgreater pressure to strikefirst. This situationcould be recreatedif bothsides were to relyon MIRVed ICBM's.

In thesecond world, the security dilemma operates because offensive and defensivepostures cannot be distinguished;but it does not operate as stronglyas in the firstworld because the defense has the advantage, and so an incrementin one side's strengthincreases its securitymore thanit decreasesthe other's. So, if bothsides have reasonablesubjective securityrequirements, are of roughlyequal power,and the variables discussedearlier are favorable,it is quite likelythat status-quostates can adopt compatiblesecurity policies. Although a statewill not be able to judge the other'sintentions from the kindsof weaponsit pro- cures,the level of arms spendingwill give importantevidence. Of coursea statethat seeks a highlevel of armsmight be not an aggressor but merelyan insecurestate, which if conciliatedwill reduceits arms, and if confrontedwill replyin kind. To assumethat the apparently excessivelevel of armsindicates aggressiveness could thereforelead to a responsethat would deepenthe dilemma and createneedless conflict. But empathyand skillfulstatesmanship can reducethis danger. Fur- thermore,the advantageous position of thedefense means that a status- quo statecan oftenmaintain a high degreeof securitywith a level of arms lower than thatof its expectedadversary. Such a statedemon- stratesthat it lacks the abilityor desireto alterthe statusquo, at least at the presenttime. The strengthof the defensealso allows statesto SECURITY DILEMMA 213 reactslowly and withrestraint when they fear that others are menacing them.So, althoughstatus-quo powers will to some extentbe threaten- ing to others,that extent will be limited. This worldis theone thatcomes closest to matchingmost periods in history.Attacking is usually harder than defendingbecause of the strengthof fortificationsand obstacles.But purelydefensive postures are rarelypossible because fortifications are usuallysupplemented by armiesand mobileguns which can supportan attack.In the nuclear era, thisworld would be one in which both sidesrelied on relatively invulnerableICBM's and believedthat limited nuclear war was im- possible.Assuming no MIRV's, it would takemore than one attacking missileto destroyone of the adversary's.Pre-emption is thereforeun- attractive.If both sideshave large inventories,they can ignoreall but drasticincreases on theother side. A worldof eitherICBM's or SLBM's in whichboth sides adopted the "SchlesingerDoctrine" would proba- bly fitin this categorytoo. The means of preservingthe statusquo would also be the means of changingit, as we discussedearlier. And the defenseusually would have the advantage,because compellence is more difficultthan deterrence.Although a statemight succeed in changingthe statusquo on issuesthat matter much more to it than to others,status-quo powers could deter major provocationsunder mostcircumstances.

In the thirdworld there may be no securitydilemma, but thereare securityproblems. Because states can procuredefensive systems that do not threatenothers, the dilemmaneed not operate.But because the offensehas the advantage,aggression is possible,and perhapseasy. If the offensehas enough of an advantage,even a status-quostate may takethe initiativerather than risk being attackedand defeated.If the offensehas less of an advantage,stability and cooperationare likely. becausethe status-quostates will procuredefensive forces. They need notreact to otherswho are similarlyarmed, but can waitfor the warn- ing theywould receiveif othersstarted to deployoffensive weapons. But each statewill have to watch the otherscarefully, and thereis roomfor false suspicions. The costlinessof the defenseand the allure of the offensecan lead to unnecessarymistrust, hostility, and war, unlesssome of the variablesdiscussed earlier are operatingto restrain defection. A hypotheticalnuclear world thatwould fitthis description would be one in whichboth sides relied on SLBM's, but in whichASW tech- niqueswere very effective. Offense and defensewould be different,but 214 WORLD POLITICS theformer would have the advantage. This situationis notlikely to occur;but if it did,a status-quostate could show its lack of desireto exploitthe other by refrainingfrom threatening its submarines. The desireto havemore protecting you than merely the other side's fear of retaliationis a strongone, however, and a statethat knows that it wouldnot expand even if itscities were safe is likelyto believethat theother would not feel threatened by its ASW program.It is easyto see howsuch a worldcould become unstable, and how spiralsof ten- sionsand conflictcould develop.

The fourthworld is doublysafe. The differentiationbetween offen- siveand defensivesystems permits a wayout of the security dilemma; theadvantage of the defense disposes of the problems discussed in the previousparagraphs. There is no reasonfor a status-quopower to be temptedto procure offensive forces, and aggressorsgive notice of their intentionsby theposture they adopt. Indeed, if theadvantage of the defenseis greatenough, there are no securityproblems. The loss of theultimate form of thepower to alterthe status quo wouldallow greaterscope for the exerciseof nonmilitarymeans and probably wouldtend to freezethe distribution of values. Thisworld would have existed in thefirst decade of the 20th century if the decisionmakers had understoodthe availabletechnology. In thatcase, the European powers would have followed different policies bothin thelong run and in thesummer of 1914. EvenGermany, fac- ing powerfulenemies on bothsides, could have made herself secure bydeveloping strong defenses. France could also have made her fron- tieralmost impregnable. Furthermore, when crises arose, no onewould havehad incentivesto strikefirst. There would have been no competi- tivemobilization races reducing the timeavailable for negotiations. In thenuclear era, this world would be onein which the superpowers reliedon SLBM's,ASW technologywas not up to itstask, and limited nuclearoptions were not taken seriously. We havediscussed this situa- tionearlier; here we needonly add that,even if our analysis is correct andeven if the policies and postures of both sides were to move in this direction,the problem of violencebelow the nuclear threshold would remain.On issuesother than defense of the homeland, there would still be securitydilemmas and securityproblems. But the worldwould neverthelessbe saferthan it has usuallybeen.