Jervis 1988 - Realism, Game Theory and Cooperation

Trustees of Princeton University
Realism, Game Theory, and Cooperation

Author(s): Robert Jervis
Source: World Politics, Vol. 40, No. 3 (Apr., 1988), pp. 317-349
Published by: Cambridge University Press
Stable URL: http://www.jstor.org/stable/2010216 .
Accessed: 22/09/2013 02:31
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
This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

All use subject to JSTOR Terms and Conditions
REALISM, GAME THEORY, AND
COOPERATION
By ROBERT JERVIS *
E VER sinceThucydides,
scholarshaveemphasizedthatinternational
politics is shaped by the anarchical contextin which it takes place.
The perniciouseffectof what JohnHerz called the securitydilemma-
the factthatmost of the ways in which a countryseeks to increaseits se-
curityhave the unintendedeffectof decreasingthe securityof others-
also is familiar,and indeed, also can be found in Thucydides: "What
made war inevitable was the growth of Athenian power and the fear
which this caused in Sparta." In view of such dynamics,how can states
cooperate? Recentanalyseshave formalizedtheseproblemsand analyzed
themby means of modern social sciencetechniques.'
The work arisesfromtheintersectionof Realism and game theory.Al-
though common interestsare stressedmore than in some formsof Real-
ism,the basic assumptionsclearlyfitwithinthisschool: thefocusis on the
stateas an actor2and on the strategiesthatcan rationallybe used to fur-
ther its interests.The studies use simple game-theorymodels, or what
Barry O'Neill calls "proto-game theory,"3in order to gain the insights
* I am gratefulfor commentsby Robert Art, JohnConybeare,Jeffrey Frieden, Joanne
Gowa, JosephGrieco, Ernst Haas, StanleyHoffmann,Stephen Krasner,Deborah Larson,
Fred Lawson, Helen Milner,RichardNelson, and JackSnyder.
JohnHertz, "Idealist Internationalism
and the SecurityDilemma," WorldPolitics2 (Jan-
uary I950), I57-80; Thucydides,The PeloponnesianWar,trans.by Rex Warner (Harmonds-
worth,U.K.: Penguin, I954), 25. The recentliteratureis summarizedand extendedin World
Politics38 (October i985), also published as Kenneth Oye, ed., CooperationunderAnarchy
(Princeton: PrincetonUniversityPress, i986). The frameworkused grows out of Robert
Axelrod, The Evolutionof Cooperation(New York: Basic Books, i984); RobertJervis,"Co-
operationundertheSecurityDilemma," WorldPolitics30 (JanuaryI978), i67-214; and Robert
Keohane, AfterHegemony(Princeton:PrincetonUniversityPress, i984). Related arguments
are made by Michael Taylor,Anarchyand Cooperation(New York: Wiley, I976). Although
the subject matteris the same as that treatedin Hedley Bull, The AnarchicalSociety(New
York: Columbia UniversityPress,I977), Bull's approachis different and his work is notcited
in thisliterature.For argumentsthatBull providesa betterfoundationforunderstandingin-
ternationalpoliticsthan does the work analyzed here, see Hayward Alker, Jr.,"The Pre-
sumptionof Anarchyin World Politics,"and RichardAshley,"Hedley Bull and theAnarchy
Problematique,"both in Alker and Ashley,eds.,AfterRealism:Anarchy, Power,and Interna-
tionalCollaboration(forthcoming).
See Charles Lipson, "Bankers' Dilemmas: PrivateCooperation in ReschedulingSover-
eign Debts," WorldPolitics38 (October i985), 200-225, for an analysisof non-stateactors
withinthisframework.
3 O'Neill, "Game Theory and theStudyof DeterrenceofWar," in RobertAxelrod,Robert
Jervis,Roy Radner,and Paul Stern,eds.,Perspectives in Deterrence,forthcoming.

318 WORLD POLITICS
and rigorthatstem fromformalization.But the formalizationis limited:
what is most importantis the basic structureof the game and the analo-
gies thatare provided.4Game theoryand Realism are generallycompat-
ible-both are structural,strategic,and rational-but each has its own
vulnerabilities.Thus, some of my criticismscan be tracedmore to prob-
lems with Realism, othersmore to problemswith game theory.But my
focus is on the work thatexemplifiesthe intersectionof the two; it is not
a full-blowncritiqueof the entiretyof eitherapproach.
Drawing on the conceptsof the iteratedPrisoners'Dilemma (PD) and
public goods,5the basic question posed by the recentwork is how self-
interestedactorscan cooperate in the faceof anarchyand importantcon-
flictinginterests.By takingtwo actorsand positingthateach of themhas
only two choices (cooperatingwithor defectingfromtheother),an inter-
estingworld emergesin which fouroutcomesare possible.They are pre-
ferredby the actorin the followingorder:first,theactordefectswhile the
other cooperates (DC), thus allowing the formerto gain an advantage;
second, both actors cooperate (CC); third,both may defect(DD), thus
producing competition;the fourthand worstoutcome would be forthe
actor to cooperate while the otherdefects(CD) and therebyexploitshim.
What makes thisconfigurationdisturbingis thateven ifeach side prefers
CC to DD (and each knows thatthisis the other'spreference),the result
can be DD because each is drivenby thehope of gainingitsfirstchoice-
which would be to exploit the other (DC) and its fear that,if it co-
operates,the otherwill exploitit (CD).
When a good idea like thiscomes along,our exuberanceat findingnew
insightsleads us to extendand applyit widely,postponinga consideration
of problemsand limitations.To some extentthisis functional:as Albert
4 For contrastingevaluationsof thepotentialof higherformsofgame theory,see ibid.,and
Duncan Snidal, "The Game Theoryof InternationalPolitics," WorldPolitics38 (October
i985), 25-57. For an excellentgeneraldiscussion,see Thomas Schelling,"What is Game The-
ory?" in Schelling,Choice and Consequence(Cambridge: Harvard UniversityPress, i984),
2I3-I4.
5 The relationshipbetweenPD and public goods is technical,complex,and subjectto dis-
pute. See JohnConybeare,"Public Goods, Prisoners'Dilemma, and theInternationalPolitical
Economy,"International StudiesQuarterly28 (March i984), 5-22, and RussellHardin, Collec-
tiveAction(Baltimore:The JohnsHopkins UniversityPress,i982), i6-30. The essentialsim-
ilaritybetween the two thatis relevanthere is thatthe equilibriumsolutionis non-optimal.
That is, in the absence of devices to avoid thisoutcome,individualself-interested rationality
leads each actor to be worse offthan he could have been ifall playershad acted differently.
This is true even thoughpublic goods are characterizedby non-rivalry of consumptionand
non-excludability-two dimensionsthat can be distinguished:see Duncan Snidal, "Public
Goods, PropertyRights,and PoliticalOrganizations,"International StudiesQuarterly 23 (De-
cemberI979), 532-66-in contrastto two-personPDs thathave bothrivalryand excludability.
When large numbersare involvedin a PD, as theyare in the Tragedy of the Commons,ex-
cludabilityis precluded and the situationcan be consideredas one of a public good thathas
the characteristicsof non-excludability but not non-rivalry.
These distinctions are important
in manyanalyses,but theyare not centralhere.

REALISM, GAME THEORY, AND COOPERATION 319
Hirschman has noted,in many cases we would not startwhat will prove
to be a fruitfulenterpriseifwe were aware of thedifficultiesat thebegin-
ning.6But ifwe ignorethe problemsin theearlystages,we may overreact
and rejectthe entireapproach when theybecome obvious later.It would
be more productiveif scholars simultaneouslyexplored the potentialof
and the problems with the approach. Since the potentialhas been dis-
cussed elsewhere(see fn. I), I will concentrateon the problems.
Three strengthsthat have made thisresearchstimulatingand impor-
tantshould be noted,however.First,it builds upon centralcharacteristics
of internationalpolitics-anarchy, the securitydilemma,and the combi-
nation of common and conflictinginterests.Second, the approach is par-
simonious and lends itselfto deductive theorizing.Third, it seeks to
bringtogetherthestudyof conflictand thestudyofcooperation,and tries
to explain a wide range of phenomena encompassingboth securityand
politicaleconomy.The deterringof exploitationmay be as relevantto the
stabilityof monetarysystemsas to armscontrol;strategiesbywhich states
can gain the benefitsof mutual cooperationmay be as importantfornu-
clear posturesas forinternationaltrade.Indeed, as thelinksto the theory
of public goods indicate, the propositionsproduced should apply to a
wide range of cases outside of internationalpolitics.
It is not a good sign,however,thatprisonersconfrontedby a District
Attorneydo not behave as the model would lead us to expect.7In order
to apply the framework,we have toassumemanyoftheelementsof world
politics that in fact are most problematical.The actor's values, prefer-
ences, beliefs,and definitionof self all are exogenous to the model and
must be provided before analysis can begin.8 If they were straight-
forward,the analysis of internationalpoliticswould be much simpler.
They thereforeneed to be investigatedand explained,nottakenas givens.
Adoption of stereotypicalRealist assumptionscan supplysome of the in-
puts the analysisrequires,but I will argue thatthisis not completelysat-
isfactory.
CHOICE AND CONTEXT
Perhaps the most importantlimitationof the work on anarchyis that

it looks at individual actors,theirpreferences,
and theirchoices,and thus
6 Hirschman'sdiscussionof what he calls the "hidinghand" is in hisDevelopment Projects
Observed(Washington,DC: BrookingsInstitution,i967), 9-34.
7 Brian Forstand Judith Lucianovic,"The Prisoner'sDilemma: Theoryand Reality,"Jour-
nal of CriminalJustice5 (Spring I977), 55-64.
8 The model also assumesthatstatescan fruitfully be consideredas unitaryactors.The de-
batesover thisissue,althoughimportant, are so well knownthattheywill onlybe touchedon
here.

320 WORLD POLITICS
blindsus to thebroadersettingin whichbehavioroccurs.Problemsarise
in a contextand out ofa history, and notall patterns
are theproductof
carefulor evenconsciouschoice.Justas in mostofoureveryday liveswe
carryout routines,so statesalso continueestablishedpolicieswithlittle
calculationor thought.9 Each new daydoes notbringa new beginning;
severerestrictionsare placedon us by theexpectations-including our
own expectations about ourselves-thatconstitute the contextwithin
whichwe mustbehave.
Whenclearpointsofchoiceoccur,theyareoftenstructured bytheset-
tingsin whichtheyarise.The timingofdecisionsand eventsis important
and at leastpartially
beyonda decisionmaker'scontrol. For example,the
productionschedulefornew nuclearsubmarineshas periodically pre-
sentedPresidentReaganwiththechoiceofwhetherto abidebytheun-
ratifiedSALT ii Treatyand destroya usablebutoldermissle-launching
submarine, or tobreakat leastpartofthisagreement.The dateon which
thesedecisionshad tobe madewas in partan arbitrary productofindus-
trialvagaries.'oFurthermore, unexpectedeventsmayoccurat thesame
timeas thedecision,and forcethemselves on thestatesmen'sfeelingsand
calculations.Thus, by strainingthe NATO alliance,Reagan'sair strike
againstLibyain Aprili986 mayhavemadeitharderforhimtobreakthe
treatyat thedecisionpointsixweekslater.In turn,thisdecisiongreatly
increasedthepressures on Reaganto takea tougherpositionat thenext
choicepointin late i986."1Similarly,
theUnitedStatesmighthavemade
more of an effortto cooperate with the Soviet Union in banning MIRVS
had it notbeensimultaneously negotiatingoverABMS.Anotherexample
is thatAmericancooperationwiththeMarielboatlift of refugeesfrom
Cuba was reducedbytheaccidentsofa changeinrelevant
laws-adopted
forunrelatedreasons-and a largesimultaneousemigrationfromHaiti
to the U.S.12
More generally,issues arise in particularhistoricalcontextsthatshape
9 JamesRosenau stressesthe role of habitsin "BeforeCooperation: Hegemons, Regimes,
Organization40 (Autumn i986),
and Habit-Driven Actors in World Politics,"International
86i-70.
- Michael Gordon, "Air Force's Delay Said to Keep U.S. to '79 Arms Limit,"New York
Times,August 29, i986.
--Michael Gordon, "Reagan Declares U.S. Is DismantlingTwo Nuclear Subs," New York
Times,May 28, i986; also see Gordon, "U.S. Still Divided on I979 Arms Treaty,"New York
Times,April 23, i986. When the U.S. did break the limits,some criticsheld thatone motive
was to show thatthecontroversy overtheIran armsdeal, whichhad justbeen revealed,would
not inhibitthe administrationfromactingdecisively.Michael Gordon,"U.S. Exceeds Limit
Set in I979 Accord on StrategicArms,"New YorkTimes,November29, i986.
12 JorgeDominguez, "Cooperating with the Enemy: U.S. ImmigrationPolicies toward
Cuba," in ChristopherMitchell,ed., Immigration Policyand U.S. ForeignRelationswithLatin

America(forthcoming).For additional examples,see Glenn Seaborg with Benjamin Loeb,
StemmingtheTide (Lexington,MA: LexingtonBooks, I987), I2, I58.

preferences and behavior.The operating incentives aregivennotonlyby

the presentcircumstances, but also by how thesecircumstances came
about.Wheretheplayersare is strongly influenced bywheretheyhave
been.'3Thus theU.S. decisionto respondwithforceimmedately to the
seizureof theMayaguezis to be explainedlargelyby thefactthatthis
eventoccurredshortly afterthefallof Vietnam,whenU.S. leadersbe-
lievedthattheyhad toactforcefully to showothersthatthepreviousde-
feathad not underminedAmericanwillingness to use force.Similarly,
one reasonwhyBritaincouldnotcooperatewithGermanyand restrain
Russiain 1914 was thatshehad doneso in thepreviouscrisisand feared
thata repetition woulddestroytheTripleEntente.In othercases,defec-
tionor cooperation in one interactionwilllead to moreofthesame.Co-
operationcan changethe situationtheactorsface,and theirbeliefs,in
ways that-makefurthercooperationmore likely.Perhapsmore fre-
quently,theprocessof mutualdefection will lead each side to fearand
disliketheother,and to developdistorted viewsof it,whichcannotbe
quicklyreversed iftheothersuddenlybeginstocooperate.'4 Actorsdo not
reactmerelyto theimmediate stimulustheyface.'5
We oftentalkofrepeatedplaysofa Prisoners' Dilemma.Butthisfor-
mulationis misleading whenthepreferences and beliefsoftheactors,and
thenatureofthegameitself, changeas it is played.Whatis at stakeand
thenatureof theissueis definedovertime,as actorsdeveloptheirposi-
tions,in partin responseto thepositions takenbyothers.As one Senator
explainedhisvotetopermittheexportofadvancedarmstoSaudiArabia:
The wholeissuechangedin thelastweekor io days.The mediabeganto
playit up as a questioninvolvingthepresident's
abilitytoconductforeign
policy.... Had it notbeenforthemediahype,theissuewouldhavebeen
strictlythe arms sale. And I would have been verycomfortable voting
againstit."''6
I For a nice example taken fromthe patternof social relationswithina

communityof
chimpanzees,see JaneGoodall, The Chimpanzeesof Gombe (Cambridge: Belknap Press of
Harvard UniversityPress, i986), 337-38.
1 See the discussionof the "spiral model" in RobertJervis,
Perceptionand Misperceptionin
International
Politics(Princeton:PrincetonUniversityPress,I976), 62-84. For argumentsthat
one or two instancesof cooperationare not likelyto be sufficient
to break thesespirals,see
Charles Osgood, An Alternativeto War or Surrender(Urbana: Universityof Illinois Press,
I962).
15For a discussionof a parallel in biologicalevolution,see StephenJayGould, "Not Nec-
essarilya Wing," NaturalHistory94 (November i985), 12-25. Also see Gould, "Of Kiwi Eggs
and the LibertyBell," NaturalHistory95 (November i986), 22-29. The argumentthatstruc-
turesshape and limitlaterdevelopmentis in sharpcontrastto the standardevolutionaryas-
sertionthat naturalselectionoperatesin a way thatprovidesthe bestfitbetweenorganisms
and theirpresentenvironment.
6 Quoted in RichardFenno, Jr.,"Observation,Context,and Sequence in theStudyof Pol-
itics,"AmericanPoliticalScienceReview 8o (March I986), I I-I2.

322 WORLD POLITICS
The moves in a game can change it. Chicken can become Prisoners'Di-
lemma if each side's behavior leads the other to believe that being ex-
ploited would be worse than mutual defection.Schellingnotes: "When a
boy pulls a switch-bladeknifeon his teacher,the teacheris likelyto feel,
whatever the point at issue originallywas, that the overridingpolicy
question now is his behavior in the face of a switch-bladechallenge."'7
Other objectionscan be grouped under threeheadings,each of which

has threecomponents.The firstsetof questionsconcernstheactors'pref-
erences,which need to be probed and explained. The second deals with
the conceptsemployedby the framework,which may turnout to be am-
biguous or troublesome. The third set involves the causes and conse-
quences of the actors' perceptions,beliefs,and values. The representa-
tions of reality employed often beg crucial questions. Like many
experiments,the models oftengain internalvalidityat thecostof external
validity:the order theyimpose is too rigid to catch the realitytheyseek
to explain.
PREFERENCES AND PREFERENCE ORDERS
As noted, the central question for the work on anarchy is how co-
operationis possiblewhen actorsare in a Prisoners'Dilemma-i.e., when
theyhave the followingpreferenceorder: exploitingor takingadvantage
of the other,mutual cooperation,mutual defection,and being exploited.
The most importantissue may be not what happens afterthese prefer-
ences have been established,but the preferencesthemselves.Much of the
explanatory"action" takes place in the formationof the preferences;we
cannot affordto leave thistopic offstage.Three questionsneed to be ad-
dressed: How do we know what the actors'preferencesare? What is the
patternof distributionof preferencesover large number of cases? How
are preferencesestablished?
WHAT ARE THE ACTORS' PREFERENCES?
The firstquestion is the most obvious one. How do we know whether

a situationresemblesa PD? The problem is even greaterforthose var-
iantsof the frameworkthatcall forcardinalratherthanordinal utilities.'8
The danger, of course, is that we will inferthe actors' preferencesfrom
ofConflict(Cambridge:Harvard UniversityPress,i960),
I7Thomas Schelling,The Strategy
265.
i8
This is trueforJervis(fn. i).

theirbehavior.Actorsrarelygive completestatements of theirprefer-

ences.In somecaseswe can usethemethodof"revealedpreferences," but
thistechniquecan be used onlywhenpreferences are stableand consis-
tent;if it is not to be tautologous,
we mustexaminea largenumberof
instancesin ordertotestthepreferences on behaviors
thatwe did notuse
toderivethepreferences in thefirstplace.
Furthermore, while the standardPD model pointsto fourpossible
outcomesthatneed to be ranked,decisionmakersmaydefinethesitua-
tiondifferently-most frequentlybyignoringthepossibility of mutual
restraint. In 1914, forexample,most leadersdid not ask themselves
whethertheypreferred peace to war becausetheydid not thinkthat
peace could be maintained.In somecases,a lack of cooperation maybe
explainedin significant measurebytheactors'inattention tothepossibil-
ityofsuchan outcome.
How COMMON Is PRISONERS'DILEMMA?
The secondproblemliesin determiningtherelativefrequency ofvar-
iousgames.How commonis Prisoners'Dilemma?HarrisonWagnerand
GeorgeDowns and hiscolleaguesnotethatstatesoftenfailto cooperate,
notbecausetheycannotsurmount thePD, butbecausetheyarein Dead-
lockand prefermutualdefectionto mutualcooperation.'9
For example,
were the statesin I914 in PD or in Deadlock? (These are not,of course,
theonlyalternatives.)Are we in PD withtheSovietUnion today?Ronald
Reagan is not alone in answeringin thenegative;much of thedebate over
whetherAmerican threatshave deterredthe Soviet Union or createdun-
necessaryconflictis a disagreementabout Soviet intentionsand prefer-
ence orders,not a dispute betweentwo contendinggeneraltheories.20
To put the questions in theirmostgeneral form:How much of inter-
national conflictis caused by the states' inabilityto make and enforce
bindingagreements?How much of internationalpoliticsis drivenbythis
problem? The model of PD may be popular, not because it catches the
most importantdynamics of internationalpolitics,but because it is in-
triguingand lends itselfto interestingmanipulations.The "law of the in-
strument"may be at least partlyat work. Furthermore,themodel is con-
gruentwith the Anglo-Americanbias of seeing world politicsas tragedy
19Harrison Wagner, "The Theory of Games and the Problemof InternationalCoopera-

tion,"AmericanPoliticalScienceReview70 (Junei983), 330-46; George W. Downs, David M.
Rocke,and Randolph M. Siverson,"Arms Races and Cooperation,"WorldPolitics38 (October
I 985), I I 8-46.
20 Jervis
(fn. I 4), 84-I I 3. Similarly,manyof theargumentsabout whetherthe failureof de-
tentewas inevitableor a matterof errorscan be phrasedin termsof whethertherelaxationof
tensionswas Pareto-superiorto a highlevel of competition.

324 WORLD POLITICS
ratherthanas evil,ofbelievingthatmostconflicts
canbe ameliorated
for
the good of all concerned.21
The empirical problems in answering these questions are relatively
clear, although extremelydifficult.Perhaps equally troublesomebut less
obvious are two conceptualdifficulties.
First,are we concernedonlywith
situationsin which the agreementscannotbe enforced,or also withthose
in which no outside agency is available to forcethe actors to reach an
agreementin the firstplace? Even where enforcementis available, as it is
in domestic society,some mutually beneficialoutcomes may be missed
because of bargaining dynamics and miscalculations.Similarly,the di-
lemma of public goods is not that actors cannot ensure that otherswill
live up to their commitments,but that it is not rational for anyone to
make a commitmentto contributein the firstplace.
A second problem is thatwhetheror not a situationis a Prisoners'Di-
lemma depends in part on how we defineits boundaries.If we look at
Japanese-Americannegotiationsin the summer and fall of 1941, the
frameworkdoes not apply: both sides preferredmutual defection(in this
case, war) to the concessions that would have been necessaryto reach
agreement.Japanwould ratherfightthan give up the effortto dominate
China and South East Asia; the United States would ratherfightthan
permitthisto happen. But if we step back and ask why Japansoughtits
sphereof influence,anarchyand the Prisoners'Dilemma emergeas cru-
cial. Domination was soughtprimarilyas a means to thegoal of autarky.
Japaneseleaders wanted theself-sufficiencythatwould go along withun-
hindered access to Asian marketsand raw materialsin order to become
relativelyimmune fromWesternpressures.22 Had it been possibleforthe
West and Japanto make a bindingagreementgivingthe lattereconomic
freedom in Asia in returnfor renouncing the use of force,a bargain
mighthave been struck.
SOURCES OF PREFERENCES
By taking preferencesas given, we beg what may be the most impor-

tantquestion on how theyare formed.When and why do actorsfindex-
ploitationparticularlyattractive,or the danger of being taken advantage
of particularlyunacceptable?When and whydo decisionmakerssee mu-
tual cooperation as desirable? Economic theorytreatstastesand prefer-
2-See Arnold Wolfers,Discordand Collaboration(Baltimore:The JohnsHopkins Univer-
sityPress,I962), chap. I5.
22 JamesB. Crowleyjapan's QuestforAutonomy: and ForeignPolicy,ig30-
NationalSecurity
1938 (Princeton:PrincetonUniversityPress,i966); Michael Barnhart,JapanPreparesforTotal
War (Ithaca, NY: Cornell UniversityPress,i987).

ences as exogenous. Analysis is thereforefacilitated,but at the cost of

drawingattentionaway fromareas thatmay containmuch of theexplan-
atory"action" in which we are interested.23
Sometimes we can deduce preferencesfromthe structureof the sys-
tem, as Realism suggests. But even a structuraltheoryof international
politicsas powerful as Waltz's has troubleproducingmany precise de-
ductions.24As the endless argumentsabout the national interestremind
us, only rarelycan descriptionsand prescriptionsof what the statewill or
should preferbe drawn fromitsobjectivesituation.25 Conybeareis able to
deduce a state'seconomic interestsfromtradetheory;26 but thisinference
onlyapplies ifwe assume thatthe stateis an actor.Such an assumptionis
often a valid guide when overridingissues of national securityare at
stake; but the relevantactorsforeconomicissuesare oftenclasses,sectors,
and groups, or even smaller and more numerous units. In the security
area, the preferenceforprotectingthe statusquo ratherthan retreating
may be deducible fromthe externalsituation,but the choice forexpan-
sion is often the result either of preferencesof importantsubnational
groups or of internalbargaining.27 More broadly,one does not have to be
a Wilsonian or a Marxistto argue thatthegoals statesseek,the coststhey
are willing to pay,and the instrumentsthatare believedappropriatecan
be deeply affectedby the natureof the state.
Transnational forcesare a thirdgeneral source of preferences.Thus,
JamesJollsees the prevalenceof Social Darwinism in the early20thcen-
turyas a major cause of the preferenceforcompetitionin world politics,
and Van Evera argues thatrampantnationalismstronglyinfluencedthe
utilitiesthat attached to various outcomes.28The maddeningly vague
"spiritof thetimes"may help to accountforthefactthat,duringtheyears
beforeWorld War I, all the major powers opted foroffensivemilitary
23See the discussionof tastesin JamesMarch,"Bounded Rationality, Ambiguity,and the

Engineeringof Choice," Bell JournalofEconomics9 (Autumn I978), 593-604.
24 Kenneth Waltz, TheoryofInternational Politics(Reading,MA: Addison Wesley,I979).
25 But see StephenKrasner,Defending theNationalInterest (Princeton:PrincetonUniversity
Press, I978). Also see David Lake, "Beneath the Commerceof Nations: A Theory of Inter-
national Economic Structure,"InternationalStudiesQuarterly28 (June i984), I45-49, and
Lake, "Power and the Third World: Toward a Realist Political Economy of North-South
Relations,"International StudiesQuarterly3I (Junei987), 22I-28.
26 JohnConybeare,Trade Wars(New York: Columbia University Press,i987).
27See JackSnyder,MythsofEmpire,forthcoming.
28 JamesJoll,"19I4: The Unspoken Assumptions,"in H. W. Koch, ed., The Originsof the
FirstWorldWar (London: Macmillan, I972), 307-28; Koch, "Social Darwinismas a Factor in
the 'New Imperialism,'" ibid., 329-54; Stephen Van Evera, "Why Cooperation Failed in
I9I4," WorldPolitics38 (October i985), 8o-I 7. For a relatedgeneralargumenton thesource
of preferences, see Aaron Wildavsky,"Choosing PreferencesbyConstructionInstitutions:A
Cultural Theory of PreferenceFormation,"AmericanPolitical Science Review 8i (March
i987), 3-22.

326 WORLD POLITICS
strategieswhen defensiveones would have been more likely to make
them reach theirforeignpolicygoals.29
Preferencesalso stem fromthe ideologiesand beliefsof individualde-
cisionmakers.Some are "hard-line,"others"soft-line"in dealing withan
adversary.Whether these orientationshold across interactionswith a
range of adversarieswe do not know, althoughsome evidence indicates
that they do.30 The effectis that under circumstancesin which some
statesmenwill believe that mutual cooperationis beneficial,otherswill
see it as a trap. Thus, because of theirbeliefsabout the natureof the ad-
versary,Ronald Reagan and his supportersbelievethatmanykindsof co-
operationwiththe Soviet Union are likelyto producegreaterSovietpres-
sures on the West ratherthan furthercooperation.
Herbert Simon's argumentabout Duverger's Law of the relationsbe-
tween voting rules and party systemsapplies to theorizingin interna-
tional politicsas well: "Most of the work is being done by propositions
thatcharacterizethe utilityfunctionof the [actor]and his or her beliefs,
expectations,and calculations.'31 Realism and game theoryare of limited
help here since thelattermustassume theactors'preferencesand utilities,
and Realism's generalizations,althoughpowerful,are oftenvague. Fur-
thermore,theseapproaches implythatpreferencesare constantwhen, in
fact,they change and thus pose a range of challenges for the anarchy
framework.32 First, and most obviously,a state's preferenceschange as
one set of decision makers replaces another.Thus the dim prospectsfor
arms control throughoutmuch of the Reagan administrationare ex-
plained by the President'sbeliefthatagreementsthatotherssaw as ad-
vantageous are in factunacceptable.Game theoryfocuseson the prefer-
29 Van Evera (fn. 28); Van Evera, "The Cult of the Offensive and the Originsof the First
World War," InternationalSecurity9 (Summer i984), 58-i07; JackSnyder,"Civil-Military
Relationsand theCult of theOffensive,I9I4 and i984," ibid.,i08-46; Snyder,The Ideologyof
theOffensive: MilitaryDecisionMating and theDisastersof 1914 (Ithaca,NY: Cornell Univer-
sityPress, i984). For a rebuttal,see ScottSagan, "1914 Revisited:Allies,Offense,and Insta-
bility,"International Securityi i (Fall i986), I5I-76, and theexchangeof lettersbetweenSny-
der and Sagan, ibid., i i (Winter I 986/87),I 87-98.
30 For a discussionof this issue in the domesticcontext, see RobertPutnam,The Beliefsof
Politicians(New Haven: Yale UniversityPress,I973).
3 Simon, "Human Nature in Politics:The Dialogue of Psychology and PoliticalScience,"
AmericanPoliticalScienceReview 79 (June i985), 298; also see William Riker,"The Heres-
theticsof Constitution-Making:The Presidencyin I787, with Commentson Determinism
and Rational Choice," AmericanPoliticalScienceReview78 (March i984), i-i6. In a different
intellectualtradition,Adam Przeworski comes to a similarconclusion:"Marxism and Ra-
tionalChoice," Politicsand SocietyI4 (No. I, i985), 379-409.
32 For a general discussionof changingpreferences, see JamesMarch and HerbertSimon,
Organizations(New York: Wiley, I958), I4I; Michael Cohen and RobertAxelrod,"Coping
with Complexity:The Adaptive Value of Changing Utility,"AmericanEconomicReview74
(March i984), 30-42; Barbara Farnham, "Value Conflictsand Political Decision-Making"
(Ph.D. diss. in progress,Columbia University).

ences thatare held at any giventime,and therefore simplyputsthis

questiontoone side;Realismimpliesthatpreferences comefromthepo-
sitionof thestatein theinternational system,and therefore impliesthat
changesin leadershipdo notmatter.Realismis lesstroubledbya second
sourceofchange:thatproducedbyshifts in theexternalsituation.
On the
simplestlevel,a previously desiredoutcomemaybecomeunacceptable
becausethestatewould no longergain by it. Thus, forinstance,states
beginto resistmutuallow tariffs as theircompetitive positiondeterio-
rates.Similarly,in the fallof i96i Americandecisionmakerscame to
doubtthewisdomoftheirown nucleartest-ban proposalwhentheysus-
pected"thattheRussiansmighthavedrawnevenwithor evenpastus in
someaspectsofthermonuclear weapons."33
In othercases,preferenceschangethroughtheprocessoftheinterac-
tionitself.When two statesare hostile,one will oftenassumethatany-
thingtheotherurgesmustbe bad; an outcomethatwas previously fa-
vored will be viewed with skepticismif the adversaryendorsesit.
Similarly,thefactthata statetakesa stronginterest in an issuecan lead
theadversary to developa contrary preference.The conflict processthen
generatesinterests
and preferences ratherthanbeingproducedbythem.34
It would otherwisebe impossibleto explainwhytheUnitedStatessup-
portsJonasSavimbiin Angola,or refusesto admitrefugees fromCuba.
Unlike other communistregimes,Cuba wantsits internalopponentsto
leave;so theU.S., counterto itsnormalpolicyofacceptingsuchpeople,
keepsthemout.35 Althoughthisinterdependence betweenstates'prefer-
encesis compatible withRealism,itleadstointerests and maneuvers that
complicate,if not contradict,
thenormalprescriptions producedby the
anarchyframework. For example,splitting up a largetransactionintoa
seriesofsmallonesmaybe ineffective in thesecircumstances.
Experienceand knowledgecan also changepreferences. Although
suchprocessesaccountformanyimportant outcomes, theyareexogenous
to game theoryand theanarchyframework. As actorsparticipate in an
arrangement, theymaycometo see thattheconsequences are quitedif-
ferentfromthosetheyexpected.Thus theexperienceof the follow-up
conferenceson the HelsinkiAgreementseemsto have convincedthe
Sovietsthatthe discussionof humanrights,whileannoying,does not
33Glenn Seaborg,Kennedy,Khrushchev, and the TestBan (Berkeley:Universityof Califor-

nia Press, 198I), I 20.
34 Of course,each statewantsto limittheother'spower and influence
and will thereforetry
to see thatanyactortheadversaryis supportingdoes notwin; butfromthiswe cannotdeduce
preferencesforspecifiedcountriesor factionsto prevail.
35Linda Greenhouse,"U.S. Assailed Again on Curbing Cuban Immigrants,"New York
Times,September27, i986.

328 WORLD POLITICS
do as muchharmas theyoncebelieved.36 In othercases,theemergence of
newscientific information maybe thesourceofchange.The U.S. position
in thetest-ban negotiations was alteredwhenpeoplerealizedthattesting
mighttake place insidea largecavern,thusmuffling theblastand un-
dercutting theverification techniquesin whichtheWestpreviously had
faith.New information or newbeliefscanalso lead topreferences forco-
operation.RobertRothstein argues,forexample,thatNorth-South con-
flictovercommodity pricingdecreasedas thedevelopingstatescame to
see thattheirpreviousproposalsforpricestabilization mightnotincrease
theirincome,as theyhad initiallybelieved.37 Negotiations proceednot
onlyby bargaining, but also bypersuasion;in somecases,thatleads to
solutionsthatneithersidehad previously thought ofand thateachcomes
to preferto itsoriginalproposal.38
Finally,preferencesmaybe unstable.Becausetheintellectual problems
are great,keydecisionsare oftendifficult, and continuedthoughtmay
produceshiftingevaluations.Althoughsuch processesare difficult to
generalizeabout,mostpeopleknowthemfromtheirownexperience: de-
cidingwhatcartobuy,whatjob totake,orwhomtomarry. Mostsharply,
decisionmakersmaycome to prefermutualcooperation to mutualde-
fection,or viceversa.Thus, theSoviet-American negotiationsforarms
controlin the Indian Ocean broke down largelybecausethe Carter
administration, afterfirstfavoring an agreement, cameto believethatit
would notbe in theAmericaninterest. Similarly, in theearlyi96os, the
UnitedStatesfavoredland-mobile then(inSALT I) proposedpro-
missiles,
hibitingthem,thenfavoredthemas stabilizing, and has recently again
calledforbanningthem.The samekindofchangeshaveoccurredin the
U.S. stancetowardantisatellite weapons.Those shifts can onlyin partbe
explainedbychangesin technological or Sovietactivities.39
possibilities
36 JohnMaresia, "Helsinki," in Alexander George, Philip Farley,and Alexander Dallin,
eds., U.S.-SovietSecurityCooperation:
Achievements,Failures,Lessons(New York: OxfordUni-
versityPress, forthcomingi988). Also see Raymond Garthoff,Detenteand Confrontation
(Washington,DC: BrookingsInstitution,i985), 480. For an argumentthatpeople's prefer-
ences are oftenformedby theirbehavior,see Daryl Bem, "Self-PerceptionTheory,"in Leon-
ard Berkowitz,ed., Advancesin ExperimentalSocial Psychology, VI (New York: Academic
Press, I972); foran applicationto internationalpolitics,see Deborah Larson, The Originsof
Containment (Princeton:PrincetonUniversityPress,i985).
37Rothstein, "Consensual Knowledge and International Collaboration:Some Lessons from
theCommodityNegotiations,"International Organization38 (Autumn i984), 732-62.
38 See Mary Parker Follett,DynamicAdministration: The CollectedPapersof Mary Parker
Follett,H. C. Metcalfand L. Urwick,eds. (New York: Harper & Row, I942); RichardWalton
and RichardMcKersie,A BehavioralTheoryofLabor Negotiations (New York: McGraw-Hill,
i965), I26-83; ErnstHaas, BeyondtheNation-State(Stanford,CA: StanfordUniversityPress,
I 964), 86-I 25.
39For a discussionof the ways in which changesin technologyinfluencedAmericanatti-
tudes toward antisatelliteweapons, particularlyby providingsatelliteswithoffensiveas well

In summary, theresearchon cooperationunderanarchyassumesthe

in
actorsare PD, whichmay not be correct.It further
takestheactors'
preferences
as givenand ignoreshow,why,and whentheychange.Game
theorycannothelphere;Realism'sanalysesofthesequestionsarelimited
at best.
CONCEPTS
The conceptsemployedbytheanarchyframework seemunproblem-

aticat firstglance.Cooperationand defection, and defense,and
offense
powerare fairlystandardconcepts,but the requirements forrigorous
analysiswithina game-theoreticapproachdemandthatwe squeezesome
of therichnessout of theseterms.The admirablegainin precisionmay
haveto be purchasedat an unacceptablyhighprice.
COOPERATION AND DEFECTION
The conceptsof cooperationand defectionare crucialto the frame-

work.These termsworkwell fora laboratory Prisoners'Dilemma,but
mostsituations are morecomplex.To startwith,are thesetheonlytwo
alternatives?Perhapswe shouldthinknotofa dichotomy, butofa con-
tinuum.But can we add thismeasureof realismwithoutsacrificing the
parsimony and deductivepowerof thetheory?Furthermore, can most
alternativesreallybe arrayedalongsucha continuum? Somepoliciesex-
pressa highdegreeof bothcooperationand defection simultaneously;
others-in some instances, thepolicyof isolation-expressneither.40In
manycasesthemostinteresting choicesare notevenon thiscontinuum.
Concretequestionsoftenariseabouthow to characterize anyindivid-
ual outcomeor policy.Sometimes thesequestionscome down towhether
theglassis halffullor halfempty, buteventhentheansweris notincon-
sequential.An exampleofan outcomeis the"chickenwar"betweenthe
UnitedStatesand theE.E.C. in theearlyi96os, whichConybeareseesas
mutualdefection becauseEuroperaisedthetariff on frozenpoultry and
the U.S. retaliatedby raising tariffson some European products.40 But
one can arguethatwhatis moresignificant
thanthedeviationfrompure
cooperationis thefactthatthedisputewas heldto one roundof action
as defensivemilitarycapabilities,see Paul Stares,TheMilitarization

ofSpace (Ithaca,NY: Cor-
nell UniversityPress, i985), and Steve Weber and SydneyDrell, "Cooperationand Discord
in the Militarizationof Space: U.S. Strategy,i960-i985," in George, Farley,and Dallin (fn.
36).
4? Thus it is not surprisingthatquantitativestudiesfindthatconflictand cooperationare
notalways inverselyrelatedto each other.
4 Conybeare(fn.26).

330 WORLD POLITICS
and counteraction. It did not producea spiralof conflict,and relations
betweenthecountries werenotembittered. An exampleofan ambiguous
military policyis thequestionwhethertheU.S. cooperated or defected
in
itsstrategicarmsprocurements betweenthemid-ig6osto thelate 1970s.
On theone hand,theUnitedStates,unliketheSovietUnion,did notde-
ployadditionalmissiles.On theother,itdid makemajorqualitativeim-
provements, mostnotablyin theformofmultiplewarheads.
How we judgea policymayalsodependon thetimespanwe examine.
Defectionin one instancecan producemutualcooperationover the
longerrun.Indeed,one implication of thetheorizing aboutanarchyis
thatcooperationis enforcedbythepossibility of defection.Thus, Presi-
dentReaganarguesthathispolicywillinduceSovietcooperation, which
willthenbe mutual.If he turnsoutto be correct, is itmoreusefultocall
thewholepolicyone of cooperation, or to divideit intodiscreterounds
and label theinitialAmericanbehavioras defection? And iftheSoviets
havebeenconciliatory in orderto weakentheWest,and willultimately
returnto an uncooperative stancefroman improvedposition, shouldthe
policynotbe characterized as defection?A simplerversionofthisambi-
guityis revealedin theCuban missilecrisis.This episodeis oftenconsid-
ered an American victory;in fact,mutual cooperationgreatlyincreased
in themonthsthatfollowed.If we takean evenlongerperspective, how-
ever,we mightlabel the outcomeas mutualdefection, becauseof the
Sovietarmsbuild-upthatwas causedpartlybythehumiliation in Cuba.
In gametheoryor in thelaboratory, we can definewhateach playofthe
gameis. But in thesituations we are tryingtoanalyze,thesedistinctions,
and theanswerto thequestionwhetherthebehavioris seenas coopera-
tiveor not,are subjectiveifnotartificial.
In othercases,theverymeaningof cooperation is unclear.It usually
denotesdoingwhattheotheractorprefers, buthowdo we characterize a
responsethatis undesiredbutis designedto benefit theother,and/orhas
thateffect?How do we classifybehaviorwhenone side desiresa high
degreeoffriction withtheotherand theotherrespondswiththesought-
forhostility?What do we sayaboutcases in whichneitherside thinks
abouttheimpacton theother?Does thenotionofdefection implysome-
thingmore than,or different from,noncooperation? The labels may
squeeze out muchoftherealitythatwe are trying to catch.A recentex-
ample displays severalof these In
ambiguities. April i986, theUnited
StatesaskedEast Germanytorestrict thetravelofLibyandiplomatswho
mightbe aidingterrorism in theWest.In response or whattheyas-
sertedto be response theEast Germansdemandedthatall diplomats
exceptthoseofBritain,France,and theU.S. showtheirpassports as they

crossed fromEast to West Berlin. (Previously,only special passes issued

by the East German Foreign Ministryhad been required.)The U.S. took
this action, not as cooperation,but as "a clearlyunfriendlynegativeact
that is intended to undermine the four-powerstatus" of Berlin.42This
interpretation is not unreasonable,but neitheris itcompelledbythe facts.
The problemof the subjectiveand politicalnatureof definitionsof co-
operationis compounded in the cold war because thetwo sides have very
differentperspectives.Most Americansthinkabout cooperatingwiththe
U.S.S.R. to maintainthe statusquo or to bringabout mutuallybeneficial
changes, particularlyin arms control.But the Soviets not only consider
the statusquo in the sense of the currentdistributionof influenceas dis-
advantageous,but believe thatit cannotbe maintainedbecause it is being
moved by the forcesof history.Thus, fortheSoviets,CC consistsof man-
aging the changes in a peaceful way. While the United States would
steadilylose ground, the Soviets argue that Washington should accept
thisprocessbecause the alternativewould be even worse.43
Even when such problemsare absent,it is easy to overlookthefactthat
what is deemed a defectionis in partrule-governed.The United States-
and mostof the world-considered theSovietemplacementofmissilesin
Cuba in i962 as stronglynoncooperative;the Soviet arms build-up of the
late i96os, which had a much greaterinfluenceon the militarybalance,
was seen as less of a defection.Although what Waltz calls "internalbal-
ancing" may not be welcomed byothers,it is acceptedas a normal partof
internationalpolitics.It is seen as legitimate;indeed,coerciveattemptsto
block it would be perceivedas illegitimateand highlythreatening.These
distinctionscannot be understoodby measuringtheobjectiveharm done
to the state.The factthatthe actorsusuallygive littlethoughtto the rules
does not reduce theirimportancein definingthe crucialconceptswe use.
Some of these difficultiesare summarized by the question a student
raised when I played a version of multipersonPrisoners'Dilemma in
class: what, she asked, are we to cooperateabout? For decision makers,
the question is never cooperationor defection,but ratherwhat goals to
seek and the tacticsthatwill be mostapt to reach them.Of courseit is not
illegitimateto impose our categorieson the behaviorof actors and the
notions of cooperation and defectionare hardly foreignto them but
42 JamesMarkham,"Allied Diplomats DefyEast GermanControls,"New YorkTimes,May

28, i986. For paralleldiscussionof an importantcase in thelate i8thcentury,
see Paul Schroe-
der,"Old Wine in New Bottles:RecentContributionsto BritishForeignPolicyand European
InternationalPolitics,1789-I 848,"JournalofBritishStudies26 (JanuaryI987), Io.
43See Garthoff(fn.36), 38-50, i069. The Japanesehad a similarconceptionof cooperation
withthe Britishin China in the I930s: see Paul Haggie, Britanniaat Bay (Oxford:Clarendon
Press,i98i), I26.

332 WORLD POLITICS
when actorsdefinetheirproblemsand choices quite differentlyfromthe
theoreticalmodel, we may be forcingdisparatebehaviorintoan inappro-
priateframework.
OFFENSE AND DEFENSE
Questions can also be raised about the conceptsof offenseand defense.

Both logic and evidence indicate that the prospectsforcooperationare
increased to the extentthatdefensivemilitarysystemsand strategiescan
be distinguishedfrom,and are strongerthan,offensiveones.44In princi-
ple, the differencebetween offenseand defense is clear: the formerin-
volves the abilityto attack the otherand seize his territory; the latterthe
abilityto repel such an attack and protectone's own territory. The issue
of whetherattackingor defendingis preferablecomes down to whether
a decision maker who believed that war was inevitablewould preferto
attack the other side, as in I914, or be on the defensivewhile the other
side attacked which,as the historyof World War I shows,would have
been a betterstrategy.45
But questions arise. Even if it is possible to say whetherthe offenseor
the defensehas the advantage,can one distinguishbetweenoffensiveand
defensiveweapons and strategies?Much of theinterwardisarmamentef-
fortswere devoted to doing so. The factthatstatesmenworked hard on
this question indicates that theybelieved that a positiveanswer was at-
tainable; thatsuch effortsyieldsuccessonlyoccasionallyindicatesthatthe
endeavor may be doomed althoughin the interwarcase thereasonmay
have been the refusalto forgooffensiveoptions ratherthan the inability
to make the relevantdistinctions.Similarly,George Downs and his col-
leagues findthatarms races are not terminatedby the shiftto defensive
systems46 which indicatesthatstatesmenare eitherunwillingor unable
to avoid threateningotherswhile seekingto maximize theirown security.
Can the conceptsof offenseand defensebe applied to nuclear weap-
ons? A frequentargumentis thatthe common-sensedefinitionshave to
be turnedon theirheads. That is,offenseis the abilityto take one's cities
out of hostage; conversely,the abilityto destroythe otherside's popula-
tion and othervalues, previouslyassociated withthe offense,is now con-
44See Jervis(fn. i), at I86-2I4; George Quester,Offense Sys-

and Defensein theInternational
tem (New York: Wiley, I977); Jack Levy, "The Offensive/Defensive Balance of Military
Technology: A Theoreticaland HistoricalAnalysis,"International StudiesQuarterly 28 (June
i984), 2I9-38; Van Evera (fns.28 and 29); Snyder(fn.29); BarryPosen,The SourcesofMilitary
Doctrine(Ithaca, NY: Cornell UniversityPress,i984); Weber and Drell (fn.39).
45See the literaturecited in fn.29 above. 46 Downs, Rocke, and Siverson(fn. i9).

sidered defensive(as long as the other side has a similarcapability)be-

cause such an act could be crediblythreatenedonly as retaliationforthe
other's attack. The obvious implicationsare that superpowerstrategies
designed to deterwar by developing counterforcecapabilitiesexacerbate
the securitydilemma and make cooperation more difficult,and that
countervaluetargetingwould permitmutual security.But such paradox-
ical reasoninghas not been accepted by all analysts,with the resultthat
thereis no consensusabout how the conceptsfitthe currentsituation.
Even if the distinctionbetween offenseand defenseworks,at least to
some extent,in the militaryarena, can it be transferred to otherdimen-
sionsof internationalrelationships?47 To findtheanswer,one would have
to ask under which conditionsbandwagoningas opposed to balancing is
likely to take place.48The formerdynamicsmake the offensivepotent;
the lattersupportthe defensive.When initialgains are expectedto create
positive feedback,each state will have strongincentivesto defect,irre-
spective of whether its ultimate intentionsare aggressive;moving first
may endanger the other,but it is necessaryin order to protectthe state.
As a result,mutual restraintwill be difficultto maintaineven ifbothsides
are satisfiedwith the statusquo. By contrast,when balancingprevailsand
gains by one side call up counteractingpressures,states can affordto
await developments,making major effortsonlyon the relativelyrare oc-
casions when another power poses a direct threat.Cooperation should
thus be more prevalentwhen bandwagoning is neitherfearednor hoped
for.
The most obvious example of bandwagoning-or at least the expec-
tationof it-is the "domino theory."When a defeatin one countryis ex-
pected to have major and deleteriousconsequences forthe state'sinflu-
ence in other areas, then even minor threatsmust be met swiftlyand
firmly.On the other hand, when one side believes thatthe other side's
local victoryis likely to be contained-for example, by the effortsof
neighborswho are alarmed by the new threat-then it can affordto try
cooperatingwith the otherbecause the costsof being taken advantage of
are relativelylow. The politicaldefensiveis potent;the securitydilemma
47For one effortat doing so, see RobertJervis,"From Balance to Concert:A Studyof In-
ternationalSecurityCooperation,"WorldPolitics38 (October i985), 58-79,at 62-64. For a fur-
ther discussion,see Jervis,"Cooperation Under Anarchy: Problems and Limitations,"in
Alker and Ashley(fn. i).
48 Arnold Wolfers,"The Balance of Power in Theory and Practice,"in Wolfers(fn. 2I),
I22-24; Waltz (fn. 24), I25-28; Stephen M. Walt, "Alliance Formationand the Balance of
World Power," International Security9 (Spring i985), 3-43,and The OriginsofAlliances(Ith-
aca, NY: Cornell UniversityPress,i987); JackSnyderand RobertJervis, eds.,Strategic
Beliefs
and SuperpowerCompetition in theAsianRimland,forthcoming.

334 WORLD POLITICS
is ameliorated; and a statecan gain a high degree of protectionwithout
proportionatelyunderminingthe interestsof itsadversary.
Argumentslike theseneed much more developmentbeforewe can be
sure thatthedistinctionbetweenoffenseand defenseis possibleand help-
ful in the politicalarena. What we have so farare essentiallyexploratory
probes; the results,while suggestive,are hardlydefinitive.Furthermore,
we have not even tried to apply such an extensionto the realm of inter-
national economic relations.
THE NATURE OF POWER
A third conceptual difficultyis created by the factthat in many en-

countersthe main stake is power-a notoriouslydifficult concept.But in
almost all formulations,power in internationalpoliticsis relativerather
than absolute.49When statesthinkabout the possibilityof an armed con-
flict,theyhave to judge how theirforcescompare to thoseof theiradver-
saries; the absolute sizes of the forceson each side are irrelevant.This is
trueformanypoliticalconflictsas well. Knowing how much leverageone
state has over anothertells statesmenand analystsverylittleunless they
also know how much leveragetheotherstatehas. Thus, itcan be rational
forstatesmento act in ways thatreduce theabsolutelevel of benefitsthey
receive.
This view is not only compatiblewith Realism, but is embedded in it.
But it shows that,as long as power is central,an elementof inherentcon-
flictwill be involved,thus complicatingstatesmen'sattemptsto establish
cooperation and undercuttingsome of the prescriptionsthatcan be de-
duced from the anarchy framework.When statescare primarilyabout
maximizing their power advantage and not about absolute gains and
losses,many of the strategiesand conditionsthatshould lead to coopera-
tionin a Prisoners'Dilemma no longerproduce thisresult.5o Thus Robert
49 Nuclear weapons mayhave changedthis,as is indicatedbythetitleand essaysin Bernard
Brodie et al., The AbsoluteWeapon(New York: HarcourtBrace, I946).
5? The implicationsof the factthatstatesoftenseek relativeratherthanabsolutegains have
been discussed by ArthurStein,"The Hegemon's Dilemma," International Organization38
(Spring i984), 355-86.Also see JoanneGowa, "Anarchy,Egoism,and Third Images: The Ev-
olutionof Cooperationand InternationalRelations,"InternationalOrganization40 (Winter
1986), 176-77, and JosephGrieco,"DistributionalUncertainty and theRealistProblemof In-
ternationalCooperation,"paper presentedto the i986 APSA meeting.For a discussionof the
issuesand a summaryoftheexperimentalliterature, see Deborah Larson,"Game Theoryand
the Psychologyof Reciprocity"(unpub., Columbia University),25-3I. More generally,Fred
Hirsch, in Social Limitsto Growth(Cambridge: Harvard UniversityPress, 1976), brilliantly
demonstratesthatmany goods in societyare "positional"-that is, theyare inherently com-
petitive.Only a fewpeople can be at the top of an establishedhierarchyof power or prestige.
Certaingoods-such as livingin an uncrowdedarea-cannot be sharedwithlarge numbers
of others.More fundamentally, in many aspectsof lifewe judge how well we are doing by
comparingourselvesto others.

Axelrod points to the ironythatthe strategyof reciprocity(Tit-for-Tat)

that proves so effectivein computertournamentsof PD can never win
any individual game.51The strategyworks well, however,because it ac-
cumulates large numbers of points when matched against fellow co-
operatorsand is not terriblybadly exploitedwhen matchedagainstmore
competitivestrategies.But if the main rewards in internationalpolitics
are for relativegains, a strategythatcan, at best,only tie will not be at-
tractive.52
The conditionsunder which statesseek to maximize theirrelativeas
opposed to theirabsolute gains need more exploration.53 These concerns
are likelyto be greaterin the securityarea than in internationaleconom-
ics, but theyare presentin the latteras well, especiallybecause military
and economic strengthsare closelylinked. Third-worldstatesoftenseek
greatereconomic equality with the developed countriesas well as abso-
lute economic growth.Furthermore,the drive to increasetheirpolitical
power vis-a-visthedeveloped stateswould make relativegains even more
important.54
Even among allies, concern forrelativegains is rarelycompletelyab-
sent.For example, would the United Statesgain or lose (or both) ifJapan
made new breakthroughsin microelectronics? The U.S. would benefitby
importingbettergoods, but would fall furtherbehind Japan in various
technologies.This, in turn,mightenable Japanto keep wideningthepro-
ductivitygap with America. More than national pride is involved,espe-
ciallysince no one can guaranteethatJapanwill maintainitspoliticalori-
entation in the future.Indeed, Stein has shown that the "hegemon's
dilemma" operates in the economic arena because a dominantstatethat
cares only about absolute gains is more likely to be overtakenby other
powers.55Aaron Friedberghas similarlydemonstratedthe importanceof
this issue in the debate about the decline of Britain'seconomic power at
theturnof thecentury.Those who were leastconcernedstressedthecon-
tinuinggrowthof the Britisheconomy; those who feltthe situationwas
dangerous concentratedon the factthat othereconomies were growing
even faster.56
Axelrod (fn. i).

52See the discussionof strategiesin Axelrod's computertournamentin Roy Behr, "Nice
Guys Finish Last-Sometimes," JournalofConflictResolution25 (Junei981), 289-300.
53 The relevantliteraturefromexperimentalpsychology is summarizedin Larson (fn.50),
28-29.
54 See RobertW. Tucker, The Inequalityof Nations(New York: Basic Books, I977); Ste-
phen Krasner,StructuralConflict(Berkeley:Universityof CaliforniaPress,i985).
55 Stein (fn.5?)-
56 Friedberg, The WearyTitan: Britainand the Experienceof RelativeDecline, i895-I905
(Princeton:PrincetonUniversityPress,forthcoming i988), chap. 2.

336 WORLD POLITICS
Waltz's argumentraises similardifficulties. He denies thatstatesseek
to maximize eitherabsolute or relativepower. Instead,theytryto maxi-
mize their security.57 But since the state's securitycan be influenced-
positivelyor negatively,depending on the situation by the power and
securityof others,this concept is likelyto be inherentlyinteractive.Un-
like the desire to maximize relativegains, thismay increasecooperation,
but the result still will complicate the standard proto-gametheoryfor-
mulations.
PERCEPTIONS, VALUES, AND SELF-INTEREST
The finalset of problemsconcernspsychology,beliefs,and values. In

increasingorder of conceptual difficulty, the issues are the psychological
impedimentsto cooperation,the role of values and the autonomyof be-
liefs,and the question of whethernarrow self-interest can explain most
international behavior. Realism has generally ignored the decision-
making level of analysis;game theorycan incorporatetheempiricalfind-
ings in thisarea only at the costof parsimonyand a revisionof manycur-
rentconclusions.To understandwhen cooperationoccurs,we need to see
how differentbeliefsand values can affectthe actors' evaluation of the
outcomes.A relatedquestion is whetherself-interest is definedcoollyand
atomisticallyor is drivenbypassionsand care forwhat happensto others.
If the latteris the case, the importantfactorswill be outsideof,ifnot de-
nied by,the anarchyframework.
PSYCHOLOGICAL IMPEDIMENTS TO COOPERATION
Robert Keohane has noted thatcooperationmay be explained in part

by the factthatpeople can operate with only bounded rationality;states-
men need to conservecognitiveresources,and shared normsand princi-
ples can be extremelyuseful in easing the burdensof prediction,choice,
and coordination.58But cooperationis decreased by otheraspects of the
way people think. People do not mechanicallyreciprocateconciliation
and defection.Instead, their behavior is mediated by their analysis of
what the otherdid and why it did it. The interpretation of others'action
is rarelyself-evident,but it is almost always important.Most behavioris
ambiguous; even more so are the underlyingintentions.Since actors' re-
sponses are linked to theirpredictionsof how the otherwill react to al-
ternativepolicies theycan pursue, theirattributionsof the causes of the
57Waltz (fn.24); see also Waltz, "Reflections
on TheoryofInternational
Politics:A Response
to My Critics,"in Robert0. Keohane, ed., Neorealismand Its Critics(New York: Columbia
UniversityPress, i986), 334.
58 Keohane, AfterHegemony(Princeton:PrincetonUniversityPress,I984), II0-32.

other's behavior will be crucial. Thus, how they react is influencedby

their inferencesas to whetherthe other intended the resultsthat were
produced and whetherthe behavioris bestexplained by the transientsit-
uation the other was in or by its enduring dispositions.59 Furthermore,
since internationalpoliticsis an interactiveprocess,a statesman'sunder-
standingof the other's behavior is influencedby how he thinkshis own
state is behaving toward the other.Indeed, perhaps the most important
psychologicalfactorthat interfereswith cooperationis thatstatesmen
and people in theireverydaylives greatlyunderestimatethe extentto
which their actions threatenor harm others. They think they are co-
operatingwhen an objectiveobserverwould say thattheyare, at least to
some extent,defecting.
Part of the reason is thatmost statesmenare only dimlyaware of the
securitydilemma. When theyare peaceful,theythink that theirinten-
tionsare clear and thatotherswill not be threatenedbythemeasuresthey
are taking in theirown self-defense.Former AssistantSecretaryof De-
fenseRichard Perle once said thatif we are in doubt about Soviet inten-
tions, we should buy arms: if the Soviets are aggressive,the build-up
will be needed, and iftheyare not, the only consequence will be wasted
money.Similarly,when U.N. troopswere movingtoward the Yalu, Sec-
retaryof State Acheson statedthattherewas no danger thatthe Chinese
would interveneout of self-defensebecause theyunderstoodthattheU.S.
was not a threatto them.60
Statesmenwho thinkwell of themselvesgenerallybelievethattheirac-
tionsare compatiblewith the reasonableinterestsof others.For example,
Raymond Garthoffshows that, during the period of detente,Ameri-
cans leaders and general public alike believed thattheircountrywas
restrainedand cooperativetoward theU.S.S.R.; actually,theU.S. contin-
ued to seek unilateraladvantage by freezingitsadversaryout of the Mid-
dle East, courtingChina, and developingadvanced militarytechnology.6'
As a resultof thisfaultyself-image,any hostilereactionby the adversary
is likelyto seem unprovoked.
This problem is compounded by a second and better-knownbias:
statestend to overestimatethe hostilityof othersand will oftensee as de-
59For furtherdiscussionof theinferenceprocessesand evidenceoftheirimportancedrawn
fromexperimentsand case studies,see Jervis(fn. I4), 32-48, and Larson (fn.50).
60 Daniel Yergin, " 'Scoop' JacksonGoes forBroke,"Atlantic Monthly 233 (JuneI974), 8i.
(The same erroris made, even more crudely,by Secretaryof Defense Weinbergerin Annual
Reportto the Congress,FY 1988 [Washington,DC: GovernmentPrintingOffice,i987], i6.)
Acheson's views are presentedin JohnSpanier,The Truman-MacArthur Controversy and the
Korean War (New York: Norton, i965), 97. Similarexamplesare discussedin Jervis(fn. I4),
67-76, and Seaborg (fn. I2), 30-3 I.
61 Garthoff(fn. 36). Also see Van Evera's discussion(fn. 28) of the role of nationalismin
preventingstatesmenfromobjectivelygaugingthe behaviorof theirown states.

338 WORLD POLITICS
fectionactionsthata disinterestedobserverwould recordas at leastpartly
cooperative.Dulles's view of Khrushchev'sarms cuts in the mid-I95os is
one such example; Reagan's view of most Soviet arms proposalsmay be
another.62
These two biases often operate simultaneously,with the result that
each side is likelyto believe that it is cooperatingand thatthe otherhas
responded by defecting.Thus, the United States says that while it has
been restrained,the Soviet Union has accumulatedmanymore arms than
it needs fordeterrence.The Sovietsprobablyhave a similarperceptionof
the American defensepostureand may thinktheyhave limitedtheirown
arms. The breakdown of detenterevealsthe same patternof perceptions
(although thatmay not provide the primaryexplanationforthe inability
of the superpowersto maintain cooperation).To take just one example,
when Brezhnev told Nixon in the springof I973 thatthe statusquo in
theMiddle East was unacceptable,and when Gromykolaterwarned that
"the fireof war [in the Middle East] could break out onto the surfaceat
any time," theymay well have thoughtthattheywere fulfillingtheirob-
ligationsunder the Basic PrinciplesAgreementto consultin the eventof
a threat to peace. The Americans perceived the Soviets to be making
threatsand, by failingto restrainthe Arabs or to notifythe U.S. of the
impending attack,to be violatingthe spiritof detente.63 The same effect
oftenappears during bargainingwhen each side thinksthatthe conces-
sions it has made are greaterthan thoseit has received.64
Furthermore,when statesmenrealize that the otherside has cooper-
ated, theyoftenbelieve that it did not have much choice. Thus, decision
makers underestimatethe abilityof othersto defect,and thereforefre-
quently believe, incorrectly,that theycan get away with some exploita-
tion.This factorplayed a role in the deteriorationof the recentdetenteas
well as in the cooling of relationsbetween England and France in the
i840s.65 The Soviet Union in the I970s, and both sides in the earlier in-
62 See the classic essay by Ole Holsti, "Cognitive Dynamics and Images of the Enemy:
Dulles and Russia," in David Finlay,Ole Holsti,and RichardFagen,Enemiesin Politics(Chi-
cago: Rand McNally, i967), 25-96. Michael Sullivan,InternationalRelations:Theoriesand Evi-
dence(Englewood Cliffs,NJ: Prentice-Hall,i976), 45-46,questionsthelinksbetweenDulles's
beliefsand Americanbehavior,however.
63 Gromykois quoted in Galia Golan, YomKippurand After (London: CambridgeUniver-
sityPress, I977), 68. The treatmentof the I973 war is a good litmustestforone's views on
detente:compare,forexample,the discussionsin HarryGelman, The BrezhnevPolitburoand
theDecline ofDetente(Ithaca, NY: Cornell UniversityPress,i984), chap. 4; Garthoff(fn.36),
chap. ii; and Alexander George, Managing U.S.-SovietRivalry(Boulder, CO: Westview
Press, i983), chap. 7.
64 For experimentalsupportforthisproposition, see Michael Ross and Fiore Sicoly,"Ego-
centricBiases in Availabilityand Attribution," JournalofPersonalityand Social Psychology37
(March I979), 322-36.
65 Roger Bullen,Palmerston, Guizot,and theCollapseoftheEntenteCordial(London: Ath-
lone, I974), 8i, 93.

stance,thoughttheycould safelyact against the otherside's interestsbe-

cause theyexaggeratedthe constraintsinhibitingthe rival'sretaliation.
Because of biases like these,analysesthatassume both sides to be per-
ceivingeach otheraccuratelyare likelyto be incorrect;strategiesthatare
based on thisassumptionare likelyto be ineffective. Axelrod shows that
the strategyof Tit-for-Tatworks quite well when thereis a one-percent
errorrate in the correctidentification of the other'sbehavior.66
But such
a figureis drasticallylower than that which can be expected in political
interactions;George Downs and his colleagueshave shownthatwhen the
errorrate is higher,thisstrategyis not likelyto yieldstablecooperation.67
It is partlyforthisreason thatLarson argues thatCharles Osgood's GRIT
(Graduated Reciprocationin Tension-reduction),which makes a major
effortto break through the adversary'sperceptual biases, may induce
more cooperationthan Tit-for-Tat.68 Strategiesthatare not tightlycon-
ditional on what the otherdoes, and thatdo not require immediateand
matchingreciprocation,may produce a change in the other's attitudes,
and so may be more effectivethan game-theoreticanalyses would sug-
gest. Strictreciprocitymay fare less well than expected. Put in slightly
differentterms,cognitivebiases decrease transparency, thus making re-
gimes harder to establishand maintain.
Cooperation is more likelyto come about when actorscorrectforthese
conflict-inducingbiases or are willing to toleratea higherlevel of per-
ceived cheatingby the otherside. But such conditionsand strategiesalso
increasethe chances thattheotherside will, in fact,cheat.One of the rea-
sons whyIsrael was taken by surprisein I973 was thatitsdecisionmakers
were highlysensitiveto the securitydilemma and believed thatthe most
probable cause of war would be Egyptianpreemptiongrowingout of an
incorrectbeliefthatIsrael was about to strike.They thereforeinterpreted
Egypt's behavior precedingthe war as evidence,not thatEgypt was de-
fecting,but that its decision makers thoughtIsrael mightbe planningto
do so. As a result,Israel did not respond by increasingits own military
readiness.69The problemsof misperceptionthusheightenthestatesman's
dilemma in deciding whetherto cooperateor not.
66
Axelrod (fn. i), i82-83-
67 Downs, Rocke,and Siverson(fn.I9), I33-34. Also see JonathanBendor,"In Good Times
and Bad: Reciprocityin an UncertainWorld,"American JournalofPoliticalScience3I (August
i987), 53I-58. Downs and Rocke discussthe implicationsof thisfindingin "Tacit Bargaining
and Arms Control,"WorldPolitics39 (April i987), 297-325.
68 Deborah Larson, "Crisis Preventionand theAustrianStateTreaty,"International Organ-
ization4I (Winter i987), 30-34.
69 JaniceStein, "Calculation, Miscalculation,and ConventionalDeterrenceII: The View
fromJerusalem,"in RobertJervis,RichardNed Lebow, and JaniceStein,Psychology and De-
terrence(Baltimore: The JohnsHopkins UniversityPress, i985), 6o-88. This is part of the
broaderdifficulty statesmenfacein decidingwhethertheotherside is an aggressorwho must

340 WORLD POLITICS
Strategiesthat are robust in the face of misperceptionare extremely
valuable, but it is far fromcertainthattheyexist.70Stable cooperationis
most likely to resultwhen the decision makers' preconceptionsprovide
an accurate fitwith what the otherside is like; thatmay be as much the
productof luck, however,as of sensitivity and statesmanship.It is clearly
importantto determinethe extentto which strategiesthat would work
well when informationis accuratecan also servein a world permeatedby
ambiguity and strong perceptual and decision-makingbiases. Neither
scholars' analyses nor statesmen'spolicies can safelybe based on the as-
sumptionthateitherside understandsthe other.
BELIEFS AND VALUES
Expected utilitiesare thevaluationan actorplaces on a courseofaction

or outcome; theyinvolve both estimatesof consequencesand judgments
about intrinsicworth.Contraryto the implicationsof manyRealist writ-
ings,theseestimatesand judgmentsare notobjective,and theyshould not
be accepted withoutinvestigatingtheirformation,as is done in most of
the work on game theoryand anarchy.In many cases, what is crucial in
determiningwhetherthe actorscooperateis theirbeliefsabout the effec-
tivenessof alternativepolicies beliefsthatoftenprove to be eitherlack-
ing in evidentiarysupportbeforethefactand/orwrongafterthefact.The
most obvious and consequentialexample is the greatexaggerationby de-
cision makers of the efficacyof the offensivebeforeWorld War I, which
meant thata crisiswas likelyto lead to war.
The general question raised by thiscase is theextentto which (and the
circumstancesunder which) the main impedimentsto cooperationare
rooted in potentiallymalleable beliefsabout the situationratherthan in
itsstructure.7'
For example,manyhistorianshave argued thattherecould
have been much greatercooperationbetween Britainand Russia in the
i9th centuryif the formerhad not greatlyexaggeratedthe hostilityand
capabilityof the latter.72More generally,decision makers are oftenun-
be met withfirmness(ifnot force),or a morereasonablestatethatcan be conciliatedor,to use

the older term,appeased. See Jervis(fn. I4), chap. 3.
7? Ibid., 109-13; RichardNed Lebow, "The DeterrenceDeadlock: Is therea Way Out?" in
Jervis,Lebow, and Stein (fn.69), i80-202.
7' See Ernst Haas, "Why Collaborate? Issue-Linkage and International Regimes," World
Politics32 (April i980), 357-405; RobertRothstein(fn.37); RobertJervis,
"SecurityRegimes,"
International Organization 36 (Spring i982), 359-60, 373-75; Jack Snyder, "Perceptions of the
SecurityDilemma in I94I," in Jervis,Lebow, and Stein(fn.69).
72 See, forexample,Norman Rich, WhytheCrimeanWar?(Hanover,NH: University Press
of New England, I 985), and R. W. Seton-Watson,Disraeli,Gladstone,
and theEasternQuestion
(New York: Norton, I972).

awareofthesecurity dilemma,and therefore actin waysthatcompound

it. Is it possibleto altersuchperceptions, and thusalterbehavior?One
reasonwhytheKennedyand Carteradministrations werefavorably pre-
disposedtowardarmscontrolwas thatmanyofthetopofficials believed
thatarmsraceswereoftendrivenbyaction-reaction cycles.Similarly,can
perceptions oftherelativeefficacy ofoffenseand defensebe alteredwith-
out changingtechnology? If militaryleadersare drivento favortheof-
fensivebecauseoforganizational and ifcivilianleadersarealso
interests,
disposedto underestimate thepowerofthedefensive,73 thenperhapsthe
security dilemmaand theresulting DD is morea creatureofbiasesand
domesticinterests thanofthestructure oftheinternational system.
This questionariseswiththecurrentconcernthatWorldWar III is
mostlikelyto startthrough"crisisinstability" approximating thatwhich
prevailedin I9I4. The vulnerability ofcommand,control, and commu-
nicationsfacilities mightlead statesmen to believethat,eventhoughany
nuclearwar wouldbe dreadful,therewouldbe realadvantagesto strik-
ing firstif the choicewere betweendoing so and beingattacked.On
closerexamination, however,itdoesnotappearthatevena well-executed
first strikewouldgreatlyreducedamagetothestate;so thedangerofcri-
sisinstability maylie lessin theobjectivesituation thanin decisionmak-
ers incorrect beliefs.74
Mosttheorists whouse theframework ofanarchytakeforgrantednot
onlypeople'sinstrumental beliefsbutalso thevaluesthattheyplace on
outcomes.Althoughwe realize thathow the Prisoners'Dilemma is
played-and indeed,whetheritis a dilemmaat all-is deeplyinfluenced
by thevalue each actorputson theother'swell-being,75 scholarsknow
littleabouttheprocessesbywhichthisevaluationis established and by
whichitcan change.It is a centraltenetin international politicsthatpeo-
ple value thesecurity and well-beingof theirown statemorethanthey
do thatofothers.The selfis definedas thenationalself.Butthisneednot
73Snyder (fn.29); Posen (fn.44).

74For a furtherdiscussion,see RobertJervis,The IllogicofAmericanNuclearStrategy (Ith-
aca, NY: Cornell UniversityPress, i984), I26-29; Jervis,"PsychologicalAspectsof CrisisSta-
bility,"in Jervis,The Implicationsof theNuclear Revolution(Ithaca, NY: Cornell University
Press, forthcoming);Richard Ned Lebow, Nuclear CrisisManagement:A DangerousIllusion
(Ithaca, NY: Cornell UniversityPress, i987); AshtonCarter,JohnSteinbruner,and Charles
Zraket,ManagingNuclear Operations(Washington,DC: BrookingsInstitution,i987).
75Morton Deutsch, "Trust and Suspicion,"Journalof ConflictResolution2 (December
1958), 265-79; Deutsch, "The Effectof MotivationalOrientationupon Trust and Suspicion,"
Human RelationsI3 (May i960), I23-39. Similarly,rationalchoice analysesof politicians'be-
havior would yield verydifferent resultsif it were assumed thatthe value being maximized
was theindividual'seconomicprospectsratherthanhispoweror votes.We would thenexpect
politiciansto caterto popularor constituency interestsonlyto amass enoughpowerthatcould
be efficientlytraded upon forpecuniarygain.

342 WORLD POLITICS
be trueforever.It maybe in thenatureofhumanbeingstoputfirsttheir
own well-beingand thatofotherswhocarrytheirgenes,butitis notin-
evitablethatpeoplewillalwayscaremoreaboutthefatesofthoseon their
side of a nationalborderthantheydo aboutsimilarindividualson the
otherside.76
The issueis oftenput in termsof self-interest versusaltruism, butit
maybe moreusefulto thinkofhow theselfis defined.Whyshouldour
attitudestowardothersbe basedon theirgeographic locationratherthan
on thevaluestheyhold?77Sharedvaluescan be one reasonfornational
butanyindividualcan havemorein commonwithmanypeople
identity,
inothercountries thanwithmanyinhisorherown.(Indeed,howAmer-
icansregardfellow-nationals seemslessdetermined bytheirraceand eth-
nicitythanby theextentto whichtheysharetheirvalues.)78 Similarly,
Alker and Shermanstressthe importance of the "scopeand depthof
normative
insecurity-relevant integration"in theinternationalsystemas
a determinant offoreignpolicy.79
The degreeofvalueintegration, and therefore thescopeofthedefini-
tionoftherelevantself,maysometimes be largerthanwe assume.States-
men usuallylike to presentthemselves as "hard-headed"and as caring
almostexclusively about theirown country. But policyis in factoften
drivenbymotivesthatare hardforRealiststoaccommodate. For exam-
ple,how can we explainthefactthattheUnitedStatesdid notconquer
Canada sometimein thepasthundredyears?80 A Realistaccountwould
76 Some observershave attributed the relativelack of concernin the home countryforthe

German hostagestaken in Beirutin Januaryi987 to theweak German nationalidentity.See
JamesMarkham, "West Germans Low-Key About Abductions,"New YorkTimes,January
i9, i987. Compare the reactionof Japan,a countrysome describeas "a huge tribalsociety,"
in a similarsituation:Clyde Haberman, "JapanOutragedat Manila AbductionofExecutive,"
New YorkTimes,FebruaryI5, i987.
77This question is an ancientone and can be tracedback at leastas faras Aristotle'sadvice
that Alexander distinguishamong his subjectsaccordingto whethertheywere Greeks or
non-Greeksratherthanaccordingto theirpersonal,individualcharacteristics; see The Politics
ofAristotle, ed. and trans.byErnestBarker(London: OxfordUniversityPress,I958), 388. For
an attemptto use sociobiologyto explainnationalloyalty,see R. Paul Shaw and Yuwa Wong,
"Ethnic Mobilizationand the Seeds of Warfare:An EvolutionaryPerspective,"International
StudiesQuarterly3I (March i987), 2I-26.
78 MiltonRokeach and Louis Mezei, "Race and Shared Beliefin Social Choice," ScienceI5I
(Januaryi966), i67-72; Rokeach, Beliefs,Attitudes, and Values (San Francisco: Jossey-Bass,
i968). Similarly,in the view of the Ottomans,"the communityof truebelievers,. . . not the
state,constitutesthe basic Muslim policy,transcendingall boundaries."See Thomas Naff,
"The OttomanEmpire and theEuropean StatesSystem,"in Hedley Bull and Adam Watson,
eds., The ExpansionofInternational Society(Oxford:Clarendon Press,i984), I43.
79 Hayward Alker, Jr.,and Frank Sherman,"CollectiveSecurity-Seeking PracticesSince
I945," in Daniel Frei, ed., ManagingInternationalCrises(BeverlyHills, CA: Sage, i982), I4I-
44. This essay draws on the work of Karl Deutsch, esp. Deutsch et al., PoliticalCommunity
and theNorthAtlanticArea (Princeton:PrincetonUniversityPress,I957).
80 William T.R. Fox, A Continent Apart:The UnitedStatesand Canada in WorldPolitics(To-
ronto:Universityof Toronto Press, I985).

be strained, havingtoconjureup power-based disincentivesthatarehard

to find.Americancooperationis betterexplainedby focusingon value
integration in threesenses.First,theuse offorceis negatively valuedby
mostAmericansunlesstheobjectis to removea menacetoAmericanse-
curityor to establisha democraticregime.Second,manyvalues are
sharedbetweencitizensofCanada and oftheUnitedStates.Even ifcon-
questwould havegiventheU.S. greaterwealthor security, thefactthat
Canadiansocietyresembles Americanmeansthatmostofwhatwe want
to see in a countryis alreadyin place in Canada. Third,largelybecause
of thecommonvaluesbetweenthetwo countries, Americansprizethe
well-beingofCanadians.Thus,usingforceagainstthemwould,bycon-
stituting an offenseagainstCanadians,diminishtheutilitiesof Ameri-
cansas well.
Commonand conflicting valuesalso helptoexplainwhoma statewill
offerto protectand whom it will oppose.8'Karl Deutschand his col-
leaguesstresstheroleofthe"compatibility ofthemainvaluesheldbythe
politically relevantstrataof all participating units"in theformation of
what theycall security-communities i.e., groupsof nationsamong
whomwar is unthinkable.82 This factoris also important in lessdrastic
formsof cooperation. The Americancommitment to Europe and still
more,to Israel cannotbe accountedforapartfromthecommonheri-
tageand valuesthatmake Americanscareaboutthefatesof thesepeo-
ples. Securityconsiderations are insufficient to explainAmericanties,
whichwould be deeplyaffected ifEuropeor Israelwereto becomefas-
cist.Similarly, Americanoppositiontocommunism in thethirdworldis
basednotonlyon nationalsecurity concerns, butalsoon an identification
withhumanbeingsinothercountries whomwe do notwanttoliveunder
tyranny.83 It maythenbe quitewrongto see Realpolitikas thesourceof
thedecisionto fightin Vietnamand to dismissas cantthedesireto save
thatcountry frommisery.Is itan accidentthattheproponents ofthewar
were wrongin theirpredictions of widespread"dominoeffects" of an
Americandefeat,butwerecorrect in theirargument thatmillionsofpeo-
ple in Indochinawouldsuffer or die froma NorthVietnamesevictory?
The most far-reachingchanges in international politicsinvolve
changesin nationalgoalsand values.Japanis nowa muchmoresuitable
partnerforcooperation thanit was intheI93os, and notonlybecauseter-
8"For a discussionof the role of ideologyin alliances,see Ole Holsti, P. Terrence Hop-
mann, and John Sullivan, Unityand Disintegration in International
Alliances(New York:
Wiley, I973), and Walt (fn.48), i8-26.
82 Deutsch (fn.79), 46.
83 For discussionsof thisargument,see RobertJervis, The Logic ofImagesin International
Relations(Princeton: PrincetonUniversityPress, I970), 244-50; George Quester,American
ForeignPolicy(New York: Praeger,i982).

344 WORLD POLITICS
ritorialexpansion is neitherpossible nor economicallynecessary.Some-
thinghas occurredthatis more basic thanchangesin instrumentalbeliefs.
Rabid nationalism and the drive to dominate have been transmuted.A
Japanese nationalistof the I930S who saw his countrytoday would be
horrified,as Mishima was. Because it took a cataclysmto produce such a
change, this patterndoes not provide an attractiveroute to a more co-
operativeworld. But it does show boththeimportanceand themutability
of values.
A less dramatic alterationmay be a more hopefulmodel. In the late
i960s, West Germans came to accept the existenceof the East German
regime. The division of the country,while stilldeeply regretted,was no
longer completelyintolerable.With the territorialstatusquo generally
accepted,tensionsand the danger of war were reduced,and cooperation
in the formof the Quadripartiteand the Helsinki Agreementsbecame
possible.
In Realism, values are generallytaken as unproblematicand constant;
theyare givens in game-theoreticstudies.Thus, althoughanalysiswithin
the frameworkof anarchycan produce accurateresults,it may make our
treatmentexcessivelystatic and distractus from importantquestions.
How values are developed, maintained,and changed may be crucial to
internationalpolitics,and may stronglyinfluencethe extentof coopera-
tion; it is a matter,however,thatcannotbe explained withinthe anarchy
framework.
NORMS: Is NARROWSELF-INTERESTENOUGH? HATRED ANDNORMS

Utilitiescan reflectpassions as well as interests.Considerationsof the-
oreticalparsimonyspeak forignoringgreed,hatred,and envy,as well as
moralityand self-sacrifice;but thatdoes not mean thatsuch motivesare
absent in the world. Witness Lord Salisbury'sanalysis of Prussia's de-
mands on France in I870 (which, it can be argued, laid the foundations
forWorld War I):
Unless[theGermans]enjoythepleasingsensationofwitnessing themor-
ofFrance,theywillthinkthattheobjectsofthewarareonlyhalf-
tification
attained.... [I]t is revenge that theydesire: not a strategicfrontier,or the
recoveryof lost "brothers"-but termsof peace which shall drive the iron
well into theirenemy's soul.84
These are the sort of impulses thatolder Realist scholarsfromThucyd-

ides to Morgenthauwarn of.They lead to destructivepoliciesand are dif-
84 Lord Robert Salisbury,"Count Bismarck'sCircular Lettersto Foreign Courts, i870,"
The Quarterly Review 129 (October i870), 553. I am gratefulto Marc Trachtenbergforpoint-
ing me to thisarticle.

ficultto theorize about. Although Realism recognizesthat cool, narrow

self-interestmay fail to guide behavior,it providesfew grounds forpre-
dictingwhen it will failor what behaviorwill followifit does.
A related issue is the role of norms in furtheringcooperation.In one
sense, norms indicate behavior that is expected,standard,or normal. In
another sense, theyindicate behavior that is approved and valued posi-
tively.85But when do normsin thefirstsense become normsin the second
sense? If theydo not,can theyreallyproduce a greatdeal of cooperation?
One of the themesof the anarchyliterature,of course,is thatnarrowself-
interestis sufficientunder many conditions,but the evidence is not very
strongand the conditionsmay be veryrestrictive.86 For example, many
scholarshave argued that in the balance of power, stabilitycan arise out
of the interplayof the states'narrownationalinterestsin survivaland ex-
pansion. The resultsare not always peaceful,but theydo maintain the
system.Paul Schroeder'sfascinatinghistoricalstudiesdenythatthisis an
accurate pictureof European balance-of-powerpolitics,however.He ar-
gues that mediating structuresof smaller states,a common understand-
ing of internationallaw, and a shared sense of the appropriatebehavior
were required ifwar was not to grow froma tool of statecraftto thedom-
inatingfactof internationalpolitics.87
In tracingmuch of thecause of World War I to thedecline in Austria's
position,Schroeder pointsto Britain:althoughit had a greatstake in the
85Kratochwil and Ruggie argue thatour standardmethodologyis inappropriteforverify-

ing theexistenceof normsin thelattersensebecause pointingto instancesin whichnormsare
violated does not establishthat theydo not existor are not important.See FriedrichKra-
tochwiland JohnGerard Ruggie,"InternationalOrganization:A Stateof the Art on an Art
of the State,"International Organization40 (Autumn i986), 766-69.A moregeneraltreatment
of normsalong theselines is FriedrichKratochwil,Rules,Norms,and Decisions:On theCon-
ditionsof Practicaland Legal Reasoningin International Relationsand DomesticAffairs(Cam-
bridge:Cambridge UniversityPress,forthcoming i988).
86 This question is raised,among otherplaces, in Alker's analysesof how people play and
thinkabout Prisoners'Dilemma in the laboratory.See Hayward Alker,Jr.,and Roger Hur-
wirtz,"ResolvingPrisoner'sDilemmas" (unpub., M.I.T.); Alker,"ReflectiveResolutionsof
SequentialPrisoner'sDilemmas," presentedat themeetingoftheSocietyforGeneral Systems
Research,May 30, i985; and Alker, "From Quantityto Quality: A New ResearchProgram
on ResolvingSequential Prisoner'sDilemmas," presentedat the i985 meetingof the Ameri-
can Political Science Association.The incentivesand settingsof laboratorysituationsare so
different fromthoseoperatingin internationalpolitics,however,thatit is farfromclear that
theseexperimentstell us much thatcan be directlytransferred.
87 See the followingessays,all by Schroeder:"The Lost Intermediaries: The Impactof i870
on the European System,"International HistoryReview 6 (Februaryi984), 1-27; "Contain-
ment Nineteenth-Century Style: How Russia Was Restrained,"SouthAtlanticQuarterly82
(Winter i983), i-i8; "World War I as Galloping Gertie,"JournalofModernHistory44 (Sep-
tember1972), 319-45; "The Nineteenth-Century Balance of Power: Language and Theory,"
paper deliveredat the 1977 meetingoftheAmericanPoliticalScienceAssociation;"The i9th-
CenturyInternationalSystem:Changes in theStructure,"WorldPolitics39 (October i986), i-
26. Also see FriedrichKratochwil,"On the Notion of 'Interest'in InternationalRelations,"
International Organization36 (Winter i982), 1-30.

346 WORLD POLITICS
statusquo and in avoiding conflict,it neveracted to limitor controldan-
gerous shiftsin the Balkans. Others also failed to maintainthe necessary
balance, although theirbehavior was logical to the extentthattheywere
willing to toleratea high risk of war. Aftershowing how much Austria
was weakened by Rumania's realignmentin i9i3, Schroederstressesthat
despite its importance,
no government addresseditselftothemostobviousand criticalquestionof
all: how was thisnew,crucialdevelopment to be managed?How, thatis,
could it be harmonizedwiththeoverallEuropeanbalance,incorporated
intotheprevailinginternational withoutraisingthealreadyfearful
system
strainsto thepointof explosion?No one thoughtaboutthisproblemor
suggesteddoinganything aboutit.88
Without conscious management-without at least some states seeking
the common interestof the membersof the system-stability,peace, and
the best interestsof the individual statescould not be maintained.A sim-
ilar argumentwas made by Lord Salisbury,laterto become Foreign Sec-
retary,when he criticizedthe refusalof the Britishgovernmentto coun-
teractthe harsh termsPrussia imposed on defeatedFrance:
If [theleaders']intentionis ... to drawall theprofit
theycan fromthear-
rangements of thegreaterinternational republic,and yetto bearno share
ofthecostand dangersofitsgovernment, we doubtnotthattheyare pre-
paringthemselves fora severecondemnation fromtheEnglishpeople.We
onlytrustthattheyare notalso preparing forEnglandthenationaldoom
thatalwayswaitsfortheselfishand timid.89
A parallel pointarisesfromone of Axelrod's computersimulations.He
finds that stable cooperation is much more likely to occur when actors
follow "metanorms" that call for punishing those who fail to enforce
norms.In otherwords,actorsmust not onlybe preparedto punish those
who defect,but also to act againstthosewho failto punishthem.90In the
same way, cooperation can be powerfullyreinforcedby injunctionsto
help others,or at least to limit the harm done to them.Thus "although
prescriptivemoralityis not usuallyexpressedin laws, in a numberof Eu-
88Paul Schroeder,"Romania and the Great Powers before1914," Revue RoumaineD'His-

toire 14 (No. I, I97, 52-53. As Schroederputs it in "World War I as Galloping Gertie":
"Everyonewanted a payoff;no one wanted to pay" (fn.87), 345.
89 Salisbury (fn. 84), 556. During the Eastern Crisis of i877, William Gladstone asked:
"What is to be the consequence to civilisationand humanity,to publicorder,ifBritishinter-
estsare to be the rule forBritishagentsall overtheworld,and are to be forthemthemeasure
of rightor wrong?" (Quoted in Seton-Watson,fn.72, p. 69.) For a relatedgeneralargument,
see ArmatyaSen, "Rational Fools: A CritiqueoftheBehaviorFoundationsofEconomic The-
ory," Philosophy and Public Affairs6 (Summer 1977), 326-41 .
9?RobertAxelrod,"Modeling the Evolutionof Norms,"AmericanPoliticalScienceReview
8o (December i986), 1095-11 1I.

ropean countries and now also in some [American] states ..., helping
othersin certainkinds of extremeneed is requiredby law."9'
Applying the same reasoning,Robert Trivers argues that evolution
should selectforpeople who reciprocatealtruism.Emotionssuch as anger
at those who do not reciprocate,and guilt over one's own cheatingare
also functionalbecause they bolster reciprocity.Short-runcalculations
might lead people to ignore the transgressionsof others(or to replyto
them only mildly),or to shirk reparationswhen theyhave been caught
cheating.But over the long run,such behavioris destructive.Dispropor-
tionate retributionfor cheating,called up by anger, can induce future
compliance: "it seems plausible ... thatthe emotionof guilthas been se-
lected forin humans partlyin order to motivatethe cheaterto compen-
sate his misdeed and ... thus to preventthe ruptureof reciprocalrela-
tionships." Because these emotions may sustain long-run cooperation
more than calculationwould,
selectionmayfavordistrusting thosewho perform altruistic

actswithout
theemotionalbasisofgenerosity or guiltbecausethealtruistictendencies
ofsuchindividualsmaybe lessreliablein thefuture. One can imagine,for
example,compensating fora misdeedwithoutanyemotionalbasisbutwith
a calculating, motive.Such an individualshouldbe distrusted
self-serving
becausethecalculatingspiritthatleadsthissubtlecheaternowto compen-
satemayin thefuturelead him to cheatwhencircumstances seemmore
advantageous.92
The issue is by no means closed. It is possiblethatnarrowself-interest

may be able to explain most instancesof cooperationin the absence of
binding authority.But we should note thatin Gouldner's classic article,
reciprocityis not only a common empirical patternand a way out of a
dilemma; it is also a moral imperative:"people should help those who
have helped them."93 Such a sense of obligation-if shared-may well
9- Ervin Staub,PositiveSocial Behaviorand Morality,I (New York: Academic Press, 1978),

3-
92 Trivers,"The
Evolutionof ReciprocalAltruism,"Quarterly JournalofBiology46 (March
1971), On p. 52, TriversanticipatesAxelrod'sargumentsabout metanorms.A similar
50-51.
argumentabout deterrenceis made by Dean Pruitt,"Some RelationshipsBetweenInterper-
sonal and InternationalConflict,"in Axelrodet al. (fn.3).
93 Alvin Gouldner, "The Norm of Reciprocity," AmericanSociologicalReview 25 (April
i960), i69-71 (emphasis added); also see Larson (fn. 50), 20-22. For a nice summaryof the
normativehold of reciprocity, see RobertCialdini,Influence(Glenview,IL: Scott,Foresman,
i985), 20-34, and especiallythe marvelousstoryon page 27. Staub notesthatexperiments in-
dicate "thatwillingnessto ask forhelp is reducedwhen people do not expectto have an op-
portunityto provide help in return"(fn. 91, p. 346). The reverseshould have been foundif
rationalcalculationwere the drivingforce.Also see Charles Kindleberger'sreviewofAfter
Hegemony,"Hierarchy vs. Inertial Cooperation," InternationalOrganization40 (Autumn
I986), 844-46.

348 WORLD POLITICS
have great "practical value," as JohnMackie notes.94Considerationsof
morality,fairness,and obligationare almost surelylarge partsof the ex-
planation for the fact that individuals in societycooperate much more
than the Prisoners' Dilemma would lead us to expect. Only economists
behave as the theorysays theyshould; othersare likelyto contributeto
public goods, especiallywhen theybelieve thatfairnesscalls forthem to
do so.95 Indeed, it is possible thatmoralityprovidestheonlyway to reach
many mutual cooperativeoutcomes. In part because of the tendencyfor
people to be self-righteousand to see theirown acts as cooperativeand
those of others as hostile,temptationsand fearsmay produce mutually
undesiredoutcomesas long as narrowself-interest is dominant.At a min-
imum, the feelingthat one is morallyobligated to reciprocatecoopera-
tion and thatotherslive under the same code-permits a wider range
and scope formutuallybeneficialexchanges.In fact,the actorsmay gain
mostwhen theydo not regardthe interactionas one of self-interested ex-
change at all. Even if thisextremeis not approached (and it is not likely
to be in internationalpolitics),withoutthe power of at least some shared
values, without some identificationwith the other,withoutnorms that
carrymoral force,cooperationmay be difficultto sustain.96
CONCLUSIONS
The queries and objections raised here are not all of the same type.
Some are assertionsthat the anarchyframeworkleads us to concentrate
JohnMackie,Ethics:InventingRightand Wrong(Harmondsworth,U.K.: Penguin,1977),

94
I19-20; see also Keohane (fn. 58), 126-27, and Keohane, "Reciprocityin InternationalRela-
tions,"InternationalOrganization40 (Winter i986), 20-24.
95Gerald Marwell and Ruth Ames, "EconomistsFree Ride, Does AnyoneElse?" Journal
ofPublic Economics15 (Junei981), 295-310. I am gratefulto JoanneGowa forreferring me
to thisinstructivearticle.Also see Charlan Nemeth,"A CriticalAnalysisof ResearchUtiliz-
ing the Prisoner'sDilemma Paradigm forthe Studyof Bargaining,"in Leonard Berkowitz,
ed., Advancesin Experimental Social Psychology,VI (New York: Academic Press, 1972), 203-
34; Daniel Kahneman, JackKnetsch,and RichardThaler, "Fairnessand the Assumptionsof
Economics,"JournalofBusiness59 (October i986), S285-300; Sen (fn.89). Transcriptsof the
deliberationsduring the Cuban missilecrisisreveal PresidentKennedy'sconcernwith per-
ceived fairness:McGeorge Bundy,transcriber, and JamesBlight,editor,"October 27, i962:
Transcriptsof the meetingsof the ExComm," International Security12 (Winter i987/88), 30-
92.
96 In his presidentialaddress to the Public Choice Society,Dennis Mueller made a similar
point:contraryto the logic of PD, in thesesituations
most of us choose the cooperativestrategymost of the time. Why? Because we were
taughtto do so. ... One is almost embarrassedto make theseobservationswere it not
thatso manyof us who work with rationalegoistmodels continuallybuild our models
on assumptionsthatignorethesetruismsfrompsychologyand everydaylife.
Mueller,"Rational Egoism vs. AdaptiveEgoism as a FundamentalPostulatefora Descriptive
Theory of Human Behavior,"Public Choice5I (No. I, i986), 5-6.

on questions thatare not central.That is true forthe discussionof pref-

erences. Even if my argumentsabout the importanceof the preference
ordersand of changes in themare correct,the frameworkcan stillbe ap-
plied, albeit at the cost of requiringus to treatas given what may be cru-
cial and problematical.But if this objection is not fatal,it also applies
more broadlyto the use of game-theorymodels in general ratherthan to
Prisoners'Dilemma in particular.Otherproblemscomplicatethepicture.
Basic conceptssuch as cooperation,offense,and power are notas straight-
forwardas theyappear in many analyses,and we may lose a great deal
when we squeeze behavior into the categoriesthe formertwo provide.
Strongperceptualbiases also underminemany of the centralpredictions
derivedfromthe PD. Althoughthenationalbehaviorthatactuallyoccurs
is oftenconsistentwith the general Realist emphasis on conflict,the rea-
sons are differentfromthose stressedby thistradition.Game theorycan
accommodate both uncertaintyand differencesbetween the perceptions
of the two sides,97but we must know what these perceptionsare; they
may be more importantand difficultto understandthanthe resultingin-
teraction.Problems thatare even more fundamentalarise ifnarrowself-
interestis not the drivingforcebehindnationalbehavior:althoughgame-
theorymodels could be built around differentpremises,many of the
Realist argumentsabout anarchywould be undermined.
While I have discussed the problemsinherentin the anarchyframe-
work, I have said littleabout its numerousstrengths.To do so would be
to recapitulatefamiliararguments;but I do want to note thatthe frame-
work is usefulforremindingus thathuman action is oftendrivenby the
twinimpulsesof fearand temptation.It providestoolsforanalyzinghow
theseimpulsescan be harnessed(ifnot tamed) in a way thatleaves all par-
ties betteroff.The conceptsof anarchyand the securitydilemma lead us
to see thatthe internationalsystemnot onlypermitsconflict,but can cre-
ate it: actors may refuseto cooperate with others,not so much because
they seek the positive gains of exploitation,but because they fear that
theirown cooperativeinitiativeswill be mistreated.For purposesof both
analysis and prescription,the frameworkyields significantpropositions
on theconditionsand strategiesthatincreasethelikelihoodofcooperative
behavior and outcomes. But its simplifications can be misleading,its as-
sumptionsrequire scrutiny,and it relegatesmanyimportantquestionsto
the background. Understandingand exploringboth the uses of the ap-
proach and its limitations,without being overwhelmed by either,will
lead to a bettergrasp of world politics.
97See, for example, JohnHarsany, "Bargaining in Ignorance of the Opponent's Utility
Function,"Journalof ConflictResolution6 (March I 962).


Jervis 1988 - Realism, Game Theory and Cooperation

Uploaded by

Copyright:

Available Formats

Jervis 1988 - Realism, Game Theory and Cooperation

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Jervis 1988 - Realism, Game Theory and Cooperation

Uploaded by

Copyright:

Available Formats

Trustees of Princeton University

Realism, Game Theory, and Cooperation

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

Jervis,Roy Radner,and Paul Stern,eds.,Perspectives in Deterrence,forthcoming.

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

CHOICE AND CONTEXT

Perhaps the most importantlimitationof the work on anarchyis that

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

Cuba," in ChristopherMitchell,ed., Immigration Policyand U.S. ForeignRelationswithLatin

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

preferences and behavior.The operating incentives aregivennotonlyby

I For a nice example taken fromthe patternof social relationswithina

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

Other objectionscan be grouped under threeheadings,each of which

PREFERENCES AND PREFERENCE ORDERS

WHAT ARE THE ACTORS' PREFERENCES?

The firstquestion is the most obvious one. How do we know whether

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

theirbehavior.Actorsrarelygive completestatements of theirprefer-

19Harrison Wagner, "The Theory of Games and the Problemof InternationalCoopera-

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

By taking preferencesas given, we beg what may be the most impor-

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

ences as exogenous. Analysis is thereforefacilitated,but at the cost of

23See the discussionof tastesin JamesMarch,"Bounded Rationality, Ambiguity,and the

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

ences thatare held at any giventime,and therefore simplyputsthis

33Glenn Seaborg,Kennedy,Khrushchev, and the TestBan (Berkeley:Universityof Califor-

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

36 JohnMaresia, "Helsinki," in Alexander George, Philip Farley,and Alexander Dallin,

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

In summary, theresearchon cooperationunderanarchyassumesthe

The conceptsemployedbytheanarchyframework seemunproblem-

The conceptsof cooperationand defectionare crucialto the frame-

as defensivemilitarycapabilities,see Paul Stares,TheMilitarization

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

crossed fromEast to West Berlin. (Previously,only special passes issued

42 JamesMarkham,"Allied Diplomats DefyEast GermanControls,"New YorkTimes,May

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

OFFENSE AND DEFENSE

Questions can also be raised about the conceptsof offenseand defense.

44See Jervis(fn. i), at I86-2I4; George Quester,Offense Sys-

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

sidered defensive(as long as the other side has a similarcapability)be-

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

THE NATURE OF POWER

A third conceptual difficultyis created by the factthat in many en-

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

Axelrod points to the ironythatthe strategyof reciprocity(Tit-for-Tat)

Axelrod (fn. i).

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

PERCEPTIONS, VALUES, AND SELF-INTEREST

The finalset of problemsconcernspsychology,beliefs,and values. In

PSYCHOLOGICAL IMPEDIMENTS TO COOPERATION

Robert Keohane has noted thatcooperationmay be explained in part

This content downloaded from 129.72.2.27 on Sun, 22 Sep 2013 02:31:14 AM

other's behavior will be crucial. Thus, how they react is influencedby