The Agent-Structure Problem in International Relations Theory

Author(s): Alexander E. Wendt

Reviewed work(s):
Source: International Organization, Vol. 41, No. 3 (Summer, 1987), pp. 335-370
Published by: The MIT Press
Stable URL: http://www.jstor.org/stable/2706749 .
Accessed: 28/09/2012 19:06

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 support@jstor.org.

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to International
Organization.

http://www.jstor.org
The agent-structure
problem
in international
relationstheory
AlexanderE. Wendt

Two theories,neorealism and world-system theory,stronglyinfluence con-

temporary academicdiscourseaboutinternational relations.Bothclaimto
provide"structural" explanations ofhowstatesbehaveintheinternational
system.Despitetheircommoncommitment to structural
analysis,however,
theirunderstandingofsystem"structure," andtherefore ofstructuralexpla-
nation,is quitedifferent.
Neorealistsdefineinternational systemstructures
in termsoftheobservableattributes oftheirmemberstates(the"distribu-
tionofcapabilities"),and as a result,theyunderstand theexplanatory role
ofthosestructures in individualist
termsas constraining thechoicesofpre-
existingstateactors.World-system theorists,on theotherhand,defineinter-
nationalsystemstructures intermsofthefundamental organizing principles
of thecapitalistworldeconomywhichunderlieand constitute states,and
thustheyunderstand theexplanatory roleofstructures instructuralistterms
as generatingstateactorsthemselves. These differences,and theirimplica-
tions,have yetto be explicatedin theinternational In
relationsliterature.1
An earlierversionof thisarticlewas presented at the 1986meeting of theInternational
I wanttothankHaywardAlker,RichardAshley,Raymond
StudiesAssociation. Duvall,Jeffrey
Isaac, BrianJob,StephenKrasner,PeterManicas,David Sylvan,JuttaWeldes,and two
anonymous fortheirhelpful
referees comments and suggestions on previousdrafts.
1. Thereare a numberof discussionsof themeanings and uses of "structuraltheory"in
neorealismandworld-system theory, butas faras I know,noneexplicitly comparesordiffer-
and world-system
entiatestheneorealist approachesto structure and structuralanalysis.On
neorealismsee,forexample,KennethWaltz,Theory ofInternationalPolitics(Reading,Mass.:
Addison-Wesley, 1979),and RobertKeohane,"TheoryofWorldPolitics:Structural Realism
and Beyond," in Ada Finifter,ed., Political Science: The State of theDiscipline (Washington,
D.C.: APSA, 1983).The bestcritiqueofneorealism's conception ofstructureis RichardAsh-
ley's "The PovertyofNeorealism,"International Organization 38 (Spring1984),pp. 225-86.
On world-system theorysee Immanuel "The RiseandFutureDemiseoftheWorld
Wallerstein,
CapitalistSystem:ConceptsforComparative Analysis,"Comparative Studiesin Societyand
History16(September 1974),pp. 387-415,andChristopherChase-Dunn andRichardRubinson,
"Towarda Structural Perspective on theWorld-System," Politicsand Society7 (no. 4, 1977),
pp. 453-76. The critiqueof world-system theorythatcomesclosestto myconcernsin this
ThedaSkocpol's"Wallerstein's
articleis probably WorldCapitalist System:A Theoretical and
HistoricalCritique,"American JournalofSociology82 (March1977),pp. 1075-90.
InternationalOrganization41, 3, Summer 1987
andtheMassachusetts
C) 1987bytheWorldPeace Foundation ofTechnology
Institute
336 International
Organization

thisarticle,I wantto beginto clarifyand contrastthenatureof structural

analysisineach ofthesetwotraditions. Myprimary interest,however,is to
critiquetheconceptionsof structural theoryfoundin each ofthem,and to
use thiscritiquetomotivate thedevelopment ofa newapproachto structural
theorizing aboutinternational relationsadaptedfromtheworkof"structura-
tiontheorists"in sociology.2 Thisapproachand theresearchagendaitim-
plies, in turn,requirea foundation in realistphilosophyof science (or
"scientific realism"3),arguablythe "new orthodoxy"in thephilosophy of
naturalscience,butas yetlargelyunacknowledged bypoliticalscientists.4
As structural theoriesof international relations,neorealismand world-
systemtheory andthusmight
differ, be compared, alonga number ofdimen-
sions:substantive claims,predictive power,scope,and parsimony, among
others.Whilethesedifferences are important, theyare, I think,strongly
conditioned by a morefundamental differenceofontology:neorealism em-
bodies an individualist ontology,whileworld-system theoryembodiesa
holisticone. A usefulway to capturethe natureand implications of this
difference is to evaluatethetwo theoriesin termsof theirunderlying as-
sumptionsabout the relationship of systemstructures to humanagents.
Despitetheircommitment to "structural"ratherthan"agentic"theorizing,
likeall structuraltheoriestheybothpresupposesome theory ofwhatis being

2. The term"structuration
theory"is sometimes narrowlyidentified
withtheworkofAn-
thonyGiddens,who has articulated
its basic problematic
in his CentralProblemsin Social
Theory(Berkeley: Universityof CaliforniaPress, 1979) and The Constitutionof Society: Out-
lineoftheTheory ofStructuration(Cambridge, U.K.: PolityPress,1984).In "On theDetermi-
nationof Social Actionin Space and Time," Societyand Space 1 (March1983),pp. 23-57,
however,NigelThrift uses the termmorebroadlyas a genericlabel fora groupof social
theorieswhichsharecertainfundamental assumptions abouttheagent-structure relationship;
thisgroupincludes,butis notlimitedto, PierreBourdieu,Outlineof a Theoryof Practice
(Cambridge: Cambridge UniversityPress,1977),Roy Bhaskar,ThePossibility ofNaturalism
(Brighton, U.K.: Harvester Press,1979),andDerekLayder,Structure, Interaction,andSocial
Theory(London:Routledge& KeganPaul, 1981).Since mypurposein thispaperis less to
advanceGiddens'sideas (indeed,I willrelymoreon BhaskarthanGiddens)thanto demon-
stratetherelevanceoftheoverallproblematic forinternational relationstheory,I shallfollow
Thrift'smoreinclusiveuse oftheterm.
3. Scientific
realism(or simply"realism")is notrelatedto politicalrealismorneorealism in
internationalrelations.
4. Whether or notscientificrealismis the "new orthodoxy" in thephilosophy of natural
scienceis undoubtedly a contentiousissueamongrealistsandempiricists, butithasinanycase
made sufficient inroadsthatthe MinnesotaCenterforthe Philosophy of Science,longan
important bastionof empiricism, helda year-long institutein 1985/86 which,amongother
things,focusedexplicitly on thatquestion.American politicalscientistsgenerallyseemto be
unawareoforuninterested inthisdebateanditspotential implications forpoliticalscience.To
myknowledge, theonlydiscussionsof scientific realismin international relationsare British:
JohnMaclean,"MarxistEpistemology, Explanations ofChangeandtheStudyofInternational
Relations,"in BarryBuzanandR. J.BarryJones,eds., ChangeintheStudyofInternational
Relations:The Evaded Dimension(London:FrancesPinter,1981),pp. 46-67, and Richard
Little, "The Systems Approach," in Steve Smith, ed., InternationalRelations: Britishand
American
Perspectives
(Oxford:Blackwell,1985),pp. 79-91.
 Agent-structure
problem 337

structured, humanor organizational agents,and oftheirrelationship to so-

cial structures.Putmoregenerally, all socialscientifictheoriesembodyan at
least implicitsolutionto the "agent-structure problem,"whichsituates
agentsand socialstructures inrelationto one another.These solutionshelp
determine a theory'sunderstanding of,and therelativeexplanatory impor-
tance it attachesto, structural analysis.Whilegenerating verydifferent
understandings ofstructural
theory, however,I shallarguethattheneoreal-
istandworld-system solutionstotheagent-structure problemare,inat least
one respect,verysimilar,and thatthissimilarity createsa commonfunda-
mentalweaknessin thesetheoriesas "structural"approachesto interna-
tionalrelations.Structuration theory, inturn,is a responseto thiscommon
weaknesswhichbothsubsumesand pointsbeyondneorealism and world-
systemtheory.
In Section1, I examinethenatureoftheagent-structure "problem"and
brieflyidentify theprincipalkindsofsolutionsto it.I argueinSection2 that
neorealismand world-system theoryembodytwo of thesesolutions,the
methodological individualistand structuralist ones, respectively. Despite
important differences betweenthem,each of theseapproachessolves the
agent-structure problemby makingeitherstateagentsor systemstructures
ontologically primitive units.The resulting effect on neorealism and world-
systemtheoryis an inability to explaintheproperties and causalpowersof
theirprimary unitsofanalysis,a weaknesswhichseriously undermines their
potentialexplanationsof stateaction.This situationcan be prevented by
adoptingan approachto theagent-structure problemwhichdoes notpre-
clude a priorimakingbothagentsand structures "problematic"or "de-
pendentvariables." In Section 3, I describethis third,structurationist
approach,and its foundations in realistphilosophyof science. Since the
utilityof structurationtheoryas a meta-theoretical framework forinterna-
tional relationsultimately depends on its abilityto enrichsubstantive
theorizing andconcreteempirical research,itsvaluecannotbe convincingly
demonstrated in a programmatic articlesuch as thisone. It is possible,
however,to indicatesomeofthechangeswhicha structurationist perspec-
tivesuggestsare necessaryinthecontemporary researchagendaininterna-
tionalrelations.Towardsthisend, in Section4, I examinesome general
epistemological and theoretical implications of structuration theoryforthe
explanation ofstateaction.In theconclusion,I return to someimplications
of scientific
realismforsocial scientific research.

1. The agent-structure
problem

The agent-structure
problemhas itsoriginsin twotruismsaboutsociallife
inquiry:1) humanbeingsand their
whichunderliemost social scientific
338 International
Organization

organizations are purposeful actorswhoseactionshelpreproduceor trans-

formthe societyin whichtheylive; and 2) societyis made up of social
relationships,whichstructure theinteractions betweenthesepurposeful ac-
tors.Taken togetherthesetruismssuggestthathumanagentsand social
structures are,inone wayor another, interdependent
theoretically or mutu-
ally implicatingentities.Thus, the analysisof actioninvokesan at least
implicitunderstanding of particularsocial relationships(or "rules of the
game") in whichtheactionis set-just as theanalysisof social structures
invokessomeunderstanding oftheactorswhoserelationships makeup the
structuralcontext.It is thena plausiblestepto believethattheproperties of
agentsand thoseof social structures are bothrelevantto explanations of
socialbehavior.Andinfact,although inverydifferent ways,neorealism and
world-system theorydo use theproperties bothofstates(powers,interests)
and ofsystemstructures relations
(polarity, ofunequalexchange)to explain
statebehavior.The "problem"withall thisis thatwe lack a self-evident
wayto conceptualize theseentities andtheirrelationship. Thisabsenceofa
single,immediately compelling conceptionof the agent-structure relation
has spawneda varietyof conceptualizations of therelationshipacrossthe
social sciences,5each reflecting theparticular philosophical and practical
commitments of itsparenttheoretical discourse.(My own adoptionof the
languageof "agents" and "structures,"therefore, is nottheory-neutral.)
Despite theirmanydifferences, however,the "agent-structure," "parts-
whole,""actor-system," and "micro-macro" problemsall reflectthesame
meta-theoretical imperative-theneed to adopt,forthe purposeof ex-
plainingsocialbehavior,someconceptualization oftheontological and ex-

problemhas, in variousguises,recently
5. The agent-structure emergedas somethingofa
throughout
cottageindustry ofthiswork:ingeography,
thesocialsciences.Fora sampling see
DerekGregory, "HumanAgencyand HumanGeography,"Transactions of theInstituteof
Geographers
British 6 (no. 1, 1981),pp. 1-18,andDerekGregoryandJohnUrry,eds., Social
Relations and Spatial Structures(London: MacMillan, 1985); in sociology, in additionto the
workofGiddensand Bhaskaralreadycited,see AlanDawe, "TheoriesofSocialAction,"in
Tom Bottomore and Robert Nisbet, eds., A History of Sociological Analysis (London:
and AaronCicourel,eds., Advancesin Social
Heinemann,1979),and KarinKnorr-Cetina
Theory:Toward an Integrationof Micro and Macro-Sociologies (London: Routledge& Kegan
see PhilipAbrams,HistoricalSociology(Ithaca:CornellUniver-
Paul,1981);in socialhistory,
sityPress,1982),and Christopher Lloyd,Explanation in Social History(Oxford:Blackwell,
1986);in thephilosophy ofsocialscience,see JohnO'Neill,ed., ModesofIndividualism and
Collectivism(New York:St. Martins,1973),andDavid-Hillel Ruben,TheMetaphysics ofthe
Social World(London:Routledge & KeganPaul,1985);inMarxisttheory, see EdwardThomp-
son's polemic in The Povertyof Theoryand OtherEssays (New York: MonthlyReview Press,
MarxismofLouis Althusser,
1978)againstthestructural andthecommentaries on thisdebate
by PerryAnderson, ArgumentswithinEnglishMarxism(London:Verso, 1980),and Nicos
Mouzelis,"Reductionism inMarxistTheory,"Telos45 (Fall 1980),pp. 173-85;andininterna-
Politics,and JamesRosenau,"Before
tionalrelations,see Waltz,Theoryof International
Cooperation:Hegemons,Regimes,andHabit-Driven ActorsinWorldPolitics,"International
Organization40 (Autumn1986),pp. 849-94.
 problem 339
Agent-structure

planatory relationship betweensocialactorsor agents(inthiscase, states)6

and societalstructures (in thiscase, theinternational system).
The agent-structure problemis reallytwointerrelated problems, one on-
tologicaland the otherepistemological. The first,and morefundamental,
issue concernsthenatureofbothagentsand structures and,because they
are in some way mutually implicating, of theirinterrelationship.In other
words,whatkindofentities arethese(or,inthecase ofsocialstructures, are
theyentitiesat all?), and how are theyinterrelated? Thereare two basic
waysto approachthisquestion:bymaking oneunitofanalysisontologically
primitive, or by givingthemequal and therefore irreducibleontological
status.Dependingon whichentity is madeprimitive, theseapproachesgen-
eratethreepossibleanswersto theontologicalquestion,whichI willcall
individualism, structuralism, and structurationism. Neorealismand world-
systemtheory embody,respectively, thefirst
twoofthesepositions, bothof
whichultimately reduceone unitofanalysisto theother.Thus,neorealists
reducethestructure ofthestatesystemto theproperties andinteractions of
its constituent elements,states,whileworld-system theorists
reducestate
(and class) agentsto effectsof thereproduction requirements of thecapi-
talistworldsystem.The structurationist approach,on theotherhand,tries
to avoidwhatI shallargueare thenegativeconsequencesofindividualism
and structuralism by givingagentsand structures equal ontologicalstatus.
Farfrombeinga mindlesssynthesis ofthe"bestofbothworlds,"however,
thestructuration projectrequiresa veryparticular conceptualization ofthe
agent-structure relationship. Thisconceptualization forcesus to rethink the
fundamental properties of (state)agentsand systemstructures. In turn,it
permits us touse agentsandstructures toexplainsomeofthekeyproperties
of each as effectsof the other,to see agentsand structures as "co-
determined" or "mutuallyconstituted" entities.
The mannerin whicha social theoryaddressestheseontologicalissues
conditions itsapproachto theepistemological aspectoftheagent-structure
problem, thechoiceandintegration ofdifferenttypesofexplanations within
theoriesofsocialbehavior.Thisproblemactuallyraisestwoepistemological
issues. The firstis the choice of the formof explanationcorresponding
respectively to agentsand structures. This choicedependslargelyon the
kindsofproperties ofagentsand structures thathavebeendeemedcausally

6. Recenttheoretical workhas conceptualized thestatebothas an agentandas a structure;

see,forexample,RogerBenjaminandRaymond Duvall,"The CapitalistStateinContext,"in
RogerBenjamin andStephenElkin,eds., TheDemocratic State(Lawrence,Kans.: University
ofKansasPress,1985),pp. 19-57.Forpurposesofthispaper,I assumewithneorealists thatthe
stateis an agent,a movewhichcan bejustified inpartbecausetheorganizing principlesofthe
statesystemconstitute statesas individual
choice-making unitswhichareresponsible fortheir
actions.My subsequent arguments aboutthewayinwhichsystemstructures constitutestates
as agentsshouldnot,however, be seenas excludinga conceptionofthestateas a structure
of
politicalauthorityin whichgovernmental agentsare inturnembedded.
340 International
Organization

significant.Thus,approachesto social inquirythatconceiveof humanbe-

ingsas reflective,goal-directedsubjects,such as rationalchoice theory,
generateagent-explanations thatare, broadlyspeaking,"interpretive"-
thatis, castintermsofthegoals,beliefs,and self-understandingsofagents.
On theotherhand,approachesthatconceiveof humanbeingsas nothing
morethancomplexorganisms processingstimuli-suchas behaviorism-
generateagent-explanations thatare moremechanistically causal in form.
The situationis similarwithrespectto "structural"explanations.Social
theoriesthat reduce systemstructures to the propertiesof individuals
usuallyconstruethe explanatory as one of constrain-
role of structures
ingthechoicesofpre-existing agents,whilethosethatconceptualize system
structures as irreducibleentitiesunderlying agentstypicallyunderstand
structures as generating or explainingagentsthemselves.The second
epistemological issue concernsthe relativeimportanceof agent-expla-
nationsand structure-explanations,ofwhatever type,in socialtheory.This
issue is of secondaryimportance in thisarticlebecause neorealistsand
world-system agreethatan adequateinternational
theorists relations theory
mustbe morestructure- thanagent-oriented. Theyunderstand thisrequire-
mentin verydifferent ways,however-a disagreement whichI willlater
showis linkedto theway theyapproachtheontologicaldimension of the
agent-structureproblem.

2. Reductionismand reification
in internationalrelationstheory

In thissectionI wantto: 1) comparetheconceptions of"structural" theory

foundin neorealismand world-system theory;and 2) show that,despite
important theseconceptionssharea commonapproachto the
differences,
agent-structureproblem,andthatthisapproachprecludesan explanation of
the essentialpropertiesof theirrespectiveprimitive units.This inability
leadsto assumptionsaboutprimitive unitsthatarewithouttheoreticalfoun-
dation,a movewhichinturnundermines thetheories'explanationsofstate
actionin theinternationalsystem.I shallarguethatthiscommonlimitation
is a functionof thebasic assumptions and internallogicof each theory's
approachto the agent-structure problem,and thattheytherefore cannot
overcomeit withinthetermsoftheirbasic ontological and epistemological
commitments.

a. Neorealism
and anti-
On the surface,at least, neorealistshave strongstructural
reductionist
commitments. In his discussionof the of
nature systemicand
 problem 341
Agent-structure

reductionist theories,KennethWaltzdefinesthe latteras theorieswhich

explaintheforeign policybehaviorofstatesexclusively intermsofcausesat
thenationallevelofanalysis.7Lenin'stheory ofimperialism, forexample,is
reductionist because it explainsexpansionist behaviorin termsof theac-
cumulation dynamicsof nationalcapitalism.Waltzcriticizessuchtheories,
correctly I think,forignoring theintervening role playedby international
systemstructures in the translationof domesticimperatives intoforeign
policybehavior.Neorealistsavoidthistypeofreductionism byadopting the
systemic(but, I shall argue,not "structural")logic and conceptualap-
paratusof micro-economic theory.8This movepermits neorealiststo inte-
gratewithina coherenttheoretical framework thestate-centric approachof
classicalpoliticalrealismwiththe systemicapproachof international sys-
temstheory,and thusto developa conceptionoftheagent-structure rela-
tionshipin international relationswhichrecognizesthecausal roleofboth
stateagentsand systemstructures.
In viewofneorealists'desireto avoidmicro-level reductionism,however,
itis ironicthattheirsolutionto theagent-structure problemis, ina different
anddeepersense,reductionist. The kindof "reductionism" whichneoreal-
istsopposeis definedas theorywhichtriesto explainbehaviorin termsof
strictlyagent-levelproperties.Thisrejectionofwhatmight be calledexplan-
atoryreductionism does notinitself,however,imposeanyparticular restric-
tionon theontological issue of how systemstructures shouldbe defined,
sincean opposition toagent-level explanationsis analytically
independent of
how systemstructures, once recognizedas causallysignificant, shouldbe
theorized.Thus,neorealists'individualist definition ofthestructure ofthe
internationalsystemas reducibleto theproperties of states9-tothedistri-
butionof capabilities-isperfectly consistentwiththeimportant rolethat

7. Waltz, Theoryof InternationalPolitics, p. 18.

Princeton
8. RobertGilpin,Warand Changein WorldPolitics(Princeton: University
Press,
1981);RichardAshley,"ThreeModes of Economism,"International StudiesQuarterly
27
(December 1983),pp. 463-96; RobertKeohane, AfterHegemony: Cooperationand Discord in
the WorldPoliticalEconomy(Princeton:
Princeton Press,1984);DuncanSnidal,
University
"The Game Theoryof InternationalPolitics," WorldPolitics 38 (October 1985), pp. 25-57.
9. Ashleythoroughly critiquestheindividualist
(andempiricist) foundationsoftheneorealist
conception of internationalsystemstructure in his "Povertyof Neorealism,"especiallypp.
238-42.Itis important tokeepinmind,however, thatinTheory ofInternationalPolitics,Waltz
startsoutwiththreedefining features
ofpoliticalstructures:1)theprinciple
according towhich
theyareorganized, 2) thedifferentiation
ofunitsandtheirfunctions, and3) thedistribution
of
acrossunits.Thisdefinition
capabilities can be usedto support a generative
approachto struc-
turaltheorizing,as JohnRuggieshowsin hisDurkheimian reconstruction
ofWaltzin "Con-
tinuityandTransformation intheWorldPolity:Towarda Neorealist Synthesis," WorldPolitics
35 (January 1983),pp. 261-85.Despitethispromising beginning, however,Waltzand other
neorealists
arguethatthefirst twofeaturesofthisdefinitiondon'tapplytointernationalpolitical
leavingus inpracticewithan individualist
structures, conception ofstructureas thedistribution
ofcapabilities.For an argument thatlinksthisresultto a lingeringneorealistcommitment to
positivism,see Little,"The SystemsApproach."
 Organization
342 International

systemstructures playin neorealist explanations ofstatebehavior.Indeed,

in bothits decision-and game-theoretic versions'0neorealism, likemicro-
economics,is characterized by "situationaldeterminism," by a modelof
actionin whichrationalbehavioris conditioned or evendetermined by the
structure of choicesituations.11 The pointis notthatneorealists engagein
explanatory reductionism (whichtheydo not),butrather thattheirdefinition
ofsystemstructure is characterizedbyontological reductionism. Thisdefini-
tionleads to an understanding ofsystemstructures as onlyconstrainingthe
agencyofpreexisting states,ratherthan,as in world-system and structura-
tiontheory,as generatingstateagentsthemselves. Thisfollowsinevitably
fromits (ontologically) reductionistdefinition of systemstructure: system
structures cannotgenerateagentsiftheyare reducedto theproperties of
agentsin thefirstplace. Despiteitsstrongly systemic focus,then,neoreal-
ism'sviewoftheexplanatory roleofsystemstructures is decidedlystate-or
agent-centric. It sees systemstructures in themannerinwhichtheyappear
to states-as given,externalconstraints on theiractions-ratherthanas
conditions ofpossibility forstateaction.
Fromtheperspective ofinternationalrelations theory, themostimportant
weaknessofneorealism's approachtotheagent-structure
individualist prob-
lemis thatitfailsto providea basisfordeveloping an explicittheoryofthe
state.It is nothardto see whythismustbe thecase. Theorizing abouta
particular kindofindividual unit,likethestate,can takeeitheroftwoforms
(or both).The first,"reductionist" strategy is to explaintheindividual in
termsof its internalorganizational structure.12Whilethisapproachmay
explainsome of the causal properties of theindividual, it neglectstheir-
reducibly social contentof manyindividual-level predicates.'3Thus,while
theinternal physiological structureofa capitalistmayexplainsomedisposi-
tionsand actions,we cannotexplainhisor herbehavioras thatofa "capi-

10. BruceBuenode Mesquita,The WarTrap(New Haven:Yale University Press,1981);

Snidal,"The GameTheoryofInternational Politics."Despiteimportant between
differences
thetwoversionsovertheconceptualization ofchoicesituations ininternational both
relations,
are based on an individualist of the structure
definition systemas the
of the international
ofcapabilities.
distribution
Determinism
11. See SpiroLatsis,"Situational inEconomics,"British JournalforthePhi-
losophyofScience23 (August1972),pp. 207-45,andthereplybyFritzMachlup,"Situational
Determinismin Economics," BritishJournalfor the Philosophy of Science 25 (September
1974),pp. 271-84.
12. Examplesofsuchan approachininternational relationsmightincludeGrahamAllision,
Essence of Decision (Boston:LittleBrown,1971)and JohnSteinbrunner, The Cybernetic
Theory ofDecision(Princeton: University
Princeton Press,1974).
13. This is probablythe mostpersistently citedproblemin the individualist program of
reducing all socialscientific to theproperties
explanations ofindividualsor theirinteractions.
See, forexample,MauriceMandelbaum, "Societal Facts," BritishJournalof Sociology6
(1955); Steven Lukes, "MethodologicalIndividualism Reconsidered,"BritishJournalof
Sociology19(June1968),pp. 119-29;HaroldKincaid,"Reduction, Explanation,andIndivid-
ualism,"Philosophy ofScience53 (December1986),pp. 492-513.
 problem 343
Agent-structure

talist"exceptin termsof theindividual'ssocial relationsto otheragents.

The secondapproach,then,is to explicateandtheorizetheserelationally or
structurally defined properties, that is, to develop a social theory of the
state.Yet theseare preciselytheproperties thatneorealism abstractsaway
fromwhenit reducessystemstructures to a distribution of properties of
preexisting individuals.The neorealist'sindividualist conceptualization of
systemstructure is thereforetooweakto support a socialtheory ofthestate:
systemstructures cannot generate agents ifthey are defined exclusively in
termsof those agentsin the firstplace. The consequenceof makingthe
individual ontologicallyprimitive, inotherwords,is thatthesocialrelations
invirtueofwhich that individual is a particular kindofagentwithparticular
causal properties mustremainforever opaque and untheorized.
One responseto thismight be to argue, as Waltz does,"4 thatthedevelop-
an
mentof explicit theory of the state is not integral to the development of
systemictheoriesof international relations. Yet clearly some model of the
stateis necessaryto buildsystemictheoriesof international relations, and
thismodelcan strongly affectthecontentofthosetheories.Thus,to argue
thatthe structure of the industrialized states'interaction withrespectto
internationaltradeis an n-person iterated Prisoner'sDilemma,andthatfree
tradeis therefore alwaysproblematic, requiresa certainunderstanding of
thosestatesand theirinterests and powers.The issue,then,is notwhether
someunderstanding of the stateis necessaryto buildsystemictheories(it
is), but whetherthatunderstanding followsfroma theory, groundedin a
coherentsetofpropositions withsomecorrespondence to reality,or simply
froma setofpre-theoretical assumptions, grounded inintuition orideology.
Whateverits advantagesin termsof analyticalconvenience,a relianceon
untheorized assumptions aboutprimitive termsleaves us unableto justify
particularconceptualizations of interaction situations and leads, therefore,
to an untenable"as if" approachto systemic theorybuilding. 15 Thus,with-
out an explicittheoryof the state'spowersand interests in international
trade,withouta theoryofthe"rulesofthegame,"itcannotbe determined
whether ornotthisgamereallyis a Prisoner'sDilemmarather than,as some

14. "Reflectionson TheoryofInternational Politics:A Responseto MyCritics,"inRobert

Keohane,ed., Neorealismand Its Critics(New York:ColumbiaUniversity Press,1986),p.
340.
15. The debateoverthevalidityof theoriesbuilton theassumption thatthesocialworld
operates"as if" certainthingsweretrueis a longone, so mysayingthatsuchreasoning is
"untenable"is, of course,contentious.The termsofthedebatewerefirst definedby Milton
Friedman's"The Methodology ofPositiveEconomics,"in hisEssaysinPositiveEconomics
(Chicago:ChicagoUniversity Press,1953),a piece whichinitiated a livelydebatewithPaul
SamuelsonandothersinthepagesoftheAmerican EconomicReviewintheearly1960s.Fora
cogentargument
particularly that"as if" reasoning evenwiththelogicalempiri-
is inconsistent
cistconceptionof scientific
explanation thatinformed Friedman'sseminalcontribution,see
TerryMoe, "On the Scientific Statusof RationalModels," AmericanJournalof Political
Science23 (February1979),pp. 215-43.
344 International
Organization

neo-Marxists mightargue,a purecooperation game.16 And,without a com-

pellingargument to thateffect,we cannotknowifa failedprediction is due
to an errorinoursystemic theoryorto a misspecificationofthe(pretheoret-
ical) rules of the game. Withouta theoryof the state,in otherwords,
neorealists'efforts to buildcompelling systemictheoriesof international
relationsare seriouslycompromised.
This consequencedoes not mean thatthe actual,micro-economic, as-
sumptionsneorealistsmake about the stateare wrongor misleading(al-
thoughI thinkthey probablyare)-just that they lack a theoretical
foundation withsomedemonstrated correspondence to reality.The resultis
a tendencyto arguethatstatesact "as if" theymaximize,forexample,
powerand wealth,and a corresponding to buildcrediblesystemic
inability
theoriesofinternational relations.Waltz'sseparation oftheoriesofthestate
and of international relationsnotwithstanding,ifneorealists wantto avoid
theseproblems theymustultimately developa socialtheory ofthestate,that
is, theymustmakethe statetheoretically "problematic."This wouldre-
quirean attempt to theorizedirectlyaboutthegenerative structures of the
worldand domesticpolitical-economy whichconstitute statesas particular
kindsofagentswithcertaincausalpowersandinterests. Neorealists already
havean implicit theoryofthesesocialrelations(iftheydidnot,theycould
notattribute anycausalpowersorinterests to stateagents),buttheycannot
makethistheoryexplicit,and therefore falsifiable,
as longas theytreatthe
stateas ontologically primitive.Since thesocialrelationswhichconstitute
statesas stateswillbe potentially unobservable andirreducible to theprop-
ertiesof statesthemselves, however,such a theoretical reorientation will
requirea non-individualist and non-empiricist understanding of system
structuresand structural analysis,an understanding of structure as some-
thingmorethana distribution ofcapabilities.

b. World-system
theory
World-system theory offers
suchan understanding
ofstructureandthus,at
leastwithrespectto itsconceptualization and structural
ofstructure analy-
problemshiftover neorealism.In one
sis, can be seen as a progressive
crucialrespect,however,world-system theoristsduplicatethe neorealist
16. As faras I know,no neo-Marxist has used game-theoreticlanguageto characterize
economicrelationsbetweentheadvancedindustrialized
international But clearly,
countries.
becauseoftheirverydifferenttheoreticalunderstanding ofthestate,neo-Marxistscholarsare
to see thoserelations
muchless likelythanneorealists andtherefore
inmercantilist, politically
terms;see, forexample,RobinMurray,"The Internationalization
fragile, of Capitaland the
Nation-State,"NewLeftReview67 (May-June1971),pp. 84-109,andJohnWilloughby, "The
Changing RoleofProtectionintheWorldEconomy,"Cambridge Journal ofEconomics6 (June
1982),pp. 195-211.Theissueinthisarticle,ofcourse,is notwhichviewis actuallycorrect,but
ratherhowto developan approachto theagent-structure problemwhichensuresat leastthe
ofdetermining
possibility whichis correct,thatis, ofdevelopinga theory ofstatesin interna-
tionaleconomicstructures.
 problem 345
Agent-structure

approachto theagent-structure problem:theyat leastimplicitly makeone

entity primitive,in thiscase thestructure oftheworldsystem,andthentry
to reduceotherentities,suchas stateand class agents,to its effects.The
resultofthisstrategy, I shallargue,is thatworld-system theorists reifythe
structure oftheworldsystemand thus,likeneorealists, are unableevenin
principle to explainitsessentialproperties. A socialstructure is reified
when
"it is treatedas an objectanalytically independent oftheactionsbywhichit
is produced.",17A solutionto theagent-structure problem, then,engagesin
reification whenitobjectifies socialstructures without recognizing thatonly
humanactioninstantiates, reproduces,and transforms thosestructures. I
shouldemphasizethatreification presupposes atleastan implicit conception
oftherelationship ofagentsto social structures: structures have reproduc-
tiverequirements which,forwhateverreason,agentspassivelyimplement.
The problemwithreification, therefore, does notconcerntheinclusionor
exclusionofagentsperse fromsocialscientific theories(sincetheymustbe
included),butratherthetermsoftheirinclusionintothosetheories.
ImmanuelWallerstein's solutionto theagent-structure problemhas the
samegeneralform,and thusthesame strengths and weaknesses,as Louis
Althusser's structuralMarxistsolution. 18 Like Althusser's insistence on the
"absoluteontologicalpriority of thewholeovertheparts,"19the core of
Wallerstein's approachis theproposition thattheonlymeaningful unitof
analysisin comparativeor international politicaleconomyis the whole
worldsystem.Moreover,bothscholarsaccepttheconceptanddiscourseof
"totality,"20 ofsocialwholesthatareirreducible, evenbycomposition laws,
to theirconstituent elements.As a result,Wallerstein, likeAlthusser, con-
ceptualizes"structure"in structuralist or generative termsratherthanin
termsoftheobservablerelationsbetween,or properties of,primitive indi-
viduals.Thatis, incontrast totheneorealist definition
ofsystemstructure as

andThomasWilson,"On theReification
17. DouglasMaynard inScott
ofSocialStructure,"
McNall and Gary Howe, eds., CurrentPerspectives in Social Theory,vol. 1 (Greenwich,
Conn.:JAIPress,1980),p. 287.
Marxistapproachto theagent-structure
18. The structural problemis discussedin Louis
AlthusserandEtienneBalibar,ReadingCapital(London:NewLeftBooks,1970),pp. 180-81,
andinStevenSmith,ReadingAlthusser Press,1984),pp. 192-200.
(Ithaca:CornellUniversity
It shouldbe noted,however,thatdespitethesimilarities betweenworld-system theoryand
Marxism
structural withrespecttotheirunderstandings they
relationship,
oftheagent-structure
differinimportant ofthecapitalist
wayson otherissues,suchas theconceptualization modeof
production. See, forexample,GaryHowe and AlanSica, "PoliticalEconomy,Imperialism,
and theProblemof World-System Theory,"in McNalland Howe, Current Perspectivesin
Social Theory,pp. 235-86.
19. Smith,Reading Althusser,p. 177.
20. Theydisagree,however,abouttheexactmeaning ofthisterm,thatis, aboutwhether
are"expressive"or"structured."
totalities see MichaelBurawoy,
Onthesedifferences, "Con-
temporary in MarxistTheory,"in ScottMcNall,ed., Theoretical
Currents in
Perspectives
Sociology(NewYork:St. Martins, 1979),pp. 16-39,andHarveyKaye,"Totality:ItsApplica-
and SocialAnalysisbyWallerstein
tionto Historical 6
andGenovese,"HistoricalReflections
(Winter1979),pp. 405-19.
346 International
Organization

thedistributionofcapabilitiesacrosspreexisting states,world-systemtheo-
ristsdefinethe structure of the worldsystemin termsof the underlying
organizing principlesoftheworldeconomy,andinparticular oftheinterna-
tionaldivisionoflabor,whichconstitute orgeneratestateandclassagents*21
The existenceand identity ofagentsas agents,andthereforeoftheircausal
powersand real interests, is produced,and therefore explained,by their
relationto thetotalityofthecapitalistworldsystem.Thus,stateagentsare
effectsof the structure of theworldsystemin muchthe same sense that
areeffects
capitalists ofthestructure ofthecapitalist
modeofproduction, or
slavesare effects of thestructureofmaster-slaverelationships.
Thisgenerative readingof world-system theorypresupposesan ontolog-
ical and explanatory distinction
between"internal"and "external"rela-
tions.22Internalrelationsare necessaryrelationships
betweenentitiesinthe
sensethattheentitiesdependupontherelationfortheirveryidentity. Stan-
dard examplesof internalrelationsare parent-childand master-slave;
neitherentityis conceivablewithout theexistenceoftheother.Thisimplies
thatan internalrelationcannotbe reducedto theproperties or interactions
of its memberelements;on the contrary, the relationship
itselfexplains
essentialproperties of each entity,and thusthecharacterof theirinterac-
tion.Externalrelations,on theotherhand,are contingent relationshipsor
interactionsbetweenentities, eachofwhichcan existwithout theother.The
factthattwostatesgo to waror signa peace,forexample,is notessentialto
theiridentityas states.Externalrelationsare importantforexplainingwhat
happensto entitiesinthecourseoftheirinteraction, buttheydo notexplain
theessentialcharacteristics ofthoseentitiesthemselves.
Generativestructuresare sets of internalrelations.To adopt a generative
approachtotheorizingaboutthestructureoftheinternationalsystem, there-
thestateas an effectofitsinternal
fore,is to understand relationsto other
statesand social formations
in the worldpolitical-economy, ratherthan
purelyas an untheorizedcause of internationalevents.The strength of a
generativeapproachto structural theorizing,then,is thatin contrastto

21. Mygenerativereading ofworld-system theory's ofstructure

conceptualization is charac-
theminority)
onlyofthe"qualitative"(and at thispoint,apparently
teristic schoolofworld-
represented,
systemtheorists forexample,byWallerstein andTerenceHopkins.Actually, the
recentdebatebetweenqualitativeand quantitative world-systemtheoristsis an interesting
a singleresearchcommunity
exampleofa quiteexplicittensionwithin betweenscientific realist
conceptions
and empiricist of theontologyand methodology analysis.On this
of structural
debatesee,forexample,RichardLittle,"The SystemsApproach,"inSteveSmith,ed., Inter-
national Relations, Britishand American Perspectives (Oxford,U.K.: Blackwell, 1985), pp.
71-91,and PeterTaylor,"The Povertyof International
Comparisons: Some Methodological
LessonsfromWorld-Systems Analysis"(DepartmentofGeography, UniversityofNewcastle-
upon-Tyne,1985).
ofthisdistinction
22. Usefuldiscussions Marx'sConcep-
includeBertellOllman,Alienation:
tionofMan in CapitalistSociety(Cambridge:
Cambridge UniversityPress,1976),pp. 26-40,
and Bhaskar, The Possibilityof Naturalism,pp. 53-55.
 Agent-structure
problem 347

neorealism'sindividualist approachit is able, in principle,to explainthe

causalpowersandinterests ofstateandclassagents,tomakethesetheoreti-
callyand empirically problematic.
A generative approachto structural analysisdoes not,however,require
thatsystemstructures be reified.World-system theorists beginto reifyso-
cial structures whentheyassert,or implyin theirconcreteresearch,not
onlythatcertainsocialrelationsare irreducible and constitute thestateand
class agentswhichare theirelements,butthattheserelationsare analyt-
icallyindependent of,andontologically priorto,thoseagents.Sucha viewis
impliedby the tendencyof world-system theorists to followAlthusserin
treatingstateand class agentsas no morethanpassive "bearers"of sys-
temicimperatives, a tendencywhichmanifests itselfin a relianceon func-
tional explanationsof state behavior.23Functionalismis evident,for
example,in world-system theorists'explanation ofgeneralwarsdirectly in
termsofthereproduction requirements oftheworld-system,24 requirements
whichbecometranslated (or translate themselves) intobellicosestateinter-
ests,as wellas in theinterpretation oftheriseof socialiststatesin sucha
way thatit is consistentwiththe reproduction requirements of theworld
system.25 Thisis notto say thatworld-system theorists wouldconsciously
argue that the reproduction of the world-system occurs withoutstate
agency-theyprobablywouldnot. But nonetheless in manyexplanations
theworldsystemineffect seemsto callforth itsownreproduction bystates;
thissuggeststhatat leastin theirconcreteresearch,world-system theorists
treatthe world-system as at some level operating independently of state
action,that in practicetheyreifytheworld-system. Whilethisresultmaybe
I it On
unintended,do notthink is accidental. thecontrary, itfollowsinevita-
the
blyfrom fundamental premiseof Wallerstein's holism-thatthewholeis
ontologically priorto itsparts.
The principalweaknessof a structuralist solutionto theagent-structure
is it
problem that,because cannot"explainanything but behavioralcon-
to
formity structural demands,' 26 it ultimately fails to providea basis for
the
explaining properties deep of structures themselves. It maybe, forex-
ample, that the division of the world system into three distinctstructural

23. Thistendencyis oneofthemostpersistently citedcriticisms

ofatleasttheearlyworkin
world-system theory.See, forexample,RobertDuplessis,"FromDemesneto World-System:
A CriticalReviewoftheLiterature on theTransition
fromFeudalismto Capitalism," Radical
HistoryReview3 (Fall 1976),pp. 3-41,andSkocpol,"Wallerstein's
WorldCapitalistSystem."
24. ChristopherChase-Dunnand JoanSokolovsky,"Interstate Systems,World-Empires
and theCapitalistWorld-Economy: A Responseto Thompson,"International StudiesQuar-
terly27 (September1983),pp. 357-67.
25. ChristopherChase-Dunn,"SocialistStatesin theCapitalistWorld-Economy," in his
Socialist States in the World-System(Beverly Hills: Sage Publications,1982), pp. 21-56.
26. DerekLayder,"ProblemsinAccounting
fortheIndividual Theo-
inMarxist-Rationalist
reticalDiscourse," BritishJournalof Sociology 30 (June 1979), p. 150.
348 International
Organization

positions(core,semi-periphery, periphery)is functional

forthereproduction
of capitalism,but this does not explainwhythe systemdevelopedthat
particular structure,nordoes itguaranteethatthatstructure willendure.27
Becauseoftheirpassiveconception ofstateandclass agency,world-system
theorists tendto fallintoan historicaldeterminism which,byignoring other
possiblehistorical trajectories,implicitlyassumesthattheevolutionofthe
worldsystemcouldnothave turnedout anyotherway. The limitations of
world-system theory'ssolutionto theagent-structure problem,and also its
similaritiesto thatofstructural Marxism,arein thisrespectno moreappar-
entthaninWallerstein's explanation ofa fundamental structuralchangelike
thetransition fromfeudalismto capitalism.28 Withouta recognition of the
ontological dependenceofsystemstructures on stateandclass agents,Wal-
lersteinis forcedintoan explanation ofthattransitionintermsofexogenous
shocksand theteleologicalimperatives of an immanent capitalistmodeof
production. Thisexplanation opposesRobertBrenner'smoreor less struc-
turationist one, whichdescribesthe transition in termsof a dialecticof
endogenousclass struggle and structural conditioning.29 This explanation
reflectsdifficulties
remarkably similarto thosestructural Marxistshavehad
in theorizing thetransitionfromone modeofproduction to another.30
World-system theorists,then,likeneorealists, treattheirprimitive units,
in thiscase thestructure oftheworldsystem,as givenand unproblematic.
This treatment leads themto separatethe operationof systemstructures
fromtheactivities ofstateandclass agents-in otherwords,toreifysystem
structures ina waywhichleadsto staticandevenfunctional explanations of
stateaction.The worldsystemis nottreatedas an historically contingent,
andtherefore continuously problematic, creationandrecreation ofstateand
class agents.I thinkthegreaterattention Wallerstein's laterworkgivesto
problemsof agencyindicatesan awarenessof thisdifficulty,31 and these
effortshave helpedto moveworld-system theoryawayfromtheexcessive
functionalism evidentin his earlycontributions and, perhaps,in mypor-
27. EmileDurkheim makesexactlythispointinTheRulesofSociologicalMethod(Chicago:
ChicagoUniversity Press,1938),p. 90,whenhe saysthat"to showhowa factis usefulis notto
explainhowitoriginated orwhyitis whatitis. Theuseswhichitservespresuppose thespecific
propertiescharacterizingit,butdo notcreatethem.The needwe haveofthingscannotgive
themexistence, norcan itconfertheirspecificnatureuponthem.Itis to causesofanothersort
thattheyowe theirexistence."
28. See, forexample,Wallerstein,TheModernWorld-System I (NewYork:AcademicPress,
1974),especiallychap. 1.
29. Brenner,"The Originsof CapitalistDevelopment: A Critiqueof Neo-Smithian Marx-
ism,"New LeftReview104(July-August 1977),pp. 25-92.
30. See, forexample,StephenResnickand RichardWolff,"The Theoryof Transitional
Conjunctures and theTransition fromFeudalismto Capitalism," ReviewofRadicalPolitical
Economics11(Fall 1979),pp. 3-22,andtheresponseinthesameissuebyHerbert Gintis,"On
theTheoryofTransitional pp. 23-31.
Conjunctures,"
ThePoliticsoftheWorld-Economy
31. See, forexample,I. Wallerstein, (Cambridge: Cam-
bridgeUniversity Press,1984),pp. 112-46.
 problem 349
Agent-structure

trayalofworld-system theory.Indeed,incontrast tothecompleteneglectof

generativestructures by neorealists,world-system theoristsseem to be
awareoftheneedto arriveat a solutionto theagent-structure problemthat
integratesgenerative structuresand stateand class agency.World-system
theory'sevolutiontowardsa greaterfocuson agencythus,in someways,
parallelsthedevelopment Marxismin thelaterworkofNicos
of structural
Poulantzas.32As withPoulantzas,however,itis difficultto see howagency
canbe brought on an ontologicalandexplanatory parwithsystemstructures
withoutexplicitly jettisoningthe strictholismand structural reification
whichcharacterize approachto the agent-structure
the structuralist prob-
lem. Nonetheless,ifonlybecause it at least recognizestheexistenceand
explanatory roleofgenerative theworld-system
structures, solutionto the
problemis considerably
agent-structure closertothatofstructuration theory
thanis theneorealistsolution.

c. Summary
In thissection,I havetriedto identify important differencesbetweenthe
neorealistand world-system understandings of "structural"explanation,
and to linkthesedifferences to theirdifferentsocialontologies.I havealso
attempted to show that,despitethesedifferences, neorealismand world-
systemtheorysharea common,underlying approachto theagent-structure
problem:theybothattempt tomakeeitheragentsorstructures intoprimitive
units,whichleaves each equallyunableto explaintheproperties of those
units,and therefore and explanatory
to justifyits theoretical claimsabout
stateaction.The obviousimplication of thisargument is thatneitherstate
agentsnorthedomesticandinternational systemstructures whichconstitute
themshouldbe treatedalwaysas givenor primitive units;theoriesofinter-
nationalrelationsshouldbe capable of providing explanatory leverageon
both.This does notmeanthata particular researchendeavorcannottake
somethingsas primitive: scientificpracticehas to startsomewhere. It does
mean,however,thatwhatis primitive in one researchendeavormustbe at
least potentiallyproblematic (or functionas a "dependentvariable")in
another-thatscientistsneed theoriesof theirprimitive units.Notwith-
standingtheirapparentaspirationto be generaltheoriesof international
relations,the individualistand structuralistontologiesof neorealismand
world-system theoryprecludethedevelopment ofsuchtheories.In contrast,
approachto theagent-structure
a structurationist problemwouldpermitus
to developtheoretical accountsofbothstateagentsand systemstructures
without engagingin eitherontological reductionism or reification.

32. State, Power, Socialism (London: Verso, 1978); see also Bob Jessop,Nicos Poulantzas:
Marxist Theoryand Political Strategy(New York: St. Martins,1985).
350 International
Organization

3. An alternative
approachto theagent-structure
problem

Structuration theoryis a relationalsolutionto theagent-structure problem

thatconceptualizesagentsand structures as mutuallyconstituted or co-
determined entities.Whatthismeanswillbecomemoreapparentlater,but
firstI wantto considerthephilosophical foundations ofstructurationtheory in
scientificrealism,andthecurrent debateofthelatterwithempiricism inthe
philosophy ofscience.Thisdigression is important to myargument because
of the continuing hegemonyof empiricist discourseon social scientists'
conceptualization of "science," and therealpossibility thatskepticsmight
use thatdiscourseto writeoffstructuration theory'sgenerativeapproachto
structural theorizingas "metaphysical."33 In contrastto empiricism, scien-
tificrealismcan,inprinciple,call an ontology "scientific"evenifitincludes
unobservable generative structures.Whilescientific realismdoes notman-
datesuchan ontology forsociallife(or,forthatmatter, anyothersolutionto
theagent-structure problem),itis a necessarycondition fortheontology of
structurationtheory.

a. Scientific
realism
The philosophy ofsciencecommunity is currently
inthemidstofa wide-
ranging debatebetweenempiricistsand scientific
realistsaboutwhatmight
be calledthe"theoryofscience."34At issuein thedebatearefundamental
questionsof ontology,epistemology, and the rationaljustification
of re-
searchpracticesinboththenaturalandsocialsciences.Ratherthanattempt
to reviewthe entiredebate,I will concentrateon contrasting the "hard
33. Thiskindof dismissalis an old individualistmove;see, forexample,May Brodbeck's
juxtaposition of methodological individualism with"metaphysical" holismin her"Method-
ologicalIndividualisms: Definition and Reduction,"in O'Neill,Modes ofIndividualism and
Collectivism, pp. 289-90.Morerecently, "analyticalMarxists"haveresurrected thisargument
tomotivate a reconstructionofMarxisttheory on "micro-foundations"; see JonElster,Making
SenseofMarx(Cambridge: Cambridge University Press,1985),pp.3-8. In thislatter
context,it
is perhapsworthnotingthata numberof social scientific realistshave arguedthatMarxist
theory is bestunderstood inrealist,ratherthanempiricist,termsandtherefore doesnotneedto
be reconstructed on microfoundations to be "scientific";see RussellKeat and JohnUrry,
Social Theoryas Science(London:Routledge& KeganPaul, 1982),pp. 96-118,and James
Farr,"Marx's Laws," PoliticalStudies34 (June1986),pp. 202-22.
34. The terms"empiricist"and "scientific realist"are thelabelstheparticipants in this
debate,mostofwhomarephilosophers ofnaturalscience,use todescribethemselves. Someof
the important contributions and overviewsare HilaryPutnam,Mathematics, Matter,and
Method(Cambridge:CambridgeUniversity Press, 1975);Bas van Fraassen,The Scientific
Image (Oxford:ClarendonPress, 1980);Ian Hacking,Representing and Intervening (Cam-
bridge:Cambridge University Press,1983);RichardBoyd,"On theCurrent StatusoftheIssue
ofScientificRealism,"Erkenntnis 19(May1983),pp.45-90;Jerrold Aronson, A RealistPhilos-
ophyofScience(New York:St. Martins,1984);Jarrett Leplin,ed., ScientificRealism(Berke-
ley: University of California Press, 1984);WesleySalmon,Scientific Explanationand the
Causal Structure oftheWorld(Princeton: Princeton UniversityPress,1984);andPaulChurch-
land and Clifford Hooker,eds., Images of Science: Essays on Realismand Empiricism
(Chicago:ChicagoUniversity Press,1985).
 Agent-structure
problem 351

core" empiricist and realistpositionson twoissuesthatare relevantto the

agent-structure problemand to structuration theoryin particular:1) the
legitimacyof ascribingontologicalstatus to unobservableentitieslike
generative structures,and2) thenatureofcausalclaimsandscientific expla-
nation.If the realistpositionson these issues seem upon readingto be
unexceptionable, thatis becausetheyare: oneoftheprincipal arguments for
scientificrealismis thatitclaimsto makebettersensethanempiricism ofthe
actualresearchpracticesofnaturaland,to a lesserextent,socialscientists.
In otherwords,realistsassumethatscientists, notphilosophers, arethefinal
arbitersofwhatis "scientific."Thiscontrasts withtheempiricist position,
whichis quiteexplicitly an artificial
reconstructionofwhatscientists are or
shouldbe doing.Indeed, it could be arguedthatneorealistsand world-
systemtheorists are,at leastin somerespects,"closet" scientificrealists.35
The explicitmetatheoretical statements ofbothremainwithinan empiricist
discourse,36 however,and thus theirresearchpracticedoes not follow
through on themethodological implicationsof the scientific
realistmodel.
Thisfailuresuggestsan ironictwiston theold behavioralargument thatthe
socialsciencesare "immature"becausetheyare not"scientific" enough:a
realistmightarguethat,farfrombeingpartof thesolution,theempiricist
conception ofnaturalscienceuponwhichmainstream socialscienceis based
is partofthecause of itstheoretical impoverishment.
The first
axis ofdebatebetweenempiricists andrealistsis theontological
statusofunobservables. Empiriciststendto "equatetherealwiththeexpe-
rientiallyknowable"in thesensethattheyare unwilling to saythatentities
existifwe cannot,at least in principle, have directsensoryexperienceof
them.Theyarguethatwe shouldremain,at most,agnosticabouttheexis-
tenceofunobservable entitieslikequarks,utilities,
orgenerative structures,
and thatwe shouldinsteadinterpret thetheoreticaltermsdescribing such
entities,and thetheoriesin whichthosetermsare embedded,"instrumen-
tally"ratherthan"realistically."37 Theoriesandtheoretical termsareuseful

35. Neorealistsmightbe seenas scientific realiststo theextentthattheybelievethatstate

interests
orutilitiesarerealbutunobservable mechanisms whichgenerate statebehavior,while
world-system theoristswouldbe realiststo theextentthattheybelievethatthestructure ofthe
world-system is a realbutunobservable entitywhichgenerates agents.
36. Themostexplicitrecentdiscussionofthephilosophy ofscienceunderlying neorealismof
whichI am awareis thesymposium aroundBruceBuenode Mesquita's"Towarda Scientific
Understanding ofInternationalConflict:
A PersonalView,"International StudiesQuarterly29
(June1985),pp. 121-36.Buenode Mesquita'semphasis ondeductive analysisandlogicalproof,
ratherthantheidentificationofpotentiallyunobservable causalmechanisms, as thefoundation
of scientific
explanation displaysa clearlyempiricistepistemologicalorientation.The explicit
statementson philosophy ofsciencebyat leastthequantitative schoolofworld-system theo-
ristsshowa similarrelianceon empiricist arguments; see, forexample,Christopher Chase-
Dunn,"The KerneloftheCapitalist World-Economy: ThreeApproaches,"inThompson, ed.,
ContendingApproaches, pp. 55-78.
37. The bestrecentdefenseofinstrumentalism moregenerally
andempiricism is vanFraas-
sen, The ScientificImage.
 Organization
352 International

instruments fororganizing andpredicting experience, butthereis too much

epistemic riskofpositing falseentitiestojustifyan "abductiveinference,"38
an inferencethattheoretical termsrefertorealbutunobservable entitiesand
processes.Empiricists, then,in effectsubordinate ontologyto epistemol-
ogy-what existsis a function of whatcan be knownexperientially.39 In
contrastto empiricists' rejectionof abductiveinference, scientific realists
arguethatsuchinferences are,inprinciple, justifiediftheentity in question
can produceobservableeffects,40or ifitsmanipulation permits us to inter-
venewitheffectin theobservableworld.41 Thus,thefactthatwe can use
theoriesaboutthe (unobservable)internalstructure of atomsto buildnu-
clearweaponswhichcan destroycitiesis a good reasonfortherealistto
believethatsuchstructures exist,as we understand themtoday.Thisthesis
is importantto structuration theorybecause,incontrast to empiricism, talk
ofunobservable and irreducible socialstructures can be scientificallylegiti-
mateinthisview.As longas theyhaveobservableeffects oraremanipulable
byhumanagents,we can, in principle, speakmeaningfully aboutthe"real-
ity" of unobservablesocial structures. "Generativestructure," in other
words,is a (potentially)scientificratherthanmetaphysical concept.
realistscommonly
Scientific adducetwobasic arguments in favorofab-
ductiveinference and,moregenerally, oftheontological statusofunobserv-
ables.42 Proponents of the"indispensability argument" arguethata realist
construalof theoretical termsis necessaryto makesenseof theactualre-
searchpracticesofnaturalandsocialscientists.4Physicists wouldnotposit
andbuildtestsaroundquarks,andsocialscientists wouldnotpositandbuild
testsaroundutilitiesor modes of production, if theythoughtthatthese
entities,despitebeingunobservable, werenotrealand causallyefficacious.
Proponentsof the "miracleargument,"in turn,go one step further by
arguing thatnotonlyis scientific realismnecessaryto makescientific prac-
ticesrationally butitis also necessaryto explaintheinstrumen-
intelligible,

Usefuldiscussions
is also knownas "retroduction."
38. Abduction arefoundin
ofabduction
NorwoodHanson,"Retroduction Discovery,"inLeonardKrimer-
andtheLogicofScientific
man, ed., The Nature and Scope of Social Science (New York: Appleton-Century-Crofts,
1969),pp. 73-83,andBoyd,"On theCurrent Realism,"especiallypp. 72-
StatusofScientific
89. Anunusually ofabductive
detailedandexplicitillustration reasoninginthesocialsciences
(andthussupporting myearliersuggestion thatsomesocialscientists scientific
are practicing
realists)is foundinElinorOstrom's"An AgendafortheStudyofInstitutions," PublicChoice
48 (no. 1, 1986),p. 19.
39. Aronson,A Realist Philosophyof Science, p. 261.
40. Bhaskar, The Possibilityof Naturalism,p. 16.
41. Hacking,Representing and Intervening; ThomasCook and Donald Campbell,"The
CausalAssumptions ofQuasi-Experimental 68 (July1986),especiallypp.
Practice,"Synthese
169-72.
42. AlisonWylie,"Arguments forScientific Realism:The AscendingSpiral,"American
Philosophical Quarterly23 (July1986), pp. 287-97.
43. Bhaskar,ThePossibility of Naturalism, p. 22. GeoffreyHellman,"RealistPrinciples,"
Philosophyof Science 50 (June 1983), especially pp. 231-32.
 Agent-structure
problem 353

talsuccessofscienceincontrolling theworld.44 Ifmaturescientific theories

did not at least partiallycorrespondto the deep structure of reality,the
successof sciencewouldbe an unexplainable "miracle."Certainly bothof
thesearguments dependon "maturetheories"fortheirforce,and thus
mightbe less convincing in the social thannaturalscientific context(al-
thoughmicro-economic theoryand Marxisttheorymightbe candidatesfor
sucha status).But therelativeweaknessofextantsocialscientific theories
does notjeopardizethescientific realist'sbasicpoint:itis a well-established
and perfectly legitimate practiceto positunobservable
scientific entitiesto
accountforobservablebehavior.Acceptanceofthispracticedoes notimply
thatanypositis a good one; scientists muststilladducedirector indirect
evidenceforthe validityof theirontologicalclaims,and thisevidenceis
alwaysrevisable.But by the same token,scientists, not philosophers of
science,are thearbitersofthatevidence.
The differences betweenempiricists and scientific realistsoverontology
fueldebate on a second axis, the natureand requirements of scientific
explanation. Traditionally, therehave been twocompeting idealsof scien-
tificexplanation:theempiricist or "nomothetic"viewthatexplanation in-
volvesthesubsumption ofa phenomenon undera lawlikeregularity; andthe
realistor "retroductive" viewthatitinvolvestheidentification oftheunder-
lyingcausalmechanisms whichphysically generated thephenomenon.45 The
resurgence in recentyearsof the scientific realistviewcoincideswithre-
newedattackson theHumeanmodelofcausation,whichsupportstheem-
piricistaccountofexplanation.46 In theHumeanmodel,a causalrelationis a
''constantconjunction"of temporallysequencedobservedevents that
standsin a relationof logicalnecessityto certaininitialconditionsand
laws.47On theHumeanview,ourinability to experience causalmechanisms
directly prevents us fromimputing anynaturalnecessity to causalrelations.
Scientific realistscriticizethismodelbecause constantconjunctions and
generalizations are not theirown explanation,and argueinsteadthatto
44. See, forexample,Putnam,Matter,Mathematics,
and Method;Boyd,"On theCurrent
StatusoftheIssue ofScientific
Realism";RichardSchlagel,"A ReasonableReplyto Hume's
Skepticism,"BritishJournalforthe Philosophyof Science 35 (December 1984), pp. 359-74.
45. See ErnanMcMullin,"Two IdealsofExplanation inNaturalScience,"inPeterFrench,
et al., eds., Causationand Causal Theories(Minneapolis:University
of MinnesotaPress,
1984),pp. 205-20,and thethree-way debatebetweenPhilipKitcher,Bas van Fraassen,and
WesleySalmonin "Approachesto Explanation,"TheJournalofPhilosophy 82 (November
1985), pp. 632-54.
46. RomHarreandEdwardMadden,CausalPowers(Totowa,N.J.:RowmanandLittlefield,
1975); Salmon, ScientificExplanation; Schlagel, "Hume's Skepticism."
47. Hencebehavioralsocialscientists'
emphasisonquantitative analysistodiscoverlaw-like
regularities,
rather
thanqualitative analysisandtheory to identify
causalmechanisms. On the
model,we cannothavesciencewithout
empiricist (relatively)
"constant"conjunctions.Fora
usefulmoreorless realistcritiqueofthismodelofcausationas itrelatesto socialscience,see
DanielHausman,"Are ThereCausal RelationsamongDependentVariables?"Philosophy of
Science50 (March1983),pp. 58-81.
354 International
Organization

makea trueexplanatory claim,it is necessaryto identify the underlying

causalmechanisms whichmakean eventnaturally necessary.The disagree-
menthereultimately hingeson thelegitimacy ofabductiveinference. Real-
istsarguethatifwe can explainthephysicaldispositions andcausalpowers
of unobservableentities,we can make a legitimate abductiveinference
abouttheexistenceof naturally necessaryrelationsbetweencause and ef-
fect,andthereby transcend Hume'sskepticism aboutcausation.48 In effect,
therealist'sfocuson causalmechanisms is an attempt to explaintheempiri-
cist'sconstantconjunctions. Thesedifferent modelsofcausation,then,gen-
erateverydifferent modelsofscientific explanation. Whereastheempiricist
explainsby generalizing aboutobservablebehavior,therealistexplainsby
showinghow the(oftenunobservable) causal mechanisms whichmakeob-
servableregularities possiblework.The scientific realist,in otherwords,
arguesthat"answersto why-questions requireanswersto how-and what-
questions."49
I wantto concludethisbriefoverviewoftheempiricist/realist debatewith
threepointsaboutitsimplications forstructuration theoryand,moregener-
ally,forsocial scientific
researchpractice.First,scientific realismattempts
to make sense of whatpracticing naturaland social scientistsin factdo,
rather thanprescribingon thelegitimacy ofcertainresearchpracticesversus
others.Abductionand the de factopositingof unobservablecausal pro-
cesses and entities,
whether thoseare quarks,utilities, or modesofproduc-
tion,go on constantly in scientificresearch,and scientific realistssee no
reasonto writethispracticeoffas unscientific. For thisreason,and thisis
the second point,in contrastto empiricism, scientificrealismcan make
scientific
senseofunobservable generative structures,ofstructures thatare
irreducibleto and generate their a
elements.Thisoutflanks keymotivation
forindividualism, namelythatstructural theorizing ofthegenerative variety
is necessarily"metaphysical" or "unscientific." Finally,although thereare
important problemsin translating the protocolsand discourseof natural
scientific
practicedirectly to the social sciences-whatRoy Bhaskarcalls
"limitsto naturalism" 50-the basic realistidea thatscientific explanation
48. Somerealistaccountsofcausation,andparticularlytheaccountofHarreandMadden,
havebeenaccusedofimplying -the explanation
"essentialism"
an Aristotelian ofobservable
phenomena in termsofoccultand impenetrable "essences"; see, forexample,David Miller,
"Back to Aristotle,"BritishJournalforthePhilosophyof Science 23 (February1972), pp. 69-
78, andFredWilson,"HarreandMaddenon Analyzing Concepts,"Philosophy
Dispositional
of Science 52 (December1985),pp. 591-607.Otherrealists,however,emphasizethatthis
objectioncan be vitiatedby explainingcausal powersin termsofthephysicalproperties and
socialrelationswhichunderliethem;Schlagel,"Hume's Skepticism."
49. RussellKeat and JohnUrry,Social Theoryas Science(London:Routledge& Kegan
Paul, 1982),p. 31.
50. Perhapsthemostdifficult problems inmaking concerntheroleofhuman
thistranslation
motivations in socialscientific
and self-understandings andtheambiguity
explanations, ofthe
notionofcausal"mechanisms" insociallife.Fora sampleoftherecentdebateamongscientific
realistson the limitsof naturalism of
in the social sciences,see Bhaskar,The Possibility
Naturalism, and Keat and Urry,Social Theoryas Science, especially the postscript.
 problem 355
Agent-structure

consistsintheidentification
ofunderlyingcausalmechanisms ratherthanin
generalizations does applyto the social sci-
aboutobservableregularities
ences, and its adoptiontherewould have importantimplicationsforthe
explanationof social action.Moreover,and perhapsmoreprovocatively,
practicemustbe "critical"to be
thisidea suggeststhatsocial scientific
Therestofthispaperbeginstoconsiderthenatureandimplica-
"scientific."
perspective
structurationist
tionsof thispointwithina specifically on the
agent-structureproblem.

b. Structurationtheory
realismprovidesa philosophical
Scientific basisfora generative approach
to structuraltheorizingin thesocialsciences,andin so doing,itprovidesa
foundation forworkingout theimplications of one of theintuitions about
sociallifewithwhichI openedthediscussionof theagent-structure prob-
lem:thatthecapacitiesandevenexistenceofhumanagentsareinsomeway
necessarilyrelatedto a social structuralcontext-thattheyare inseparable
fromhumansociality.The implications of thisinsight,however,can be
workedout in at least two different ways-one reifiesthe social relations
thatconstitute agents,and one does not.I arguedearlierthatworld-system
theory embodiesa structuralist approachtotheagent-structure problemthat
is proneto reificationand determinism. Structuration theoryattemptsto
preservethegenerative and relationalaspectsofstructuralism whiletaking
explicitconceptualand methodological stepsto preventtheanalyticalsep-
arationof generative structures fromtheself-understandings and practices
ofhumanagentsto preventstructural reification.
It maybe usefulto prefacethediscussionwithsomecomments on what
structurationtheory,as a theory, is about.Structuration theoryis an "ana-
lytical"ratherthan"substantive"theory,in thesensethatit is aboutthe
analysisrather thanthesubstanceofthesocialworld.51 Structuration theory
sayssomething aboutwhatkindsofentities thereareinthesocialworldand
howtheirrelationship shouldbe conceptualized, and as suchit providesa
conceptualframework or meta-theory forthinking aboutreal worldsocial
systems,but it does not tell us whatparticular kindsof agentsor what
particularkindsofstructures to expectinanygivenconcretesocialsystem.
Structuration theory,then,does notcompetedirectly withneorealismor
world-system theory,but insteadwiththeirindividualist and structuralist
approachesto the agent-structure problem-thatis, withtheirsocial on-
tologies.As a social ontology,however,structuration theorydoes have
in "The Statusof Structuration
distinction
51. Ira Cohenmakesthisparticular Theory:A
Replyto McLennan,"Theory,Culture,and Society3 (no. 1, 1986),pp. 123-34.NigelThrift
theoryis moremeta-theory
makesa similarpoint,arguingthatstructuration thantheoryin
"Bear and Mouse or Bear and Tree? AnthonyGiddens'Reconstitution of Social Theory,"
Sociology 19 (November 1985), pp. 609-23.
356 International
Organization

implications forthe potentialcontentof substantive theoriesabout real-

worldsocial systems,and forthemethodology thatsocialscientists should
use fo studythose systems.These implications in turndefinea research
agendaforsocial inquiry.I shallsuggestin Section4 how neorealismand
world-system theoryrelateto thisresearchagenda,and indicatesome of
thegapsin contemporary internationalrelationsresearchthatemergefrom
thiscomparison.
In keepingwithstructurationtheory'scharacter as a conceptualapproach
ratherthana substantively-definedtheory, thegroup of"structuration theo-
rists"is also quitediverseand, indeed,mayonly recognizable such
be as
fromoutsidethestructurationist problematic. Thrift, forexample,identifies
fivemajorsocial theorists (PhilipAbrams,Roy Bhaskar,PierreBourdieu,
Anthony Giddens,andDerekLayder)as structurationists, eventhough most
ofthesewouldprobablyresistthestructurationist label(perhapsbecauseit
is Giddens's).52 Despitetheirinternal
differences, however,theyall share
fourbasicanalytical objectivesthatcan be seenas defining the"hardcore"
ofthestructuration researchprogram.53
theyaccepttherealityand explana-
1) In oppositionto individualists,
toryimportance ofirreducible unobservable
and potentially social
thatgenerateagents.
structures
and stress
theyopposefunctionalism
2) In oppositionto structuralists,
"the needfora theoryofpracticalreasonand consciousnessthat
can accountforhumanintentionalityand motivation."54
3) These oppositionsare reconciledbyjoiningagentsand structuresin
a "dialecticalsynthesis"thatovercomesthe subordination of one
to theother,whichis characteristicofbothindividualismand struc-
turalism.55
4) Finally,theyarguethatsocial structures fromspatial
are inseparable
and temporalstructures,and thattimeand space musttherefore be
incorporated and explicitly
directly and concreteso-
intotheoretical
cial research.56

52. Thrift,"On theDetermination ofSocial Actionin Space andTime,"p. 30.

53. Adaptedfromibid.,pp. 28-32.
54. Ibid.,p. 30.
55. This synthesisrequiresthedevelopment of mediating conceptsthatcan linkstructure
and agencyin concretesituations, and as suchis probablythekeysourceof disagreement
amongstructuration But whether
theorists. thislinkageis established through a "position-
practicesystem"(Bhaskar),a "habitus"(Bourdieu),or a "system-institution" nexus(Gid-
dens),theyall servethesametheoretical function inconcreteresearch, namely, bindingagents
and structuresintomutually implicatingontological andexplanatory roles.
56. This pointis morethana ritualadmonition to be sensitiveto the
forsocial scientists
andgeographical
historical contextoftheirsubjects:substantive mustbe about
"socialtheories
thetime-space ofsocialstructures
constitution rightfromthestart."(Thrift, "On theDetermi-
nationofSocial Action,"p. 31, italicsin original.)
 Agent-structure
problem 357

The following discussionelaboratesthesepointsby firstdiscussingthena-

tureofsocialstructures, thenofagents,andfinally oftheirinterrelationship.
My accountreliesprimarily on Bhaskar'swork,whichofthefivetheorists
displaysthemostexplicitly scientific
realistorientation.S7
Structuration theoristsstartout much like structuralists by defining
"structure" ingenerative termsas a setofinternally relatedelements.58 The
elementsofa social structure couldbe agents,practices,technologies, ter-
ritories-whatever can be seen as occupyinga positionwithina social or-
ganization.The factthattheseelementsare internally relatedmeansthat
theycannotbe defined or evenconceivedindependently oftheirpositionin
thestructure. Thus,in contrastto theneorealistdefinition of international
systemstructures as consistingof externallyrelated,preexisting, state
agents,a structurationist approachto thestatesystemwouldsee statesin
relational
termsas generated orconstituted byinternal relations ofindividu-
ation(sovereignty) and,perhaps,penetration (spheresofinfluence). In other
words,statesarenotevenconceivableas statesapartfromtheirpositionina
global structure of individuated and penetrated politicalauthorities.The
natureand configuration of the internalrelationsthatcomprisea social
structure,inturn,definea setofpossibletransformations orcombinations of
itselements.As a set ofpossibletransformations, socialstructures are,by
definition,notreducibleto therelationships betweena structure's elements
thatare observedin a givenconcretecontext.Structures make a given
combination or instantiation of elementspossible,but theyare not ex-
haustedby whateverparticular manifestation is actual.
Structuration theorists arguethescientific realistthesisthatbecause so-
cial structures generateagentsand theirbehavior(in the sense thatthey
makethe latterpossible),thatbecause social structures have observable
effects,we can potentially claimthattheyare real entitiesdespitebeing
possiblyunobservable.This thesisraisestheissue of whenwe can legiti-
matelyclaimthata social structure exists.The keyweaknessofabductive
inferenceis thedangerofcircularreasoning andself-confirmation; we assert
thata structure existsbecauseithas theobservedeffects whichwe posited
forthestructure in thefirstplace. Thisweaknessis, I think,at theheartof
57. In his Profilesand Critiquesin Social Theory(Berkeley: Universityof CaliforniaPress,
1982),Giddensindicates(p.14) thathe also acceptsa realistconception of science,buthis
realismis generally
less explicitandthusmoreattenuated thanBhaskar's.A moreimportant
reasonforrelyingon Bhaskarrather thanGiddens,however, is thelatter'sweakerconception
ofsocialstructureas rulesandresourcesratherthanas a setofrealbutunobservable internal
relations,a conceptionwhichis arguablyultimately voluntaristin its implications;
see for
example,AlexCallinicos,"Anthony Giddens:A Contemporary Critique,"Theory andSociety
14 (March1985),pp. 133-66.
58. See, for example, Bhaskar, The Possibilityof Naturalism,especially pp. 47-56; Peter
Manicas, "The Concept of Social Structure,"Journalfor the Theoryof Social Behavior 10
(July1980),pp. 65-82; Keat and Urry,Social Theoryas Science,p. 121; AndrewSayer,
Method in Social Science: A Realist Approach (London: Hutchinson,1984), pp. 80-87.
358 International
Organization

thefrequently heardcomplaintof mainstream social scientists, who claim

thatMarxistand otherstructural theoriesin socialscienceare,inprinciple,
"non-falsifiable." The appropriate responseto thisproblemis to findevi-
dencefora structure's existenceor workings thatare independent of the-
particularobservations fromwhichthestructure was abduced,59 andtorec-
ognizeand attempt to controlfortheradicalopennessof social systems.60
This is, of course,whatnaturalscientific practiceis all about: scientists
workbackwardfroman observedphenomenon to a postulatedentityor
causal mechanism, and thentheytryto developtestsin closed systemsto
determine independently whetherthe inference is justified.Clearlysuch
independent evidenceis hardertoobtaininthesocialsciences,butthisdoes
notinvalidatethe basic idea of usingunobservablestructures and causal
mechanisms to accountforsocialbehavior.Indeed,in viewoftheindivid-
ualist'srejectionofgenerative structuresas "metaphysical," itis ironicthat
thegenerative approachis muchclosertotheconceptualization ofstructure
inmathematics andthenaturalsciencesthanis theindividualist's definition
as a distributionofunit-level properties.61
Up tothispoint,thestructurationist understanding ofstructure is identical
to thestructuralist's.Each conceivesofstructure in combinatorial termsas
an irreducibleentitythat"generates"itselementsandtheirpossibletrans-
formations. Structuration theoristsdivergefromstructuralists, however,in
arguing thatsocialstructures inatleasttwofundamental
differ respectsfrom
naturalstructures, and thata recognition ofthesedifferences is essentialto
avoidthereification ofsocialstructures characteristicofstructuralism. The
firstdifferenceis that"social structures, unlikenaturalstructures, do not
existindependently of the activitiestheygovern."62Whileit may make
senseto saythata naturalstructure has an existenceapartfromthebehavior
59. Keat and Urry,Social Theory as Science,postscript.
60. Anopensystemis oneinwhichinvariant constant conjunctions do notobtain.Although
thecomplexity andopen-endedness ofopensystems limitthepossibilitiesfordecisivetestsof
socialscientificclaims(see Bhaskar,ThePossibility ofNaturalism, pp. 164-65),thisproblem
afflictsnotonlythosetheorieswhichreferto unobservable entities.For an interesting
and
explicitlyrealistargument abouthowopensystems might, in somecases,be studiedina way
thatwouldpermitrelatively controlled tests,see Cook and Campbell,"Quasi-Experimental
Practice."
61. On thedefinition of structure in mathematics see, forexample,MarcBarbut,"On the
MeaningoftheWord'Structure' in Mathematics," A Reader
in M. Lane, ed., Structuralism:
(London:Jonathan Cape, 1970),MichaelResnick,"Mathematics as a Scienceof Patterns:
Ontologyand Reference,"Nous 15 (November1981),pp. 529-50,and StewartShapiro,
"Mathematics andReality,"Philosophy ofScience50 (December1983),pp. 523-48.Modern
physics, inturn,is basedongrouptheory (themathematical theory ofbinary systems),whichis
explicitlycombinatorial andpossibilistic initsviewofstructure. I shouldprobably note,how-
ever,thatalthough I emphasizethissimilarity in socialand naturalscientificconceptions of
structure,I amnotsayingthatsocialscienceshouldbe socialphysics.I amonlytrying tojustify
a certainkindofthinking andexplanation in socialsciencebypointing outthatitpervadesthe
naturalsciencesas well.
62. Bhaskar,ThePossibility ofNaturalism, pp. 48-49; on thedifferences betweennatural
andsocialstructures, see also Giddens,StudiesinSocialandPoliticalTheory (London:Hutch-
inson,1977),pp. 118-19.
 Agent-structure
problem 359

of its elements,social structures are onlyinstantiated by thepracticesof

agents.The deep structure ofthestatesystem,forexample,existsonlyin
virtueof the recognition of certainrulesand the performance of certain
practicesby states;if statesceased suchrecognition of performances, the
statesystemas presently constituted wouldautomatically disappear.Social
structures, then,are ontologically dependentupon (althoughtheyare not
reducibleto) theirelementsin a way thatnaturalstructures are not. The
seconddifference is that"social structures,unlikenaturalstructures, do not
existindependently oftheagents'conceptions ofwhattheyaredoingintheir
activity."63 In otherwords,social structures have an inherently discursive
dimension in thesensethattheyare inseparablefromthereasonsand self-
understandings thatagentsbringto theiractions.This discursivequality
does notmeanthatsocialstructures arereducibleto whatagentsthinkthey
are doing,since agentsmaynotunderstand the structuralantecedents or
implications oftheiractions.Butitdoes meanthattheexistenceandopera-
tionof social structures are dependent uponhumanself-understandings; it
also meansthatsocial structures acquiretheircausal efficacy onlythrough
themediumofpracticalconsciousnessand action.
Justas social structures are ontologicallydependent uponand therefore
constituted by the practicesand self-understandings of agents,the causal
powersand interests ofthoseagents,in theirownturn,are constituted and
therefore explainedby structures. The structures
thatconstitute agentsare
oftwodistinct kinds:external, or social,structures;
andinternal, or organi-
zational,structures. Each typeexplainsa distinctset ofthecausal powers
and interestsof agents-social and intrinsic ones, respectively.Thus, all
agentspossessthreeintrinsic capacitiesor powersinvirtueoftheirinternal
organizational structure or "anatomy":641) to have a theoretical under-
standing(howeverinaccurate)of its activities,in the sense thatit could
supplyreasonsforits behavior;2) to reflexively monitorand potentially
adaptits behavior;and 3) to makedecisions.These causal powersdiffer-
entiateagentsfromthe non-sapient elementsthatcomprisenaturalstruc-
tures,and to theextentthatstatescan be consideredgoal-directed unitsof
action,theycan be consideredagentsby thisdefinition. Internalorganiza-
tionalstructures arealso important, however,forexplaining thesubjectively
perceivedinterests ofagents.Individualandorganizational decision-making
pathologiesin the state,forexample,maybe crucialfordetermining how
social structural or objectiveimperatives forcompetentstatepractice-
a state's "real interests"-translateinto subjectiveinterestsand actual
performance.65

63. Bhaskar, The Possibilityof Naturalism,pp. 48-49.

64. Adapted fromGiddens, The Constitutionof Society, pp. 5-6.
65. Structurationtheoristshave yetto tacklein a sustainedway the natureand role of
in socialscientific
interests Although
explanations. someofthemorematerialisticallyinclined
theorists
structuration mightrejecttheexplanatory I aminclined
altogether,
use ofinterests to
360 International
Organization

The importance of theirinternal organizational structures notwithstand-

ing,however,othercausalpowersandtherealinterests ofagentsaredepen-
dentupon and thusexplainableonlyby the externalor social structural
contextin whichtheyare embedded.For example,the causal powersof
capitalists(forexample,to investanddisinvestcapital,to extracta surplus
fromlabor)and theirrealinterests (to maximizeprofits, and so forth)are a
function of the organizing principlesof the capitalistmodeof production
whichdefinetheirpositionand interestsas a capitalist's.Similarly, the
causal powersof the state-to maintaincontrolover the resourcesand
violencepotential ina giventerritory, to actinan international environment
freeoflegalcompulsion, and so on-are conferred uponitbythedomestic
and international social structures in virtueofwhichitis a statein thefirst
place. These structural relationsmaybe as generalas theorganizing princi-
ples oftheinterstate system(forexample,sovereignty, penetration) which
constitutestatesas such,or theymayconsistofthemorelocalizedorganiz-
ingprinciples ofconcreteinternational systems,likethebalanceofpower,
whichdefineparticular kindsofstates.Thus,the"balancer"ina balance-of-
powersystem,or a core statein thecapitalistworldeconomy,has certain
powers,responsibilities, andinterests whichitpossessesonlyinvirtueofits
socialstructuralposition.66 Social structures, then,constitute theconditions
of existenceof statesand stateaction;indeed,withoutsocial structuring
principlesone could nottalkmeaningfully aboutthefundamental building
blocksof international relations:"states," "state powers," "foreignpol-
icy," andso forth. Putinanother way,international anddomesticstructures
generatethe "rules of thegame" (broadlydefinedto includestateagents
themselves) withinwhichstatesinteract.
Structuration theory, then,conceptualizes agentsand structures as mutu-
allyconstitutive yetontologically distinct entities.Each is in somesensean
effectof the other;theyare "co-determined." Social structures are the
resultoftheintended andunintended consequencesofhumanaction,justas
thoseactionspresupposeor are mediatedby an irreducible con-
structural
text.Thisunderstanding oftheagent-structure relationship is madepossible
by conceptualizing each fromthestartas ontologically dependent uponthe
other,byconceptualizing agentsintermsoftheinternal relations thatdefine
themas such, and by conceptualizing social structures as existingonly
through themediumoftheagentsandDracticesthattheyconstitute. Thisis

thinkthattheiragent-structure framework presupposesat leastan implicit between

distinction
"subjective"and "real" interests. The best overviewof the variousconceptualizations of
"interest,"and ofthedifficultiesin explaining is probablystillWilliamConnolly's
interests,
"Interestsin Politics,"in his book, The Termsof PoliticalDiscourse(Princeton: Princeton
University Press,1974),pp. 45-84.
66. For a discussionofthebalanceofpowerthatis consistent in itssubstance,ifnotin its
philosophical withtheinterpretation
rationale, see Ashley,"The Poverty
I suggest, ofNeoreal-
ism,"pp. 276-79.
 problem 361
Agent-structure

whatGiddensmeansby the "dualityiof structure," that"the structural

properties ofsocial systemsare boththemediumand outcomeoftheprac-
ticeswhichconstitute thosesystems." 67 Structuration theoryis therefore
morethanan attempt to introduce a greater balanceofstructure andagency
in socialtheorythanis foundin individualism and structuralism.Its social
ontologyradicallyreconceptualizes the fundamental properties of agents
and social structuresin sucha wayto makethemontologically interdepen-
dent,and it is onlyvirtueof thisreconceptualization thatthe "errors"of
reductionand reification characteristic of individualism and structuralism
are avoided. As I shall indicatein the nextsection,the ontologicaland
conceptualinterdependence ofagentsand structures in structurationtheory
has important implications forthe explanationof social action.Put very
generally, it forcesus to see agentsand structures as simultaneously in-
volvedin theproduction of socialphenomena.In Bhaskar'swords:
Thussocietyis nottheunconditioned creationofhumanagency(volun-
tarism),butneitherdoes it existindependentlyofit (reification).
And
individual
actionneithercompletely determines (individualism)
noris
completelydetermined by (determinism) socialforms.68

4. Implicationsforinternationalrelationstheory

The discussionofstructuration theoryso farhas focusedon itssocialontol-

ogy,on its conceptualization of the natureand relationship of humanor
organizationalagentsand socialstructures. Whilestructurationtheory does
notby itselfgenerateclaimsor hypothesesaboutparticular international
systemstructures or thecauses of stateaction,therealist/structurationist
problematic does havebothepistemological andtheoretical for
implications
thestudyofinternational relations.
Thus,on theonehand,structuration the-
ory'ssocialontologystrongly conditions itsapproachto theexplanation of
stateaction.This idea is consistentwiththeeffort realiststo
of scientific
reversethesubordination ofontology whichis characteris-
to epistemology,
tic of empiricism,and insteadto maketheformof scientific explanations
dependent on thenatureand causal properties ofentities.Beyondthisgen-
eralconcernwiththeformof explanations, however,structurationtheory
also has implicationsforthecontentof substantive internationalrelations
theoriesor,perhapsmoreprecisely, forthenatureandscopeoftheresearch
agendaswhichunderliethosetheories.In particular, structurationtheory
suggeststhat,whileneorealism andworld-system theoryprovideimportant
insightsintothestructureanddynamics ofinternationalsystems,theyleave

67. Giddens, CentralProblems in Social Theory,p. 69.

68. Roy Bhaskar,"Emergence,Explanation, in Paul Secord,ed., Ex-
and Emancipation,"
plainingHumanBehavior(BeverlyHills:Sage Publications,
1982),p. 286.
362 International
Organization

importantgapsinthetheorization ofthetwobasicbuilding
blocksofinterna-
systemstructures.
tionalrelationstheory,statesand international

a. Epistemological implications
Relativelylittleempiricalresearchhas beenexplicitly informed by struc-
turationtheory, whichmight illustrate fortheexplanation
itsimplications of
stateaction.69In thisarticle,I cannotdevelopan extendedempirical illustra-
ion of myown; instead,I shalladaptmaterials fromthefewstructuration
andrealistphilosophers
theorists70 ofsocialsciencewhohavetackledissues
ofsocialscientificexplanation inorderto maketwogeneralepistemological
arguments:1) thatstructural and agent-basedanalyseshave distinctand
irreduciblefunctions in theexplanation ofsocialaction,butthat2) theyare
bothnecessaryelementsof a completeexplanation of social action.These
two arguments have important implicationsforour understanding of the
natureandlimitsofstructural andwhatI shallcallhistoricalexplanations, as
wellas fortheirintegration in "structural-historical"analysis.
Explanations areanswersto certainkindsofquestions.Whatcountsas an
adequateexplanationtherefore dependson theobjectof thequestion,on
whatis takento be problematic.71 Froma structurationistperspective, two
kindsof questionsare particularly relevantto theexplanation of socialac-
tion:"How is actionX possible?"and "WhydidX happenrather thanY?"
The domainsofthesetwoquestions,andtherefore thekindsofanswerswe
wouldexpect,are different. "How-questions"are concernedwiththedo-
mainof the possible,whereas"why-questions" are concernedwiththe
domainoftheactual.To remainclearon thenatureand limitsofstructural
explanation,an explicitepistemological andmethodological distinction must
be maintained betweenthelogicof thesequestions:"structural"analysis
explainsthepossible,while"historical"analysisexplainstheactual.His-
toricalanalysisfocuseson whatactuallyhappenedor willhappen,andthus
takesas unproblematic thepossibility thatthoseeventscan happen.Actual
behavior,ratherthantherangeofpossiblebehaviors,is theexplanandum.
Whilethishistoricalanalysisofconjunctural causesis an essentialcomponent

69. Themostextensive use ofan explicitly perspective

structurationist inempirical research
is probablyAllanPred,Place, Practice,and Structure(Cambridge, U.K.: PolityPress,1986).
In hisExplanationinSocial History,however,Lloydargues(p. 306)thattheworkofa number
of prestructuration has a distinctly
theorists "structure,"
structurationist including,forex-
ample,theworksofBarrington Moore,Social OriginsofDictatorship andDemocracy(Boston:
Beacon Press,1966),and AlainTouraine,TheSelf-Production of Society(Chicago:Chicago
University Press,1977),and Abrams,HistoricalSociology.
70. Sayer,Methodin Social Science;SylvanandGlassner,A Rationalist Methodology.
71. Theimplications
oftheepistemological betweendifferent
distinctions kindsofquestions
are brought in Alan Garfinkel,
out systematically FormsofExplanation(New Haven: Yale
University Press, 1981),especiallypp. 21-48. Despiteits explicitlyanti-realist
ontological
perspective,van Fraassen'sTheScientificImageis also quitegoodon thelogicor "pragmat-
ics" ofdifferenttypesofexplanations.
 problem 363
Agent-structure

of the explanationof stateaction,however,it is onlya necessary,nota

condition
sufficient, forscientific
explanation. Within realistepis-
a scientific
temology, why-questions requireanswersto how-questions; thatis, to ex-
plainwhya statedid X ratherY, we need to knowhowthatstateand its
choiceswerepossibleinthefirst necessaryto makethe
place. It is therefore
actualbehaviorand properties of statesand statesystems"problematic"
ratherthansimplyacceptingthemas given.It is necessary,in otherwords,
to engagein structural analysisto explainthecausal properties ofstatesin
virtueofwhichtheiractionsarepossible.The following discussionofstruc-
turalandhistorical analysisexpandson thedistinct, butnonetheless interde-
pendent,epistemological rolesofthesetwoformsofexplanation.
Structural researchstartswithactualevents,withhistory, and by a pro-
cess ofcritiqueand abduction-thatis, byaskingwhatmustexistforthose
eventsto happen-abstractsto thesocialand internal organizational struc-
tureswhichmakethose eventspossible.These structures mightthenbe
modeledformally withqualitativetechniques (mathematical orgrammatical)
whichdescribetheirpossibilisticrelationship to observableevents,72 al-
thoughsuchmodelingmaynotalwaysbe possiblein opensystems.In any
case, structural explanationscontribute to the explanationof observable
eventsby showingthattheyare instancesofthepossiblewaysofactingof
socialagents,wherethosepossibilities aredefined bythestructurally deter-
minedcausal powersand interests ofthoseagents.Putanotherway,struc-
turalexplanations revealtheconditions ofexistenceor "rulesofthegame"
of social action.In thissense structural theoryis necessarily"critical,"
sinceitforcesus to lookbeyondgivenappearancesto theunderlying social
relationships thatgenerate(in a possibilisticsense) phenomenalforms.
Whilestructural analysesare thuspartofa completeexplanation of actual
events,however,theydo notexplainthoseeventsdirectly; theyonlyan-
swerthequestionofhowtheyare possible,ofwhatcombinations or trans-
formations of a structure'selementsare consistentwithits organizing
principles.Althoughstructural analysesmay uncover"tendencies"for
structures to be actualizedin certainways,neither generalizationnorpoint
prediction is an importantaspectofstructural explanations, andanyattempt
touse themto accountdirectly fortheproduction ofparticular eventswould
riskoverextending thembeyondtheirproperexplanatory domain.73
Historicalresearch,on theotherhand,"studiesactualeventsandobjects
as 'unitiesofdiversedeterminations,' each ofwhichhavebeenisolatedand
examinedthroughabstract[structural] research."74 Historicalexplana-
72. Anexcellentintroductionto someoftheformalmethodsthatcouldbe usedingenerative
structuralanalysesis foundin chaps. 5 and 6 of Sylvanand Glassner,A Rationalist
Methodology.
73. In Methodin Social Science,Sayerargues(p. 217) thata failureto recognizethese
analysisis responsible
ofstructural
limitations orwhathe calls"pseudo-
forthedeterministic,
concrete,"qualityofmuchMarxistresearch.
74. Sayer,Methodin Social Science,p. 216.
 Organization
364 International

andcausalpowersofagentsas given(orreconstruct
tions75taketheinterests
themwithout tryingto explainthem),andthenattempt to explainparticular
eventsby focusingon how thosepowersand interests are affectedby the
incentives facingactors.Neorealismproceedsat thislevel;it stipulates the
structural contextand theinterests and causal powersof agentsand then
attempts to answerthequestion"Whydid stateX do Y ratherthanZ?"
Beyondthis,however,it is important to notethatone of theintendedor
unintended effectsof stateactionis to produceor reproducesystemstruc-
tures;consequently, historicalanalysisis necessaryto explainthe emer-
gence and persistenceof the structural conditionswhichconstitutethe
mediumandconditions ofpossibilityforstateaction.Thisrecursive quality
ofstructural andhistoricalexplanations is the"unity"beneaththeir"diver-
sity,"and thusit is theultimatebasis of theirepistemological interdepen-
dence.It is nonetheless necessaryto maintain thedistinctionbetweenand
autonomy ofeachexplanatory mode:eachultimately explainstheproperties
ofthecentralobjectsoftheother.
The respectiveexplanatory limitations and historicalanal-
of structural
ysessuggest,however,thata completeexplanation ofstateaction-thatis,
one thatexplainsbothhowthatactionwas possibleandwhythatpossibility
was actualizedina particularformat a givenmoment-willhavetocombine
thesemethodologies intoa "structural-historical"or "dialectical"analy-
sis.76This combination willrequireabstractstructural analysisto theorize
and explainthecausal powers,practices,and interests of states,and con-
cretehistoricalanalysisto tracethecausallysignificant sequenceofchoices
and interactions whichlead to particularevents(andto thereproduction of
Giventhedifficulty
social structures). ofdoingstructural and historical
re-
searchsimultaneously, analysismayrequire"brack-
structural-historical
eting"first one andthentheotherexplanatory mode,77thatis, takingsocial
75. Bymyuse oftheterm"historical"todescribethisform I do notmeanto
ofexplanation,
suggestthatthisis theexplanatorymodehistoriansalwaysuse,orthattheresearch practiceof
astructural
is necessarily
historians itseemstomethatjustas
On thecontrary,
oratheoretical.
goodhistory
goodsocialscienceis historical, is structural I am onlytrying
andtheoretical. to
arguethat"historical"and"structural"explanations distinct
areepistemologically butinterde-
pendent formsofinquiry, ofwhouses them.
regardless
is fromFernandoCardosoand Enzo Faletto,Depen-
76. The term"structural-historical"
dencyand Developmentin Latin America (Berkeley: Universityof CaliforniaPress, 1978),pp.
ix-xiv,while"dialectical"is fromSylvanandGlassner, A RationalistMethodology,pp. 154-
between"abstract"and"concrete"research
59; bothtermsparalleltherelationship inSayer's
Methodin Social Science. Although he does notuse eitherof theseterms,PeterManicas
ofthelogicandimplications
providesa goodillustration ofthisform ofinquiry of
inhiscritique
ThedaSkocpol'sStateandSocialRevolutions (Cambridge: Cambridge UniversityPress,1979);
see hisreviewinHistoryand Theory 20 (no. 2, 1981),pp. 204-18.
77. Giddens,CentralProblemsinSocial Theory, pp. 80-81.Thisnotionof"bracketing" is a
focalpointofsomeofthemajorcritiques ofstructuration theory;see, forexample,Margaret
Archer,"Morphogenesis versusStructuration:On Combining and Action,"British
Structure
Journalof Sociology33 (December1982),pp. 455-83,and NickyGregson,"On Dualityand
Dualism:The Case ofStructuration TheoryandTimeGeography," ProgressinHumanGeog-
raphy10(June1986),pp. 184-205.
 problem 365
Agent-structure

structuresand agentsin turnas temporarilygivenin orderto examinethe

explanatory effectsof theother.This methodologicaldifficulty,
however,
shouldnot obscurethe epistemological interdependenceof structural and
analysis,thefactthattherespectiveexplanatory
historical rolesof agents
cannotbe understood
and social structures apartfromtheirinterrelation-
ship.This conclusionfollowsdirectlyfromthe ontologyof structuration
theory.Agentsare inseparable inthesensethattheir
fromsocialstructures
actionis possibleonlyin virtueof thosestructures,and social structures
cannothave causal significanceexceptinsofaras theyare instantiated by
agents.Social action,then,is "co-determined"by theproperties of both
agentsand social structures.

b. Theoreticalimplications
Whilethisdiscussionof some epistemological implications of structura-
tiontheory is admittedlyverygeneral,itis nonetheless relevant tothescope
andcontent ofsubstantive internationalrelationstheories.A keyimplication
oftheargument in Section2 abouttheagent-structure relationship was that
theoriesofinternational relationsmusthavefoundations in theoriesofboth
theirprincipalunitsof analysis(stateagentsand systemstructures). Such
theoriesare morethansimplyconvenient or desirable:theyare necessary
to explainstateaction.This requirement followsdirectlybothfromthe
scientific
realist'sconceptionof explanationas identifying causal mecha-
nisms,and fromthe ontologicalclaimsof structuration theoryaboutthe
relationshipofagentsand structures. If theproperties ofstatesand system
structuresare boththought to be causallyrelevantto eventsin theinterna-
tionalsystem,and ifthoseproperties are somehowinterrelated, thentheo-
reticalunderstandings of boththoseunitsare necessaryto explainstate
action.Waltz'ssuggestion thatthetheoryofthestateis notintegral to the
taskofdeveloping systemic theories ofinternationalrelations musttherefore
be rejected.Structurationtheoryprovidesa conceptualand methodological
framework to overcomethisseparation,and as suchit definesa research
agendafortheorizing aboutbothstateagentsand thesystemstructures in
whichtheyare embedded.The core ofthisagendais theuse of structural
analysistotheorizetheconditions ofexistenceofstateagents,andtheuse of
analysisto explainthegenesisand reproduction
historical of social struc-
tures.Although evenpreliminary remarks aboutthepossiblecontent ofsuch
theorieswouldrequireanotherarticle,I can indicatesomeofthedirections
andbodiesofresearchwhichmightbe relevantto sucha researchagenda.
"Theorizingthe state" impliesa researchendeavorwhichseeks to de-
velopa theoreticallyandempirically grounded understanding ofthecausally
significant
properties (suchas powers,interests, practices)ofthestateas an
organizational agentor entity.Ideallysucha theorywoulddefineexhaus-
tivelythepossiblewaysofactingofstateagents,rather thangeneratedeter-
 Organization
366 International

minatepredictions aboutparticular statebehaviors.The possibleways of

actingof an agentare constituted and therefore explainedby the social
structural contextinwhichitis embeddedandbyitsinternal organizational
structure, and as such theyare amenableto structural Thus,
explanation.
structural analysiscan be used to explicatethesocial structural organizing
principles whichgeneratethestateas a particular kindofsocialactor,that
is, in virtueof whichthestateis a statein thefirstplace. Thisuse would
recognizethestateas an inherently socialentity,rather thanas a Hobbesian
primitive individual. Structural analysiscouldalso revealtheinternal organi-
zationalstructures of the state whichconditionits perceptionsand re-
sponses to social structural imperatives and opportunities. International
relationstheorists have tendedto discounttheimportance of suchinternal
organizational structures in explainingstateaction,78 but giventhatthese
structures willbe theproximate cause ofanystateaction,theyarelikelyto
constitute important mechanisms in theproduction of statebehavior.
To elaboratea littlefurther, at leastfoursocialstructures might constitute
states: domestic-economic, domestic-political, international-economic,
and international-political structures.79 Recentworkconsistentwiththe
critical-structuralorientation ofstructuration theoryhas examinedall these
structures, althoughmuchremainsquitepreliminary and unintegrated with
otherbodies of structural research.The mostsustainedanalysesof the
state'sstructural conditionsof existenceare foundwithintheneo-Marxist
tradition and withinthedebatesof thelatterwithWeberians.80 Whilethe
neo-Marxist literatureis concernedprimarily withthe relationship of the
(capitalist)state to domestic economic structures, world-system theorists
have concentrated on thestate'srolein international economicstructures.
Althoughthe world-system understanding of the structure of globalcapi-

78. StephenKrasner,"AreBureaucracies Important?"ForeignPolicy7 (Summer1972),pp.

159-79;RobertArt,"Bureaucratic PoliticsandAmerican ForeignPolicy:A Critique,"Policy
Sciences4 (December1973),pp. 467-90.
79. This multiplicity of structuresimpliesa rejectionof whatmightbe called structural
monism, thatis, theviewthatthereis onlyone setofunderlying organizing suchas
principles,
thoseoftheeconomy,thatcan be explicated termsandtherefore
ingenerative of
constitutive
agents.Thisanti-monism withthecritiqueof structural
is consistent Marxismdevelopedby
post-Althusserians likeBarryHindessandPaulHirstinModeofProduction andSocialForma-
tion(London:MacMillan,1977),and ErnestoLaclau and ChantalMouffein Hegemonyand
SocialistStrategy (London:Verso,1982).Buttheirdiscourse-theoretic solutiontotheproblem
ofstructuralmonism inmanywaysfundamentally opposesmysuggestion thatwe buildtheories
ofmultiple socialstructures realism.
on thebasisofscientific
80. Prominent examplesof neo-Marxist statetheoryincludeJohnHollowayand Sol Pic-
ciotto,eds.,Stateand Capital:A MarxistDebate(London:EdwardArnold,1978);Poulantzas,
State,Power,Socialism;andGoranTherborn, WhatDoes theRulingClassDo WhenIt Rules?
(London:New LeftBooks, 1978).WeberiancritiquesincludeThedaSkocpol,"PoliticalRe-
sponseto Capitalist Crisis:Neo-Marxist TheoriesoftheStateandtheCase oftheNew Deal,"
Politicsand Society10 (no. 2, 1981),pp. 155-201,and MichaelMann,"The Autonomous
PoweroftheState,"EuropeanJournalofSociology25 (no. 2, 1984),pp. 185-213.
 problem 367
Agent-structure

talismis arguablymarredby an overemphasis on exchangerelationships,81

boththe neo-Marxist and world-system literatureofferimportant insights
intotheeconomicconditions of existenceof thestate,and therefore some
of its causal powersand liabilities.Less research,I think,has been done
froma critical-structuralperspectiveon thepoliticalstructures thatmight
constitute the state.Nonetheless,at the domesticlevel,innovativework
is beingdonewhichtakesas itsstarting pointa Gramscianrejectionofthe
economismof mostneo-Marxist theoryand insteadattemptsto theorize
politicalformsin critical-structural terms.82 This line of analysisis now
beingextendedto the international level by a numberof scholarswho
have focusedon thenatureand implications of suchfundamental organiz-
ingprinciples ofthestatesystemas sovereignty, thebalanceofpower,and
hegemonicdomination forthe conceptualization of the stateand explana-
tionof stateaction.83
Purelyschematicthoughtheseremarksare, I thinkall thesebodies of
researchwouldpotentially contribute to a singleoverallproblematic gener-
atedby a structurationistapproachto international relations-thedevelop-
mentofa criticaltheoryofthecausalpowersand interests ofthestate.An
implication of a rejectionof structural monism,however,is thatit willbe
necessarytotheorizetherelationships or "articulations" betweenthediffer-
ent structures whichconstitute the state. In otherwords,a "structure-
structure"problememergesfromthe structurationist problematic.This
problemis at the core of the literature on the "articulation of modesof
production,"84andis implicit inmuchoftherecent"post-Marxist" debate.
Ironically,thisproblemis strongly reminiscent of J. D. Singer'soriginal

81. Examplesofthe"productionist" critiqueofworld-systemtheoryincludeBrenner, "The

Origins ofCapitalist
Development," andHoweandSica, "PoliticalEconomy, and
Imperialism,
theProblemof World-System Theory."The alternativeconceptualization
ofthestructureof
thecapitalistworldeconomy(in termsof a globalmodeof production) has beenmostfully
developedby the "internationalization of capital"schoolof Marxistpoliticaleconomy;see
ChristianPalloix,"The Self-Expansion ofCapitalon a WorldScale," ReviewofRadicalPolit-
ical Economics9 (Summer1977),pp. 1-28.
82. Poulantzas,State,Power,Socialism;Nicos Mouzelis,Politicsin theSemi-Periphery
(NewYork:St. Martins,1986);SamuelBowlesandHerbert Gintis,
Democracy andCapitalism
(New York:Basic Books, 1986).
83. BruceAndrews,"Social RulesandtheStateas a SocialActor,"WorldPolitics27 (July
1975),pp. 521-40;RobertCox, "Gramsci,Hegemony, andInternational
Relations: AnEssayin
Method,"Millenium12 (Summer1983),pp. 162-75;Ruggie,"Continuity and Transforma-
tion";Ashley,"The Povertyof Neo-Realism,"and "Social Willand International Anarchy:
BeyondtheDomesticAnalogyin theStudyofGlobalCollaboration," in HaywardAlkerand
Ashley,Anarchy, Power,Community: Understanding International
Cooperation(forthcom-
ing).Despitethepotential usefulnessofthisresearchto thestructurationist
problematic,how-
ever,someof thesescholarswouldprobablyrejectassociationwiththattheory,especially
insofaras itis groundedin realistphilosophyofscience.
84. AidanFoster-Carter, "The Modes ofProduction Controversy,"New LeftReview107
(January-February 1978),pp. 47-77;HaroldWolpe,ed., TheArticulationofModesofProduc-
tion(London:Routledge& KeganPaul, 1980).
368 International
Organization

discussionofthelevelsofanalysisproblem,85 although thelatterwouldbe

seenhereas one instanceofa moregeneraltheoretical and methodological
problemof apprehending therelationships betweendifferent structures of
whatever type(politicalor economic)orlevelofanalysis(domesticor inter-
national).
Although as a theoretical and methodological issuethe"structure-
structure
problem"pointsbeyondstructuration theory, itstreatment clearly
has implications fortheattempt to buildsatisfactory theoriesand explana-
tionsofstateaction.Each structure inwhichthestateis embeddedwillhave
itsownlogic,reproduction requirements, andthusprescriptions forcompe-
tentstatepractice.Developinga structural theoryof the stateand state
action,then,involvesmorethansimplyexplicating thedifferent structuring
principles whichgeneratestates;italso requiresisolating and assessingthe
causalroleofandinterrelationships amongdifferent andsometimes compet-
ingstructural determinations of stateaction.
The needfora theoryofthestateininternational relationsis mirrored by
theneedfortheoriesofthesystemstructures whichconstitute thestate.In
generalterms,thesetheorieswouldhave at leasttwomainelements:1) a
synchronic modelof theorganizing principles,logic,and reproduction re-
quirements of thestructure in question,and 2) an historical accountofthe
genesisand reproduction of thestructure. Structural theorists have gener-
ally been reluctantor unableto graspthe contingent natureof structural
genesisand reproduction, and have tendedinsteadtowardsfunctional or
teleologicalreadingsofthatprocess.Thistendency can be corrected bythe
emphasisof structuration theoryon the historicalspecificity and contin-
gencyofthestructuring ofsocialstructures. Giventhestructurationist con-
ceptualization ofsocialstructure as onlyinstantiated andreproduced bythe
activitiesof social agents,an historical analysisof social structuring must
beginwiththeintendedand unintended consequencesof stateaction(and
theactionof otheragents).Althoughtherelevantmethodological toolsto
suchan analysisare potentially quitevaried,thegame-theoretic methodol-
ogycharacteristic ofrecentworkinneorealism is potentiallywell-adapted to
thisanalyticaltask. The analysisof iteratedgames and the "new insti-
tutionalism" in thestudyof politicalinstitutions in particularhave proven
usefulingenerating insightsintotheemergence ofandreproduction ofsocial
institutionsas theunintended consequencesof strategic interactions,86and
thereis no a priorireasonwhywe cannotextendthelogicofsuchanalyses
to theanalysisofgenerative structures. We mustrecognize,however,that

85. J.D. Singer,"The Levels of AnalysisProblemin International Relations,"in Klaus

Knorrand SidneyVerba,eds., The International System:Theoretical Essays (Princeton:
PrincetonUniversityPress,1961),pp. 77-92.
86. MichaelTaylor,Anarchyand Cooperation (New York:Wiley,1976);RobertAxelrod,
TheEvolution ofCooperation(NewYork:Basic Books,1984);JamesMarchandJohanOlsen,
"The New Institutionalism:
OrganizationalFactorsin PoliticalLife,"AmericanPoliticalSci-
enceReview78 (September1984),pp. 734-48.
 problem 369
Agent-structure

game-theoretic
modelsfocusattention
on thetechnical of
decisionproblems
given agents,and thattheytherefore
tendto neglectthewaysin whichthe
structureof social interactions constitute or empowerthoseagentsin the
firstplace. The use ofgametheory to developan historicalunderstanding of
the emergenceof social structures, therefore, wouldhave to be comple-
mentedby a generativeunderstanding of the construction of agentsand
situationsof strategic interaction.
These remarkson theimplications of structurationtheoryforthescope
and contentof international relationstheoriesare obviouslypurelysche-
maticand are intendedonlyto illustrate thekindsofresearchthatmightbe
relevantto a structurationist approachto international relations.Indeed,I
shouldemphasizethatstructuration theory byitselfcannotgeneratespecific
theoretical claimsaboutinternational relations.The theory has epistemolog-
icalimplications fortheformwhichexplanations ofstateactionshouldtake,
and it suggestsa researchagendaforsubsequenttheorizing, butitdoes not
makea directcontribution to oursubstantive understanding ofinternational
relationsperse. Thispointraisestheissueofthecriteria bywhichstructura-
tiontheoryshouldbe evaluatedbyscholarsofinternational relations.Given
thatits analyticalor meta-theoretical qualitypreventsan empiricalassess-
mentof the theory,it seems to me thatstructuration theoryshouldbe
evaluatedon pragmatic grounds,on itsabilityto solveproblemsin existing
substantive theories,to suggestnew areas of theoretical and empiricalin-
quiry,or to integrate different bodiesofresearch.By thiscriterion, I think
structuration theoryclearlyimpoveson its principalcompetitors, individ-
ualismand structuralism. It providesa framework forexplaining theproper-
ties of both state agentsand systemstructures whichis deniedto the
individualist and structuralist ontologiesof neorealismand world-system
theory, anditdefinesa researchagendaforinternational relationsthatorga-
nizesand subsumesundera singleproblematic a potentiallygreatervariety
of extantsocial scientific research.The potentialcontribution to interna-
tionalrelations researchis there,butwe cannotassess theimportance ofthe
contribution untiltheoristsattempt to use a structurationistperspectiveto
groundand inform theirtheoretical and empirical research.

Conclusion

Ratherthantrying to summarize a longargument abouttheagent-structure

problemand structuration theory,I wantto concludethisarticlebypicking
up its othermainthread,the implications realismforsocial
of scientific
scientific
research.Whether theoryprovidesa satisfac-
or notstructuration
toryresolutionto theagent-structureproblem,socialscientists'
adoptionof
a scientific
realistperspectiveon ontologyand epistemology could have
potentiallyrevolutionaryconsequencesfortheirtheoretical and empirical
370 International
Organization

research.The hegemonyof empiricist discoursein social sciencehas led

social scientistsintoan apparentdichotomy between"science" (thatis,
empiricistscience) and the allegedly"un-scientific" paradigmsof her-
meneuticsand criticaltheory.87 Whateverthe limitsto naturalism in the
socialsciences,scientificrealismundermines thisdichotomy bychallenging
thecore of theempiricist's argument, theinterpretation of naturalscience
uponwhichherappropriation of themantleof "science" rests.Scientific
realism,then,offers an alternativeto thestandard positionsinthePositivis-
musstreit, one whichenjoinssocial scientists to think"abductively"about
"causal mechanisms"to buildtheirtheories,insteadof trying to findlaw-
likegeneralizations aboutobservableregularities. Amongthemoreimpor-
tantconsequencesof such an ontologicaland epistemological shiftis a
scientificmotivation forstructuraltheorizing in thegenerative or relational
sense.Thismethodological prescription is inherently"critical"sinceit re-
quiresa critiqueandpenetration ofobservableforms totheunderlying social
structures whichgeneratethem.88 An implication realism,then,
ofscientific
is that"criticaltheory"(ina broadsense)is essentialtothedevelopment of
social science,and by extensioninternational as a "science."
relations,

87. BrianFay,Social Theoryand PoliticalPractice(London:Allen& Unwin,1975).

88. Roy Bhaskar,"Scientific
ExplanationandHumanEmancipation," RadicalPhilosophy
Dandeker,"TheoryandPracticein Sociology:The Critical
26 (1980),pp. 16-26;Christopher
ImperativesofRealism,"JournalfortheTheory ofSocialBehavior13(July1983),pp. 195-210.