Trustees of Princeton University

The Level-of-Analysis Problem in International Relations Author(s): J. David Singer Reviewed work(s): Source: World Politics, Vol. 14, No. 1, The International System: Theoretical Essays (Oct., 1961), pp. 77-92 Published by: Cambridge University Press Stable URL: http://www.jstor.org/stable/2009557 . Accessed: 07/08/2012 12:23
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

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

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

http://www.jstor.org

THE LEVEL-OF-ANALYSIS PROBLEM IN INTERNATIONAL RELATIONS

TN anyareaof scholarly there alwaysseveral are waysin inquiry, the under and 1 which phenomena study be sorted arranged may for purposes systemic of analysis. Whether the in physical social or sciences, theobserver choose focus may to or upontheparts uponthewhole, upon componentsuponthesystem. may, example, the or He for choose between flowers thegarden, rocks thequarry, trees the or or the the or theforest, houses theneighborhood, carsor thetraffic the or the jam, the delinquents thegang, legislators thelegislative, so on.1 or the or and Whether selects micro- macro-level analysis ostensibly he the of or is a mere matter methodological conceptual of or convenience. the Yet choice often turns to be quitedifficult, maywell become out and a central issuewithin discipline the concerned. and The complexity significance theselevel-of-analysis of decisions readily are suggested by the long-standing controversies between social psychology and sociology, personality-orientedculture-oriented and anthropology, or micro- macro-economics, and to mention a few.In thevernacular but of general systems theory, observer always the is confronted a with system, sub-systems, their its and respective environments, while and he maychoose hissystem cluster phenomena as any of from most the minute organism theuniverse to itself, choice such cannot merely be a function whimor caprice, of habitor familiarity.2 responsible The scholar mustbe prepared evaluate relative to the utility-conceptual and methodological-of various the alternatives to him,and to open appraise manifold the implications the level of analysis of finally selected. itiswith So international relations. But whereas prosand cons of the various the possible levelsof have analysis beendebated in exhaustivelymany thesocialsciences, of theissuehas scarcely beenraisedamongstudents our emerging of
1 As KurtLewinobserved his classic in contribution thesocialsciences: to "The first prerequisite a successful of in observation any science a definite is about understanding what size of unit one is going to observeat a given time."Field Theoryin Social

2For a usefulintroductory statement the definitional taxonomic on and problems in a generalsystems see approach, the papersby Ludwig von Bertalanffy, "General System and KennethBoulding,"GeneralSystem Theory," Theory:The Skeletonof Science,"in Societyfor the Advancement General SystemsTheory, General of Ann Systems, Arbor, Mich.,I956, I, partI.

Science, New York, I95I, P. I57.

78

THE INTERNATIONAL

SYSTEM

discipline.3 tranquillity be seenbysome a reassuring as Such indimay cation that issue notgermane ourfield, byothers evithe is to and as dencethatit has already beenresolved, thiswriter but the perceives quietude witha measure concern. is quitepersuaded its He of of relevance certain it has yetto be resolved. and that it Rather, is contended theissuehas beenignored scholars steeped the that in still by intuitive artistic and tradition thehumanities enmeshed the of in or webof"practical" We and policy. have, ourtexts elsewhere, roamed in up anddowntheladder organizational of complexity remarkable with abandon, focusing uponthetotalsystem, international organizations, regions, coalitions, extra-national associations, nations, domestic pressure social groups, and as classes, elites, individuals theneeds themoment of And most us havetended settle required. though of to uponthenation as ourmost comfortable resting place, haveretained propensity we our forvertical drift, failing appreciate valueof a stable to the pointof focus.4 Whether lackofconcern a function therelative this is of infancy ofthediscipline thenature theintellectual of or traditions whence from
it springs, nevertheless it remains significant a variablein the general sluggishness whichcharacterizes development theory thestudy the of in of relations amongnations. is thepurposeof thispaperto raisethe It issue,articulate alternatives, examinethe theoretical the and implicationsand consequences two of themorewidelyemployed of levelsof analysis:the international system and the nationalsub-systems.
I.
THE REQUIREMENTS OF AN ANALYTICAL MODEL

Priorto an examination the theoretical of implications the level of of analysis orientation or in employed our model,it mightbe worthwhile to discuss uses to whichany such model mightbe put,and the therequirements whichsuchusesmight of expect it. we Obviously, would demandthatit offer highly a accurate descriptionofthephenomena underconsideration. Therefore scheme the must as and present complete undistorted picture thesephenomena is a of as possible;it mustcorrelate withobjective reality and coincidewithour to referents the highest empirical possibledegree.Yet we know that
3An important to pioneering attempt deal withsomeof theimplications one'slevel of of analysis, is N. however, Kenneth Waltz,Man, theState,and War,New York,I959. But Waltz restricts himself a consideration theseimplications theyimpinge to of as on thequestion thecausesof war. See also thiswriter's of reviewof Waltz, "International Conflict: ThreeLevelsof Analysis," xii WorldPolitics, (April i960), pp. 453-6i. 4Even duringthe debatebetween"realism"and "idealism"the analytical implicationsof the variouslevelsof analysis received onlythe scantiest attention; ratherthe seemsto have been at the two extremes pragmatic emphasis of policyand speculative

metaphysics.

THE LEVEL-OF-ANALYSIS PROBLEM

79

suchaccurate of a and representation complex wide-ranging of body phenomena extremely is difficult. a illustration be Perhaps useful may borrowed cartography; oblate from the which planet the spheroid earth most isnot closely represents transferable two-dimensional tothe surface of a map without some distortion. Thus, the Mercator projection exaggerates distance distorts and direction an increasing as we at rate move north south or the from equator, whilethepolargnomonic projection suffers these from same debilities wemove as toward equator. the Neither offers therefore a wholly accurate presentation, eachis true yet enough reality be quiteuseful certain to to for The specific purposes. samesort tolerance necessary evaluating analytical of is in any model for study international the of if relations; we must sacrifice repretotal sentational accuracy, problem to decide the is where distortionleast is dysfunctional where and suchaccuracy absolutely is essential. Thesedecisions in turn, function thesecond are, a of requirement of anysuchmodel-a capacity explain relationships to the amongthe phenomena under investigation. ourconcern notso much Here is with of accuracy descriptionwith as validity explanation. model of Our must havesuchanalytical capabilities to treat causalrelationships as the in a fashion which notonly is validandthorough, parsimonious; but this latter requirement is often overlooked, itsimplications research yet for strategy notinconsequential.5 are It should asserted be herethatthe primary purpose theory to explain, of is and whendescriptive and explanatory requirements in conflict, latter are the ought be given to even priority, at thecost some of representational inaccuracy. Finally, may legitimately we demandthatany analytical model offer promise reliable the of prediction. mentioning requirement In this last, there no implication it is themostdemanding difficult is that or ofthethree. Despite popular the belief thecontrary, to prediction demands of one'smodelthandoesexplanation evendescription. less or For example, informed any layman predict pressure the can that on
5For example, one criticof the decision-making model formulated RichardC. by Snyder, W. Bruck,and BurtonSapin,in Decision-Making an Approachto the H. as Studyof International Politics(Princeton, N.J.,I954), pointsout thatno singleresearcher could deal withall the variables thatmodel and expectto complete in more thana very comparative few studies his lifetime. Herbert in See McClosky, "Concerning Strategies a Scienceof International for viii (January Politics," WorldPolitics, I956), pp. 28I-95. In defense, however, one mightcall attention the relativeease with to whichmanyof Snyder's categories could be collapsed intomoreinclusive ones,as was apparently done in the subsequent case study(see note ii below). Perhapsa more telling criticism themonograph McClosky's of is comment that"Untila greater measure of theory introduced theproposal is into and therelations amongvariables specified are moreconcretely, is likelyto remainlittlemore thana setting-out categories it of and, like any taxonomy, fairly limited its utility" 29I). in (p.

80

THE INTERNATIONAL

SYSTEM

car its accelerator a slowly of moving will increase speed; thatmoreor less of the moon will be visibletonight than last night; or that the normalhumanwill flinch when confronted withan impending blow. Thesepredictions notrequirea particularly do or elegant sophisticated but model of the universe, theirexplanation demandsfar more than mostof us carry aroundin our minds.Likewise,we can predict with impressive reliability anynationwill respond military that to attackin and kind,but a description understanding the processes of and factors leadingto such a response considerably are more elusive,despitethe grosssimplicity theactsthemselves. of Having articulated ratherbriefly requirements an adequate the of analytical turnnow to a consideration thewaysin model,we might of whichone's choiceofanalytical focusimpinges upon sucha modeland affects descriptive, its and explanatory, predictive adequacy.
II.
THE INTERNATIONAL SYSTEM AS LEVEL OF ANALYSIS

with the systemic level of analysis, findin the total Beginning we international system partially a familiar and highlypromising point offocus. First all, itis themostcomprehensive thelevelsavailable, of of the encompassing totality interactions of whichtake place withinthe and system itsenvironment. focusing the system, are enabled on By we to studythe patterns interaction of which the system reveals, and to generalizeabout such phenomenaas the creationand dissolution of the coalitions, frequency duration specific and of powerconfigurations, in modifications its stability, responsiveness changesin formal its to and thenormsand folklore political institutions, whichit manifests as a societal In system. other the words, systemic levelof analysis, only and thislevel,permits to examineinternational us relations the whole, in with a comprehensiveness is of necessity when our focusis that lost to shifted a lower,and more partial,level. For descriptive purposes, it both then, offers advantages disadvantages; former and the flowfrom itscomprehensiveness, thelatter and from necessary the dearth detail. of As to explanatory the system-oriented capability, model poses some In genuine difficulties. the first place, it tends to lead the observer into a position whichexaggerates impactof the system the upon the nationalactorsand, conversely, discounts impactof the actorson the the system. This is, of course, no meansinevitable; by one could conlookupon thesystem a rather ceivably as passiveenvironment which in states out their act dynamic relationships rather thanas a socio-political with a dynamic its own. But thereis a naturaltendency of entity to endowthatupon whichwe focusour attention withsomewhat greater

THE LEVEL-OF-ANALYSIS PROBLEM

81

and pain, or gain and loss, nationsmay differ widelyin what they consider be the nationalinterest, we end up havingto break to and down and refine largercategory. the Moreover, Professor Morgenthau finds himself compelled go stillfurther disavowthe relevance to and of bothmotives and ideological preferences nationalbehavior, in and theserepresent of the more usefuldimensions differentiating two in among the severalnationsin our international system. eschewing By anyempirical concern withthedomestic internal and variations within the separate nations, system-oriented the approachtendsto producea sortof "black box" or "billiardball" conceptof the nationalactors.7 By discounting-ordenying-the differences among nations,or by
6 Hans J. Morgenthau, PoliticsAmongNations,3rd ed., New York, i960, pp. 5-7. Obviously, model does not precludethe use of power as a dimension his for the differentiationnations. of 7 The "black box" figure comesfrom someof thesimpler versions S-R psychology, of in whichthe observer more or less ignores what goes on withinthe individualand concentrates upon the correlation betweenstimulus and response;theseare viewed as empirically verifiable, whereas cognition, and perception, other mental processes have to be imputed the individual to witha heavyreliance theseassumed"intervening on variables." The "billiard ball" figure seemsto carry same sortof connotation, the and is bestemployed Arnold by in Wolfers "The Actors International in in Politics" William T. R. Fox, ed., Theoretical Aspectsof International Relations, Notre Dame, Ind., I959, pp. 83-i06. See also, in thiscontext, RichardC. Snyder, "International Relations Theory-Continued," xiii (January World Politics, i96i), pp. 300-I2; and J.David Singer, "Theorizing AboutTheory International in Politics," Journal Conflict of Resolution,iv (Decemberi960), pp. 431-42. Both are reviewarticles dealingwith the Fox anthology.

potential it might than be to normally expected have.Thus,we tend to move,in a system-oriented away fromnotions model, implying muchnational and of autonomy independence choiceand toward a more deterministic orientation. this Secondly, particular of analysis level almost inevitably requires thatwe postulate highdegree uniformity theforeign a of in policy operational ofournational codes actors. definition, allowlittle By we room divergence thebehavior ourparts for in of when focus we upon thewhole. is no coincidence ourmost It that prominent theoreticianand oneof thevery text few writers focusing upontheinternational system-should "assume that[all] statesmen think and act in terms ofinterest defined power."6 this as If single-minded behavior interbe preted and literally narrowly, havea simplistic we imagecomparable to economic or sexual man man,and ifit be defined broadly, are we no better thanthepsychologist off whose human model pursues "selfrealization" "maximization gain";all suchgross or of models suffer from samefatal the weakness theutilitarian's as "pleasure-pain" principle. as individuals widely what Just differ in they deem be pleasure to

82

THE INTERNATIONAL SYSTEM

differences of of observing many these positing near-impossibility the with image one homogenized them,8 concludes a highly atwork within And this in system. though maybe an ofournations theinternational it uponwhichto base anycausalstatements, inadequate foundation More statements. adequatebasisforcorrelative offers reasonably a between correlations and it permits to observe measure us specifically, and seem impinge uponthenation which to certain forces stimuli or of whichare the apparent consequence these the behavior patterns impliedin the word the stimuli. But one muststress limitations of stimulus what thought be theconsequence a given is to "apparent"; the and or mayonlybe a coincidence artifact, untilone investigates how persuasive the majorelements the causal link-no matter in of not deductive speak only correlation, ofconsequence. logic-onemay of the pitfalls intra-nation Moreover, avoiding multitudinous by emof oronerous sophistication ingas itdoeslittle themethodological beneath behavioral one the externalities piricism called when probes for in beensuggested theintroduction, oftheactor. Finally, hasalready as as orientation should to satisfactory thesystemic prove be reasonably the is even beyond a basisfor prediction, if suchprediction to extend regardand anticipatory statements characteristics system attempt ofthe of that are this themselves; assumes, course, theactors ing theactors grossand in predicted relatively characterized theirbehavior and terms. general of of implicationsa model are significant These, then, some themore as a whole.Let us turn which focuses system upontheinternational
the of now to the morefamiliar our two orientations, nationalstate itself. III. THE NATIONAL STATE AS LEVEL OF ANALYSIS witha singularly model,requirmanageable one observation, emerges

The other in to levelof analysis be considered thispaperis the national This is relations. state-our primary actorin international students, is theone and focus among Western clearly traditional the dominates which all in employed English-speaking almost ofthetexts and colleges universities. Itsmost significant differentiation obvious advantage that permits is it in our system. Because doesnotreit among actors theinternational the of it to actors, encoursimilaritythenational quire attributiongreat
for thatit is "futile" searchformotives because to 8Morgenthau observes, example, are of beyond as are,frequently data, they "themostillusive psychological distorted they and of alike" (op.cit.,p. 6). recognition, the interests emotions actorand observer by

THE LEVEL-OF-ANALYSIS PROBLEM

83

ages the observer examine The favorable detail. to themin greater of results suchintensive cannot overlooked, it is only be as analysis whentheactors studied somedepth are in we are ableto make that really validgeneralizationsa comparative of And though nature. the systemic model doesnotnecessarily and preclude comparison contrast amongthenational in it usually eventuates rather sub-systems, gross comparisons on relatively based crude dimensions characteristics. and On theother is hand, there no assurance thenation-oriented that approach produce sophisticated will a modelforthecomparative study offoreign with policy; perhaps exception theHaas andWhiting the of study,9 ofourmajor none texts a serious successful to makes and effort and describe explain national behavior terms most thesignifiof in of cantvariables whichsuchbehavior by might comparatively be anaBut wouldseem be a function, ofthelevel analysis lyzed. this to not of employed, of our general but with unfamiliarity the othersocial sciences whichcomparison a majorpreoccupation) of the (in is and retarded ofcomparative state and government politics,field which a in mostinternational relations specialists likely have had some are to Butjustas thenation-as-actor permits toavoid inaccurate focus us the homogenization often which flows from systemic the focus, alsomay it leadus intotheopposite ofdistortion-a type marked exaggeration of thedifferences among sub-systemic Whileit is evident our actors. that neither these of extremesconducive thedevelopmenta sophistiis to of catedcomparison foreign of policies, suchcomparison and requires a balanced preoccupation both with similarity difference, danger and the seems be greatest to when succumb thetendency overdifferenwe to to tiate;comparison contrast proceed and can onlyfrom observed uniformities.'0 experience.

Oneoftheadditional liabilities which flow turn in from pressure the to overdifferentiate of Ptolemaic is that parochialism. Thus,in overemphasizing differences the amongthe manynationalstates, the observer prone attribute is to many whathe conceives be virtues of to tohisownnation thevices others, and to especially adversaries the of themoment. Thatthisethnocentrism no meansan idlefearis is by borne byperusal themajor out of international relations published texts

9ErnstB. Haas and Allen S. Whiting, Dynamicsof International Relations, New York,1956. 10A frequent by-product thistendency overdifferentiate of to is whatWaltz calls the "second-image in fallacy," which one explainsthe peacefulor bellicosenatureof a nation's foreign policy in exclusively terms its domestic of economic, or political, social characteristics (op.cit., chs.4 and 5).

84

THE INTERNATIONAL SYSTEM

in the UnitedStatessincei945. Not onlyis the worldoften perceived the of through prism theAmerican nationalinterest, an inordinate but of degree attention notspleen) is directed towardtheSovietUnion; (if it would hardly amissto observe be thatmostof thesemightqualify in equallywell as studies American The scientific inadeforeign policy. quaciesof thissortof "we-they" orientation hardly requireelaboration, yet theyremaina potentdanger in any utilizationof the national actormodel. Another significant implication thesub-systemic of orientation that is it is onlywithin particular its framework we can expectanyuseful that application thedecision-making of approach.1" all of us, of course, Not willfind inapplicabilitymajorloss; considering criticism its a the which has been leveledat the decision-making and the failureof approach, mostof us to attempt application, mightconcludethatit is no its one lossat all. Buttheimportant thing notehereis thata system-oriented to modelwould notoffer hospitable a framework sucha detailedand for comparative approachto thestudy international of relations, matter no what our appraisalof the decision-making approachmightbe. Another perhapsmoresubtle and implication selecting nation of the as our focusor level of analysis thatit raisesthe entire is question of goals,motivation, purpose national and in policy.'2 Thoughit maywell be a peculiarity the Westernphilosophical of tradition, seem to we exhibit, when confronted with the need to explainindividual color lective behavior, strong a proclivity a goal-seeking for approach.The questionof whethernationalbehavioris purposive not seems to or require discussion twodistinct in (butnotalwaysexclusive) dimensions. Firstly, there the moreobviousissueof whether is thosewho act on behalf thenationin formulating executing of and foreign policyconsciously pursue rather concrete goals.And it would be difficult deny, to forexample, thattheserole-fulfilling individuals envisage certain specificoutcomes which theyhope to realize by pursuinga particular
1"Its mostwell-known and successful is statement foundin Snyderet al., op.cit. Much of thismodel is utilizedin the textwhich Snyderwrotewith Edgar S. Furniss,Jr., American ForeignPolicy:Formulation, and Principles, Programs, New York, I954. A morespecific application foundin Snyder is and GlennD. Paige, "The United States Decision to Resist Aggression Korea: The Application an Analytical in of Scheme," Administrative ScienceQuarterly, (DecemberI958), pp. 34i-78. For those iii interested this approach, in very usefulis Paul Wassermanand Fred S. Silander, Decision-Mating: Annotated An Bibliography, Ithaca,N.Y., 1958. 12 And if thedecision-making version thismodelis employed, issueis unavoidof the able. See the discussion motivation Snyder, of in Bruck, and Sapin,op.cit., pp. 92-I17; note that25 of the 49 pages on "The Major Determinants Action"are devotedto of motives.

THE LEVEL-OF-ANALYSIS PROBLEM

85

In strategy. this sense,then,nationsmay be said to be goal-seeking organisms whichexhibit behavior. purposive However,purposiveness may be viewed in a somewhatdifferent light, askingwhether is not merely intellectual by it an construct that man imputes himself reasonof his vain addiction thefree-will to to by doctrine he searches characteristics as for whichdistinguish from him physicalmatter and the lower animals.And having attributed this conscious behavior himself an individual, maybe goal-pursuing to as it arguedthatman thenproceedsto projectthisattribute the social to organizations whichhe is a member. of The questionwould seem to distill downto whether man and his societies own pursuegoals of their choosingor are moved towardthose imposedupon themby forces whichareprimarily their control.13 beyond Another the way of stating dilemmawould be to ask whether are concerned we with the ends whichmen and nationsstrive or the ends towardwhich theyare for impelledby the past and presentcharacteristics theirsocial and of physical milieu.Obviously, areusingtheterms we "ends,""goals,"and "purpose" two rather in distinct ways; one refers thosewhich are to consciously envisagedand more or less rationally pursued,and the other thoseofwhichtheactor little to has knowledge toward but which he is nevertheless propelled. Taking a middlegroundin whatis essentially specific a case of the freewill vs. determinism debate,one can agree that nationsmove toward outcomes whichthey of have little knowledge and overwhich have less control, thattheynevertheless prefer, therethey but do and foreselect, particular outcomesand attempt realize themby conto sciousformulation strategies. of Also involvedin the goal-seeking problemwhen we employthe nation-oriented modelis thequestionof how and why certain nations pursuespecific sortsof goals. While the questionmay be ignoredin thesystem-oriented model or resolved attributing by identical goals to all nationalactors, nation-as-actor the approachdemandsthatwe investigate processes whichnational the by goals are selected, internal the and external factors thatimpingeon thoseprocesses, and the institutionalframework fromwhichtheyemerge.It is worthy note that of the despite strong predilection the nation-oriented for model in most
13 A highly but suggestive, moreabstract treatment thisteleological of is question in TalcottParsons, The Structure Social Action, 2nd ed., Glencoe, of Ill., I949, especially in his analysis Durkheim Weber.It is interesting notethatforParsons act of and to an implies, interalia, "a future stateof affairs towardwhich the processof action is and he therefore oriented," comments that "in this sense and this sense only,the schema action inherently of is teleological" 44). (p.

86

THE INTERNATIONAL SYSTEM

of our texts, or empirical evendeductive of analyses theseprocesses are few.'4Again, one mightattribute conspicuously theselacunae to the and methodological conceptual inadequaciesof the graduatetraining which international relations But in specialists receive.'5 traditionally any event, goals and motivations bothdependent are and independent variables, and if we intendto explain a nation'sforeign policy,we cannotsettle themerepostulation thesegoals; we are compelled for of to go back a step and inquireinto theirgenesisand the processby which theybecomethe crucialvariables thattheyseem to be in the behavior nations. of There is still anotherdilemma involvedin our selectionof the nation-as-actor model,and thatconcerns phenomenological the issue: do we examineour actor'sbehavior termsof the objectivefactors in whichallegedly influence behavior, do we do so in terms the that or of actor'sperception these"objective of factors"? Though thesetwo approaches notcompletely are exclusive one another, proceed of they from greatlydifferent and oftenincompatible assumptions, and produce markedly modelsof nationalbehavior.'6 divergent The first theseassumptions of concerns broadquestionof social the causation.One view holds thatindividuals and groupsrespondin a quasi-deterministic fashionto the realitiesof physicalenvironment, theactsor powerofother individuals groups, similar or and "objective" and "real" forces stimuli. oppositeview holds thatindividuals or An and groups notinfluenced their are in behavior suchobjective by forces, but by the fashion whichtheseforces perceived in are and evaluated, however distorted incomplete or suchperceptions be. For adherents may of thisposition, onlyreality thephenomenal-that the is whichis discernedby thehumansenses;forces thatare not discerned not exist do
14 Among the are exceptions Haas and Whiting, chs. 2 and 3; and some of op.cit., thechapters RoyC. Macridis, Foreign in ed., Policyin WorldPolitics, Englewood Cliffs, N.J., I958, especially that WestGermany Karl Deutschand Lewis Edinger. on by 15 As early I934, EdithE. Warenoted as that". . . thestudy international of relations is no longerentirely subjectfor politicalscienceor law, but thateconomics, a history, sociology, geography-all socialsciences-arecalledupon to contribute the towards the .. understanding . of theinternational See system." The StudyofInternational Relations in the UnitedStates, New York, 1934, p. 172. For some contemporary suggestions, see Karl Deutsch, "The Place of Behavioral Sciences Graduate in Training International in Relations," Behavioral Science, (July1958), pp. 278-84; and J.David Singer, III "The Relevance the Behavioral of Sciencesto the Studyof International Relations," ibid.,VI (Octoberi96i), pp. 324-35. 16The fatherof phenomenological philosophy generally is acknowledged be to EdmundHusserl(i859-I938), author Ideas: General of Introduction PurePhenomento ology, New York,1931, trans. W. R. BoyceGibson;theoriginal published 1913 by was in underthetitleIdeen zu einerreinen Phdnomenologie Phdnomenologischen und Philosophie. Application thisapproach socialpsychology comeprimarily of to has through the workof Koffka Lewin. and

THE LEVEL-OF-ANALYSIS PROBLEM

87

nized and appraised, mustconcedethatperceptions certainly one will affect manner whichsuchforces responded As has often the in to. are beenpointed out,an individual will fall to the groundwhen he steps outofa tenth-story windowregardless hisperception gravitational of of forces, but on the otherhand such perception a major factorin is whether nothe steps ofthewindowin thefirst or out place."7 The point hereis thatif we embracea phenomenological view of causation, we willtendto utilizea phenomenological modelforexplanatory purposes. The second assumption which bears on one's predilection the for phenomenological approach is more restricted, and is primarily a one. Thus, it may be arguedthat any description methodological of nationalbehaviorin a given international situation would be highly were it to ignorethe link betweenthe external incomplete forcesat workupon thenationand itsgeneral foreign policybehavior. Furthermore,if our concernextendsbeyondthe mere description "what of happens"to therealmof explanation, could be contended it thatsuch omission the cognitive of and the perceptual linkagewould be ontologically disastrous. How, it might asked,can one speakof "causes" be ofa nation's policies whenone has ignored mediabywhichexternal the conditions factors translated and are into a policydecision?We may observecorrelations betweenall sortsof forcesin the international system the behavior nations, theircausal relationship and of but must remainstrictly deductive and hypothetical the absenceof empirical in investigation into the causal chain which allegedlylinks the two. Therefore, evenif we are satisfied withthe less-than-complete descriptivecapabilities a non-phenomenological of model,we are stilldrawn to it ifwe are to make anyprogress explanation. in The contrary view would hold thatthe above argument proceeds froman erroneous comprehension the natureof explanation of in socialscience. One is by no meansrequiredto traceevery perception, and transmission, receipt betweenstimulus and response inputand or in output order explain behavior thenation anyother to the of or human group.Furthermore, is to saythatempirical who observation-subject
17 This issuehas beenraisedfrom timeto timein all of thesocialsciences, foran but excellent discussion it in termsof the present of see problem, Harold and Margaret Sprout, Man-Milieu Relationship Hypotheses the Contextof International in Politics, Princeton University, Center International of Studies, 1956,pp. 63-71.

in for actor, those do exist soonly the that and in fashion which that do they perceived. that are the Though is difficultaccept position an it to as individual,group, a nation affected suchforces climate, a or is by distance, a neighbor's or are physical as power only insofar they recog-

88

THE INTERNATIONAL SYSTEM

as it is to a hostof errors-is better basisof explanation a any than informed Isn'tan explanation deduction, inference,analogy? or which flows modeljust as reliable logically from coherent a as theoretical onebasedupona misleading elusive most which of and of body data, is susceptible analysis to onlyby techniques concepts and to foreign political science history? and This leads,in turn, thethird thepremises to of relevant one's to stand thephenomenological arethedimensions characon issue: and teristicsthepolicy-makers' of phenomenal empirically field discernible? Or, moreaccurately, if we are convinced their even that perceptions andbeliefs constitute a crucial in variable theexplanation a nation's of foreign policy, they observed an accurate can be in and systematic ?18 fashion Furthermore, we notrequired thephenomenological are by modelto go beyond classification description suchvariables, a and of and be drawn thetangled ofrelationships ofwhich into web out they If emerge? we believe these that phenomenal are variables systematically observable, explainable, canbe fitted ourexplanation are and into ofa nation's behavior theinternational in system, there a further then is tendency embrace phenomenological to the approach. not, ifwe If or are convinced thegathering suchdatais inefficientunecothat of or nomical, willtend shy we to clear it. of The fourth in thephenomenological issue dispute concerns very the nature the of nation an actor international as in relations. orwhat Who is it thatwe study? it a distinct Is socialentity withwell-defined boundaries-a untoitself? is it an agglomeration indiunity Or of viduals, institutions, customs, procedures? shouldbe quite and It whoviewthenation thestate an integral evident those that or as social unitcouldnotattach much to utility thephenomenological approach, ifthey particularly areprone concretize reify abstraction. to or the Such abstractions incapable perception, are of cognition, anticipation or of the (unless, course, reification so faras to anthropomorphize goes and assign the abstraction attributes will,mind,or perto such as On if sonality). theother hand, thenation state seenas a group or is within institutional of individuals an operating framework, it then makesperfect to on sense focus thephenomenal of those field indiwhoparticipate thepolicy-making viduals in In process. other words, people capable experiences, are of images, expectations, instiand while
18This is another the criticisms of leveledat the decision-making approachwhich, almostby definition, seemscompelledto adopt some formof the phenomenological model.For a comprehensive treatment the elements of involved humanperception, in see Karl Zeneret al., eds., "Inter-relationships Between Perception Personality: and A
Symposium,"Journalof Personality,xviii
(1949),

pp. I-266.

THE LEVEL-OF-ANALYSIS PROBLEM

89

sense. tutional in abstractions not, are except themetaphorical Thus,if our actorcannot evenhavea phenomenal is point field, there little in employingphenomenological a approach.19 which phenomenothe These, then, some thequestions are of around of wouldseemto revolve. logical issue Thoseof us who think social forces operative as that regardless theactor's of who awareness, believe need explanation notinclude of thesteps a causalchain, in who all aredubious thepracticalitygathering of of phenomenal or who data, visualize nation a distinct the as entity apart from individual its members, will tendto reject phenomenological the approach.20 Logically, who disagree onlythose witheach of the abovefourassumptions be would compelled adopt approach. to the with Disagreement anyone wouldbe sufficient for grounds so doing. and some The aboverepresent of themoresignificant implications fascinating problems raised theadoption oursecond of model. by They seem indicate thissub-systemic to that orientationlikely produce is to richer description more and satisfactory theempiricist's of (from point view)explanation international of relations, though predictive its power would no appear greater thesystemic than orientation. thedescripBut tiveand explanatory advantages achieved are onlyat the priceof considerable methodological complexity.
IV. CONCLUSION

Having discussed ofthedescriptive, some explanatory, predictive and capabilities these of two possible levelsof analysis, might it now be useful assess relative to the utility thetwoandattempt general of some statement to their as prospective contributionsgreater to theoretical in growth thestudy international of relations. In terms description, findthat systemic produces of we the level a more comprehensive totalpicture international and of relations than does national sub-systemic On theother or the level. hand, atomized the coherent andless level analysis someimage produced thelower by of is whatbalanced itsricher and by detail, greater depth, moreintensive portrayal.21to explanation, seemslittle As there doubtthatthesub19Manyof theseissuesare raisedin the ongoingdebateover "methodological indiand in vidualism," are discussed cogently Ernest Nagel, The Structure Science, of New York,i96i, pp. 535-46. 20 Parenthetically, holders thesespecific of viewsshouldalso be lessinclined adopt to or modelin thefirst thenational sub-systemic place. 21In a reviewarticle dealingwith two of the more recentand provocative efforts towardtheory and Processin International (MortonA. Kaplan,System New Politics, York, 1957, and GeorgeLiska, International Equilibrium, Cambridge, Mass., 1957), P. adds Charles Kindleberger a further-if altogether not persuasive-argument favor in

90

THE INTERNATIONAL SYSTEM

systemic actor more or orientationconsiderably fruitful, is permitting as it doesa morethorough of theprocesses which investigation by foreign policies made. are the Herewe areenabled go beyond limitato tions imposed thesystemic andtoreplace by mere level correlation with the more significant causation. interms prediction, orientaAnd of both tions seemto offer similar a of Here theissueis a degree promise. function whatwe seekto predict. of Thusthepolicy-maker tend will toprefer predictions thewayin which about nation or y willreact x to a contemplated on his own nation's move whilethe scholar part, willprobably prefer either generalized predictions the regarding behavior a given of classof nations those or regarding system the itself. Does this summary up toan overriding forone or another add case of thetwomodels? wouldseemnot.For a staggering It of variety reasons scholar be more the may interestedonelevelthan in another at anygiven time willundoubtedly hisorientation and shift to according hisresearch needs. theproblem really oneof deciding So is not which levelis most valuable thediscipline a wholeandthen to as demanding thatit be adhered from to now untoeternity.22 it Rather, is one of realizing there this that is preliminary conceptual andthat must issue it be temporarily resolved prior anygiven to research And undertaking. it must alsobe stressed we havedealthereonlywithtwoof the that morecommon orientations, thatmanyothers available and are and perhaps evenmorefruitful potentially either thoseselected than of here.Moreover, international the system givesmanyindications of prospective change, itmaywellbe that and existing institutional forms willtakeon newcharacteristics newoneswill appear take or that to their place.As a matter fact, incapacity perform functions of if to its leadstothetransformation ordecay an institution, mayexpect of we a steady deterioration evenultimate and disappearance thenational of state a significant in theworld as actor political system. However, evenif thecaseforone or another thepossible of levels ofanalysis cannot madewithanycertainty, must be one nevertheless maintain continuing a awareness to their We mayutilize as use. one level here another and there, we cannot but afford shift orientato our in tion themidst a study. when do in fact of And we makean original
of thelower, level sub-systemic of analysis: "The totalsystem infinitely is with complex everything One can discussit intelligently, interacting. therefore, only bit by bit." "Scientific International WorldPolitics, (Octoberi958), p. 86. xi Politics," 22 It shouldalso be keptin mind thatone could conceivably developa theoretical model whichsuccessfully embraces bothof theselevelsof analysis without sacrificing conceptual and clarity internal In consistency. thiswriter's view,suchhas notbeendone to date,thoughKaplan's System and Processin International Politicsseemsto come fairly close.

THE LEVEL-OF-ANALYSIS PROBLEM

91

selection replace with or one we another appropriate at do times, must so witha fullawareness thedescriptive, and of explanatory, predictive implicationssuch of choice. A final point remains be discussed. to this Despite lengthy exegesis, onemight be prone inquire still to whether is notmerely sterile this a exercise verbal in gymnastics. itmight asked, thedifference is What, be between twolevelsof analysis the empirical the if remain referents essentially same?Or,toputit another is there difference the way, any between international relations comparative and Perforeign policy? hapsa fewillustrations illuminate subtle important will the but differences which emerge whenone'slevel analysis of One shifts. might, for example, postulate whentheinternational that is system characterized bypolitical between of itsmost conflict two is powerful there actors, a strong tendency the for system bipolarize. isa systemic-oriented to This proposition.sub-systemic A proposition, dealing with samegeneral the empirical referents, state when powerful finds would that a actor itself in political conflict another approximate with of parity, willtendto it exert pressure itsweaker on neighbors joinitscoalition. proposito Each tion, assuming is true, theoretically by itself, each is it is useful but verified a different by intellectual operation. Moreover-and is this thecrucial for thing theoretical development-one notadd these could twokinds statements of together achieve cumulative to a of growth empirical generalizations. To illustrate further, could, thesystemic postulate one at level, that whenthedistributionpower theinternational of in is system highly itis diffused, more stable than when discernible the clustering wellof defined coalitions occurs. And at thesub-systemicnational or level, thesame empirical phenomena wouldproduce sort proposition: this of whena nation's decision-makers it difficult categorize find to other nations as or readily friend foe,they tendto behave toward in a all moreuniform moderate and fashion. Now, taking these two setsof propositions, muchcumulative how usefulness wouldarisefrom atto tempting merge codify systemic and the proposition thefirst from illustration thesub-systemic with proposition thesecond, vice from or versa? different ofanalysis couched different Representing levels and in frames reference, woulddefy of they theoretical integration; may one wellbe a corollary theother, they notimmediately of but are combinable. prior A translation onelevelto another from must takeplace. This,itis submitted, is quitecrucial thetheoretical for development ofourdiscipline. all ofthecurrent With emphasis theneedfor on more and empirical data-gathering research a prerequisitetheory-buildas to

92

THE INTERNATIONAL SYSTEM

ing,onefinds concern therelationship with these little among separate and discrete data-gathering Even if we wereto declare activities. a moratorium deductive speculative on and research thenext for decade, and all ofus weretolabor in of diligently thevineyards historical and contemporary thestate international data, of relations theory would probably no more be advanced that at than is now, time it unless such empirical activity becomes moresystematic. "systematic" far And is usedhere indicate cumulative to the of growth inductive deductive and generalizations an impressive into array statements of conceptually related oneanother flowing some to and from common frame referof ence.Whatthat frame reference of should or will be, cannot be, be saidwith much certainty, it doesseemclearthat must but it exist. As longas weevade some these of a crucial priori decisions, empiricism our willamount little to more thanan ever-growing potpourri discrete, of disparate, non-comparable, isolated ofinformation and bits orextremely low-level generalizations. as such, And, they makelittle will contributionto thegrowth a theory international of of relations.