You are on page 1of 29

Three Puzzles in Search of a Researcher: Presidential Address

Author(s): Dina A. Zinnes


Source: International Studies Quarterly, Vol. 24, No. 3 (Sep., 1980), pp. 315-342
Published by: Wiley on behalf of The International Studies Association
Stable URL: http://www.jstor.org/stable/2600250
Accessed: 31-07-2015 18:12 UTC

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.

Wiley and The International Studies Association are collaborating with JSTOR to digitize, preserve and extend access to
International Studies Quarterly.

http://www.jstor.org

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
ThreePuzzles
in Searchof a Researcher
Presidential
Address

DINA A. ZINNES

Although everyoneis for"moretheory," mostofus haverather littleunderstandingof


howto get"moretheory." Thisessaysuggests one approachto theory development:
thinkingin termsofpuzzles.To developtheconceptof"puzzlement," threebodiesof
literature
are reviewed,and setsof conflicting
evidenceare presented as puzzles.The
solutions ofIR researchers.
to thepuzzlesareleftto thenextgeneration

Preface
I havebeenintrigued in recentyearsbyan apparent contradiction.
In attempting to understand international
politicsit seemsincredibly
to think
difficult intheoreticalterms.
WhenI ask"whydo warsoccur,"
I am immediately answer,an "ifthen"typeof
led to a "push-pull"
descriptive statement:"alliancesproducewars"or "thedistribution
of powercauseswars."It appearsdifficult to thinkin processterms,

EDITOR'S NOTE: Thisis theaddressdelivered


byISA President
Dina A. Zinnes
at theTwenty-First
AnnualConventionoftheInternational
StudiesAssociation
inLos
Angeles, March20, 1980.
California,
AUTHOR'S NOTE: Supportforthisresearchwasgranted
bytheDefenseAdvanced
ResearchProjectsAgency,
OfficeofArmyResearch. Contract
MDA-903-80-C-149.

INTERNATIONAL STUDIES QUARTERLY, Vol. 24 No. 3, September1980 315-342


? 1980 I S.A.
315

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
316 INTERNATIONAL STUDIES QUARTERLY

to formulate setsofassumptions aboutwhatistakingplacethatwould


makespecific hypotheses naturalconclusions.
In contrast,however, I havediscovered thatI do thinkintheoretical
termsin othercontexts.BecauseI am attempting to sella house,I
noticerealtysigns.As I cycleto schoolone day,I noticea "forsale"
realtysign.SeveraldayslaterI discovera small"sold"signsitting on
topoftherealty sign.A fewdaysthereafter I notethatthe"sold"sign
hasdisappeared buttherealty signremains. I ampuzzled.Thehouseis
apparently backon themarket.
Why?I beginto tellmyself stories.Perhapstherewasan enormous
stormthenightbeforeand thesoldpartofthesignthatsitsatopthe
realtysignwas simplyblownaway.Butifthestormhadbeenstrong,
I shoulddiscoverevidences of it elsewhere. I see no felledbranches,
tossedgarbagecans,etc. So I dismissthatstory.Perhapsthebuyers
putina bidthathadbeenaccepted; the"sold"signwentup,thebuyers
attempted togeta loanandwererejected, thesigncamedown.... What
is intriguingisthattotellthestories toourselves is relatively
easy.Why
can we notdo likewise whenobserving international phenomena?
Perhapstheanswerlies in thewaywe haveaskedthequestions.
I beginto tellmyself storiesbecausesomething doesnotworktheway
I anticipated. I haveobserved "forsale,""sold,"and"forsale"signsin
rapidsuccession.Thesecontrasting observations do not fitintomy
normalconception oftherealestatemarket, so I am pressedto make
senseoutofthem.Therearetwoingredients here:(1) myobservations
containinformation, I knowsomething whenI beginmystorytelling;
(2) I amforcedto tella storybecausetheinformation doesnotfitinto
a preconceived pattern.
Thesefactors describea puzzle:piecesofinformation, thebeliefthat
thepiecesfittogether intoa meaningful picture-buttheinability to
fitthepiecestogether Whenwe attempt
initially. to understandinter-
nationalphenomena, we typically do nothavetheseingredients. We
askbroad,open-ended questions. Wedo notusepiecesofinformation,
and we do notsee ourproblems as problems in fittingthepiecesto-
gether.Perhapsifwe thought in termsofpuzzles-notgeneralques-
tions-we wouldthinktheoretically morenaturally. Thisis themoti-
vationbehindmy'paper.
Sincetheconceptof "puzzle"withrespectto thestudyof inter-
nationalphenomena is nottotallyclear,noris theprocessbywhich
thepiecesofa potential puzzleareassembled obvious,thegoalofthis
paperwas to determine-through directexperience-whether opera-

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
Zinnes/ PRESIDENTIAL ADDRESS 317

tionalizing"puzzle" was even feasible.I pickedthreeareas thathave


long been of interestto me and triedto assembleour empiricalknowl-
edge withineach in a formthatwould suggesta puzzle. The exercise
was noteasy.Not onlywas thequantityofrelevantliterature extensive,
but, in addition,therewas a continuingworrythat the pieces being
assembledwerenotreallycomparable.Data sets,operationalmeasures,
researchdesign,and statisticaltestsweresufficiently diverseto suggest
that contrastingresultscould be the consequence of different pro-
cedures.Whilesome mayarguethatone or moreofthepuzzlesis more
methodologicalthansubstantive, at thisstagemyresponseis thateven
thisis of interest:Whyare resultsnot invariantover methodologies?
In the followingpages I presentwhatappear to be contrasting and
conflicting piecesof informationin theareas of (1) dyadicinteraction,
(2) the relationshipbetweennationalattributesand conflictbehavior,
and (3) systemicstructure and war. One consequenceof theexerciseis
theconclusionthatin each oftheseareas thereis significant evidenceof
cumulation.But anotherequallypressingresultis thattheredo appear
to be some genuinepuzzles.

"What puzzles you?" Jim Rosenau asked. We were havingdrinks


followingdinnerat the University of South Carolina's conferenceon
"The FutureoftheInternationalStudiesAssociationand International
Studies." Rosenau had made the openingpresentationfor the con-
ferenceby describingwhat he consideredthe necessaryprerequisites
for "ThinkingTheoryThoroughly,"as the titleof his paper put it.
Aftersketchingseveral avenues to theorybuildingthat he feltwere
dead ends, Rosenau (1979,p. 17) made a veryinterestingobservation:

To thinktheoretically one mustbe genuinelypuzzled by international


phenomena. . . I do not use the notionof "genuinepuzzles" casually.
Theyare not simplyopen-endedquestions,butreferratherto perplexity
over specificand patternedoutcomes.

Over drinksthatsame eveningJimwas suggesting thatgenuinepuzzle-


mentwas a surprisingly difficultgoal to reach. To demonstratehis
pointhe raisedthechallengeto me: "Tell mewhatitisthattrulypuzzles
you about internationalphenomena."
Since thateveningI have thoughtfrequently aboiit both the state-
mentthatpuzzlementis not easy to achieveand thechallenge:"What
puzzles you?" At firstblush the questionseemsincrediblynaive,and

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
318 INTERNATIONAL STUDIES QUARTERLY

one's initialreactionto thestatement "puzzlesare noteasyto come by"


is one of considerabledisbelief.Whydo nationsgo to war,whatis the
relationshipbetweenbordersand conflict,how do nationalattributes
relateto foreignpolicydecisions-one could go on and on. We would
seemto have an incrediblearrayof unanswered,importantquestions.
Butinherent in thisveryresponseliesthedifficulty. Thesearequestions,
theyare not reallypuzzles. Puzzles are questions,but not everyques-
tionis necessarily a puzzle.The issue,then,is one ofsortingthepuzzles
fromthe questions.
Or is it? The above questions about war, conflict,borders,and
foreignpolicyare of centralimportanceto the field.Researchin any
waydirectedtowardthestudyof theseissuesis of criticalsignificance.
So whybothersortingpuzzlesfromquestions?We have enoughwork
to do just studyingthe questions.Who needs puzzles?The answeris
inherentin Rosenau's "ThinkingTheoryThoroughly."Attempting to
answer the above how, why, what questions will certainlyprovide
knowledge,butit is lessclearthattheposingofthesekindsofquestions
will produce theories.Puzzlementmay also not get us there,but,as
Rosenau implies,thereis somethingabout theposingof a puzzle that
forcesone to beginthinkingin processterms.
When confrontedwith a puzzle, we are almost requiredto begin
postulatingunderlyingprocessesthat mightbringthe puzzle to reso-
lution.Like a detectiveconfronted witha murderin a roomwithdoors
locked fromthe insideand no possibleweapon withinsight,we must
begin to spin stories."Let us suppose that the murdererwas in the
room and that followingthe shooting. . ." Puzzles seem to make us
thinklike detectives, and thinkinglikedetectivesmake us look forthe
not necessarilyobvious or even visibleoperatingprinciples.
Thus the distinctionbetweena puzzle and a questionwould seem
to be of some importance.Undoubtedly,puzzles are not theonlyroad
to theory.But if puzzlementprovideseven one way of forcingus to
thinktheoretically, thenit is clearlyimportant thatwe understandhow
one becomespuzzled. Not onlywould such an understanding helpus
as researchers, it mightalso providecluesas to how we can bettertrain
students.If we understandthe dynamicsof puzzlement,thenperhaps
we can move studentsmorequicklyand surelybeyondthe inevitable
"I'm interested in X" to themoreuseful"whyis itthatA occurredhere
but not in thisrelativelysimilarcontextover here?"
My goal, then,is to accept Rosenau's challenge.Can I in factcon-
structsome genuinepuzzlesin international politics?I willmake three

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
Zinnes / PRESIDENTIAL ADDRESS 319

such attempts,tracingthe originof the puzzlementand what we, as


researchers qua detectives,currentlyknowabout theproblem.In none
of the cases do I know who the murdereris, nor will I provideany
speculationas to what the underlyingprocess mightbe. Eventually
I hope to be able to do thisforat least one or two of thepuzzles. But
the purposehereis less ambitious.First,I wantto see ifsuch a task is
possible.Second, I wantto suggesta fewplaces in theliterature where
the clues reallydo seem to come togetherin intriguing and not very
straightforward ways.Third,I wantto see whetheror notintheprocess
ofcreatingpuzzlesI can cometo termswithwhatitmeansto be puzzled:
Are thereclues that would help one distinguishbetweenpuzzles and
questions?

Do Nations Interact?
The firstpuzzle comes out of mycurrentresearchon international
crises.To understandwhyit is indeeda puzzle,it is necessaryto sketch
thebackgroundof thatstudy.One ofthegoals ofthatprojectis to find
modelsthatadequatelydescribethepatternof interactions of pairsof
nationspriorto an international crisis.Threeverysimplemodelswere
proposed.The firstpostulatedthattwonationspay no attentionto one
another,thatthebehaviorthateach directedtowardsthe otheris only
a functionofwhateach did previously. Thus,thismodelessentially says
thatnationsdo not interactin precrisisperiods.We can represent this
model witha simpleset of lineardifference equations:

x(t) - x(t-1) = alix(t-1) + g [1]

y(t) - y(t-1) = a22y(t-I) + h [2]

where
x(t) - x(t-1) representsthe changein behaviorthatX directstoward
Y; y(t)- y(t-1) represents
thechangein behaviorthatY directstoward
X; x(t-1) representsthe behaviorthatX directedat Y at (t-1); y(t-1)
representsthe behaviorthat Y directedat X at (t-1); and al and a22
indicate the impact that the previousdirectedbehaviorhas on the
changein behavior.

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
320 INTERNATIONAL STUDIES QUARTERLY

In contrast,a second model proposed that each nation'sdirected


behaviortoward the otherwas a directfunctionof the behaviorre-
ceived fromthe opponent.Thus, thismodel assumes thata nation's
behaviorin precrisisperiodsis solelya consequenceof thebehaviorof
theothernation,thatnationsact and reactas stimulus-response
mecha-
nisms.This model can also be representedin lineardifferenceequa-
tionformwherethevariablesand parametershave thesame meaning
as above:

x(t) - x(t-1) = al2y(t-l) + g [3]

y(t) - y(t-1) = a2lx(t-1) + h [4]

Thus, the change in X's behavioras directedtowardY is a function


of the behaviorthat Y directedat X at the previoustimepoint;and
likewisefornationY.
The thirdand finalmodel suggestedthatthe directedbehaviorsof
nationsin precrisisperiodsmightbe a functionboth of whata nation
did previouslyand what the opponent did previously.Combining
Models I and II, we can write:

x(t) - x(t-1) = alix(t-1) + al2y(t-l) + g [5]

y(t) - y(t-1) = a2lx(t-1) + a22y(t-l) + h [6]

Having proposed the three models, we next identifiedfourteen


internationaleventsthathave been definedby othersas "crises."The
selection of the crises was based on an extensiveconsiderationof
definitionsof "internationalcrisis"and a varietyof listsof eventsthat
various researchershave compiled of "crises" based on these defi-
nitions.The analysis of only fourteencrisesfromsome of the more
extensivelistswas dictatedby data concerns.
The variablesx(t) and y(t),thebehaviorthateach nationdirectsat
the otherduringa specifiedtime period,were operationalizedusing
WEIS coded data. As thosefamiliarwiththisdata set will recall,the
WEIS data consistof directedeventscoded by initiatornation,target
nation,date, and in termsof twenty-two different categoriesof be-
havior, twelve of which reflecthostile behaviorsand ten of which
representcooperativebehaviors.Althoughthereis an implicitscale
inherentin thesecategories-certaintypesof eventsare clearlymore

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
Zinnes / PRESIDENTIAL ADDRESS 321

hostile in nature than others-the data themselvesare not scaled.


For our purposeswe simplyconsideredthefrequency ofall hostileacts
and the frequencyof all cooperativeacts. In short,we collapsed the
twenty-two categoriesinto two basic ones: hostilityand cooperation.
Using the WEIS data, we then developed threemeasuresforthe
variablesx(t) and y(t). The firstwas a simplefrequencycount of the
numberof hostileacts thatX (or Y) directedtowardsY (or X) during
a given time period. The second measure expressedthe frequency
count as a proportionof the total numberof acts that X directed
towardsY (i.e., hostileplus cooperative).The thirdtook thedifference
betweenthefrequency of hostileacts and thefrequencyofcooperative
acts, a measureof residualor "net" hostility.
The finaldecisionconcernedtheoveralltimeframeto be examined
prior to the crisis event. We triedfourvariations:six monthsand
twelvemonthsbeforethemonthin whichthecrisiseventoccurredand
six and twelvemonthsbeforethe day on whichthe crisisoccurred.
I apologize for the many details, but it is importantto note the
varietyin the way thingswere measuredand analyzed,and to make
absolutelyclear the appropriateinterpretation of the variablesin the
model, i.e., that x(t) and y(t) are frequenciesof directedhostile/
cooperativebehaviors.Only againstthisbackgrounddoes the puzzle
reallyseem genuine.For the puzzle quite simplyis this:Model I pro-
vides verygood fitsfor manyof the crisiscases usingvarious opera-
tionalmeasuresand timesframes.Model II, on theotherhand,provides
unbelievablyand almost universallypoor fits.AlthoughModel III
providesslightly betterresultsthanModel I, indicatingthatinteraction
does in some cases accountforsome of the behaviorpriorto a crisis,
the increasein the adjusted R2 is not appreciable.Furthermore, an
examinationrevealsthatthe standarderrorof the coefficients forthe
interactiontermin Model III is typicallylarge.Thus,mostofthegood
resultsforModel III are attributable to theModel I component.(For a
more detailed presentationsee "Selection of AppropriateModels for
DescribingPrecrisisBehavior,"1979,CenterforInternationalPolicy
Studies.)
Let me more dramaticallyillustratewhat is happeningin a pre-
crisisperiod,accordingto theseresults.We have two nationssending
nastymessagesto each other,but themessagesthatX sendsto Y never
reachY. A clericalassistantintercepts each messagefromX and filesit.
The same sort of thingis happeningover in the statedepartmentof
nation X. At each subsequenttime point nationX looks at what it

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
322 INTERNATIONAL STUDIES QUARTERLY

said to nation Y last time,and this makes X even angrier;whereas


beforenationX senta note of protest,it now sendsa notecontaining
demands.Nation Y is going throughsimilarbehavior.Each nation's
own previousbehavioris makingit madder and madderat the op-
ponent.It is almostas ifeach nationwerehavingitsownprivatetemper
tantrum.Finally,each nation gets so worked up over its own past
actionsthat one of the nationstakes a "punch"at the other-and so
we have a crisis.The puzzle would seem to be the variancebetween
these resultsand the way we normallythinkabout nations before
a crisis.
One mighthave a varietyof reactionsto theseresults.Given the
factthat the contrastbetweenModels I and II is so pervasiveacross
nearlyall crises,measures,and timeframes,one could concludethat
our intuitivefeelingabout how nations behave in precrisisperiods
has been wrong-whereas we have always assumed that therewas
interaction,thereis in factno interaction.Anotherreactionis to be
suspicious of the overall researchdesign-the way in which (1) the
variablesweremeasured(e.g., theuse of unscaledfrequencies),(2) the
timeframesused (e.g., the factthatfrequencieswereaggregatedover
groupsof days), (3) the statisticalanalysiswas performed(e.g., could
theequationsbe estimatedsimultaneously ratherthanindependently?).
While I would be more inclinedtowardthe second than the firstre-
action, thereis yeta thirdpossiblereaction.Variationson Models I,
II, and III have now seen an extensiveseriesof applicationsin inter-
national relations.To what extentare these resultsconsistentwith
whatothershavefoundindifferent arenasofinternationalinteractions?
Consider the arms-raceliterature.In six relativelyrecentstudies
(since 1974), variationson Models I, II, and III have been testedfor
a largenumberofpairsofcountriesusingseveraldifferent data sources,
differentoperational measures of the variables,and different time
frames.The countrypairsthathave been examinedincludetheUnited
States and the USSR, the individualcountriesof NATO and the
Warsaw Pact, various combinationsof the Middle East countries
(Israel,Egypt,Iran,Jordan,Syria,Iraq), India and Pakistan,and Great
Britain versus various European countries during the nineteenth
century.Althoughthetimeperiodsare largelycontemporary, typically
coveringthepost 1948period,thestudyoftheBritishnavalracecovered
century.Althoughmoststudiesopera-
the last half of the nineteenth
tionalize the variables in termsof annual defenseexpenditures,one
study(Rattinger,1976) developedindicesbased on thecharacteristics

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
Zinnes / PRESIDENTIAL ADDRESS 323

of theweapons.The majorityof studiesuse theSIPRI yearbooks,but


theseare augmentedbyothersourcessuchas theStateman'sYearbook,
Institutefor StrategicStudies, and Departmentof Defense military
expenditures.We would appear, then,to have a reasonablyextensive
appraisal of Models I and II. Do these resultstell us that nations
interact?
With two exceptions,the overalland overwhelming resultis that
nationsdo not interactin thisdomain. One mightsay thatthereare
no arms races in the standardmeaningof "race." Rattinger(1975),
forexample,makesa directcomparisonbetweenModels I and II, and
whilehisversionof Model II differs fromtheone we examined,
slightly
the resultsare completelyconsistent:Model I does far betterthan
Model II. In a similarstudy,Ostrom(1977) comparesa moreelaborate
versionof Model I withan econometricversionof Model III. Both
modelsprovideverygood fits,buttheinteraction componentof Model
III is not statistically Like our results,thefitforModel III
significant.
appears to be mostlydue to the Model I component.Hollist(1977a,
1977b)providesa seriesof further testsof Model III. Like Ostrom,he
findsthat,whenthe model fitswell,it is most oftenthe case thatthe
reaction term is not statisticallysignificantwhile the Model I co-
efficientis.
It would seem, then,that the arms-raceliteratureis largelycon-
sistentwiththe crisisresults.But thereare two small fliesin the oint-
ment:Rattinger's(1976) studyof pairs of Middle East countriesand
Moll's (1974) analysisof the Britishnaval budgetduringthe last part
of the nineteenthcentury.In both of thesetwo studiesit is possible
to conclude that thereis some reactiontakingplace. The Rattinger
study differsfromall the othersbecause of his use of indices that
measure weapon capacities. The analyses are then done separately,
not onlyforvarious pairs of countries,but also withindifferent types
of weapon systemsand for different timeperiods. For certainpairs
of countries,time periods,and weapon types,it is possible to assert
thata countryreactsto its opponent:"from1956to 1967 . . . Israel is
seenas reactingto theArab nations'armaments, whereasArabreaction
occursonlyin the Israeli-Syrian missile-boatrace."On theotherhand,
"Egyptneverreactedto Israel" (Rattinger,1976: 526-527).
Moll's analysisalso differsfromtheotherstudies.All otherstudies
fitequationsto data setsand thenassessthedegreeoffitand thevalues
of the parameters.Moll turnsthe problemaround, postulatesthe
values of theparameters, and thenuses thesein a versionof Model III

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
324 INTERNATIONAL STUDIES QUARTERLY

to predictBritishnaval budgets.When comparingthe predictedwith


theactual budgets,he findsa veryclose correspondence. The difficulty
withhis resultsfor our purposes,of course,is thatwe have no clear
wayof assessingwhichcomponentof Model III is producingthegood
results.Nevertheless, the resultsat least suggestthatinteractionmay
be a factor.
The consistencybetweenresultsobtained for events in precrisis
periods and arms budgets over the postwar period would seem to
encourage more confidencein what at firstappeared to be rather
strangeresultsforthe precrisisperiods.Perhapsnationsreallydo not
interact.Yet importantdifferences exist betweenthe arms budgeting
process on the one hand and directedhostile-cooperative eventson
the other.Armsbudgetsare the resultof a seriesof complexinternal
decisions,as wellas a possibleinteractive component.It does notseem
so strange,then,to discoverthat arms budgets,like so manyother
budgetingprocesses,are primarily a functionofwhathappenedduring
the previoustimeperiod. But to discoverthatdirectedhostileacts are
also onlya functionof previousdirectedhostilitycannot be so easily
understood.Let us turn,then,to a setofanalysesthathavelookedmore
specificallyat the behavioralacts of nations.
Four studieshave examinedour questionusingconflictactsbetween
pairs of Middle Easterncountries.Unlikethearms-raceliterature that
containsspecifictestsof Model I versusModel II, only one of these
studiesexplicitlymakesthiscomparison.Thus, an overallcomparison
across the studiesbetweenModels I and II requiresthatinferences be
made about the results.In addition,whilethe studiesare unifiedby
theirfocuson the Middle East, theydiffer withrespectto data sources,
pairs ofcountriesexamined,operationalization and statisticalanalysis.
Wilkenfeldet al. (1972) examinedall possiblepairsbetweenEgypt,
Iraq, Israel,Jordan,Lebanon, and Syriafrom1949to 1967,coded the
data intothirty-one categoriesofconflictinteraction fromfourdifferent
sources,aggregatedfrequenciesby month,and used factoranalyses
to scale the events.The stepwiseregressionsuggeststhatinputfrom
othernationsis a far more importantfactorin determining directed
conflictbehaviorthanis a nation'sown pastbehavior.Milstein's(1972)
studyof fiveformsof violentacts betweenIsrael on theone hand and
each of the Arab statesof Egypt,Jordan,Iraq, Saudi Arabia, Syria,
and Lebanon on the other,between1948and 1967,usingtwo sources
and aggregating acts byweeks,showsthatnationsdo interact.Several
tests of versions of Model II provide very respectablecorrelation
values.

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
Zinnes / PRESIDENTIAL ADDRESS 325

Burrowesand Garriga-Pico(1974) also examinetheeventsbetween


Israel and the Arab countriesof Jordan,Syria,Saudi Arabia,and the
UnitedArab Republicfortwoand a halfyearsbeforetheSix Day War,
collectingdata fromten sources,and codingand scalingthedata into
eightlevels of conflictand eightlevels of cooperation.Their results
confirmthose of the firsttwo studies:thesepairs of nationscan be
shownto be reactingto one another.McCormick(1975), like Ostrom
and Rattinger,explicitlycontrastsModels I and II. Usingthe scaled
eventsfromthe COPDAB data set,aggregatedby one-weekperiods,
McCormickexaminespairsof nationsfortheSuez Crisisof 1956-1957
and the Six Day War crisisof 1966-1967.A comparisonof the beta
weightsfor Models I and II generallyshows Model II to be superior
over the fulldurationof both crises.Whenthe two crisesare broken
down into less intenseand moreintensephases, Model II is foundto
be considerablysuperiorin the intensephase, thoughthe two models
are roughlyequivalentin the less intensephase.
Thus, we seemto have,in a seriesof Middle East cases, ratherdirect
evidenceto contradictour precrisisfindings.Butthereare threestudies
notdone on the Middle East thatcontradicttheMiddle East analyses.
The mostsignificant oftheseis thestudydone by Duncan and Siverson
(1975). Duncan and Siversonpose a seriesof modelswhichtheyterm
closedand open,and whichcorrespondto Models I and II, respectively,
thoughtheyare formulatedas various formsof Markov processes.
Using a set of data collectedby Smoker(1969), in whichthe number
of communications betweenIndia and China was coded by weekfrom
1959through1964,theyshowthatwhilethedeductionsobtainedfrom
both models do not provide verygood fitsto the data, the closed
Model I versionis definitely superiorto the open Model II version.
These resultsare consistentwithan earlierstudyby Hilton(1971), in
whichMarkov analysesare performedon the coded perceptionsand
expressionsofhostility duringthe 1914crisis.The closedmodelshowed
that previousexpressionsof hostilitywere a strongfactorin deter-
miningsubsequentexpressionsof hostility.In a less directstudyof
Models I and II thatalso used Markovmodels,Zinnesand Wilkenfeld
(1971) showed,usingnondirectedconflicteventsbetweennationsfrom
1955-1960,that previousforeignconflictwas a major determinant in
the amountof subsequentforeignconflict.
The analysis of precrisisperiodssuggestsnationsdo not interact.
Extensiveanalysesof armsbudgetingseem to confirmthisresult,yet
studiesof hostilebehaviorsbetweenMiddle East countriescontradict
both the precrisisand arms-racestudies.Now we see that Markov

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
326 INTERNATIONAL STUDIES QUARTERLY

analyses of such divergentdata as Sino-Indiancommunicationsand


1914 precrisiscommunicationsseem to contradictthe Middle East
results.
We have examinedarms races,conflictinteractions in the Middle
East, and Markovprocessesdefinedoverdiplomaticexchanges.There
is one last area in whichthesemodelshave recently beentested:Soviet-
U.S. interactions duringthetestban negotiationsof 1962-1963.Using
a scaled contentanalysisof the negotiationtexts,and aggregating the
data by month,a seriesof threestudiesexaminesvariousapplications
of Models I, II, and III. In thefirstofthese,Hopmann(1972) examines
a numberof different linksbetweenthe various coded categoriesof
"attitudes,behaviorsand responses,"forthe UnitedStates and Great
Britain,each paired withthe Soviet Union. Whilethe resultsare pre-
sentedin variousgraphs,it is possible,throughcarefulattentionto the
definitionsof the coded categories,to interpret theseresultsin terms
of Model I or II. The resultsare mixed.For GreatBritain(and to some
extentthe Soviet Union), previousdirectedbehaviorspredictsub-
sequentbehavior.For theUnitedStatesand theSovietUnion,however,
a strongreaction process can be observedin termsof a varietyof
categories.These latterresultsare further confirmedin a moredirect
test of Model II in Hopmann and King (1976) and Hopmann and
Smith(1977).
So whereare we? Armsraces,directedconflict-cooperative events
priorto crises,and diplomaticexchangessuggestthatnationsin these
settingsdo not interact.Conflictbehavioramong the Middle East
countriesand statements made duringthe U.S.-Soviettest-bannegoti-
ationssuggestthatthereare some arenas in whichnationsdo interact.
Are we pickingup differences in timeperiods,typesof interactions,
operationalinterpretations of variables,different
formalizedversions
of Models I, II, and III, or differences in analysisprocedures?What
mightexplain whyarms races, precrisisperiods,and diplomaticex-
changesall fall togetherand yetare distinguishable fromMiddle-East
conflictand test-bannegotiations?

WhyAre Some Nations War Prone?


My second puzzle is a consequence of reviewingthe empirical
literatureon internationalviolence (Zinnes, forthcoming).But, in

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
Zinnes / PRESIDENTIAL ADDRESS 327

attemptingto assess whatwe did and did not knowabout whyinter-


nationalviolenceoccurs,a seriesof contrastingthemesemerges.First,
althoughit is clearlythe case thata small subsetof nationsaccounts
forthe greatestnumberof wars(Singerand Small, 1972; Richardson,
1960; Bremer,1975) in recenttimes-i.e., the numberof wars is not
evenlydistributedover the nationsin the system-it is equally clear
that thereare veryfew nations that have not engaged in any wars
(Richardson, 1960). Some nations go to war more frequently than
others,but almost everynationhas engagedin war at some time.
This suggeststhatinternational violenceis probablynot the result
of special conditionsbut rathertheconsequenceof certainattributes
of nations.If warsare theresultof nationshavingconflictsoverterri-
tory,for example, then why do such conflictsmainlyoccur among
onlya selectfew?It mustbe thecase thatan attributeofthesenations-
e.g., theirpower-makes themmoreproneto engagein territorial con-
flict.If we were to assertthatparticularconditionsalways produced
internationalviolence,we would have to explain whysome nations
are more subjectto those conditionsthan are othernations.On the
otherhand, since all nationsfindthemselvesin conflictsituationsat
some pointin time,it mustbe thecase thattheattributewhichmakes
nations conflictprone must vary over nations; some nations must
have moreof thisattributethan others.
Significantly,manystudieshave addressedpreciselythisquestion.
Surprisingly,most of them have been unable to identifysuch an
attribute.In 1964Rummelrana seriesofmultiplecorrelations between
attributesof nationsas measuredby an earlierfactoranalysisof vari-
ables measuringtechnology,demographyand size (Berry,1961),and
variousmeasuresofforeignconflict.The resultswerealmostuniversally
nil. In 1965 Haas used some of the Rummeldata togetherwithdata
obtained fromTanter (1966) and Banks and Textor (1963) to show
thatno relationship existedbetweenregimetype,levelofdevelopment,
degree of urbanization,or populationdensityon the one hand and
conflictbehavioron the other.Haas extendedtheseanalysesin 1974
by specificallycorrelatingwar frequencywitha varietyof additional
attributeslike GNP, population,level of education,agriculturalwork
force,and so forth.The resultswere still nonexistent.Meanwhile,
Rummel (1968) had provided an additional set of cross-sectional
analyses.Lookingat all nationsduringthe 1950s,Rummelfoundthat
veryfewofthe235 variablesexaminedwererelatedto 13different ways
of measuringforeign-conflict behavior.

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
328 INTERNATIONAL STUDIES QUARTERLY

Anothersetofstudiesexamineda different typeofnationalattribute,


domesticturmoilor unrest.Rummel(1963) studiedthe relationship
betweenforeignconflictbehaviorand such measuresas riots,strikes,
and demonstrations.He could find no relationship.Tanter's(1966)
replicationover anothertimeperiodproducedsimilarfindings.Using
both the Rummel and Tanter data, Zinnes and Wilkenfeld(1971)
showedthatchangesinforeign-conflict behaviorcouldnotbe attributed
to the level of domesticunrestwithincountries.Collins (1973) used
measurescomparableto thoseemployedby Rummeland Tanterin a
studyof thirty-three independentAfricanstatesfrom 1963 to 1965.
His overallconclusionwas thesame: no relationshipbetweeninternal
and externalconflictbehavior.Haas (1968) looked at ten essentially
Europeancountriesfrom1900to 1960and did notdetectrelationships
between suicide, homicide, and alcoholism rates and war aggres-
siveness.
The above studies were all cross-sectionalin scope, aggregating
variables over time periodsand, withthe exceptionof Haas (1968),
focusingon data for the 1950s and 1960s. But, if one suspectsthat
thesecircumstancescould account forthe largelynull findings,these
suspicionsare quicklyput to restby a set of longitudinalstudies.Haas
(1968) foundno relationship betweenwarfrequency and levelofindus-
trializationfor nine European countriesfrom1900 to 1960. Bremer
et al. (1973) found no relationshipbetweenpopulationdensityand
changesin populationdensityon the one hand and war participation
on the otherforEuropean nationsbetween1816and 1965. Small and
Singer (1976) show that thereis no difference in war-participation
frequencybetweendemocraciesand nondemocraciesover the time
period from 1816 to 1965. Ray (1974, 1978) examinedten countries
(Great Britain,France, Spain, Germany,Australia,Russia, Turkey,
Italy,Poland, and Rumania) from1816 to 1970 and could not detect
any relationshipbetweenvarious measuresof a nation'spowerand
diplomaticstatuson theone hand and theamountofwarinvolvement
or war initiationon the other.In anotherset of longitudinalanalyses
involvinga more complex set of indicesand hypotheses,Gochman
(1975) furthersubstantiatesthesefindings.
Thus,a largenumberofstudiesusinga varietyofattribute measures,
operationalizationsof international
violence,different nations,various
timeframes,and alternateresearchdesignsall seemto contradictour
initialbeliefthatthereare attributesof nationsthatmakethemmore
war prone.But thepictureis not yetcomplete.Thereis yetanotherset
of studiesthat would seem to confirmour initialsuspicionand clash
withthe resultsalreadyseen.

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
Zinnes / PRESIDENTIAL ADDRESS 329

One of the strongestsets of positiveresultsoccursbetweendefense


expendituresand the likelihood that a nation will engage in inter-
nationalviolence.Althoughnearlyall ofRummel's(1968) 235 attributes
did not correlatewithforeignconflictbehavior,a seriesof indicators
involvingdefenseexpendituresdid producesome of thelargestcorre-
lations (range of .35). Naroll's (1969) studyof 2000 yearsof history
suggestsa strongrelationshipbetweenarmamentand war. Weede
(1970) foundreasonablecorrelations betweenvariousindicesofdefense
and foreignconflict.Stuckeyand Singer(1973) showedthata military-
industrialindexwas relatedto a nation'swarbehavior.Sylvan's(1976)
analysis of 15 Asian countriesfrom1956-1970showed that military
assistanceto thesecountriesincreasedthe likelihoodthatthesestates
would be involved in internationalviolence. Newcombe and Wert
(1973) founda relationshipbetweena "tension"indexbased on expected
defenseexpendituresand war frequency.Choucri and North(1975)
found relationshipsbetweendefenseexpendituresand violence for
most of the 6 countriesexaminedfrom1871to 1914. Wallace (1977)
showed that nationsinvolvedin arms races have considerablyhigher
prospectsof becominginvolvedin a war thando nationsnot engaged
in such competition.
The defensepostureof a countrywould thusappear to have a very
importantimpacton its war behavior.Such a resultis surprisingnot
simplybecause it contradictsthe usual notionsof deterrencetheory,
namely,thatthe best war-prevention measureis a strongdefense,but
also because it does not seem consistentwith the previouslycited
studiesofno relationship betweenattributes and international
violence.
The defensepostureof a countrymustbe relatedto otherattributes
of thatcountry.A nationcannotunrestrainedly pump moneyintoits
defensesector.Such decisionsare constrainedbyresources.Ifthelevel
ofdefenseis closelyrelatedto international violence,one wouldantici-
pate thatvariousmeasuresoftheresourcebase, GNP, levelofdevelop-
ment,urbanization,and thelikewould also be relatedto international
violence.Given the multitudeof studiesand the varietyof measures
investigated,one would suspect that at least some resource-base
measurewould relatewell to violentbehavior.The factthatnone do
and thatnumerousstudiesof defensepostureand violenceshow posi-
tiverelationshipsposes a most intriguing contrast.
But thisis onlythe beginningof the puzzle. Anotherset of studies
has uncovered an additional correlate of internationalviolence.
Although,as noted earlier,Rummel (1963) and Tanter (1966) were
unable to show any relationshipbetweendomesticunrestand inter-
national violence behavior, both studies did uncover very strong

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
330 INTERNATIONAL STUDIES QUARTERLY

relationships,between variousformsof international violence.Thus,


nationsthatthreaten andsendprotests arealso verylikelyto mobilize
troopsand engagein wars.Thesefindings are supported by Zinnes
and Wilkenfeld (1971),whofoundthatpreviousforeign conflictbe-
haviorwas a maindeterminant ofsubsequent violentbehavior.
At firstglance,sucha resultdoesnotseemunreasonable. Nations
thatengageinoneformofviolencecouldeasilymoveto another form
ofviolence. Butifoneformofviolencebegetsanother formofviolence,
howdoestheprocessbegin?Whatsetsthehostility in motion? As we
havealreadyseen,wecannotconcludethatsomeattribute triggersthe
hostility.Norcanweconclude thatsomenationsareinherently hostile.
Thiswouldbe consistent withtheinitialobservation thata fewnations
accountformanywars,butit is notconsistent withtheotherhalfof
thatobservation thatnearlyall nationsengagedin warat somepoint
intime.Ifsomenationsarealwayshostile, andtheirprevious hostility
feedstheirsubsequent hostility,howdo we accountforthosenations
thatonlyengagein a fewwars?
A thirdsetofstudieswouldseeminitially to offer atleastthebegin-
ningof a solutionto thepuzzle.Richardson (1960)showedthatthe
number ofborders correlatedwellwiththenumber ofwars.Midlarsky
(1975) further confirmed thisfinding whenhe correlated thelog of
thenumberofbordersforeachofthenonislandcentralpowerswith
thenumber ofwars.StarrandMost(1976)extended thedatacollection
on borders, payingspecialattention to thedefinition andclassification
ofinternational borders.Theirresultsconfirmed thoseofRichardson
and Midlarsky, thougha subsequentstudy(1977) showedthatthe
resultsareaffected bytimeperiodsandtypesofborders. Weede(1970)
discovered thatcontiguous dyadsweremostlikely toengageinviolence.
Thisresultwas confirmed byGarnham (1976a).
Bordersarean attribute, bordersproducecontact, contact generates
conflict, conflictleads to international violence.It wouldseemthat
thepuzzleis on itswaytowarda solution.The difficulty is thatthis
explanationcan onlyaccountforsomesubsetof nations.England
has no borders, yetitengagedinthegreatest number ofwarsbetween
1816and 1965.Switzerland, on theotherhand,haslotsofborders but
accountsforveryfewwars.In part,thissaysthatthereare deviant
casesandthattheborder-war relationship doesnotholdperfectly. But
it also suggeststhattheremustbe differences betweennationswith
respectto theirviolentbehavior.Yetwe havebeenunableto findan
attribute thataccounts forthedifference between EnglandandSwitzer-
landwithrespectto international violence.

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
Zinnes/ PRESIDENTIAL ADDRESS 331

Is Polarizationa Precondition
for War?
Mythirdpuzzleis intimatelyrelatedtothesecondpuzzle.It,too,is
concerned largelywiththeissueofinternational conflict andtheout-
breakof violence.Butthethirdpuzzlelooksat thisquestionfroma
differentperspective:fromthe more"bird's-eye" viewof theinter-
nationalsystem. Thepuzzlement arosebecauseI thought I hadsuccess-
fullysolvedone puzzle-onlyto discoverthatthesolutionwas not
consistentwiththeevidence.
I have long been intrigued by the relationship betweensystem
structureand thepropensity forinternationalviolence.Steepedinthe
mystiqueof the balanceof powerliterature, I have alwaysfeltit
reasonableto arguethatdifferent system structures couldinhibitor
stimulateinternationalconflict.
Although I was neverconvinced that
balanceofpowerisnecessarily an optimalsystem forrestraining war,it
hasappearedintuitively plausibleto methatvariousconfigurations of
alliances,distributionsof power,and numbersof nationsin an
international systemhave a directbearingon the likelihoodthat
wouldoccurandthatwarsmight
conflicts result.
Thustheproblem was
notto determine whether sucha relationshipexisted,butto findthe
appropriate combination ofsystem structureanddynamics thatwould
permit one to concludesomething aboutconflict.
Thedebatebetween Kaplan(1971)andWaltz(1964)concerning the
relativemeritsofbipolarandmultipolar systemshasbeenofparticular
Kaplanputsforth
interest. thestandard balanceofpowerposition that
assertsthatmultipolarsystems are lessconducive to conflictand war
thanarebipolarones.Waltzproposes equallyplausible reasonswhythe
reversemightbe true.Thus,Kaplan(1971:94-95)argues:

Clearlytwoessentialstates. . . wouldbe unstable... anymomentary


advantagewouldpermit onetoconquer....Evena system ofthreewould
beunstable....Asthenumber ofstatesisincreased
therecomesa pointat
whichtheprobability of a futurecoalitionthatwouldguardagainst
dismemberment becomes sufficiently
high....Maximizingthegainsfrom
wartheninvolves risksthataretoo great. . .

Consequently, Kaplanconcludes,whenthereare a numberof large


equal-sizedstates,theriskofwaris great,andthisveryfactkeepsthe
system peaceful.
In contrast,Waltzobservesthatthe"bipolarworld. . . hasshowna
remarkable andgoesontodemonstrate
stability" whytheconventional
wisdomthat"a worldof manypowersis morestablethana bipolar

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
332 INTERNATIONALSTUDIES QUARTERLY

world" is wrong. The argumentcontains the followingingredients


(Waltz, 1964: 882-884):

(1) Withonlytwoworldpowers therearenoperipheries.... Anyeventelicits


theinterestoftheother."
(2) "Withthecompetition bothserious
andintense,a losstoonecouldeasily
appearas a gainto theother."
(3) "Therangeoffactors includedinthecompetition is extendedas thein-
ofthecompetition
tensity increases.
Increasedintensity isexpressedina
reluctanceto acceptsmallterritorial
losses."
(4) "Thethirddistinguishingfactorinthebipolarbalance.. . is thenearly
constantpressure andtherecurrence ofcrises.... Rathera largecrisis
nowthana smallwarlater.""In a bipolarworld... thereis onegreat
. . . theknowledge
certainty ofwhowillopposewhom."

In short,Waltz proposes, bipolar worlds are unambiguousworlds


wherethe opponentis clearlyidentifiedand where,furthermore, any
gain by one side is automaticallya loss for the other;i.e., it is by
definitiona zero-sumsituation.The lack ofambiguityand thezero-sum
settingforcebothsidesto counteranymovebytheother.Thisproduces
crises,butthecriseslessentheprospectsforwar. Whereas(Waltz 1964:
884),

...crisesalso occurin a multipolar


world,butthedangersare diffused,
responsibilitiesunclear,and definition easilyobscured.
ofvitalinterests
... Whenpossibleenemies areseveral... unity to
ofaction.. . isdifficult
secure. . .

And so, the implicationis thatmultipolarworldsare less stable.


Both argumentsare convincing,yetit is difficult to see how both
could be correct.Whilebotharguments couldbe wrong,i.e.,theremight
well be no relationshipwhateverbetweenpolarityon theone hand and
on the other,it is seeminglyimpossibleforbipolar
conflict/stability
worldsto be bothunstable,a la Kaplan, and stable,a la Waltz,nordoes
it seem reasonable to assertthat multipolarworlds are stable, a la
Kaplan, but unstableaccordingto Waltz.
Bueno de Mesquita(1978) presentsan interesting that
interpretation
wouldappearto resolvethiscontradiction. He suggeststhatthebipolar
and multipolarpositionsshare a commonassumption,namely,"they
agree that multipolarstructuresproduce more uncertainty than do
bipolar structures" betweenthe two positions
(p. 244). The difference

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
Zinnes / PRESIDENTIAL ADDRESS 333

liesin theconsequences thateachderivesfromthisbasicassumption.


The Kaplanpositionassertsthat"uncertainty encourages caution"on
thepartofthedecisionmakers, whiletheWaltzposition concludes that
uncertainty makesdecisionmakersmorewillingto tryriskymoves.
Thus, Bueno de Mesquitasuggeststhatthe multipolar argument
containsan implicitassumptionthat decisionmakersare "risk-
avoidant," whilethebipolarargument containstheimplicit assumption
thatdecisionmakersare "risktakers."
Thisresolution workswithin thecontext ofmultipolar systems, but
howdoes theassumption aboutdifferences between decisionmakers
helpus in understanding a bipolarworld?To maintainconsistency,
Waltzwouldhaveto saythatdecisionmakersremainrisktakers, but
thatinbipolarworldsthereareno riskstobetaken.So farsogood.But
whatwouldKaplanargue?Assuming thatKaplanagreesthatbipolar
systems areunambiguous, thentheassumption thatdecision makers are
notrisktakerswouldhavetoleadtotheconclusion thatbipolarsystems
arestable.Ifbipolarsystems containno riskand,inaddition, decision
makersdo not like to take risks,thenthe conclusionwouldseem
inescapable thatbipolarsystems mustbe highly stable.Indeed,evenif
Kaplan were not to agree with Waltz that bipolarsystemsare
unambiguous, ifweassumethatdecision makers arerisk-avoidant, then
whywouldtheynotremainrisk-avoidant in bipolarsystems?
A somewhat different
resolutiontothetwoarguments ispresented by
delvingintothenatureoftheuncertainty and ambiguity in thesetwo
systems. A basicassumption behindKaplan'smultipolar qua balanceof
powerworldisthatnationsarecontinually attempting toincreasetheir
powerposition.Equallyimportant is theassumption thatall nations
mustcontinually attempt to thwartan aggressive movebyanyone
nation.Note thesimilarity betweenthesepointsand thestatements
madebyWaltz.Eachnation istryingtogetmoreforitself, buteachknows
thatitmustnotletothers besuccessful inthissameattempt-asituation
notdissimilar to Waltz'snotionofzero-sumness inbipolarity settings.
Thus,bothpositions wouldprobably agreeonthenature ofthedynamic
within internationalsystems.
Thereis, however, a significantdifference in theoperationofthe
dynamic intwo-versusn-nation systems. Assuming thatsuccessinwaris
a directfunctionofa nation'spower, andthatnations arefullyinformed
aboutthepowerpositions ofothersand arerational, thenwarsshould
onlyoccurintwo-nation systems whenoneofthenations becomesmore
powerful thantheother.Twoconclusions canbedrawnfrom this.First,

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
334 INTERNATIONAL STUDIES QUARTERLY

if nationsstayequal in power,no war willoccur.This is preciselythe


conclusion drawn by Waltz. Second, however,the power balance
betweenthe two nationsis criticaland tenuous.Any sudden (even a
slight)surgeof poweron thepartofone sidehas devastating for
effects
theother.Consequently, as longas thetwonationsmaintainthebalance
betweenthem,warswillnot occur,but thesystemcontainsno cushion
factorshouldone ofthenationsgeta littleextrapower(a technological
breakthrough, forexample).In thissense,then,thesystemis "unstable,"
as balance of powertheoristsargue.Pollard(1923: 62-63)forexample,
providesthe followinganalysis:

therewouldbe fiveGreatPowers,actingindependently as a rulebut


combining wheneverthe need arose to suppressany claimantto
supremacy balancethere
overtherest.... Inthismultiple was... noneed
ofstrict equalityand no necessarydisturbanceifonestategrewstronger
... so longas thegrowthwasnotso great theunited
as tothreaten strength.
. . . Thebalancewasnotthena delicateone... butgradually. . . thismul-
tiplepassedintoa simplebalanceof power. . . in a simplebalancethe
twoscalesmustbe even... themoreperfect thebalancethemoreeasilyit
is upset.

For Waltz, stabilitymeans thatas long as the two nationsare equal,


therewill be no war; forbalance of powertheorists, means
instability
that such an even distributionof power is extremelydifficultto
maintain.
The argumentbecomescleareras we movefroma twoto an n-nation
system.In a two-nationsystem,whenA attacksB thereis no way in
whichB can increaseits power. In a threenationsystem,thereis the
possibilitythattheattackednationB could gethelpfroma thirdnation
C. Nation C becomes the cushionfactor,and its existencemakesan
attackbyA on B lesslikelyto succeed.Ifthereare onlytwostatesinthe
system,thenA's uncertainty is onlywithrespectto whetheritspoweris
sufficientto overcomeB. Whenwe add nationsto thesystem, we add an
additionalriskto A's 'calculations,namely,whetheror not any of the
other states will combine forceswith B. In a multipolarsystem,a
potentialattackerdoes not knowwho, or how powerful, theopponent
will be. For this reason,the multipolartheoristarguesthatpotential
attackerswill be deterred.
Thus,thedifference betweenthetwo argumentsliesin theprobabil-
itiesthatare beingcompared.Thereare essentiallyfourprobabilities
inherent intheseanalyses.The firstis theprobability
thatone nationwill

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
Zinnes / PRESIDENTIAL ADDRESS 335

attackanother whenthereareonlytwonationsinthesystem andthey


areequalinpower.Callthisprobability pi.Letusassumethatpi= 0. The
secondprobability ofonlytwoequalmajorpowerswill
is thata system
maintain thisdistribution ofpower.Callthisprobability
p2,andassume
thatitis positive.
A thirdprobability,p3, is theprobability
that,ifthere
areonlytwonationsinthesystem, andtheyareunequalinpower,then
themorepowerful nationwillattack.Assumethisprobability iscloseto
p4, is theprobability
1. The fourthprobability, thatin a systemofthree
ormorenationsonenationwillattackanother.
Themultipolar
theorists
focuson probabilitiesp2,p3, and p4. The beliefthatp2 iS positiveleads
ofpowerwillbecomeunequal.This
thatthedistribution
themtosuspect
leads themto comparep3 and p4. Althoughp4 is positive,theybelieve
facingan attacking
thattheriskysituation nationgivenan n-nation
systemmakes p4 smallerthan p3. The bipolarargument,on the other
hand,makesa verydifferent comparison. The comparisonhereis
betweenpi and p4. Sincepi is zero,thebipolartheorist necessarily
concludesthatmultipolar systems have a greater likelihoodof war.
Thus,Waltzis comparing thelikelihoodof warbetweentwomajor
powers, undertheassumption thatthetworemain roughly equal,with
thelikelihood ofwaramongn equal nations.
Ifthisanalysisis correct,it suggeststhatwecannotreallycompare
bipolarsystems with multipolar oneswithout beingclearas tohowthe
probabilityofsystem changeenters thepicture. It impliesfurther that
bipolarand multipolar theoristsmight wellagree on therelationshipin
multipolarsystems between the number ofpoles and the likelihoodof
war.The argument implicit in themultipolar argument indicatesthat
theprobabilityforwar should declineas thenumber ofpolesincreases;
as we add nationsto the system,we increaseA's riskof being
unsuccessfulinan attackagainstB. Whilethismaynotprevent war,it
shoulddecreaseitsfrequency.
The puzzlecomesin examining thefewempirical studiesthathave
the
analyzed relationship between polarityand international violence.
Inthe most relevant Haas
analysis, (1973,1974) studied 21 international
systems spanning 1649through 1898.Within eachofthesesystems, he
thenumber
identified ofmajorpowers/ polesusingcriteria developed by
Rosecrance (1963).Thisnumber wasthencorrelated withthenumber of
warsoccurring withinthatsubsystem. The correlation was .58,indi-
catingthatas thenumber ofpolesincreased,so didthefrequency ofwar.
Thisresultcouldbe an artifact ofsystem duration, sincesomesystems
lastedlessthantenyearswhileotherscoveredmorethanfifty years.

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
336 INTERNATIONAL STUDIES QUARTERLY

Clearly,thelongera system thegreater


persists, itschancesforwar.
theresultis discouraging.
Nevertheless,
Even moresurprising is the finding of anotherstudy.Buenode
Mesquita(1975,1978)determined thenumber ofpolesperyearforthe
totalinternationalsystem from1816to 1965through anexamination of
theextentto whichnationssharealliancepartners, a procedure for
thenumber
identifying ofpolesthatistotally different
from thatusedin
theHaas study.Buenode Mesquitafindsno relationship ineither the
nineteenthor twentieth century betweenthismeasureof numberof
polesand theexistence or nonexistence ofwar.
Two otherstudieshavelookedat therelationship between polarity
andwar,butfromsomewhat differentvantage points.Singer andSmall
(1968)developedan indexof polarity bydetermining thenumberof
dyadicrelationships possibleina system composedofn-nations. They
thendetermined how manyof thoserelationships in factexistedor
couldnotexist,basedonthetargets implicit inexisting alliances.Their
interest
wasintheextent towhichallpossiblelinkswere"usedup"either
byexistingtiesorimpossibilities. Ifalltieswereusedup,theyconcluded
thatthesystem was completely polarized.The extentto whichsome
linkswerenotusedupwasa measure ofhowdeviant thesystem waswith
respecttobipolarization. Thus,theindependent variable inthisinstance
is notthenumber ofpoles,buttheextent towhicha system is oris not
polarized.Nevertheless, the correlation betweenthisindexand the
number ofwarsbeginning inthefollowing threeorfive-year periods was
nonexistent.
essentially
Thelastoftherelevant studieswasdonebyWallace(1973).LikeSinger
and Small,andunlikeHaas or Buenode Mesquita, Wallace'sconcern is
withthe degreeto whicha systemdeviatesfrompolarity.Usinga
complexprocedure fordetermining whatis essentially thenumberof
poles,Wallacedevelopsan indexthatmeasures theextent towhicheach
nationdeviatesfromthesegroupings. Wallace'sdependent variableis
notthefrequency ofwar,but"amountofwar"as measured bythetime
eachnationspendsinwarandthetotalbattledeathsofeachwar.Using
whathe termstheunweighted polarity index,thecorrelations arenot
unlikethosefoundby Bueno de Mesquitaand Singerand Small,
namely,nil. Usinga weighted indexthattakesintoaccountwhich
nationsarethestrong deviants,thecorrelations increase, goingas high,
forsomeperiods, as .46.Wallacereports thata fewofthesecorrelations
are significant
at the.05 level.

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
Zinnes / PRESIDENTIAL ADDRESS 337

The surprisingfeature,however,is that these correlationsare


negative.Thus, forcertainperiodsthereappears to be a relationship
betweendeviationsfrompolarityand amountofwarin thesystem,but
theresultsindicatethatthegreaterthedeviationfrompolarity,thatis,
the less polarizedthesystem,theless war thereis in thesystem.While
increasingdeviationsfrompolaritycannotbe equated withincreasing
multipolarity, one wouldneverthelesssuspectfromtheHaas studythat
as a systembecomes more polarizedit should have morewar. Thus,
while the Haas studyprovidesresultsthatcompletelycontradictour
argument,the Wallace analysis providessome meager evidence in
supportof the argument.
The difficulty,then,is twofold.Whilewe seemto have resolvedthe
analyticconflict,theconclusionswarrantonlyverymeagerand indirect
supportfromthe work of Wallace. In addition,thereare obvious
conflictsamong the empiricalstudies. Haas suggeststhat as poles
increase,so do wars.Buenode Mesquitasaysthatno relationship exists.
Singerand Small showthatdeviationsfrombipolarity haveno effect
on
thefrequency ofwar.Yet Wallace indicatesthat,usinga weightedindex
ofdegreeofpolarity,thegreaterthedeviationfrompolaritythelesswar
thereis in the system.

So Whatis Puzzlement?
If I have been successfulin sketchingthreedifferentpuzzles,then
whatis itthattheyhave in common?Whatis theprocessofpuzzlement?
In each case, puzzlementseemsto occurbecausesomethingdoes notfit
withsomethingelse.To discoverthattwonationsare directing hostility
at one another,butnotlisteningto whateach has said,does notseemto
correspondwith what we read in the newspapersand the way we
normallythinkabout theactivitiesofnations.The surpriseis enhanced
whenwe notethatthisoccursjust beforetheoutbreakofa majorcrisis.
The puzzle becomes more complex whenwe thenfindthatthereare
otherarenasin whichnationsseemnotto interact. Yet forcertaintypes
of nationsand certainkindsof activities,one can show thatthereare
situationsin whichinteractiondoes take place. Whyis thisthe case?
Whydoes interactiontake place in some instancesand not others?
Ifsomenationsgo to warmorefrequently thanothers,itsuggeststhat
thereis someattribute unequallysharedamongnationsthatenhancesa

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
338 INTERNATIONAL STUDIES QUARTERLY

nation'swar proneness.Yet we can not findsuch an attribute.Or ifwe


do (borders),itonlyaccountsforsomecases. The problemseemsfurther
complicatedby the discoverythat a strongdefenseposturemakes a
nationmorewar prone.Buthow can thisbe thecase whentheresource
base thatwould determinedefensespendingbears no relationshipto a
nation'swar proneness?
Assumingthatwe have resolvedthedebatebetweenthetwopolarity
arguments, facedbythestrangeempirical
we are nevertheless The
results.
"combined"theorypredictsthatin movingfroma threeto an n-polar
system,one oughtto discovera decreaseintheamountofwar.Butonly
one analysiseven slightly supportsthisconclusion.Two othersfindno
relationship, and one findsthe reverserelationship.
Clearly,then,puzzlesrequireinformation. Thiswouldseemto be the
major difference betweena puzzle and a simplequestion.To be truly
puzzled, we must already know something.We can be extremely
uninformed and yetask a question.Butpuzzlementrequiresmorethan
just pieces of information.Puzzlementarises because thingsdo not
workthewaywe anticipated.We believenationsdo interact, yeta rather
extensiveset of data and sophisticatedstatisticalanalysessay theydo
not. Thus puzzlementarises because thereis a clash of evidence.This
would seem to be the key to how puzzlementarisesand whyit is so
important.
If to become trulypuzzled we need to know something,thenit is
obviousthatwe mustexpandour researchhorizonsto look beyondthe
confinesof specificresearchquestions.We needto becomemoreaware
of how our researchfitswiththeworkofothers.And we needto do this
withoutbeing constrainedby subjectmatteror methodology.When
asking whethernations interact,we must be ready to see this as a
problemin manydifferent arenas,regardlessofwhether theanalysisis a
multipleregressionor Markov process.This meansthatwe should be
better"intellectualarcheologists," throughtheremainsofstudies
sifting
past forclues that will help us to understandbroaderquestions.To
become puzzled,then,requiresthatwe searchforrelationships across
studies,thatwe broadenthe scope of a problembeinganalyzed.
Butifthisis how one becornespuzzled,itis obviousthatconsiderable
work is required.Is it worththe effort?Why mightwe expect that
puzzlementis one possible road to theorybuilding?The clash of
evidencerequiresus to pose a processthatwould permitus to account
forthedifference. To explainwhynationsinteractin one arenaand not
anotherrequiresthatweconsidertheinherent processofinteractionand

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
Zinnes / PRESIDENTIAL ADDRESS 339

whythe arenas differ.The test-bannegotiationsrequireface-to-face


interaction, a settingthatis certainlydifferent
fromtwo bureaucracies
attemptingto determinedefensebudgetlevels.The Middle East has
been a scene of continualviolence.Perhaps it requiresthisdegreeof
intensityof interactionbefore nations really pay attentionto one
another.Many precrisisperiodsdo not have thischaracteristic.
It is thistypeofreasoningprocess,thesortingand sifting ofevidence,
that would seem to be most valuable with regardto puzzlement.
Withoutpuzzlement,the questionsbecome too open-ended:whydo
warsoccur;whendo nationsinteract? Withpuzzlement, wecan beginto
narrowthescope and use cluesto helpus putpiecestogether. Thereis no
guaranteethatpuzzlementwilllead us to theory.The difference between
greatdetectivesand poor ones lies ultimately in theabilityto makethe
creativeleap fromtheevidenceto thefullpicture.Butsurely,assembling
as manycluesas possiblein as coherenta way as possibleprovidesthe
best possiblebase fromwhichto make such leaps.

REFERENCES

BANKS, A. and R. B. TEXTOR (1963) A Cross-PolitySurvey.Cambridge,MA: MIT


Press.
BERRY, B.J.L. (1961) "Basic Patternsof Economic Development,"pp. 110-119inNor-
tonGinsbury(ed.) Atlas ofEconomicDevelopment.Chicago:Univ.ofChicago Press.
BREMER, S. A., J. D. SINGER, and U. LUTERBACHER (1973) "The population
densityand war pronenessof Europeannations,1816-1965."ComparativePol. Studies
6 (October): 329-348.
BUENO de MESQUITA, B. (1978) "Systemicpolarizationand theoccurrenceand dura-
tion of war." J. of ConflictResolution22 (June):241-268.
---(1975) "Measuringsystemicpolarity."J. of ConflictResolution19 (June): 187-
216.
BURROWES, R. and J. GARRIGA-PICO (1974) "The road to the Six Day War: rela-
tionalanalysisof conflictand cooperation."Peace Sci. Society(International)Papers
22: 47-74.
CHOUCRI, N. and R. C. NORTH (1975) Nationsin Conflict:Population,Expansion
and War. San Francisco:Freeman.
COLLINS, J. N. (1973) "Foreignconflictbehaviorand domesticdisordersin Africa,"
pp. 251-293in J. Wilkenfeld(ed.) Conflictand Linkage Politics.New York: David
McKay.
DUNCAN, G. T. and R. M. SIVERSON (1975) "Markov modelsforconflictanalysis:
resultsfromSino-Indian relations,1959-1974."Int. Studies Q. 19 (September):
344-374.

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
340 INTERNATIONAL STUDIES QUARTERLY

GARNHAM, D. (1976a) "Dyadic international war, 1816-1965:theroleofpowerparity


and geographicalproximity." WesternPol. Q. 29: 231-242.
--- (1976b) "Powerand lethalinternationalviolence,1969-1973."J.ofConflictReso-
lution20 (September):379-394.
GOCHMAN, C. S. (1975) "States,powerand interstate conflict:themajorpowers,1820-
1970."Deliveredat themeetingoftheInternational StudiesAssociation,Washington,
DC.
HAAS, M. (1974) InternationalConflict.Indianapolis:Bobbs-Merrill.
--- (1973) "Societal developmentand international conflict,"pp. 191-199in J. Wil-
kenfeld(ed.) ConflictBehaviorand Linkage Politics.New York: David McKay.
--- (1970) "Internationalsubsystems:stabilityand polarity."Amer. Pol. Sci. Rev.
64 (March): 98-123.
---(1968) "Social changeand nationalaggressiveness 1900-1960,"pp. 215-244inJ.D.
Singer(ed.) QuantitativeInternationalPolitics.New York: Macmillan.
--- (1965) "Societal approachesto thestudyof war." J. of Peace Research2, 4: 307-
323.
HILTON, G. (1971) "Expressionsof hostility in crisis."J. of Peace Research8, 3-4:249-
262.
HOLLIST, W. (1977a) "Alternativeexplanationsof competitivearms processes:tests
on fourpairs of nations."Amer.J. of Pol. Sci. 21 (May): 315-340.
--- (1977b) "An analysisofarmsprocessesintheUnitedStatesand theSovietUnion."
Int.StudiesQ. 21 (September):
503-528.
HOPMANN, P. T. (1972) "Internaland externalinfluenceson bargainingin armscon-
trolnegotiations:the partialwar testban," pp. 313-337in B. M. Russett(ed.) Peace,
War and Numbers.BeverlyHills: Sage.
--- and T. KING (1976) "Interactionsand perceptionsin thetestban negotiations."
Int. Studies Q. 20 (January):105-142.
HOPMANN, P. T. and T. C. SMITH (1977) "An applicationof a Richardsonprocess
model: Soviet-Americaninteractionsin the testban negotiations1962-1963."J. of
ConflictResolution21 (December): 701-726.
KAPLAN, M. A. (1971) "Intervention in internalwar: some systemicsources,"pp. 275-
303 in J. N. Rosenau (ed.) The ScientificStudyof ForeignPolicy.New York: Mac-
millan.
McCORMICK, J. M. (1975) "Evaluatingmodelsofcrisisbehavior:someevidencefrom
the Middle East." Int. Studies Q. 19 (January):17-45.
MIDLARSKY, M. (1975) On War. New York: Macmillan.
MILSTEIN, J. S. (1972) "Americanand Soviet influence,balance of powerand Arab-
Israeliviolence,"pp. 139-166in B. M. Russett(ed.) Peace, Warand Numbers.Beverly
Hills, CA: Sage.
MOLL, K. (1974) "International conflictas a decisionsystem."J. ofConflictResolution
28 (December): 555-577.
NAROLL, R. (1969) "Deterrencein history," pp. 150-164inD. G. Pruittand R. C. Snyder
(eds.) Theoryand Researchon the Causes of War. EnglewoodCliffs,NJ: Prentice-
Hall.
NEWCOMBE, A. G. (1969) "Initiativesand responsesin foreignpolicy."Peace Research
Reviews3 (June).

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
Zinnes / PRESIDENTIAL ADDRESS 341

--- and J. WERT (1973) "The use ofan inter-nation tensiometer forthepredictionof
war." Peace Sci. Society(International)Papers 21: 73-83.
OSTROM, C. W., Jr. (1977) "Evaluatingalternativeforeignpolicy decision-making
models:an empiricaltestbetweenan armsrace modeland an organizationalpolitics
model." J. of ConflictResolution21 (June):235-266.
POLLARD, A. F. (1923) "The balanceofpower."J.oftheBritishInstituteofInt.Affairs
2 (March): 53-64.
RATTINGER, H. (1976) "Fromwarto warto war."Int. StudiesQ. 20 (December):501-
531.
--- (1975) "Armaments,detenteand bureaucracy:the case of the armsrace in Eu-
rope." J. of ConflictResolution 19 (December): 571-595.
RAY, J. L. (1978) "Status inconsistency and 'aggressive'war involvementin Europe,
1816-1970."Presentedat the InternationalScience Associationmeeting,February
1978.
--- (1974) "Statusinconsistency and war involvement in Europe, 1816-1970."Peace
Sci. Society(International)Papers 23: 69-80.
RICHARDSON, L. F. (1960) Statisticsof Deadly Quarrels.Pittsburgh, PA: Boxwood.
ROSECRANCE, R. (1963) Actionand ReactioninWorldPolitics:International Systems
in Perspective.Boston: Little,Brown.
ROSENAU, J. (1979) "Thinkingtheorythoroughly." Presentedas thekeynoteaddressat
the seminaron InternationalRelationsTheoryorganizedby the Centreof Interna-
tionalPoliticsand Organizations,School of International Studies,JawaharlalNehru
University, New Delhi, India, May 14-17.
RUMMEL, R. J. (1968) "Nationalattributes and foreignconflictbehavior,"pp. 187-214
in J. D. Singer(ed.) QuantitativeInternationalPolitics.New York: Macmillan.
--- (1964) "Testingsome possiblepredictorsofconflictbehaviorwithinand between
nations."Peace ResearchSociety(International)Papers 1: 79-111.
--- (1963) "Dimensionsof conflictbehaviorwithinand betweennations."General
SystemsYearbook 8: 1-50.
SINGER, J. D. and M. SMALL (1972)The WagesofWar, 1816-1965: A StatisticalHand-
book. New York: JohnWiley.
---(1968) "Allianceaggregationand theonsetofwar,1815-1945," pp. 245-286inJ.D.
Singer(ed.) QuantitativeInternationalPolitics.New York: Macmillan.
SMALL, M. and J. D. SINGER (1976) "The war pronenessof democraticregimes."Je-
rusalemJ. of Int. Relations 1: 49-69.
SMOKER, P. (1969) "A time seriesanalysisof Sino-Indianrelations."J. of Conflict
Resolution13 (June): 172-191.
STARR, H. and B. A. MOST (1977) "A return journey:Richardson,'frontiers,'and wars
in the 1964-1965era." (mimeo)
--- (1976) "The substanceand studyofbordersininternational relations."Int.Studies
Q. 20 (December): 581-620.
STUCKEY, J. and J. D. SINGER (1973) "The powerfuland the war-prone:ranking
thenationsbyrelativecapabilityand warexperience,1820-1964."Presentedat Mexico
CityConferenceon Poder Social: AmericaLatina en el Mundo, May.
SYLVAN, D. A. (1976) "Consequencesofsharpmilitary assistanceincreasesforinterna-
tionalconflictand cooperation."J. of ConflictResolution20 (December): 609-636.

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions
342 INTERNATIONAL STUDIES QUARTERLY

TANTER, R. (1966) "Dimensionsofconflictbehaviorwithinand betweennations,1958-


1960."J. of ConflictResolution10 (March): 41-64.
WALLACE, M. D. (1977) "Armsracesand escalation:somenewevidence."Thisarticleis
a revisedversionof a paperpresentedat theSocial ScienceHistoryAssociation,Ann
Arbor,Michigan,October.
--- (1973) "Alliance polarization,cross-cutting
and internationalwar 1815-1964:a
measurement procedureand some preliminary evidence."J. of ConflictResolution
27 (December): 575-604.
WALTZ, K. (1964) "The stabilityof a bipolarworld."Daedalus 93 (Summer):881-908.
WEEDE, E. (1970) "Conflictbehaviorof nationstates."J. of Peace Sci. 3: 229-236.
WILKENFELD, J. (1972) "Models fortheanalysisofforeignconflictbehaviorofstates,"
pp. 275-298in B. M. Russett(ed.) Peace, Warand Numbers.BeverlyHills,CA: Sage.
--- V. L. LUSSIER, and D. TAHTINEN "Conflictinteractions in the Middle East,
1949-1967."J. of ConflictResolution2 (June): 135-154.
ZINNES, D. A. (forthcoming) "Why war?"in T. R. Gurr(ed.) Handbook of Conflict
Theoryand Research.New York: Macmillan.
--- (1979) "Selectionofappropriatemodelsfordescribingprecrisisbehavior."Center
forInternationalPolicy Studies,Indiana University.
--- and J. WILKENFELD (1971) "Analysisofforeignconflictbehaviorofnations,"
pp. 167-213in W. F. Hanrieder(ed.) ComparativeForeignPolicy:TheoreticalEssays.
New York: David McKay.

This content downloaded from 164.77.114.99 on Fri, 31 Jul 2015 18:12:32 UTC
All use subject to JSTOR Terms and Conditions

You might also like