Johns Hopkins University Press

An Introduction to the Structural Analysis of Narrative

Author(s): Roland Barthes and Lionel Duisit
Source: New Literary History, Vol. 6, No. 2, On Narrative and Narratives (Winter, 1975), pp. 237
-272
Published by: Johns Hopkins University Press
Stable URL: http://www.jstor.org/stable/468419
Accessed: 19-10-2015 19:42 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 [email protected].

Johns Hopkins University Press is collaborating with JSTOR to digitize, preserve and extend access to New Literary History.

http://www.jstor.org

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
 An Introductionto the
StructuralAnalysisof Narrative*
RolandBarthes

HERE ARE COUNTLESS FORMS of narrativein the world. First

of all, there is a prodigiousvarietyof genres,each of which
branchesout into a varietyof media, as if all substancescould
be relied upon to accommodate man's stories. Among the vehicles
of narrativeare articulatedlanguage,whetheroral or written,pictures,
stillor moving,gestures,and an orderedmixtureof all thosesubstances;
narrativeis presentin myth,legend, fables,tales, shortstories,epics,
history,tragedy,drame [suspensedrama], comedy,pantomime,paint-
ings (in Santa Ursula by Carpaccio, for instance), stained-glasswin-
dows,movies,local news,conversation.Moreover,in thisinfinitevariety
of forms,it is presentat all times,in all places, in all societies;indeed
narrativestartswith the very historyof mankind; thereis not, there
has never been anywhere,any people without narrative; all classes,
all human groups,have theirstories,and veryoftenthose storiesare
enjoyed by men of different and even oppositeculturalbackgrounds:
narrativeremains largely unconcernedwith good or bad literature.
Like life itself,it is there,international,transhistorical, transcultural.
Are we to inferfromsuch universality that narrativeis insignificant?
Is it so commonthatwe can say nothingabout it, except fora modest
descriptionof a few highlyparticularizedspecies, as literaryhistory
sometimesdoes? Indeed how are we to controlsuch variety,how are
we to justifyour rightto distinguishor recognizethem? How can we
tell the novel fromthe shortstory,the tale fromthe myth,suspense
drama from tragedy (it has been done a thousand times) without
referenceto a common model? Any criticalattemptto describeeven
the mostspecific,the most historicallyorientednarrativeformimplies
such a model. It is, therefore, understandablethat thinkersas early as
Aristotleshould have concernedthemselveswiththe studyof narrative
forms,and not have abandoned all ambitionto talk about them,giving
* Originally published in Communications,8 (1966), as "Introduction l'analyse
A
structuraledes r&cits."
i It will be recalled that such is not the case with either poetry or the essay,
which relyon the cultural level of the consumer.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
238 NEW LITERARY HISTORY

as an excuse the factthat narrativeis universal. And it is normalthat

structuralism,in theearlystages,shouldhave made narrativea primary
concern. For is it not one of structuralism's main preoccupationsto
controlthe infinitevarietyof speech acts by attemptingto describethe
language or langue fromwhich theyoriginate,and fromwhich they
can be derived? Faced with an infinitenumberof narrativesand the
manystandpointsfromwhich theycan be considered(historical,psy-
chological, sociological, ethnological,aesthetic,etc.), the analyst is
roughlyin thesame situationas Saussure,who was facedwithdesultory
fragmentsof language, seekingto extract,fromthe apparent anarchy
of messages,a classifying principleand a centralvantage point forhis
description. To confine myselfto the currentperiod, the Russian
formalists,Propp, and ILvi-Strauss have taught us to identifythe
followingdilemma: eithernarrativeis a randomassemblageof events,
in which case one can onlyspeak of it in termsof the narrator's(the
author's) art,talent,or genius-all mythicalembodimentsof chance;2
or else it shares with other narrativesa common structure,open to
analysis,howeverdelicateit is to formulate.There is a world of differ-
ence betweenthe fortuitous, in itsmostcomplexforms,and thesimplest
combinativeor obligatoryscheme: forno one can producea narrative
withoutreferring himselfto an implicitsystemof unitsand rules.
Wherethenshouldwe look forthe structureof narrative?No doubt
in the narrativesthemselves.All the narratives?Many commentators,
who admit the idea of a narrativestructure,are nevertheless reluctant
to cut loose literaryanalysis from the model used in experimental
sciences: they boldly insistthat one must apply a purelyinductive
methodto the studyof narrativeand that the initialstep mustbe the
studyof all narrativeswithina genre,a period, a society,if one is to
set up a general model. This commonsenseview is, nonetheless,a
naive fallacy. Linguistics,which only has some three thousand lan-
guages to contendwith,failedin the attempt;wisely,it turneddeduc-
tive,and fromthatday on, incidentally, it foundits properfootingand
proceeded with giant steps, even managing to anticipatefactswhich
had not yetbeen discovered.3What then are we to expectin the case
of the analysisof narrative,faced with millionsof narrativeacts? It is

2 There exists, of course, an art of the storyteller: it is the ability to generate

narratives (messages) based on the structure (the code); this art corresponds to
the notion of performanceas defined by Chomsky,and it is far remote from the
notion of authorial "genius," Romantically conceived as a personal, hardly explica-
ble, secret.
3 See the historyof the Hittite "a," postulated by Saussure and discovered in fact
fiftyyears later, E. Benveniste,Problemes de linguistiquege'nrale (Paris: Gallimard,
1966), p. 35.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 239

obviouslycommittedto deductiveprocedures;it is compelled to con-

ceive, first,a hypotheticalmodel of description (which American
linguistscall a "theory"), and then to proceed gradually fromthat
model down, towardsthe species,which at the same time partake in
and deviatefromthe model. It is onlyat the level of such conformities
or discrepancies,and equipped with a single tool of description,that
the analystcan turnhis attentiononce moreto the pluralityof narrative
acts, to their historical,geographical,and cultural diversity.4
In order to describeand classifythe infinitenumberof narratives,
one needs then a "theory" (in the pragmaticsense that we are here
intending), and we must turn to the task of searchingfor one and
sketchingit out.5 The workingout of such a theorymay be made much
easier if we proceed froma model that can provide the initial terms
and principles. In the currentstate of research,it seems reasonableto
elect linguisticsitselfas a basic model for the structuralanalysis of
narrative.6

I. The Language of Narrative

1. Beyond the sentence

As everyoneknows, linguisticsstops at the sentence; it is the last
unit thatfallswithinits scope; forif the sentence-being an orderand
not a sequence-is not reducible to the sum of its words, and con-
stitutesthereforean originalunit, an enunciation,on the otherhand,
is nothingbut the successionof the sentencesit contains. From the
point of view of linguistics,there is nothingin discoursethat is not
matched in the sentence. "The sentence," writes Martinet, "is the
smallest segment that is perfectlyand systematicallyrepresentative
of discourse."' It followsthat linguisticscannot conceivablyadopt for

4 Let us keep in mind today'sconditionsof linguisticdescription:"Linguistic

structureis always relatednot only to the data of the corpus,but also to the
grammaticaltheorywhich describesthese data" (E. Bach, An Introductionto
Transformational Grammars[New York, 1964], p. 29). And also the following,
fromBenveniste(Problimes,p. i ig) : "It has been recognizedthatlanguagemust
be describedas a formalstructure, but that this descriptionrequired,as a pre-
requisite,the establishment of adequate proceduresand criteriaand that,in the
finalanalysis,the realityof the object was not separablefromthe methodchosen
to defineit."
5 The apparent"abstract"characterof the theoreticalcontributions found in
Communications, 8 (1966), is due to a methodological preoccupation: that of
rapidlyformalizing concreteanalyses:formalization thatdiffers
is a generalization
fromothergeneralizations.
6 But not indeed imperative(see Claude Bremond'scontribution, based on
logicalratherthan linguistic approach,in Communications, 8 [g966],60-76).
7 "R6flexions surla phrase,"Languageand Society(Copenhagen,1961), p. 113-

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
240 NEWLITERARY
HISTORY

its object anythingsuperior to the sentence, because beyond the

sentence,all there can ever be is more sentences: having described
the flower,the botanist cannot concern himselfwith describingthe
"bouquet."
And yet it is obvious that discourseitself (as an arrangementof
sentences) is organized, and that, through this organization,it is
perceived as the message of another "language," functioningat a
higherlevel than the language of linguistics:8 discoursehas its units,
its rules, its "grammar." Because it lies beyond the sentence,and
though consistingof nothingbut sentences,discoursemust naturally
be the object of a second linguistics.This linguisticsof discoursehas
fora verylong timehad a famousname: rhetoric.But as a resultof
an intricatehistoricalprocess,rhetoricwas switchedover to the hu-
manitiesthathad become separatedfromthe studyof language. It has
become necessary,of late, to take a freshlook at the problem: the
new linguisticsof discourse has not yet developed, but it has been
postulated by linguiststhemselves.9This fact should not be over-
looked: althoughdiscourseconstitutesan autonomousobject of study,
it mustbe studiedfromthe vantage point of linguistics.If a working
hypothesisis to be assignedto an analysisburdenedwiththe enormous
task of dealing with an infinityof materials,it is most reasonable to
postulatea homologousrelationbetweensentenceand discourse,assum-
ing thata similarformalorganizationencompassesall semioticsystems,
whatevertheirsubstancesor dimensions. Discourse would then be a
large "sentence" (whose unitsdo not necessarilyhave to be sentences)
in the same way that a sentence,allowingforcertainspecifications, is
a small "discourse." This hypothesisfits in well with certain pro-
positionsof currentanthropology.Jakobson and Livi-Strauss have
pointed out that the human status could be definedas the abilityto
createsecondary,"self-multiplying" systems(toolsto make tools,double
articulationof language, incest taboo conducive to the extensionof
families), and the Soviet linguistIvanov supposes that artificiallan-
guages cannot be acquired prior to the developmentof natural lan-
guages. It is thereforelegitimateto postulatea "secondary" relation
betweensentenceand discourse-a homologousrelation-to reflectthe
purelyformalcharacterof correspondences.
The generallanguage of narrativeis but one of many idiomswithin
8 It goes without saying, as Jakobson did not fail to notice, that between the
sentence and the space beyond it, there are transitions: coordination,for instance,
may reach beyond the sentence.
9 See, in particular: Benveniste, Problemes, Ch. Io; Z. S. Harris, "Discourse
Analysis,"Language, 28 (1952), I-3o; N. Ruwet, "Analyse structuraled'un po'me
francais,"Linguistics,3 (1964), 62-83.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 24 I

the scope of the linguisticsof discourse,10and consequentlyit comes

under the homologoushypothesis.Structurally, narrativebelongswith
the sentencewithouteverbeingreducibleto the sum of itssentences: a
narrativeis a large sentence,just as any declarativesentenceis, in a
certainway,the outlineof a littlenarrative.The main categoriesof the
verb (tenses, aspects, modes, persons) have their equivalent in nar-
rative,except that they are expanded and transformedto match its
size, and are equipped with signifiersof their own (often extremely
complex ones). Moreover,the "subjects" themselves,in theiropposi-
tion to verbal predicates,also tend to conformto the sentencemodel:
the actantial typologyput forwardby A. J. Greimas sees the great
number of charactersto be found in narrativeas equivalent to the
elementaryfunctionsof grammaticalanalysis." The kind of homology
here suggestedis interesting not merelyforits heuristicvalue, but also
because it implies an identitybetween language and literature(inas-
much as it is a sort of privilegedvehicle for narrative). It is hardly
possibleany longerto conceiveofliteratureas an artwhichwould stand
free of any relationto language, having once used the latter as an
instrument to expressideas, passion,or beauty: language neverceases
to accompany discourse,holdingup to it, as it were, the mirrorof its
own structure.Doesn't literature,more particularlyin our day, turn
the veryconditionsof language use into a language of its own?12

2. The levels of meaning

From the very first,linguisticsprovided the structuralanalysis of
narrativewith a decisiveconcept,because it pointedout the essentials
forany systemof meaning,namelyits organization;linguisticsmade it
possibleat once to spell out how narrativediffersfroma mere series
of propositions,and to clarifythe enormousmass of elementsthat go

Io It would be preciselyone of the tasks of the linguisticsof discourse to lay the

foundation of a typologyof discourse. On a temporarybasis, one can recognize
three broad types of discourse: metonymous (narrative), metaphorical (lyrical
poetry, sapiential discourse), enthymematic (intellectual discursive).
II See "El1ments pour une theorie de l'interpretationdu recit mythique," III,
I, Communications,8 (1966).
12 Mention must be made here of Mallarmr's insight,at the time he was con-
templatinga project in linguistics: "Language has appeared to him as the instru-
ment of fiction: he will follow the method of language (to be determined).
Language, as it were, mirrored. Finally, fiction seems to him to be the very
procedure of the human mind-it is fiction which causes all method to be
broughtinto play, and man is reduced to his will" (Oeuvres compltes [ed. Pleiade],
p. 851). One will recall that for Mallarmr fiction and poetry are synonymous
("situated at the converging point of other arts, generated by them, and gov-
erningthem,thereis Fiction or Poetry,"ibid., p. 335).

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
242 NEWLITERARY
HISTORY

into the makingof a narrative. Such a concept was that of the level
ofdescription."
It is well knownthata sentencecan be described,in linguisticterms,
on several levels (phonetic, phonological,grammatical,contextual);
theselevelsstand in hierarchicalrelationto each other,forif each has
its own units and its own correlations,thus making an independent
descriptionmandatory,thennone can, of itself,produce any meaning.
No unit pertainingto a certainlevel can be endowed with meaning
unlessit can be integratedinto a superiorlevel: a phoneme,although
perfectlydescribable,means nothingby itself;it partakesin meaning
only if integratedinto a word; and the word itselfmust in turn be
integratedinto the sentence.14The theoryof levels (as enunciatedby
Benveniste)providestwo typesof relations: distributional(if the rela-
tionsbelongon thesame level), integrative(if theystraddletwo levels).
It followsthat distributionalrelationsalone are unable to account for
meaning. Thus, in orderto carryout a structuralanalysis,it is necessary
firstto distinguishseverallevelsof description[instancede description]
and to place theselevelswithina hierarchical(integrative)perspective.
Levels are operations.15Thus it is normal that linguisticsshould
tend to multiplythem as it progresses.For the time being,analysisof
discourse can only operate at rudimentarylevels. In its own way,
rhetorichad assignedat least two planes of descriptionto discourse:
dispositioand elocutio.16Nowadays, in his analysisof the structureof
myth,LUvi-Strauss has alreadyspecifiedthatconstitutiveunitsof mythi-
cal discourse(mythemes)becomesignificant onlybecause theyappear in
clusterswhich in turn combine among themselves;17 and Tzvetan
Todorov, takingover the distinctionof the Russian formalists, suggests
workingon two large levels, each of which may be broken down
further: the story(the argument),which consistsof a logic of actions
and a "syntax"of characters,and discourse,comprisingtenses,aspects,

13 "Linguistic descriptions are never monovalent. A description is not correct

or incorrect,it is better or worse, more useful or less useful" (M. A. K. Halliday,
"Linguistique gendrale et linguistiqueappliqude," ltudes de linguistique appliquie,
I [Ig62], p. 12).
14 The levels of integration were postulated by the Prague School (see J.
Vachek, A Prague School Reader in Linguistics [Bloomington,Ind., 19641],p. 468),
and have been adopted since by many linguists. We think that Benveniste (Prob-
lemes,Ch. I o) gave this theoryits clearest formulation.
15 "Loosely defined,a level can be considered as a systemof symbols,rules, etc.,
which must be used to representexpressions" (E. Bach, Introduction, pp. 57-58).
I6 The third part of rhetorics,inventio,did not concern language: it had to do
with res,not with verba.
StructuralAnthropology(New York, 1962), p. 233.
17

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 243

and modes pertainingto narrative.'" Whateverthe numberof levels

one proposesto study,and whatevertheirdefinition, thereis no doubt
that narrativeis a hierarchyof levels or strata. To understanda nar-
rative is not only to follow the unfoldingof the storybut also to
recognizein it a number of "strata," to project the horizontalcon-
catenationsof the narrativeonto an implicitlyvertical axis; to read
a narrative(or listento it) is not only to pass fromone word to the
next,but also fromone levelto the next. Let me introduceat thispoint
a kind of apologue: in The Purloined Letter,Poe pungentlyanalyzes
the failureof the Police Inspectorto lay his hands on the letter. His
investigations were perfect,"withinthe scope of his specialty,"to quote
Poe's words. The inspectordid notomita singlelocation,he completely
"saturated" the level of the "search"; but in order to find the letter,
protectedas it was by itsveryprominence,one had to switchto another
level,in otherwords,to substitutethe relevanceof the concealerforthe
relevance of the police agent. In similar fashion,however complete
the "search" mightbe when it came to bear on a horizontalset of nar-
rativerelations,in orderto be efficient, it mustalso be directed"verti-
cally": the meaning does not lie "at the end" of the narrative,but
straddlesit. Thus, meaningeludes any unilateralinvestigation, no less
than the purloinedletteritself.
Many trialsand errorsare to be expected beforethe levels of nar-
rativecan be identifiedwith certainty.The ones we are offeringhere
constitutea tentativeprofilewhose principalmeritis, forthe moment,
almostexclusivelydidactic: throughthemwe can situateand classify
problems,withoutincurringdisagreementwith the few analyses that
have taken place.19 We propose to distinguishthree levels in any
narrativework: the level of "functions" (in the sense Propp and
Bremondgave to thisword), the level of "actions" (in the sense used
by Greimas when he writesof charactersas actants), and the level
of "narration" (which is roughlythe level of "discourse" as seen by
Todorov). Attentionis again called to the fact that those levels are
bonded togetheraccording to a mode of progressiveintegration: a
functionhas a meaningonlyinsofaras it takes its place in the general
line of action of an actant; and thisaction in turnreceivesits ultimate
meaningfromthe fact that it is being told, that is, entrustedto a dis-
course which possessesits own code.

i8 See Tzvetan Todorov, "Les categories du r6cit litteraire,"Communications,8

(1966), 125-51.
19 My main concern, in the introduction,has been to interfereas little as possible
with currentresearch.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
244 NEW LITERARY HISTORY

II. Functions

1. The determination of units

Since any systemcan be definedas a combinationof unitspertaining
to certainknown classes,the firststep is to break down the narrative
and determinewhateversegmentsof narrativediscoursecan be dis-
tributedinto a limitednumberof classes; in otherwords,to definethe
smallestnarrativeunits.
Accordingto the integrativeperspectivehere defined,a purelydis-
tributionaldefinitionof units will not do: meaning must be, from
the veryfirst,the criterionby which units are determined. It is the
functionalcharacterof certainsegmentsof the storythat makes units
of them,hence the name of "functions,"earlyattributedto thosefirst
the practicehas been to regard
units. Since the Russian formalists,20
as a unitany segmentof thestorywhichpresentsitselfas the termof a
correlation. The "soul" of any functionis, as it were, its seedlike
quality,which enables the functionto inseminatethe narrativewith
an elementthat will latercome to maturity,on the same level,or else-
whereon anotherlevel. If, in Un Coeur simple,Flaubertinformsthe
readerat a certainpoint,nonchalantlyas it seems,thatthe sous-prc'fet's
daughtersin Pont-l'Eveque owned a parrot,it is because thisparrotis
to play an importantrole in Fd1icit5'slife: the enunciationof this
detail (whicheverlinguisticformit may assume) constitutes a function,
or narrativeunit.
Is everythingfunctionalin a narrative? Is everything, down to the
most minutedetail, meaningful? Can narrativebe integrallybroken
down into functionalunits? As will soon become apparent, there
are no doubt severalkindsof functions,for thereare several kinds of
correlations.The fact remains,however,that a narrativeis made up
in one way or another,is significant.
solely of functions: everything,
It is not so much a matterof art (on the part of the narrator) as it is
a matterofstructure.Even thougha detailmightappear unequivocally
trivial,imperviousto any function,it would nonethelessend up point-

20 See, in particular, B. Tomachevski, "Th6matique" (1925), Thdorie de la

littirature (Paris: Seuil, 1965). A little later, Propp defined a function as "an
act of a character, defined fromthe point of view of its significancefor the course
of the action" Morphologyof the Folktale (1928), tr. Laurence Scott (1958; Austin
and London, 1968), p. 21. Todorov's definition,"The meaning (or function)
of an element in the work is its ability to enter into correlationswith other ele-
ments in this work,and with the work as a whole," is to be found in Communica-
tions, 8, which also includes precisions contributedby A. J. Greimas, who comes
to define a unit by its paradigmatic correlation,but also by its position within the
syntagmaticunit to which it belongs.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 245

ing to its own absurdityor uselessness: everythinghas a meaning,or

nothinghas. To put it in a different way, Art does not acknowledge
the existenceof noise (in the informationalsense of the word).21 It is
a pure system: thereare no wasted units,22 and therecan neverbe any,
howeverlong, loose, or tenuousthe threadswhich link themto one of
thelevelsofthestory.23
From a linguisticpoint of view, the functionis obviouslya content
unit: it is "what an utterancemeans," not the way it is made, which
constitutesit as a functionalunit.24This essentialsignifiedcore may
have a varietyof signifiers, some of them quite devious. If we are
informed(in Goldfinger)that "James Bond saw a man in his fifties,"
such informationinherentlycontains two simultaneousfunctions,re-
flectingan unequal degreeof urgency: on the one hand, the age of the
characterfitsinto a certainportrait(whose relevanceto the remaining
part of the storyis not negligible,but diffuse,or delayed), and on the
otherhand, the immediatesignificationof the utteranceis that Bond
does not know his futureadversary. The unit thus implies a very
strongcorrelation(the openingof a threatcoupled with an obligation
to identify). In order to determinethe initial narrativeunits, it is
therefore necessaryneverto lose sightof the functionalcharacterof the
segmentunder consideration,and to be preparedin advance to recog-
nize that those segmentswill not necessarilycoincide with the forms
traditionallyattributedto the various parts of narrative discourse
(actions, scenes, paragraphs,dialogues, inner monologues,etc.), and
still less with "psychological"classes (behaviors, feelings,intentions,
motivations,rationalizationsof characters).
Similarly,since the langue of narrativeis not the langue of articu-
lated language-though it oftenuses the latteras its vehicle-the nar-

21 This is preciselywhat distinguishesit from"life," which offersonly a "blurred"

communication. The "blurred" effect (that which limits the view) may exist in
art, but then only as a coded element (Watteau, for instance); and the "blurred"
effect,for that matter, does not exist in the written code, which inevitably calls
forclear delineation.
22 At least in literature,where the freedom of notation (due to the abstract
nature of articulated language) implies a much stronger commitment than in
the "analogical" arts,such as movie making.
23 The functionalityof the narrative unit is more or less immediate (hence
noticeable), depending on the level where it operates: whenever the units are
positioned on the same level (in the case of suspense, for instance), the functionality
is quite noticeable; much less, however, when the function becomes saturated at
the narrational level: a modern text, with a lower degree of significanceon the
anecdotic level, achieves its full impact only at the level of icriture.
24 "The syntacticunits (beyond the sentence) are in fact content units" (A. J.
Greimas, Cours de semantique structurale,cours roneotype,VI, 5). The explora-
tion of the functionallevel is thereforea part of general semantics.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
246 NEW LITERARY HISTORY

rativeunitsare independentof linguisticunitswithregardto substance.

They may indeed coincide,but onlyoccasionallyratherthan systemati-
cally; functionswill be representedat timesby unitslarger than the
sentence (groups of sentencesof yaryinglength,up to the work as a
whole), at timesby lesserunits (the syntagm,the word, and even in
the word, only certainliteraryelements25). When we are told that
Bond, upon hearing the telephone ring while on duty in his Secret
Service office,"picked up one of the fourreceivers,"the monemefour
constitutesin itselfa functionalunit, for it refersto a concept which
is necessaryto the storyas a whole (one of a highlytechnical bu-
reaucracy). In fact,in thiscase, the narrativeunitis not the linguistic
unit (the word), but onlyitsconnotativevalue (linguistically,theword
four never means "four"). This explains why, on occasions, certain
functionalunitscan be smallerthan the sentencewhile stillbelonging
to discourse;such unitsreach out beyondthelevel of denotationwhich,
like the sentence,belongsto linguisticsproper,even thoughthe units
may be materiallyconfinedby the sentenceof which theyare a part.

2. Classes of units
These functionalunits must be distributedinto a small number of
formalclasses. If one is to determinethese withoutrelyingon their
content (psychologicalsubstance,for instance), one must again con-
sider the various levels of meaning: some units correlatewith units
on the same level,while otherscannot be fulfilledwithoutswitchingto
anotherlevel. Hence the necessityto provide,at the outset,two broad
classes of functions,distributionalon the one hand, integrativeon the
other. The formercorrespondto Propp's functions,revivedby Bre-
mond among others,but which we intend to considerhere in much
greaterdetail than theydid. To thesealone we shall assignthe name
of "functions"(although the otherunitsare no less functional). The
model has become a classical one afterTomachevski'sanalysis: the
purchase of a gun has, for its correlate,the momentwhen it is put
to use (and if it is not used, the functionis invertedto designate
vacillation,etc.) ; pickingup thephone has foritscorrelatethe moment
when it is laid down; the intrusionof the parrotinto Fdlicite'shome
correlateswiththe stuffing episode,itsworship,and so on. The second
broad class of units,integrativeunits,comprisesall the "indices" or

25 "One must not consider the word, as a primary,indivisibleelement of literary

art, like a brick used in the constructionof a building. It can be broken down
into more tenuous 'verbal elements'" (J. Tynianov, quoted by Todorov, Langages,
6[1971], 18).

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 247

"indicators" (in the broader sense of the word).26 In that case, the
unit, instead of referringto a complementaryand consequentialact,
refersto a more or less diffuseconcept which is nonethelessnecessary
to the story: personalitytraitsconcerningcharacters,information with
regard to theiridentity,notationsof "atmosphere,"and so on. The
relationbetween the unit and its correlateis no longer distributional
(often several indices point to the same signifiedand the order of
occurrencein discourseis not necessarilyrelevant) but integrative;in
orderto understandwhat purposean index [indice] or indicatorserves,
one mustpass on to a higherlevel (actions of the characteror narra-
tion), foronly therecan the "index" be clarified. The administrative
power that lies behind Bond, suggestedby the numberof lines on his
phone, does not have any bearingon the sequence of actionstriggered
by the act of answeringthe phone; it only takes on value on the level
of a general typologyof character (Bond is on the side of Order).
Indices, because theirrelationsare, as it were, verticallyoriented,are
trulysemanticunits,forunlikeproperlydefined"functions"that refer
to "operations,"indicesreferto a signified,not to an "operation." The
sanctionof indicesis "higher-up,"sometimesit is even virtual,outside
the explicitsyntagm(the personalitytraitsof a charactermay never
be verbalizedand yetrepeatedlyindexed), it is a paradigmaticsanction.
By contrast,the sanction of "functions"is always "furtheron," it is
a syntagmaticsanction.27Indeed, the distinctionbetween functions
and indices bears out another classical distinction: functionsimply
metonymicrelata,indicesmetaphoricrelata; the formerare functional
in termsof action,the latterin termsof being.28
These two main classes of units,functionsand indices,account for
of narratives.Some narrativesare predominantly
a certainclassification
functional (such as popular tales), while some others are predomi-
nantlyindicial (such as "psychological"novels). Between these two
opposites,we have a whole spectrumof intermediary forms,deriving
theircharacteristicsfromhistory,society,or genre. But that isn't all:
withineach of thosetwo broad classes,two subclassesof narrativeunits
can readilybe determined. Referringback to the class of functions,
its units are not equally "important": some constituteactual hinges

26 These designations,and the ones subsequentlyintroduced,may all be temporary

ones.
27 This does not preclude the possibilitythat,ultimately,the syntagmaticdispersion
of functionsmay come to express a paradigmatic relation between separate func-
tions, as has been generallyacknowledged since L6vi-Strauss and Greimas.
28 One cannot reduce functions to actions (verbs) nor indices to modifiers
(adjectives), for there are actions with indicial value, "signaling" a personality,
an atmosphere,etc.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
248 NEW LITERARY HISTORY

of the narrative(of a fragmentthereof); othersdo no morethan "fill

in" the narrativespace separatingthe hinge-typefunctions. Let us
call the formercardinal functions(or nuclei), and the latter,in view
of theircomplementary nature,catalyses.In orderto classifya function
as cardinal,all we need verifyis thatthe actionto whichit refersopens
(or maintainsor closes) an alternativedirectlyaffectingthe continua-
tion of the story,in otherwords,that it eitherinitiatesor resolvesan
uncertainty.If in a fragmentof narrativethe telephonerings,it is
equally possibleto answer or not to answer the call, proceduresthat
are bound to carrythe storyalong different paths. On the otherhand,
between two cardinal functions,it is always possibleto bring in sub-
sidiarynotations,which clusteraround one nucleus or another,with-
out modifyingits alternativenature: the space separating"the tele-
phone rang" from "Bond picked up the receiver" can be saturated
with countlessminor incidentsor descriptions,such as "Bond made
his way to the desk, picked up the phone, put down his cigarette."
These catalysesare stillfunctional,insofaras theyenterinto correla-
tionswith a nucleus,but theirfunctionality is toned down, unilateral,
parasitic. The functionalityinvolved is purelychronological(what is
describedis what separatestwo momentsof a story),whereasthe link
between two cardinal functionspossessesa double functionality, at
once chronologicaland logical: catalysesare no morethan consecutive
units,while cardinal functionsare both consecutiveand consequential.
Indeed, thereis a strongpresumptionthat the mainspringof the nar-
rative activityis to be traced to that veryconfusionbetweenconsecu-
tivenessand consequence,what-comes-after being read in a narrative
as what-is-caused-by.Narrative would then be a systematicappli-
cation of the logical fallacy denounced by scholasticismunder the
formulapost hoc, ergo propterhoc, which may well be the motto of
Destinywhose "language," afterall, findsits expressionin narrative;
and this "telescoping"of logic and temporalityis mainlyachieved by
the frameworkof cardinal functions. These functionsmay at first
glance appear quite trivial. What makes them crucial is not their
spectacularquality (the importance,the volume, the unusual nature,
or the impact of the enunciatedaction), but ratherthe riskinvolved:
the cardinal functionsare the risk-ladenmomentsof narrative. Be-
tween the disjunctivepoints,or "dispatchers,"the catalysesopen up
areas of security,rest,or luxury; such "luxuries," however,are not
useless. It should be stressedagain that,fromthe point of view of the
story,catalysisremainsfunctional,even if only marginally. Were it
purely redundant (in relation to its nucleus), it would nevertheless

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 249

partake in the economy of the message. But it is not redundant.

Though a particularnotationmay seen expendable, it retainsa dis-
cursivefunction: it precipitates,delays, or quickens the pace of dis-
course,sums up, anticipates,and sometimeseven confusesthe reader.29
Since what is noted alwaystendsto be seen as what is "worthnoting,"
catalysisconstantlyreactivatesthe semantictensionof discourse,for-
ever saying: there has been, there is going to be, meaning. The
enduringfunctionof catalysisis, then, in the final analysis,a phatic
function(to use Jakobson'sterm): it maintainscontact betweenthe
narratorand the reader. To sum up, one cannot delete a nucleus
withoutalteringthe story,but then again one cannot delete a catalysis
withoutalteringthe discourse.
With regardto the second broad class of narrativeunits (indices),
the units they contain have this in common: they can be saturated
(completed) only on the level of characters,or on the level of nar-
ration. They are part of a parametricalrelation,30whose second term,
on account of its implicitnature,remainscontinuouslyactive, affect-
ing a whole episode, a character,or the work as a whole; however,a
distinctioncan be made betweenindicesproper,referring to a person-
a
alitytrait, feeling,or an a
atmosphere(e.g., suspicion), philosophy,
and, on the otherhand, bitsof information used to identifyor pinpoint
certainelementsof timeand space. To say that Bond is on dutyin his
officewhile, throughhis open window, heavy billowing clouds can
be seen obscuringthe moon, is to index a stormysummer night,a
deductionwhich can in turnbe translatedinto an atmosphericalindex
pointingto the heavy, anguish-ladenclimate of an action as yet un-
known to the reader. It followsthat an index always signifiesim-
plicitly,whileinformants do not,at least on the level of the story: they
provide pure, locally relevantdata. Indices imply a decipheringac-
tivenessand consequence,what-comes-after being read in a narrative
with a character or an atmosphere; informantsbring with them a
ready-madeknowledge. Like catalyses,theyare marginallyfunctional
yet stillfunctional: whateverthe "flatness"in relationto the restof
the story,the informant(e.g., the preciseage of a character) is there
to authenticatethe realityof the referent,to root fictionin the real
world. Whateverservesas informantis a realisticoperator,and to that

29 Val6ryspokeof "dilatorysigns." The detectivemakesextensiveuse of these

"deceptive"units.
30 Accordingto Ruwet,a parametricalelementis an elementwhich remains
constantthroughout the durationof a musicalpiece (forinstance,the tempoin a
Bach allegro,or themonodiccharacter ofa solo).

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
250 NEW LITERARY HISTORY

extent,it possessesof an undeniable functionality, if not on the level

of the story,at least on the level of discourse.31
Nuclei and catalyses,indices and informants(again, the names are
immaterial) are, it seems, the initial classes into which the units of
the functionallevel can be distributed.Two remarksshould be ap-
pendedto thisclassification.
First of all, a unit can at the same time belong to two different
classes: to drink.a whiskey(in the hall of an airport) is an action
that can pass offas a catalysisto the (cardinal) notationof waiting,
but it is also, and at the same time,an index to a certainatmosphere
(modernity,relaxation,reminiscence,etc.): in other words, certain
unitscan be mixed units. This opens up a whole range of possibilities
in the economyof narrative;in the novel Goldfinger,Bond, having to
conduct a search in his opponent's room, receives a pass from his
associate: the notation is a clear-cut function (cardinal). In the
filmversion,thisdetail is changed. Bond laughinglysnatchesa set of
keysfroman uncomplainingchambermaid; the notationis no longer
merely functional,but also indicial, pointing to Bond's personality
type (his devil-may-careways and his success with women). In the
second place-more on the subject later-it should be noticed that
the four classes just mentionedare subject to anotherdistribution-
closer to the linguisticmodel, incidentally.Catalyses,indices,and in-
formantsindeed have one characterin common: theyare expansions
in theirrelationto the nuclei. Nuclei (as will be shown shortly)form
togetherfinitesets combiningveryfew terms; theyare logicallycon-
trolled,at once necessaryand sufficient.Once thisframeworkhas been
constituted, the otherunitsfillit in accordingto a mode of proliferation
which has no theoreticallimits. As everyoneknows,that is what hap-
pens to thesentence,whichis made up of simplepropositions, yetkeeps
sproutingany numberof duplications,paddings,convolutions,and so
forth. Like the sentence,narrative can give forthany number of
catalyses. Mallarme bestowed so much importanceon this type of
structurethat he made it the organic principleof Jamais un coup de
des, which may well be considered,complete with its "nodes," its
"antinodes,"its "nodal words," and "lace-words,"as the veryblazon
of all narrativeform-of all language.

31 In Communications, 8 (1966), I52-63, G. Genette establishes two types of

description: the ornamental and the meaningful. The latter relates of course to
the level of the storyand the ornamental to the level of discourse, which explains
why, for a long time, it made up a perfectlycoded "piece" of rhetorics: descriptio
or ekphrasis,a highlyregarded exercisein neo-rhetorics.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 25 I

3. Functional syntax

How, accordingto what "grammar,"are the different unitslinked

togetherin the narrativesyntagm?What are the rulesof thefunctional
"combinative"or obligatoryscheme? Informantsand indicescan com-
bine freelyamong themselves: such is the case in the portrait,which
presents,side by side,withoutrestrictions, personalbiographicalrecords
and personalitytraits. A simple implicativerelation binds together
nuclei and personalitytraits: a catalysisnecessarilyimpliesthe exist-
ence of a cardinalfunctionon to whichit can depend,but the implica-
tionis not reversible.As forcardinalfunctions, theyare bound together
in a solidarityrelation: a functionof this type combinesselectively
withone of its own kind,and vice versa. This solidarityrelationmust
engage our attentionfurther: first,because it helps definethe very
frameworkof the narrative(expansions are optional,nuclei are not),
second,because theyare the principalconcernof researcherswho seek
to give a structureto narrative.
It has alreadybeen pointedout thatnarrative,on account of its very
structure,tends to establisha confusionbetween consecutivenessand
consequence,betweentimesequence and logic. In that ambiguitylies
the centralproblemof narrativesyntax. Is it possibleto uncover,be-
hind the temporalsequence of the narrative,an atemporallogic? This
point has been a divisiveissue among researchersuntil quite recently.
Propp, who has been creditedwithopeningthe way to presentstudies,
adamantly defendedthe principlethat the chronologicalorder is ir-
reducible: to him timeis theverystuffof realityand forthisreason,he
insistedon rootingthe tale in temporality.Yet Aristotle,even as he
contrastedtragedy (defined by its unity of action) to the narrated
story(definedby a pluralityof actionswithinone temporalscheme),
was already stressingthe primacyof logic over chronology.32And so
have modernresearchers(LIvi-Strauss,Greimas,Bremond,Todorov),
all of whom (while possiblydivergingon otherpoints) would probably
subscribeto thispropositionby UIvi-Strauss: "The chronologicalorder
of successionis reabsorbedby an atemporalmatrix."33 Contemporary
analysis tends to "dechronologize" the narrativecontinuum and to
"relogicize" it, subjectingit to what Mallarme used to call, referring
to the French language, "the primitivethunderboltsof logic." 4 To
be more precise, the goal is to give a structuraldescriptionto the
chronologicalillusion;it is up to narrativelogic to accountfornarrative

32 Poetics, 1459a.
33 Quoted by Bremond, "Le Message narratif," Communications,4 (1964).
34 Quant au livre (Oeuvres completes[ed. Pl1iade], p. 386).

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
252 NEW LITERARY HISTORY

time. To put it anotherway, temporalityis no more than a structural

class of narrative(understoodas discourse), just as in ordinarylan-
guage, timeexistsonlyin the formof a system.From the pointof view
of narrative,what we call timedoes not exist,or at least it only exists
functionally, as an elementof a semioticsystem: timedoes not belong
to discourseproper,but to the referent.Both narrativeand language
can only referto semiologicaltime; "true" time is only a referential
illusion,"realistic,", as Propp's commentaryshows. It is in thisrespect
onlythatstructuraldescriptioncan presumeto come to termswithit.35
What then is the logic that regulatesthe principalfunctionsof nar-
rative? Establishingsuch a logic has been an activelypursued and
mostwidelydebated goal in currentresearch. Referenceis here made
to contributionsby A. J. Greimas, Claude Bremond, and Tzvetan
Todorov, publishedin Communications,8, all of which deal with the
logic of functions.Three main trendsof researchare emerging,set
forthby Todorov in his article. The first, initiatedby Bremond,is more
properlylogical in its approach: the goal is to reconstruct
the syntax
of human behavioras exemplifiedin narrative,to trace the succession
of "choices" which this or that characterinevitablyhas to face36 at
various pointsin the story,and thus to bring to lightwhat could be
called an energeticlogic,37since charactersare caught at the moment
when theychoose to act. The second model is linguistic(Levi-Strauss,
Greimas): the essentialpreoccupationof this researchis to identify
paradigmaticoppositionsin the functions,and then to "project" such
oppositionsonto the syntagmaticaxis of narrative,according to the
Jakobsoniandefinition of the "poetic" principle(evidencewillbe found
in Communications,8 of new developmentsin Greimas'thinkingwhich
tend to corrector completehis paradigmaticapproach to functions).
The thirddirectionof research,sketchedout by Todorov, is somewhat
different,forit sets up the analyticalprocesson the level of "actions"
(that is to say, of characters),and triesto figureout the ruleswhich

35 In his own way, keenly perceptive as always though not driven to its con-
clusions, Valery has correctlyformulatedthe status of narrative time: "The belief
in time as an agent and a guiding thread is based on the mechanism of memory
and that of combinative discourse" (Tel Quel; emphasis ours): the illusion is
indeed a product of discourse-itself.
36 This conception bears a certain resemblance to one of Aristotle'sviews: the
proairesis, a rational choice of potential actions, is the foundation of praxis, a
practical science which, unlike poiesis, does not produce any work distinct from
its agent. In these terms,one may say that the analyst triesto reconstructthe praxis
which operates withinnarrative.
37 This logic based on choice (to do this or to do that) has the advantage of
accounting for the dramatization process which is usually embodied in narrative.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 253

attendthe combinations,variations,and transformations, in narrative,

of a certainnumberof fundamentalpredicates.
No attemptis heremade to choose among thoseworkinghypotheses;
theyare not rivalbut paralleltheories,stillin the processof elaboration.
The only point on which we could venturea few complementaryob-
servationsconcernsthe dimensionsof the analyticaleffort.Even if we
set aside the indices,informants, and catalyses,therestillremainsin a
narrative (especially if it is a novel and not a tale) a considerable
numberof cardinalfunctions;manycannotbe controlledby the above-
mentionedanalyses,which have been dealing thus far with the larger
articulationsof narrative. Provisionmust be made, however,for a
detaileddescription,
sufficiently accounting units,in-
forall narrative
thesmallest
cluding as onewillrecall,
Cardinalfunctions,
segments.
cannotbe determined but onlyby the inter-
by their"importance,"
nature
locking oftheir
implicative call,"how-
relations:a "telephone
it mayappear,on theonehandcomprises
everfutile a fewcardinal
functions
(ringing,pickingup the phone,speaking,putting down
thephone),buton theotherhand,thesametelephone
call,considered
a chainofimplications,
at leastthrough
mustbe linked,
as a whole, to
of the anecdote.The widespan of functional
thelargerarticulations
in narrative
arrangement basedon relays,
imposesan organization
whosebasic unitscan be no otherthan a smallgroupof functions,
whichwillbe referred withBremond's
to as a sequence(in conformity
terminology).
A sequence ofnuclei,
isa logicalstring linked bya solidarity
together
relation:
38 thesequenceopens when one of itsterms
is lackingan
antecedentof the same kin, and it closes when another of its terms
no longerentailsany consequentfunction.To take a deliberatelytrivial
example, consecutivefunctionslike orderinga drink,receivingit, con-
sumingit, and paying for it, constitutean obviouslyclosed sequence,
for it is not possible to mention anythingprior to the orderingor
posteriorto the paying,withoutmovingaway fromthe homogeneous
set designatedas consommation.Indeed, a sequence is always name-
able. When determiningthe largerfunctionsof the tale, Propp, then
Bremond,foundit convenientto name them (Fraud, Treason, Struggle,
Contract,Seduction, etc.); the naming process is also inevitablefor
trivial sequences, those "micro-sequences,"as they might be called,
ofwhichthefinegrainofthenarrativetextureis made. Does the opera-
tion of naming sequences belong exclusivelyto the analyst? In other

38 In theHjelmhnslevian
senseof doubleimplication, twotermspresuppose
whereby
each other.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
254 NEW LITERARY HISTORY

words,is it purelymetalinguistic?Surelyit is, since it deals with the

narrativecode, yet one could argue that it is part of a metalanguage
elaborated by the reader (or listener) himself,as he apprehendsany
logical sequence of actions as a nominal whole: to read is to name;
to listenis not only to perceivea language, but also to constructthat
language. The sequence titlesare fairlysimilarto the cover-wordsof
translationmachines which cover, quite adequately, a great variety
of meanings and nuances. The conventionalnarrativelanguage in-
ternalizedby the reader comes readily equipped with such essential
head-words. The self-containedlogic which structuresa sequence is
inextricably tied to its name: any functionwhich initiatesa seduction
imposes, from the momentit appears, by virtueof what is conjured
up by the name, the whole processof seduction,as we have learned
through all the narrative acts thathave fashionedin us the "language"
of narrative.
However minimalits importance,the sequence, made up as it is of
a small numberof nuclei (which means,in fact,"dispatchers"),always
involvesmomentsof riskwhich make it worthyof analysis: it might
sound futileto set up as a sequence the logical successionof trivialacts
which go into the offeringof a cigarette(offering,accepting,lighting
up, smoking). Yet preciselyat each of those points,a choice, hence
a "freedom"of meaning,becomes possible: du Pont, Bond's special
agent,offersto lighthis cigarettewithhisown lighter,but Bond refuses;
the meaningof thisdeviationfromthe normis that Bond instinctively
shrinksfrom a booby-trappedgadget."9 One may say, then, that a
sequence is a potentiallyincompletelogical unit. As such, it is justified
withinthe local context,but it is also rooted in the larger context.
Because it is self-contained with regardto its functions,and bracketed
under a name, the sequence can be apprehendedas a unit, ready to
functionas a simple term in another,broader sequence. Take the
followingmicro-sequence: extendingone's hand, shaking hands, re-
leasing the handshake. This Greetingbecomes a simple function:
looked at in a certainlight,it assumesthe role of an index (du Pont's
flabbinessand Bond's shrinking fromit). Consideredas a whole,how-
ever, it constitutes one term along a broadersequence,subsumedunder
the name of Encounter,whose other terms (drawing near, stopping,
hailing,greeting,settlingdown together) can be micro-sequenceson

39 It is quitepossibleto identify,
evenat theinfinitesimal
level,an oppositionofa
paradigmatictype,if not betweentwo terms,at least betweentwo poles of a
sequence: thesequenceoffering ofa cigarette,
spreadsout-even as it suspendsit-
the paradigmDanger/Safety (broughtto lightby Cheglovin his analysisof the
SherlockHolmescycle),or Suspicion/Protection, Aggressiveness/Friendliness.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 255

theirown. A whole networkof subrogationsthus binds togetherthe

narrative,fromthe smaller matricesup to larger functions. We are
dealing here,of course,with a hierarchythat stillfitswithinthe func-
tional level. It is only when the narrativehas reached a greater
expansion, one connection leading to another-from the cigarette
offeredby du Pont to Bond's fightwith Goldfinger-that the analysis
of functionscan be consideredcomplete. The pyramidof functions
then yieldsto the next level (the level of Actions). There is indeed,
at the same time, a syntaxwithinthe sequence, and a (subrogating)
syntax regulatingfunctionsamong themselves. The firstepisode of
Goldfingerthus presentsitselflike a "stemma":

Request Aid

Encounter Solicitation Contract Surveillance Capture Punishment

Approach Hailing Greeting Installation

hand extended hand shaken hand released

This representation is obviouslyanalytical. The reader, by contrast,

perceives a linear succession of terms. But what calls for special at-
tentionis that some termsbelongingto several sequences can easily
dovetail into each other. Before a sequence is completed,the initial
termof a freshsequence can be introduced: sequencesproceedaccord-
ing to a contrapuntalpattern." Functionallythe structureof narrative
is that of the fugue: narrative"pulls in" new material even as it
"holds on" to previousmaterial. It is conceivablethat,withinthesame
work,thisdovetailingof sequencesmay suddenlybe interrupted, some-
wherealong the line, at a clean break-offpoint,yetthiscan only hap-
pen if the few independentblocks (or "stemma") that now make up
the work are recovered,as it were, on the upper level of Actions (of
characters). Goldfingeris made up of threefunctionallyindependent
episodes,since the stemmascease to interlockon two occasions. No
sequential relationexistsbetween the episode in the swimmingpool
and thatof Fort Knox; but thereremainsan actantialrelation,forthe
characters (hence the structureof their relationship) are the same.
Here we recognizethe epic pattern("a whole withmultiplefables"):

40 This counterpoint has been anticipated by the Russian formalists,who have

roughed out its typology; it is not unlike the principal "gnarled" structuresof a
sentence (see below, V. I).

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
256 NEW LITERARY HISTORY

the epic narrativeis brokenon the functionallevel but remainsone on

the actantial level (this can be verifiedin the Odyssey or Brecht's
theater). It is necessarythento top the level of functions(which sup-
pliesthe major part of the narrativesyntagm)witha higherlevel from
which,one afterthe other,the unitsof the firstlevel derivetheirmean-
ing, and thatis the levelof Actions.

III. Actions

1. Towards a structural status of characters

In Aristotelianpoetics,the notionof characteris secondary,entirely
subordinatedto the notionof plot. There can be fableswithoutchar-
acters,accordingto Aristotle,but therecannot be characterswithout
fables. This view has been upheld by classicaltheoreticians(Vossius).
Later, the character,which untilthen had been nothingbut a name,
the agent of an action,41took on psychologicalconsistency, became an
a a
individual, "person," fully constituted "being," even though he
might remain idle, and of course, even before he acted.42 Character
was no longersubordinatedto action; it became the instantembodi-
ment of a psychologicalessence; such essencescould lend themselves
to inventorieswhich have found theirpurestexpressionin the list of
traditional"roles" of the bourgeoistheater (the coquette, the noble
father,etc.). From the veryfirst, structuralanalysisshowedthe utmost
reluctanceto treat the characteras an essence,even for classification
purposes; as T. Todorov remindsus in his article,Tomachevskiwent
so far as to deny character any narrativesignificancewhatsoever,a
point of view which he toned down subsequently. Withoutgoing so
far as to ignore charactersin his analysis,Propp reduced them to a
simple typology,based not on psychologybut on the homogeneous
nature of the actions assignedto them by the narrative(giver of the
magic object, Assistant,Villain, etc.).
Since Propp, the characterhas kept challengingstructuralanalysis
with the same problem: on the one hand the characters (whatever
the names given to them: dramatis personae or actants) constitute
a necessaryplane of the description, outsideof whichthe commonplace

41 It will be kept in mind that classical tragedydoes not as yet use "character,"
but only "actors."
42 The "person-character" dominates the bourgeois novel; in War and Peace,
Nicolas Rostov is, from the outset, a nice, loyal, courageous young man; Prince
Andrew is high-born,disillusioned,and so forth: what happens to them illustrates
them,but does not make theminto what theyare.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 257

"actions" that are reportedcease to be intelligible,so that it may

safelybe assumed that there is not a single narrativein the world
without"characters,"43or at least without"agents." Yet on the other
hand, thesenumerous"agents" cannot be eitherdescribedor classified
in termsof "persons,"whetherone considersa "person" as a purely
historicformrestrictedto certaingenres (no doubt the best known to
us), thus puttingone under obligationto considerseparatelythe case,
quite considerableindeed, of all the narratives(popular tales, con-
temporarytexts) using agents,but not persons; or whetherone takes
the view that the "person" is but a convenientrationalizationsuper-
imposed by our epoch on otherwisepure narrativeagents. Structural
analysts,scrupulouslyavoiding to define the character in terms of
psychologicalessences,have done theirbest until now, experimenting
with various hypothesesto definethe characternot as a "being" but
as a "participant." To Claude Bremond,each charactercan be the
agent of action sequences that are properlyhis own (Fraud, Seduc-
tion) ; when a singlesequence involvestwo characters-it is the normal
case-the sequence implies two perspectives,or, if one prefers,two
names: what is Fraud to one is Dupery to the other. What it comes
to is that each character,even a secondaryone, is the hero of his own
sequence. Tzvetan Todorov, analyzing a "psychological"novel (Les
Liaisons dangereuses), startsnot fromcharactersbut fromthe three
broad relationshipsin which they are apt to become involved and
which he calls basic predicates (love, communication,assistance).
These relationshipsare examinedby two sortsof rules: the derivation
rules, when other relationshipshave to be taken into account, and
action rules, when it comes to describingthe transformation of the
original relationships in the course of the story. There are many
charactersin Les Liaisons dangereuses,but "what is said of them"
(their predicates) is classifiable.Finally,A. J. Greimas has proposed
to describe and sort out charactersin narrativenot on the basis of
what they are but on the basis of what theydo (hence the name of
actants), inasmuchas theypartakein threemain semanticaxes, which
incidentallyhave their replica in the sentence (subject/object,attri-
butive clause, circumstantialclause), namely communication,desire

43 If a portionof contemporaryliteratureradically interfereswith the "character,"

it is not in order to destroy it (which is not possible), but to depersonalize it
(which is quite different). A novel devoid of any characters, such as Drame, by
Philippe Sollers, turns entirelyaway from the person, to the benefit of language,
but retains neverthelessa fundamental interplayof actants bearing on the speech
acts themselves. This type of literaturedoes not do away with the "subject," but
the "subject" is, fromnow on, the linguisticsubject.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
258 NEW LITERARY HISTORY

(or the quest), and ordeal." Since this participationfalls into pairs
of opposites,the infiniteworld of charactersalso comes under the con-
trol of a paradigmaticstructure(subject/object,giver/recipient, ad-
jutant/opposer),projected on the syntagmatic axis of the narrative;
and since an actant servesto definea class, its role can be filledby
differentactors, mobilized according to rules of multiplication,sub-
stitution,or by-passing.
These threeconceptionshave many pointsin common. The main
point,which should be stressedonce more,is that theydefinea char-
acter by his participationin a sphere of actions,such spheresbeing
limitedin number,typical,and subject to classification.That is the
reason whythe second level of description,thoughconcernedwith the
characters,was called the level of Actions: the word action then is
not to be understoodhere in the same sense as thoseminoracts which
formedthe textureof the firstlevel,but ratheras designatingthe larger
articulationsof praxis (to desire,to communicate,to struggle).

2. The problemof the subject

The problemsraised by a classificationof charactersin narrative
are stillpartiallyunresolved.There is surelya large measureof agree-
ment on the fact that the innumerablecharactersin narrativecan be
subjectedto rulesof substitutionand that,even withinone work,one
single figurecan absorb differentcharacters.45On the other hand,
the actantial model proposed by Greimas (and furtherdeveloped by
Todorov in a different perspective)seems to have withstoodthe test
of accommodatinga great numberof narratives: like any structural
model,its meditdoes not lie so much in its canonic form(a six-actant
matrix) as it does in the regulatedtransformations (by-passes,con-
fusions,duplications,substitutions) to which this model lends itself,
thusraisingthehope ofestablishing an actantialtypologyof narratives.46
However, when the matrix has good classifyingpotential (which is
the case with Greimas' actants), it has more difficultyaccountingfor
the multiplicityof participatoryacts as soon as one startsanalyzing

44 Se'mantiquestructurale(Paris: Larousse, 1966), pp. 129ff.

45 Psychoanalysishas widely accredited these operations of condensation. Mal-
larm6 had already written,in his time, referringto Hamlet: "Secondary figures,
those characters [comparses]must inevitablybe! for,in the ideal mode of picturing
peculiar to the stage, everythingmoves according to a symbolicreciprocitybetween
types or relative to a central isolated figure" (Crayonnli au thadtre,Oeuvres [ed.
Pl6iade], p. 301).
46 For instance: narrativeswhere object and subject are mergedinto one character,
as in narrativescentered on the quest of oneself, of one's identity (L'Ane d'or);
narratives where the subject pursues successive objects (Madame Bovary), and
so forth.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 259

themin termsof perspectives;and whentheseperspectives are respected

(as in of
Bremond'sdescription),the system charactersends up being
too fragmented.The reductionproposedby Todorov avoids both pit-
falls,but it has thus far been applied only to one narrative. It seems
that mostof thesedifficultiescan be smoothedover fairlyrapidly. The
real difficultyone runsinto when classifyingcharactersis the location
(hence the existence) of the subject in any actantial matrix,whatever
its formulation.Justwho is the subject (the hero) of a narrative? Is
thereor is therenota privilegedclass of actors? The Frenchnovelseems
to have built up in us a tendencyto emphasize,one way or another,
sometimesin a devious (negative) way,one particularcharacteramong
others. But thisprivilegedstatusonlyhas a limitedapplicabilitywhen
one considersthe whole of narrativeliterature.Thus, for example, a
great many narrativesset up two opponentsat odds with each other
over the possessionof a stake, and this oppositionhas the effectof
"equalizing" theiractions. The subject then is actually a double sub-
ject, and it cannot be furtherreduced by substitution.This may even
be a widelyused archaic form,as if narrative,emulatingthe practice
of certain ancient languages, recognized as in Greek, a "dual" in
persons. This "dual" is all the more interesting because it pointsout
the affinitybetweennarrativeand the structureof certain (quite mod-
ern) games in whichtwo equal opponentsset out to conquer an object
placed in circulationby a referee. This scheme recalls the actantial
matrixproposedby Greimas,an analogy that is not surprisingif one
pauses to realizethatplay,consideredas a language,possessesthe same
symbolic structureas that found in language and narrative. The
procedureof playingcan be analyzed in the same manner as a sen-
tence.47If we must retaina privilegedclass of actors (the subject of
the quest, of desire,of action), one should at least make it more re-
sponsiveby subjectingsuch an actant to the specificcategoriesof the
grammatical person, not the psychological. Once more it will be
necessaryto draw closer to the linguisticmodel in order to describe
the personalstance (I/you) as distinctfromthe apersonalstance (he)
of the action, each of these two categoriesbeing furtherdescribable
as singular,dual, plural. It is quite possible that the grammatical
categories of the person (accessible through our pronouns) will
eventuallyhold the keyto the actional level. But since thesecategories
can onlybe definedin relationto discourse,not in relationto reality,48

47 The analysis of the James Bond cycle (Communications, 8 [1966], 77-93)

concernsitselfmore with play than with language.
48 See the analyses of the person by Benveniste (Problemes).

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
260 NEW LITERARY HISTORY

the characters,consideredas unitson the actantiallevel,can only find

their meaning (their intelligibility)if they are integratedinto the
thirdlevel of the description,which we shall call, for the purpose of
this study,the level of Narration (as distinctfrom Functions and
Actions).

IV. Narration

1. Narrative communication

Just as there is, within the narrative,a large exchange function

(enacted by giverand recipient),similarly,in homologicalfashion,the
narrative,viewed as object, is the basis of a communication: thereis
a giver of narrativeand a recipientof narrative. In linguisticcom-
munication,I and you are presupposedby each other; similarly,a
narrativecannot take place without a narrator and a listener (or
reader). This is a banal statement,yet one that has been so far in-
sufficientlyused. No doubt the part of the addresserhas been abun-
dantly paraphrased (commentatorshave studied the "author" of a
novel withoutbeing too concerned incidentallywhetherhe is really
the "narrator"), but when it comes to the reader,literarytheoryshows
more pronouncedmodesty. In fact, the real problemis not how to
probe the narrator'smotivesor measure the effectsthe narrationmay
have on the reader,but ratherto describethe code throughwhich the
narrator'sand thereader'spresencecan be detectedwithinthe narrative
itself.The signsof the narratorseem, at firstglance, more visibleand
more numerousthan the signsof the reader (a narratorsays I more
oftenthan he saysyou) ; in actual fact,the latterare simplyharderto
detectthan the former.Thus each time the narratorstops "represent-
ing" and recountsfactswhichhe knowsperfectly well,thoughtheyare
unknownto the reader,thereoccurs,througha suspensionof themean-
ingfuldimension,a sign of the reading act, for there would not be
much sense in the narrator'sgiving himselfinformation."Leo was
the boss of thisjoint,"49we read in a first-person
novel: thisis a sign
of the reader'spresence,a close approximationof what Jakobsoncalls
the conativefunctionof communication.In the absence of any known
inventoryof such signs,we shall forthe momentset aside the signsof

49 Double Bang in Bangkok. The sentence functionsas "a wink to the wise,"
as if the reader himselfwere being addressed. By contrast,an utterance like "So
then, Leo had just left a few minutes ago" is a sign of the narrator,for it is part
of a line of reasoningfollowedby a "person."

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 261

the reception (importantas they are) to say a word of the signs of

narration.50
Who is the giverof the narrative? Three conceptionsseem to have
been formulatedso far. The firsttakes the view that the narrative
emanatesfroma person (in the fullypsychologicalsenseof the term):
the personhas a name, it is the author,who is the locus of a perpetual
exchange taking place between the "personality"and the "art" of a
perfectly identifiedindividualwho periodicallytakesup the pen to write
a story.The narrative(especiallythe novel) is then no more than the
expressionof an I who existsindependentlyof it. The second con-
ceptionsees the narratoras a sortof omniscient,apparentlyimpersonal,
consciousnessthat tells the storyfrom an all-encompassingpoint of
view, that of God: 51 the narratorstands at the same time inside his
characters (since he knows all that happens in them) and outside
them (since he never identifieswith one more than the other). The
third conception,the most recent (Henry James, Sartre), declares
that a narratormustlimithis storyto what the characterscan observe
or know: the assumptionis that each of the charactersis, in turn,the
transmitter of narrative. All threeconceptionsare inadequate in that
they seem to considerthe narratorand the charactersas real, "living"
persons (the unfailingpotencyof this literarymythis well known),
assumingfurtherthat narrativeis originallyconstitutedat the refer-
ential level (these again are "realist" conceptions). Now, at least
fromour viewpoint,both narratorand charactersare essentially"paper
beings." The livingauthor of a narrativecan in no way be mistaken
forthe narratorof that narrative;32 the signsof the narratorare em-
bedded in the narrative,hence perfectlydetectableby a semiological
analysis.But in orderto argue that the author himself(whetherhe is
obtrusive,unobstrusive, or surreptitious)has signsat his disposalwhich
he can scatterthrough work,one mustpositbetweenthis "person"
his
and his language a strictcomplementaryrelation which makes the
author an essentialsubject, and narrativethe instrumentalexpression
of that subject. This assumptionstructuralanalysisis loath to make.
The one who speaks (in the narrative) is not the one who writes(in
real life) and the one who writesis not the one who is.53
50 In "Les categories," Todorovdealswiththenarrator's and thereader'simages.
51 "Whenwillsomeonewritefromthepointof viewof a joke, thatis to say the
way God sees eventsfromabove?" (Flaubert,Prefacea la vie d'ecrivain[Paris:
Seuil, 19651,p. 91).
52 A distinction given the wide scope of this analysis,
all the more necessary,
because, historically, a considerablemass of narrativesis withoutauthors (oral
narrative,folktales,epic poemsentrusted to bards,to recitors,
etc.).
53 J. Lacan: "Is the subjectto which I referwhen I speak the same as the
onewhospeaks?"

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
262 NEW LITERARY HISTORY

In fact, like language, narrationproper (or the narrator'scode)

admits of only two systems: personal and apersonal (impersonal).
These two systemsdo not necessarilybenefitfromthe linguisticmarks
attached to the person (I) and to the nonperson(he). For example,
some narratives,or at least some episodes,can verywell be writtenin
the third person, although their real stance is neverthelessthe first
person. How are we to decide? All one has to do is to rewritethe
narrative(or the passage) fromthe he to the I: as long as thisopera-
tiondoes notentailany alterationof thediscourseotherthan thechange
of grammaticalpronouns,we can be certainthatwe are stillin a person
system. The beginningof Goldfinger,althoughwrittenin the third
person,is in fact "spoken" by Bond. When testingwhetherthe stance
has changed, the decisive factoris that the rewritingthen becomes
impossible;thus the sentence"he saw a man in his fifties, stillyoung
looking . . ." is perfectlypersonal, in spite of the he ("I, James Bond,
saw . . ."), but the narrative utterance "the tinkling of the ice cubes
againstthe glass seemedto awaken in Bond a sudden inspiration"can-
not be consideredpersonal,on account of the verb "to seem," which
becomes a sign of apersonality(not on account of the he). There is
no doubt thatthe apersonalmode is the traditionalmode of narrative,
language havingworkedout a whole tensesystempeculiar to the nar-
rative (articulatedon the aoristm), designedto eliminatethe present
of the personwho is speaking. "In narrative,"writesBenveniste,"no-
body speaks." Yet the personal stance (under various guises) has
gradually found its way into narrative,narrationbeing broughtto
bear upon the hic et nunc of the locutionaryact (indeed thisis exactly
the definitionof the personal system). As a result,narratives,even
some of the most common types,will be found to interminglethe
personal with the apersonal mode at a veryfast tempo, oftenwithin
the limitsof one sentence. For example, the followingsentencefrom
Goldfinger:
His blue-gray apersonal
eyes personal
into
werelookingintently
duPont'seyesmakinghim personal
lose hiscountenance
forthissteadygaze evokeda
mixtureofingenuousness,irony apersonal
and self-depreciation.

54 Benveniste,Problimes.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 263

This intermingling of the two systemsis obviouslyfeltas a faciledevice.

Such a practice can become faking. A detectivestory by Agatha
Christie (Five twentyfive) manages to keep the riddle alive only by
cheating on the person of the narration: a character is described
fromwithin,eventhoughhe is alreadythemurderer.55 Everything hap-
pens as if a witnessingconsciousness,belongingto discourse,could be
made to coincide withina single person,with a murderer'sconscious-
nessinherentin the referent.Only throughthistrickyjugglingwiththe
two systemscan the riddlebe keptalive. It is thusunderstandablethat,
at the other pole of literature,writersof fictionshould have made
the commitmentto a rigorousand consistentsystemof narrationone
of the necessaryconditionsof a work, without always having been
able, however,to meet the challenge.
This rigor-sought after by certain contemporarywriters---is not
necessarilyan aestheticimperative; what is generallycalled a psy-
chological novel is usually characterizedby a mixtureof the two sys-
tems,mobilizingin turn the signs of the nonpersonand the signs of
the person. Indeed, "psychology"-there lies the paradox-cannot
long surviveon a pure person system,for if the whole narrativeis
reducedto the narrationalstance,or if one prefers,to the illocutionary
act, then the very contentof the person is threatened: indeed, the
psychologicalperson (belongingto the referentialorder) has nothing
to do with the linguisticperson, which is never defined by natural
dispositions,intentions,or personalitytraits,but only by its (coded)
pointof insertionin thediscourse. It is thisformalpersonwhichtoday's
writersare tryingto express. We are faced here with an important
subversion(confirmedincidentallyby the readingpublic,who has the
impressionthat no one writesnovelsany more), forit is aimed at con-
vertingthe narrativefrom the order of pure observation (which it
occupied until now) to the performativeorder, wherebythe mean-
ing of a speechact becomestheveryact bywhichit is uttered.56Today,
writingis not "telling"; ratherit signifiesthat one is telling,thereby
making the whole referent("what is being said") contingentupon
this illocutionaryact. This is why part of contemporaryliteratureis
no longerdescriptivebut transitive,strivingto achieveso pure a present

55 Personal mode: "It even seemed to Barnaby that nothing looked changed,
etc." The device is even more blatant in The Murder of Roger Akroyd,since the
murdereris simplymade to say I.
56 On the performativemode, see Todorov, "Les cat6gories." The classical
example of a performativeis "I declare war," a speech act which "records" or
"describes" nothing, but derives its entire meaning from the fact that it is being
uttered (by contrast: "The king declared war," actually records or describes
something).

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
264 NEW LITERARY HISTORY

in speech that the whole of discoursebecomes identifiedwith the act

that deliversit, the whole logos being reduced-or extended-to a
lexis.57

2. The narrative situation

It can then be said that the narrationallevel is occupied by the
signs of narration,which reintegratefunctionsand actions into the
narrativecommunication,the latterbeing articulatedby its giverand
its recipient. Some of the these signs have already been studied. In
oral literatures,certain codes of recitationhave been figured out
(metricformulae,conventionalprotocolswithregardto presentation),
and it is knownthat the "author" is not the one who inventsthe most
beautifulstories,but the one who achieves the greatestmasteryover
the code he shares with his audience. In these oral literatures,the
narrationallevel is so clear-cut,its rulesso binding,that it is difficult
to conceive a "tale" withoutthe coded narrativesigns ("Once upon
a time,"etc.). In our writtenliteratures, the "formsof speech" (which
are in fact narrationalsigns) have been identifiedearly: among them
the classificationof modes of authorialinterventions,
outlinedby Plato,
continuedby Diomedes,58the coding of beginningsand endingsof nar-
rative, the definitionof various stylesof representation(the oratio
directa,the oratio indirecta,with its inquit, the oratio tecta),59the
studyof "pointsof view," and so forth. All theseelementsare part of
the narrationallevel. To these,of course,must be added the writing
processas a whole, for its role is not to "transmit"the narrative,but
to makeit conspicuous.
Indeed, it is in that self-emphasisof narrativethat the units at the
lowestlevel take on theirfullsignificance.This ultimate,self-designat-
ing, formof narrative[i.e., the narrationallevel] transcendsboth its
contentsand its properlynarrativeforms(functionsand actions). This
explainswhythe narrationalcode should be the last level to be reached
by our analysis; going any furtherwould be oversteppingthe limit
of narrative-as-objector transgressingtheimmanencerulewhichunder-
lies thisanalysis. Narrationcan indeed receiveits meaning onlyfrom
the world which makes use of it: beyondthe narrationallevel begins

57 On the opposition between logos and lexis, see Genette's "Frontieres" (Com-
munications,8).
58 Genus activum vel imitativum (no interferencewith discourse on the part
of the narrator: the theater,for instance); genus ennarrativum(the poet alone is
entitled to speak: aphorisms, didactic poems); genus commune (a mixture of
the two: the epic poem).
59 H. Sorensen,Melanges Jansen,p. 150.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 265

the externalworld,that is to say othersystems(social, economic,ideo-

logical) which no longer include narrativesonly, but elements of
another substance (historical facts, determinations,behaviors, etc.).
Justas linguisticsstops at the sentence,the analysisof narrativestops
at the analysis of discourse: fromthat point on, it is necessaryto
resortto anothersemiotics. Linguisticsis aware of this kind of limit
which it has already postulated-if not really explored-under the
name of situations.Halliday definesthe "situation" (in relationto the
sentence) as the body of nonassociatedlinguisticfacts;6 Prieto,as the
body of factsknown by the receiverat the momentof the semic act
and independentlyof this act.61 In the same way, one can say that
any narrativeis contingentupon a "narrativesituation,"or body of
protocolsaccordingto which the narrativeis "consumed." In the so-
called "archaic societies,"the narrativesituationis coded to a very
high degree;62 nowadays, only "avant-garde" literaturestill dreams
of providingprotocols,spectacularprotocolsin the case of Mallarme,
who wanted the book to be recitedin public according to a precise
combinatorialscheme. So, too, Butorprovidestypographicalprotocols,
punctuatinghis books with his own signs. But, as a rule, our society
tends to de-emphasizethe coding of the narrativesituationas much
as possible: there are innumerablenarrationaldevices which tryto
naturalizethe ongoingnarrative,artfullypresentingit as the product
of natural circumstances,and divestingit, as it were, of its decorum.
Epistolarynovels,so-calledrediscoveredmanuscripts,authorswho hap-
pen to have met the narrator,filmswhich run the beginningof their
storybeforeidentification of the cast, all are devices for naturalizing
the narrative. This reluctanceto dramatize its codes is peculiar to
bourgeoissocietyand the mass cultureto which it has givenrise: both
insiston having signs that do not look like signs. Yet this is only a
structuralepiphenomenon,as one mightsay: howevercommonplace,
howevercasual the gesturethe reader or writermakes upon opening
a novel or a newspaper or turningon a televisionset, nothingcan
preventthis modestact fromimplantingin him, all of a sudden and
in its entirety,the narrativecode that he is going to need. In this
way the narrationallevel plays an ambiguous role: contiguouswith
the narrativesituation (and even sometimesincluding it), the nar-
rational level opens out into the world where the narrativeis con-

60 M. A. K. Halliday, "Linguistique genbrale,"p. 6.

61 L. J. Prieto, Principes de Noologie (The Hague: Mouton, 1964), p. 36.
62 A tale, as Lucien Sebag points out, can be told in any place at any time, but
not a mythicalnarrative.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
266 NEW LITERARY HISTORY

sumed. Yet, at the same time,acting as a keystoneto the preceding

levels,thislevel closesthe narrative,constitutes
it once and forall, like
the speech act of a language which anticipatesand even carries its
own metalanguage.

V. The Systemof Narrative

Language propercan be definedby the concurrenceof two funda-

mental processes: the processof articulation,or segmentation,which
produces units (this correspondsto form,according to Benveniste),
and the processof integration,which collectstheseunitsinto units of
a higherrank (this constitutesthe meaning). This double processhas
its counterpartin the language of narrative,which also recognizesan
articulationand an integration,a formand a meaning.

1. Distortion and expansion

Form in narrativeis markedessentially by two governingforces: the
dispersionof signs throughoutthe storyand the insertionof unpre-
dictable expansions among them. These expansions appear as op-
portunitiesfor freedom; nevertheless, it is in the nature of narrative
to absorbsuch "discrepancies"as a part of itslanguage.63
Sign distortionsexist in language, and Bally analyzed them in his
comparativestudyof Frenchand German;64dystaxie[dystaxy]occurs
as soon as the signs (of a linguisticmessage) are no longerjuxtaposed,
as soon as the linear (logical) order is disturbed (for instance the
predicateprecedingthe subject). One typicalformof dystaxyoccurs
when the different partsof one sign are separatedby othersignsalong
the chain of the message (for instancethe negativene jamais and the
verb a pardonnein: elle ne nous a jamais pardonne) : the signbeing
fractured,its signifiedis distributedamong severalsignifiers,separated
fromeach other,none of which can be understoodby itself. As we
have seen when dealing with the functionallevel, that is exactlywhat
happens in the narrative: the unitsof a sequence may forma whole at
the level of this particularsequence, and yet be separated fromeach
otherbytheinsertionof unitsfromothersequences.As notedpreviously,

63 In Valery's terms, "From a formal standpoint, the novel is similar to the

dream: both can be defined as embodying this curious property: all their dis-
crepancies are organic to them."
64 Charles Bally, Linguistique ge'ndraleet linguistiquefrangaise,4th ed. (Berne,
I965).

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 267

the structureof the functionallevel is that of a fugue.65Accordingto

the terminologyof Bally, who contrastssyntheticlanguages, where
dystaxypredominates(as in German), with analyticallanguages that
are morerespectful of the logical linearorderand monosemy,narrative
would be a highlysyntheticallanguage, based essentiallyon a whole-
within-the-whole and overlappingsyntax. Each point in the narrative
radiates in several directionsat a time: when James Bond orders a
whiskeywhile waiting for the plane, this whiskeyconsideredas an
index takes on a polysemicvalue; it is a sortof symbolicnode which
attractsand combinesseveral signifieds(modernity,wealth, leisure).
But consideredas a functionalunit, the orderingof a whiskeymust
workitsway throughseveralrelays(consumption,waiting,departure)
beforeit reachesitsfinalmeaning: the unitis "claimed" by the whole
of narrative,yet on the other hand, the narrative"hangs together"
only throughthe distortionand irradiationof its units.
Generalized distortiongives the language of narrativeits unmis-
takable character: because it is based on a relation,oftena distant
one, and because it mobilizesa sortof implicittrustin one's intellective
memory,distortionis a purelylogical phenomenon,and as such,it con-
stantlysubstitutesmeaning for the pure and simple facsimileof nar-
rated events. In "life," when two people meet,it is veryunlikelythat
one person's requestto "have a seat" would not immediatelybe fol-
lowed by the other person'stakingthat seat; in narrative,these two
units,proximatefroma mimeticpoint of view, may well be separated
by a long sequel of insertionspertainingto quite different functional
spheres. Thus a sort of logical time comes to prevail, bearing little
resemblanceto real time, the apparent fracturingof units being still
closelysubordinatedto the logic whichbinds togetherthe nuclei of the
sequence. Suspense is evidentlybut a privileged,or, if one prefers,an
exasperatingformofdistortion:on the one hand,by keepinga sequence
open (throughemphatic devices such as delays and reactivations),it
securesthe contactwith the reader,thus managing an obviouslycom-
municativefunction;on the otherhand, it holds over him the threat
of an uncompletedsequence, of an open paradigm (if, as we believe,
all sequences have two poles), that is to say, a logical disorder. It is
this disorderwhich is consumed with that particularanguish tinged
with delight (the more to be savored, since it is always straightened
out in the end). Suspenseis, therefore, a way of gamblingwithstruc-

65 Cf. Levi-Strauss (Structural Anthropology,p. 234): "Relations originating

in the same cluster may appear at wide intervals,when viewed froma diachronic
perspective." A. J. Greimas, for his part, insisted on the dispersed nature of func-
tions (Se'mantique structurale).

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
268 NEW LITERARY HISTORY

ture,withthe ultimategoal being,as it were,to riskand to glorifythe

structure.Suspenseis the intelligiblemade problematic;by represent-
ing order (not the serial typeof order) in its fragility,
it achieves the
very idea of language. What is the
ostensibly most patheticis also the
most intellectual: the appeal of suspense is to the "mind," not to
the "bowels."66
Dispersed along the narrative,the functionalnuclei leave interstitial
gaps betweenthem,which may be filledalmostindefinitely; interstices
can accommodatea greatnumberof catalyses. However,at thispoint
a new typologymay be introduced,for the catalysticfreedomcan
be regulated,firstaccordingto the contentof the functions(some func-
tions are more apt to develop catalyses than others, for instance,
waiting67); second, according to the substance of narrative (writing,
as a medium,has a potentialfordieresis-hence catalysticpossibilities
-far superiorto that of a film: it is easier to "freeze" an enunciated
gesture than its visualized counterpart68). The catalysticpotential
of narrativefindsa corollaryin its elliptical potential. On the one
hand, a function ("he had a substantialmeal") can economically
replace all the virtualcatalysesit containsimplicitly(the detailsof the
meal) ;69 on the otherhand, it is possibleto reduce a sequence to its
nuclei,and again 'a whole hierarchyof sequencesto its principalterms,
withoutalteringthe meaningof thestory.A narrativecan be identified
even ifone reducesitstotalsyntagmto itsactantsand major functions.70
In other words, narrativelends itselfto summary (what used to be
called the argument). At firstglance, this seems to be the case with
any kind of discourse;but each type of discoursehas its own type of
summary. A lyricalpoem, for instance,is a vast metaphorpossessing
66 J. P. Faye writes,referringto Baphomet by Klossovski: "Rarely has fiction(or
narrative) so clearly revealed what it always is, by necessity: an experiment of
'thought'on 'life' " (Tel Quel, 22 [1955], 88).
67 Logically waiting comprisesonly two nuclei: (i) the setting up of the terms
of waiting; (2) the fulfillmentor frustrationof the waiting process; but the first
nucleus is subject to extensive catalysis, sometimes a self-perpetuatingcatalysis
(Waiting for Godot): a furtherinstance of gambling with structure,this time
carried to the extreme.
68 In Val&ry's terms, "Proust separates-and gives one the feeling of being
able to separate indefinitely-what other writersare accustomed to leap over."
69 Here again, specificationsaccording to substancemust be anticipated: literature
has an unmatched potential for ellipsis-which the movie lacks.
70 Such a reduction does not necessarilycorrespond to the division of the book
into chapters; on the contrary,chapters seem to assume more and more a dis-
junctive role, setting up break-offpoints, that is, built-in suspense devices (a
favoritewith serial publications).

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 269

a single signified;71to sum it up means to reveal the signified,an

operationso drasticthatit causes the identityof the poem to evaporate
(when summarized,lyricalpoems are reduced to the signifiedsLove
and Death), hence the widespreadbeliefthat paraphrase or summary
of a poem is impossible. By contrast,the summaryor paraphrase of
a narrative (if carried out according to structuralcriteria) preserves
the individualityof the message. In otherwords,narrativeis reducible
without fundamentaldamage. What remains untranslatableis de-
terminedonly at the last level, the narrationallevel. The signifiers
of narration [narrativite],for instance,cannot easily be transferred
fromnovelto film,forthelatterhardlyever72makesuse of the personal
treatment.As for the last layer of the narrationallevel, namely the
idiosyncraticmode of writing[ecriture],it cannot be translatedfrom
one language to another (or, if it can, the resultsare poor at best).
The translatability of narrativeis inherentin the structureof its lan-
guage; it would then be possible,if one proceeded in reverse,to find
one's way back to this by isolatingthe (unequally) reducible
structure,
elementsfrom the irreduciblein narrative. The existencetoday of
differentand concurrentsemiotics (literature,movies, comics, radio
broadcasting) would greatlyfacilitatethisanalyticalprocedure.

2. Mimesis and meaning

In the "language" or narrative,the second importantprocess is
integration: what has been disjoined at a certainlevel (a sequence,
forinstance) is joined togetheragain at a higherlevel (whetherit be a
sequence elevated in the hierarchy,a signifiedsubsumingwidelyscat-
teredindices,or an action affectinga whole class of characters). The
complexityof a narrativecan be compared to that of an organigram
[organigramme],73 capable of integratingbacktrackingand forward
leaps; or, morecorrectly,integrationmakesit possibleto compensatefor

71 According to Ruwet ("Analyse structurale,"p. 82): "A poem can be under-

stood as the result of a series of transformationsapplied to the proposition 'I love
you.' " As it happens, Ruwet is here referringto the analysis of the paranoiac
fantasy made by Freud in relation to President Schreber (Five psychoanalyses).
72 Once more, there is no relation between the grammatical "person" of the
narrator and the "personality" (or subjectivity) which a film producer may
incorporateinto the presentationof a story: the I-camera (continuouslyidentifying
with the eye of a character) is an exceptional case in the historyof movie-making.
73 [Organigram: a stemma (or diagram) that enables one to grasp visually the
various kinds of relationshipsthat bind togethermembersof a complex,hierarchically
structured organization. Whereas a diagram represents some structure which
exists in the physical sense (part of a machine, or plant, or organ), an organigram
sets forthan abstract set of relations, such as the hierarchy of command in the
armed forces,or in a large corporation,or in the judiciary system. Tr.]

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
270 NEW LITERARY HISTORY

the seeminglyuncontrollablecomplexityof unitssituatedon one level.

Integrationhelps directthe comprehensionof fragmentedelements,at
once contiguousand heterogeneous(as theyoccurin thesyntagmwhich
respondsonlyto one dimension: succession). If, withGreimas,we call
isotopythe signifyingunit (for instance,one which pervades a sign
and itscontext),we shall thensay thatintegration is an isotopicfactor:
each (integrated) level impartsits isotopyto the units of the lower
level,and preventsthe meaningfrom"hangingloose"-a consequence
that would surelytake place if one did not perceivethe overlappingof
the levels. Yet narrativeintegrationdoes not offerthe appearance
of smooth regularity,like that of a fine architecturaldesign which
would lead, from the infinitevarietyof simple elementsthrougha
symmetricalnetworkof detours,up to a few complex masses; a unit
oftenappears as a singleunit,yetit may have two correlates,one at a
certainlevel (a functionwithina sequence), the other at a different
level (an index pointingto an actant). Narrativethus appears as a
succession of tightlyinterlockingmediate and immediate elements;
dystaxyinitiatesa "horizontal"reading,whileintegration superimposes
on it a "vertical" reading. There is a sort of structural"limping,"a
constantinterplayof potentials,whose "falls" impart"tone" or energy
to the narrative. Each unit is perceivedas a surfacetexture,while an
in-depthdimensionis maintained,and in this way narrative"moves
along." Throughthe concurrentuse of thesetwo dimensions,structure
branches out, proliferates, becomes exposed-then folds upon itself:
what is new neverceases to be what is expected. There is, of course,
a sort of freedomof narrative (similar to the freedomexperienced
by any speakerwith regardto language), but thisfreedomis limited
in a literalsense: by the stringent code of language at one end, by the
stringent code of narrativeat the other,witha sortof slack in between:
thesentence. If one triesto encompassthe whole rangeof writtennar-
rative,one finds at firstthe most systematiccoding procedure (the
phonematic,or even meristematic level), next a progressiverelaxation
till one reaches the sentence (which representsthe highestdegree of
combinatorialfreedom), then tensionis progressively resumed,with
smallergroupsof sentences(micro-sequences)stillenjoyinga measure
of freedom,and culminateswith broader actions which formamong
themselvesa stringentand restricted code: the creativityof narrative
(at least under its mythical,"life-stimulating" appearance) would
thenbe situatedbetweentwo codes, the linguisticcode and the trans-
linguisticcode. That is why it can be said, paradoxically,that art
(in the Romanticsense of the term) is a matterof enunciatingdetails,
whereasimaginationinvolvesa masteryof the code: "In brief,"said

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
AN INTRODUCTION TO THE STRUCTURAL ANALYSIS OF NARRATIVE 271

Poe, "it willbe seen thattheingeniousman is alwaysfullof imaginative

potentialand that the trulyimaginativeman is neveranythingbut an
analyst. . . ."74
Thus the claim that "realism" is the prime motivationof narrative
must be largelydiscounted. As Bond, on duty in his office,picks up
a telephonecall, he "said to himself,"so we are told by the author:
"Calls fromHong-Kong are just as bad as ever and just as bad to get."
Now, the real informationdoes not lie eitherin the "he said to him-
self" or in the bad quality of the telephoneservice; perhaps this con-
tingencywill make thingslook more "alive," but the trueinformation,
the informationthat will springup fromits seed later, is the tracing
of the call back to its origin,namelyHong Kong. So in any narrative,
imitationremains contingent.75The functionof narrativeis not to
"represent";it is to put togethera scene which still retainsa certain
enigmaticcharacterforthe reader,but does not belong to the mimetic
order in any way. The "reality" of a sequence does not lie in the
"natural" order of actions that make it up, but in the logic that is
unfolded,exposed,and finallyconfirmed, in the midstof the sequence.
To put it anotherway, the originof a sequence is not the observation
of reality,but the necessityto vary and to outgrow the firstform
that man ever came by, namelyrepetition: a sequence is essentiallya
wholewithinwhichnothingis repeated. Logic heretakeson an emanci-
pating value--and so does all the narrative,which restson it. Men
may keep reinjectinginto narrativewhat theyhave known,what they
have lived; but if theydo, it is througha formwhich has conquered
repetitionand instituteda model for a "becoming." Narrative does
not make people see, it does not imitate;the passionthat may consume
us upon readinga novelis not thatof a "vision" (in fact,strictlyspeak-
ing, we "see" nothing). It is the passion to discovermeaning,it is a
strivingtowards a higher order of relation, which also carries its
emotions,its hopes, its threats,its triumphs. What goes on in a nar-
rative is, fromthe referential(real) point of view, strictlynothing.76
What does "happen" is language per se, the adventureof language,
whose advent never ceases to be celebrated. Althoughwe know little
more about the originsof narrativethan we know about the origins
of language, it can reasonablybe argued that narrativeis contempo-

74 Le double Assassinat de la rue morgue,tr. Charles Baudelaire.

75 Genette rightlyreduces mimesis to bits of inserted dialogues; and even at
that, there is always a semantic, rather than mimetic, function lying hidden in
dialogue.
76 "A dramatic work displays the succession of the outer effectsof human acts,
so that no moment in that succession can ever retain its reality and, all things
being considered, nothing happens" (Mallarme, Crayonne, p. 296).

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions
272 NEW LITERARY HISTORY

raneous with monologue,whose emergenceseems to be posteriorto

that of dialogue. In any case, even withoutstretching
the phylogenetic
theory,it may be significantthat man's offspringshould have "in-
vented,"at thesame time (around the age of three), both thesentence
and Oedipus' narrative.

]COLE PRATIQUE DES HAUTES ETUDES,

PARIS

(Translatedby Lionel Duisit)

Note from the translator: The translator of modern critical theory is often
caught between his desire to respect the integrityof an author's original text,
particularlyhis terminology,and the necessityto be understood without imposing
on the reader the use of a specialized glossary. In the presentcase, whenever faced
with termsthat have no English equivalent, I have tried to avoid using approximate
substituteschosen fromalready existingEnglish terms. It is hoped that a carefully
controlled context, with an occasional substitutegiven in brackets at the time of
firstoccurrence, has aided the understanding of such terms as actants, indices,.
informant,atemporal scheme, organigrams,etc., which have, since 1966, gained
currencyamong the proponentsof structuralanalysis.
Although this essay was originally published in Communications, 8 (1966),
it has remained one of the key documents in the study of narrative. The present
authorized translationis the firstto be published in English.

This content downloaded from 138.51.244.230 on Mon, 19 Oct 2015 19:42:27 UTC
All use subject to JSTOR Terms and Conditions