Introduction to the Structural Analysis of Narratives

Introduction to the Structural Analysis of Narratives

ROLAND BARTHES

The article reprinted here was originally published as the introduction to no. 8 of Communications, perhaps the most memorable issue of the pathbreaking French journal and one generally considered to be a manifesto of the French structuralist school. This issue, wholly devoted to the structural analysis of narrative, included seminal essays by A.-J. Greimas, Claude Bremond, Christian Metz, Tzvetan Todorov and Grard Genette. In their semiological work, these critics were indebted to a variety of sources: structural linguistics, the Prague School, Russian formalism, structural anthropology and so on. But their most direct influences were Vladimir Propp Morphology of the Folktale ( 1928) and Lvi- Strauss' Structural Anthropology ( 1958, trans. 1963).

In his introductory essay, Roland Barthes proposed his own deductive model for the structural analysis of narrative at discourse level, closely following the example of generative linguistics. Rejecting all kinds of thematic approach, he aims at the construction of a 'functional syntax' theoretically capable of accounting for every conceivable type of narrative. He bases his model on Propp's concept of 'function' as the structural unit governing the logic of narrative possibilities, the unfolding of the actions performed by the characters and the relations among them. Barthes' model improves on Propp's in that it offers the notions of 'levels of description' and the logic of vertical ('hierarchical') integration of narrative instances, which prefigure those of Genette and Bal. Barthes also contends that traditional classifications of character types are unsatisfactory because they rely excessively on the privileging of one particular kind of character: the subject. In line with Todorov and Greimas, he proposes to void the notion of 'character' of its humanistic connotations in favour of the functional notion of agent or 'actant'. Anticipating the importance given

by reader-response criticism to the narratee, Barthes defines narrative communication as an exchange between narrator and listener. He stresses the peculiarities of literary enunciation and insists on the differentiation between narrator (who speaks in the narrative), implied author (who writes), and real author (who is). Barthes' later phase is generally considered to veer towards a post-structuralist concern with desire, the pleasure of the text, the critique of cultural stereotypes and a looser, more contextualized and particularized approach, for instance, in his reading of a short story by Balzac in S/Z.

The narratives of the world are numberless. [ . . . ] Able to be carried by articulated language, spoken or written, fixed or moving images, gestures, and the ordered mixture of all these substances; narrative is present in myth, legend, fable, tale, novella, epic, history, tragedy, drama, comedy, mime, painting (think of Carpaccio Saint Ursula), stained glass windows, cinema, comics, news item, conversation. Moreover, under this almost infinite diversity of forms, narrative is present in every age, in every place, in every society. [ . . . ] Caring nothing for the division between good and bad literature, narrative is international, transhistorical, transcultural: it is simply there, like life itself. [ . . . ]

Faced with the infinity of narratives, the multiplicity of standpoints -- historical, psychological, sociological, ethnological, aesthetic, etc. -from which they can be studied, [ . . . ] the Russian Formalists, Propp and Lvi-Strauss, have taught us to recognize the following dilemma: either a narrative is merely a rambling collection of events, in which case nothing can be said about it other than by referring back to the storyteller's (the author's) art, talent or genius -- all mythical forms of chance -- or else it shares with other narratives a common structure which is open to analysis, no matter how much patience its formulation requires. There is a world of difference between the most complex randomness and the most elementary combinatory scheme, and it is impossible to combine (to produce) a narrative without reference to an implicit system of units and rules.

Where then are we to look for the structures of narrative? Doubtless, in narratives themselves. [ . . . ]

Thus, in order to describe and classify the infinite number of narratives, a 'theory' [ . . . ] is needed and the immediate task is that of finding it, of starting to define it. Its development can be greatly facilitated if one begins from a model able to provide it with its initial terms and principles. In the current state of research, it seems

reasonable ¹ that the structural analysis of narrative be given linguistics itself as founding model.

1. The Language of Narrative

1. Beyond the sentence

As we know, linguistics stops at the sentence, the last unit which it considers to fall within its scope. [ . . . ]

And yet it is evident that discourse itself (as a set of sentences) is organized and that, through this organization, it can be seen as the message of another language, one operating at a higher level than the language of the linguists. ² Discourse has its units, its rules, its 'grammar': beyond the sentence, and though consisting solely of sentences, it must naturally form the object of a second linguistics. For a long time indeed, such a linguistics of discourse bore a glorious name, that of Rhetoric. As a result of a complex historical movement, however, in which Rhetoric went over to belles-lettres and the latter was divorced from the study of language, it has recently become necessary to take up the problem afresh. The new linguistics of discourse has still to be developed, but at least it is being postulated, and by the linguists themselves. ³ This last fact is not without significance, for, although constituting an autonomous object, discourse must be studied from the basis of linguistics. [ . . . ]

Structurally, narrative shares the characteristics of the sentence without ever being reducible to the simple sum of its sentences: a narrative is a long sentence, just as every constative sentence is in a way the rough outline of a short narrative. Although there provided with different signifiers (often extremely complex), one does find in narrative, expanded and transformed proportionately, the principal verbal categories: tenses, aspects, moods, persons. Moreover the 'subjects' themselves, as opposed to the verbal predicates, readily yield to the sentence model. [ . . . ] Language never ceases to accompany discourse, holding up to it the mirror of its own structure -- does not literature, particularly today, make a language of the very conditions of language?

2. Levels of meaning

From the outset, linguistics furnishes the structural analysis of narrative with a concept which is decisive in that, making explicit immediately what is essential in every system of meaning, namely its organization, it allows us both to show how a narrative is not a simple sum of propositions and to classify the enormous mass of elements which go to make up a narrative. This concept is that of level of description.

A sentence can be described, linguistically, on several levels (phonetic, phonological, grammatical, contextual) and these levels are in a hierarchical relationship with one another, for, while all have their own units and correlations (whence the necessity for a separate description of each of them), no level on its own can produce meaning. A unit belonging to a particular level only takes on meaning if it can be integrated in a higher level; a phoneme, though perfectly describable, means nothing in itself: it participates in meaning only when integrated in a word, and the word itself must in turn be integrated in a sentence. ⁴ The theory of levels (as set out by Benveniste) gives two types of relations: distributional (if the relations are situated on the same level) and integrational (if they are grasped from one level to the next); consequently, distributional relations alone are not sufficient to account for meaning. In order to conduct a structural analysis, it is thus first of all necessary to distinguish several levels or instances of description and to place these instances within a hierarchical (integrationary) perspective.

The levels are operations. ⁵ It is therefore normal that, as it progresses, linguistics should tend to multiply them. Discourse analysis, however, is as yet only able to work on rudimentary levels. In its own way, rhetoric had assigned at least two planes of description to discourse: dispositio and elocutio. ⁶ Today, in his analysis of the structure of myth, Lvi-Strauss has already indicated that the constituent units of mythical discourse (mythemes) acquire meaning only because they are grouped in bundles and because these bundles themselves combine together. ⁷ As too, Tzvetan Todorov, reviving the distinction made by the Russian Formalists, proposes working on two major levels, themselves subdivided: story (the argument), comprising a logic of actions and a 'syntax' of characters, and discourse, comprising the tenses, aspects and modes of the narrative. ⁸ But however many levels are proposed and whatever definition they are given, there can be no doubt that narrative is a hierarchy of instances. To understand a narrative is not merely to follow the unfolding of the story, it is also to recognize its construction in 'storeys', to project the horizontal concatenations of the narrative 'thread' on to an implicitly vertical axis; to read (to listen to) a narrative is not merely to move from one word to the next, it is also to move from one level to the next. [ . . . ] It is proposed to distinguish three levels of description in the narrative work: the level of 'functions' (in the sense this word has in Propp and Bremond), the level of 'actions' (in the sense this word has in Greimas when he talks of characters as actants) and the level of 'narration' (which is roughly the level of 'discourse' in Todorov). These three levels are bound together according to a mode of progressive integration: a function only has meaning insofar as it occupies a place in the general action of an actant, and this action in turn receives its final meaning from the fact that it is narrated, entrusted to a discourse which possesses its own code.

II. Functions

1. The determination of the units

Any system being the combinations of units of known classes, the first task is to divide up narrative and determine the segments of narrative discourse that can be distributed into a limited number of classes. In a word, we have to define the smallest narrative units.

Given the integrational perspective described above, the analysis cannot rest satisfied with a purely distributional definition of the units. From the start, meaning must be the criterion of the unit: it is the functional nature of certain segments of the story that makes them units -- hence the name 'functions' immediately attributed to these first units. Since the Russian Formalists, ⁹ a unit has been taken as any segment of the story which can be seen as the term of a correlation. The essence of a function is, so to speak, the seed that it sows in the narrative, planting an element that will come to fruition later -- either on the same level or elsewhere, on another level. [ . . . ]

Is everything in a narrative functional? Does everything, down to the slightest detail, have a meaning? Can narrative be divided up entirely into functional units? [ . . . ] A narrative is never made up of anything other than functions: in differing degrees, everything in it signifies. This is not a matter of art (on the part of the narrator), but of structure; in the realm of discourse, what is noted is by definition notable. [ . . . ] To put it another way, one could say that art is without noise (as that term is employed in information theory). ¹⁰ [ . . . ]

From the linguistic point of view, the function is clearly a unit of content: it is 'what it says' that makes of a statement a functional unit, not the manner in which it is said. This constitutive signified may have a number of different signifiers, often very intricate. If I am told (in Goldfinger) that Bond saw a man of about fifty, the piece of information holds simultaneously two functions of unequal pressure: on the one hand, the character's age fits into a certain description of the man (the 'usefulness' of which for the rest of the story is not nil, but diffuse, delayed); while on the other, the immediate signified of the statement is that Bond is unacquainted with his future interlocutor, the unit thus implying a very strong correlation (initiation of a threat and the need to establish the man's identity). [ . . . ]

Functions will be represented sometimes by units higher than the sentence (groups of sentences of varying lengths, up to the work in its entirety) and sometimes by lower ones (syntagm, word and even, within the word, certain literary elements only). When we are told that -- the telephone ringing during night duty at Secret Service headquarters -- Bond picked up one of the four receivers, the moneme four in itself constitutes a functional unit, referring as it does to a concept necessary to the story (that of a highly developed bureaucratic technology). In fact, the narrative unit in this case is not the linguistic unit (the word) but only its connoted value (linguistically, the word /four/ never means 'four'); which explains how certain functional units can be shorter than the sentence without ceasing to belong to the order of discourse: such units then extend not beyond the sentence, than which they remain materially shorter, but beyond the level of denotation which, like the sentence, is the province of linguistics properly speaking.

2. Classes of units

The functional units must be distributed into a small number of classes. If these classes are to be determined without recourse to the substance of content (psychological substance for example), it is again necessary to consider the different levels of meaning: some units have as correlates units on the same level, while the saturation of others requires a change of levels; hence, straightaway, two major classes of functions, distributional and integrational. The former correspond to what Propp and subsequently Bremond (in particular) take as functions but they will be treated here in a much more detailed way than is the case in their work. The term 'functions' will be reserved for these units (though the other units are also functional), the model of description for which has become classic since Tomachevski's analysis: the purchase of a revolver has for correlate the moment when it will be used (and if not used, the notation is reversed into a sign of indecision, etc.) [ . . . ] As for the latter, the integrational units, these comprise all the 'indices' (in the very broad sense of the word), ¹¹ the unit now referring not to a complementary and consequential act but to a more or less diffuse concept which is nevertheless necessary to the meaning of the story: psychological indices concerning the characters, data regarding their identity, notations of 'atmosphere', and so on. The relation between the unit and its correlate is now no longer distributional (often several indices refer to the same signified and the order of their occurrence in the discourse is not necessarily pertinent) but integrational. In order to understand what an indicial notation 'is for', one must move to a higher level (characters' actions or narration), for only there is the indice clarified: the power of the administrative machine behind Bond, indexed by the number of telephones, has no bearing on the sequence of actions in which Bond is involved by answering the call; it finds its meaning only on the level of a general typology of the actants (Bond is on the side of order). [ . . . ] Functions and indices thus overlay another classic distinction: functions involve metonymic relata, indices metaphoric relata; the former correspond to a functionality of doing, the latter to a functionality of being. ¹²

These two main classes of units, functions and indices, should already allow a certain classification of narratives. Some narratives are heavily functional (such as folktales), while others on the contrary are heavily indicial (such as 'psychological' novels); between these two poles lies a whole series of intermediary forms, dependent on history, society, genre. But we can go further. Within each of the two main classes it is immediately possible to determine two subclasses of narrative units. Returning to the class of functions, its units are not all of the same 'importance': some constitute real hingepoints of the narrative (or of a fragment of the narrative); others merely 'fill in' the narrative space separating the hinge functions. Let us call the former cardinal functions (or nuclei) and the latter, having regard to their complementary nature, catalysers. For a function to be cardinal, it is enough that the action to which it refers open (or continue, or close) an alternative that is of direct consequence for the subsequent development of the story, in short that it inaugurate or conclude an uncertainty. [ . . . ] Between two cardinal functions, however, it is always possible to set out subsidiary notations which cluster around one or other nucleus without modifying its alternative nature: the space separating the telephone rang from Bond answered can be saturated with a host of trivial incidents or descriptions -- Bond moved towards the desk, picked up one of the receivers, put down his cigarette, etc. These catalysers are still functional, insofar as they enter into correlation with a nucleus, but their functionality is attenuated, unilateral, parasitic; it is a question of a purely chronological functionality (what is described is what separates two moments of the story), whereas the tie between two cardinal functions is invested with a double functionality, at once chronological and logical. Catalysers are only consecutive units, cardinal functions are both consecutive and consequential. [ . . . ]

Were a catalyser purely redundant (in relation to its nucleus), it would nonetheless participate in the economy of the message. [ . . . ] Since what is noted always appears as being notable, the catalyser ceaselessly revives the semantic tension of the discourse, says ceaselessly that there has been, that there is going to be, meaning. Thus, in the final analysis, the catalyser has a constant function which is, to use Jakobson's term, a phatic one: it maintains the contact between narrator and addressee. A nucleus cannot be deleted without altering the story, but neither can a catalyst without altering the discourse.

As for the other main class of units, the indices, an integrational class, its units have in common that they can only be saturated (completed) on the level of characters or on the level of narration. They are thus part of a parametrical relation ¹³ whose second -- implicit -- term is continuous, extended over an episode, a character or the whole work. A distinction can be made, however, between indices proper, referring to the character of a narrative agent, a feeling, an atmosphere (for example suspicion) or a philosophy, and informants, serving to identify, to locate in time and space. [ . . . ] Indices involve an activity of deciphering, the reader is to learn to know a character or an atmosphere; informants bring ready-made knowledge, their functionality, like that of catalysers, is thus weak without being nil. Whatever its 'flatness' in relation to the rest of the story, the informant (for example, the exact age of a character) always serves to authenticate the reality of the referent, to embed fiction in the real world. [ . . . ]

Nuclei and catalysers, indices and informants (again, the names are of little importance), these, it seems, are the initial classes into which the functional level units can be divided. This classification must be completed by two remarks. Firstly, a unit can at the same time belong to two different classes: to drink a whisky (in an airport lounge) is an action which can act as a catalyser to the (cardinal) notation of waiting, but it is also, and simultaneously, the indice of a certain atmosphere (modernity, relaxation, reminiscence, etc.). [ . . . ] Secondly, it should be noted [ . . . ] that the four classes just described can be distributed in a different way which is moreover closer to the linguistic model. Catalysers, indices and informants have a common characteristic: in relation to nuclei, they are expansions. Nuclei (as will be seen in a moment) form finite sets grouping a small number of terms, are governed by a logic, are at once necessary and sufficient. Once the framework they provide is given, the other units fill it out according to a mode of proliferation in principle infinite. [ . . . ]

3. Functional syntax

How, according to what 'grammar', are the different units strung together along the narrative syntagm? What are the rules of the functional combinatory system? Informants and indices can combine freely together: as for example in the portrait which readily juxtaposes data concerning civil status and traits of character. Catalysers and nuclei are linked by a simple relation of implication: a catalyser necessarily implies the existence of a cardinal function to which it can connect, but not vice-versa. As for cardinal functions, they are bound together by a relation of solidarity: a function of this type calls for another function of the same type and reciprocally. [ . . . ]

The functional covering of the narrative necessitates an organization of relays the basic unit of which can only be a small group of functions, hereafter referred to (following Bremond) as a sequence.

A sequence is a logical succession of nuclei bound together by a relation of solidarity: ¹⁴ the sequence opens when one of its terms has no solidary antecedent and closes when another of its terms has no consequent. [ . . . ] The sequence indeed is always nameable. Determining the major functions of the folktale, Propp and subsequently Bremond have been led to name them (Fraud, Betrayal, Struggle, Contract, Seduction, etc.). [ . . . ] Yet at the same time they can be imagined as forming part of an inner meta-language for the reader (or listener) who can grasp every logical succession of actions as a nominal whole. [ . . . ]

However minimal its importance, a sequence, since it is made up of a small number of nuclei (that is to say, in fact, of 'dispatchers'), always involves moments of risk and it is this which justifies analysing it. It might seem futile to constitute into a sequence the logical succession of trifling acts which go to make up the offer of a cigarette (offering, accepting, lighting, smoking), but precisely, at every one of these points, an alternative -- -and hence a freedom of meaning -- is possible. Du Pont, Bond's future partner, offers him a light from his lighter but Bond refuses; the meaning of this bifurcation is that Bond instinctively fears a booby-trapped gadget. ¹⁵ A sequence is thus, one can say, a threatened logical unit, this being its justification a minimo. It is also founded a maximo: enclosed on its function, subsumed under a name, the sequence itself constitutes a new unit, ready to function as a simple term in another, more extensive sequence. [ . . . ] What is in question here, of course, is a hierarchy that remains within the functional level: it is only when it has been possible to widen the narrative out step by step, from Du Pont's cigarette to Bond's battle against Goldfinger, that functional analysis is over -- the pyramid of functions then touches the next level (that of the Actions). There is both a syntax within the sequences and a (subrogating) syntax between the sequences together. The first episode of Goldfinger thus takes on a 'stemmatic' aspect:

Etc.

Obviously this representation is analytical; the reader perceives a linear succession of terms. What needs to be noted, however, is that the terms from several sequences can easily be imbricated in one another: a sequence is not yet completed when already, cutting in, the first term of a new sequence may appear. Sequences move in counterpoint; ¹⁶ functionally, the structure of narrative is fugued: thus it is this that narrative at once 'holds' and 'pulls on'. Within the single work, the imbrication of sequences can indeed only be allowed to come to a halt with a radical break if the sealed-off blocks which then compose it are in some sort recuperated at the higher level of the Actions (of the characters). Goldfinger is composed of three functionally independent episodes, their functional stemmas twice ceasing to intercommunicate: there is no sequential relation between the swimming-pool episode and the Fort Knox episode; but there remains an actantial relation, for the characters (and consequently the structure of their relations) are the same. [ . . . ] The level of functions (which provides the major part of the narrative syntagm) must thus be capped by a higher level from which, step by step, the first level units draw their meaning, the level of actions.

III. Actions

1. Towards a structural status of characters

[ . . . ] Structural analysis, much concerned not to define characters in terms of psychological essences, has so far striven, using various hypotheses, to define a character not as a 'being' but as a 'participant'. For Bremond, every character (even secondary) can be the agent of sequences of actions which belong to him (Fraud, Seduction); when a single sequence involves two characters (as is usual), it comprises two perspectives, two names (what is Fraud for the one is Gullibility for the other); in short, every character (even secondary) is the hero of his own sequence. Todorov, analysing a 'psychological' novel (Les Liaisons dangereuses), starts not from the character-persons but from the three major relationships in which they can engage and which he calls base predicates (love, communication, help). The analysis brings these relationships under two sorts of rules: rules of derivation, when it is a question of accounting for other relationships, and rules of action, when it is a question of describing the transformation of the major relationships in the course of the story. There are many characters in Les Liaisons dangereuses but 'what is said of them' (their predicates) can be classified. Finally, Greimas has proposed to describe and classify the characters of narrative not according to what they are but according to what they do (whence the name actants), inasmuch as they participate in three main semantic axes (also to be found in the sentence: subject, object, indirect object, adjunct) which are communication, desire (or quest) and ordeal. ¹⁷ Since this participation is ordered in couples, the infinite world of characters is, it too, bound by a paradigmatic structure (Subject/Object, Donor/ Receiver, Helper/Opponent) which is projected along the narrative; and since an actant defines a class, it can be filled by different actors, mobilized according to rules of multiplication, substitution or replacement.

These three conceptions have many points in common. The most important, it must be stressed again, is the definition of the character according to participation in a sphere of actions, these spheres being few in number, typical and classifiable; which is why this second level of description, despite its being that of the characters, has here been called the level of Actions: the word actions is not to be understood in the sense of the trifling acts which form the tissue of the first level but in that of the major articulations of praxis (desire, communication, struggle).

2. The problem of the subject

[ . . . ] The real difficulty posed by the classification of characters is the place (and hence the existence) of the subject in any actantial matrix, whatever its formulation. Who is the subject (the hero) of a narrative? Is there -- or not -- a privileged class of actors? The novel has accustomed us to emphasize in one way or another -- sometimes in a devious (negative) way -- one character in particular. But such privileging is far from extending over the whole of narrative literature. Many narratives, for example, set two adversaries in conflict over some stake; the subject is then truly double, not reducible further by substitution. [ . . . ] If therefore a privileged class of actors is retained (the subject of the quest, of the desire, of the action), it needs at least to be made more flexible by bringing that actant under the very categories of the grammatical (and not psychological) person. [ . . . ] It will -- perhaps -- be the grammatical categories of the person (accessible in our pronouns) which will provide the key to the actional level; but since these categories can only be defined in relation to the instance of discourse, not to that of reality, ¹⁸ characters, as units of the actional level, find their meaning (their intelligibility) only if integrated in the third level of description, here called the level of Narration (as opposed to Functions and Actions).

IV. Narration

1. Narrative communication

[ . . . ] In linguistic communication, je and tu (I and you) are absolutely presupposed by one another; similarly, there can be no narrative without a narrator and a listener (or reader). Banal perhaps, but still little developed. Certainly the role of the sender has been abundantly enlarged upon (much study of the 'author' of a novel, though without any consideration of whether he really is the 'narrator'); when it comes to the reader, however, literary theory is much more modest. In fact, the problem is not to introspect the motives of the narrator or the effects the narration produces on the reader, it is to describe the code by which narrator and reader are signified throughout the narrative itself. [ . . . ]

Who is the donor of the narrative? So far, three conceptions seem to have been formulated. The first holds that a narrative emanates from a person (in the fully psychological sense of the term). This person has a name, the author, [ . . . ] the narrative (notably the

novel) then being simply the expression of an I external to it. The second conception regards the narrator as a sort of omniscient, apparently impersonal, consciousness that tells the story from a superior point of view, that of God: the narrator is at once inside his characters (since he knows everything that goes on in them) and outside them (since he never identifies with any one more than another). The third and most recent conception ( Henry James, Sartre) decrees that the narrator must limit his narrative to what the characters can observe or know, everything proceeding as if each of the characters in turn were the sender of the narrative. All three conceptions are equally difficult in that they seem to consider narrator and characters as real -- 'living' -- people (the unfailing power of this literary myth is well known), as though a narrative were originally determined as its referential level (it is a matter of equally 'realist' conceptions). Narrator and characters, however, at least from our perspective, are essentially 'paper beings'; the (material) author of a narrative is in no way to be confused with the narrator of that narrative. ¹⁹ The signs of the narrator are immanent to the narrative and hence readily accessible to a semiological analysis; but in order to conclude that the author himself (whether declared, hidden or withdrawn) has 'signs' at his disposal which he sprinkles through his work, it is necessary to assume the existence between this 'person' and his language of a straight descriptive relation which makes the author a full subject and the narrative the instrumental expression of that fullness. Structural analysis is unwilling to accept such an assumption: who speaks (in the narrative) is not who writes (in real life) and who writes is not who is. ²⁰

In fact, narration strictly speaking (the code of the narrator), like language, knows only two systems of signs: personal and apersonal. These two narrational systems do not necessarily present the linguistic marks attached to person (I) and non-person (he): there are narratives or at least narrative episodes, for example, which though written in the third person nevertheless have as their true instance the first person. How can we tell? It suffices to rewrite the narrative (or the passage) from he to I: so long as the rewriting entails no alteration of the discourse other than this change of the grammatical pronouns, we can be sure that we are dealing with a personal system. [ . . . ]

2. Narrative situation

The narrational level is thus occupied by the signs of narrativity, the set of operators which reintegrate functions and actions in the narrative communication articulated on its donor and its addressee.

Some of these signs have already received study; we are familiar in oral literatures with certain codes of recitation (metrical formulae, conventional presentation protocols) and we know that here the 'author' is not the person who invents the finest stories but the person who best masters the code which is practised equally by his listeners: in such literatures the narrational level is so clearly defined, its rules so binding, that it is difficult to conceive of a 'tale' devoid of the coded signs of narrative ('once upon a time', etc.). In our written literatures, the 'forms of discourse' (which are in fact signs of narrativity) were early identified: classification of the modes of authorial intervention (outlined by Plato and developed by Diomedes), ²¹ coding of the beginnings and endings of narratives, definition of the different styles of representation (oratio directa, oratio indirecta with its inquit, oratio tecta), ²² study of 'points of view' and so on. All these elements form part of the narrational level, to which must obviously be added the writing as a whole, its role being not to 'transmit' the narrative but to display it.

It is indeed precisely in a display of the narrative that the units of the lower levels find integration: the ultimate form of the narrative, as narrative, transcends its contents and its strictly narrative forms (functions and actions). This explains why the narrational code should be the final level attainable by our analysis, other than by going outside of the narrative-object, other, that is, than by transgressing the rule of immanence on which the analysis is based. Narration can only receive its meaning from the world which makes use of it: beyond the narrational level begins the world, other systems (social, economic, ideological) whose terms are no longer simply narratives but elements of a different substance (historical facts, determinations, behaviours, etc.). Just as linguistics stops at the sentence, so narrative analysis stops at discourse -- from there it is necessary to shift to another semiotics. Linguistics is acquainted with such boundaries which it has already postulated -- if not explored -- under the name of situations. Halliday defines the 'situation' (in relation to a sentence) as 'the associated non-linguistic factors', ²³ Prieto as 'the set of facts known by the receiver at the moment of the semic act and independently of this act'. ²⁴ In the same way, one can say that every narrative is dependent on a 'narrative situation', the set of protocols according to which the narrative is 'consumed'. In so-called 'archaic' societies, the narrative situation is heavily coded; ²⁵ nowadays, avant-garde literature alone still dreams of reading protocols -- spectacular in the case of Mallarm who wanted the book to be recited in public according to a precise combinatory scheme, typographical in that of Butor who tries to provide the book with its own specific signs. Generally, however, our society takes the

greatest pains to conjure away the coding of the narrative situation: [ . . . ] epistolary novels, supposedly rediscovered manuscripts, author who met the narrator, films which begin the story before the credits. The reluctance to declare its codes characterizes bourgeois society and the mass culture issuing from it: both demand signs which do not look like signs. Yet this is only, so to speak, a structural epiphenomenon: however familiar, however casual may today be the act of opening a novel or a newspaper or of turning on the television, nothing can prevent that humble act from installing in us, all at once and in its entirety, the narrative code we are going to need. Hence the narrational level has an ambiguous role: contiguous to the narrative situation (and sometimes even including it), it gives on to the world in which the narrative is undone (consumed), while at the same time, capping the preceding levels, it closes the narrative, constitutes it definitively as utterance of a language [langue] which provides for and bears along its own metalanguage. [ . . . ]

Notes

See especially B. TOMACHEVSKI, ''Thmatique'' ( 1925), in Thorie de la littrature, ed. T. Todorov ( Paris: 1965), pp. 263-307. A little later, Propp defined the function as 'an act of a character, defined from the point of view of its significance for the course of the action', Morphology of the Folktale ( Austin, Tex. and London: 1968), p. 21.

This is what separates art from 'life', the latter knowing only 'fuzzy' or 'blurred' communications. 'Fuzziness' (that beyond which it is impossible to see) can exist in art, but it does so as a coded element (in Watteau for example). Even then, such 'fuzziness' is unknown to the written code: writing is inescapably distinct.

These designations, like those that follow, may all be provisional.

Functions cannot be reduced to actions (verbs), nor indices to qualities (adjectives), for there are actions that are indicial, being 'signs' of a character, an atmosphere, etc.

N. RUWET calls 'parametrical' an element which remains constant for the whole duration of a piece of music (for instance, the tempo in a Bach allegro or the monodic character of a solo).

In the Hjelmslevian sense of double implication: two terms presuppose one another.

It is quite possible to identify even at this infinitesimal level an opposition of paradigmatic type, if not between two terms, at least between two poles of the sequence: the sequence Offer of a cigarette spreads out, by suspending it, the paradigm Danger/Safety (demonstrated by Cheglov in his analysis of the Sherlock Holmes cycle), Suspicion/Protection, Aggressiveness/Friendliness.

This counterpoint was recognized by the Russian Formalists who outlined its typology; it is not without recalling the principal 'intricate' structures of the sentence.

A. J. GREIMAS, Smantique structurale ( Paris: 1966), pp. 129f.

See the analyses of person given by Benveniste in Problmes.

A distinction all the more necessary, given the scale at which we are working, in that historically a large mass of narratives are without authors (oral narratives, folktales, epics entrusted to bards, reciters, etc.).

J. LACAN: 'Is the subject I speak of when I speak the same as the subject who speaks?'

Genus activum vel imitativum (no intervention of the narrator in the discourse: as for example theatre); genus ennarativum (the poet alone speaks: sententiae, didactic poems); genus commune (mixture of the two kinds: epic poems).

H. SRENSEN in Language and Society (Studies presented to Jansen), ( Copenhagen: 1961), p. 150.

M. A. K. HALLIDAY, ''General linguistics and its application to language teaching'', Patterns of Language ( London: 1966), p. 4.