A narrative is often defined, as we have done here, as "the semiotic representation of a series of events." But there is another more restricted definition which is equally common: according to Bal, "a narrative text is a text in which an agent tells a story" (Narratology 119). Semiotic representation through signs is always the work of an agent, and the narrator is, in this sense, the agent who enunciates the narrative text. The narrative text, then, is a linguistic enunciation like many others. We will draw a basic opposition between the subjects of the enunciation, the characters in the text, and the subject of the enunciating, the instance whose words represent those characters and the rest of the textual universe.
We could at this point draw on a linguistic analogy to introduce an important analytical concept, narrative person. According to Jakobson, the verbal category of person characterizes the protagonists of the enunciation (spoken about) with reference to the protagonists of enunciating (the addresser and the addressee).
A first person form, such as "I," means that the addresser, the main protagonist of the activity of enunciating, is positing himself as the subject of both enunciating and enunciation. A second person form, "you," equally locates the person spoken about with respect to the speaking situation: if "I" does the speaking, "you" is present in some way or other. On the other hand, "he" or "she" are characterized by their absence from the speech situation, if still by reference to it. This verbal category translates easily into narratology, according to the structuralist principle that narrative is an expanded verb or sentence. Just as we have first, second and third person pronouns, we have works of literature in the first and in the third person, and even in the second (many love songs, and even long narratives like La Modification).
So far we might have characterised the enunciative structure of narrative. But here we are dealing with literary narrative, and this adds some further complications. Literature is a linguistic game, a peculiar mode of enunciating. The difference between the subject of the enunciating and the subject of the enunciation, the difference between the speaker and the person spoken about, is not as clear-cut here as it is elsewhere. Who does the enunciating in literary narrative? Sometimes we shall find that it is a character, as in Great Expectations, and sometimes we seem to hear the voice of the author himself, as in El Quijote or Fielding's novels. There is a clear difference in principle between the subject of the enunciating and the subject of enunciation, since all novels have authors and all have characters. But we shall often need to posit an intermediary figure, the narrator, who shares some characteristics of both author and character. The narrator is a bridge between the enunciation and the enunciating, and one of the tasks of literary analysis is to determine the extent to which each of these poles is the more relevant in his composition. The narrator will be defined as the enunciator of a narrative text which nevertheless does not account for the full complexity of the literary work.
The basic problem of this complexity begins to appear at a linguistic, non-literary level, in the phenomenon of quoted speech. Plato was the first to analyze the fact that the "I" of a narrative text need not refer to the author. He separates "simple" narrative from "imitative" narrative. In imitative narrative, the speaker "speaks through somebody else's mouth" and "tries to conform as far as possible to the language of the person in whose name he speaks."
Drama uses imitative narration, dithyrambs use simple narration, and epic poetry uses both: its mode is, therefore, mixed. This is the first critical passage devoted both to the issue of narrative persona and to the question of direct and indirect style, and we must recognize that the problems have some structural as well as historical relation.
Plato did not appreciate imitative narration, since imitations may degrade the speaker if the imitated object is unworthy. But Aristotle will use opposite criteria. If Plato criticized Homer because of his extensive use of direct speech, Aristotle praises him for the same reason. According to him, "the poet should speak as little as possible in his own person, for it is not this that makes him an imitator" (Poetics 63, XXIV.7). This idea that the narrator should efface himself, to let the action unfold dramatically before the audience, to let the characters expand freely with a minimum of narratorial control, returns again with the development of the realist novel: we find different versions of this creed in Spielhagen, in Henry James, in Joyce.
However, we can hardly say that Aristotle has a definite concept of a narrator as an entity different in any way from the poet. In the case of fiction, the difference is obviously easier to establish in the case of first-person narratives. Thus, Wordsworth speaks of the "dramatic parts" of his poems, "those parts when the poet speaks through the mouth of his characters."
These parts must preserve a strict decorum (a decorum of psychological realism, of course--after all, it is a Romantic who speaks). An excessive linguistic elaboration would go against verisimilitude, and is to be reserved in any case for the voice of the poet. Wordsworth wants his lyrical ballads written in the language of men, and that is why he uses narrators so often. But still his narrators are introduced by the voice of the poet. A more advanced concept of a narrator is introduced by Spielhagen, who differentiates in this respect first and third person novels:
In the language of art, we call a novel in which the hero appears as the narrator of his fate a first person novel, in opposition to other novels, where the hero is a third person and we are told of his adventures by the writer.
We see that the aim here is to tell apart those fictions which are told by a fictional character from those in which an authorial voice is in charge. The difference established here, then, is not so much one of author versus narrator as (once again) one of author versus character. There is a "narrator" only when a fictional character tells the story.
Tomashevski further develops the notion of "narrator": in his view, there is a narrator different from the author in those novels which are written in imitation of an oral narrative (skaz), where a specific fictional character tells a story with a language that characterizes him as a specific individual in his own right, not a neutral, self-effacing and transparent medium, or "abstract narrative" (Teoría 253-4).
We have to wait until the New Critics and other immanent students of literature to find the opposition author / narrator extended to the point of being applied to all kinds of novels, in first or third person, with a neutral or an obstrusive speaker. The New Critics consider the literary text as a self-sufficient object, which in order to be understood does not require a knowledge of the author's context or ideas (other than the one provided by the language of the text). Therefore there will be no more talk of authors: instead we find only the implied image of the author provided by the text. This is no longer a flesh-and-blood person properly speaking, but a textual construct, which is called by the critics in a number of ways: dramatic speaker, lyrical subject (in the case of poetry), implied author, author, narrator. Terminology, once again, is confusing, and we should look into a critic's assumptions in this respect, not merely into his set of favourite terms. A typical pronouncement is given by Wolfgang Kayser: "the narrator is not the author . . . ; the narrator is a fictional being the author has turned into."
And for Genette, in a fictional narrative,
the role of narrator is itself fictive, even if assumed directly by the author. . . . The narrator of Père Goriot 'is' not Balzac, even if here and there he expresses Balzac's opinions, for this author-narrator is someone who 'knows' the Vauquer boardinghouse, its landlady and its lodgers, whereas all Balzac himself does is imagine them; and in this sense, of course, the narrating situation of a fictional account is never reduced to its situation of writing. (Narrative Discourse 213).
So, for modern criticism, the very act of writing literature carries along with it a fictionalization of the speaker. The real self of the author becomes to some extent irrelevant, and we understand the work in terms of his "official" self in the institution of literature, the image of the author which emanates from his works.
However, this fact does not rule out the simple phenomenon of narration through the mouth of a fictional character. The real author is not the implied author, all right, but this does not always mean that the implied author is always in charge of producing the narrative text, of being the immediate subject of enunciating. Therefore we have three candidate figures to fill in this subject position: the author, the implied autor, and the narrator. They are perhaps first identified by Barthes, when he argues that in literature "he who speaks (in the narrative) is not he who writes (in life) and he who writes is not he who is."
Booth had already observed that "'Narrator' is usually taken to mean the 'I' of a work, but the 'I' is seldom if ever identical with the implied image of the artist."
--which in turn is seldom if ever identical with the artist.
We can study narrators in many possible ways. Since narrators are, in part, characters, we can study their personality in the way we would study the personality of any other character. Greimas and Courtés provide a systematic framework for the study of the competence of any discursive subject.
We can study the different modalities that bear upon the textual subject and constitute him as such. This analysis can be applied to characters and narrators alike, although the results will obviously vary because of their different positions in the textual structure.
MODALITIES : Virtualising Actualising Performative
Exotatic : MUST CAN DO
Endotaxic : WANT KNOW BE
But we shall postpone this study until we introduce a further level of analysis, the narrative as literary work, in which a new textual subject, the implied author, is introduced. As we shall see, it is helpful to characterize the narrator and the implied autor differentially, with respect to each other's stance and competence. We see that Genette does not introduce this kind of considerations just as he does not introduce the concept of the implied author-a choice of limits which is legitimate for a particular essay, but too restrictive for narrative theory as a whole.
What Genette does very well is to systematize the varieties of narrators according to purely formal criteria: their structural position with respect to the fabula and the different enunciative levels of the work. The two criteria he uses result in the fourfould characterization of narrators into extradiegetic / intradiegetic on one hand, and homodiegetic / heterodiegetic on the other. Before we examine these concepts further, it will be convenient to remember an important difference between different analytical problems: fictionality on one hand, and enunciative hierarchy on the other.
The relationship of fictionality is the one established between real and fictional phenomena. Fiction is a kind of parasitic or alternative reality, one which is grounded in a real world with respect to which it is defined as fiction. So a fictional event or a fictional world can be represented as a framed section in the middle of reality: the frame, by definition, cannot be crossed:
Author / Inventor
Fictional world Fictional
characters and happenings
Fictional characters and happenings
The relation of fictionality is recursive: it can be applied again and again to the object it produces.
Any relationship of embedding can therefore multiply in two directions inside a text:
€ Horizontally, a number of fictional worlds may coexist at the same fictional level. That is, we find a recursivity in length, an enchaining of embedded texts:
€ Vertically, the signified world of a text may include another text or semiotic element which introduces an embedded fictional reality, which in turn may contain a further fictional reality, and so on. That is, we can find fictional words where characters give rise to other fictional worlds, by dreaming them, imagining them or writing about them; the levels of fictionality assume here the form of a Chinese box:
These two modalities of embedding (enchained embeddings or embedded embeddings) can be combined in an infinite number of ways. The fictional relation can establish multiple embeddings, enchainings and hierarchies until a complex pattern of relationships is constituted between the different realities of the discursive activity. An author or inventor, real or fictional, may invent different fictional worlds which are independent from each other, or he may establish further relations of fictionality between those fictional worlds.
Real world Real author
The other hierarchic relationship we mentioned, enunciative hierarchy, is established between a main text and a subordinate text which is embedded inside it. The simplest instance is the use of direct speech, with an introductory speech verb (the main text) and a quoted sentence:
John said, "I can't find my umbrella."
At textual level, this relationship is established between whole texts, and not just between sentences. For instance, the stories in The Canterbury Tales are hierarchically dependent on the main story, which frames them by telling us about the circumstances of their telling and the identity of the different narrators. When the characters in the Miller's Tale speak, their enunciation is hierarchically inferior to the enunciation of the tale by the Miller, just as the Miller's enunciation is hierarchically inferior to Chaucer's. This kind of embedding is also potentially recursive: in Lost in the Funhouse John Barth exploits this multiple embedding with comical effects.
Enunciative embedding can be developed in the horizontal and the vertical relation in the same way as fictional embedding, and it may combine those two modalities of development in just the same way.
It is obvious that these two kinds of embedding are different in nature. The contents of an embedded text may be fictional with respect to those of the main text, but they may also refer to the same (real or fictional) world. Therefore, a change in enunciative level, the introduction of a speaker through the words of another, does not necessarily involve a change in fictional level. Conversely, a change in fictional level does not necessarily involve a change in enunciative level. A fictional world must certainly consist in a semiotic representation of some kind: it is something referred to, signified, rather than something which is present in itself. But this representation need not be made by means of language. A fictional world may appear in a dream, a picture, a film, not only in a literary narrative. The same thing happens when we project the fictional relationship to the inside of a literary narrative: when the narrator tells us of a character's dream, we enter a second-degree fictional world without entering a second-degree narrative.
There is still a third kind of semiotic embedding which appears in narrative texts and which should be kept in mind, even if it not directly related to the discursive position of the narrator. Not all embedded semiotic structures must assume the form of discourses. When speaking of fiction we have already mentioned the possibility of pictures, dreams, etc., which appear as elements of the main fabula. Some of these elements can be used to introduce an embedded fragment which nevertheless refers to the same fictional world of the main narrative (for instance, the description of a photograph in a novel). There is a change in level here, but it is not in enunciative level nor in fictional level. We shall call these changes in semiotic level. Of course, direct speech is also a semiotic device, but it is used to represent speech--itself, in a way. A photograph, on the other hand, cannot be quoted the way a letter is quoted.
Furthermore, if we look back to the concept of perspective or focalization, we shall soon see another possible kind of embedding, an embedding which does not involve semiotic artifacts present in the fabula (verbal or other). What is embedded are different kinds of cognitive structures, or perspectives. Bal speaks in this respect of changes in the level of focalization, usually introduced by verbs of perception or cognition and structured very much like the shifts in enunciative level introduced by speech activity verbs.
We have already introduced the general notion of enunicative level, relative to the use of direct speech in a text. Narrative levels are simply enunciative levels where the quoted speech is a narrative. A character in a story tells a story (about his past, for instance, or a fictional story) and that story is situated at a narrative level which is secondary with respect to the main story.
What Genette calls intradiegetic story is a story within a story,
not only in the sense that the first frames it with a preamble and a conclusion . . . but also in the sense that the narrator of the second narrative is already a character in the first one, and that the act of narrating which produces the second narrative is an event recounted in the first one. (Narrative Discourse 228).
Although the identity of the intradiegetic narrator and his status as a character need not always be that clear, this might be taken to represent the standard situation. It should be kept in mind, however, that many other types of embedded enunciations can be found, in which the embedded element is not a narrative. It may be a poem or a piece of statistics, and even if it is a narrative it can be very different from the main one: for instance, a piece of news report, or a letter, embedded in a novel.
An intradiegetic story can contain another story which is intradiegetic with respect to it. In absolute terms, from the point of view of the complete structure of the work, this story will therefore be intradiegetic in the second degree (metadiegetic, according to Genette). The "main" narrator, the one who introduces the hierarchically superior level, is situated in an extradiegetic position--insofar as he is a narrator; he may, of course, be at the same time a character inside the story and be inside the diegesis in that sense, but Genette will use there the term homodiegetic. There is always an extradiegetic narrator of some kind or other in a narrative work, though there need not be any intradiegetic narrators. It is important to separate this issue of narrative level from the question of narrative person we have mentioned before: an extradiegetic narrator may tell the story in the first or in the third person; in Genette's terms, he may be either homodiegetic or heterodiegetic. The same goes for intradietetic narrators. That is, the opposition intradiegetic / extradiegetic situates the narrator with respect to the whole narrative hierarchy of the work, whereas the opposition homodiegetic / heterodiegetic defines the narrator in terms of his own narration--which need not encompass the whole work.
Genette observes the peculiar use of intradiegetic stories in a whole tradition of narrative writing, and proceeds then to an examination of the main relationships between the embedded narratives and the main text. In his Nouveau discours, he adds some indications by John Barth to distinguish five main types:
1) Causal relationship: when both narratives refer to the same fabula, the intradiegetic narrative may be analeptic and explicative of the events in the main fabula.
2) When we find an intradiegetic proleptic narrative, the function is of prediction.
3) The third type of relationship is purely thematic. This often happens in the case of fictional embedded stories. An extreme form, in Genette's view (Narrative Discourse 233) is the structure of mise en abyme.
4) Sometimes the thematic relationship is made explicit by one of the characters, and the story acquires an explicit exemplary, persuasive value. This is the case, for instance, of exempla in medieval narrative.
5) Sometimes it is not the story but its narrating which establishes the significant connection with the main story: Genette's favourite example is the Thousand and One Nights, where only the narrating keeps the intradiegetic narrator alive. In other cases, the function of the narrating may be purely distractive (as in the Decameron).
These five types are classified according to a greater importance of the narrative act itself. And all of these values may be mixed in a variety of degrees in concrete narratives.
We have said that by definition the barrier between reality and fiction cannot be crossed. This is also the case for the other kinds of barriers that we have studied, deriving mainly from the difference between the sign and the referent. The sign cannot suddenly become its referent--in those cases when it comes, we feel that it should not.
But we have alsoo witnessed the prolifereation of barriers and distinctions within those fictional, represented or quoted worlds. And these barriers, intradiegetic or fictional in the second degree, are no longer impassable. Although they lay a claim to the same logical status as the border between the first and the second level, they are in fact very different: they are textual constructions which can be modified and transgressed at will--as long as we do not care so much about verisimilitude. Genette uses the term metalepsis to refer to the transition from one narrative level to another. In spite of this definition, he seems to include there the transitions from one semiotic level to another and from a fictional level to another as well. But we could divide these "metalepses" into as many kinds as the barriers they overstep. So, if an intradiegetic narrator suddenly becomes extradiegetic (as it happens at one moment in Marguerite de Navarre's Heptaméron) the phenomenon is analogous to, but not to be confused with an illegal mixture of fictional worlds--for instance, in Marguerite Yourcenar's "Comment Wang-Fô fut sauvé", where a painter avoids execution by painting a picture and escaping through that fictional landscape.
We have already introduced the concept of narrative person. Jakobson's definition of the grammatical category of person gave us a starting point:
First person signals the identity of one of the protagonists of the enunciated process with the agent of the enunciating process, and second person his identity with the actual or potential patient of the enunciating process. ("Embrayeurs" 182)
It would be tempting to transpose these grammatical categories directly into the field of literary narrative. But in fact it is impossible, since just as we saw there is no one-to-one connection between verbal tense and narrative temporality, there is no one-to-one connection between grammatical person and narrative person. The most usual terminology opposes first-person narratives to third-person narratives, meaning presumably that the main character is referred to in the first or in the third person. Genette introduces a distinction along different lines: homodiegetic narratives are told by a narrator who is present (though not necessarily as a protagonist) in the story he relates; heterodiegetic narratives are told by a narrator who is absent from that story. The purest form of homodiegetic narrative is autodiegetic narrative (what is normally understood by "first-person" narrative, where the narrator is also the protagonist).
It is to be noted that a heterodiegetic narrator need not be an authorial narrator--nor an "omniscient" one. From Genette's definition, indeed, it is not ruled out that the narrator belongs to the same world (fictional or otherwise) as his characters; it is only required that he must not figure as a character in his narrative. But, as Genette points out, these categories cannot be rigid, since the concept of identity itself is not rigid, but manipulable to some extent at least through discursive activity. And in some kinds of narrative, the problematic borderline between different identities is already given from the start: in autodiegetic narrative, the same "person" is split into two completely different textual roles: hero and narrator. The concepts of homodiegetic and heterodiegetic narration are nevertheless useful as ideal poles, as fixed reference points against which we can measure the ambiguities or displacements that will inevitably occur in most narrative texts.
(End of Narrative Theory)
More Structural Narratology
Cf. Genette, Narrative Discourse 214 n.
 Roman Jakobson, "Les embrayeurs, les catégories verbales et le verbe russe" 183.
 Plato, República 101, III.
 William Wordsworth, "Preface to the Second Edition of Lyrical Ballads" 439.
 Friedrich Spielhagen, Beiträge zur Theorie und Technik des Romans 131. Translation mine.
 Wolfgang Kayser, "Qui raconte le roman?" 72.
 Roland Barthes, "Introduction à l'analyse structurale des récits" 26.
 Wayne Booth, The Rhetoric of Fiction 73.
 A. J. Greimas and J. Courtés, Sémiotique 124.
 Mieke Bal, Narratologie 38; "The Laughing Mice: Or, on Focalization" 203ff.
 The original use of"metalepsis" as defined by Genette's source Pierre Fontanier, covers a range of rhetorical phenomena much wider than the one Genette implies.