In the literal sense, the term “narrator” designates the inner-textual (textually encoded) highest-level speech position from which the current narrative discourse as a whole originates and from which references to the entities, actions and events that this discourse is about are being made. Through a dual process of metonymic transfer and anthropomorphization, the term narrator is then employed to designate a presumed textually projected occupant of this position, the hypothesized producer of the current discourse, the individual agent who serves as the answer to Genette’s question qui parle? The narrator, which is a strictly textual category, should be clearly distinguished from the author (Schönert → Author) who is of course an actual person.
A narrator is a linguistically indicated, textually projected and readerly constructed function, slot or category whose occupant need not be thought of in any terms but those of a communicative role. Terms designating this role include discursive function or role, voice, source of narrative transmission, producer of current discourse, teller, reporter, narrating agent or instance. The position occupied by this presumed inner-textual originator of the discourse functions as a logico-linguistic center for all spatio-temporal and personal references occurring in the discourse, i.e. as highest-level center of the discourse. An inner-textual narrator can in principle be assigned to any narrative text, not just a fictional one, and such ascription does not require any knowledge about the actual world producer of the words of the text, be it a human being or a computer program. The linguistically projected narrator and the actual world producer will be confronted at a later stage (3.6).
Good reasons, stemming from text linguistics, philosophy, narratology and common sense, can be adduced for the necessity or at least advisability of granting the narrator category as defined above a central place in the description and interpretation, both informal and professional, of literary narratives. In Benveniste’s (1966) and Jakobson’s (1971) text linguistics, any utterance is described as consisting of two indissoluble components: the speech event (énonciation, saying) and that which is said (énoncé) to which correspond, respectively, the sayer (sujet de l’énonciation) and the one spoken of (sujet de l’énoncé). Since narrative utterances are a subset of the universe of utterances, they too must therefore contain a sayer. For narrative, the terms thus translate into narration, narrated event, narrator and narrated agent(s), respectively. A narrator can thus be defined as the sujet de l’énonciation of one or more utterances that represents an event (Coste 1989: 166). In terms of linguistic pragmatics or speech act theory, any narrative, regardless of its length, is a macro speech act of the constative type, claiming that such and such happened. For a claim to be made, there needs to be an agent who makes this claim, hence the narrator. If narrative is a report of acts and events, we need a reporter behind it, and if it is a tale, we need a teller. In terms of communication theory, any act of communication consists of a sender sending a message to a receiver. A narrative consists of someone telling someone else that something happened, and no such act can be imagined without a sender-narrator position. Even a failed, confused or contradictory act of reporting presupposes a narrator no less than a successful one.
Plato was the first to claim that the underlying difference between narrative and drama as basic types of discourse consists in the difference between directly showing and indirectly telling or reporting, rooted in the absence or presence respectively of a mediating instance between the characters’ speech and the audience. And the narrator is precisely this mediating instance. Modern arguments for mediacy as the generic hallmark of narrative can be found in Friedemann ( 1965) and Stanzel ( 1971). In contemporary narratology it is customary to distinguish between three functions which are essential to give rise to any narrative: doing, seeing and saying (Bal 1981: 45). Thus, characters do certain things which are viewed from a certain perspective, and what is seen is then reported. To these three functions there correspond three roles: narrative agent, focalizer (which has been a subject of scholarly controversy) and narrator. Baxtin’s ([1934/35] 1981) influential theory of the novel, which can be generalized to all narrative, regards the novel as the site of interplay between two kinds of utterances: those stemming from the characters and those stemming from an inner textual narrator. The whole essence of narrative would be missed if one were to deny the textual existence of a narrator as a stylistic and ideological position. Finally, psychonarratology (e.g. Bortolussi & Dixon 2003) has shown that readers process literary narratives in the same way as they do ordinary communication insofar as they assume a textually encoded conversational partner responsible for the contents of the narrative. This mimetic-illusionist assumption a bout the nature and status of the narrator has recently come under scrutiny by cognitively-oriented narratologists (Nünning 2001; Fludernik 2003; Herman → Cognitive Narratology). On this view, both narrated world and narrator are not inherent to the text, but rather constructed in readers’ minds at the point of intersection of individual textual data and general cognitive categories possessed by these readers .A literary narrative is consequently a text capable of creating in the reader’s mind the representational illusion of observing an ongoing process of narrative communication in which a more or less personalized narrator plays a key role. Identifying and characterizing such a narrator is an optional naturalization or meaning creation strategy open to the reader and building upon two kinds of input: textual signals and storytelling scenarios (frames, schemes) the reader already possesses from his real-life experience and which are activated once a certain number of narrator indicators have been identified in the text (Emmott & Alexander → Schemata). Works which destroy the illusion of an independently existing narrated domain may still produce a powerful representational illusion of narrative activity with a narrator figure behind it. One can say in conclusion that the notion of narrator has been approached and defined in terms of three distinct theoretical frameworks (Grall 2007): rhetoric (speech act, communication); narratology (mediation, interplay of utterances); and cognitive science (reader psychology and models of text comprehension).
Some narrators are more marked and individuated than others. But what are the minimal textual conditions under which one could identify a distinct narrating position or voice? Such conditions could be represented as a hierarchical series. The text must be capable of being naturalized as representing one or more reporting utterances or speech acts stemming from one or more agents. Some texts, classified as narratives in our culture, such as unframed interior monologues (Schnitzler’s Fräulein Else) or textes-limites of modernism or postmodernism, do not satisfy this requirement and consequently cannot be considered as possessing any inscribed originators. The second condition is that it should be possible to demarcate the utterances of which the text consists and assign each of them to a distinct voice or originator. (It is only in rare cases that all utterances recorded in a text were originally made by one speaker at one time.) The third condition is that one should be able to determine the hierarchical relations between the different utterances and their originators, as defined by such questions as who can quote whom, who can refer to whom and who can report about whom (Margolin 1991), but also to determine the total number of such originators and levels of speech in the text. Finally, and most crucially, one should be able to identify a single, highest-level originator of all originators, so to speak: one general, primary or global textual narrating voice, such that (a) the text as a whole can be seen as a macro speech act or utterance emanating from that voice, and (b) all textually occurring utterances originating with other speakers are embedded within this macro speech act, that is, are merely quoted or mentioned in it. There is no algorithm for deciding whether any or all of the above conditions are satisfied by a given text even though readers make such decisions semi-intuitively all the time. The muse who provides the answer to the epic question at the beginning of the Iliad is the earliest Western example of such a global narrator, but this occurs also with the anonymous voices relating the whole of War and Peace or Père Goriot. When it is not possible to identify a single highest-level narrator, we are dealing with multi-narrator narratives (Faulkner’s As I Lay Dying or The Sound and the Fury, Orhan Pamuk’s My Name is Red) in which different textual segments consist of reports stemming from different speakers, none of whom occupy a position higher than the others. “Narrator” in the prototypical sense, however, designates the single, unified, stable, distinct human-like voice who produces the whole narrative discourse we are reading. In general, although not universally, this discourse assumes the shape of an account of independently existing and known facts. Going one step further, the narrator can be envisioned as a fictional agent who is part of the story world and whose task it is to report from within it on events in this world which are real or actual for him (Thomson-Jones 2007: 78).
Richardson (2006) described the difficulty in pinning down and defining a single or unified or stable highest-level narrator position in many postmodern texts, even though they contain numerous signs of narrator and narrational activity alike. In such texts, of which Beckett’s Trilogy is the showcase, it is sometimes impossible to locate a constant highest-level narrator, and even if one is locatable, this utterer has no voice of his own or is mimetically impossible. The first case involves either a constant reversal of levels between quoter and quoted where “the one you invented has invented you” (Beckett), or an open-ended regression of levels where whenever we think we have finally reached the primary textual speaker, the unquoted quoter, it turns out that this discourse, too, is in fact being quoted by a still higher-level voice. In the second case, the highest-level speaker is a mere conduit or “mouth” (Beckett) voicing a discourse whose inscribed originator is someone else, so that all tokens of “I” in this discourse designate not the utterer, but that “cantankerous other” (Beckett). The net result is that “I seem to speak, it is not I; about me—but it is not about me” (Beckett). The supposedly highest-level voice ends up lacking all identity, as it is merely a “ghost writer” for another or the mere conduit for another’s discourse or an impersonator speaking as another (Margolin 1986/87). In the mimetically impossible case (Richardson 2006: 103–05), the primary speaker turns out to be a number of distinct voices which merge without any explanation, which contain so much incommensurable information that they cannot be unified into one speech position or whose level is indeterminate and floating between the character, narrator and persona of the biographical author, as when such a narrator claims to have invented figures in other texts by the same author (e.g. Beckett’s Trilogy). Finally, a specific highest-level individual voice cannot be identified in a discourse consisting of a verbal collage of recycled clichés from media reports, advertising and the like (Petersen 1993: 138).
When a primary global narrator can be defined for a given narrative, the discourse as a whole can be viewed as its macro speech act. Individuating the narrator in a literary fictional context means constructing or inferring an image of the utterer with the sole means for so doing being the verbal record of his speech act. This task needs to be guided by two theoretical frameworks: linguistic pragmatics, which seeks to define the time, place, and context of utterance and the utterer’s capabilities, beliefs and communicative intentions; and the cognitive psychological theory of attribution, which seeks to infer from a behavior, including verbal, the dispositions and attitudes of the agent (Margolin 1986). Now literary texts vary enormously as regards the kinds and the amount of clues they provide for this purpose and the resultant textual markedness of the narrator or “degree of narratorhood” (Chatman’s term). At degree zero we have the impersonal or transparent mode of narration associated with an anonymous voice or covert (effaced, imperceptible) narrator coming from nowhere and announcing categorically that “once upon a time there was.” At the other end stands the perceptible, dramatized or personal mode of narration associated with an overt narrator who could say things like “Living now in my old age in the city of NN, I still remember with great affection what X did 30 years ago.” Obviously, the greater the number and diversity of the textual elements available for speaker indication, the richer the resultant speaker image. Once again, the two extremes would be a mere voice with no psychological person behind it and a concrete figure with both an inner life and a body.
One major source of data for building the image of the narrator is claims occurring in his/her discourse that go beyond the strict reporting of individual facts. These include summaries, analyses, comments, and generalizations of various kinds, all concerning the narrated domain. Chatman (1978) has proposed a useful typology of such claims in ascending order from set descriptions and temporal summaries to reports of what characters did not do, say or think, then to explanations, interpretations and judgments of reported actions or characters, and ending with generalizations of any kind, including purported general truths, maxims and norms of action which go well beyond the reported events. The extent of such claims varies enormously from one author to the other, two extremes being Hemingway and Henry James. The aesthetic desirability of such narratorial “intrusions” or “telling” beyond mere “showing” has been the object of heated critical debate since the 19th century (e.g. Otto Ludwig , Friedrich Spielhagen  1967, Käte Friedemann  1965, Percy Lubbock  1947 and Wayne C. Booth  1983). Critics for whom narratorial mediation is a mere handmaiden for showing camera-like what happened would advocate the avoidance of all such material and consider it a mere deviation detracting from the effectiveness of the narration. Conversely, those for whom mediation is the very essence of narrative as distinguished from drama would consider such material as radical enrichment of “mere reporting.”
The types of utterances just mentioned help us individuate the narrator as a mind, so to speak. But what about him/her as a person in a communicative situation? Here linguistic features play the major role. Doležel (1967) has outlined several such features, again in hierarchical order. First is the use of first- and second-person pronouns to indicate the presence of the originator and the inscribed addressee of the current speech event, both of whom are absent in third-person discourse. Next is the use of all three major tenses, especially of the present tense, to indicate the current communicative transaction relative to which all narrated events are temporally ordered. In pure third-person past-tense narration, on the other hand, the past tense is not related to any particular speech situation, but is more aspectual, merely indicative of the narrated events already having taken place. Third is the use of deictics (demonstratives, indexicals, shifters) of time and place such as “now,” “here,” “lately,” “yesterday,” etc. relating the narrated events to a present speaker and his embodied space-time position. Another major element is address to the inscribed narratee, such as the famous “Dear reader,” consisting of questions and admonitions and providing the speaking voice with immediacy, projecting an ongoing communicative exchange (telling) in addition to what is being narrated (told). Such address is part of the rhetorical strategy employed by the narrator, and embodies his/her communicative intentions. Equally important is the use of subjective semantics, expressing the narrating instance’s attitudes and reactions to the narrated events, which both adds a strong personal element and functions as part of the teller’s rhetorical strategy vis-à-vis the addressee. A final individuating feature is a personal style of narration, indicative of a particular mind style.
Narratorial comments are focused on the narrated, while the linguistic features listed above may be indicative of the narrated or the narration. The fullest systematic picture of elements in the communicative situation (narration) which help characterize the narrator can be provided by using Jakobson’s model of verbal communication (1960), five of whose six functions are concerned with enunciation. The expressive function is concerned with the speaker’s self-reference, self-characterization, and expression of emotions and attitudes. The conative or appellative functions may create the illusion of face-to-face communication where the addressee is urged to listen, understand, sympathize, etc., not only with respect to the narrated, but also regarding the narrator and his current activity. Metalinguistic references to the medium employed (oral or written) and its limitations again highlight the narrator’s present act of telling, and so do discussions of the appropriateness and potentialities of the type of discourse selected (letter, diary, confession, report). And finally, there are of course references to the current narrating activity and its linguistic embodiment as it is being produced.
As Prince (1982) and Nünning (1989) have noted, the greater the number of signs of the narration compared to those of the narrated, the more marked the narrator and his activity become. An extreme example is provided by postmodern narratives where hardly any story gets told, since most of the text is concerned with the process of telling and its difficulties and with the figure of the teller and his struggle to tell (Neumann & Nünning → Metanarration and Metafiction). Finally, when the telling process is foregrounded and presented as durative (taking days, months or years), it is possible to draw conclusions not only with respect to some of the narrator’s mental and physical traits, but also as regards possible changes to these features in the course of the narration.
Once a certain amount of individuating information about the narrator has been garnered from the textual data listed above, one could attempt to draw an image of this narrator as a human or human-like figure. Now in principle any physical, mental or behavioral aspect of the narrator could enter such a picture, but as regards those aspects most closely tied to the defining teller role, the following have been suggested by various narratologists: degree and kind(s) of knowledge possessed; reliability; relation to various components of the speech act performed; articulateness; attitude towards the narrated (straightforward, ironic, sympathetic, etc.); projected teller role.
Once a global narrator has been identified in a discourse, all information about the narrated domain, including characters’ direct discourse, originates with that narrator. Now the knowledge a narrator may have about any of the characters may be restricted to what can be garnered from sense impressions, or it may include direct access to their minds, something not possible outside fiction (® focalization). Even if restricted to external data, a narrator may know more, the same as or less than one or more of his characters, and he may also withhold information from his addressee. One egregious example is Agatha Christie’s The Murder of Roger Ackroyd, where the narrator withholds the crucial information that he himself is the murderer. Some, but by no means all, anonymous narrating voices telling their story in the third-person past tense are endowed with omniscience: “Familiarity, in principle, with the characters’ innermost thought and feelings; knowledge of past, present and future; presence in locations where characters are supposed to be unaccompanied […]; and the knowledge of what happened in several places at the same time” (Rimmon-Kenan  2003: 96). And such panoramic or Olympian knowledge can be fully authoritative, not open to any challenge or enquiry. This is the maximal degree and kind of knowledge any narrator can possess, and the possibility of any narrating instance possessing such knowledge is the most basic constitutive convention of all fiction writing. As soon as the narrator becomes personalized, knowledge claims begin to be restricted in scope and kind to the humanly possible (unless the speaker is a supernatural entity) and are open to modalization (“it seems,” “probably,” “as far as can be known”) and thus the challenge of limited epistemic authority. Because of their rhetorical needs, authors sometimes endow personalized narrators with intermittent omniscience. The highly personalized narrator of Proust’s first-person novel À la Recherche du temps perdu can thus on occasion report with certainty about what another person thought or what happened when someone was all by himself.
Personalized narrators, and only personalized ones, may on occasion be deemed by the reader as unreliable, meaning that the validity of some or even all claims made by them is low or non-existent, that these claims need consequently to be rejected and, if possible, replaced by more valid, reader-formulated ones regarding the given topic. (Notice, though, that if the narrator is cast in the role not of a reporter of facts but of an inventor of tales, unreliability is inapplicable [Walton 1990: 374–75].) Following Phelan & Martin (1999), one can distinguish three axes of unreliability: facts and events of the narrated domain; the interpretation of such facts (i.e. supplied inferences, explanations or motivations); moral, practical, aesthetic, etc. judgments and evaluations of these facts. While the first two kinds of reliability are epistemic, the third is clearly axiological and normative. Moreover, unreliability of factual claims is the most radical, since it may prevent us from figuring out what the narrative world was “really” like. A narrator may himself alter the reliability of any of his claims by citing lack of information or inability on his part to fathom things. There are numerous indicators of narratorial factual unreliability (cf. D’hoker & Martens eds. 2008) including paratextual and intertextual elements such as title (Diary of a Madman) or a narrator figure falling clearly under a codified unreliable literary type (picaro, scoundrel). In multiple narrator texts (3.4), conflicts between the reports on the same events by different narrators indicate that at least one of them is unreliable. In realistic literature, a major clash between our world knowledge (extra-textual information) and claims made by the narrator may also serve as such an indicator (Hansen 2007). Inner-textual indicators of factual unreliability are inconsistency and incongruity between claims made by the narrator regarding the same events, while illogicality, invalid or non-sequitur inferences as well as explanations and generalizations lacking any evidence are grounds for deeming narratorial interpretations of fact unreliable. Strong conflict with the moral or aesthetic norms held by the reader are grounds for rejecting narratorial judgments. In the factual and interpretative cases, one also assumes that the events of the narrated domain are in and by themselves amenable to a consistent description and that valid generalizations and explanations of this domain are possible. Narratorial unreliability is ultimately a readerly computational hypothesis adopted in order to explain the origin of inconsistencies and incongruities in the narrated world, a crucial point first made by Yacobi (1981). To claim that the narration of a given story is unreliable is to assume the existence of a personalized mediator with human-like cognitive and sensory capabilities whose erroneous or aberrant mind can serve as the source of all textual incongruities with respect to the narrated domain (Marcus 2007).
Once we are ready to psychologize the narrator, we could seek for mental explanations for the unreliability of some or all of his claims. Depending on the particular text, such grounds could be the narrator’s lack of knowledge or experience, mental deficiencies ranging from limited intelligence to insanity or drug-induced hallucinations, self-deception (in cases of autobiographical narration), a particular mental disposition (the chronic liar), and a deliberate deceptive strategy. Creating a narrator figure whom readers will deem unreliable redirects attention from the told to the telling and the teller, from what is known and evaluated to the circumstances and activities of informing and judging, and to the person failing to perform them properly.
From the speech act of narration one can construct an image of its performer along three major axes: status, involving the speaker’s relation to his speaking activity; contact, involving the speaker-audience relation; and stance, involving the relation between the speaker and the topic of his discourse. Such is the key thesis of Lanser (1981), the most comprehensive account to date of the narrator in terms of speech act theory. Status covers, among other things, social identity, extent of knowledge, presentation of the told as report or invention, and “mimetic authority” encompassing sincerity and honesty or their absence, trustworthiness (both intellectual and moral), and competence or skill at telling. Contact includes the teller’s attitude towards his inscribed addressee: formality to intimacy, deference to contempt; self-reference and direct address or the absence of both; the teller’s attitude towards his activity including self-confidence or hesitancy, consciousness of this activity of telling and reference to it or lack of both. Stance is a more heterogeneous category, but most important probably is the narrator’s relation to his characters: adopting or not adopting their language and/or spatio-temporal perspective and/or values (Lanser 1981: 224). Lanser’s pragmatics of narration follows in the footsteps of classical rhetoric where a speaker is regarded as a human subject with various emotions (pathos), values (ethos) and intentions and who, through the organization and manner of delivery of his discourse, seeks to mold in particular ways the attitudes, emotions and judgments of his addressees (Grall 2007: 253–54).
Under this heading is understood the manner of telling, especially those stylistic choices that help characterize the speaker’s discourse and, by metonymic transfer, the speaker’s mind as sophisticated, abstract, complex and rational or their opposite, finely nuanced or simplistic, emotional and immediate or rational and distanced, and so on. While such qualifications cannot be strictly defined in any systematic and exhaustive manner, they form an important part of our personality sketch of the narrator as perceiver, chronicler and analyst of the narrated world. Our corresponding judgment of him as intelligent and perceptive or not will have a decisive influence on our assessment of his credibility and ultimately on how much of what he claims about the narrated domain we are ready to accept.
Equally incapable of formal definition and failsafe determination, yet every bit as important, is the narrator’s attitude towards the told, as manifested in the way characters and events are represented. An open-ended list of qualifiers would include neutral vs. judgmental, sympathetic vs. detached, involved vs. distanced, cynical, sentimental, emotionally charged, curious, amused, bewildered, and so on. The relation between the tone or manner of telling and its subject matter can itself serve as the basis for second-order characterization of the speaker. Speaking in a cold, distanced manner about an atrocity may lead us to characterize the speaker as heartless or as doing his best to hide his emotions, depending on the context (Margolin 1986). The drawing of such inferences is not an exact science, for it depends on the specific inner-textual contexts as well as on the reader’s cultural context; even so, such inferencing plays an important role in any portrait of the narrator drawn by the reader.
The last key aspect of the narrator’s image is his/her textually projected role. Is the narrator presented as a reporter (chronicler, biographer, historian, eye witness) who vouches for the truth of his assertions regarding the narrated? Or as an editor or publisher transmitting and vouching for the prior existence and/or authenticity of the documents (letters, diaries) he is presenting (though not necessarily for the veracity of the claims made in them)? Or as an author-fabricator, a storyteller engaged in the invention of stories, perhaps with a playful attitude? Or maybe as an oral teller, as in the skaz tradition, presenting a story to a live audience with a focus on the performative or transmissive aspect, on oral address and unmediated audience response? (For the underlying functions, see Ryan 2001; for the key properties of the narrator in his teller role, see Booth  1983 and Petersen 1993.)
Some narratives do not have a general or global narrator, so that the events on the narrated level are related by numerous independent partial narrators, neither of whom refers to the discourse of the others, thus creating “narrative parataxis” (Coste 1989: 173). Now these partial narrators need not be participants in the narrated events, as when three contemporary historians tell the story of Napoleon’s defeat in Russia. Furthermore, each of them may narrate a different phase of the total action sequence, a pattern labeled “narrative relay” (Coste 1989: 173), or the same events may be covered by all of them in converging or diverging ways. In fictional narratives, one encounters both patterns, but with the difference that the narrators are normally also participants in the events they narrate. Since each character-narrator possesses his own perspective or “take” on the events, the net result is multi-perspectival narration where there exist two or more narrating instances who portray the same events in different ways, each from his own standpoint (Nünning 2001: 18). An epistolary novel consisting entirely of correspondence between two or more persons is a plurivocal narration in which each letter writer reports on and discusses events concerning himself, his addressee or some third party. An epistolary novel with a framing editor’s discourse turns this editor into the global narrator, since all the embedded letters are basically quoted by him, the text as a whole constituting a two-level narrative.
In general, a narrative can comprise several hierarchically ordered levels of narration, each with its own narrator. In such cases, the primary narrator is the one who introduces or quotes all the others, without himself being introduced by any of them; the secondary narrator is introduced or quoted by the first and introduces in his turn all lower-level narrators, and so on. This story-within-a-story phenomenon has been described by Nelles (2005: 134) as a “structure by which a character in a narrative text becomes the narrator of a second narrative text framed by the first one,” i.e. where one narrator’s discourse embeds that of another at a subordinate level. While the primary narrator may remain a disembodied voice, all lower-level narrators are characters with respect to the primary one and must therefore be individuated to some degree with respect to verbal, mental, behavioral and physical features. Embedded narrators (Coste & Pier → Narrative Levels), too, can function either as reporters, in which case issues of reliability are paramount, or as storytellers, where their skill at story telling and its impact on their destiny are key (Walton 1990: 369–72).
When a narrator employs tokens of “I” and “you” in his discourse, these tokens automatically refer to him in his current speaker role and to his inscribed addressee as participants in the ongoing communicative transaction. But these tokens may also refer to speaker and addressee as entities existing beyond the sphere of narration as objects of telling (=characters, narrative agents) in the narrated sphere. And as characters (Jannidis → Character), they may be located at points in space and time beyond the narration’s here and now. Insofar as narrators have themselves as narrative agents, they are engaged in producing a first-person narrative, whereas if it is their addressees who act as narrative agents, a second-person narrative is being produced. If the entities referred to in the narrator’s discourse are not part of the current communicative situation, then a third-person narrative is produced. (Note that it is quite possible to have a third-person narrative in which the speaker and the addressee in their communicative roles are quite prominent.) Put differently, the referents of first- and second-person narratives participate in both story and discourse systems and those of third-person narratives in the story system only. Using the narrated system as our point of departure, the main distinction is between narratives in which the narrator also participates in the narrated events (first-person narrative) and those in which he does not (second- and third-person narratives).
Several unusual forms of narration merit special attention with regard to the narrator-character relation. One is the impersonal “one” form where the pronoun can designate anyone and everybody who is or would be in the situation portrayed, including the narrator himself. But this particular pronoun endows narration with a depersonalized aura. The “you” form automatically picks out the inscribed addressee and can pick out any reader who is ready to put himself imaginatively in this addressee’s situation. But what if the narrator’s claims are about a “you” in a separate narrated sphere, possibly also distinct in space and time from the narrational sphere? Why tell the addressee his own life story? And how possible is it for a personalized narrator to have access to this “you’s” interiority? No one motivation is possible, only a series of local context-dependent ones (Fludernik ed. 1994). “We”-narrative concerning not speaker and addressee, but rather the speaker and other(s) in a distinct narrated sphere, is especially tricky. “We” is always I+other(s). So is it the whole group speaking in unison, like the chorus in Greek tragedy, or one speaker only? And if this speaker is one, is he an authorized spokesperson for the group? “We”-narratives may serve as tools for constructing a group’s sense of cohesion and identity, but mental access by the we-narrator is necessarily curtailed. Since we=I+other(s), whenever a text using a first-person plural pronoun seeks to depict the thoughts of other(s) beyond the speaker, it necessarily straddles the line between first- and third-person narration: a character discloses that which can only be known by an external, impersonal intelligence, that is, an omniscient narrative voice. Such narratives are thus simultaneously first- and third-person discourses, transcending this basic narratological divide (Richardson 2006: 60).
When speaking about his own discourse, the narrator normally adopts his own current epistemic and evaluative perspective, although he can adopt the presumed perspective of his inscribed addressee, as in: “Is it ever boring, listening to you.” When making claims about the narrated domain, the narrator can engage his own perspective, but alternatively he may take on the perspective of a character, speaking “[a]s though he himself were […] in the epistemological position he attributes to a character, reporting what he takes this character to know” (Walton 1990: 379). In the case of the autodiegetic (=autobiographical) narrator, the character whose epistemological position he adopts may be himself at a different time, usually in the past, but possibly also a projected future version of himself. In his study of Dostoevskij’s poetics, Baxtin ( 1984) showed the myriad ways in which a character’s perspective can be incorporated into the narrator’s discourse, ranging from harmony up to sharp internal dissonance and parodic inversion. Free indirect discourse, one of the hallmarks of fiction writing, is a linguistic form combining the narrator’s deictic position and the character’s idiom and semantics. Finally, a narrator can speak of himself qua narrative agent as of another, that is, in the second or third person (e.g. Caesar in De bello gallico). The reasons for such a deictic shift are numerous and local, but the transfer can never encompass the whole text; otherwise, it will not be identifiable.
The narrator figure presented so far is the one postulated or constructed in standard, classical narratology. As we have seen, it is based on three assumptions, namely, that for every work of narrative fiction:
(a) there exists a specifiable inner-textual, highest -level speech or communication position functioning as the point of origin of the current discourse. In other words, all narrative fiction is communication.
(b) There is always an individual figure or agency occupying this speech position and thus backing the assertions contained in the narrative discourse or presenting the fictional world to us (Kania 2005).
(c) This individual figure or agency exists on the strictly fictional level, and is a distinct entity within the fictional universe projected by the text.
Thus, in writing down his text, the flesh and blood author gives rise to a substitutionary speaker who performs the macro speech act of reporting and who is solely responsible for all claims, specific or general, made in this report (Ryan 1981, Martínez-Bonati 1996). In writing down his actual text and communicating it to actual readers, the author thus projects or evokes the image of an act(ivity) of narrative communication between a fictional narrator and his fictional narratee(s). (Schmid 2005: 45-6). This fictional narrator is assumed to be a constituent of every work of narrative fiction and hence a universal, indispensable component of any narratological model. Note that the three claims listed above form a hierarchical order, so that one cannot assert (c) without asserting (a) and (b), or assert (b) without asserting (a). Conversely, one can reject (c) and yet maintain (a) and (b), or deny both (c) and (b) and still keep (a). And of course one can deny all three claims. Over time, and even more so in the last decade, challenges to one or more of these three assumptions or claims have been issued by linguists (Banfield, Kuroda), philosophers (Hamburger, Currie, Wilson, Kania), and literary scholars (most prominently Patron , but also Walsh  and Koeppe & Stuehring ). All of these challenges deny at least the pan-narrator claim, the claim about the “ubiquity of the non-actual fact telling narrator” (Alward), no matter how textually unmarked or effaced, by turning such a narrator into a mere option within the narratological model, to be applied to a given narrative only if warranted by the existence in the text of certain textual features. Hamburger  1993 for example has argued on philosophical grounds that one can meaningfully speak of a narrator figure only in first person narratives, while in all other cases the narrator is a mere metonymy for a presentational textual function. Banfield (1982) has argued on linguistic grounds that the notion of narrator is meaningful and warranted only in cases of overt, foregrounded narration similar to the oral one, such as the skaz (which is of course third person narration).
As soon as the universality of the fictional narrator is rejected, a uniform treatment of all varieties of narrative fiction is no longer possible, and the one, universal model is replaced with a whole set of options, alternatives or partial models, each of them being deemed the most appropriate or warranted in some case or another. But why stop with (c)? In fact, assumption (b) and even (a) can, and have been, rejected by some scholars at least for some (types of) texts. Sylvie Patron for example (Patron 2010) claims that not all works of narrative fiction can be justifiably regarded as acts of communication, thus denying the universality of (a). Some works of narrative fiction are similar to Benveniste’s histoire, and hence better regarded as non-communication, so the argument goes, since it is not possible to define in them a global, inner textual speech position functioning as point of origin of the discourse as a whole. In such works, one might add, the marginalized or non- existent communicative function is replaced by the dominance of the presentational one. And on this non-communicative (or semantic-oriented) view it is expressions by themselves that can refer, and the entities in the narrated domain can be established without recourse to a particular speech position. Accepting just assumption (a) as universal would imply viewing narrative texts as consisting of an interplay of two kinds of discourses, defined by such hierarchical (hence anti-symmetrical) relations as dominant and subordinate, embedded and embedding, quoting and quoted, referring and referred to. The dominant discourse is associated with the highest -level speech position and is for convenience’s sake referred to as the discourse of the narrator, while the subordinate discourses are associated with the speech positions of characters. Yet one deals here with linguistic and discursive functions or roles only, and stops short of any attempt to anthropomorphise them, to identify and characterize any specific human- or human-like individuals who occupy the respective positions. Stylistic and ideological features, rather than pragmatic or individual- psychological ones, are the ones to be associated with each speech position. (Baxtin).
Accepting assumptions (a) and (b) while rejecting (c) opens up three options as to the identity of the occupant of the narrator position. One is obviously (c) itself, the maximalist view which is precisely the one being rejected. The other one is the minimalist: the occupant of the highest- level speech position in a work of narrative fiction is always the actual author, but in a ludic or make-believe guise, feigning the making of true assertions, and sometimes also pretending to be someone else. And there is also the middle position: if we replace essentialism with instrumentalism and universal claims with qualified existential ones, we can regard the author in a make-believe mode and the fictional narrator as two co-existing options. In some cases the first would be the better warranted by the text, while in others the second would be more appropriate. The choice is thus between a fictional individual who reports seriously of facts in the narrated domain as known to him, and the actual author-performer merely feigning, pretending or playing the role of a reporter of facts, or a maker of true factual assertions, while in actuality he is their inventor (Searle’s illocutionary pretense. See Searle 1975). In terms of rules of procedure or methodological norms, two opposing norms can be envisioned. The first would claim that the default case of the originator of the narration is the fictional narrator, and good reasons should be provided whenever one rejects this option in favour of the author-cum-fabulator one. The opposite norm, advocated by some philosophers, is that the default case is the author as fabulator-pretender, and good reasons should be provided whenever one posits instead a fictional individual as teller-reporter. The most crucial case is that of heterodiegetic (third person), impersonal narration, where the highest textual speaker position is occupied by an anonymous, unindividuated voice (“geistig und abstrakt” in Thomas Mann’s words) or, in other words, where the speaker position is unmarked. It is precisely in such cases, several scholars have argued, that it is totally unwarranted to fill the teller slot with a fictional individual figure of an “effaced” narrator. In such cases, so the argument goes, it makes much more sense to make the actual author in his role as pretender the originator of the discourse. Such narratives are hence “narratorless” in that they do not satisfy (c), while still possessing (b).
If we adopt an instrumentalist view of theories, regarding them as cognitive tools rather than ontological commitments, once could now quickly assess the relative cost/benefit of postulating a fictional narrator vs. an author as pretender in cases of third person impersonal narration. Quite obviously, the advantages of one position are the shortcomings of the other and vice versa. The advantages of the author as pretender are as follows: this position conforms with Occam’s dictum that entities (and, one might add, especially fictional ones) should not be multiplied beyond necessity. It also conforms to David Lewis’ principle of minimal departure for fictional worlds, which states that a fictional world should be assumed to be as similar as possible to the actual one unless explicitly specified otherwise. (Lewis 1978). And this view further enables us to tackle in a straightforward manner the issue of narrative style and composition. It is thus the actual author in his role as pretend-reporter who makes all of the stylistic and compositional choices regarding the narration. And finally, adopting this view provides continuity with a long tradition harking back to Antiquity. Conversely, sticking with the always a fictional narrator position, even in the case of impersonal third person narration, preserves the absolute distinction between the fictional and the actual, as well as providing a uniform treatment along a continuum for all varieties of narrative fiction, instead of splitting the domain into radically heterogeneous sub-domains. And it is also simpler, since it involves a semantic consideration only, and does not require pragmatic considerations about actual people playing specific, culturally defined pretend roles. Arguments for and against each position keep being offered in the research literature, but it may be our deeply held fundamental views on the relation between art and actuality, rather than methodological considerations, which ultimately make us adopt one of the two positions.
(a) The image of the textual speaker as constructed in the context of fiction writing should be examined in its relation to the projected speaker image in lyrical poetry (persona) and in non-fictional narratives. (b) It is assumed here that the diarist and letter writer are narrators, yet Chatman (1978) denies this: is it because he implicitly identifies narrator with global narrator? (c) Can narrators be focalizers, and if so, when and to what extent? This problem has not been touched upon here, yet is the subject of extensive debate in the critical literature. (d) This entry makes no use of the notion of ® implied author, which the present writer finds redundant in a communication -based model. However, the implied author appears in almost every discussion of the narrator. Should this be the case? (e) Narrator unreliability as regards judgments and evaluations has been treated here entirely as a matter of readers’ criteria, unlike factual unreliability, for which there are objective inner-textual indicators. Why has this view emerged only recently, and is the resistance to it associated with the implied author postulate?