1

Who Is the Narrator?

Richard Walsh
University of York

Who is the narrator? Today most literary critics are happy to regard the narrator as an

inherent feature of narrative, although the coherence of any distinct concept of such a

narrating agent remains debatable, to say the least.1 In calling the narrator into

question, I want also to question the broad assumptions that have sustained the

concept in critical practice; I do not think of it as a purely narratological matter, but

one that has large consequences for our understanding of fiction. Indeed, the

narrator‟s promotion from representational accidence to structural essence has

occurred specifically in response to the qualities of fiction, not narrative per se; and

the concept has only been put to the most cursory use outside the fictional context.

This is because the narrator, thus understood, functions primarily to establish a

representational frame within which the narrative discourse may be read as report

rather than invention.2 In other words, it defines the extent to which we can set aside

our knowledge that the narrative in hand is indeed fictional. By conceiving of a

fictional narrative as issuing from a fictional narrator, the reader has cancelled out its

fictionality, negotiated a mode of complicity with representation, and found a

rationale for suspension of disbelief. I want to suggest, though, that certain dubious

critical tendencies are perpetuated by this model of fiction. Firstly, critical

interpretation tends, in point of detail, to be confined within the narrative‟s

representational frame, rather than attending to its rhetorical import—with the

common result that criticism indulges too far in collaboration with the fiction‟s own

rhetoric of representation. Secondly, the representational frame induces a kind of

critical double vision that separates this intrafictional perspective from a larger sense
 2

of the fiction as a literary work (characterized by its style, technique, themes,

symbolism, etc.); and the effect of this dichotomy is that such literary considerations

become the belated response to a naive primary reading experience. I would want to

argue that as the basis for reading fiction, a willing suspension of disbelief will not

do: disbelief is essential to reading a work of fiction as fictional, and only by doing so

can we apprehend the effects it achieves by means of fiction‟s own particular literary

resources, including the involvement to which the phrase “suspension of disbelief”

testifies. One of the consequences of rejecting the concept of the narrator is that the

representational frame, as an impassable barrier between the creative and (putatively)

informative aspects of fiction, is breached. It is with this in mind that I am going to

question the idea that the narrator, as a distinct and inherent agent of fictional

narrative, is a logical, or even plausible, construct.

I‟d like to approach the problem schematically, in the first instance, by invoking

two of Gérard Genette‟s distinctions: between homodiegetic and heterodiegetic

narrators (a matter of person: that is, in place of the common distinction between first-

and third-person narrators, a more exact contrast between involvement and non-

involvement in the story); and between intradiegetic and extradiegetic narrators (a

matter of level: that is, the distinction between a narrator who narrates within a larger,

framing narrative, and one whose narration itself constitutes the primary narrative).

Between them, these distinctions produce four classes of narrators (Genette 1980:

248): my intention is to show that none of them require a distinct narrative agent. The

two intradiegetic classes are relatively straightforward: these narrators are simply

characters, within a narrative, who relate a story in which (respectively) they are and

are not themselves involved. Marlow in Heart of Darkness, sitting aboard the “Nellie”

on the sea reach of the Thames and narrating his journey to the farthest point of

navigation, is intradiegetic and homodiegetic; In Sarrasine Mme de Rochefide‟s

 3

unnamed admirer, who tells her the sculptor‟s story on the evening after the Lanty

ball, is intradiegetic and heterodiegetic. The extradiegetic categories are more

difficult. Genette maintains that extradiegetic narrators, being outside any diegesis,

cannot be characters—“for that would be meaningless” (Genette 1988: 85): yet an

extradiegetic homodiegetic narrator like Huck Finn is, of course, identified with a

character in the story. So the extradiegetic homodiegetic case seems to establish a

clear distinction, even within the fictional frame, between characters and narrators.

But obviously many such narrators—Huck Finn, Tristram Shandy, Humbert Humbert,

Molloy—are at least as strongly characterized in the telling of their tales as they are in

the role of protagonist. How then are they different from their intradiegetic

counterparts? Genette has himself acknowledged that the distinction between

extradiegetic and intradiegetic is relatively unimportant, given that “all that is needed

to convert an extradiegetic narration into an embedded narration is a sentence of

presentation” (1988: 95). He illustrates the point with a playful revision of A la

recherche du temps perdu, a favoured extradiegetic homodiegetic narration. I shan‟t

quote in full, as the crux is simply this: “Marcel cleared his throat and began: „For a

long time I used to go to bed early,‟ etc.” (1988: 95). Very well, but consider the
vastly different effect of this: “The ironic spinster cleared her throat and observed, „It

is a truth universally acknowledged, that a single man in possession of a good

fortune,‟ etc.” This second case is the transformation of an extradiegetic

heterodiegetic narration, and involves something that the first did not—namely the

creation of a character (we might want to call her Jane). I shall return to the case of

extradiegetic heterodiegetic narration in a moment, but first I want to pursue the

implications of this difference for Genette‟s homodiegetic example. My point is that,

in such a case, the only necessary effect of the transformation from extra- to

intradiegetic is a specification of the narrating instance. This cannot amount to a

 4

change in level, as for Genette a narrating instance is implied by every narrative: “the

main point of Narrative Discourse, beginning with its title, reflects the assumption

that there is an enunciating instance—the narrating—with its narrator and its narratee,

fictive or not, represented or not, silent or chatty, but always present in what is indeed

for me, I fear, an act of communication” (1988: 101). Indeed, to concede that a

narrating instance is not implied by every narrative would be to concede to the

linguistic arguments for non-narrated narratives advanced by Ann Banfield (1982)

and others. Discourse, as an act of communication, is action; in fiction, the

represented discourse of a homodiegetic narrator is therefore represented action. And

what is action but the illustration of character? Extradiegetic homodiegetic narrators

are indeed characters, and if there is any meaninglessness lurking in that formulation,

it can be located in the concept of the extradiegetic itself. Narrators are always outside

the frame of the stories they tell: “extradiegetic” appears to have the additional force

of placing the narrator outside representation. But if the narrator is fictional, where

would that be? In such cases the telling of the story is itself a represented event, as

clearly represented as any act of speech, thought or writing in the story: we could

legitimately put quotation marks around the whole.

The purpose of my attention to the extradiegetic homodiegetic narrator has been

to establish this preliminary point: there is nothing about the internal logic of fictional

representation that demands a qualitative distinction between narrators and characters.

Such narrators, being represented, are characters, exactly as intradiegetic narrators

are. But of course it is the fourth class of narration, the extradiegetic heterodiegetic,

that constitutes the real issue. In this class fall those narratives that we might want to

call “impersonally narrated,” such as The Ambassadors, The Trial, or Mrs Dalloway;

as well as what is sometimes called “authorial narration”—Tom Jones, Vanity Fair or

Middlemarch. The one irreducible fact underlying the impulse to attribute such
 5

narratives to a narrator is that these narratives are fictional: despite the token gestures

of narratologists whose bias towards fiction sits uneasily with their claims for the

more general bearing of narratology, there is no more reason to posit a narrator for

historical or biographical narrative than to attribute every non-fictional discourse to a

textual agent.3 Genette‟s own early statement of the underlying assumption is

representative: “the narrator of Père Goriot „is‟ not Balzac,” he says, “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” (1980: 214). The function of the narrator is to allow

the narrative to be read as something known rather than something imagined;

something reported as fact rather than something told as fiction.4 But this view of the

matter suffers the embarrassment that some of the things such a narrator is required to

“know” are clear indices of the narrative‟s fictional status. The most obvious of these

occurs with internal and free focalization—that is, the narrative‟s access to the mind

of another: “her heart, like a larded partridge, sweltered before the fire of a burning

desire to shake off the shroud of Vauquer and rise again as Goriot. She would marry

again, sell her boarding-house, give her hand to this fine flower of citizenship . . .”
(Balzac 1991[1835]: 16-17). The only way to account for such knowledge of

characters‟ minds in terms of the narrator model is to take quite literally the figurative

concept of “omniscient” narration: in order to know rather than imagine, the

(evidently superhuman) agent of narration must indeed have such powers.

“Omniscience,” I would suggest, is not a faculty possessed by a certain class of

narrators but, precisely, a quality of imagination. Even when authors self-consciously

dwell upon their own omniscience with regard to their creations, the power itself is

fanciful. The reader is not obliged to hypothesize a narrator who really is omniscient
 6

within the terms of a given fiction, because the authorial imaginative act doesn‟t

merely initiate a fiction, but pervades it.

There are other aspects of focalization, even where omniscience is apparently

renounced, that prove unassimilable to the concept of the narrator as the one who

“knows.” Consider external focalization, which prohibits any access to the characters‟

thoughts: here “the focus is situated at a point in the diegetic universe chosen by the

narrator, outside every character” (Genette 1988: 75). But this focus cannot be

understood as a heterodiegetic narrator‟s own perspective, because that would make

the narrator homodiegetic—even if anonymous and perhaps non-corporeal. Seymour

Chatman, discussing the issue of focalization, has rightly insisted upon the radical

difference between narratorial “slant” and character “filter”: the narrator “is a

reporter, not an „observer‟ of the story world in the sense of literally witnessing it. It

makes no sense to say that a story is told „through‟ the narrator‟s perception since

he/she/it is precisely narrating, which is not an act of perception but of presentation

or representation” (1990: 142).5 He draws the necessary conclusions for the category

of narration we are considering here: “The heterodiegetic narrator never saw the

events because he/she/it never occupied the story world. . . . Even for so-called
„camera-eye‟ narration it is always and only as if the narrator were seeing the events

transpire before his very eyes at the moment of narration” (ibid.: 144-45). How are we

to understand this “as if”? We cannot resort again to omniscience, unless we

compromise it by assuming a sustained narratorial reticence about the characters‟

thoughts, and other such matters. This reticence could only be disingenuous: in

Genette‟s terms it would have to be described paradoxically, as a defining paralipsis.

So the only way to construe external focalization would be as the work of the

narrator‟s imagination: again the narrator‟s rationale, as the one who “knows,” is

undermined.
 7

Of course, the point isn‟t really that the narrator “knows” at all; it‟s that the author

can‟t know. The purpose of the narrator is to release the author from any

accountability for the “facts” of fictional narrative. Genette has codified this principle

in the double formula “A = N → factual narrative and A ≠ N → fictional narrative,”

where the equal sign symbolizes “the serious commitment of the author with regard to

his narrative assertions” (1990a: 766, 770). This narrow definition of identity is

adopted in preference to “onomastic or biographical identity” because the narrator of

a manifestly fictional story may indeed be identified with its author in these terms, as

is the case in Borges‟ “El Aleph,” or in Tom Jones. It is perfectly clear that Fielding

“does not in the least vouch for the historical veracity of the assertions of his

narrative”; but Genette argues further that he does not “identify with the narrator who

is supposed to have produced it, any more than I, good citizen, family man, and free-

thinker, identify with the voice that, through my mouth, produces an ironic or playful

statement such as, „I am the Pope!‟ ” (1990a: 768). The assumption is that fiction and

irony are “nonserious” speech acts, and so require a distinction between their actual

and pretended speakers. Genette is following John Searle, whose account of fictional

utterances as pretended acts of assertion is the canonical speech act treatment; and if
the implication of a narrator is not quite self-evident in Searle‟s pretence formula, it

may arguably be present in another description, “imitating the making of an

assertion,” which he offers as equivalent (Searle 1975b: 324). Searle approaches

fiction with priorities very different from those of the literary theorist, however: he

remains in broad sympathy with J. L. Austin‟s view that such matters “fall under the

doctrine of the etiolations of language” (Austin 1975 [1962]: 22). Searle‟s somewhat

hasty and dismissive response to fiction is motivated by the evident need to

distinguish it from mere falsehood: if a fictional utterance is read simply as an

authorial assertion, then it must be taken as infelicitous—an error or a lie—which

 8

hardly satisfies our sense of how fiction works. But his own pretence theory is equally

unsatisfactory from a literary point of view because, far from using speech act theory

to explain fiction, it disqualifies fiction in order to protect speech act theory. To

classify fiction as a “nonserious” speech act is simply to disallow it: the problem of

fictionality is not accounted for, but merely displaced. So, instead of a real act of

asserting something fictional, Searle gives us an imitated act of assertion: that is, a

fictional act of assertion, since fictionality (unlike falsehood) is an ontological

property, not just a property of propositions. A fictional act of assertion would not

seem to be any less problematic than an act of asserting something fictional: it has

merely expelled fictionality from the domain of speech act theory. And if this account

of fiction as authorially disavowed assertion amounts to the creation of a narrator, as

Genette assumes, then its question-begging is even more starkly exposed. Either the

narrator is fictional, or the narrator asserts something fictional: in either case such an

account can have no bearing at all on fictionality, which remains to be explained. The

pretence formula can only accommodate fictionality by invoking a narrator

simultaneously inside and outside the fiction.

Genette‟s own response to Searle goes some way towards addressing the problem
by arguing that the description of fiction as pretended assertion does not exclude the

use of fictional utterances to perform some other, serious illocutionary act. The aim of

his intervention “is by no means to replace Searle's „Fictional texts are pretended

assertions,‟ but to complete it approximately as follows: „. . . which hide, under

indirect speech acts, fictional speech acts that are themselves illocutionary acts sui

speciei, serious by definition” (Genette 1990b: 66). Indirect speech acts (among

which Genette includes figurative utterances, as simply indirect speech acts with an

unacceptable literal meaning) are those in which one illocutionary act serves as the

vehicle for another. “You‟re standing on my foot” is also a request that you get off;
 9

“Hegel is a dead horse” is also an assertion that it is no longer worth disputing with

him.6 To be understood, indirect speech acts need to be considered in relation to their

contexts, on the basis of a set of accepted rules for cooperative communication such

as H. P. Grice has outlined. Grice‟s “Cooperative Principle” states the criteria for the

successful performance of a serious speech act in a few general maxims: one of these

is the first maxim of Quality, “do not say what you believe to be false”; another is the

maxim of Relation, “be relevant” (Grice 1975: 46). The literal illocution of an indirect

speech act manifestly disregards the Cooperative Principle, typically by being

irrelevant, but also by being false: if an indirect speech act were exhausted in its

literal illocution, it would have to be regarded as infelicitous. But because this

disregard for the Cooperative Principle is so blatant, we are led to suppose that the

maxims are not just being violated, but exploited. This process, by which we are

prompted to look for a nonliteral illocution that will successfully relate the speech act

to its context, is what Grice terms “conversational implicature.” So, the maxim of

quality may be furtively violated (as it is in lying), in which case the Cooperative

Principle breaks down; but it may also be flouted (blatantly violated, as in irony, or in

a work of fiction), in which case we are able to assume it is being exploited in the
interests of conversational implicature, and so conclude that the Cooperative Principle

is being maintained indirectly.

For Genette, fictional texts are indirect speech acts that imply, by means of

pretended assertions, acts in the category of “declarative illocutions with an institutive

function” (1990b: 64): that is, acts declaring the existence of a fictional world. The

illocutionary act of establishing this fictional world, with the agreement of an

audience—Genette cites “Coleridge‟s durable phrase” (ibid.: 63)—is the serious

element of the fictional utterance. I have two objections: firstly, if fictional utterances

are indirect speech acts, they must do something more than institute a fictional world.
 10

Genette sometimes appears to regard this declarative illocution as itself a fictional

speech act, in which case it remains within the frame of fictionality, and adds nothing

to Searle‟s account of the authorial speech act; but in any case it confines the creative

act to the existential matter of fiction, without any sense of the extent to which this act

concerns meaning, in language, rather than existence. This sits uncomfortably with

Genette‟s claim that the novelist thereby creates “a work of fiction,” as he seems to

acknowledge when he says (with more than a hint of circularity) that the successful

achievement of a fiction‟s illocutionary intention consists “at least” in having its

fictional intention recognized (1990b: 62, 67-68). His own account seems to allow

this much at most, in which case it cannot be a sufficient general formula; nor, I think,

is it sufficient even on occasion.

Secondly, if fictional utterances are to have an indirect illocutionary force then the

literal speech acts by which they achieve this cannot be pretended, but must be

seriously performed. Genette can only interpret pretended speech acts as indirect

speech acts by blurring two distinct concepts: he notes that Searle himself explicitly

refuses to consider fiction as figurative utterance “in the name of a distinction, to my

mind rather fragile, between „nonserious‟ and „nonliteral‟ ” (1990b: 66). But Searle is
right in this respect: as he defines it, seriousness attaches to illocutionary intention,

whereas literalness attaches to sentence meaning. This is why he makes it clear that

“to pretend” is itself an intentional verb: if you didn‟t intend to pretend, you didn‟t

pretend (Searle 1975b: 325). Searle distinguishes between serious and nonserious

(pretended) speech acts according to whether or not the illocutionary act was actually

performed. If there is no actual performance, but only a “pseudoperformance” (1975b:

325), then there is no possibility of a serious indirect speech act because the felicity

conditions (or maxims, after Grice) normally attaching to the speech act are

suspended, in which case they cannot even be violated, much less flouted in the
 11

interests of conversational implicature. Genette seems to interpret seriousness as

sincerity, which allows him to say that any nonliteral illocution is obviously also, in

its literal sense, nonserious (read insincere). But sincerity is just a condition upon the

success of serious speech acts—it doesn‟t even arise unless the speech act is serious.

If pretence is understood in the sense it has in Searle‟s account, then indirect speech

acts and pretended speech acts are incompatible. And even if Genette has silently

reformulated Searle‟s pretence account as “pretending to be the narrator,” this

translates the model for third-person fictional utterance into that for first-person

fiction; in which case the discourse itself is intrafictional, and excludes the possibility

of any indirect speech act declaring its fictionality.7 If the indirect speech act model

applies to fictional utterance, then the pretence model does not; but if pretended

speech acts are not involved in fiction, neither is the narrator.

My argument is that pretence can have no role in an account of fictional utterance,

either alone or in combination with indirect speech acts. I don‟t wish to repudiate the

idea of fictions as indirect speech acts: the broad outlines of Genette‟s account strike

me as highly suggestive. But the model requires that fictional utterances are serious

authorial speech acts, and this excludes any possibility of a default narrator. If, when
Genette declares “I am the Pope,” we assume that he is neither deceitful nor deluded

but adhering to the Cooperative Principle, then we note the literal absurdity of the

statement and understand, perhaps, “I acknowledge a fondness for issuing the

occasional bull”—or some other relevant implicit meaning. There is no phantom

voice here, because this is a serious speech act, the felicity of which is provided for

indirectly, by conversational implicature. So it is with fiction: an author can seriously

narrate a fictional narrative, because its relevance is not a matter of information; its

falsehood, or indeed any adventitious veracity, is beside the point. Fiction may be

related to the indirect speech act model in the following way: it is a series of
 12

illocutionary acts of assertion (typically) which, whether true or false, are literally

irrelevant because they cannot be taken as informative; but which imply, by virtue of

their context—being presented as a novel, a romance, a tall tale, a shaggy-dog story—

the illocutionary act of displaying a narrative. This implied act is normally

transsentential, comprising as it does the whole narrative, and might better be

described as a “discourse act”; but in any case it fulfils the criterion of relevance as an

invitation to contemplate, to interpret, to evaluate; as something worthy of display,

something “tellable.” That is, its relevance is not informative but exhibitive: the

question of its truthfulness is therefore not applicable as a felicity condition.8

The answer I am proposing to my original question, “who is the narrator?” is this:

the narrator is always either a character who narrates, or the author. There is no

intermediate position. The author of a fiction can adopt one of two strategies: to

narrate a representation, or to represent a narration. I say this in full awareness of

Genette‟s criticisms of the concept of representation (1988: 42): indeed his point, that

the term equivocates between “information” and “imitation,” is borne out by my own

antithesis. “Representation” is a matter of (fictional) information in “to narrate a

representation,” but a matter of (discursive) imitation in “to represent a narration.”

But I persist in the usage on the grounds that this disjunction of means does not at all

undermine the unity of the rhetorical end that I take “representation” to signify.

Where does unreliable narration fit into this scheme? The need for a concept of

unreliable narration arises when we wish to explain inconsistencies in the narrative

without blaming the author. This is not to say that we do not sometimes find the

author culpable: when we discover Sancho, in chapter 25 of Don Quixote, riding the

ass that was stolen from him in chapter 23, we can dismiss it as an oversight on

Cervantes‟ part. We need more substantial reasons than inconsistency alone if we are

to identify unreliable narration. To be interpreted as unreliable, a narrative must

 13

provide some logic by which its inconsistencies can be explained—some means of

accounting for the narrator‟s self-contradictions or manifest distortions. That is,

unreliability cannot simply be attributed to an impersonal narrator: it must be

motivated in terms of the psychology of a narrating character.9

But perhaps Chatman‟s concept of narratorial “slant” suggests a more subtle,

evaluative form of unreliability. Perhaps it is possible for the language of a novel in

its own right, without implying any disjunction between the narration and the “facts”

of the narrative, to cohere into an idiom, register, attitude or ideology that requires a

distinction between author and narrator. That is after all the situation with first-person

narration, although there are simpler and more obvious representational grounds for

identifying a narrator when that narrator is homodiegetic. Is the narrative language

alone sufficient to betray a narrator? Dorrit Cohn argues that just such a situation may

arise when “reportive” narrative is interrupted by commentary. Her example is from

Death in Venice, and prompts us, she thinks, to “personalize the source of the weighty

intervention . . . as a rather narrow and opinionated moralist”—in other words, not

Thomas Mann (Cohn 1990: 797). I don‟t actually find the passage she quotes very

provoking in this respect, but in any case I have strong reservations about the
possibility of such narratorial characterization in principle. Clearly the difference

between authorial and narratorial personality must be established in textual terms (it is

Mann as author, not Mann as public figure, who concerns us here), yet the absence of

textual indicators such as inconsistency necessitates an appeal to the author‟s

personality as already known, prior to the text. Authorial personality can be regarded

as an intertextual phenomenon, to be abstracted from a writer‟s whole corpus; but

there still remains the unwarranted assumption that this personality is uniform, for

otherwise there is no reason why the narration in question should not be taken to

exhibit another aspect of authorial personality. Personality, after all, is not monolithic;
 14

not timeless, not unitary, not even necessarily coherent. Indeed novelists, who are

perhaps rather less straightforward than academics, are quite likely to attitudinize in

diverse ways in their writing: such mannerism remains an aspect of self-presentation,

and should not be granted an independent identity. Cohn recognizes that works like

Death in Venice may indeed be taken as authorial narration, but thinks her

interpretation preferable “for readers intent on salvaging the aesthetic and ideological

integrity of the work in question” (1990: 799). The integrity she is actually concerned

with is the author‟s, not the work‟s: I would suggest that the issue of integrity only

arises because the work in question has been illegitimately excluded from a prior

interpretation of authorial personality. In general, I think the possibility of

distinguishing between authorial and narratorial personality depends upon that

distinction being available to interpretation as a meaningful aspect of the text‟s own

representational rhetoric.10 At that point, I suspect, the passages of commentary to

which Cohn appeals would have effectively cohered into a homodiegetic frame

around the narrative.

I want, nonetheless, to consider one of the conclusions Cohn draws from the

possibility that narrative language alone can characterize a narrator. For my purposes
here that claim in itself causes no difficulty: it‟s a characterization, involving creative

work, and not something inherent in narrative as such. But since this characterization

is occasional (it only occurs in passages of commentary within a continuous narrative)

it must, in my terms, imply an intermittent character. For Cohn, the only logical way

to account for this is to conclude that the narrator is always present, sometimes overt,

and sometimes covert. “By extension and analogy,” fictions like The Castle or A

Portrait of the Artist can then be taken to have covert narrators throughout (1990:

797-98). This covert narrator, wholly uncharacterized, is exactly the kind of pure

narrative agent I am trying to eradicate.

 15

Leaving aside the validity of “extension and analogy,” can there be such a thing as

covert narration, even between passages of overt narration? I want to suggest that the

underlying assumption here—that narrating characters must have continuity of

being—is an instance of criticism internalizing a literal model of the logic of

representation, and then using it against the text itself. In representational terms any

narrating character is the source of the narrative language, certainly: but then

representation itself is only a product of the fictional deployment of the same

language. We need to understand these relations hierarchically, in that the language of

fiction is its means of representation, and representation is its means of ascribing that

language to a narrator. So to treat a represented instance of narration as ontologically

prior to the language doing the representing is to press the logic of representation

beyond representation itself, and make the subordinate term superordinate—that is, to

assert a paradox in the name of logic. Yet this is exactly what the idea of covert

narration demands: even when the representation of a narrator is not sustained, the

whole discourse is interpreted as a unified narrating instance because the narrator, a

local representational issue of the language, is translated into its global, literal source.

We should keep in mind the fact that representational “logic” is actually a fictional
rhetoric: it should not be made to exceed its brief. If a (hypothetical) novel‟s language

invokes a narrator in the interest of some local effect, then to interpret this effect as

indicative of a ubiquitous but otherwise covert narrator is to miss this rhetorical

subtlety completely in our rage to impose a uniform representational logic upon the

novel.

The idea of an intermittent narrating character, on the other hand, would fit such a

novel very well; and I think it entirely consonant with the rhetoric of fictional

representation. Consider the situation of homodiegetic narrators: they are far from

being ubiquitous presences, even if we discard such categorical aberrations as

 16

Madame Bovary (Flaubert as Charles Bovary‟s classmate in the opening chapter) or

Vanity Fair (Thackeray making his characters‟ acquaintance at Pumpernickel in

chapter 62). As Genette has noted, Marcel has a striking propensity to disappear as

narrator of A la recherche du temps perdu whenever Proust‟s purposes demand the

omniscience his narrator denies him. (Genette 1980: 250-52). But if Marcel is too

literary a narrator to make the argument, try Huck Finn. Huck is as strongly realized,

and ingenuous, a narrator as you could wish for; yet Twain put an explanatory note in

front of his novel drawing attention to the different dialects it contains, because he

didn‟t want readers to “suppose that all these characters were trying to talk alike and

not succeeding” (Twain 1966[1884]: 48). He didn‟t do this to emphasize Huck‟s

talents as a mimic: nor was it an oversight on his part—he‟d paid particular attention

to it, he was proud of it, and he wanted to make sure we notice his fine ear for dialect.

The conclusion must be that, in those parts of the novel where Twain is accurately

representing the various dialects of the Mississippi valley, the narrating Huck Finn is

not merely covert, but entirely absent. I should emphasize that Twain‟s note is not

essential to this point: it only makes starkly explicit the truth that a conflict of

representational objectives is likely to arise in any mediated narrative. There is an

inherent tension between the representational needs of the narrative transmission and

those of the narrative events; and in the case of direct speech, it is almost always the

character‟s language itself that is represented, not the narrator‟s representation of that

language. At such points, the situation is a mirror image of our hypothetical novel:

local elimination of the narrator rather than local creation of a narrator.11

There is another figure who threatens to intervene in this discussion of the

narrator: having made an appeal to the author, I need to take account of the implied

author. Wayne Booth originally advanced the concept as a way of talking about

authorial personality and intention without co-opting, or being encumbered by, the
 17

author‟s actual biography—for reasons that are well founded in the history of

criticism. But as his choice of term indicates, he objectified this concept as a distinct

agent situated “between” the narrator and the author. If I am saying that in fact the

narrator and the author are sometimes one and the same, I appear to have simply

disregarded this intermediate figure. A short answer may be to observe that, as the

conventional wisdom has it, the implied author (being implied) cannot actually be the

narrator. In that case, perhaps the issue does not arise here: the “implied author” is

just the author implied behind a narrating character; and when it is the author who

narrates, the implied author obviously need not be invoked. But the argument might

be pressed the other way round: if the locus of textual intent is definitionally the

implied author, then the narrator cannot be simply the author—and so must be a

distinct agent. A closer inspection of the implied author is needed to dismiss this

objection.

It is possible, on reading Booth‟s original discussion in The Rhetoric of Fiction, to

extract two pertinent motives for distinguishing between the real author and the

implied author. The first is a matter of authorial personality: against Ford Madox

Ford, Booth insists that Fielding, Defoe and Thackeray cannot be accused of
insincerity on the basis of external evidence: “A great work establishes the „sincerity‟

of its implied author, regardless of how grossly the man who created that author may

belie in his other forms of conduct the values embodied in his work. For all we know,

the only sincere moments of his life may have been lived as he wrote his novel”

(Booth 1983 [1961]: 75). As the second sentence suggests, Booth‟s defence here

actually hovers between two strategies: to declare a separation between the real author

and the implied author, or simply to refuse the uniformity that Ford‟s rigid “sincerity,”

regardless of the diversity of its occasions, seems to impose upon personality. The

second option, it seems to me, is quite sufficient. Booth‟s second motive has to do
 18

with authorial intention: in order to explain our “apprehension of a completed artistic

whole” as a textual phenomenon, we need “a term that is as broad as the work itself

but still capable of calling attention to that work as the product of a choosing,

evaluating person rather than as a self-existing thing” (1983 [1961]: 73, 74). Again,

the distinction is blurred by equivocation: as Chatman has observed, “this definition

straddles the fence of „intentionality,‟ half accepting and half rejecting its relevance to

textual structure. On the one hand, Booth disallows the intention of the real author,

but on the other, he wishes to avoid calling texts „self-existing things‟ ” (Chatman

1990: 81).

Chatman‟s own defence of the implied author proposes to redefine the concept in

order to “resist the anthropomorphic trap” due to which it continually gravitates back

towards the real author in Booth‟s usage (Chatman 1990: 88). Accordingly, he takes

“the anti-intentionalist view that a published text is in fact a self-existing thing. . . .

The text is itself the implied author” (ibid.: 81). His argument is founded upon a

distinction between oral and written narrative: the oral situation is straightforward,

thanks to the actual presence of the author; but in the case of a published fiction “the

real author retires from the text,” and the implied author is invoked “on each reading”
as the textual principle of invention and intent (1990: 75, 74). Chatman emphasizes

his concept‟s freedom from anthropomorphic assumptions by offering alternative

terms: “ „text implication‟ or „text instance‟ or „text design‟ or even simply „text

intent‟ ” (1990: 86). Yet even these terms indicate the tension in his argument. If the

text is to be a self-existing thing, divorced from authorial intention, then there are no

grounds for appealing to a concept of intent at all—it is no longer required. Chatman

insists that “the act of a producer, a real author, obviously differs from the product of

that act, the text”; but then he can only explain textuality by reinventing that act of

production as itself immanent in the text: “If all meanings—implicit as well as

 19

explicit—are the products of the text‟s activity, and if this activity always presupposes

agency, then we have to posit some such text principle or agent as the implied author”

(1990: 83, 90). He conceives of the written text as manifesting, on each reading, its

own intentional agency—that is, a virtual oral authorship equivalent to the actual

presence of the author. This seems to be just a more subtle version of the

anthropomorphic trap from which he claims to have escaped. If we want to talk about

intent in fiction, we should accept that in doing so we are necessarily invoking the

author. Of course our idea of the author of a written narrative is no more than an

interpretation; but this is equally true with oral narrative.

It will be clear by now that I subscribe to Genette‟s principle that “agents should

not be multiplied unnecessarily” (1988: 148). Genette has himself rejected the concept

of the implied author, reasoning that it has been “constituted by two distinctions that

remain blind to each other: (1) IA is not the narrator, (2) IA is not the real author, and

it is never seen that the first is a matter of the real author and the second is a matter of

the narrator, with no room anywhere for a third agent that would be neither the

narrator nor the real author” (1988: 145). It‟s a nice knockdown argument, and I

would only want to qualify it by extending its scope: there is no room anywhere for a
third agent that would be neither a character nor the real author.

My argument against the narrator, then, comes down to this: fictions are narrated

by their authors, or by characters. Extradiegetic homodiegetic narrators, being

represented, are characters, just as all intradiegetic narrators are. Extradiegetic

heterodiegetic narrators (that is, “impersonal” and “authorial” narrators), who cannot

be represented without thereby being rendered homodiegetic or intradiegetic, are in no

way distinguishable from authors. This assertion is unaffected by the fictionality of

the narrative, since that is best accounted for by the function of conversational

implicature in maintaining the felicity of speech acts; nor is it affected by issues of

 20

unreliability, because unreliability always requires characterization; nor by covert

narration, because that concept is an abuse of the logic of representation; nor by the

implied author, because the senses in which that term conflicts with my argument are

themselves bogus.

To repudiate the narrator as a distinct narrative agent intrinsic to the structure of

fiction is to repudiate the idea of a closed border between the products of

representation and the real-world discourse of the author. The narrator, postulated

simultaneously inside and outside representation, dissociates the author from the act

of representation: the concept accordingly divides critical attention between the events

and characters of the fictional world in their own right, and the literary ends they

serve as representations. The former mode of criticism requires that critics suspend

their awareness of the narrative‟s fictionality, this awareness being reserved for the

latter activity. But when the narrator disappears, so does this division in critical

attention. By insisting that fictional representation is an authorial activity, I keep the

fictionality of the narrative always in view: my critical attention is always to the

literary act, the representational activity that is fiction. Instead of attending to

representational content and artistic form by turns, I can integrate them at every point
as aspects of a fiction‟s argument: that is, the end to which a particular fiction directs

its rhetorical resources.

Having referred to the concept of argument as the instantiation of the “rhetoric of

fiction,” I want briefly to clarify my use of that phrase, so dear to Booth. In his own

usage, Booth distinguished between a narrow sense, the (overt) rhetoric in fiction, and

a broad sense, fiction as rhetoric, “an aspect of the whole work viewed as a total act

of communication” (Booth 1983 [1961]: 415). Even in this broader sense, though,

rhetoric is in the service of representation: it is the means by which the author tries “to

impose his fictional world upon the reader” (ibid.: xiii). Chatman goes further,
 21

distinguishing between this interior, “aesthetic” orientation and an “ideological” one:

“Rhetoric working to ideological ends suades us of something outside the text,

something about the world at large” (Chatman 1990: 197). This last sense is the one

that interests me; but unlike Chatman I do not see it as distinct from his “aesthetic”

sense, which retains representation as an end. Such aesthetic-rhetorical appeals serve

to establish the rightness of a fictional representation, and this rightness is itself the

“ideological” end to which that representation is being employed. By “the rhetoric of

fiction,” then, I mean the entire resources of fiction as a rhetoric, in itself, for

emotional and rational effect in real-world discourse; and by “argument” I mean

simply the end to which these resources are used by a particular fiction. I would

contend that the advantages of reading fiction this way far outweigh any regrets that

might attend the demise of the narrator.

 22

Notes

1
. The ubiquity of the narrator is a fundamental assumption for Gérard Genette (1980; 1988), Frank

Stanzel (1984), Gerald Prince (1982), Shlomith Rimmon-Kenan (1983) and, despite having

entertained ideas of non-narration in Story and Discourse (1978), Seymour Chatman (1990).

Notable dissenters, on linguistic grounds, have been Käte Hamburger (1973 [1957]), Ann

Banfield (1982) and S.-Y. Kuroda (1976). My own objections to the narrator are based upon

representational rather than linguistic criteria: hence, I shall be arguing that certain “narrators” are

outside representation, not that certain narratives function outside communication.

2
. Some narrators, of course, flaunt their inventiveness: an instance would be Flann O’Brien’s At

Swim-Two-Birds. In such a case the representational frame endowed with an aura of fictional

reality is coextensive with the personality and environment of the narrator himself. It is worth

noting that if fictionality does indeed imply a narrator, such novels would require a second-order

narrator to sanitize the inventiveness of the first. Critics have generally refrained from such

follies.
3
. Dorrit Cohn (1990: 791-800) proposes to make quite explicit the way the author/narrator

distinction operates as a basic criterion for segregating fictional from historical narrative.
4
. Stanzel is equally emphatic on this point: “while the authorial narrator and the first-person

narrator can be differentiated according to their position in regard to the represented world of the

characters, they cannot be distinguished according to their relationship to the apparatus of

narrative transmission. . . . They originate in that primal motivation of all narration, to make the

fictional world appear as reality” (Stanzel 1984: 17).

5
. This observation is particularly salutary in relation to Stanzel’s rather equivocal account of the

status of reflector-characters: “first-person narrators who are actualized only as an experiencing

self, and who therefore restrict themselves to the reflection of experiences not overtly

communicated, are reflector-characters”; “Since [a reflector-character] does not narrate, he cannot

 23

function as a transmitter in the above sense [that of a teller-character]”; “The epistemological

difference between a story which is communicated by a teller-character and one which is

presented by a reflector-character lies mainly in the fact that the teller-character is always aware

that he is narrating, while the reflector-character has no such awareness at all” (Stanzel 1984: 145,

146, 147).
6
. Searle would not count the second example as an indirect speech act: his reason is that the literal

assertion of a figurative utterance is defective (because evidently false), whereas the literal

illocution of an indirect speech act is not. I would maintain that indirect speech acts are always

literally defective, in terms of relevance. Searle acknowledges that they are indeed often defective

in this respect, but denies that they are necessarily so (Searle 1975a: 70-71): it seems to me that,

where utterances may be taken as literally felicitous, to precisely that extent they are ineffective as

indirect speech acts.

7
. I would want to argue further, against Genette and Searle, that first-person narration does not

conform to a pretence account. For Genette, first-person narrative must “finally come down to the

dramatic mode (a character speaks) and consist of serious illocutions more or less tacitly posited

as intrafictional. The pretence here consists, as Plato and Searle agree, in a simulation, or

substitution, of identity (Homer pretends to be Chryses, Doyle pretends to be Watson, as

Sophocles pretends to be Oedipus or Creon)” (1990b: 68-69). I can accept the first sentence, and

note that in saying so Genette appears to have retracted his claim that the first-person narrator is

not a character; but against the second sentence I would argue that authors do not pretend to be

narrating characters, they represent narrating characters. The possibility of unreliable narration

demands this, because when such unreliability occurs the narratorial slant itself (rather than the

events of the narrative) is the object of the author’s representational rhetoric: the distance between

author and narrator is essential to interpretation. In first-person narration, authors do not imitate

the narrating character, nor “the making of an assertion,” but a discursive idiom.
 24

8
. See Mary Louise Pratt (1977). Pratt’s concept of “narrative display text” seems very close to what

is needed here: it is notable that her only reason for ultimately subordinating this concept to the

imitation speech act model is the assumption that all fictions have narrators (1977: 173, 207-8).

My brief reformulation is only a sketch of the possible result of abandoning this assumption, and

obviously needs clarification at several points: I am aware, for example, that the implied act I

posit does not comfortably fit within the category of indirect speech acts as defined by Searle, or

even Genette.
9
. For a far more systematic analysis of unreliable narration, see Tamar Yacobi (1981; 1987), who

places it in the context of alternative means of resolving interpratative incongruities—categorized

as the genetic, generic, existential and functional principles. I am in broad sympathy with

Yacobi’s account, which I do not consider to be seriously undermined by my dissent from its

declared premise: “Insofar as fictionality characterizes the discourse as well as the world of

literature, literary communication is always mediated” (1987: 335). On the criteria for

unreliability, see also Marie-Laure Ryan (1984: 127-28).

10
. According to Yacobi, “To become unreliable, [the narrator] must be exposed as such by some

definite norm of congruity and to some definite effect. . . . In the absence of concrete grounds—or

what appears to be so on the surface—even if the distinction between author and narrator still

holds in theory, then for all practical reading purposes it gets blurred, almost to the point of

disappearance” (1987: 346-7). This hedged dichotomy between the practical and the theoretical

(which exercises Yacobi again on page 357) is obviated once it is admitted that the mediation of a

narrator is not inherent in fiction.

11
. Obviously this provides no basis for a qualitative distinction between extradiegetic and

intradiegetic narrators: Marlow and Mme de Rochefide’s admirer are subject to the same

constraints. Accordingly, it doesn’t provide for any such distinction between narrating and other

characters either. There is a recursiveness about the act of narration, compared to other
 25

represented acts, that tends to highlight the representational contingency of character; but this

contingency applies to any character, as recent innovative fiction has shown (I take the

disintegration of Slothrop in Gravity’s Rainbow to be a quite explicit example). Representation

involves no commitment to the continuity of characters except insofar as this is itself a privileged

representational objective—which of course, in any broadly realist fiction, is the case most of the

time.
 26

References

Austin, J. L.

1975 [1962] How to Do Things with Words, edited by J. O. Urmson and Marina

Sbisà, 2nd edition (Oxford: Oxford University Press).

Balzac, Honoré de

1991 [1835] Old Goriot, translated by Ellen Marriage (London: Everyman).

Banfield, Ann

1982 Unspeakable Sentences: Narration and Representation in the Language of

Fiction (Boston: Routledge & Kegan Paul).

Booth, Wayne C.

1983 [1961] The Rhetoric of Fiction, 2nd edition (Chicago: University of

Chicago Press).

Chatman, Seymour

1978 Story and Discourse: Narrative Structure in Fiction and Film (Ithaca:

Cornell University Press).

1990 Coming to Terms: The Rhetoric of Narrative in Fiction and Film (Ithaca:
Cornell University Press).

Cohn, Dorrit

1990 “Signposts of Fictionality: A Narratological Perspective,” Poetics Today

11(4): 775-804.

Genette, Gérard

1980 [1972] Narrative Discourse, translated by Jane E. Lewin (Oxford: Basil

Blackwell).

1988 [1983] Narrative Discourse Revisited, translated by Jane E. Lewin (Ithaca:

Cornell University Press).

 27

1990a “Fictional Narrative, Factual Narrative,” Poetics Today 11(4): 755-74.

1990b “The Pragmatic Status of Narrative Fiction,” Style 24(1): 59-72.

Grice, H. P.

1975 “Logic and Conversation,” in Syntax and Semantics, Vol. 3: Speech Acts,

edited by Peter Cole and Jerry L. Morgan, 41-58 (New York: Academic

Press).

Hamburger, Käte

1973 [1957] The Logic of Literature, translated by Marilynn Rose, 2nd edition

(Bloomington: Indiana University Press).

Kuroda, S.-Y.

1976 “Reflections on the Foundations of Narrative Theory—From a Linguistic

Point of View,” in Pragmatics of Language and Literature, edited by Teun

A. van Dijk, 107-40 (Amsterdam: North-Holland).

Pratt, Mary Louise

1977 Toward a Speech Act Theory of Literary Discourse (Bloomington: Indiana

University Press).

Prince, Gerald
1982 Narratology: The Form and Functioning of Narrative (The Hague:

Mouton).

Rimmon-Kenan, Shlomith

1983 Narrative Fiction: Contemporary Poetics (London: Methuen).

Ryan, Marie-Laure

1984 “Fiction as a Logical, Ontological, and Illocutionary Issue,” Style 18(2):

121-39.
 28

Searle, John R.

1975a “Indirect Speech Acts,” in Syntax and Semantics, Vol. 3: Speech Acts,

edited by Peter Cole and Jerry L. Morgan, 59-82 (New York: Academic

Press).

1975b “The Logical Status of Fictional Discourse,” New Literary History 6(2):

319-32.

Stanzel, Frank K.

1984 [1979] A Theory of Narrative, translated by Charlotte Goedsche

(Cambridge: Cambridge University Press).

Twain, Mark

1966 [1884] The Adventures of Huckleberry Finn (Harmondsworth: Penguin).

Yacobi, Tamar

1981 “Fictional Reliability as a Communicative Problem,” Poetics Today 2(2):

113-26.

1987 “Narrative Structure and Fictional Mediation,” Poetics Today 8(2): 335-

72.