Jahn, Manfred. 1983.
"Narration as Non-communication: On Ann Banfield's Unspeakable Sentences". Revised version of a paper originally published in Kölner Anglistische Papiere, 23 (1983).

Note: This is scanned text and may contain some scanner-typical errors.

1. "Showing/telling" and their linguistic correlates

The claim that the techniques developed by modern linguistics can be applied to literary data and used in literary theory to arrive at significant insights is nowadays often discredited as pure idealism. Yet Banfield's (1982) book is idealistic in just that sense. Approaching typically literary sentences with the system of Chomskyan generative grammar, Banfield explores stylistic features of narrative texts that were hitherto considered beyond linguistic analysis. The results have prompted one recent reviewer to go so far as to say "The book should be required reading for literary critics (and also for linguists)" (Epstein, 1982, p. 1280). Indeed, linguistic theory may find itself indebted to her for her analysis of free indirect discourse and the supporting evidence for the theory, first proposed by S.-Y. Kuroda (1973), that there are sentences which express something without at the same time communicating anything, in other words, that expressing something is an autonomous function of language (cf. Lyons, 1982); whereas the literary theorist may be brought to reconsider his position on traditional notions such as point of view, the mediatedness of narrative texts, and "dual voice" interpretations. Then again, the present formulation of the theory still provokes many counterexamples, an effect which Banfield's theses, developed in a number of articles from 1973 onwards, have produced in the past.

One interesting correspondence between linguistics and literary theory concerns the literary opposition "showing/telling" (Booth, 1961). Banfield considers "showing" to be closely related to the autonomous use and appearance of the linguistic expressive function. "Telling", on the other hand, may, but need not be, part of the linguistic activity of communicating. The major thesis in Banfield's theory is that sentences in narration are essentially non-communications -- they either express or tell something noncommunicatively. This is what the seemingly paradoxical title of her book refers to: narration consists of "unspeakable", non- communicated sentences, even if there is a speaker (e.g. a first person narrator).

This provocative stand challenges the widely held view that narrative literature belongs within a setting of embedded communication situations involving sender-receiver configurations such as author-reader, narrator-fictional addressee, implied author-implied reader etc. However, a few related approaches concerning narration exist, notably those of Emile Benveniste (1954) and Käte Hamburger (1973), whose supporting arguments Banfield discusses in some detail (in her chapter 4).

The main theses, both linguistic and literary, of Banfield's theory may be briefly summed up as follows. (1) Not all functions of language operate in the setting of language in communication. (2) One such independent function is the "expressive function", i.e., the use of language to represent "a picture of the activities and states of the mind" (p. 210). (3) This expressive function of language can be observed in isolation from a communication situation, particularly in the sentences of free indirect discourse -- "represented speech and thought" in Banfield's terminology. (4) The communication criterion divides the set of narrative forms into two distinct groups: (a) texts based on a communication situation and (b) texts essentially independent of a communication setting. One major point of Banfield's theory is that the bulk of narrative literature belongs to group (b).

2. The sentences of fiction

How can sentences reflect a fictional world? Subdividing "fictional world" into the usual ready-made concepts, Banfield provides the following answer: facts and events can be narrated; speech can be quoted (directly or indirectly) or represented (i.e., rendered three different ways); thought can be reported or represented; and sense data can be represented. "Representing" in Banfield means the mirroring of activities of the mind in the medium of language (cp. pp. 268, 273) and is the main mode to make use of the expressive function of language. Ultimately, narration is produced by the two activities of narrating and representing.

The building blocks of a fictional world are illustrated in the following examples. Most of the sentences are from Banfield, but I have occasionally substituted my own. Additionally, as grammatical person turns out to be an important feature, the paradigms have been arranged so as to present the two main options, third person and first person. Despite the fact that many literary critics might be tempted to arrange the categories on a mimetic scale, the order presented here is strictly and intentionally nominal.

Direct Speech
(1) She said to him/me, "I am tired." (p. 23)

The quotation proper -- the sentence within quotation marks -- may be conjoined with an optional introductory phrase which contains a verb of communication and identifies the communicants. Postposition or interposition of the introductory phrase, i.e. its use as a parenthetical as in ""I am tired", she said" is also possible.

Indirect Speech
(2a) Mrs. Dalloway said she would buy the flowers herself. (Woolf, Mrs. Dalloway/Banfield, p. 66)

(2b) And then I told Joe that I felt very miserable. (Dickens, Great Expectations, p. 65)

Indirect Thought
(3a) It occurred to Mr. Winkle that this advice was very like that which bystanders invariably give to the smallest boy in a Street fight 17...] (Dickens, Pickwick Papers, p. 30)

(3b) It was much upon my mind [...] that I ought to tell Joe the whole truth (Dickens, Great Expectations, p. 37)

Represented Speech
(4a) He'd so looked forward to going to Chartres with both of them, he said. (Dos Passos, l9l9/McHale, 1978, p. 252)

(4b) What I really wanted was rivets, by heaven! (Conrad, "Heart of Darkness"/Banfield p. 122)

The category "represented speech" identifies one of the two subsets of free indirect discourse; the other is illustrated in (5). (4a) shows the use of an optional disambiguating parenthetical specifying verbal communication.

Represented Thought
(5a) Not that she blamed the girl, and the marriage had been happy enough, she believed. (Woolf, To the Lighthouse/Banfield, p. 76)

(5b) If he was dead, I thought, I would see the reflection of the candles on the darkened blind 17...] (Joyce, Dubliners/Banfield, p. 95)

Both variants here show an optional parenthetical indicating reflective consciousness. Whereas the subject of the parenthetical in (4a) denotes a speaker, the subjects of the parentheticals in (5a) and (Sb) refer to a thinking subject of consciousness.

Represented Non-reflective Consciousness
(6) Now she/I could see a slice of the sky. (p. 201)

The category subsumes represented perception without being restricted to it: it also covers feelings, beliefs and other nonverbal sense data of a subject of consciousness.

Pure Narration
(7a) Then during the meal Mr. Arnoldsen gave a toast. (Hamburger/Banfield, p. 316 n. 1)

(7b) I fell unconscious. (p. 162)

(7a) and (7b) are "objective" sentences in contradistinction to the more "subjective" sentence types exemplified by (4), (5) and (6). No you referring to an addressee, no present tense or here/now referring to a current speech act could be inserted without producing a type change. In addition, the tense used in French would be the passe simple (aorist) as opposed to the impaifait which would occur in translations of (4), (5), and (6). A sentence of pure narration may have a speaker as in (7b) in a first person context; according to Banfield it would still remain "objective". Both (7a) and (7b) tell of events, but statements of fact, such as "He was stupid", are not by definition excluded (pp. 263, 269). (7a) is, in fact, an example coined by Käte Hamburger to demonstrate that the truth of such sentences may not be doubted in a fictional context. Banfield goes one step further by claiming the same for (7b).

These then are the main sentence types which are treated in Banfield's book. Actual classification plays only a minor role, as Banfield is "not attempting a taxonomy of narrative sentences" (p. 18), and consequently no systematic table of sentence types is offered. A systematic gap, namely the category "direct" or "quoted thought" is quite apparent here, a point which will be taken up below.

(The typical sentences of narration, according to Banfield, correspond to types (4) to (7); i.e., are sentences that either represent consciousness or narrate objectively. How do the first three types, direct speech, indirect speech, and indirect thought fit into the system? All can and do appear in narration. Direct speech is exceptional in that it introduces a new text (cf. below) which sets its own frame of reference as a representation of independent discourse. Although Banfield never discusses this point, there is some reason to assume that sentences of indirect speech and indirect thought are technically sentences of pure narration. But this assumption has unwelcome consequences. It might perhaps be claimed that the indirect speech sentence (2a), for instance, is non-communicated, non-spoken, because it occurs in narration. But whatever indirect speech is, it is not unspeakable. Thus something is at fault: either the above classification or Banfield's terminology. Of course it is also a terminological paradox to admit speakers to texts consisting of unspeakable sentences.)

Four further terms are important to Banfield's argument: Self, Speaker, Addressee and Text. To emphasize their special assigned meanings Banfield consistently capitalizes them. The Self is a subject of consciousness (a reflector, in literary terminology) and a necessary element of sentences (4) to (6), but not (7); section 4 explains what happens to the Self of the Speaker in (7b). The Self is an agent capable at least of consciousness, usually also of thought and speech, so that a preverbal infant (p. 213), for example, or an animate machine in a science fiction context would qualify. The remaining three terms are defined by simple formal linguistic properties. The Speaker is the agent who is identified by the deictic "I"; an Addressee is the referent of the second person ("you"). Finally, a Text is a sequence of sentences with an identical Speaker.

Relating these definitions to the sample sentences above, one notices that our awareness of the function of the agents in their fictional world can be a source of confusion. By definition, there are two Speakers in (1), the referents of "me" and "I", respectively. In (4a) there is a represented speaker ("he"), but he is no Speaker since there is no "I". Mrs. Dalloway in (2a) is an indirectly quoted speaker, but, by the same reasoning, she is no Speaker, either. Similarly (and, as will be seen, significantly), as none of the sentences (1) to (7) contains a "you",none of them contains an Addressee. This is entirely independent of the fact that we recognize that "him/me" in (1) are addressees. (If any of this appears confusing, check capitalization.)

The categories and definitions reported so far will generally excite little comment as most of them have been known and recognized for some time, albeit sometimes under different names. Some of Banfield's terms are immediately convincing, others perhaps less so. The term "represented non-reflective consciousness", for instance, seems rather unwieldy. "Pure narration" is somewhat ambiguous as "pure" could erroneously be taken to imply a value judgment or an indication of speakerlessness (it actually seems to mean non-subjective). On the other hand, Banfield's terms for the phenomenon hitherto variously identified as free indirect discourse, style indirect libre, erlebte Rede, narrated monologue, etc., are a welcome relief. Both represented speech and represented thought are explicit where the traditional terms are vague or misleading: they avoid the inadequate association with the indirectness of indirect speech (p. 70), they make no use of the ambiguous word "free" (cf. Bonheim, 1982, p. 60; Banfield n. 14, p. 277), and, above all, "represented thought" does not imply a representation of speech where there is no speech to be represented. Nor is clarity gained at the price of two terms for one, as the two terms very easily can, and in Banfield's treatment almost always are, combined in the compound category "represented speech and thought". Further compound categories in Banfield are "indirect speech and thought" (p. 36) and "represented consciousness" (which combines represented speech, represented thought, and represented non-reflective consciousness) (p. 203). Of course, compound categories are potentially dangerous if one forgets that they were made up from independent and, as in these cases, mutually exclusive terms. Even though Banfield generally uses compound or basic terms as appropriate, the following quotation shows an instance where she has failed to decompound "represented speech and thought":

Actually, of course, "represented speech" cannot be used to represent a character's thoughts, just as "represented thought" cannot be used to represent a character's speech; it is therefore misleading to say that "represented speech and thought" could represent a character's thought. Interestingly, saying (more correctly) that "the natural mode to represent thought is "represented thought"" sounds even more tautological than the original wording and may well lead an unsuspecting reader to accept the statement as reasonably true -- unless one pauses to consider possible "less natural" modes. Indeed, consider a parallel question. What would be "the natural mode" for representing speech? A reasoned answer would very likely avoid the terminological trap and opt for direct speech, not represented speech. Thus, the question remains why the theory does not admit a category "quoted thought" and how it copes with sentences such as (8):

(8) Let me alone! screamed Anthony silently. Let go of me! (Metalious, The Tight White Collar, p. 134)

Here, to all appearances, the "quoted" parts closely resemble direct speech, which of course they aren"t -- as witness the information carried by the parenthetical. Often the parenthetical or an equivalent introductory phrase is missing, and we have an interior monologue if we encounter a long sequence of sentences of quoted thought. But neither quoted thought nor interior monologue find due systematic recognition in Banfield's theory.

3.Quoted thought

It must be admitted that both concepts, quoted thought and interior monologue, are often challenged. Both imply that thought is a kind of "inner speech" and neglect preverbal thinking and non-verbal mental states or activities. For Banfield the assumption of "inner speech" is a fallacy, since "thought is not linguistic in form" (p. 80), "thought can be indirectly but not directly quoted" (p. 36), and "we cannot say "John thought this" without thereby implying, perhaps falsely, that John's thinking consists of inner speech" (p. 80). The terminological gap, the missing category "quoted thought", is therefore intentional. On the other hand, it can hardly be denied that the style exemplified by (8) exists in fiction and can in no way be called anomalous. Is the fallacy of inner speech really so serious as to discredit the category quoted thought? What category other than quoted thought could be made to cover a sentence such as (8)? Banfield's answer is to make recourse to the term "direct speech" and to use it in a double capacity. Arguing that all relevant examples such as the paradigmatic "Mary thought to herself: "I must be late"" are syntactically indistinguishable from direct speech (p. 78), Banfield's conclusion is that they must be considered to be a special case of direct speech. Sometimes even that qualification is dropped and sentences such as the one from To the Lighthouse -- "He thought, women are always like that" (pp. 35, 78) -- are unequivocally labelled direct speech even though the actual context confirms what the verb says, that this is thought, not speech. Clearly, this treatment does not exactly avoid the fallacy that thought is speech. It would also appear that the category error of calling thought speech is far more serious than the assumption that thought is quotable. One recalls the hint that the "natural mode" of representing thought is the technique represented thought. Although Banfield explicitly denies that she prefers represented thought over quoted thought -- there is a footnote that says "This should not be taken to imply that represented speech and thought is of greater aesthetic value than interior monologue" (p. 300, n. 14, note again the redundant and misleading "speech and"), her refusal to assign the style a proper name really speaks for itself.

Since there is no denying the fact that the style exists, it seems only sensible to assign a proper name to it such as "quoted thought" and assume that represented thought and quoted thought are in fact variant andlor complementary rather than rival techniques. In fact there is some reason to stress their complementary character even in cases where one can be easily translated into the other and vice versa. The two forms produce different effects of distance and focus; very generally speaking, represented thought appears to be more remote and less focussed than quoted thought. A significant point in this context is that there are cases where represented thought cannot be translated into quoted thought and vice versa. Banfield herself, reviewing the findings of a previous article, shows that neither style is a transformational derivation of the other. The unique applicability of represented thought can be appreciated in the following passage. Note particularly the concluding sentence.

The phrase "and for the most part in pictures" can be added without in the least clashing with what was expressed before, and this fact demonstrates how represented thought avoids what would be (in this case) a problematic, even impossible, inner speech act. For the same reason it would be anomalous to construct a quoted thought version such as

(9) "I will return to the earth in male human shape", thought U Po Kyin, for the most part in pictures.

On the other hand, even though represented thought is flexible enough to encompass a variety of phenomena such as hesitations, repetitions and even incomplete sentences, Banfield herself shows that certain words and constructions cannot be accommodated in represented thought. One such construction is the imperative, and this is the reason why there is no represented thought variant to (8) such as "Let him alone! screamed Anthony silently", with "him" intended to refer to Anthony himself.

A further point in support of the term quoted thought can be adduced by taking a closer look at its counterpart, direct speech. Naturally, there is no inner speech problem for direct speech, and neither can it be claimed that direct speech is inaccurate (another objection often raised against quoted thought). In fact, as many linguistic studies show, a written representation of direct speech allows highly exact transcriptions of speech utterances, not only word for word quotations but indications of emphasis, pronunciation, intonation etc. Even if such super-realistic detailing of direct speech is apparently possible, it is striking how little use is actually made of this possibility in fictional texts, where (excepting the occasional marking of pronunciation and emphasis) direct speech is hardly ever an attempt at exact transcription of speech. On the contrary: in fiction as well as in almost any other context direct speech involves filtering out the performance errors which occur in actual speech, and other linguistic paraphernalia such as intonation and pronunciation are generally left to the recreative competence of the reader. Direct speech is therefore an "imitation of idealized speech" as Banfield convincingly puts it (p. 248), not a true copy or reproduction (such as a recording). If one accepts the conclusion that people do not speak exactly as they are generally quoted, then a similar consideration should apply to quoted thought. If quoted thought is considered as an imitation of idealized thought the major objection against the style loses a good deal of ground. Both direct speech and quoted thought are then imitations of idealized phenomena in a written medium. This assumption makes the concept quoted thought accessible to an analysis unburdened by considerations of naturalness and accurateness.

4. Subjective expressions, deictics, and the division of narrative forms

In order to analyse the linguistic properties of her sentence types, and to determine how, if at all, they differ from non-literary sentences, Banfield pays particular attention to their behaviour in connection with subjective expressions and deictics. Subjective expressions must originate in a Self, and their presence in sentences of fiction point to a character or a narrator as the source of an individual point of view. The following list distributes the main types of subjective expressions into three groups (cf. Banfield, pp. 28, 53, 90, 114).

Subjective expressions are a powerful testing material for pointing up the differences between most of the basic sentence types. For instance, whereas direct speech can freely incorporate elements of all groups, indirect speech cannot incorporate elements of A or B (these list "non-embeddable" expressions of subjectivity only). represented speech and thought manages the items of groups A and C but fails with those of group B. Represented non-reflective consciousness cannot include items of A or B without producing a type change in the direction of represented thought; and pure narration does not work with items of any of the groups.

(As some of these findings were published in earlier articles by Banfield, a number of counterexamples have been proposed. McHale (1978, p. 255) cites examples from Dos Passos which appear to refute the claim that indirect speech or indirect thought cannot contain items of group B, e.g., "He thought to himself he was damn lucky to get away from [...] that sonofabitchin" foreman". Even though Banfield's discussion of McHale's examples appears to be rather disoriented (she does not seem to recognize them as indirect speech or indirect thought), oddly enough, their dismissal on the assumption of "partial quotations" (p. 115) can be considered plausible. Since Banfield says "I have not found any unquestionable cases of represented thought with phonetic material" (p. 115), I submit "Ah, it was great, real great man, to just sit and dig the radio and smell the car, that special CADILLAC smell and not have those ghuddamn houserats all ovuhya" (Selby, Last Exit to Brooklyn, p. 275), at the same time admitting its exceptional character.)

Since a sentence can, theoretically at least, contain an arbitrary number of human agents, each with his or her own point of view, an obvious question is:
How are the subjective expressions of a sentence correlated to one or many Selves? And is there a limit on the number of Selves that can appear in a sentence? On the evidence produced by Banfield it appears that all subjective expressions fasten to one Self; and it follows that a sentence cannot express more than one Self. Banfield calls this the principle of "1 EXPRESSION! 1 SELF" (p. 93), read: for every independent sentence, which may optionally contain subjective expressions, there is at most one Self. It is this principle which clashes with traditional assumptions of embedded communication levels in narrative fiction, and also "dual voice" theories of certain sentence types, especially third person represented thought. To illustrate, consider first a direct speech sentence liberally loaded with subjective expressions:

(10) "Yes, by God, howl hate that fool of a doctor!" John said.

All subjective expressions here attach to the Self of the Speaker, John, the referent of "I". If John's statement is to be reported by somebody using indirect speech, the subjective expressions all have to be dropped as indirect speech only reports the "propositional content" of an utterance (although the tendency of the expressive force could be added descriptively):

(11) John said that he hated the doctor, whom he considered a fool.

Represented thought, as the main contestant for the dual voice theory, is particularly interesting in its behaviour towards subjective expressions. As the subjective expressions in the original sentence (9) all come from groups A and C they can all be incorporated in a represented thought version:

(12) Yes, by God, how he hated that fool of a doctor! John thought.

As predicted by Banfield, again all subjective expressions attach to one Self, namely John's. If the dual voice theory of represented thought were correct it should at least theoretically be possible to include a second "voice", i.e. inject a subjective expression by a narrator, for instance. Indeed, a crucial test case obtains if we assume that there is a narrator who believes the doctor to be nice and John's judgment to be quite wrong. Would the following sentence present this state of affairs?

(13) How he hated that fine fellow of a doctor! John thought.

This does not work. If (13) can be interpreted at all, one is forced to assume that "fine fellow" is a schizophrenic contradiction on John's part.

As it is apparently impossible to display differently anchored subjective expression in one sentence, the dual voice theory of represented thought is severely challenged. If, in order to articulate a "dual voice" sentence, one of the voices must be devoid of any subjective expressivity then the meaning of "dual voice" is quite unclear.

The principle "1 Expression/1 Self" is only superficially threatened by two apparent counterexamples. The first is a case of direct speech such as

(14) The hell she said, "To hell with it!" (p. 57)

Here we encounter the Selves of both the quoting and the quoted speaker, correlated to the subjective expressions "The hell" and "To hell", respectively. However, the rules of the grammatical system employed by Banfield analyse (14) (and likewise (1)) as consisting of two independent sentences (in deep structure) so that the principle may be taken to hold for each. The second counterexample appears in sentences like

(15) But to help him, they reflected, was impossible. (Woolf, Mrs. Dalloway/ Banfield, p. 96)

Although attached to several agents ("they"), this is nevertheless a "single point of view"; in other words, a Self can "be plural -- just as the first person may be plural" (p. 96).

The sentence types (1) to (7) also differ in the ways in which they use the deictics -- words like I, you, now, here, etc. In an ordinary discourse situation with a communication setting these as well as the tenses all relate to the act of discourse in which I is the Speaker, you the Addressee and the time/place deictics are all anchored to the current speech act. Direct speech presents an image of a discourse situation and thus freely employs all deictics. Third person represented speech and thought as well as non-reflective consciousness cannot include the Addressee deictic you. Here and now can appear in represented consciousness, but instead of necessarily relating to a speech act, they more generally relate to an act of consciousness which may be speech (as in represented speech), but can also be an act of thinking (represented thought) or an act of sense data processing (non-reflective consciousness). Within represented consciousness the past tense significantly co-occurs with the deictics here and now and thus excludes the occurrence of the present tense with its own associated heres and nows. In third person pure narration there is neither a speech act nor any other act of consciousness to which a now could relate; a past progressive conveying simultaneity with an act of consciousness would also be excluded. The third person versions of the sentences therefore systematically lack Speakers, Addressees and the present tense of a current act of discourse, and it is thus difficult or impossible to maintain that a communication situation still holds. As these sentence types are "characteristic" in narration (pp. 164, 180), Banfield concludes that the feature of noncommunication should be a defining property of narration.

(Stated thus, the argument still looks vulnerable. Even granted that the characteristic sentences in narration are those of pure narration and represented consciousness, it is neither obvious nor logically necessary to assume that the construct narration has the properties of its characteristic or even essential sentences. The conclusion also seems incautious in view of the fact that a text of narration usually does not exhaustively consist of sentences of pure narration and represented consciousness; direct speech, for instance, belongs to neither.)

Banfield's conclusion, if true, supports the division of the field of narration as proposed by Hamburger. But whilst Hamburger sets a categorical dividing line between third person and first person texts, Banfield goes one step further in an attempt to include the majority of first person texts in a common domain of narration.

First person narratives employ sentences containing a Speaker identifiable by the use of the first person. Such narratives can obviously not be called speaker-less or narratorless. What Banfield questions is whether the presence of a Speaker necessarily presupposes a complete communication setting. Whilst she does not deny that a speaker is a necessary part of a communication situation, the point is whether Speaker presence is a sufficient condition. If this were so, an I would automatically entail a you, an utterance act would be established and this in turn would sooner or later trigger the present tense which relates to it. Claiming that this is generally not the case in narration, Banfield goes on to argue that it is the presence of a you rather than an I that automatically presupposes all the other elements of a communicative framework.

To establish this thesis in more detail consider the typical case of a Speaker who talks mainly about past events he has himself experienced. Such a Speaker must have two distinguishable Selves -- a current Self, or the Self of the "narrating I", and a past Self (the Self of the "experiencing I"). The principle 1 Expression! 1 Self also holds for the two I's; it is impossible for both Selves to make their presence felt via subjective expressions in any one sentence.

Banfield establishes this point with the following inadmissable sentence (note particularly the expressive How-construction, the progressive form indicating simultaneity, and the deictic now):

(16) How my heart was beating now, I remember now (p. 160)

Neither, argues Banfield, can two nows be anchored to two points in time in one sentence, nor can a "Self-in-the-Past [...] coexist with a Self-in-the-Present" (p. 160). However, since Selves are by definition originators of point of view and identifiable by subjective expressions, there are typical sentences for each of the two Selves of a Speaker. One is

(17) How my heart was beating then, I remember now

in which the present tense in the parenthetical relates to the time of an utterance, the expressive construction attaches to the subjectivity of the current Self, and a you could easily be present (as in "I can tell you now"). The other sentence is

(18) How my heart was beating now, I realized then

Here the deictic now relates to an act of consciousness on the part of the past Self. There is neither a present tense relating to a current utterance, no possible you relating to an Addressee, and none of these could be inserted. It is sentences like (18) which show that the presence of an I does not necessarily entail the presence of a you and that first person sentences are not by definition based on a complete communication setting.

Banfield goes on to assert that the sentence type usually encountered in first person narration has non-communication characteristics like the one exemplified by (18). In spite of the presence of a Speaker no Hearer is addressed; the sentence is just as "unspeakable" as its third person counterpart. The first person in first person narration, like the one in (18), is "a SPEAKER whose SELFness is suppressed; this is the narrator of first person narration" (p. 160).

(This singling out of sentence type (18) over (17) as the paradigm for first person narration seems rather restrictive. Banfield goes on to deduce that the second person and the present tense cannot occur in first person narration, points that will be discussed in the next section. The original sentence from Dubliners ("An Encounter"), of which (16) to (18) are synthetic modifications, is "How my heart beat as he came running across the field to me!" This is cited on p. 158 with the comment that the exclamatory force of the How-construction is attributable to the sentence's Self. Which Self is that? The "expressivity" of the sentence represents a present reaction to a past event on the part of the speaking or narrating Self rather than of his past Self" (p. 160), in other words, Joyce's sentence corresponds to (17) rather than to (18). Quoting the same evidence in an earlier article Banfield (1978, p. 430) states that "the exclamation expresses a present response to a past eventt and adds: ""a reading appropriate to a story about childhood on the part of a mature narrator". Thus the original model sentence, taken from a text of first person narration, does not show what is supposedly the essential feature of first person narration, namely a Speaker with a suppressed Self.)

Banfield's reasoning leads to a division of narrative texts in which narration, either third or first person, is distinct from discourse because its characteristic sentences are noncommunicative. Additionally, however, a deviant first person narrative text is allowed which is considered to be no text of narration: this is the type of text called skaz, which is firmly tied to a complete communication setting, has a Speaker (a "story-teller" or "raconteur", p. 172) and presupposes an audience of possible interlocutors.

5.Skaz and first person narration

The original meaning of the word skaz is "speech", but the Russian formalists, with whom it originated, use it mainly to denote

These definitions suggest that skaz is quite closely related to the dramatic monologue (Banfield confirms this in n. 25, p. 306), so much so, in fact, that one might perhaps simply say skaz is a dramatic monologue in prose.

Banfield substitutes "storyteller", "letter writer" etc. for the term narrator in order to avoid the implication that narration (in the sense of her definition) is involved. One consequence of the definition is that epistolary novels must be considered to be a form of skaz or perhaps an ordered sequence of skaz texts. As to the currency of the genre Banfield states that whilst "Eastern European literature abounds in examples of skaz narratives" (p. 172), non-epistolary skaz texts are "rare in English". "Commonly cited examples" are Ring Lardner's "Haircut", Mark Twain's "The Celebrated Jumping Frog" and Huckleberry Finn, and the frame stories in some of Conrad's novels. To this Banfield adds the "Cyclops" episode from Ulysses ("hitherto unnoticed") and the first version of Faulkner's "Spotted Horses", of which she presents a detailed analysis.

Skaz is understood to have the following distinctive formal features: the use of the second person for the potential interlocutor, references to the current speech act and possibly indications of pronunciation or spelling (p. 174, cf. n. 24, p. 306). All first person narratives that systematically exhibit these features, particularly the first, must by definition be skaz texts.

Generally speaking however, the presence of you in first person narratives is hardly a rarity -- unless, of course, these occurrences are unsystematic and accidental. Consider the following excerpts.

In these examples an explicit you or the you of an imperative conspicuously cooccurs with the present tense, which to all appearances refers to a current speech act ("speech act" to include an act of writing). Moreover, the referent of you is in some cases indeed a potential, if not actual interlocutor. This is obviously so in (a) and (b) where the narrators (story-tellers?) react to the presence of their addressees. The actual reader of (a) has not said, and could not possibly say (at this point) that the narrator is mad, and he cannot be flattered into believing he knows the nature of the narrator's soul as suggested in (b). The you addressed must be somebody else. The first two or three excerpts quoted admittedly also have the flavour of a dramatic monologue. They conjure up a mental picture of their respective speech situations. This is particularly obvious in (b) where the smug narrator addresses his story about his murder of Fortunato to an apparently equally smug friend. Leaving aside (c) for a moment, cases (d) to (f) seem different from (a) and (b) in appearing to address the actual reader. In fact, he could hardly pretend he is not being spoken to when asked to "take that first sentence" (e) or take the following episode "for instance" (f). If he did not do as requested, he might just as well put the text aside.

This analysis seems to argue for a dividing line between cases (c) and (d) according to whether the Addressee is a fictional agent or the actual reader and allot (a) to (c) to skaz and (d) to (f) to first person narration. That Catcher in the Rye should be considered a skaz text is, among other things, supported by the affinities it bears to Huckleberry Finn. This division is, unfortunately, based on non-formal and as yet vague ideas about actual readers and Adressees -- clearly not a satisfying procedure. The alternative is to accept a formal criterion only. Since the major formal criterion is presence of you, all cases quoted must be skaz. This conclusion however, is a case of proving too much: skaz would then clearly no longer be rare in English and would indeed subsume a great number of ordinary, if not typical first person narrations.

In view of this it is not surprising that problems also arise with the formal properties ascribed to first person narrations. Consider the following two statements by Banfield:

(19) When a present NOW occurs, the sentence is, of course, an intrusion of discourse (p. 163)

(20) In first person narration, there is no moment of narration, no speech act and hence, no PRESENT. (p. 164)

The assertion in (20) is that first person narration cannot contain the present tense. But unless all examples (a) to (f) are considered skaz, some of them, say (d) to (f), are clearly counterexamples. Or could they, in the light of (19) be dismissed as (insignificant) "intrusions of discourse"? Banfield herself says, though in a different context, that "the true narrating I never intrudes" (p. 212). Now, what follows? If (d) to (f) are not intrusions then the theory fails to account for them. Alternatively, the narrators are no "true" first person narrators, whatever that means.

Would it help to turn to what Banfield believes to be an uncontestable "classic first person narration", namely David Copperfield (p. 171)? Consider its first sentence (incidentally, and perhaps significantly, (a) to (f) are also all first-page excerpts):

Here, right from the beginning, we have current reflections of the narrating I in the present tense. There is no explicit you, but its consideration of a potential reader, and that is a potential you, is quite tangible (these pages must show whom?). Traditional literary theory would hold that passages such as these are a highly characteristic feature of any fictional autobiography and that the sentiments and judgments expressed in these "intrusions" are of considerable importance.

6.The narrative present

As was demonstrated in the last section, present tense passages from first person narrations are on the whole not difficult to find, and Banfield's theory, although it acknowledges their existence, shows a tendency towards relegating them to non-narrational forms like skaz. Apart from the present tense in narratonal comment, there is also a present tense used as a narrative tense, the so- called historical present, which gives rise to similar problems.

Although Banfield claims Käte Hamburger's support on a number of points, e.g. the question of narratorless third person narration, the definition of the sentence of pure narration, and the characteristics of the "epic pretenite", she refrains from consulting her on the question of the historical present. Hamburger, discussing the stylistic and aesthetic aspects of the present tense in narration, summarily condemns the tense as having no true function, and as being redundant. This harsh judgment is rooted in her theory that the epic preterite in fiction, having lost its past tense meaning, fulfills the main function of the historical present, a point that is subscribed to by Banfield. But Hamburger's critical assessment is obviously not shared by authors. In fact, the popularity of the present as a narrative tense in German fiction is so considerable that one can hardly justify treating it as a parasitic or subsidiary tense. That is why F. K. Stanzel has recently (1982, p. 135) simply discarded the misleading "historical" and coined the term "narrative present" in order to emphasize its independent character, particularly as opposed to the epic pretenite.

Of course, Banfield could not adopt Hamburger's normative stance, even if she wanted to. The present tense occurs in the data, and the theory must either exclude this feature or else account for it somehow. That this is a problem for Banfield's approach follows from the fact that the present tense, the present progressive and deictics such as now are all understood as referring to the moment of a speech act in a discourse situation (p. 133). But the theory also claims that a discourse situation is foreign to narration. The present tense is therefore barred from certain typical sentences, particularly those representing consciousness, which do not "normally allow the present tense at all" (p. 98). In represented speech and thought "no non-generic present tense appears" (p. 100; possibly explaining "normally" in the preceding quotation), and "It is [...] the absence of any utterance in the represented E[xpression] which explains the absence of the tense indicating the moment of utterance" (p. 121). The present tense is therefore practically restricted to the sentences of pure narration, i.e. sentences without subjective expressions, deictics, and the progressive form (p. 164) asserting facts or events.

The sheer number of counterexamples that can be presented here makes this position untenable. Even Hamburger quotes a German literary example of present tense erlebte Rede, and Pascal (1977, p. 41 ff.) presents several present tense free indirect speech passages from Goethe. The examples are admittedly all German, but Banfield's theory aims to be universal.

If needed, English counterexamples can also be found. The present tense chapters of Edwin Drood, for instance, yield examples of (i) present tense in conjunction with what Banfield would consider to be an odd "now" in a narrative statement, (ii) a dialogue in present tense represented speech and (iii) an instance of present tense represented thought, all predicted as non-occurring.

These counterexamples cannot be dismissed as exceptional or anomalous, and exclusion of the data on aesthetic grounds is also Out of the question. However, it would certainly be interesting to determine whether the theory could cope with the narrative present as an alternative narrative tense. For a detailed treatment of the present tense in English narration see Casparis (1975) who provides further counterexamples.

7.Narratorless texts

There are no "third person narrators"; even the notion is contradictory, as Tamir (1976, p. 415) has pointed out. Every narrator is a first person narrator. The crucial question is, can there be narrators in third person texts without endangering the concepts "third person text" and "narrator"? Banfield narrowly defines a narrator as the referent of recurring "I's" outside sentences of direct speech (again suppressing that disagreeable category, interior monologue). Furthermore, a narrator is the originator of "his" text and "every sentence is attributed to this fictional persona and his point of view" (p. 212). His text is an "intentional object" (p. 212), he is "the single unifying voice" (p. 183), "the locus of the text's meaning" (p. 184). Note well that all this applies to narrators of first person texts and, according to Banfield, to this type of narration only.

By contrast, Banfield (following Hamburger and Benveniste) considers third person texts, if they consist of sentences lacking the first person outside direct speech, as being narratorless. Still, in order to locate the unity, meaning and point of view of a third person text, most critics posit a narrator, in analogy to the narrator in first person narration, even if there is no overt "I" in the text. Most critics would indeed distinguish several types of narrators in third person narration with respect to their degree of presence, obtrusiveness, omniscience etc. Even an "effaced narrator" would still be endowed with a shadowy presence as selector, arranger or simply, and conveniently, as unifying principle. This practice more or less naturally leads to "polyphonic" or dual voice interpretations.

Here Banfield notes that the assumptions have become circular. The reasoning starts off by positing (rightly) that the narrator in first person narration is the integrator of the text's unity; they conclude by saying (possibly wrongly) that because a text has unity it must have an integrator in the form of a narrator. But just as the dual voice theory fails to give an adequate account of third person represented speech and thought (in which no narrator's voice can actually occur), there is no compelling reason to posit a narrator for the other types of sentences occurring in third person narration. A third person narration needs no narrator as an abstract unifying construct. The critic must look for unity in the text itself and proceed from the narrative facts, to which any text is bound by conditions of narrative consistency. In short, he must turn to "the objeetivized [third person] text, which must be held together by some other hypothesis than that of the narrator's voice" (p. 222).

This approach requires a qualified response -- with some third person narratives it appears quite sensible, with others it doesn"t seem to work equally well. First, there are third person texts where the supposed narrator does no more than insert parentheticals such as "she thought" in appropriate places, or shift tenses and pronouns to produce represented speech and thought. Banfield convincingly shows that this "operational" definition of represented speech and thought is inadequate. But even if it were right one would wonder how such an automatism should prove to be the key to the text's integrity. Surely nobody would accept a tape recorder, or the person who transcribed a taped event and mechanically added speaker identifications as such an integrator. Yet the decision on whether a third person narration has a narrator or not is often based on mechanical trivialities such as these, which may actually only reflect certain housekeeping routines of the narrative medium.

On the other hand, Banfield's approach to third person texts encounters problems whenever the text contains "authorial intrusions". According to Banfield, these are "special cases where a first person sentence lays claim to being the author's voice" (p. 211). Why the author's voice? Banfield chides literary critics for their insufficient differentiation between the concepts author and narrator (p. 184), but her own use of the pair is not quite clear either. In the instance just quoted she appears to refer any first person in third person contexts to "the author's voice". On another occasion, discussing Lawrence's "England, My England" and Lady Chatterley's Lover (both third person texts) she freely uses the terms "Lawrence's narrator", "the text's first person subject of consciousness", "narrating I" etc. (1973, p. 35). But irrespective of whether "authorial intrusions" originate with the author or with a narrator, they are in any case irrelevant to the essence of third person narration, according to Banfield and her precursors and followers. Thus Käte Hamburger disarmingly asserts that "there is also no such fictive narrator in cases where this impression might be awakened by interspersed first-person flourishes such as "I", "we", "our hero", and the like" (1973, p. 139). Kuroda (1976, p. 140, n. 8) fabricates so-called "local narrators" so that "some sentences of a narrative, or even only some constituents of sentences (e.g. clauses modifying a noun), may be attributed to narrators of various sorts, without, however, the entire narrative being attributed to a narrator or narrators". Finally, Banfield herself considers authonial intrusions as extraneous formulae, "special cases" separable from the main text, certainly no more than what the term says -- i.e. intrusions.

The question of the relevance of authonial intrusions was one major bone of contention between Käte Hamburger and Franz K. Stanzel in their critical debate in the fifties and sixties. It may still be instructive to view Stanzel's traditional approach in the light of Banfield's new perspective. In Stanzel's recent reformulation of his theory, third person texts are broadly divided up into two types -- "authonial narratives" which are mediated and integrated by an "authorial narrator" who is the originator of (and identified by) his authorial intrusions (typical case: Tom Jones) and "figural narratives" such as Mrs. Dalloway which, in their pure form, are narratorless. Thus, there is at least a partial correspondence between the two theories in that narratorlessness is admitted in both.

However, given Stanzel's position, the concept of a narratorless figural narration promptly runs into trouble. Since Stanzel is an adherent of the dual voice approach and consequently uses a set of fairly permissive criteria of what constitutes narratonial presence, he is at pains to isolate truly pure (i.e. truly narrator- less) examples of figural novels. Stanzel himself readily points out narratorial traces in supposedly figural novels, and in other so-called figural texts apparently only selected passages qualify, such as chapter 5 (which chapter 5?) of To the Lighthouse. Although Stanzel glosses over this by saying that he is only presenting a system of ideal types, of abstract and intentionally "weltfremd" constructs whose systematic characteristics may never be fulfilled by any one text, it will be observed that neither authonial nor first person narrations are at all weltfremd. The problem with the category figural novel may be that the true figural novel has not yet been written. Alternatively, the source of the problem may lie with the definition of what is and what is not a narratorial trace.

A comparison of the two theories leads to the conclusion that Banfield's approach does no justice to authorial narrations, while Stanzel's approach fails to show that figural narrations actually exist. Banfield all too radically posits narratorlessness for all third person texts, and Stanzel all too easily finds narratorial traces in practically all third person texts.

The question, then, is whether the two approaches preclude mutual recognition of individual strengths and weaknesses, and also, of course, whether they are compatible in their strong points. Banfield's theory would certainly benefit from Stanzel's concept of a distinct category of third person authorial narration. The adoption of this concept would mean that the relevance of authorial intrusions need no longer be denied. On the other hand, in view of Banfield's well- argued attack on dual voice interpretations, Stanzel might reassess his understanding of narratorial traces. To be specific, such a reassessment could proceed from the following considerations.

According to Stanzel, among the traces of vestigial narratorial presence are the following (page references for Stanzel, 1982):

(1) Short passages of "dual voice" erlebte Rede (pp. 247, 255)

(2) Outside views as in "He clasped his hands and raised them towards the white form" (p. 84, from Portrait of the Artist)

(3) Language level of narrative report higher than that of reflector (p. 256)

(4) Joint reflectors (p. 215)

(5) Non-pronominal reflector identification such as "Mrs. Dalloway" (p. 213)

(6) Anonymous or missing reflectors (p. 225)

If these features signal narratorial presence then it is difficult to see how any text at all could ever qualify as a "pure" figural text. But all narratorial implications are categorically denied in Banfield's system. (1) Stanzel's dual voice interpretation of erlebte Rede would be refuted on the strength of the principle 1 Expression! 1 Self which posits that a narrator cannot intrude into a sentence of represented speech and thought. (2) Although supported by Cohn (1978, p. 102), the example quoted might just as well be a representation of Stephen Daedalus's consciousness (as the vagueness of "white form" indeed suggests) than of what is visible to an outside narrator. (3) Although a number of narrative sentences, particularly those representing consciousness, approach the language level of the reflector in a "mind style", the language level of represented consciousness is in principle independent of the language level of the reflector; it is a representation of something, not a quotation, and not a symptom of somebody else's (i.e. the narrator's) style (cp. Banfield, p. 213 f.). (4) A plural Self is allowed in Banfield, and it is not obvious why it should signal a narrator. (5) Banfield admits that reference to the Self would be pronominal in represented speech and thought, but claims that a proper name is possible in sentences of non-reflective consciousness; cf. the views of Banfield and Stanzel's on "Father Conmee was wonderfully well indeed" (p. 208), which could well be "speech represented [...] in a non-reflective form", e.g. heard speech (p. 208, Stanzel, p. 229). (6) Because Banfield does not accept the position that narrative texts are necessarily mediated, absence of a reflector does not automatically entail presence of a narrator.

Banfield's points seem to be well taken. Adopting her position would have two consequences: whilst the dual voice hypothesis would have to be curtailed or abandoned, the concept of figural narration would profit considerably. Figural narration would no longer be a type of text that is "out of the world"; instead, Portrait of the Artist, Mrs. Dalloway and many others could be admitted as fair and square examples of the type rather than borderline cases between uncertain categories.


Banfield, Ann. 1973.
"Narrative Style and the Grammar of Direct and Indirect Speech". Foundations of Language, 10, 1-39.
---. 1978. "Where Epistemology, Style, and Grammar Meet Literary History: The Development of Represented Speech and Thought". New Literary History, 9, 415-454.
---. 1982.
Unspeakable Sentences: Narration and Representation in the Language of Fiction. London: Routledge & Kegan Paul.
Benveniste, Emile.
1959. "Les relations de temps dans le verbe francais". Problemes de linguistique generale, chap. XIX. Paris: Gallimard, 1966.
Bonheim, Helmut.
1982. The Narrative Modes: Techniques of the Short Story. Cambridge: Brewer.
Booth, Wayne C. 1961.
The Rhetoric of Fiction. Chicago: U. of Chicago Pr.
Casparis, Christian Paul.
1975. Tense Without Time: The Present Tense in Narration. Bern: Francke.
Cohn, Dorrit. 1978.
Transparent Minds: Narrative Modes for Presenting Consciousness in Fiction. Princeton: Princeton U.P.
Epstein, E.L. 1982.
"Communicating the conscious". Times Literary Supplement, Nov. 19, 1982, p. 1280.
Hamburger, Käte. 1973.
The Logic of Literature. London: Indiana Univ. Pr.
Kuroda, S.-Y. 1973.
"Where Epistemology, Style and Grammar Meet: A Case Study from the Japanese". Festschrift for Morris Halle, ed. P. Kiparski and S. Anderson. New York: Holt, Rinehart & Winston.
Lyons, John. 1982.
"Deixis and Subjectivity: Loquor, ergo sum?". Speech, Place, and Action, ed. Robert J. Jarvella and W. Klein. Chichester: Wiley.
McHale, Brian. 1978.
"Free Indirect Discourse: A Survey of Recent Accounts". Journal for Descriptive Poetics and Theory of Literature, 3,249-287.
Pascal, Roy. 1977.
The Dual Voice. Manchester Univ. Pr.
Stanzel, Franz K. 1982.
Theorie des Erzählens. 2nd rev. ed. Göttingen: Vandenhoeck & Rupprecht.
Tamir, Nomi. 1976.
"Personal Narrative and Its Linguistic Foundation". PTL: Journal for Descriptive Poetics and Theory of Literature, 1, 403-429.


Dickens, Charles. Pickwick Papers. London: Everyman, 1972.
Great Expectations. London: Everyman, 1971.
David Copperfield. Harmondsworth: Penguin, 1971.
The Mystery of Edwin Drood. Harmondsworth: Penguin, 1974.
Drabble, Margaret.
The Millstone. Harmondsworth: Penguin, 1971.
Metalious, Grace.
The Tight White Collar. London: Pan, 1966.
Orwell, George.
Burmese Days. Harmondsworth: Penguin, 1973.
Poe, Edgar Allan.
Tales of Mystery and Imagination. London: Everyman, 1971.
Salinger, J.D.
The Catcher in the Rye. Harmondsworth: Penguin, 1979.
Selby, Hubert.
Last Exit to Brooklyn. London: Calder & Boyars (Corgi), 1974.
Sillitoe, Alan.
"The Fishing-Boat Picture", The Loneliness of the Long-Distance Runner. London: Pan, 1969.
Thackeray, William M.
Barry Lyndon. Harmondsworth: Penguin, 1975.

[21 Aug 2000]