[Note: I want to thank Jon Erickson for helping me to prepare this article. The following abbreviations will be used: FID = free indirect discourse, POV = Point of view, RST = represented speech and thought, RST3 = third person RST, RST1 = first person RST.]
1. The literary and linguistic reception of Banfield (1982)
When Ann Banfield's Unspeakable Sentences was published in 1982 it was unanimously felt to be an important contribution to both linguistics and narratology, and one reviewer (Epstein 1982: 1280) even called it "required reading for literary critics (and also for linguists)". But few attempts were subsequently made to integrate Banfield's findings into existing narratological schemes. Most literary theorists quickly put the book aside because Banfield's formidable formalisms and her less than lucid style seemed overwhelming and often incomprehensible. Those who fought their way through her book soon found that the seemingly waterproof argumentation not only proceeded from a set of highly questionable assumptions as to "what counts as evidence" (McHale 1983: 18), but also that its main axioms and conclusions were assailable on the basis of numerous counterexamples. Brian McHale's by then longstanding quarrel with Banfield came to a head in his influential review article of 1983, [End of p. 348] in which he flatly denied that Banfield's theory made any sensible contribution to the study of narrative at all: "Banfield is, I believe, finally wrong - wrong not in this or that detail of her treatment of narrative sentences, but wrong in principle, wrong in her orientation" (1983: 17).
The linguistic reception of Banfield was generally less critical because her book seemed to show, at least initially, a new and promising way to approach the fertile corpus of narrative texts. All commentators acknowledged Banfield's painstaking research; but most also increasingly found her self-imposed restriction to the sentence level and her denial of a narratorial voice too narrow. Thus Violi (1986), though primarily a favorable evaluation, asserted the need to establish and proceed from a textual framework. Yamaguchi (1989) argued for the reinstatement of the narrator and suggested bringing back the communication setting, allowing an interpretation of FID as a "mentioned" form. A third commentator, Toolan (1988), applauded the formal rigidity of Banfield's theses, but went on to reaffirm McHale's counterexamples and ultimately condemn Banfield's enterprise as an instructive failure - "a reminder of the perils of decontextualized formalism", as he has more recently said (1990: 75).
On the whole, the linguistic reception of Banfield did leave a glimmer of hope that some aspects of her theory, or at least some of her findings, could be salvaged within a less rigidly "grammatical" framework. However, it is only recently, eight years after Unspeakable Sentences, that a first full-scale attempt at a discourse-analytical extension of Banfield's theory has been proposed. Taking Virginia Woolf's Mrs. Dalloway and To the Lighthouse as her "data base", Ehrlich (1990) analyzes the linguistic markers and context-sensitive features that lead readers to their intuitive and presumably invariant POV assignments. Because Banfield's "represented speech and thought" (RST) is seen to be intimately linked with POV, I will begin by briefly outlining an approach to RST that takes into account many of Banfield's less dogmatic findings and partly also integrates the critical comments and the many counterexamples that have appeared in the wake of Unspeakable Sentences. In section 3 I will then take a close look at Ehrlich's new proposals for contextualizing RST.
2. Represented Speech and Thought reconsidered
Banfield's term "represented speech and thought" has been accepted by some theorists (Violi 1986, Yamaguchi 1989, and Ehrlich 1990), while others prefer to stick to the more traditional FID. This difference in terminology - for convenience, I will distinguish between "FID users" and "RST users" - actually points up a major divergence in the analysis of the standard non-direct forms listed in (1):[End of p. 349]
a. He should have stayed at home and read his book. [FID, RST]
b. He should have stayed at home and read his book, thought Peter Walsh. [RST + parenthetical] (Mrs. Dalloway, 185)
c. Peter Walsh thought that he should have stayed at home and read his book. [indirect thought]
Whereas all will acknowledge (1a) as a specimen of FID/RST, most FID users reject (1b) because it is accompanied by the parenthetical phrase thought Peter Walsh. Parentheticals are a subset of inquit or consciousness phrases that sometimes accompany representations of speech/thought acts in an interpolated or final position. However, "free", according to most FID users, means free not only from introductory, preposed, inquits, but also from parentheticals. If a parenthetical is added to a FID sentence, as is the case in (1b), thus allegedly binding it, making it non-free, all its significant features (such as its relative closeness to a hypothetical original speech or thought act) do in fact remain intact. Since parenthetical clauses do not dominate, (1b), and in particular its represented thought section, is still independent (in this sense still "free") - unlike (1c), which is a proper, subordinating, indirect thought sentence. In view of (1b) and (1c) it is thus slightly inaccurate to think that FID, shorn of its F, turns into ID (indirect discourse). RST and parentheticals, on the other hand, seem to co-exist well in an almost symbiotic relationship. In fact, Ehrlich (1990) shows that parentheticals are important and often necessary RST markers. In multi-perspective texts, such as Woolf's, the reader would plainly be lost without them, an insight often disregarded by FID users. The question of the exact nature of the dovetailing between parentheticals and RST - whether parentheticals are plain narratorial exposition, for instance, as most commentators assume - will briefly be taken up below, in the discussion of Ehrlich's handling of parentheticals.
If both "free" and "indirect" in FID point in the wrong direction, so does "discourse", because what is represented in FID very often, (1a) is an example, is a thought process, not a discourse oriented utterance. Thus it appears that RST is indeed, on all three counts, so to speak, an improvement over its more customary competitor. One must not overlook, however, that it, too, contains one potentially dangerous element, namely the word and. Even if it is true that speech and thought have very similar textual representations, they do remain mutually exclusive categories, both ontologically and in their fictional existence. Nothing, strictly speaking, is ever a representation of speech and thought. Thought, as opposed to speech, is non-discursive, private, non-communicative, non-pragmatic and semi-verbal, to list just a few differential properties. The abbreviation "RST" (which Banfield 1982 avoids) hides this difference and may even be the cause of a category error. "RST" ought therefore to be used with caution: it is an acceptable generalization emphasizing the syntactic similarities between two representational techniques; [End of p. 350]however, in a discussion of specific cases the terms represented speech or represented thought, dependent on the circumstances, are certainly preferable.
In addition to the types exemplified in (1), Banfield also presents a borderline case with an introductory inquit phrase and an intonation break, (2a), noting "a superficial resemblance [. . .] between sentences of represented speech and thought and those of indirect speech" (1982: 71). Other cases such as (2b) and (2c) she classifies as ungrammatical:
a. He said: oh was he tired.
b. *He said oh was he tired.
c. *He said that oh was he tired.
(2b) and (2c) Banfield cites as *He said (that) oh was he tired (1982: 71), implying that there is no significant difference in ungrammaticalness between the two options. However, McHale (1983) and Toolan (1988) argue that according to their, and probably most other native speakers' intuition, (2b) is not seriously ungrammatical and at any rate certainly possible in narrative contexts. On this view many of Banfield's arguments from grammaticalness or ungrammaticalness become suspect. Here is a selection of problematical cases as quoted from Banfield by McHale (1983: 24-25):
a. *The consul asked himself why then should he be sitting in the bathroom. (Banfield 1982: 29)
b. *He [. . .] thought to himself he was damn lucky to get away from [. . .] that sonofabitchin' foreman. (Banfield 1982: 114)
c. *No, sir, he could not obey his order, he told the officer. (Banfield 1982: 114)
These sentences are all assumed to be ungrammatical by Banfield because the indirect speech samples (3a)-(3b) contain direct speech constructions such as inverted questions or indications of pronunciation and the RST sample (3c) makes use of an addressee-oriented deictic (sir). McHale, an expert on Dos Passos, points out that these sentence types are in fact Dos Passos's "staple sentence[s]" (1983: 31). He also shrewdly argues that none of Dos Passos's critics and reviewers "ever mentioned his ungrammaticalness" (1983: 31). Also discussing Banfield's examples, Toolan (1988) helpfully suggests that sentences like (3a) and (3b) might be termed "creative indirect speech" (1988: 143).
The most salient differential feature of RST is its shiftedness, i.e., a (usually) backshifted tense and shifted pronouns as in (4). It is also useful to distinguish whether RST occurs in a third person or a first person context, for which the abbreviations RST3 and RST1, respectively, might be suggested.[End of p. 351]
a. She was not interested in his memoirs.
b. He had once been a farmer, if I could believe it.
Assuming that (4a) is RST3 and (4b) RST1, the shiftedness of RST can be illustrated by reconstructing an original or "preshifted" utterance or thought, a procedure which Cohn (1978: 100) calls the "litmus test" of RST. As a simple notational convention I shall suggest the following in (5), which explicate the RST cases in (4):
a. ~I am not interested in your memoirs.
b. ~I was once a farmer, if you can believe it.
In this notation, the "approximately equal" symbol indicates the explicator's approximate reconstruction of a preshifted "direct" utterance or thought. It must be stressed at once that this notation is a purely explicatory device; in particular, it can neither be claimed that the assumed original is exactly determinable nor, in fact, that such an original even exists. This complication is illustrated in (6) and (7):
a. He was a writer of her caliber, Smith said. (Banfield 1982: 26)
b. ~He is a writer of her caliber.
c. ~You are a writer of Dorothy's caliber.
d. ~I am a writer of your caliber.
[a] Probably his good works would take the form of building pagodas. [b] Four pagodas, five, six, seven - the priests would tell him how many - with carved stonework, gilt umbrellas and little bells that tinkled in the wind, every tinkle a prayer. [. . .]
[c] All these thoughts flowed through U Po Kyin's mind swiftly and for the most part in pictures. (Orwell, Burmese Days, 7)
(6b)-(6d) and many others could serve as possible preshifted originals for (6a). As Banfield has shown, this fact makes it impossible to assume any kind of derivational relation between the preshifted "bases" and their RST realizations. In (7) the reader is explicitly advised that "all these thoughts" go through the reflector's head "for the most part in pictures". Note that we can still test and, up to a point, explain the RST character of (7a) and (7b) by construing preshifted fragments such as ~Probably my good works will take . . . the priests will tell me how many . . .; yet at the same time we must also take into account the text's reference, in (7c), to the semi-verbal quality of U Po Kyin's thoughts. Incidentally, the instructional character of (7c) clearly indicates that there is a narrator at work here, a circumstance that was emphatically denied by Banfield for RST3 and for third person texts in general.[End of p. 352]
The shiftedness assumption of RST brings out an interesting borderline case. As is well known, RST can easily accommodate incomplete sentences. But what happens if, in an elliptical RST, there is no pronoun and no tense? Consider (8) and (9):
[a] Where was that boat now? [b] Mr. Ramsay? [c] She wanted him. (Woolf, To the Lighthouse, 230; Ehrlich 1990: 79)
[a] Where was that boat now? [b] Mr. Ramsay? [c] I want him.
Are (8b) and (9b) represented thought or direct thought? This question can obviously only be answered by observing actual reading strategies. It seems natural to say that since (8b) is framed by represented thought, (8b) is read as represented thought as well. But note that this strategy does not help in (9). Here we have the sequence represented thought - X - direct thought and the question is, does X (9b) inherit represented thought status from (9a) or direct thought status from (9c)? Possibly authors can use an ambiguity like this for shading from repesented thought to direct thought and vice versa. On the other hand, disambiguation would even in this case be possible if in a given context either represented or direct thought is firmly entrenched as the unmarked (or default) representational technique. There may be a difference, for instance, in the markedness status of RST in Mrs. Dalloway, where it is presumably the unmarked form, as opposed to that in chapter 2 of Ulysses, where, it may be argued, direct thought is the unmarked mode.
Banfield made an important discovery when she drew attention to the fact that RST was handicapped in certain respects. Most literary theorists work on the assumption that RST is really more a style than just a technique of rendering speech or thought events; indeed Stanzel even suggests that erlebte Rede may inform whole texts and be more or less equivalent to what he calls figural (impersonal) narrative situation (1982: 254; also Cohn 1981: 172). However, Banfield (1982: 113) has shown that RST is virtually unable to render imperatives and (despite McHale 1983: 24) it is generally true that it is all but helpless vis-a-vis many "addressee-oriented" elements. Thus sentences like those in (10a) and (10b) have no natural RST realizations:
a. Let me alone! screamed Anthony silently. Let go of me!
(Metalious, The Tight White Collar, 134)
b. Would it bore you to come with me, Mr. Tansley? (To the Lighthouse, 12)
Banfield uses this fact - that RST cannot properly address a you - as evidence [End of p. 353] for her argument that RST is not a phenomenon that one would expect in an ordinary communication setting, and this, for her, is crucial evidence for the non-communicative character of narrative. A less controversial implication of the evidence of (10) is that RST really depends on being supplemented by the other representational techniques. Paradoxically, RST can also be shown to be more versatile in certain respects than Banfield expected:
He will write to her? [~You will write to me?]
He will write to her every alternate day, and tell her all his adventures. [~I will write to you . . .] (Dickens, Edwin Drood, 174)
Have I heard, she wants to know, from poor Blanche?
[~Have you heard from poor Blanche?] (Delafield, Diary of a Provincial Lady, 431)
Although Banfield claims that in RST "no non-generic present tense appears" (1982: 100), both examples, (11) is RST3 and (12) is RST1, comfortably use the present tense as a narrative tense - not just as an occasional generic, or even historical, present. (Note, incidentally, that the existence of present tense RST is a good reason for calling RST a "shifted" rather than a "backshifted" form.)
Although it is certainly true that RST is primarily a literary phenomenon, Banfield may have overemphasized its literariness. Yamaguchi (1989: 586) points out that there are known cases of RST in oral story-telling, and McHale quotes several sources discussing non-literary RST, but adds that RST is always felt to introduce a "fictional" element (1978: 282f.). The following example, (13), contains a case of represented speech in what is perhaps a more obvious non-literary context, namely the minutes of a military staff conference:
The feature just noted, that RST retains almost all of the expressivity of [End of p. 354] direct speech, leads Banfield to argue that the subjective expressions (evaluative adjectives, contrastive stress, kinship terms, exclamations, questions etc.) are pointers to a source consciousness, a "Self". This Self Banfield equates with the concept of "point of view" as used in literary discussions of narrative texts (1982: 68). One of Banfield's main principles is that within the scope of a single sentence - which she calls EXPRESSION - at most one Self can be expressed. It is this principle ("1 E[XPRESSION]/1 SELF") that runs counter to the so-called "dual voice" approach to RST/FID which is practised by many literary theorists. For "dual voicers" (as Toolan 1988: 136 calls them), FID is the prototypical case where both the voice of a character and that of the narrator may be heard. Opinions on this matter vary considerably, however, and inconsistencies are not rare. Thus Toolan at one point states that he sympathizes "with Banfield's underlying assumption that an utterance cannot emerge from two distinct subjectivities at once" (1990: 78); yet he also claims that RST exhibits "many marks of the narratorial voice and presence" (1988: 128). Ehrlich (1990), on the other hand, strictly follows Banfield in assuming that an RST sentence can only have one Self or POV. This she does even for a sentence such as (14), which she classifies as "semantically anomalous":
Yes, she disliked that sweet man immensely, Lily said. (Ehrlich 1990: 69)
In spite of the "contradictory information" contained in (14), Ehrlich (1990: 69-70) assigns both disliked and sweet to one Self, the character identified in the parenthetical. However, since RST in many respects remains close to its preshifted original it probably also retains the original's intonational quality; and there appear to be sharply divergent intonations for ~I dislike that sweet man immensely. In particular, in addition to representing an inconsistent, perhaps pathological, utterance, (14) can also be read (and intonated) as containing an ironical judgment. In this latter reading, in which sweet is the ironic echo of another character's subjective expression, "1 E[XPRESSION]/1 SELF" is violated, a fact blandly dismissed by Banfield by stating, "With irony, we have passed beyond the jurisdiction of grammar" (1982: 221). The same problem lies at the heart of the clash between Banfield's claim in (15) and McHale's invented counterexample (16):
In "Yes, she could hear his poor child crying now," the yes cannot be the expression of "her" point of view and poor of his. (Banfield 1982: 94)
She was about fed up with both of them, father and daughter. Above all, she was sick and tired of hearing him moan about his poor child. His poor child this, his poor child that: enough already! Yes, she could hear his poor child crying now. (McHale 1983: 35-36)
[End of p. 355]
Contrary to Banfield's reasoning in (15), (16) creates a context that favors a dual voice reading based on subjective expressions anchored on two Selves. But, as may be expected from an ironical utterance, McHale's female character now only metalinguistically mentions the key phrase poor child. Since such ironic echoing is very often taken to be the prime example of dual voicedness, McHale's counterexample, though valid, actually highlights the fact that Banfield's principle remains quite strong on ordinary, non-mentioned linguistic evidence. Yamaguchi (1989: 587), on the other hand, draws a different conclusion and proposes that all RST be considered a mentioned form. This assumption may indeed help to explain the narratorial irony that is often taken to be associated with RST; however, it clearly serves no useful purpose for the apparently much more frequent case of ordinary, non-ironic RST.
Sperber and Wilson, who were the first to analyze irony on the basis of the "use-mention distinction", explicitly allow RST as a mentioning technique (1981: 305). Since they do not quote an ironic RST example, I will provide the following:
"When do you leave?" she asked.
She said nothing more. Strangely enough, a tinge of melancholy had settled over her spirits. No doubt the proximity of the town was the cause of this. (Orczy, The Scarlet Pimpernel, 358)
~No doubt, the lady here thinks, the proximity of the town is the cause of my melancholy; but reader and narrator know that this is just her counterfactual rationalizing and that the real cause of her state of mind is that her husband is leaving her. Note that RST is only incidental here, as this type of irony (dramatic irony) works in exactly the same fashion with any other mentioning technique (e.g., direct speech or direct thought). In addition, the irony in this case appears to establish itself without having to rely on any dual-voiced elements.
Since the real RST/FID attraction for dual voicers lies in the mingling of a narrator's and a character's voice, one had best look at passages where both an intrusive narrator and a reflector may be encountered, as is sometimes the case in Thackeray or Dickens. Consider the following, slightly modified example from Edwin Drood:
[a] He will soon , poor youth that he is, be far away, and may never see them again, he thinks. [b] Poor youth! Poor youth! (Dickens, Edwin Drood, 177; italicized phrase added)
Here the expressive poor youth is tied in one place (18a) to the character and [End of p. 356] in another (18b) to the narrator. But, obviously, the self-pitying poor within represented thought carries none of the awareness of tragic irony expressed in the narrator's emotive exclamations. Again, despite extremely favorable conditions, no dual voice reading is obvious or necessary.
Banfield is aware of the fact that, apart from irony, "1 E[XPRESSION]/1 SELF" also runs into difficulties with two further special cases. One is the echo question, which is discussed at length in Banfield (1982: 123) and commented on by Yamaguchi (1989: 587). The other is "heard speech", which, Banfield thinks, is exclusively a feature of first person narration:
[a] Did I see it? [b] I saw it. [c] What more did I want? [d] What I wanted was rivets, by heaven! (Banfield 1982: 122)
This is dialogic RST1 in which the interlocutor's represented speech - (19a) and (19c) - contains subjective expressions (questions) aligned with his Self. Fettered by her principle, Banfield is forced to assume that the sequence of turns in (19) represents alternating POVs, despite the fact that the narrator's I - normally a salient POV indicator - is explicitly present in each sentence. This is rightly challenged by Yamaguchi (1989: 587) as an ad hoc solution. For heard speech, clearly some sort of POV has to be assigned to the perceiving subject. Perception is a Self-oriented activity; and what is heard is often different from what has been said. Indeed, a number of cases which Banfield did not fully foresee can be adduced as further evidence of this division and propagation of separate POVs:
He worked hard - seven hours a day; his subject was now the influence of something upon somebody - they were walking on and Mrs. Ramsay did not quite catch the meaning, only the words, here and there . . . dissertation . . . fellowship . . . readership . . . lectureship. (Woolf, To The Lighthouse, 13)
but one question still remained, which the faces of the Jews pretty significantly suggested, - was I that person? [~Is he that person?] (De Quincey, Confessions of an English Opium-Eater, 25)
a. Oh how extraordinarily nice I was! (*she thought). (Yamaguchi 1989: 583; ungrammaticalness assigned by Yamaguchi)
b. Oh how extraordinarily nice I[i] was! (*she[j] thought). (Banfield 1982: 94; ungrammaticalness assigned by Banfield)
(20) represents part of Mr. Tansley's conversation with Mrs. Ramsay as heard speech in a third person context. Mrs. Ramsay's consciousness leaves its mark of subjectivity on the heard represented speech, and we get through to [End of p. 357] Mr. Tansley's POV only via a representation of Mrs. Ramsay's flawed perception of it. (21) appears to be quite close to the first person heard speech case in (19), except that this is not represented speech, but represented thought; and here the speech-thought disparity happens to be crucial. Of course, the experiencing I cannot read other minds; what he does is guess at the Jews' thoughts on the evidence of their facial expressions. The subjective expressions in this represented thought - the question construction and the italicized emphasis - all denote the Jews' POV. But the context makes it quite clear that this is a speculation of the I, who does not passively perceive something that has been said within earshot, but actively constructs what he thinks goes on in his interlocutors' minds. (22a) is a misquotation of one of Banfield's sentences, here reproduced as (22b). Yamaguchi forgets to indicate the reference conditions that Banfield meant to be the cause of the sentence's assumed ungrammaticalness. (22a), as it stands, is a perfectly grammatical specimen of direct thought. Oddly enough, Banfield's original (22b), in which the I refers to the speaker/narrator and the she to a character, a combination thought to be inconceivable, is also not ungrammatical because it can be read as a speculative reconstruction, by the first person narrator, of the thought content of another mind, i.e., exactly along the lines of (21).
Banfield's rule that a sentence can materially express only one POV is largely responsible for her practice of viewing narrative sentences in isolation. If two neighboring sentences show subjective elements anchored on different Selves, then, Banfield argues, they simply produce a sequence of different POVs; in fact, on her view, a text may be a sequence of alternating POV representations. What her grammar of expressive sentences precludes is that POV might be arranged hierarchically, i.e., that one POV might be presented from another POV (even if perhaps it is impossible for both POVs to be explicitly present in the subjective expressions of a single sentence). A case in point is (23), which looks deceptively simple:
He wanted to tell Arthur Winner that he knew; he knew; he hadn't meant to! All he did was make everybody trouble - (Cozzens, By Love Possessed, 71)
Apparently (23) begins as indirect thought and then "ungrammatically" shades into represented thought, exhibiting expressive constructions and an ellipsis indicative of the POV of the third person, he. A contextualized reading of this passage shows that this is in fact far off the mark. In By Love Possessed there is only one centre of consciousness, namely the small-town lawyer here identified as Arthur Winner, and thus what (23) actually represents is Arthur Winner's speculation about a business partner's telephone call that did not reach him. His partner, he has reason to believe, probably wanted to say, apologetically, ~I know; I know; I hadn't meant to . . . etc. In other words, [End of p. 358] (23) shows a hypothetical POV reconstructed and represented from Arthur Winner's controlling POV. Thus a sentence may be RST indicating the POV of a character; or perhaps it may be remembered (speculation of (alien thought)) or some such nested POV construct (for similar nested configurations see Cohn 1978: 133). The expressive elements, if any, apparently attach to the innermost POV; indeed, the innermost level is the one that will be represented in the first place. (23) demonstrates how little point there is in analyzing such sentences in isolation. Once context is considered, POV assignment is more than the answer to the question Whose subjective expressions? - it is a consequence of textual interpretation and reading strategy.
3. Ehrlich's discourse-analytical approach to RST and POV
Although crediting Banfield's "profound" influence (vii), Ehrlich prudently distances herself from Banfield's total denial of any speaker presence whatsoever in third person texts and particularly RST3. Strangely enough, Ehrlich does not acknowledge any of the theorists involved in the debate over Banfield's book. Possibly Toolan (1988) and Yamaguchi (1989) came too late to be considered; but that McHale's (1978) and (1983) contributions should be missing is a serious oversight. At any rate, just as Banfield's critics unanimously proposed, Ehrlich's approach now explicitly recognizes the "limitations of a sentence based approach" (16), and she also, without further ado, reinstates the narrator for third person texts.
For the Woolfian narrator, Ehrlich assumes the unobtrusive, withdrawn, minimal, objective kind of the species, who has an occasionally audible voice and a POV of his/her/its own, but may not use the first person (10). This latter (generally too rigid) restriction was imposed on "impersonal" texts by Tamir (1976); the occurrence of the first person, plural, arguably including a self-reference of the narrator, in To The Lighthouse II.3, goes unnoticed by Ehrlich. Apart from Tamir, Ehrlich's literary authority on POV is Genette (1980); the more recent discussion on POV, focalization etc. (e.g. Chatman 1986) has not been incorporated. According to Ehrlich, in an RST passage "the speaker (narrator) totally identifies with a character of the narrated events" (118, n. 5), whereas narration proper consists of "sentences expressing the POV of the narrator (i.e. where the narrator assumes a position equi-distant from all characters)" (16). For an RST passage this seems to pave the way for a dual voice interpretation in the sense of Pascal (1977) or Cohn (1978). But the question of whether a passage is narration or RST, i.e., exhibits the POV of the narrator OR that of a character (this disjunction creeps in on p. 9 and takes further shape on p. 20 of Ehrlich's book), is taken to be either obvious or the result of disambiguation (24). And the ambiguity in [End of p. 359] question is taken to be a razorblade, either/or (but-NOT-both) ambiguity, not a tolerant, malleable, dual-voice oriented, literary, and/or ambiguity.
Due to its lack of reliance on the critical discussion of Banfield's work, Ehrlich's exposition of RST vs. indirect speech/thought vs. direct speech/thought is heavily dependent on many of Banfield's questionable ungrammatical paradigms. As could be expected, some of the types of "ungrammatical" sentences quoted by Ehrlich actually do occur in Woolf:
a. *She asked what had he wanted to tell her? [p. 8; ungrammaticalness assigned by Ehrlich]
b. *She thought that he was an awful prig - oh yes, an insufferable bore. [p. 9; ungrammaticalness assigned by Ehrlich]
a. [she] said did he mind her just finishing what she was doing to her dress? (Mrs. Dalloway, 46; cp. Banfield 1982: 281, n.7, also McHale 1983: 27)
b. Maisie Johnson positively felt she must cry Oh! (Mrs. Dalloway, 31)
Cognizance of McHale's counterexamples would have prevented this embarrassing oversight, and Toolan's "creative indirect speech" could have been put to excellent use, because, unlike in Banfield, nothing of importance hinges on the particular distinction between RST and indirect speech in Ehrlich's approach. In fact, it is Ehrlich's concern to widen the scope of RST as far as possible. And she is quite right in pointing out that "a large set of sentences" (19) which do not formally display the typical features of RST (direct speech syntax, exclamations, subjective expressions etc.) should sometimes, given the right context, be interpreted as RST. Thus
He took a day off with his wife and played golf.
may not at first glance look like RST, yet in its actual context happens to be "a back-shifted version" (24) of a speech act of Dr. Holmes in Mrs. Dalloway. We know this from contextual features only; otherwise the sentence could also "plausibly be interpreted as conveying the narrator's viewpoint" (20). It is crucial that Ehrlich here adduces the shifted nature of RST. Quite a number of sentences, even those that formally look like narration or other types of speech or thought representation correctly become amenable to an RST reading under these conditions.
From here, unfortunately, Ehrlich goes on to a far less principled, and therefore less plausible, attempt to extend the scope of RST much further. Thus the following examples, borrowed from Banfield (1982: 157), are also considered to be possible RST: [End of p. 360]
a. A few drops of rain were falling. (21)
b. She saw the moon. (22)
Examples (27a) and (27b), Ehrlich claims, are potential representations of perception acts - which is true - and are for that reason potential RST (21 f.) - which does not at all follow. Unless a character can be imagined to think or say ~A few drops of rain are falling, or ~I see the moon - somewhat unlikely scenarios - Cohn's litmus test fails; hence these examples are not RST. If represented perception were part of RST, RST should be called something else.
To make matters worse, RST is further extended by Ehrlich to cover even direct thought and indirect thought (cp. examples 2 and 6 on pp. 7 and 84, respectively). This radical extension appears to be necessary in order to establish a general "narration/RST distinction" (26, 52) in fiction. This distinction is attributed to Banfield, but Banfield actually differentiated "represented perception" and RST, then subsumed both under "represented consciousness", and only then arrived at the far more acceptable global opposition "narration/represented consciousness". One of the immediate effects of Ehrlich's now seriously overextended RST is that, cuckoo-like, it has crowded out all its siblings. Of course, direct speech, direct thought, indirect speech and indirect thought all appear even in Ehrlich's sample passages, RST-biased as they are, yet she hardly ever has cause to mention them by name. But, it may be asked, is it not exactly Woolf's conscious modulation, her orchestration of all of the techniques, that produces the remarkable depth-effects and rhythmic quality of her novels? Extended RST not only provides no answer; the question itself is pointless if almost everything is RST. This surely is an extremely inauspicious starting point for what, at least in part, professes to be an Analysis of Literary Style.
In its radically extended scope, RST serves to attribute episodes to a character's consciousness ("Andrew's RST", "the RST of Mr. Tansley" and so on). For Ehrlich, the next logical step is to read such RST episodes as expressing the POV of their source consciousness (75). Phrases like "a sentence of RST, reflecting James's point of view" (72) or "particular points of view (i.e. RST units)" (40) suggest that she, like Banfield, assumes that there is an almost total conceptual overlap between RST and character POV. Analyzing the coherence conditions holding in an episode, Ehrlich also discusses transitions between characters' RSTs (POV switching), and between RST and passages that are attributable to the narrator's POV (78). This discussion, under such headings as referential linking, semantic connectors, alignment to the narrative time axis, time deictics, the function of paragraph indentations, etc., turns up a number of useful criteria that generally correctly predict readers' intuitive POV assignments. Ehrlich's postulated reading [End of p. 361] strategy for marking off episodes of RST and episodes of narratorial exposition (cp. p. 40 ff.) appears to proceed roughly along the lines of (28):
a. If a sentence contains character oriented subjective expressions or is accompanied by a parenthetical indicating a reflector, then consider this sentence to be the beginning of an RST episode anchored on the reflector-character's POV.
b. If the following sentence is referentially, semantically and/or temporally linked to the preceding sentence, or if it re-satisfies (a), then continue attributing this character's POV to it.
c. If substrategies (a) and (b) fail, assume a narratorial POV equidistant from that of all the characters currently present.
As an illustration of this strategy let us assume that on a rough, "first pass" reading of a paragraph it is found to contain sentences of narration (n), sentences unambiguously recognizable as character RST (c) and sentences that are ambiguous between n and c, symbolized as "n-over-c". A hypothetical sequence of sentences may look like in (29):
n n n n n c n c c c
According to (28a), the fourth element, the single c, would be identified as the first unambiguous RST sentence. The following two n-over-c's would be disambiguated on the strength of (28b); and for the last sentence (28a) and (28b) may be assumed to fail. Thus a second pass reading on the basis of (28) would read (29) as nnncccn. Note that, on the basis of (28), the first ambiguous n-over-c sentence in (29) must be disambiguated as narratorial POV. A passage from To the Lighthouse, the beginning of III.13, may serve as a full example:
Mr Ramsay had almost done reading. One hand hovered over the page as if to be in readiness to turn it the very instant he had finished it. He sat there bareheaded with the wind blowing his hair about, extraordinarily exposed to everything. He looked very old. He looked, James thought, getting his head now against the Lighthouse, now against the waste of waters running away into the open, like some old stone lying on the sand;
Ehrlich comments that the text here "begins as narration" and "later on becomes James's RST" (122 n. 6). She then, somewhat off-handedly, concedes that readers may also "retroactively identify the first half of this paragraph as James's RST". Indeed this appears to be a necessary step, and therefore (28a) should be modified to read as follows: [End of p. 362]
If a sentence contains character oriented subjective expressions or is accompanied by a parenthetical indicating a reflector, then backtrack and test whether the preceding sentences can also be read as representations of this character's POV. If so, move the beginning of this RST episode back by a suitable number of sentences.
Under this revised strategy, the sequence of the abstract example in (29) would now be read as nnccccn (instead of nnncccn). It is interesting to note that Ehrlich, albeit half-heartedly, here comes very close to McHale's so-called integrational model of the reading process, "whereby sentences give rise to interpretative reconstructions which in turn affect the interpretation of subsequent and even, retrospectively, of preceding sentences" (1983: 39).
A reading strategy like the one detailed in (28) and (31a) can be a powerful discourse-analytical tool. However, as noted above, many of Ehrlich's assumptions are either too general or inherently implausible; and therefore it is not surprising that there are passages where her model does not operate smoothly, and sometimes conspicuously fails to work at all. To begin with, POV attribution in passages containing representations of dialogue is hampered by Ehrlich's assumption that RST and POV are equivalent. Consider the following passage from Mrs. Dalloway where Rezia is in consultation with Sir William Bradshaw:
Sir William explained to her the state of the case. He [Septimus] had threatened to kill himself. There was no alternative. It was a question of law. He would lie in bed in a beautiful house in the country. The nurses were admirable. Sir William would visit him once a week. (Mrs Dalloway, 107; Ehrlich, 75)
The first sentence is followed by a series of RST sentences that "derive their interpretation as RST from their semantic dependence on the predicate of the [...] preceding sentence Sir William explained to her the state of the case" (75). Since RST is closely linked to character POV, Bradshaw's RST is an expression of his POV. We may perhaps add that even the first sentence is already infected with doctor's jargon (the state of the case) and approximates Bradshaw's professional view of a patient, as opposed to Rezia's view of her husband. (The narrator, at this point, is neither "equidistant from all characters", nor does he/she "totally identify" with Bradshaw; and the reader is, of course, well advised to keep his/her critical distance as well.) Within the RST section, as must be expected, all value judgments (e.g., beautiful house, admirable nurses) clearly originate with Bradshaw, although the reader, and apparently Rezia, too, recognize them as insincere feints.
While this augmented reading is still reasonably compatible with Ehrlich's assumptions, it only works up to a point. As soon as we reach the last [End of p. 363] sentence in (32), presumably a shifted version of ~I will visit him once a week, the non-pronominal speaker identification Sir William strongly suggests that this is now not so much what Bradshaw says, but what Rezia hears. If this is true we have an interesting conflict between one character's RST and another character's perception (= also RST, in Ehrlich's model). Thus, unnoticed by Ehrlich, the passage apparently displays one of the shifts in POV typical of Woolf (many similar shifts occur in the same scene). The last sentence is still RST; and Bradshaw is still its subjective-expressive Self; however, a new and additional POV now originates with Rezia, and a change in orientation has been accomplished fluidly, without any of the more noticeable formal breaks such as an explicit parenthetical or a new paragraph or any of the other violations of coherence conditions detailed by Ehrlich. Ehrlich's failure to recognize this POV switch is due to her inability to place the phenomenon "heard speech" within her concept of extended RST. As argued above, heard speech allows (and sometimes manifests) two POVs: that of its speaker and that of its hearer. The fact that one character's speech may be another character's perception also reaffirms what should have been obvious from the outset: that the alleged logical link between RST and POV is, within a heuristic framework, purely accidental.
Another problematic area is the assumed role of parentheticals or, as Ehrlich calls them, "sentences containing a parenthetical" (SCPs). Suppose a passage is liberally saturated with parentheticals, as is often the case in Woolf, facilitating or maintaining an RST reading. Since Ehrlich ultimately assigns parenthetical phrases to the narrator's POV (this comes rather late in the book: p. 110), it may be asked whether such a passage is homogeneous RST or a heterogeneous mixture of character's and narrator's POVs, or possibly even a case of switching between POVs. Although not directly addressing the question, Ehrlich always calls such passages plain RST, irrespective of whether they do or do not contain parentheticals. In point of fact, it remains an unanswered question whether parentheticals are always reader-conscious expository RST markers and thus bound to a narratorial POV. This is quite apparent in the following passages (not quoted in Ehrlich):
'I should like to see Mr. Dubonnet,' said Hugh in his curt worldly way. (Mrs. Dalloway, 126)
'Armenians,' he said: or perhaps it was 'Albanians.' (Mrs. Dalloway, 132)
'Oh, Lucy,' she said, 'the silver does looks nice![']
'And how,' she said, turning the crystal dolphin to stand straight, 'how did you enjoy the play last night?' 'Oh, they had to go before the end!' she said. 'They had to be back at ten!' she said. 'So they don't know what happened,' she said. (Mrs. Dalloway, 43)
[End of p. 364]
In (33) the phrase in his curt worldly way is not an objective report, rather, it reflects Richard Dalloway's slightly critical view of Hugh Whitbread. In (34) the parenthetical clearly does not assert that what is represented as direct speech was actually said. In (35) the plethora of parenthetical she said phrases (more follow in the immediate context) plus their needlessly ambiguous pronominal reference - all pronouns refer to Mrs. Dalloway, incidentally - obstructs rather than aids the reading process. The parentheticals here strikingly ignore, rather than satisfy, the reader's need to identify "the source of the speech or thought" (11). Unless this passage is a stylistic scandal (surely a last resort judgment) these parentheticals must have another function. Obviously neither the functional range nor the possible POV assignments of narrative parentheticals has yet been satisfactorily determined.
Ehrlich is only rarely concerned with purely narrative passages, although she does discuss the beginning of To The Lighthouse (98) and the sequence describing the route of the "Prime Minister's car" in Mrs. Dalloway (77) as typical instances. But, in general, it is the RST passages that get major coverage, and episodes of narration usually only come up when the material cannot be attributed to one of the characters. Here is one of several examples:
[...] so he turned deliberately in his chair and looked out of the window and said, all in a jerk, very rudely, it would be too rough for her tomorrow. She would probably be sick. (To The Lighthouse, 100; Ehrlich, 69)
Here, according to Ehrlich, all in a jerk and very rudely stick out because they
Still, he wished he had known how to answer Miss Briscoe properly; he wished it had not come out all in a jerk like that. 'You'd be sick.' (To the Lighthouse, 101)
How much of a surprise is (37)? Are we now hurriedly readjusting our understanding of (36)? Surely a more adequate reading strategy should have allowed for this possibility from the very start. For a related example consider (38), which presents Lily Briscoe trying to evade Mr. Ramsay:
She fetched herself a chair. She pitched her easel with her precise, old-maidish movements on the edge of the lawn, not too close to Mr. Carmichael, but close enough for his protection. (To the Lighthouse, 167-8; Ehrlich, 72)
This time it is the expressions with her precise, old-maidish movements that
- all except myself, thought Lily, girding at herself bitterly, who am not a woman, but a peevish, ill-tempered, dried-up old maid presumably. (To the Lighthouse, 174)
So, just as Mr. Tansley is aware of being rude, Lily thinks she looks like a "dried-up old maid", which, of course, is not at all inconceivable, nor does it in any way mean that she is one. It is very likely, I think, that a competent reader (who, at this stage, must have quite a detailed mental picture of Lily) will confidently - or perhaps, I grant, provisionally - read the whole of (38) as an expression of Lily's POV. In fact, not reading it so is a serious error, not only because the character orientation is supported five pages later, in (39), but because it amounts to misconstruing Lily, denying her the ability of critical self-assessment. Indeed the subjective bitterness of this expression of her inferiority complex actively engages our sympathy in a way no "objective", [End of p. 366] factual, narratorial statement (about an unattractive spinster?) could. Again Ehrlich's seemingly rigorous and "descriptively adequate" (2) approach supports what is demonstrably a misreading. Wrong here, how often is it wrong elsewhere?
Credit is due to Ehrlich for incorporating RST, POV and a set of reading strategies within an integrated framework which carries considerable explanatory power. Unfortunately, many of her root assumptions are based on speculative and questionable generalizations. What is urgently needed is a definition of RST that is aware of the danger involved in the compounding of speech and thought, a better understanding of the coexistence of RST and parentheticals, and an account of the interplay between RST and other forms of speech, thought and sense data representation. As in most literary studies, POV turns out to be an interesting and useful concept, but only if it is held to be of a different order than RST. Finally, as the discussion of Ehrlich's POV assignments has shown, a lack of interpretative foundation can result in seriously inadequate readings. Of course, explaining ordinary or natural reading strategies is a legitimate, even necessary, aim of discourse analysis; but it will not get far with sophisticated literary texts unless it takes into account the intuitions of informed and competent readers.
A. Theoretical Texts
[20 Aug 2000]