This essay is a continuation of an earlier one (Jahn, 1996) in which I suggested a revision of Genette's theory of focalization by linking its original question "who sees?" to Henry James's vision of windows in the House of Fiction and, more metaphorically, of windows provided by story-internal "reflector" characters. The main strands of James's complex metaphor were taken up in the term "windows of focalization," a narratological concept grounded on a linguistic-cognitive model of the reading process (Jackendoff, 1987) and on Iser's (1976) and Wolf's (1993) reception-orientated accounts of narrative aesthetic illusions. The present essay continues this integrative bent, first, by situating traditional point-of-view-concepts within the narratological framework of Genettean and post-Genettean focalization theory, second, by reconceptualizing criterial aspects of focalization as "vectored indicators of subjectivity," third, by moving from Genette's (1980 ) and Stanzel's (1984 ) typologies of focalizations/perspectivizations to a four-type gliding-scale model, and fourth, by extending Ryan's (1987) concept of "story-line windows" to focalized contexts and discussing patterns of window shifting and window overlap. [End of p. 86]
1. Standard models of focalization
As is generally known, Genette divides narratology into the three fields of Tense, Mood, and Voice. Tense treats the possibilities of temporal arrangement and presentation (Order, Speed, and Frequency); Voice deals with narrators, embedded narratives and the choice of grammatical person; Mood analyzes "the regulation of narrative information" (1988, 41), subsuming (a) modes of presenting action, speech, and thought, and (b) modes of selection and restriction. The latter subfield Genette terms focalization -- or, rather, he promotes an already existing word (both in French and English) to a theoretical term.
For his definition of focalization, Genette draws on four traditional approaches: Brooks and Warren's (1959/1943) point of view approach, Pouillon's (1946) vision approach, Blin's (1954) field approach, and Todorov's (1966) knowledge approach. Brooks and Warren build on the question "Who sees the story?" (1959, 659); Pouillon (1946, 69-114) distinguishes three main vision modes, vision avec ("vision with," i.e., vision through a character's eyes), vision par derrière ("vision from behind," i.e., from an omniscient narratorial vantage) and vision du dehors ("vision from outside"); Blin (1954, part II) treats Stendhal's use of subjectively restricted fields [restrictions de champ]; and Todorov asks whether the narrator knows more than, as much as, or less than the character (1966, 126).
Genette's own contribution to this is, roughly, fourfold. First, he introduces a sharp focalization-narration distinction by setting "who sees?" against "who speaks?". Second, he defines focalization by combining and adjusting the four pre-narratological models just listed; third, he details a typology comprising three major types (zero, internal, and external focalization), roughly equivalent to Pouillon's vision and Todorov's knowledge categories. Fourth, he explicates two main types of alterations, that is, deviations from normal restrictions, such as a text's providing for too much or too little information (dubbed "paralipsis" and "paralepsis," respectively). Modestly, Genette avers that his account is basically "a reformulation" (1988 , 65).
Post-Genettean focalization theory is largely influenced by Bal's (1983 ) critique of Genette and her introduction of various new terms and definitions. Basically, Bal proposes three major modifications: (1) she throws out Genette's category of external focalization -- mainly, she argues, because it rests on a confusion of subject and object, a confusion of "who sees?" and "what is seen?"; (2) she adds the concepts of "external" and "internal focalizers," in effect re-introducing narratorial point of view (via the external focalizer); and (3) she initiates an inquiry into the nature and epistemological restrictions of perceptible and imperceptible objects. Although Bal's modifications have found the approval of many theorists, Genette himself remains largely unconvinced, passing the general verdict that "my study of [End of p. 87] focalizations has caused much ink to flow -- no doubt, a little too much" (1988, 65).1
For the present purpose, I will use the Genettean axioms as points of departure, adding, subtracting, and modifying as I go along. Specifically, I want to reassess the formula "who sees?" and the metaphor of "visual field" that defined focalization in Genette's original (later to be discarded) account. Taking these concepts quite literally, initially, I will use the following "mental model" of vision to derive a general model of focalization:
(1) A model of vision (cf. Jahn 1996, 242)
Although (1) is blatantly reductive in many respects, it integrates a number of essential concepts in an iconic frame structure. The model represents a field, an angle, a direction, and an extent, but it also suggests the deictic relations associated with a field of perception. Especially important is its incorporation of two types of foci, which correspond to two distinct meanings of the word focus in ordinary language: (i) the burning point of a lens or a mirror; (ii) an area of attention or interest. In (1), focus-1 is the burning point of an eye's lens, usually located in a person's head, and focus-2 is the area of attention which the eye focuses on to obtain maximum sharpness and resolution (foveal vision, in physiological terms). Literally or (not too) metonymically, by nature or close association, focus-1 represents a point of view, an origo, a deictic centre (Bühler), a point of origin (O'Neill), a subject of focalization (Bal). Focus-2, on the other hand, is roughly equivalent to an object of focalization; more properly speaking, an object of focalization is an object in F2. Since each point in V and F2 [End of p. 88] uniquely "vectors" from focus-1 (a cue first used by Jost [1989, 106]), V and F2 can be regarded as vector spaces, i.e., fields defined by vectors radiating from F1. It is this technical abstraction that allows one to generalize (1) beyond its purely visual denotation, promoting it not only to a general model of perception, but to a general model of focalization itself.
A focus-1 is usually located in a person's head, excluding, for the moment, any of the "camera eye" (or "hypothetical observer" or "empty deictic centre"2) options. In Genette's orginal account, the focus-1 is usually provided by a story-internal "focal character"; in other accounts, this agent is variously labelled centre of consciousness, reflector (James), refractor (Brooks and Warren, 1959 , 663), figural medium (Stanzel, 1984), filter (Chatman, 1986), SELF (Banfield, 1982), or focalizer (Bal, 1983). Without entering into any detailed terminological discussion, I will use the Jamesian term reflector for a story-internal subject of consciousness, and like Bal, Lintvelt, Cohn, Rimmon-Kenan, Toolan, Nünning, and many others (Genette and Chatman notably excepted), I am assuming that a focus-1 may also belong to a narrator.
In Narrative Discourse Revisited, Genette briefly acknowledges that his own original formula "who sees?" was too "purely visual, and hence overly narrow," and he replaces it, first, by "who perceives?", and then by "where is the focus of perception?" (1988, 64). Considering, however, that "[i]nternal focalization is fully realized only in the narrative of 'interior monologue'" (Genette, 1980, 193), Genette's modified definition still fails to apply to who thinks. In any event, the basic mold of the formula whose ___ orients the narrative perspective? suggests a more general fill-in like "subjectivity" rather than Genette's own "point of view" (Genette 1980, 186), which is not only tautological but unhappily reintroduces the term "point of view" that "focalization" is meant to replace. Indeed, supposing "subjectivity" to express a central commonality, one can quickly factor out the following more comprehensive (though still not exhaustive) list of criterial aspects:
(2) Criterial aspects of focalization
Note that categories A, B and C have been arranged so as to represent a scalar progression toward increasing amounts of what one might term "conceptuality." Apart from that, the categories exhibit considerable overlap and concurrence, and many of them are inherently very fuzzy. Yet (2) is sufficient, for the time being, to identify the types of subjectivity states, processes and data that can be assumed to vector from an origo-like source. In fact, it is the textual provision and maintenance of such origos that creates a Jamesian window on the story world.
Unfortunately, from a mainstream narratological vantage, (2)'s multiplication of focalization indicators is likely to appear suspect in several respects. Especially problematic is (2C)'s inclusion of voice since Voice was, of course, the original razor used by Genette to separate focalization from narration. But as I have argued (Jahn, 1996, 243-51), this is a division that bears rethinking. Briefly, for one thing, thought indubitably has a voice quality, and Genette himself makes a point of calling interior monologues "immediate speech" (1980, 231). Second, a reflector is, of course, capable of speech and thus in principle has a voice just as the narrator does. Third, there is good reason to assume that the narrator, like a reflector, is conceived of as a thinking and perceiving agent (an "external" or "narrator-focalizer" in post-Genettean terms). In short, whether related to speaking or thinking, to discourse or perception, conceptualization is a powerful subjectivity indicator, vectoring from and pointing back to an origo.3
Another suspect candidate in (2), again from a mainstream narratological point of view, is imaginary perception (B.ii). Mainstream narratology tends to treat a character's imaginary perception either as embedded (metadiegetic) narration (Genette's proposal, 1980, 231) or as embedded (hypo-) focalization (Bal's proposal, 1983, 255). As I have attempted to show elsewhere, both accounts are about equally questionable (Jahn, 1996, 260-1). Here, following the precept developed in a deliberately constructivist branch of cognitive science (Neisser, 1967; Jackendoff, 1987), imaginary perception is considered as being fully co-equal with ordinary perception. This view makes a virtue of the fact that it is often impossible to tell whether a perception is based on "real" sensory input, on imaginative processes, or on a combination of the two. [End of p. 90] In fact, the very lack of exclusivity in this case usefully stresses the functional interplay between ordinary and imaginary perception.
Imaginary perception is of crucial importance also on the levels of narrating and reading. As for narrating, it may suffice, for the moment, to draw attention to the role of recollection in homodiegetic narration (I will turn to other instances of narratorial imaginary perception in section 2, below). Reading is most profitably seen as a series of bidirectional processes linking, in bottom-up direction, visual input and conceptual structure, and in top-down direction, conceptual structure and mental imagery (Jackendoff, 1987, 259). Reading a narrative, in particular, involves both the building of conceptual structure and the triggering of imaginary perception (Jahn, 1996, 3.2). Indeed, the close interaction of these levels is a good reason for considering narrative a "representational" medium. Something in the nature of a disclaimer is in order, however, and I will quickly acknowledge that narrative discourse need not, and in general does not, represent story events mimetically (i.e., without any intermediate levels of "cognitive construction," cf. Fauconnier, 1990, 153), that narrative discourse is a "grainy" medium, at best (Jackendoff, 1983, 224), and that the mental imagery effected by it is less vivid, less detailed, and more indeterminate than the images of, for instance, pictorial or cinematic representations (cf. Iser, 1971, 288). Even Genette, who is an avowed anti-representationalist (1988 , 42), is prepared to admit that a sentence like "Once upon a time there was a little girl who lived with her mother at the edge of the forest" appeals to and practically presupposes "the reader's imaginative cooperation" (1993, 39-40). On this view, a reader's textually conditioned imaginary perception is not merely an incidental by-product of some exceptionally picturesque style, but part and parcel of the normal processing of narrative statements.
Of course, there is also a long, albeit non-narratological tradition that has always accepted textual representation as an activating rather than a reproductive mode, as poiesis (creation) rather than mimesis (imitation). According to Karl Bühler's deictic field theory, whose narratological relevance has often been noted,4 textually induced imaginary perception amounts to a "transposition to the Phantasma" (1965 , 124-36). Ohmann (1971, 14) argues that a literary work's necessary pragmatic condition is that it "leads the reader to imagine a speaker, a situation, a set of ancillary events." As will be shown below, Iser and Wolf take readerly imaginary perception to be the key condition of various types of "aesthetic illusions." All these intuitions demand a more detailed narratological deliberation, and this may well begin with the question of how readerly imaginary perception correlates with fields of focalization and why it might be profitable to say that narrators, talking about what they imaginatively perceive, enable readers to adopt (transpose to) fictional points of view. [End of p. 91]
2. Aesthetic illusions and windows of focalization
The following textual evidence has mainly been chosen because it exhibits typical combinations of basic narrative parameters -- narrators and reflectors, setting and character, story-internal and story-external contexts, description and diegetic report. All excerpts come from first chapters, (3), (5), and (6) being actual incipits. Consider, first, (3) and (4), which are narratorial descriptions:
(3) is a block description of a City and the setting of incipient action. Some of the detail is perception orientated, as when the City is characterized as "a set of mislaid dentures," suggesting a view from afar and above (a bird's eye view, a view from an airplane), or when its growth is likened to that of "a tree, or a disease." Evoking this imagery, the text's heterodiegetic narrator executes some rapid shifts of point of view, but it is not in fact difficult for the reader to assume that the narrator is talking about what he or she mentally sees or to execute a similar series of shifting points of view in order to visualize the existents described. The narrator's language is characterized by a perceptible critical slant stressing the arbitrariness of the place, the monotony of the landscape, the City's sinfulness, and it is at times couched in collocations that seemingly pull in opposing directions. On the whole, however, it is not difficult to trace these switches in a judgmental direction; readers will readily accept this as the narrator's idiosyncratic style and rightly expect similar things to come.
(4) is a narratorial block characterization, and here, again, what is presented issues from a heterodiegetic narrator for whom Mr. Bounderby's facial and cranial features are obviously present in [End of p. 92] imagination -- and since the imagery is so graphic, in ours as well. Like (3), (4) is characterized by a pronounced critical metaphoricity. Both (3) and (4) are narrative pauses spent on expositorial description. Following Wolf (1993, 102), they are instances of the primary illusion of existents, which is concerned with setting, characters, and things.
The following two examples have also been paired off because they exemplify the two main subtypes of Wolf's primary illusion of story events. (5) is a narrator's report of events while (6) mediates action through reflectorial perception:
The streams of the great Black Forest of Germany bubbled in playful melody no more, but rushed on with deafening din, mingling their torrent-roar with the wild creaking of the huge oaks, the rustling of the firs, the howling of the affrighted wolves, and the hollow voices of the storm.
The dense black clouds were driven restlessly athwart the sky; and when the vivid lightning gleamed forth with rapid and eccentric glare, it seemed as if the dark jaws of some hideous monster, floating high above, opened to vomit flame ....
It was indeed an appalling sight! (Reynolds, Wagner the Wehr-wolf, 5)
(6) begins medias in res in the typical fashion of what Stanzel terms a figural novel. To use Stanzel's own words, "[i]n the case of a narrative beginning with a reflector-character, the reader is obliged to [End of p. 93] forego all preliminaries and to place himself in the position of the reflector-character, experiencing the narrated events in actu. The absence of a teller and of narrative preliminaries and the reference to the reflector-character by means of a pronoun which lacks an antecedent are the narrative conditions under which this transfer is most quickly and completely effected" (1984, 160). Large portions of (6) represent the reflector's perceptions -- things he sees, feels, and hears: the "pine-needled floor"; the "gently" sloping ground; the wind blowing "overhead." The first paragraph effectively crosses and triangulates visual, haptic, and auditory perception, all crisply oriented from the reflector's origo. Indeed, if the reader chooses to stick to this prototypically "figural frame" (Jahn 1997), the two speech acts in paragraphs 2 and 3 may well be construed as representing the reflector's awareness of his own question and his hearing the other character's reply rather than objective facts.
Lastly, (7) below may serve as an instance of what Wolf terms secondary aesthetic illusion (1993, 102):
The novel's homodiegetic narrator here pauses in the act of telling to comment on his present environment, to name the current date, and to allude to his present high-strung "state of mind." The time/place coordinates establish the narrator's temporal and mental distance towards characters and events to be recounted. Additional detail (style, diction and syntax, references to "the above lines," "my little study," etc.) slowly build up a mental impression of this narrator, his voice, and his discourse here-and-now. Although divorced from story-time and story-line (excepting the leitmotif foreshadowing of the protagonist's fate), the passage is quite as origo-orientated in its perceptional, affectual, judgmental and stylistic features as the preceding ones.
A good deal more could be said about (3)-(7), particularly concerning how these passages deploy their "perception indicators" (Fehr, 1938), how they filter perceptional data, how they encode judgments, affects, and so on. At this point, however, it is sufficient to point out that [End of p. 94] (3)-(7) are all focalized passages, or, in terms of figure (1), predicated on vector fields originating from a focus-1 associated with a narrator's or a reflector's Self. Adapting Henry James's metaphor, I will say that a passage that is focalized in this manner represents a focalization window, either on story events and existents as illustrated in (3)-(6), or on the discourse here-and-now as illustrated in (7). Passages like (3)-(7) invite the reader to execute a Bühlerian transposition to fictional points of view. Focalization, in short, is a matter of providing and managing windows into the narrative world, and of regulating (guiding, manipulating) readerly imaginary perception.
3. Four types and a scale of focalization
In discussing (3)-(7) above, I argued that all of these passages are focalized. I did not, however, claim that they are all focalized in the same manner or to the same degree, and I will now turn to the question of whether the notion of degrees of focalization admits of a more precise definition.
There have been many attempts to devise a system of graded categories of "perspective," usually based on properties such as angle of vision, restriction of field (Blin, 1954), or amount of knowledge (Todorov, 1966). Generally speaking, such concepts are compatible with the iconic properties of figure (1); in fact, they can be easily implemented by drawing different angles, differently-sized fields, etc. Beyond that, however, they are notably resistant to more systematic definition.
One of Genette's focalization criteria is "selection of information with respect to omniscience" (1988 , 74). Exploiting the gradability of the terms selection and restriction, he arranges his types of focalization in a scalar typology in which omniscience marks a polar reference point ("zero or non-focalization"), subject to no restriction whatsoever, while internal and external focalization involve increasingly restrictive degrees of selection. Apart from the category of internal focalization, however, Genette's typology has not met with general approval. "External" focalization, for instance, is not based on "who perceives?" (or any of Genette's alternate focalization questions), but on how something is perceived (namely, without access to, or mention of, inside views). Zero focalization, in its final awkward definition, is "variable, and sometimes zero, focalization" (1988, 74). Discussing a case of internal focalization, Genette continues to insist that it might "just as well" be considered external focalization (1988, 75). The main problem, perhaps, is that Genette's typology mixes too many heterogeneous ingredients. Among the parameters that go into it is the number and identity of focal agencies (fixed vs. variable focalization, focal character vs. disembodied observer), the distance from which something is seen (close or far), the knowledge potential of narrators and characters (features of characterization involving quantitative and qualitative aspects), perceptional limitations (access or non-access to inside views), scope (whole texts or individual passages), and combinatorial constellations (variable or multiple). [End of p. 95]
A stalemate of a different kind arises in Stanzel's account of perspective (1984 , ch. 5). Taking a narrative text's treatment of spatial existents and relations to be particularly indicative, Stanzel distinguishes two main paradigms: "texts with distinctly perspectival spatial presentation and texts with aperspectival spatial presentation" (1984, 117). The first type occurs mainly in camera-eye novels and figural texts such as (6) or Joyce's Portrait of the Artist (Stanzel, 1984, 117-8). For a typical instance of aperspectivism, Stanzel quotes the following passage from Trollope:
Focusing on the passage's spatial descriptions, Stanzel points out that the exact nature and location of the individual pieces of furniture seems to be largely insignificant. The reader is not given, and cannot easily derive, a "concrete impression" of the room. The text's prominent authorial narrator, Stanzel says, is simply not particularly interested in evoking an "experienced perception" of spatial relations (1984, 121). Extrapolating from this, Stanzel goes on to delineate a historical pattern according to which fiction up to and including the Victorian period falls under the general rule of aperspectivism, while perspectivism becomes the "predominant stylistic trend" (1984, 122) of the modernist novel, after which aperspectivism and perspectivism join to create the disillusionary games of postmodernism. Stanzel is careful to add that these are general tendencies only and that there may be numerous exceptions.
As always, it is difficult to fault Stanzel on his textual analyses, which, in contradistinction to a good deal of narratological practice, are all the more plausible for their determined awareness of reading effects. And yet the proposed perspectivism-aperspectivism opposition is too rigid to be entirely convincing (cf. Cohn, 1981, 175). Its problematic [End of p. 96] nature typically comes out in a classroom situation. Tell a group of students that (6) is a "perspectivized" passage, and everybody will generally nod and agree; tell them that (8) is "aperspectival," and the general reaction will be one of doubt and incomprehension. And indeed, Stanzel, moving from the perspectivism of spatial presentations to perspectivism (or the lack of it) in general, modifies the meaning of perspective in midstream. Although he is aware of the fact that "the figurative meaning of the term [perspective] in the sense of view of a thing as it presents itself from the personal, subjective point of view of a novel character or narrator is at least equally important" (1984, 123), he does not fully realize that this "figurative" meaning also affects terms like perspectivization and perspectivism. This explains why Stanzel's perspectivism-aperspectivism opposition is so wobbly and why it is difficult to accept that the Victorian novel, or the novel of the Eighteenth Century, or really any known novel whatever, should fall into an "aperspectivism" slot. If Edmiston (1991) had accepted Stanzel's scheme (as fortunately he did not) he could hardly have written his insightful monograph about "Focalization in Four Eighteenth-Century French Novels."
In fact, what Stanzel's treatment of (8) unintentionally suggests is that judgments on perspectivization (focalization) have to be based on a wider set of focus-orientated subjectivity features such as those listed in (2). While one can easily grant that (8) is not as "sharply" perspectivized as (6), one can hardly fail to notice that (8) is full of indicators of oriented perception, judgment and affect, hence no paradigm case of "aperspectivism" at all. In fact, in order to capture what one might call (8)'s attenuated perspectivism one needs a finer-grained multi-category focalization scale such as the following.
(9) A scale of focalization
[End of p. 97] On (9)'s continuum (shown as an arrow proceeding from a zero point, on the left), I have placed four main categories graphically symbolizing a morphological progression. The rightmost type, strict focalization, replicates the vector-field shape of figure (1), but in order to indicate its more general nature, its "eye" has been removed and its focus-1 has been indicated by a "+" representing a point of origin. In strict focalization, F2 is perceived from (or by) F1 under conditions of precise and restricted spatio-temporal coordinates. In ambient focalization, the field of subjectivity is shown as an ellipse: like a geometrical ellipse, which has two foci, ambient focalization is based on two (or more) F1's, depicting a thing summarily, from more than one side, possibly from all sides, considerably relaxing the condition of specific time-place anchoring, and allowing a mobile, summary, or communal point of view. In weak focalization, all F1's, and with them all spatio-temporal ties, disappear, leaving only a focused object (F2). Lastly, in zero focalization the focused object itself disappears, as possibly do the limits of the perceptional field itself (which for this reason is shown as a dotted shape).
The foregoing descriptive definitions suggest that three basic properties are sufficient to differentiate the four types identified in (9): presence or absence of focus-1; the nature of the spatio-temporal link between F1 and F2; presence or absence of focus-2. Indeed, the intuitive rationale for deciding whether a passage is more focalized than another passage involves the following comparative judgments:
Passage X is more focalized than passage Y if ...
(b) X has one F1 and Y has two or more (strict vs ambient);
(c) X's spatio-temporal orientation is more determinate than Y's (strict vs ambient);
(d) X has an F2, and Y has none (weak vs zero).
Of the texts discussed so far, only the Hemingway (6) and the Mann (7) exhibit strict focalizations. Their F1's are situated in a relatively determinate here-and-now, the main frequency mode of their diegetic statements is singulative, and the events roll by in scenic sequence. Cases (3) to (5), in contrast, are less strictly focalized. [End of p. 98] Although they all have F1 origos, they use relaxed time-place deictics, their points of view are mobile, and they frequently employ the durative and iterative frequencies of summary or the static mode of description, where exact temporal sequence is not crucial. In other words, they are instances of ambient focalization. In fact, reassessing (8) in light of (2), (9) and (10) now confirms what I take to be most readers' intuition, namely that this is no case of "aperspectivism," but a case of ambient focalization. Specifically, its perceptional and judgmental point of view is that of an authorial narrator who, as Stanzel himself points out, "adopts" (1984, 121) the collective point of view of the two visitors who perceive the same things -- in other words, it has two F1's (or perhaps three, counting the narrator) and one F2. Fittingly, too, the reflectors' progress in space remains vague, and the passage's temporal sequence is considerably clouded by the narrative mode of summary.
Generally speaking, strict and ambient focalizations appear to predominate in the narrative corpus. In fact, in order to document the two "exceptional" types, I have to turn to decidedly marginal material. I will assume weak focalization to occur in passages presenting a sequence of bare facts, independent of an individual point of view or specific spatiotemporal conditions. Genette's Water boils at 100 degrees C (1988, 101) might be a suitable example, even though it falls short of narrative import. Perhaps the general idea is better illustrated by passages such as (11) and (12):
(11) presents bare story facts before their passing through a focalization filter. (12) exhaustively describes all possible return paths between four objects -- door, fire, window and bed -- and is quite obviously constructed from a nonperspectival spatial model.
It is even m ore difficult to imagine anything less focalized than (12)-(13), since a zero focalized text must have no perspective, no spatial order, no origo, no foregrounded center of interest, and no obvious organizing principle (excepting, let us say, an entirely serial or random arrangement). Instructively, anything that lacks all these characteristics succumbs to what in statistical theory is termed the zero hypothesis, that is, the hypothesis of random distribution. Hence, a random collection of narrative sentences such as (13) and a monotonously serial arrangement such as (14) can serve as suitable instances of zero focalization: [End of p. 99]
Krak! - - - - - - - Krek! - - - - Krek! - - Krik! - - Krik! - - Krik! - [. . . ] Krak! Krek! Krik!
The continuous ordinal scale presented in (9) locates two common and two marginal categories, and it defines the relation more focalized than with the help of a preference rule system, (10), which specifies a set of cognitive parameters predicting intuitive comparative judgements. In combination, (9) and (10) constitute an explicit theoretical account allowing one to describe the range of narrative focalization. An additional advantage of preference rule systems is that they facilitate extrapolation, especially when faced with borderline or intermediate cases. Since intermediate cases between strict and ambient focalization are of particular importance, consider the following passage, which is an instance of collective focalization (Banfield's "plural SELF" scenario, 1982, 96):
This passage depicts two reflectors, James and Cam, sitting in a boat at opposite ends. Between them, they see common objects such as Mr. Ramsay, Macalister, and Macalister's boy. Both James and Cam think the same thoughts about them, and thus far the scenario clearly approximates the frame of ambient focalization. On the other hand, (15) is clearly more focalized than, for instance, (8): the reflectors' perceptions and thoughts are firmly locked in a determinate story here-and-now, and the passage proceeds scenically, singulatively, and in actu, that is, in the typical spatial fixedness and temporal rhythm of strict focalization. In other words, the preference rule system detailed in (10) enables us to recognize and differentiate important intermediate cases such as these and to account for what is surely no rarity in literary texts -- something that is partly A and partly B. [End of p. 100]
4. Window shifting and deictic diffusion
The sample passages adduced so far are all assumed to be uniform with respect to focalization. Uniform focalization is one of their main coherence factors, and all categories defined are focalizations that are "local" in scope. I will now turn to passages which are no longer uniform, either because they shift from one window to another, or because one window overlaps or overlays another window.
For both scenarios, I will follow Ryan's (1987) approach to "story-line windows" and her exploitation of film and computer metaphors. Postulating a "narrative screen" (1987, 69) and defining a story-line window as the continuous "take" of an imaginary narrative "camera" (1987, 62), Ryan introduces a powerful array of window concepts including actual and virtual windows, open and closed windows, successor and continuator windows, merging and splitting windows, and many more (1987, 74). Interestingly, Ryan briefly deliberates whether the windows concept could be extended to focalized passages. She discards this intuition, mainly for two reasons: first, because many of her textual examples are not strictly or perspicuously focalized, and second, because she assumes standard focalization to be complex, involving an internal focalization embedded in an external (narratorial) focalization, "two imaginary cameras, one focused on the other" (1987, 77). Neither of these reasons is entirely compelling: on the view developed here, less-than-strict focalization can be accounted for in the guises of ambient or weak focalization, and the notion of overlapping windows is easily naturalized within the computer screen analogy used by Ryan herself. Significantly, too, the "shifting" conditions detailed by Ryan are equally as valid for story-line windows and focalization windows. According to Ryan, a window shift occurs when there is a "referential break" in the text's continuity. Referential breaks come in two forms: either "the camera goes offline" and shifts to a different location, or the "narrative clock is reset, either forward or backward" (1987, 63). Applying this to focalization windows yields two sufficient conditions: shifting from one origo to another implies a shift in location but may or may not be accompanied by a temporal shift; conversely, a text may shift from one location to another, or it may gap a period of time or turn to a previous period in time and yet remain anchored in the same origo. (16) and (17), below, illustrate the respective scenarios:
Apparently, a window shift also occurs when the text shifts from a representation of ordinary perception to one of imaginary perception. The following passage contains two shifts which are not expressly signalled, and it is interesting to observe the heuristics that enable one to recognize the shift into, and out of, the dream window:
Here, the spatial criterion alone appropriately suggests that a new window opens when the scene shifts from bed to printing office, and, similarly, that this window closes and gives way to the previous window when the scene shifts back to the reflector's bed (and ordinary perception). While it might be thought that such shifts should be identified on the strength of a qualitative difference, it is probably an advantage that our present set of shifting principles is so poor as not to include any qualitative criteria. If one posited that a shift from ordinary perception to dreaming was necessarily attended by a dream quality of some sort -- (19)'s dream window obviously has this quality -- then one would be hard put to account for verisimilar dreams. However, dream, recollection, and vision qualities might well enter the window shifting preference system as additional, non-necessary clues.
Galbraith (1995, 46-9) has suggested treating shifts into imaginary perception with reference to a hierarchy of "ontological planes" (1995, 47). On this view, texts and readers variously "push into" or "pop out of" the text's current ontological "stack" (a LIFO, or "last in, first out" stack, in artificial intelligence parlance). For instance, the reader pushes into a "higher" ontological level by moving from a reflector's ordinary perception to imaginary perception, as in the transition from (19a) to (19b). Alternatively, the current window may "pop off" the stack and return reader and text to an underlying level, as in the transition from (19b) to (19c). Similar "pops" also occur when a text shifts from quotational inset to quotational frame, from figural perception to narratorial report, from narratorial report to comment, and so on.8
It is obvious from the foregoing examples that a theory of shifting [End of p. 103] can be used to explain the various arrangement, grouping, and cutting patterns of focalization that a text exhibits on a macronarrative level of analysis. Shifting also has the effect of partitioning an extended heterogeneous passage into homogeneously focalized windows. There is another type of focalization heterogeneity, however, that seems to be due not to shifting but to mutual interference between concurrently active windows. In this case, it seems sensible to follow Ryan's (1987, 78) suggestion (again making use of the computer and film screen analogy) that windows can coexist, overlap, or overlay. In fact, a text's deictic elements which, according to (2C), are criterial focalization indicators, often imply more than one origo. Consider the following synthetic sentence (cf. Hamburger, 1977, 59-72; Banfield, 1982, 154-67; Galbraith, 1995, 25):
In (20), "now" supposedly indicates a story-now and "she" indicates a figural deictic center. At the same time, "she" has the deictic import of 'not-I-who-is-speaking', just as the past tense has the deictic import of 'not-now-as-I-am-speaking' so that both references are anchored in a deictic center other than the current reflector. Hence the deictic vectors in (20) pull in different directions, creating an effect of "deictic diffusion."
In the literature, several suggestions for dealing with deictic diffusion have been brought forward. According to one view, deictically diffuse passages present a "dual perspective" (Hernadi, 1973); according to another, they are a "dual voice" phenomenon (Pascal, 1977); in a third, they are due to "allophonic focalization" (Füger, 1993, 48). Persuasive as these proposals may seem in restricted areas of application, they reduce the matter either to a purely perceptional or to a vocal issue, leaving the precise character of the "duality" undefined. Pursuing a more radical tack, some deictic shift theorists (Duchan et al, 1995) follow Hamburger (1977 ) and Banfield (1982) in assuming a "speakerlessness" position that conveniently filters out all complicating narratorial noise. However, as Cohn (1990, 795-97) and many others have pointed out, this view not only ignores the deictic implications of pronouns and tenses that were noted in (20), it also fails to account for the very real option of a dialogic coexistence of authorial narrators and third-person reflectors. Against these proposals, a more viable solution would be to assign deictic elements different weights or orders of precedence (cf. Cohn, 1990, 797) and to posit, for instance, that a present and active figural deictic center and an explicit story-now take precedence over a narratorial deictic center and an implicit discourse-now. This stipulation amounts to a preference rule that captures a common case such as (20) and at the same time happens to be fully compatible with Galbraith's notion of a stack of ontological planes. Since the current window is always the one on top of the stack, the lower levels may be taken to be partially occluded (as on a computer screen) or backgrounded, but leaving some "deictic residue." Deictic residue not only [End of p. 104] accounts for a past tense without a prominent past meaning,9 but also for the reduced perceptibility of "covert" narrators (Chatman, 1978, ch. 5). Backgrounded deictic residue can also serve as a much-needed "excuse" -- in the non-pejorative sense of frame theory (Minsky, 1979  , 18) -- for treating third-person, past-tense passages such as (6) as fully "internally focalized" segments. Similar excuses are, of course, also needed for retaining a given focalization frame in the presence of nonconforming conceptual material such as quoted, heard, mentioned or imagined linguistic matter.
More active interferences between narratorial and reflectorial conceptualizations usually result in a phenomenon variously termed substitutionary perception (Fehr, 1938) or discordant psychonarration (Cohn, 1978). In substitutionary perception, a reflector "lends his sight" while a narrator "substitutes" his words (Fehr, 1938, 97-8); in discordant psychonarration, the narrator's diction and style transcend the reflector's perceptional or conceptual ability. Such passages may well be experienced as juxtapositions of two windows, one anchored in a reflectorial "fallible filter" (Chatman, 1986), the other on a knowledgeable and well-spoken narrator. Discordant psychonarration is most obvious when the reflector is an infant (What Maisie Knew) or an animal. Consider (21), below, in which we witness the shooting of a lion from the animal's point of view.
Even though (21) initially makes some effort to create a concordant representation of the animal's perception, plausible lion concepts such as "the object," "a thing" (i.e., the hunters' car), "a man figure," a "biting" pain, etc., heavily collide with "a .30-06 220-grain solid bullet," a technical term accessible only to the narrator and momentarily foregrounding the window of the secondary aesthetic illusion playing the narrator's voice.
A similar type of conceptual diffusion underlies the stylistic device termed "colouring," (Hough, 1978, 49-60; Fludernik, 1993, 334-8), "contamination," or "reflectorization" (Stanzel, 1984, 168-84; Fludernik, 1996a, ch. 5.1; Fludernik, 1996b). In colouring, reflectorial diction is non-ironically integrated into narratorial sentences, as in the following passage: [End of p. 105]
In (22), a few snippets from the character's idiom ("betteranever," "dug the cats") colour not only directly quoted and free indirect discourse passages (where colouring is expected and normal), but also the diegetic statements themselves (where it is not).10
Obviously, the examples adduced in this section do not exhaust all shifting and overlapping patterns, nor do the windows concepts introduced here provide a means for addressing all features of narrative discourse. However, what has hopefully become apparent is that the present approach provides some promising initial heuristics for analyzing a wide range of micro- and macro-narrative patterns of focalization.
This paper began by reviewing the original axioms of focalization and embracing its narratological basis. In order to evolve further, however, the theory must outgrow any mere "reformulation" (Genette, 1988 , 65) of traditional accounts. Nor can it remain satisfied with "diadic" or "triadic" typologies (Cohn, 1981, 175) based on too few or too many criteria. Ultimately, the present shortcomings of the theory were shown to be mainly due to a tendency toward overprotection. Genettean narratology tends to deny the representational effect of narrative for fear of committing the mimetic fallacy, and it implements its categorical distinction who speaks? / who sees? in order to avoid the fallacy of treating characters as narrators. There may have been good reasons for introducing such fallacy stoppers in the seventies, yet it is apparent today that the consequent compartmentalizations closed the door to many promising avenues of inquiry. This paper attempts to reopen that door, primarily by stressing the role of imaginary perception in the process of reading and in the reader's construction of fictional worlds. Methodologically, it executes the "cognitive turn" called for by Ibsch (1990), and it does so by applying a set of interdisciplinary concepts such as mental models, frames, stacks, and preference rules.
At this point, two main projects suggest themselves for further treatment. First, a number of theoretical questions need to be addressed in greater detail, particularly as concerns the phenomenon here identified as "deictic diffusion." It remains to be seen [End of p. 106] whether the solutions tentatively suggested -- assuming a context of overlapping windows or of "excusing" the diffusion on a rationale of deictic weight -- will ultimately prove useful. Second, we must obviously also push beyond the dryly theoretical business and see whether our extended toolbox (preferably used with "intelligence, ingenuity, and tact," as was once said of Narrative Discourse [Prince 1980, 413)] enables us to undertake new types of textual analyses, to arrive at new insights about themes, styles, techniques, and effects.
1. For post-Genettean accounts of focalization see Bal (1983 ; 1985, ch. 7), Lintvelt (1981, 47), Vitoux (1982), Rimmon-Kenan (1983, ch. 6), Cordesse (1988), Toolan (1988), Kablitz (1988), Nünning (1989, 53-60), Jost (1989), Nelles (1990), Edmiston (1991), Delayto (1996 ), O'Neill (1994), Herman (1994), Ronen (1994, ch. 6), Jahn (1996), Nieragden (1996).
2. On the "empty deictic centre" option, see Banfield (1987) and Fludernik (1996a, 5.2).
3. Saying this, I am not, of course, suggesting that we reset the notorious point-of-view trap which originally motivated Genette to oppose "who speaks?" and "who sees?" Dichotomizing speakers and seers/perceivers effectively inhibits Booth's point-of-view blunder of "christening 'narrator' a focal character" (Genette,1988, 65); however, it also prevents us from recognizing that narrators are "capable of both speaking and seeing, and even of doing both things at the same time" (Rimmon-Kenan, 1983, 72).
4. See, for instance, Hamburger (1977, 104-11), Stanzel (1984, ch. 4.5), Fludernik (1993, ch. 1.3), Galbraith (1995).
5. "Ich überlese die vorstehenden Zeilen und kann nicht umhin, ihnen eine gewisse Unruhe und Beschwertheit des Atemzuges anzumerken, die nur zu bezeichnend ist für den Gemütszustand, in dem ich mich heute, den 27. Mai 1943, zwei Jahre nach Leverkühns Tode, will sagen: zwei Jahre nachdem er aus tiefer Nacht in die tiefste gegangen, in meinem langjährigen kleinen Studierzimmer zu Freising an der Isar niedersetze, um mit der Lebensbeschreibung meines in Gott ruhenden -- o möge es so sein! -- in Gott ruhenden unglücklichen Freundes den Anfang zu machen" (Mann, Doktor Faustus, 9).
6. I must point out here that both (13) and (14) are in fact embedded texts, and that their contexts largely recuperate them for more standard types of focalization. For instance, the frogs' chorus in (14) is Watt's perception.
7. For an inventory of such indicators, see Fehr (1938), Lethcoe (1969, ch. 2), Ehrlich (1990), Zubin/Hewitt (1995), Wiebe (1995).
8. See Ryan (1991, ch. 9) for a convincing application of the LIFO stack concept to embedded narratives.
9. For a discussion of the past or non-past meaning of the "epic preterite," see Hamburger (1957, 59-72), Stanzel (1959), Fludernik (1993, 47-51, 198-9).
10. "Reflectorization" more generally comprises both characterial diction and figural perception.