Tense, Mood, and Voice in Film:
(Notes after Genette)

Brian Henderson

Narrative Discourse^[*] is a translation of "Discours du récit," which is the major part of Figure III (1972) by Gérard Genette. Genette's topics are traditional ones of literary theory and criticism: order, duration, frequency, mood, and voice in classical fiction and the subversion of these in modern texts. Genette offers no new method of literary analysis but a clarification, systematization, and extension of older ones, amounting to a transformation of them. The book is immensely stimulating and has already been quite influential.

Aside from a reference or two, Genette does not treat film, and his work cannot be directly applied to film. Each of his categories must be rethought on the ground of film analysis, a project whose value is argued below.

What does Narrative Discourse do? It studies Proust's A la recherche du temps perdu and it studies narrative discourse; it is at once the criticism of one text and a theory of narrative. To do both without subordinating one to the other is a paradox; but it permits "a refreshing rotation and mutual entertainment between theoretical dryness and critical meticulousness." Genette compares the project to an insomniac turning over and over in search of a better position.

His study of narrative discourse is limited to the narrative text itself. He does not compare the Recherche to actual events, as one might a history, nor to the circumstances of its production, as might a biography of Proust. Rather he addresses the relationship between a narrative discourse and the events that it recounts and the relationship between the discourse and the act that produces it. But it is the narrative discourse alone that informs us both of the events that it recounts and of the activity that supposedly gave birth to it.

Narrative Discourse, An Essay in Method , translated by Jane E. Lewin (Ithaca: Cornell, 1980). Figures of Literary Discourse , translated by Alan Sheridan (New York: Columbia University Press, 1982) comprises translations of Genette essays from Figure I (1966), Figure II (1969), and an additional essay from Figure III . See also Seymour Chatman, Story and Discourse: Narrative Structure in Fiction and Film (Ithaca: Cornell, 1978), especially pp. 63-79.

Genette distinguishes three aspects of narrative reality within the narrative text: story refers to the narrative content, that is, to the chronological order of events implied by the narrative; narrative refers to the narrative discourse or text; narrating refers to the producing narrative action.

Having made these distinctions, Genette turns to method. Since any narrative is a linguistic production, telling of one or several events, he argues that it is legitimate to treat it as the development given to a verbal form, in the grammatical sense of the term: the expansion of a verb. "I walk, Pierre has come are for me the minimal forms of narrative, and inversely the Odyssey or the Recherche is only, in a certain way, the amplification (in the rhetorical sense) of statements such as Ulysses comes home to Ithaca or Marcel becomes a writer ." This perhaps authorizes us to formulate the problems of analyzing narrative discourse according to categories borrowed from the grammar of verbs. Genette reduces these to three basic classes of determination: tense, mood, and voice. Tense, which includes order, duration, and frequency, has to do with temporal relations between narrative and story. Mood has to do with the modalities, i.e., the forms and degrees of narrative representation. Voice has to do with the mode of action of the verb in its relations with the subject of the statement, not an actual writer or speaker, but the subject of the enunciating itself; more broadly, it has to do with what Genette calls the narrative situation and its two protagonists: the narrator and his audience, real or implied. These terms are borrowed from the grammar of verbs for narrative analysis; Genette claims no rigorous homology between the two instances.

To study the tense of a narrative is first of all to compare the order in which events are arranged in the narrative discourse with the order of succession these events have in the story. Types of order include straight chronology and anachronies, a general name for deviations from straight chronological order. Anachronous narratives begin, in one way or another, in medias res then go backwards (analepses) or, much less frequently, forwards (prolepses). Analepses are external when their entire extent happens before the beginning of the first narrative, commonly used to give background to the present action. They are internal when they fill in gaps occurring later than the start of the novel. There are also lateral ellipses, in which the narrative sidesteps an element within a chronological passage. Genette calls this paralipsis, ellipsis to the side; like a temporal ellipsis it may be filled in later.

It is not difficult to apply Genette's basic concepts of order to cinema. It seems that the majority of films of every era have been told in straight chronological order whereas, according to Genette, such order is the exception among novels. Prolepses are relatively rare in the novel but they are even rarer in cinema. Flashforwards are sometimes used by Resnais and Roeg, on the whole integrally to their projects; in some sixties and early seventies films they were used

rather unsystematically as instances of "mind" or "cinema" or both, e.g., Easy Rider .

Anachronies in cinema come down mainly to analepses. Such figures are extremely rare in the first twenty years of film history, although one does find there dreams, visions, and even "reviews" of material shown earlier in some sequels, episodes of serials, etc. In Broken Blossoms (1919), Griffith introduces Battling Burrows then flashes back to his recent victory in the ring to explain his bad temper at having to keep in training; he introduces Lucy then flashes back to her encounters with a weary wife and with some prostitutes to show the options that hem her in. In the manner of classical fiction, The Cabinet of Dr. Caligari (1919) casts its main story as a long analepsis inside a framing story set in the present; but like Last Year at Marienbad (1961), this "past" may be delusional and therefore the "remembering" itself a strictly present action. Victor Sjöstrom's The Phantom Chariot (1920) exhibits a remarkable mastery of cinematic tenses, at one point presenting a flashback within a flashback.

The greatest period for analepsis in classical cinema was undoubtedly 1941–1957, that maturity of the sound film marked by Citizen Kane (1941) and How Green Was My Valley (1941). The latter begins with an older Huw leaving the valley and remembering it as it was; the rest of the film presents that remembering and never returns to the framing story, recurring at the end to memory images from the early part of the analepsis iteself, in a kind of looping of the past. More usual is the analeptic structure of All About Eve (1950), which returns to the framing story at the end and carries it forward for an additional scene or two. Citizen Kane, Double Indemnity (1944), and Sunset Boulevard (1950) follow this pattern, as indeed do most analeptic films of the classical period.

Analepses in the cinema have nearly always involved the use of language, sometimes explicit dialogue setting up a return to the past, more often the direct intervention of language in the form of titles or voice-over. Without such linguistic cues, and quite often with them, analepses in classical cinema are usually marked redundantly by plot, indirect dialogue, music (dreamlike or "mental"), and optical devices such as blurs, rippling, fades, or dissolves. Classical cinema reacts to a tense shift as though to a cataclysm; the viewer must be warned at every level of cinematic expression, in sounds, in images, and in written language, lest he/she be disoriented.

Why is this? Cinema has no built-in tense system as language does. One cannot write a sentence without indicating tense but one can apparently make a shot, and therefore perhaps a film, without indicating tense. In written discourse the tense structure must be renewed with every sentence and, concomitantly, a single sentence may subvert that structure; Genette pauses over the tense structure of several individual sentences by Proust and their implications

for the narrative as a whole. In the cinema, one is tempted to say, if the camera keeps running or if there is a cut to another angle within the scene, continuous temporality is the result; the cut to another sequence is read as straight chronological order, unless otherwise marked. One must do something, indeed a great deal, if a comprehensible shift of tenses is to be achieved in cinema. So, at any rate, classical cinema has persuaded us by making straight chronological order seem natural and inevitable. Cinema, unlike language, is, in Barthes's terms, a "complex system in which different substances are engaged," in which "the senses are subjected to the concerted action of a collection of images, sounds, and written words." Language need only change a word-ending to indicate tense shifts; cinema must pass from one complex conjunction of communication channels to another and provide a plausible transition between them, all without an explicit tense system.

This is one reason why films like Wild Strawberries (1957) and 8 1/2 (1963) made such an impact when they appeared and have had such influence. They went back and forth between past and present without explicit linguistic cuing and with a subtlety as to other signals that made the old ways seem heavy-handed. These films seemed to abolish the rhetoric of tense shifts in cinema; in fact they substituted new rhetorics for old. 8 1/2 in particular convinced viewers that it had achieved the seamless interaction between past and present that it had aimed for. Any spoken version of "I remember" would have defeated the film's psychological subtlety and cinematic fluidity. But other, well-hidden signals prepare the audience for tense shifts and guide it back smoothly. These include the context of setting and plot and Fellini's isolation of Guido by reframing or character movement just before a return to the past. There are also at times different optical treatment of past and present, wind sounds, fading of present voices, etc. Rota's music is used to bridge time shifts expertly as is Mastroianni's detachment and bemusement, his inwardness amidst public scenes, slight modulations of which carry him (and us) from something in the present to something in the past. He is passive yet alert to his own visions, as though receiving them from outside and shaping them at the same time. Indeed Fellini brings in a pair of mind-readers to keep his hero from introducing the past on his own ("Asa Nisi Masa").

Certain films undermine our ability to distinguish analepses from straight chronological order and thereby subvert narrative order itself. These include works by Buñuel, Duras, and Rainer, not to mention the semi-narrative films of the avant-garde. Made well before Marienbad —and more lucidly—Maya Deren's Meshes of the Afternoon (1943) tempts us to discover a single narrative order; our effort is necessary to the film's operation but is of course doomed to fail.

Perhaps the most formidable subversion of order specifically, of order as order, is Not Reconciled (1965) by Jean-Marie Straub, an adaptation of

Mastroianni as Guido in  8 1/2

Heinrich Böll's long Billiards at Half-Past Nine (1962) into a 53-minute film. The story deals with three generations of a Cologne family and the resistance of some of its members to militarism in World War I, in the thirties, and in the postwar period in which it is set; it all takes place on Heinrich Fähmel's eightieth birthday, a Ulysses -like day on which various characters remember the past. Straub actually straightens out the novel's jumbling of the events of the day and eliminates most of its flashbacks. Nevertheless the film is vastly harder to follow than the novel because Straub has omitted most of the signals that cue viewers as to time shifts.^[*] He uses straight cuts to go to the past and to return; there are no optical devices and no music cues. Voice-over is used for the past

The film begins with a three-second shot of a man playing billiards, saying to a bellboy, "Tell me what, boy?", which is followed by eight shots with voice-over concerning boys involved in political conflicts at a high school. Then there is a shot of the billiard player ordering a cognac and five shots of the bellboy going to get it, including two flashbacks concerning his life at the hotel. A shot back in the billiard room, in which the bellboy asks if there is more to the story, is followed by twelve more shots, with voice-over, of one of the schoolboys being beaten, being smuggled out of the country, etc. At this point, less than seven minutes into the film, first- and even second-time viewers have no idea where they are and have no way to regain their footing for the rest of the film. For a fuller treatment, see Richard Roud, Jean-Marie Straub (New York: Viking, 1972), which remains the best introduction to its subject.

material itself, but not to bridge present-past shifts. Plot context and period detail are probably what viewers most rely upon to construct the narrative order of a work. Here too Straub withholds orientation: he makes his thirties shots and his fifties shots virtually indistinguishable in dress, environment, gestures, mode of speech, etc., which is also a thematic statement. (Some of the World War I–era shots have period flavor.) Also, there is never enough context provided to know with certainty which shots belong to which period, let alone how the various events depicted relate to one another. It is only upon multiple screenings or upon reading about the film, or both, that we realize that some of the characters are the same characters at different ages; there are two Roberts, two Schrellas, two Johannas, and two Heinrichs. Most remarkable of all is the temporal coherence of Not Reconciled when one has studied it carefully. Its order is straight chronology with three major analepses (Robert's two-part story of the thirties and Heinrich's story of 1910–1918) and two minor ones (Hugo the bellboy's life at the hotel). Straub is right in calling it a simple film but it takes a great deal of work to understand that.

All of our examples, from Griffith to Straub, have stayed within Genette's basic categories of straight chronology and analepsis. It is notable that he himself does not stay within these simple concepts. Thus he finds in Proust "repeating prolepses" (playing a role of advance notice), "prolepsis on analepsis," "analepsis on prolepsis," "analepsis on paralipsis," etc. Our discussions indicate that cinema has not (yet) developed the complexity of tense structures found in literary works. They also indicate that in cinema analysis the interesting questions have to do not with narrative order itself, but with how films indicate this order, both straight chronology and deviations from it.

For Genette, duration is a more difficult question than that of order or frequency because the time needed to read a text depends upon factors external to it. There is no "normal" reading speed so one cannot pose a hypothetical equality between narrative time and story time, against which to measure variations. One can, however, compare parts of the narrative discourse to each other to determine their relative speeds. This method applies only to the large narrative articulations of a work and not to detailed studies of rhythm, which, Genette notes, would lack rigor. Thus he divides Proust's text according to its principal temporal and spatial breaks, determines the number of pages devoted to each, and dates these sections by the story time they cover. He then expresses these findings in an overall ratio: the Combray episode devotes 140 pages to ten years, the Gilberte episode devotes 200 pages to two years, etc. On the basis of this table Genette assigns a range of variations to the Proustian text: from 150 pages for three hours to three lines for twelve years. He also notes a durational evolution in the overall text: a gradual slowing of the nar-

rative through longer and longer scenes and a corresponding increase in the number and length of ellipses between such scenes. Genette calls such variations of speed, which all narratives have, anisochronies or effects of rhythm.

Films, of course, have fixed viewing times. One may compare a length of film, measured in minutes and seconds or in feet, to a portion of story, or vice versa, and one may assemble these findings in an exact table. One may examine the large narrative articulations of a whole film or the microrhythms of very short passages. Genette's account of duration does not enlarge these methodological options; he is dealing with a less exact case. But by making the question of duration newly interesting, he provides new motivation to make such analyses.

Duration in cinema is a very delicate matter, as anyone knows who has shown Straub or Michael Snow to an unreceptive audience. Even viewers sophisticated in other respects become angry if their sense of proper filmic duration is challenged.

If Straub's transgressions in Not Reconciled have to do with order, those of his Chronicle of Anna Magdalena Bach (1968) have to do with duration. The 93-minute film consists almost entirely of images and sounds of musicians in period costume playing Bach's works. An account of the circumstances of the composing and playing of these works, of Bach's financial difficulties, of the births and deaths of their children—and of little else—is spoken over the images and music by Anna Magdalena. She usually speaks over the ending of one work and the beginning of another or in very brief intervals between them. She also speaks very rapidly, as though speaking of family matters and external circumstances hurriedly between musical performances. Straub calls it a Marxist film because it presents the actual production of music (directly recorded), to which Bach's waking hours were massively devoted, as well as his marginal economic status and his struggle for the right to select his musicians. Despite superb performances of well-chosen Bach works, Chronicle still has the power to empty theaters and classrooms.

In the commercial realm, Frank Capra reports in his autobiography that the first version of Lost Horizon (1937) was too long and had too much background to the main story. As a result, the preview audience was restless throughout the screening. A durational miscalculation at the beginning threw off the reception of the entire film. Capra's response was to "Burn the first two reels," his chapter title for the incident, an indication of how strongly the durational code is enforced.

Film's multi-channelled textuality raises durational problems, and opportunities, not found in literature. Image, dialogue, voice-over, music, sound effects, and written materials may contribute, complementarily or redundantly, to a single duration; or they may create multiple, simultaneous, or contradic-

The Chronicle of Anna Magdalena Bach

tory temporalities. Modern films have exploited such possibilities in a variety of ways. In Le Petit Soldat (1960), like The Lady from Shanghai (1948) before it, a final shot of the hero is held while the voice-over summarizes a much larger time period:

It was after killing Palivoda that I learned of Veronica's death. Only one thing was left to me: learn not to be bitter. But I was happy, because I had a lot of time in front of me.

Similarly, Badlands (1973) opens with a long take of Holly in bed playing with her dog, over which we hear her voice-over:

My mother died of pneumonia when I was just a kid. My father had kept their wedding cake in the freezer for ten whole years. After the funeral he gave it to the yard man. He tried to act cheerful but he could never be consoled by the little stranger he found in his house. Then one day, hoping to begin a new life away from the scene of all his memories, he moved us from Texas to Fort Dupree, South Dakota.

The classical cinema is not without its durational contradictions either, although one must look more carefully to find them. How Green Was My Valley (1941) covers many years of its novel source in less than one year, including births, deaths, two sets of emigrations, the transformation of the valley from garden to slagheap, etc. In the same year Huw begins school, graduates, and goes to work in the mines. The decision to use Roddy McDowall as Huw

Le Petit Soldat

throughout the film imposes this durational contradiction: Huw does not age while events around him suggest years going by. The other players do seem to get older—there is even an excuse for his mother's hair turning white suddenly. The duration of the plot and that implied by the performance level are contradictory, something that could not happen in literature.

Genette continues his discussion of duration and prepares for his discussion of frequency by proposing a general system of narrative speeds. In theory there is a continuous gradation from the infinite speed of ellipsis, defined as story time without discourse time, to the absolute slowness of descriptive pause, defined as discourse time without story time. In fact the narrative tradition has reduced that liberty to four basic relationships that Genette calls the canonical forms of novelistic tempo. Besides ellipsis and descriptive pause, there are the scene, most often in dialogue, realizing an equality of time between narrative and story, and the summary, a form with variable tempo—that of the other three is fixed—that covers the entire range between scene and ellipsis. Summary remained, to the end of the nineteenth century, the most usual transition between scenes, the background against which scenes stood out, and thus the connective tissue par excellence of novelistic narrative, whose fundamental rhythm was defined by the alternation of summary and scene. Genette ingeniously shows that, despite appearances, there are almost no summaries in Proust and no descriptive pauses. Proustian descriptions are always grounded in a character's perception and therefore are parts of scenes. Moreover, narrative cutting in Proust is never accomplished by Balzacian summaries but by a quite different kind of synthesis, that of the frequentive or iterative, which is used in conjunction with ellipses.

The cinematic equivalent of Genette's scene is clear enough; indeed the notion and the prestige of the dramatically present moment were borrowed by the novel from theater. Summary we will consider in relation to the iterative, below. Ellipsis in cinema is a vast topic; aside from single-take films on the Lumière model or virtual long takes like Hitchcock's Rope (1948), there can be no cinema without ellipsis. In Rossellini, Godard, and other figures of modern cinema, as in the later Recherche , ellipses tend to become longer and more abrupt. The scenes between gaps are not necessarily longer, as in Proust, but may be denser in synthesized content (Rossellini) or in cinéma-vérité detail (Godard). Rossellini's ellipses trace the movements of an argument unconcerned with conventional diegesis, though what he does show is usually "neorealist" in its detail and objectivity. Similarly, but differently, Godard likes to balance the ethnographic detail of actual locations, direct sound, and observed gestures with the arbitrary jumps and dénouements of fiction, often expressed through ellipses. (In Masculin-féminin [1966], Paul's death occurs between the next-to-last and the last sequences.) Generalizing is hazardous, but both Rossellini and Godard tend to use long takes around their ellipses, as though their abruptness and force would be lost amidst montages.

Description poses special problems for film analysis because every shot serves a descriptive function, whatever else it may do. At the same time no shot is entirely descriptive, therefore cannot be a true descriptive pause, because the fixed time of film viewing makes it dramatic also. Even if no action occurs in this shot or in this setting, the time devoted to them builds expectations for action to come; they too are ticks on the dramatic clock. Indeed few things build more expectancy than silent shots of objects in a narrative film.

Speaking very broadly, some films seem to invent actions to fill out projects descriptive of a time or place (Ford's Judge Priest ); others describe an environment so that we will understand the action to follow (every heist film). The opening scenes of classical narrative films generally describe a setting while introducing us to the protagonists of the action. The trick in classical script construction, as in the theater before it, was to accomplish this while already advancing the action. The opening sequence of How Green Was My Valley , told in the frequentive tense, describes a complex setting in detail, introduces us by name and personality to eight or nine characters, indeed to an entire village, and immerses us in a nostalgic vision of a way of life that will begin to deteriorate in the next sequence.

Description provides a bridge to frequentive or iterative narrative, for as Genette says, "The classic function of iterative narrative is thus fairly close to that of description, with which, moreover, it maintains very close relations." Narrative frequency has to do with the relations of repetition between the narrative and the story. Genette proposes a system of relationships between the

capacities for repetition on the part of both the narrated events of the story and the narrative statements of the text. This system is reducible to four virtual types. One may narrate once what happened once; one may narrate n times what happened n times; one may narrate n times what happened once; one may narrate one time what happened n times. The last is the frequentive or, as Genette prefers to call it, iterative narrative, in which a single utterance takes upon itself several occurrences of the same event: "For a long time I used to go to bed early."

In classical narrative, iterative sections are almost always functionally subordinate to singulative scenes, for which they provide a sort of informative frame or background. In this the classical function of iterative narrative is fairly close to that of description: both are at the service of the narrative as such, which is the singulative narrative. The first novelist to liberate the iterative from its functional dependence was Flaubert, who constructed entire passages in its mode. But no novel has given the iterative a scope, importance, or technical elaboration comparable to Proust's. Genette develops these points in a series of fine analyses, including a classification of kinds of iteration that need not concern us here. Suffice it to say that his elaboration of the iterative is one of the most original things in Narrative Discourse .

In classical cinema, as in the classical novel, the iterative nearly always has an information or background function subordinate to singulative scenes. The exterior shots of trains or buses in backstage musicals signify numberless such journeys. If the words "Baltimore," "Philadelphia," "New Haven" appear under a series of train shots, as in The Bandwagon , then there is a narrative summary as well as iterative narrative; a specific itinerary is signified as well as numerous rides on trains. The montages of nightclub signs and clinking champagne glasses in romantic comedies are iterative in function. If they stand for a series of such evenings, they are what Genette calls generalizing or external iteration; if for one of them, what he calls synthesizing or internal iteration.

In Trouble in Paradise (1932), Lubitsch shows Mme. Colet's typical activities as head of a company by cutting from her on a stairway to four shots of servants and assistants down below, who say, respectively: "Yes, Mme. Colet," "No, Mme. Colet," "No, Mme. Colet," "Yes, Mme. Colet." This rhyme is itself rhymed later when Gaston has taken over direction of the company. Lubitsch shows this change (summary) and the typical obedience that Gaston now receives (iteration) by cutting from him to another series of servants and subordinates: "Yes, M. LaValle," "No, M. LaValle," "No, M. LaValle," "Yes, M. LaValle." Needless to say, the repeated yesses and nos also echo and mock the sexual oscillation that is going on in the plot.

The opening sequence of How Green Was My Valley presents a large and sustained figure of iteration; it takes up over half of the film's entire spoken

narration by the older Huw, whose face we never see. As he is leaving it for the last time, Huw remembers his valley as it was when he was a boy, describing a typical day through a repeated use of the word "would," the principal verb form for the frequentive in English.

Someone would strike up a song and the valley would ring with the sound of many voices . . . Then came the scrubbing . . . Most [of the coal] would come off but some would stay for life . . . There was always a baron of beef or a shoulder or leg of lamb before my father. There was never any talk while we were eating . . . My mother was always on the run . . . After dinner, when dishes had been washed, the box was brought to the table, for the spending money to be handed out. No one in our valley had ever seen a bank. We kept our savings on the mantlepiece.

At this point, when Huw gets his penny, hurries to the candy shop, and meets Bron, his future sister-in-law, the voice-over narration suffers an odd sea-change, dropping almost imperceptibly from the iterative to the singulative mode.

It was on this afternoon that I first met Bron—Bronwen. She had come over from the next valley for her first call on my father and mother.

I think I fell in love with Bronwen then.

Which afternoon is "this," since the voice-over has been describing the typical activities of many days but of no particular day? Thematically one might call this a fall into time: out of the frequentive mode, which is preserved, as in Proust, from the ravages of time, into the singulative mode, the order of irreversible time and change, in which the family will be dispersed and the valley destroyed. The fall into time is also a fall from grace, corresponding to Huw's first love attachment outside of the family as well as to his brother's marriage, the first actual division of the family.

Most interesting, theoretically, is how and why such a slippage is possible in cinematic iterative; it would not be in literary narrative. Are the images neutrally either iterative or singulative, depending upon a voice-over to define their temporal status? If so, then the continuity of the images may sustain us through a shift in tense, while the changed verbal tense tells us how to read the images. Or is there a shift of reference here? The images have seemed to follow the voice-over narration, as though illustrating it; perhaps, its introductory function accomplished, the voice-over now refers to the images. "This afternoon" seems to refer to the images we see before us. This shift may prepare us for the

withdrawal of the voice-over in favor of the images, which will now be furnished with synchronized dialogue; this is what happens in How Green Was My Valley . After Huw describes his love for Bronwen and the scene shifts to her introduction to the family, from which young Huw is pointedly excluded, his voice-over ceases—except for seven or eight more interjections of a clearly summarizing sort throughout the rest of the film.

Our discussion of How Green Was My Valley raises the problem of what Genette calls the pseudo-iterative:

scenes presented, particularly by their wording in the imperfect, as iterative, whereas their richness and precision of detail ensure that no reader can seriously believe they occur and reoccur in that manner, several times, without any variation.

This happens occasionally in Proust; in cinema it is pervasive because "richness and precision of detail" seem to be inherent in photography and sound recording, at least in those practices of them engaged in by most films. Thus Donald Crisp and company walk to the pay window in precisely this way only once, as recorded in the take included in the release print of the film. Some might take an ultra-literal position and say that this makes the iterative impossible in cinema; its images and sounds are always singulative. But, as noted, cinema is a "complex system"; it is the "concerted action" of images, sounds, including voice-overs, and written words that create meaning in cinema, including iterative constructions. And, as Metz once argued, the fact to be understood is that films are understood, including the cinematic iterative.

Nevertheless, the singulative tendency of recorded sounds and images can have a backlash effect on iterative constructions. How Green Was My Valley goes from the iterative to the singulative by virtue of a tense change in the voice-over. A much more frequent figure is a scene that begins in the iterative, by virtue of a title or a voice-over, then becomes singulative, as though reverting to the singulative in the absence of continued linguistic definition to the contrary. A title in Lady Windermere's Fan (1925) tells us that the way a gentleman rings a lady's doorbell reflects the state of their relationship and we see a tentative finger finally engage the button. "But when he knows her well . . ." says a later title, and we see an impatient finger ringing and ringing. Under the pressure of the title, these images are iterative: they narrate once what happened many times; but in the scenes that follow each image, singulative events occur. One might say that the doorbell images are iterative and the rest of each scene singulative; or better, that the doorbell images have a dual status, belonging both to an iterative construction and to a singulative one.

There is also the opposite problem, what Genette calls "a sort of contamination by the iterative" of a singulative scene. In Proust this is produced by a few iterative sentences within a singulative passage. In cinema, as in the other performing arts, there are many ways to suggest habit, frequency of occurrence, and typicality within a singulative context. Indeed, how many plays and films, from King Lear to On the Waterfront , open on a scene of habitual order just as it is about to be changed irrevocably.

Genette's chapters on mood and voice painstakingly and brilliantly distinguish them. Perhaps for that reason, his treatment of the two topics is very much intertwined. Strictly speaking, the mood of narrative statements can only be the indicative since the function of a narrative is to report facts, real or fictive, not to give an order (the imperative), to express a wish or state a condition (the subjunctive), etc. Admitting that he is extending the linguistic metaphor, Genette argues that within the indicative there are differences between degrees of affirmation as well as different points of view from which an action may be looked at. One can tell more or less what one tells and one can tell it according to one point of view or another; this capacity and the modalities of its use are what Genette's category of narrative mood concerns. Thus the chief sub-categories of mood are distance and perspective. The narrative can furnish more or fewer details, rendered in a more or less direct way, and can thus seem to keep at a greater or lesser distance from what it tells. That distance need not be fixed throughout a narrative but may vary according to the knowledge of one or more participants in the story, adopting what is usually called their point of view. Thus the narrative seems to take on, with regard to the story, one or another perspective.

Genette's discussion of distance entails an extended contrast between mimesis and diegesis. Mimesis is direct imitation, as are words and actions on a stage. Written narration cannot be mimetic since it presents words and actions described by an intermediary—the author or, more correctly, the narrative discourse itself. An exception is direct speech, presented in the manner of drama, in which the mediation of the narrator seems to disappear. Genette cleverly shows that the interior monologue, which he prefers to call immediate speech, pushes mimesis to its limit by "obliterating the last traces of the narrating instance and giving the floor to the character right away." Some analysts have called cinema or particular films or passages mimetic, others have called them diegetic; sometimes the terms are used more or less interchangeably. Genette's definitions might make possible a new clarity regarding cinematic instances, although, again, cinema is a more difficult case because it is a "complex system." Films usually present the direct imitation of speech and action but do so in a mediated, or diegetic way. Films may combine mimetic and diegetic ele-

ments in a large variety of ways and may also oppose them through slight discrepancies or gaps, as in Bresson's Diary of a Country Priest (1951).

Narrative perspective is a second mode of regulating information, turning on the use or absence of a restrictive point of view. Genette confronts at the outset a widespread confusion between mood and voice, between the question: Whose point of view orients the narrative perspective? and the question: Who is the narrator? Or, more simply, between "Who sees?" and "Who speaks?" Genette's clarification of this large, muddied realm is, with his original discussion of frequency, his most conspicuous achievement.

Genette prefers the term "focus of narration" or "focalization" to "point of view." Ishmael and Strether occupy the same focal positions in Moby-Dick and The Ambassadors , respectively. That Ishmael also narrates while an author absent from the story narrates The Ambassadors is a question of voice, not of mood. Wayne Booth's implied author and narrator, the narrator who is dramatized or undramatized, reliable or unreliable, are all categories of voice.

Genette distinguishes nonfocalized narratives, generally the classical narrative, from internally and externally focalized narratives. Internal focalizations are fixed if we rarely leave the point of view of one character; variable if we go back and forth between more than one character; or multiple if the same event is evoked several times according to the point of view of different characters, e.g., The Sound and the Fury , certain epistolary novels, or Rashomon . In external focalization (some of Hammett and Hemingway), the hero performs in front of us without our knowing his thoughts or feelings. A focalization is not necessarily steady; a work may combine many types of focalization, even all of them, although particular passages are usually focalized in one way rather than another. The division between variable focalization and nonfocalization is sometimes difficult to establish, since the nonfocalized narrative can most often be analyzed as a narrative that is multifocalized ad libitum .

Older Huw is the narrator of How Green Was My Valley —it is literally his voice that we hear; but this does not control the question of mood. We almost never see things from young Huw's visual perspective. How Green is nonfocalized in that no sequence, let alone the whole film, is shot from any one character's perspective. It is variably focalized in that it frequently borrows a character's perspective for one or more shots—sometimes for dramatic reasons, sometimes opportunistically because it yields a stunning image or an efficient view of the action. Huw's arrival at school is shot in a non-focalized way; he is small and timid in the hallway but no character is looking at him in this way. When Huw opens the door, Ford cuts to his point of view of the girls in the class looking hostile, then to the boys in the class, farther away, looking even more hostile. (Both shots cannot be from Huw's perspective, strictly speaking.) Ford then cuts to a shot of the teacher from Huw's point of view

but when he is summoned to the front of the room, his point of view is dropped in favor of, more or less, the point of view of the class. Later on, a few shots of Huw's fight in the school yard are taken from the perspective of a girl who is sympathetic to Huw but not otherwise an important character. Ford gives us a few shots from Huw's perspective to register the first impact of school, then proceeds in a nonfocalized or variably focalized way to present the events that happen there.

Several of the film's shots are classical Griffith-Ford iconic images, for some of which, anyway, there can be no character perspective, unless it be that of God or America. Such shots include those of the entire family assembled, for which there can be no character perspective because there is none who is not part of the scene. Many of the final memory images in How Green , though not of the entire family, are also of this kind.

Genette's account of voice is particularly long and complex; our remarks will be partial and abbreviated. Genette equates voice with "the narrating instance," that is, with the presence of the narrator in the story he tells. The narrator is not the author and not a character. Even where the narrating is identified with a character, the instances are distinct. The elements of the narrating situation include "time of the narrating," "narrative level," and "person."

The tense structure of language requires a narrator to situate himself temporally in relation to the story he tells—but not spatially. Narrators almost never describe the place where they are narrating; moreover, the narration itself, as opposed to the story, has no apparent duration. In the classical novel, in Proust, "narrating involves an instantaneous action."

Voice-over narrators in cinema are generally shown: Joe Gillis floating dead in a pool in Sunset Boulevard; Walter Neff addressing the dictaphone in Double Indemnity; Addison DeWitt and Karen Richards sitting at the banquet table in All About Eve . In How Green we see the older Huw pack his things in order to leave the valley but his face is hidden, as though sharing the shame of the besmirched valley and the degraded present. We do not see the writer of A Letter to Three Wives (1949), who narrates the film; a more remarkable transgression of this sort is Le Plaisir (1951) by Max Ophuls. Over a blank screen, the author, presumably Maupassant (voice of Peter Ustinov), speaks to us:

I have always loved the night, the hours of darkness. That's why I am so grateful to be able to speak to you in the dark. They wanted to photograph me; after all, this is a photographic medium, can one say that? But that I didn't allow. An author's pleasure is to be heard, not seen. I thought the best thing might be if I just told you these stories myself, as if I were sitting beside you and, well, who knows, maybe I am. I will try to speak English, but I have not had as much practice as I would like.

Perhaps you can guess my anxiety. These tales are rather old and you are so very modern, as we all call ourselves while we are still alive. Anyway, be patient with me. Here's the first story.

The instances mentioned all realize that instantaneity of narration of which Genette speaks: Huw packing his things, DeWitt and Richards at the banquet table, and Joe in the pool remember it all and speak it all in a flash. Not Reconciled does something different in this respect also. Robert Fähmel, the narrator of the two-part analepsis about the thirties, pauses in the middle of his story for a cognac and the film devotes five shots to the fetching of the cognac. This includes two brief flashbacks concerning the duties of Hugo, Robert's listener, in the hotel. In Straub, even the listener has his own material situation and even listening must be produced; narrating and listening usually take place in an idealist utopia, free of physical and economic constraints.

Genette on person rejects the distinction between "first-person" and "third-person" narratives. Every narrating is, by definition, presented in the first person; this is the only way a narrator can be in his narrative. The real question is whether or not the narrator will use the first person to designate one of his characters; but even when he does, the narrating instance and the character remain distinct; they do not merge. For one thing, the narrator almost always knows more than the hero, even when he is himself the hero. "For the narrator focalization through the hero is a restriction of field just as artificial in the first person as in the third." Marcel the narrator knows everything that will happen to Marcel the character from the moment he begins to narrate. "Between the information of the hero and the omniscience of the novelist is the information of the narrator, who disposes of it according to his own lights and holds it back only when he sees a precise reason for doing so." Advance notices in a narrative, and more generally all prolepses, cannot be the hero's doing, they must be the narrator's.

Thus older Huw knows in advance everything that will happen to the Morgans in the film and withholds information in order to reveal it gradually, as do DeWitt and Richards in Eve and other narrators in their films. This follows from the nature of the narrating instance, but something is wrong with this statement. Maybe we do not believe that Huw, DeWitt and Richards, etc., really narrate their entire stories. There is first the fact that many important scenes in each film could not be known by their narrators: the wedding party for Bronwen and Ivor, the scenes between Angharad and Mr. Gryffyd, the town gossip scenes, and others in How Green; the intimate scenes between Margo and Bill and between Lloyd and Eve, all the scenes of Eve alone, and others in Eve. Genette defines as "alterations" instances of narrators giving more information (or less) than is within their competence, saying that such

momentary infractions of the code governing the narrational context may not call the code itself into question. But these instances go well beyond this definition and make us question what the code of the voice-over narrator in classical films is.

There is also the fact that the narrators of How Green, Eve , and other films are actually silent for most of the film. In both instances, there is a good deal of narration at the beginning of the film in order to set the scene and to introduce the main characters; once that is done, the film proceeds in usual scenic fashion, the voice-over brought back from time to time to provide narrative summaries and to bridge ellipses in the long period of time covered by each film. (This is a long year in the case of How Green; a theatrical season, October to June, in the case of Eve .) The conclusion seems unavoidable that voice-over narration in classical cinema has little in common with character narration in fiction; it is a narrative convenience used and dropped by the film to suit its purposes. Character narration in fiction is often consistent for the whole text; if a character's voice is dropped, it is usually in favor of another character's voice. There are switches from subjective voice to objective voice, as in the final section of The Sound and the Fury , but there is usually not then a return to subjective voice. The voice-over in film is usually picked up and dropped by the film at will, according to no principle but expediency. Above all, of course, voice-over narration in cinema does not comprise the whole text. It does not pervade the text or inflect every detail of it, color it, shape it, lend it its entire flavor, as character narration in fiction does. In the "complex system" of cinema, it is just one element among many elements, to be juggled along with them, often in shifting combinations.

A closer look at the functions of the narrators in How Green and Eve will confirm and amplify these points, since their patterns are typical of voice-over narration in classical cinema generally. Both films begin with long voice-over narrations at the beginning of the film: to describe the setting and general situation and to introduce the characters by name, occupation, and relation to the other characters, as the camera shows us their faces and their physical environment. Both films have an unusually large number of important characters and both attempt to situate them, complexly and fully, in a distinctive milieu, a world. To attempt this through usual, objective exposition seems difficult, if not impossible—though that may be because the devices work so well and are used so skillfully in these two films. It seems undeniable, however, that the convention of the voice-over made possible more complex, "novelistic" projects than formerly.

The function of the voice-over narration in How Green and Eve is not only to introduce a complex situation but to bridge the transition of the narrative from

All About Eve

present to past, a shift that will hold good for most or all of the rest of the film. (All of it in How Green ; all but the final scene in Eve , which is a kind of epilogue and a looping of the narrative, oddly parallel to the looping of visuals at the end of How Green .) If we recall the crisis of order posed to classical cinema by an analepsis, we appreciate the importance of this function. One wonders, indeed, if bridging the transition from present to past may not be the primary function of voice-overs in the classical period; that is, whether "voice" is not determined by "order" in classical narrative films, at least in part.

Often, as in How Green and Eve , a voice-over will continue for a certain duration after the bridge from present to past: to insure continuity, to provide orientation and exposition within the world of the flashback, and to reintroduce characters and plot situation within the flashback. Then, when the analepsis is well launched, the film characteristically drops the voice-over and proceeds for the main part of the film in an objective way. About two-thirds of Huw's total narration in How Green is spoken in the opening sequence of the film, until the formal introduction of Bronwen to the family, a duration of about ten minutes. The first 14 of the 25 voice-over speeches in Eve occur in the opening sequence of the film, until we see Eve in the theater alley as she approaches Karen Richards. Also, in both films, the voice-over recedes just before the first big, objective scene, in both cases a vivid set-piece serving to make us forget the introductory voice and to launch the world of the flashback on its own: the wedding of Bronwen and Ivor and the party that follows in How Green ; the introduction of Eve to Margo and company in Eve . Thereafter, in both films, the voice-over returns occasionally to provide brief narrative summaries and to bridge

transitions within the analepsis. This happens 8 times in How Green , 11 times in Eve; for example:

Such transition or scene-setting speeches turn their narrators into speaking title cards, pragmatic in the case of How Green , ornate and witty in the case of Eve . But what a comedown from the high-flown openings of both narrations, which even lead us to believe that the films might be about their speakers! Instead Huw is a bystander in most of the scenes in which he appears and is missing altogether from many others. Addison is a bystander at the party scene and others and is missing altogether from even more scenes. Interestingly, each film gives its forgotten narrator one important scene or sequence later: the school sequence for Huw in How Green ; Addison's "killer to killer" confrontation with Eve in Eve .

The voice-over narrator in classical film is a puppet of the narration. One might say the same of character-narrators in fiction, but they at least are always onstage and are built to withstand observation; they also serve all the functions of the narration. The puppet narrators of cinema are jerked on and off stage in a manner that is quite undignified. They may have integrity as characters but they have no integrity as narrators, no resistance to the demands placed upon them; they are nothing but the functions they serve, a collection of odd jobs. They are reeds buffeted by the winds of the narrative; their voices resonate but they are hollow. An important part of their function is to mimic the novelistic voice, the guarantor of experience, subjectivity, and seriousness, with its "I remember" and "I suffer." In fact these inflated "I"s of classical cinema "open" for the main story; they are second-rate acts that

warm up the audience with their sonorities. They are ludicrous stand-ins for the novelistic "I" because, though they wind up elaborately, they have nothing to pitch. For precisely these reasons, of course, voice-over narrators in cinema are interesting; they ruthlessly expose the mechanisms of narration in classical cinema.

This is not to disparage the convention of the voice-over in cinema, which has figured in so many excellent films. It is rather to indicate that if we seek to define voice in cinema, in Genette's sense, we will have to look beyond these voices.