The Classical Hollywood Cinema - Cap 5

The Classical Hollywood Cinema. Film Style and Mode of Production to 1960
David Bordwell, Janet Staiger and Kristin Thompson
5

Space in the classical film
The motion picture industry for many years has been trying to remove the one dimension of the screen. By lighting, with lenses of inexplicable complexity, through movement, camera angles, and a variety of other techniques, the flatness of the screen has largely been overcome. 1
Ranald MacDougall, 1945
In making narrative causality the dominant system in the film's total form, the classical Hollywood cinema chooses to subordinate space. Most obviously, the classical style makes the sheerly graphic space of the film image a vehicle for narrative. We can see this principle at work negatively in the prohibitions against 'bad' cuts. 'The important subjects should be in the same general area of the frame for each of the two shots which are to be cut together, ' but 'as long as the important subject is not shifted from one side of the screen to the other, no real harm is done.' 2 In describing the classical cinema's use of space we are most inclined to use the term 'transparent, ' so much does that cinema strive to efface the picture plane. 'The screen might be likened to a plate-glass window through which the observer looks with one eye at the actual scene.' 3 We need, however, a fuller account of how classical narration uses image composition and editing to create a powerful representation of three-dimensional space.

The image: composition
While recognizing that Hollywood cinema subordinates space to narrative causality, we ought also to acknowledge that the classical spatial system is, in a strictly logical sense, arbitrary. We could imagine other systems that privileged different devices (e.g., decentered framings, discontinuity editing) but which were equally coherent and equally supportive of causality. Historically, however, the classical construction of space appears far from arbitrary, since it synthesizes many traditions which have dominated various Western arts.
Post-Renaissance painting provided one powerful model. Cinematographers and directors constantly invoked famous paintings as sources. Cecil B. De Mille claimed to have borrowed from Doré, Van Dyck, Corot and one 'Reubens.' 4 Robert Surtees cited the Impressionists, Leon Shamroy imitated Van Gogh. Discussions of lighting invariably invoke Rembrandt. 5 To a point, such assertions are simply hyperbole. Allan Dwan remarked: 'Once in a while we would undertake the imitation or reproduction of something artistic-a famous painting, let's say.' 6 (Staged replicas of famous pictures were also a convention of theatrical melodrama.) But in a more significant sense, Hollywood did perpetuate many precepts of post-Renaissance painting. The very name 'film studio' derives from the term for the workroom of the painter or sculptor. While no major cinematographers were professional painters, many (Charles Rosher, Karl Struss, Stanley Cortez, James Wong Howe) had been portrait photographers, a field in which academic rules of composition and lighting prevailed. And occasionally a cinematographer would articulate principles of filmmaking that directly echo those of academic painting. 7 We ought not to be surprised, then, that Hollywood's practices of composition continue some very old traditions in the visual arts.
An outstanding example is the Hollywood cinema's interest in centered compositions. In post-Renaissance painting, the erect human body provides one major standard of framing, with the face usually occupying the upper portion of the
-50
picture format. The same impulse can be seen in the principle of horizon-line isocephaly, which guarantees that figures' heads run along a more or less horizontal line. 8 Classical cinema employs these precepts. While extreme long shots tend to weight the lower half of the image (this derives from landscape painting traditions), most shots work with a privileged zone of screen space resembling a T: the upper one-third and the central vertical third of the screen constitute the 'center' of the shot. This center determines the composition of long shots, medium shots, and close-ups, as well as the grouping of figures (see figs 5.1 through 5.8). In widescreen films, the center area is proportionately stretched, so even slightly off-center compositions are not transgressive (especially in a balanced shot/reverse-shot cutting pattern). Classical filmmaking thus considers edge-framing taboo; frontally positioned figures or objects, however unimportant, are seldom sliced off by either vertical edge. And, as the illustrations indicate, horizon-line isocephaly is common in classical filmmaking. Thus the human body is made the center of narrative and graphic interest: the closer the shot, the greater the demand for centering.
But how to center moving figures? The classical style quickly discovered the virtues of panning and tilting the camera. The subtlest refinement of this practice was the custom of reframing. A refraining is a slight pan or tilt to accommodate figure movement. Every film in the UnS contained some reframings; after 1929, one out of every six shots used at least one reframing. The chief alternative to reframing is what Edward Branigan has called the frame cut. 9 Within a defined locale, a figure leaves the shot, and, as the body crosses the frame line, the cut reveals the figure entering a new shot, with the body still crossing the (opposite) frame line (see figs 5.9 through 5.14). Frame-cutting is extraordinarily common in classical cinema, partly because it is the least troublesome match-on-action cut to make but also because it confirms the importance of the center zone of the screen. In a frame cut, the image's edge becomes only a bridge over which figures or objects pass on their way to center stage.
With centering comes balance, but the complex and dynamic equilibrium of great Western painting is usually lacking in Hollywood compositions. Overall balance and an avoidance of distractingly perfect symmetry generally suffice. Once centered, the human body provides enough slight asymmetries to yield a generally stable image, and camera viewfinders, engraved with cross-hatchings, enabled cameramen to balance the shot. When balance is lost, the results leap to the eye. In figures 5.15 and 5.16, from The Bedroom Window (1924), William C.deMille's practice of multiple-camera shooting has pushed the shots off-center and off-balance. Of course, such imbalance can be causally motivated, as in Harvey (1950), for which cinematographer William Daniels had to frame the shots asymmetrically to include the invisible rabbit. 10 The value of balance in the classical cinema can be seen in the way that a vacancy in the frame space will be reserved for the entry of a character; that figure will complete the balanced composition (see figs 5.17 through 5.19).
Both centering and balancing function as narration in that these film techniques shape the story action for the spectator. The narrational qualities of shot composition are also evident in the classical use of frontality. Renaissance painting derived many principles of scenography from Greek and Roman theater, so that the idea of a narrative action addressed to the spectator became explicit in Western painting. The classical film image relies upon such a conception of frontality. The face is positioned in full, three-quarter, or profile view; the body typically in full or threequarter view. The result is an odd rubbernecking characteristic of Hollywood character position; people's heads may face one another in profile but their bodies do not (see figs 5.20 and 5.21). Standing groups are arranged along horizontal or diagonal lines or in half-circles; people seldom close ranks as they would in real life (see figs 5.22 and 5.23). The dyspeptic Welford Beaton was one of the few critics who noticed this practice: 11
In most of our pictures the directors make their characters face the camera by the simple expedient of turning them around until they face it, no matter how unnatural the scene is made thereby. In Gentlemen Prefer Blondes [1928], there is an exhibition of flagrant disregard of common sense in grouping characters. Ruth Taylor, Alice White, and Ford Sterling are shown seated at a round table in a
-51-
restaurant. Instead of forming a triangle, they are squeezed together so closely that Sterling, in the center, scarcely can move.
Yet complete frontality-e.g., direct address to the camera-is rare; a modified frontality requires that a wedge be driven into the space, opening up the best sightlines.
Frontality constitutes a very important cue for the viewer. When characters have their backs to us, it is usually an index of their relative unimportance at the moment. George Cukor points out a scene from Adam's Rib (1949) in which Katharine Hepburn was turned from the camera: 'That had a meaning: she indicated to the audience that they should look at Judy Holliday.' 12 Groupings around tables often sacrifice a good view of the least significant character in the scene. One UnS film, * Saratoga (1937) vividly illustrates how troubled the film's space becomes when frontality is disrupted. Jean Harlow died in the course of the film's production, before several scenes were shot. In those scenes, Harlow was replaced by a double who never faces the camera, resulting in the odd phenomenon of having no portrayal of the heroine's expressions during climactic moments of the action.
Most important, frontality can be lost if it is then regained. Over-the-shoulder shot/reverseshot cutting decenters a figure and puts his or her back to us, but the reverse shot reinstates that character front and center. Once the figures are arranged for us in the image, editing can introduce new angles, but then closer shots will typically be centered, balanced, and frontal in their turn. Even if one minimizes editing, as Orson Welles and William Wyler are often thought to do, the deep-focus composition cannot forfeit frontality-indeed, in films like The Magnificent Ambersons (1942) and The Little Foxes (1941), classical frontality is in fact exaggerated (see figs 5.24 and 5.25).
The most obvious way that the classical cinema works to treat the screen as a plate-glass window is in the representation of depth. Probably the most important depth cue in cinema is movement. When a figure moves and creates a continuous stream of overlapping planes and receding shapes, when the camera glides through or across a space-under these circumstances it becomes very difficult to see the screen as a flat surface. This is perhaps one of the reasons that modernist and avant-garde films have often suppressed the kinetic depth effect by such devices as flicker, still images, and graininess.
Classical Hollywood space is created in planes through various depth cues. To the usual cues of visual overlap (the object that overlaps must be closer) and familiar size, the classical image adds pattern, color, texture, lighting, and focus to specify depth. Geometrical patterns and colors, especially of costumes, stand out from plainer backgrounds (see figs 5.26 and 5.27). Even in black-and-white filming, set designers painted sets in different colors to create planes in depth. 13 More dense and concentrated textures were reserved for the figures in the foreground, and cinematographers would diffuse the light on backgrounds to make them more granular. Lighting is particularly important in establishing depth. Cinematographers were careful to alternate planes in contrasting keys and half-tones (a silhouetted foreground, a bright middle ground, a darker background). 14 Hollywood's standardized three-point lighting system (key, fill, and backlighting), supplemented by background lighting, eye lights, and other techniques, had as its effect the careful articulation of each narratively relevant plane. The importance of backlighting cannot be overestimated here. Commonly thought of as a Griffith cliché or a sudden lyrical effect, backlighting is in fact one of the most common ways the Hollywood filmmaker distinguishes figure from background: A pencil-line of light around the body's contour pulls the figure forward (see figs 5.28 and 5.29). 15 Edge lighting of figures remained common even after fast film stocks and color films enhanced figure separation (see fig 5.30). Low-key lighting could be very effective in picking out planes if edge-lighting supplemented it (see fig 5.31). Finally, the planes of the classical image also usually get defined by selective focus, an equivalent of aerial perspective in painting. In framings closer than medium shot, the characters are in focus while other planes are not. 16 Variations are possible-in deep-space compositions, a figure in the foreground might be out of focus while another in the background is in focus-but the principle generally holds good. No classical films throw figures out of focus to favor insignificant objects (kegs, stoves) in the manner
-52-
of Ozu's films or of certain avant-garde works. 17
Stacked planes are not enough; the classical style stresses volumes as well. Cinematographers valued 'roundness' as much as depth, using highlights to accentuate curves of face and body or to pick out folds in drapery. 18 As early as 1926, the cinematographer was compared to the sculptor: 19
It is chiefly by the use of such lighting equipment that the sculptor-director seeks his worshipped 'plasticity.' Failing a true stereoscopic effect in film, he models his figures to a roundness with lights behind and above and on either side, softening here and sharpening up for accent elsewhere with a patience and skill inevitably lost on the layman.
Make-up was designed to enhance the roundness of faces. Likewise, a set had to be represented as a volume, a container for action, not a row of sliced planes. Designers often built three-dimensional models of sets in order to try out various camera positions. Even the ceiling, which usually could not be shown, had to be implied through shadow. 20 Camera movement could endow the set with a sculptural quality too, as Dwan observed: 'In dollying as a rule we find it's a good idea to pass things in order to get the effect of movement. We always noticed that if we dollied past a tree, it became solid and round, instead of flat.' 21
The importance of planes and volumes in defining classical scenographie depth makes academic perspective rather rare. Developed during the Renaissance as a revision of ancient Greek perspective, central linear perspective organizes planes around the presumed vantage point of a stationary monocular observer. The impression of depth results from the assumption that parallel lines receding from the picture surface seem to meet at a single point on the horizon, the vanishing point. 22 Now it is indisputable that certain aspects of Hollywood film production, such as set design and special-effects work, frequently draw upon principles of linear perspective. 23 But images in the Hollywood cinema seldom exhibit the central vanishing point, raked and checkered floorplans, and regular recession of planes characteristic of what Pierre Francastel calls the 'Quattrocento cube.' 24 (Such conventions are far more common in pre-classical films; see fig 5.32.) The classical shot is more usually built out of a few planes placed against a distant background plane-in a long shot, the horizon; in a closer view, the rear wall of a room (see figs 5.33 and 5.34). A limited linear perspective view can be supplied by the corner of a room or ceiling or the view out of a window. Sometimes, especially in 1940s films, a more explicit sense of perspective emerges; an occasional establishing shot exhibits a deep recessional interior (see fig 5.35) or a skewed vanishing point (see fig 5.36). But in medium-long and medium shots (the majority of the shots in a film), linear perspective remains of little importance, and pronounced depth is achieved by interposing figures and objects on various planes.
Such art-historical traditions would not seem easily applicable to the scenographie space constructed by the soundtrack. But the classical cinema modeled its use of sound upon its use of images. (Chapter 23 examines how this occurred historically.) As one technician wrote: 25
With the two-dimensional camera, which bears the same psychological relation to the eye as monaural sound does to the ear, the illusion of depth can be achieved by the proper use of lighting and contrast, just as by the manipulations of loudness and reverberation with the microphone. And just as the eye can be drawn to particular persons or objects by the adjustment of focal length, so can the ear be arrested by the intensification of important sounds and the rejection of unimportant ones.
What Hollywood technicians called 'sound perspective' was the belief that the acoustic qualities of dialogue and noise had to match the scale of the image. Engineers debated how to convey 'natural' sound while granting that strictly realistic sound recording was unsuitable. Microphones had to be rotated in the course of conversations; musical numbers had to be prerecorded; some dialogue had to be post-synchronized; and, most importantly, sounds had to be segregated onto separate tracks for later mixing. In the theater, the speakers were placed behind the screen, as centered as were the figures in the frame. The same conceptions of balance, centrality, and spatial definition were applied to stereophonic sound in the early 1950s. 26
-53-
Thus in the Hollywood cinema the space constructed by the soundtrack is no less artificial than that of the image. Alan Williams points out that like visual perspective, sonic perspective is narrational, yielding not 'the full, material context of everyday vision or hearing, but the signs of such a physical situation.' 27 He shows how selective the sonic space of a Hollywood locale is in comparison with that of the racketfilled café in Godard's Two or Three Things I Know About Her (1966). Similar effects occur in the dense, layered montage of offscreen sound in Rainer Werner Fassbinder's Third Generation (1980) and In a Year of Thirteen Moons (1980), during which radios, television sets, and several conversations compete for our attention. In this sense, classical sound technique articulates foreground (principal voice) and background (silence, 'background' noise, music 'under' the action) with the same precision that camera and staging distinguish visual planes.
Centering, balancing, frontality, and depth-all these narrational strategies-encourage us to read filmic space as story space. Since the classical narrative depends upon psychological causality, we can think of these strategies as aiming to personalize space. Surroundings become significant partly for their ability to dramatize individuality. Hence the importance of doors: the doorway becomes a privileged zone of human action, promising movement, encounters, confrontations, and conclusions. The classical film also charges objects with personal meanings. Props (guns, rings, etc.), and especially representational props (photographs, dolls, portrait paintings) all bear an ineluctable psychological import. (How many classical films convey a lover's disgust by violence against the picture of the beloved.) Shot scale is also geared to expressivity, with the plan américain (the knees-up shot) and the medium shot the most common ones because they 'retain facial expressions and physical gestures-partially lost in the long shot-and relate these, dramatically, to the action involved' 28 A close-up, which can theoretically show anything, becomes virtually synonymous with the facial close-up, the portrait that reveals character. It is significant, however, that extreme facial close-ups-framings closer than full facial shots-are almost absent from the classical cinema, as if cutting the face completely free of the background made the close-up too fragmentary. (Compare the frequency of enlarged portions of faces in the Soviet cinema of the 1920s.) Lighting brings out the personality of the character, while diffusion distinguishes women by spiritualizing them. 29 In the sound cinema, the voice parallels the face as a vehicle of personalization. In all these ways, the classical cinema declares its anthropocentric commitment: Space will signify chiefly in relation to psychological causality.
Classical narration of space thus aims at orientation: The scenography is addressed to the viewer. Can we then say that a larger principle of 'perspective' operates here-not the adherence to a particular spatial composition but a general 'placing' of the spectator in an ideal position of intelligibility? 30 Certainly Hollywood's own description of its work emphasizes the camera as an invisible witness, just as the soundtrack constitutes an ideal hearing of the scene. This aesthetic of effaced present is anthropocentric (camera and sound as eye and ear) and idealist (the witness is immaterial, an omniscient subject), hence also ideological. Yet the viewer is not wholly a passive subject tyrannized by a rigid address. Analogies with perspective, being spatial, tend to neglect the spectator's activities. Just as the viewer must meet causal and temporal systems halfway, the viewer must contribute something in order to make classical space work. That contribution includes the sort of hypothesisforming and -testing that I have emphasized in earlier chapters. That we tend to anticipate data, that we frame our hunches as more or less likely alternatives (or paradigmatic choices), that we retroactively check our hypotheses-all these activities operate in our construction of classical space.
So, for instance, centering procedures quickly lead the viewer to perform certain operations. Confining significant narrative action to any constant zone of screen space effectively insures that attention paid to other areas will not be rewarded. Moreover, psychologists have long known that it is hard to read a configuration as three-dimensional if we are markedly aware of the edges of the image: our eye tests for consistency, and the depth of the represented space conflicts with the boundary of the picture. 31 Centered film compositions, either static or
-54-
moving, draw our attention away from the frame edge. Even the viewing situation encourages this, since black masking on the theater screen conceals the aperture line. Cinematographers often darkened the edges of the image to avoid a glaring contrast between the picture and the theater masking. 32 Distracting our attention from the edge thus discourages us from testing the image as a flat space. Compare, however, the flattening effect of edge-framed compositions in non-Hollywood traditions (see fig 5.37).
Similarly, frontality functions as a strong cue for the spectator. Since the classical Hollywood cinema is predominantly anthropocentric, the representation of the expressive body arouses in us an interest nourished not only by art but by everyday life. Our principal information about people's mental states is derived in large part from posture, gesture, facial expression, and eye movement (as well as voice), so that if classical cinema is to represent psychological causation in its characters, narrational space must privilege these behavioral cues. Moreover, as Gombrich points out, some objects give a more exact feeling of frontality than do others. We are remarkably sensitive to anglings of body, face, and especially eyes, and we tend to orient ourselves to postures and gazes with a precision that we do not apply to walls or trees. 33 In addition, of course, 'normal' camera height, standardized at between 5 and 6 feet, corresponds to a gaze from an erect human body, a position canonized not only in art but also in culture generally. 34 Imagine a classical film with only one difference: it is entirely shot from straight above the characters. The consistent bird's-eye view would destroy the expressive basis of the narrative because the classical filmmaker lacks schemata for rendering such an orientation and the film viewer has no appropriate repertoire of expectations.
And what of the spectator's construction of depth? The various depth cues, most prominently movement, require an act of spatial integration on the viewer's part. If classical space does not pose the visual paradoxes of images in some German Expressionistic cinema or in abstract film, that is partly because we scale our expectations to a limited set of possibilities. But consider the baffling space of figure 5.38, from Griffith's Trying to Get Arrested (1909). A tiny man runs in at the lower left corner. The cue of familiar size dictates that he looks small because he is far away, but the receding planes of the shot seem to deny this. Is the man then a leprechaun? No, he is indeed in the distance, as a later frame (fig 5.39) makes clear. The peculiarity of this primitive shot arises from the way the image foils those expectations about planes and volumes that the classical cinema would have confirmed by composition and framing. Certainly seeing an image as deep is 'easier' in cinema than in other arts, but even film depth must be achieved to some degree, relying upon what Gombrich has called 'the beholder's share.' 35

Continuity editing
Theorists are still a long way from fully understanding how the viewer contributes to the creation of classical space, but some consideration of the process of editing may help. Certainly editing can work against the orientation achieved within the image, as it does in the films of Eisenstein, Ozu, Nagisa Oshima, Godard, and other filmmakers. 36 Classical continuity editing, however, reinforces spatial orientation. Continuity of graphic qualities can invite us to look through the 'plate-glass window' of the screen. From shot to shot, tonality, movement, and the center of compositional interest shift enough to be distinguishable but not enough to be disturbing. Editors seldom discussed graphic continuity, but the procedure was explained as early as 1928 by two visitors to the Hollywood studios, who claimed that either the point of interest in shot B should be on the screen 'almost' where the point of interest of shot A ended, or B should continue A's movement: 37
This has no reference to the story itself, but merely to the making of the pictures considered only as spots of colour and centres of pictorial interest. The eye should be led a gentle dance, swaying easily and comfortably from side to side of the picture, now fast, now slow, as the emotional needs of the story demand.
Compare the graphically gentle cut of the typical shot/re verse-shot series, which only slightly shifts the center of interest (see figs 5.40 through 5.43) with the graphically jarring cut which alters that
-55-
center of interest quite drastically (see figs 5.44 and 5.45).
Once graphic continuity is achieved, the editing can concentrate upon orienting us to scenographie space. Crosscutting creates a fictive space built out of several locales. As Chapter 4 points out, classical crosscutting presupposes that shifts in the locale are motivated by the story action. More often, editing fulfills the narrational function of orienting us to a single locale (a room, a stretch of sidewalk, the cab of a truck) or to physically adjacent locales (a room and a hallway, the rear of the truck). Thus the principles and devices of continuity editing function to represent space for the sake of the story.
André Bazin has summarized the basic premises of classical continuity editing: 38

1 The verisimilitude of the space in which the position of the actor is always determined, even when a close-up eliminates the decor.


2 The purpose and the effects of the cut are exclusively dramatic or psychological.
In other words, if the scene were played on a stage and seen from a seat in the orchestra, it would have the same meaning, the episode would continue to exist objectively. The changes of point of view provided by the camera would add nothing. They would present the reality a little more forcefully, first by allowing a better view and then by putting the emphasis where it belongs.
Besides spelling out the classical assumptions about consistent spatial relations and the determining role of character psychology, Bazin reveals the extent to which classical editing continues and elaborates the scenography of nineteenth-century bourgeois theater. Bazin's mobile-yet-stationary spectator in the orchestra personifies the viewpoint created by the classical '180°' or 'axis-of-action' system of spatial editing. The assumption is that shots will be filmed and cut together so as to position the spectator always on the same side of the story action. Bazin suggests that the 'objective' reality of the action independent of the act of filming is analogous to that stable space of proscenium theatrical representation, in which the spectator is always positioned beyond the fourth wall. The axis of action (or center line) becomes the imaginary vector of movements, character positions, and glances in the scene, and ideally the camera should not stray over the axis. In any scene, explains Robert Aldrich, 'You have to draw the center line…. You must never cross the line.' 39 If we assume that two conversing characters are angled somewhat frontally (as is usual), the classic 180° system will be as laid out in diagram 5.1. Camera positions A, B, C, and D (and indeed any position within the lower half-circle) will cut together so as to orient the viewer, while camera position X (or any position on the other side of the center line) is thought to disorient the spectator.

The 180° principle governs all the more specific devices of continuity editing. Analytical editing moves the spectator into or back from a part of a total space. A cut from position A to position B (or vice versa) would be an analytical cut, respecting the axis of action. Shot/reverse-shot cutting assumes that the series of shots alternates a view of one end-point of the line with a view of the other. Thus cutting from camera position C to that of D would be a shot/reverse-shot pattern. Typically, shot/reverse-shot editing joins shots of characters facing one another, but it need not.
-56-
The same principle applies to vehicles, buildings, or any entities posited as being at opposite ends of the axis of action. Eyeline-match cutting uses character glance as a cue to link shots. The assumption is that the eyeline runs parallel to the axis, so the camera positions will remain on one side of the line. Shots C and D when cut together will yield correct eyeline matches in a way that, say, shots X and D would not. A comparatively uncommon case of eyeline-match cutting, point-of-view cutting, reveals the limits of permissibility in the 180° system. The first shot shows the character looking at something offscreen; the second shot shows what the character is seeing, but more or less from the character's optical vantage point. Remarkably, critics continue to reduce shot/reverse-shot cutting to point-of-view cutting. A recent monograph defines shot/reverse shot in a conversation scene as taking the second shot 'from the first character's point-of-view.' 40 Hollywood shot/reverse- shot cutting is more properly what Jean Mitry calls semi-subjective: we are often literally looking over a character's shoulder. 41 (Edward Branigan has shown that camera angle is the critical variable here: camera distance is often inexact in classical point-of-view cutting. 42 ) But even the point-of-view shot remains within the 180° convention because it represents a camera position on the axis itself (e.g., position E on the diagram). The power of the 180° system may also be seen in what we may call the 'earline-match' cut, in which a character listens from outside the space of the scene. The assumption is that the sound travels in a straight line, which constitutes the axis of action. If a listener at a door cocks his ear to screen left, a cut to someone inside the room walking to that door must show the character moving screen right.
Obviously, across a series of shots all these editing devices work smoothly to reinforce each other, so that an establishing shot will be linked by an analytical cut to a closer view, and then a series of shot/reverse shots will follow. But the system, being part of a stylistic paradigm, has a certain latitude as well, so that one can use the shot/reverse-shot schema if one character has turned his back to the other, if there are five or six characters present, and so on.
One more device of the 180° system deserves mention, not least because it dramatizes the extent to which the system defines a coherent but limited field for the spectator. Editing for directional continuity translates the imaginary line into a vector of movement. If a character or vehicle is moving left to right in shot 1, it should continue to do so in shot 2. Directional continuity cutting is like eyeline cutting: just as two shots of figures looking in opposite directions imply that the figures are looking at each other, so two shots of figures moving in opposite directions lead us to expect the figures to meet. Directional continuity also resembles point-of-view cutting in that one can show the movement from a position on the axis of action-i.e., either a heads-on or a tails-on shot of the action. (A shot from this position can function as a transition if one wants to cross the line.) Directional continuity is often used within a circumscribed space, as when a character goes from the window (exit frame left) and comes to the desk (enter frame right). In these cases, Hollywood directional continuity depends upon the frame cut. What is more revealing, though, is that directional continuity can be maintained across separate spaces, for in that case the 180° system presupposes that the ideal spectator is situated on one side of an axis perhaps miles long! The closed chamber-space of the theater has been left behind, but Bazin's spectator-in-the-orchestra and his or her relation to proscenium space remain intact.
The devices of continuity editing are best seen as traditional schemata which the classical filmmaker can impose upon any subject. As King Vidor wrote: 'The filmmaker should be consciously aware of this 180° rule throughout the whole field of film action. It is not only beneficial in sports, but in chase sequences, with cowboys, Indians and cavalry, animal pursuits, moon landings, dinnertable conversations, and a thousand other movie subjects.' 43 Most film critics are aware of these schemata but consider them simply a neutral vehicle for the filmmaker's idiosyncratic themes or 'personal vision.' What makes the continuity devices so powerful is exactly their apparent neutrality; compositional motivation has codified them to a degree of rigidity that is still hard to realize. In each UnS film, less than 2 per cent of the shot-changes violated spatial continuity, and one-fifth of the films contained not a single violation. No wonder that, of all Hollywood stylistic practices, continuity editing has been considered a set of firm rules.
-57
As with other classical techniques, continuity editing cues form a redundant paradigm. Conventional 180° editing assumes that the establishing shot and the eyeline match cut and directional continuity of movement and the shot/reverse-shot schema will all be present to 'overdetermine' the scenographie space. The redundancy of the paradigm becomes evident when we watch a non-classical filmmaker simply remove one or two cues. In Dreyer's Day of Wrath (1943), the characters' eyelines in medium shot often violate the 180° axis, but there are frequent establishing shots to orient us. Conversely, in Bresson's Procès de Jeanne d'Arc (1961), the eyelines respect the axis of action, but scenes frequently lack establishing shots. 44 In neither film do we lose our bearings (although, since each filmmaker exploits his devices systematically, the result is significantly different from the space of the classical scene).
What are the narrational consequences of spatial continuity editing? One answer might be based on a broad conception of perspective. In perpetuating the playing space of post-Renaissance bourgeois theater, classical editing makes the spectator an ideally placed onlooker. To paraphrase Bazin, the action and the viewer are separate ('the episode would continue to exist objectively'), yet the narration acknowledges the onlooker by implicitly addressing her or him ('by allowing a better view'). In sum, the intelligible orientation created within the single shot is kept consistent across shots by positing a spectator that can be moved only within the limits of a theatrical space of vision.
This account is certainly correct as far as it goes. Its drawbacks are the passivity it imputes to the spectator and its neglect of certain significant irregularities in the continuity system. For one thing, the space constructed by continuity editing is rarely a total one, even on the favored side of the axis of action. Not only do we seldom see the fourth wall of the typical interior, but areas immediately in front of the camera remain relatively undefined. Films of the late teens and the 1920s sometimes have holes in their scenographie space; the establishing shot may not show all adjacent areas from which characters may emerge. And Hollywood practitioners have long employed the aptly named 'cheat cut, ' in which the shift of camera distance and angle during a cut covers a distinct change in character position (see figs 5.46 through 5.49). The cheat cut works to enhance balance, centering, or frontality: 45
'Cheating' is the great game between the camera operator and the Continuity girl. To compose a foreground or a background the operator will sometimes move or substitute objects, or have the artiste raised or lowered in relation to his surroundings. Actually, after a long while in pictures, I realised that such 'cheating' is seldom noticeable to an audience, but in the studio it often seems fantastic.
The viewer's willingness to ignore unshown areas of space and to overlook cheat cuts suggests that the viewer actively forms and tests specific hypotheses about the space revealed by the narration. The always-present pockets of non-established space are, in the absence of cues to the contrary, assumed to be consistent with what we see. (We assume that there is more wall, a door, etc.) If a technician or a lighting unit peeped into the shot, that would provoke us to revise such assumptions. The cheat cut suggests that a process of hierarchical selection is at work. Since we are to attend to story causality, the fact that a character is first three feet and then suddently two feet from another character becomes unimportant if our expectations about the action are confirmed from shot to shot. Of course, there are limits to how much the cut can cheat before the operation distracts us from story causality, and these warrant psychophysical study. 46
Our hierarchical selection of what to watch is evident from the very schemata of classical cutting. For example, the repetition of camera position becomes very important. Typically, any classical series of shots will include several identical camera set-ups. The reestablishing shot will usually be from the same angle and distance as the establishing shot; shot and reverse-shot framings may be repeated several times. Such repetitions encourage us to ignore the cutting itself and notice only those narrative factors that change from shot to shot. In a similar way, the first occurrence of a set-up often 'primes' us for a later action. In *The Caddy (1953), Harvey hides from dogs in a locker room. A plan américain reveals him leaning on the door; on the right of
-58
the frame are clothes lying on a coat rack. Cut: the dogs outside the door wander off. The next shot repeats the plan américain of Harvey, but now Harvey notices the clothes. The first set-up unobtrusively asked us to hypothesize that Harvey would disguise himself, and the guess is confirmed by keeping set-ups constant. A similar process occurs in figures 5.50 through 5.53. This priming of later actions does not occur in films by Eisenstein and Godard, for instance, who seldom exactly repeat set-ups and who thus demand that we reorient ourselves after every cut.
The phenomenon of priming illustrates Gombrich's point that schemata set the horizon of the viewer's expectations. Classical editing is organized paradigmatically, since any shot leads the viewer to infer a limited set of more or less probable successors. For example, an establishing shot can cut away to another space or cut in to a closer shot; the latter alternative is more likely. An angled medium shot of a character or object is usually followed by a corresponding reverse shot. Cutting around within a locale is most likely to be based upon eyeline matches and upon shot/ reverse-shot patterns, less likely to be based upon figure movement, and least likely to be based upon optical point-of-view. (In this respect, Hitchcock relies upon point-of-view cutting to an almost unique degree.) The classical construction of space thus participates in the process of hypothesis-forming that we saw at work in narration generally. Julian Hochberg compares the viewer's construction of edited space to 'cognitive mapping': 'The task of the filmmaker therefore is to make the viewer pose a visual question, and then answer it for him.' 47
The process of viewer expectation is particularly apparent in the flow of onscreen and offscreen space. Consider again the shot/reverse-shot schema. The first image, say a medium shot of Marilyn, implies an offscreen field, foreshadowing (by its angle, scale, and character glance) what could most probably succeed it. The next shot in the series, a reverse-angled view of Douglas, reveals the narratively significant material which occupies that offscreen zone. Shot two makes sense as an answer to its predecessor. This backing-and-filling movement, opening a spatial gap and then plugging it, accords well with the aims of classical narration. Furthermore, shot/reverse-shot editing helps make narration covert by creating the sense that no important scenographie space remains unaccounted for. If shot two shows the important material outside shot one, there is no spatial point we can assign to the narration; the narration is always elsewhere, outside this shot but never visible in the next. This process, which evidently is at work in camera movement and analytical cutting as well, is consistent with that unself-conscious but omnipresent narration described in Chapter 3. 48*
Classical offscreen space thus functions as what Gombrich calls a 'screen, ' a blank area which invites the spectator to project hypothetical elements on to it. 49 Given classical viewing priorities, we are more concerned with the distinct persons and things visible within space than with the spaces between and around them. If a shot shows a person or object that was implicit in the previous shot, we check the new material against our projection rather than measuring the amount of space left out. Since Hollywood scenography seldom represents a locale in its entirety, we must construct a spatial whole out of bits. And if those bits not only overlap in what they show but agree with the fields we have inferred to be lying offscreen, we will not notice the fuzzy areas that have never been strictly accounted for. Classical editing supports orientation according to Gombrich's negative principle of perspective: A convincing image need not show everything in the space as long as nothing we see actually contradicts what we expect. 50 If classical cinema makes the screen a plate-glass window, it is partly because it turns a remarkably coherent spatial system into the vehicle of narrative causality; but it is also because the viewer, having learned distinct perceptual and cognitive activities, meets the film halfway and completes the illusion of seeing an integral fictional space.

The Classical Hollywood Cinema - Cap 4

The Classical Hollywood Cinema. Film Style and Mode of Production to 1960
David Bordwell, Janet Staiger and Kristin Thompson
4

Time in the classical film
Our examination of exposition has shown that the narrational aspect of plot manipulates story time in specific ways. More generally, classical narration employs characteristic strategies for manipulating story order and story duration. These strategies activate the spectator in ways congruent with the overall aims of the classical cinema. We shall also have to pay some attention to how narration uses one device that is commonly associated with the Hollywood style's handling of time: crosscutting.

Temporal order: the search for meaning
After dramas supposedly without endings, here is a drama which would be without exposition or opening, and which would end clearly. Events would not follow one another and especially would not correspond exactly. The fragments of many pasts come to bury themselves in a single now. The future mixed among memories. This chronology is that of the human mind. 1
Jean Epstein, writing in 1927, thus describes his film La Glace à trois faces. Hollywood cinema, however, refuses the radical play with chronology that Epstein proposes; the classical film normally shows story events in a 1-2-3 order. Unlike Epstein, the classical filmmaker needs an opening, a threshold-that concentrated, preliminary exposition that plunges us in medias res. Events unfold successively from that. Advance notice of the future is especially forbidden, since a ftashforward would make the narration's omniscience and suppressiveness overt (see Chapter 30 on alternative cinemas' use of the flashforward). The only permissible manipulation of story order is the flashback.
Flashbacks are rarer in the classical Hollywood film than we normally think. Throughout the period 1917-60, screenwriters' manuals usually recommended not using them; as one manual put it, 'Protracted or frequent flashbacks tend to slow the dramatic progression'-a remark that reflects Hollywood's general reluctance to exploit curiosity about past story events. 2 Of the one hundred UnS films, only twenty use any flashbacks at all, and fifteen of those occur in silent films. Most of these are brief, expository flashbacks filling in information about a character's background; this device was obviously replaced by expository dialogue in the sound cinema. In the early years of .sound, when plays about trials were common film sources, flashbacks offered a way to 'open up' stagy trial scenes (e.g., The Bellamy Trial, Through Different Eyes, The Trial of Mary Dugan, Madame X, all 1929). Another vogue for flashbacks ran from the late 1930s into the 1950s. Between 1939 and 1953, four UnS films begin with a frame story and flash back to recount the bulk of the main action before returning to the frame. Yet those four flashback films still comprise less than 10 per cent of the UnS films of the period. What probably makes the period seem dominated by flashbacks is not the numerical frequency of the device but the intricate ways it was used: contradictory flashbacks in Crossfire (1947), parallel flashbacks in Letter to Three Wives (1948), open-ended flashbacks in How Green Was My Valley (1941) and I Walked With a Zombie (1943), flashbacks within flashbacks within flashbacks in Passage to Marseille (1944) and The Locket (1946), and a flashback narrated by a dead man in Sunset Boulevard (1950).
It is possible, of course, to present a shift in story order simply as such, with the film's narration overtly intervening to reveal the past.
-42-
In The Ghost ofRosie Taylor (1918), an expository inter-title announces that it will explain how the situation became what it is; the title motivates the flashback. The Killing (1956) uses voice-over, documentary-sty le narration to motivate 'realistically' its jumps back in time. The rarity of these overt intrusions shows that classical narration almost always motivates flashbacks by means of character memory. Several cues cooperate here: images of the character thinking, the character's voice heard 'over' the images, optical effects (dissolve, blurring focus), music, and specific references to the time period we are about to enter. If we see flashbacks as motivated by subjectivity, then the extraordinary fashion for temporal manipulations in the 1940s can be explained by the changing conception of psychological causality in the period. Flashbacks, especially convoluted or contradictory ones, can be justified by that increasing interest in vulgarized Freudian psychology which Chapter 2 has already discussed.
Classical flashbacks are motivated by character memory, but they do not function primarily to reveal character traits. Nor were Hollywood practitioners particularly interested in using the flashback to restrict point-of-view; one screenwriters' manual suggests that 'unmotivated jumping of time is likely to rattle the audience, thereby breaking their illusion that they participate in the lives of the characters.' 3 Even the contradictory flashbacks in Through Different Eyes or Crossfire serve not to reveal the teller's personality so much as they operate, within the conventions of the mystery film, as visual representations of lies. Jean Epstein's aim in La Glace à trois faces-to reflect the mixed temporality of consciousness, fragments of the past in a single now-is far removed from Hollywood's use of flashbacks as rhetorical 'dispositions' of the narrative for the sake of suspense or surprise. Nor need the classical flashback respect the literary conventions of firstperson narration. Extended flashback sequences usually include material that the remembering character could not have witnessed or known. Character memory is simply a convenient immediate motivation for a shift in chronology; once the shift is accomplished, there are no constant cues to remind us that we are supposedly in someone's mind. In flashbacks, then, the narrating character executes the same fading movement that the narrator of the entire film does: overt and self-conscious at first, then covert and intermittently apparent. Beginning with one narrator and ending with another (e.g., I Walked With a Zombie), or compelling a character to 'remember' things she never knew or will know (e.g., Ten North Frederick [1958]), or creating a deceased narrator (e.g., Sunset Boulevard)-all these tactics show that subjectivity is an arbitrary pretext for flashbacks.
Classical manipulations of story order imply specific activities for the spectator. These involve what psychologists call 'temporal integration, ' the process of fusing the perception of the present, the memory of the past, and expectations about the future. E.H. Gombrich points out that temporal integration depends upon the search for meaning, the drive to make coherent sense of the material represented. 4 The film which challenges this coherence, a film like Not Reconciled (1964), Last Year at Marienbad (1961), or India Song (1975), must make temporal integration difficult to achieve. In the classical film, however, character causality provides the basis for temporal coherence. The manipulations of story order in Not Reconciled or Marienbad are puzzling partly because we cannot determine any relevant character identities, traits, or actions which could motivate the breaks in chronology. On the other hand, one reason that classical flashbacks do not adhere to a character's viewpoint is that they must never distract from the ongoing causal chain. The causes and effects may be presented out of story order, but our search for their connections must be rewarded.
Psychological causality thus permits the classical viewer to integrate the present with the past and to form clear-cut hypotheses about future story events. To participate in the process of casting ever more narrow and exclusive hypotheses, we must have solid ground under our feet. Therefore, through repetition within the story action and a covertly narrated, 'objective' diegetic world, the film gives us clear memories of causal material; on this basis we can form expectations. At the same time, the search for meaning of which Gombrich speaks guides us toward the motifs and actions already marked as potentially meaningful. For example, motifs revealed in the credits sequence or in the early scenes accumulate
-43-
significance as our memory is amplified by the ongoing story. Kuntzel suggests that these reinscribed motifs create a vague déjà-vu that becomes gradually more meaningful: 'The entire itinerary of The Most Dangerous Game is to make its initial figure readable, to progressively reassure the subject plunged ex abrupto into the uncertainty of the figure.' 5 The classical aesthetic of 'planting' and foreshadowing, of tagging traits and objects for future use, can be seen as laying out elements to be recalled later in the cause-effect logic of the film. If temporality and causality did not cooperate in this way, the spectator could not construct a coherent story out of the narration.
Our survey of narration has shown that the viewer's successive hypotheses can be thought of as a series of questions. Hollywood cinema's reliance upon chronology triggers the fundamental query: What will happen next in the story? Each shot, wrote Loos and Emerson, 'is planned to lead the audience on to the next. At any point, the spectator is wondering how things will come out in the next scene.' 6 The forward flow of these hypotheses may be related to the irreversibility of the film-viewing experience; Thomas Elsaesser has speculated that the channeling of chronology into causality helps the viewer 'manage' the potentially disturbing nature of the film-viewing situation. 7 The relatively close correspondence between story order and narrational order in the classical film helps the spectator create an organized succession of hypotheses and a secure rhythm of question and answer.

Duration, deadlines, and dissolves
Like order, classical Hollywood duration respects very old conventions. The narration shows the important events and skips the intervals between them. The omitted intervals become codified as a set of punctuation marks: expository inter-titles ('The Next Day') and optical effects. From 1917 to 1921, fade-ins and -outs and iris-ins and -outs were the most common optical transitions between scenes. Between 1921 and 1928, the iris fell into disuse, replaced by the fade as the most common transition. In the sound era, fades and dissolves were the most common signs of temporal ellipsis. Wipes enjoyed a vogue between 1932 and 1941 and appeared occasionally thereafter. Such optical punctuation marks were often compared with theatrical or literary conventions (curtain, end of chapter). Within a scene, of course, some of the same ellipses could be used. After the late 1920s and until the early 1950s, scenes often began with a shot of a building or a sign and then dissolved to the action proper. In the same period, a wipe, either hard- or soft-edged, might follow a character moving from one sub-scene to another. (Not until the late 1950s did a few films begin to eliminate such internal punctuation and simply use the straight cut to link scenes and subscenes. 8 ) Such a clear set of cues creates an orderly flow of action; compare the disruptive effect, in the films of Eisenstein and Godard, of beginning a scene's action and then, part of the way through, interrupting the action with a title that tells us when the action is occurring.
Punctuation marks enable the narration to skip unimportant intervals by simple omission. The montage sequence lets the narration represent, however briefly, those intervals. The montage sequence does not omit time but compresses it. A war, a prison sentence, or a career can be summed up in a few shots. Films which cover a great length of time may make heavy weather of montage sequences, as does *High Time (1960), which employs montages of seasons and semesters to cover four years on a college campus. The montage sequence was especially important in literary adaptations, since the plots of novels tended to cover extensive periods. 9 So critical were montages to temporal construction that they were also called 'time-lapse' sequences.
The classical film creates a patterned duration not only by what it leaves out but by a specific, powerful device. The story action sets a limit to how long it must last. Sometimes this means simply a strictly confined duration, as in the familiar convention of one-night-in-a-mysterioushouse films (The Cat and the Canary [1927], Seven Footprints to Satan [1929], *One Frightened Night [1935], *Sh! The Octopus [1937]). More commonly, the story action sets stipulated deadlines for the characters.
The mildest and most frequent form of the deadline is the appointment. This is most evident in the romance line of action, wherein a suitor will invite a woman out for dinner, to a dance, etc.
-44
If the film makes romance primary, the acceptance, rejection, or deferral of such invitations forms a significant part of the drama (e.g., *Interlude [1957], *The King and the Chorus Girl [1937]). The very title of *Appointment for Love (1941) conveys the same idea. Even if the film does not rely completely upon the romance line of action, many scenes include the making of appointments for later encounters. Just as motifs anticipate future actions, so appointments gear our expectations toward later scenes.
The deadline proper is the strongest way in which story duration cooperates with narrative causality. In effect, the characters set a limit to the time span necessary to the chain of cause and effect. Over three-quarters of the UnS films contained one or more clearly articulated deadlines. The deadline may be stipulated in a line of dialogue, a shot (e.g., a clock), or crosscutting; whatever device is used, it must specify the durational limit within which cause and effect can operate. Most frequently, the deadline is localized, binding together a few scenes or patterning only a single one. Scenes in *Miss Lulu Bett (1921) are structured around the repeated deadline of the family's dinner hour. A series of short episodes in *High Time (1960) are governed by the fact that the freshmen must build a bonfire by seven o'clock. The localized deadline is of course most common at the film's climax. In *Fire Down Below (1957), one of the protagonists is trapped in the hold of a ship; it is on fire and sinking, and the suspense is predicated upon the slow drainage of time until the situation becomes hopeless. *The Canterville Ghost (1944) presents the climactic scene of the ghost and young William proving their courage by towing a ticking bomb across the landscape. When William says, If it'll hold for twenty seconds more!' the Ghost starts to count the seconds off. The conventional last-minute rescue is the most evident instance of how the classical film's climax often turns upon a deadline.
A deadline may also determine the entire structure of a classical film. The protagonist's goal can be straightforwardly dependent upon a deadline, as when in *Roaring Timber (1937), Jim agrees to deliver eighty million feet of lumber in sixty days. *The Shock Punch (1925) gives the protagonist the task of finishing construction of a building by a certain date; the film's last scene occurs on the deadline day. In 1940s films, the use of the flashback can also limit the duration of the story action. For example, *No Leave, No Love (1946) begins with the protagonist rushing to a maternity ward; while he waits for news of his child's birth, he tells another husband the story of how he met his wife. By halting the action at a point of crisis and flashing back to early events, the film makes those events seem to operate under the pressure of a deadline. (See also The Big Clock [1948] and Raw Deal [1948].)
*Uncertain Glory (1944) offers a clear example of how appointments mix with deadlines to unify the duration of the classical Hollywood film. The film's action takes place in France under the Nazi Occupation. The first six scenes present the escape of the convict Jean and his capture by the police detective Bonet; in these portions, alternating point-of-view creates suspense. When Bonet has captured Jean, we learn that the Gestapo will shoot one hundred hostages if a partisan saboteur does not surrender in five days. This long-term deadline structures the bulk of the film, as Bonet tries to convince Jean to pose as the saboteur, help the Resistance, and save the hostages. While the deadline hovers over the action, the two men quarrel, villagers conspire against them, Jean falls in love with a village woman (entailing small-scale appointments), and Jean tries several times to escape from Bonet. Finally, in the penultimate scene, at five o'clock Jean decides to surrender himself: 'Deadline's six o'clock, isn't it?' He turns himself in.
It should be evident that deadlines function narrationally. Issuing from the diegetic world, they motivate the film's durational limits: the story action, not the narrator, seems to decide how long the action will take. Planning appointments makes it 'natural' for the narration to show the meeting itself; setting up deadlines makes it 'natural' for the narration to devote screen time to showing whether or not the deadline is met. Moreover, appointments and deadlines stress the forward flow of story action: the arrows of the spectator's expectations are turned toward the encounter to come, the race to the goal. When, in * Applause (1929), the sailor from Wisconsin asks April for a date, we expect to see the date; when he says he has only four days of leave, we are not surprised that he should ask her to marry him before his leave is up. Deadlines and appoint
-45-
ments thus perfectly suit classical narration's emphasis upon eliciting hypotheses about the future.
As a formal principle, the deadline is one of the most characteristic marks of Hollywood dramaturgy. Alternative styles of filmmaking can often be recognized by their refusal to set such explicit limits on the duration of story action. The alternatives vary. Ozu structures his films by repeated routines and cycles of family behavior. Jacques Tati uses a fixed duration (a week, a day or two) simply as a block of time without a deadline. Eisenstein often composes a film of separate, durationally distinct episodes (e.g., Ivan the Terrible [1945]). The 'art cinema' of Federico Fellini, Ingmar Bergman, or Michelangelo Antonioni is characterized partly by its refusal of deadlines, its replacement of appointments by chance encounters, and its 'open' endings that do not allow the audience to anticipate when the chain of cause and effect will be completed. A Hollywood version of L'avventura (1960) would be sure to include a scene in which someone says: 'If we don't find Sandra in three days, her supply of food will run out.'
Within the classical scene, the viewer assumes durational continuity unless signals say otherwise. The individual shot is assumed to convey a continuous time span which only editing can disrupt. Yet the classical cinema is a cinema of cutting; the single-shot sequence is very rare. Thus classical editing strategies have to signal temporal continuity. Match-on-action cutting is the most explicit cue for moment-to-moment continuity. If a character starts to stand up in one shot and continues the movement in the next shot, the classical presumption is that no time has been omitted (see figs 4.1 and 4.2). Editors are warned that if they mismatch action, audiences will be confused about temporal progression. 10 But the match-on-action cut, expensive and timeconsuming, is relatively rare; of all the shot-changes in a classical film, no more than 12 per cent are likely to be matches on action. In the absence of information to the contrary, spatial editing cues, such as eyeline-match cutting, imply durational continuity.
The adoption of synchronized sound-on-film had a very powerful effect on how the classical cinema represented story time, as Chapter 23 will show in detail. Diegetic sound created a concrete perceptual duration that could aid editing in creating a seamless temporal continuity. If two characters are talking, the sound editor could make the continuous sound conceal the cut. A British editor summarized American practice: 11
This flowing of sound over a cut is one of the most important features of the editing of sound films-in particular, of dialogue films. The completely parallel cut of sound and action should be the exception rather than the rule. … Most editors today make a practice of lapping the last one or two frames of modulation on the soundtrack of the shot they are leaving over onto the oncoming shot.
That is, the shot change precedes the dialogue change by a syllable or a word. This 'dialogue cutting point' (Barry Salt's term) became standard by 1930. 12 On other occasions, of course, the sound can lead the image; very commonly a classical film will motivate a cut by an offscreen sound. The noise of a door opening, a character starting to speak, the music of a radio from another room-these can all help sound flow over a cut.
Another way of using sound to secure durational continuity is to employ diegetic music. Of course non-diegetic music, as accompaniment, had been present in the silent cinema, but there its quality as narration made it temporally abstract. In the sound film, diegetic music could cover certain gaps at the level of the image while still projecting a sense of continuous time. For example, in Flying Fortress (1942), a couple sit down to dinner in a restaurant while a band is playing. The meal is abbreviated by means of dissolves, creating ellipses on the visual track; but the band's music continues uninterrupted. The bleeding of music over large ellipses suggests how easily the temporal vagueness of music can make sound fulfill narrative functions.
The dissolve, the most common indication of duration, affords us an instructive example of how classical narration does its temporal work. Visually, the dissolve is simply a variant of the fade-a fade-out overlapped with a fade-in-but it is a fade during which the screen is never blank. 'To the layman or the average theatregoer, a lap dissolve passes unobtrusively by on the screen without his being aware that it had happened. A
-46
lap dissolve serves the purpose of smoothly advancing the story.' 13 The dissolve was quickly restricted to indicating a short, often indefinite interval, if only a few seconds (e.g., a dissolve from a detail to a full shot). This makes the dissolve a superb way to soften spatial, graphic, and even temporal discontinuities. The dissolve could blend newsreel footage with studio shots, cover mismatched figure positions or screen direction, or blend an extreme-long shot with a close-up (see figs 4.3 through 4.5). Filmmakers of the 1920s in Europe and Russia showed that the dissolve opens up a realm of sheerly graphic possibilities, but Hollywood severely curtailed these: apart from a few exceptions (such as Josef Von Sternberg's work), the Hollywood dissolve became, as Tamar Lane puts it, 'a link…. It bridges over from one situation to another without a jarring break of action and without need for explanatory matter.' 14
After 1928, the dissolve on the image track was accompanied by a sound transition as well. At first, the procedures of sound editing and the uncertainties of sound perspective made technicians puzzled. Imagine switching abruptly from the blast of a jazz orchestra to a flash of a whispered conversation, then to the rush of a train and back to the silken vampire sleeping peacefully in her boudoir. Such a rush of conflicting sound ought to leave an audience as nervous as a doe at a waterhole.' 15 Sound dissolves were declared distracting; while a closeup of a face could dissolve to a long shot of a crowd, to mix even briefly the character's speech with the crowd's babble would result in cacophony. Instead, the character would complete the dialogue and pause; the crowd noise would then be sneaked in over the dissolve. Like the offscreen sound that motivates the cut to a new space, the sound bridge here may sometimes very slightly anticipate the next image. Both image and sound dissolving procedures show how, once a transition became codified, it could provide a continuous and unself-conscious narration.
Like our experience of story order, the viewer's experience of story duration depends upon a search for meaning. Gombrich writes: 'We cannot judge the distance of an object in space before we have identified it and estimated its size. We cannot estimate the passage of time in a picture without interpreting the event represented.' 16 In the classical cinema, the narration's emphasis upon the future gears our expectations toward the resolution of suspense. It is this that determines what periods the narration will eliminate or compress. When this does not happen, when the narration dwells upon 'dramatically meaningless intervals, ' duration comes forward as a system in the film and vies with causality for prominence. (See the various critiques 17 of Hitchcock's use of the long take in Rope [1948].) Time in the classical film is a vehicle for causality, not a process to be investigated on its own. Hence the stricture that a walk without dialogue is 'dead' or wasted time. (Compare the durational importance of the silent walk in Dreyer, in Antonioni, and, from a different culture, in the Navajo films described by Sol Worth and John Adair. 18 )
More generally, classical narration's insistence upon closure rewards the search for meaning and makes the time span we experience seem a complete unit. Even from shot to shot, our expectation of causally significant completion controls how we respond. 'We hardly realize that we look at two different shots if the first one shows the beginning of an action and the next one its continuation.' 19 The match-on-action cut, the bleeding of sound over a cut, the use of dissolves and diegetic music all confirm our expectation of completion. The viewer's ability to test hypotheses against a film's unfolding cause and effect means that duration again becomes secondary to a search for narrative meaning.
Hollywood has also exploited our search for temporal meaning by shaping the felt duration of our experience. Narrative 'rhythm' can be thought of as a way in which narration focuses and controls successive hypotheses. Camera movement, especially if it is independent of the figures and closely timed to music, can create a moment-by-moment arc of expectation. 20 Editing was the earliest rhythmic realm which the classical cinema systematically exploited; by 1920, scenarists were recommending using short shots to increase excitement. 21 Rhythmic editing is still far from clearly understood theoretically, but certainly the time needed to grasp a new shot depends partly upon expectation. It appears that if the viewer is prepared and if the shot is graphically comprehensible, the viewer requires between half a second and three seconds to adjust to the cut. 22 Slowly paced editing leaves a
-47-
comfortable margin, so that the new shot is on the screen quite long enough for the viewer to assimilate it. But in Hollywood's use of accelerated editing, the viewer is primed to expect a very narrow range of alternative outcomes and the shots then flash on the screen so quickly that the viewer can 'read' them only in gross terms: do they confirm or disconfirm the immediate hypothesis? This process is evident in the last-minute rescue, when all the viewer wants to know is whether the rescuers will arrive in time, so the accelerating editing builds excitement by confining each shot to posing, retarding, and eventually answering this question. The ability of rapid editing to funnel the spectator's hypotheses into very narrow channels is confirmed by Robert Parrish's claim that fast pace can cover story problems. Asserting that The Roaring Twenties (1939) works like 'one big ninety-minute montage, ' Parrish notes: 'The audience never gets a chance to relax and think about the story holes. They're into the next scene before they have time to think about the last one.' 23

Crosscutting
Strictly speaking, crosscutting can be considered a category of alternating editing, the intercalation of two or more different series of images. If temporal simultaneity is not pertinent to the series, the cutting may be called parallel editing; if the series are to be taken as temporally simultaneous, then we have crosscutting. For example, if the film alternates images of wealth and poverty with no temporal relation to one another, we have parallel editing; but if the rich man is sitting down to dinner while the beggar stands outside, we have crosscutting. Griffith's Intolerance (1916) uses both types: parallel editing makes abstract analogies among the four epochs, while crosscutting within each epoch depicts simultaneous actions. In the classical Hollywood cinema, parallel editing is a distinctly unlikely alternative, since it emphasizes logical relations rather than causality and chronology.
Crosscutting is a narrational process: two or more lines of action in different locales are woven together. Our hero gets up in the morning; cut to the boss looking at the clock; cut to our hero eating breakfast; cut to the boss pacing. Christian Metz has pointed out that such a sequence manipulates both order and duration. 24 Within each line of action, the events are consecutive; but between the lines of action taken as wholes, the temporal relations are simultaneous. The hero gets up somewhat before the boss looks at the clock, but across the whole sequence, we understand that while the hero gets up and comes to work the boss waits for him. There is yet another factor involved, which Metz does not mention: usually, crosscutting creates ellipses. If we cut from hero waking up to boss to hero leaving, the shot of the boss covers all the time it takes our hero to dress, wash, etc. Crosscutting almost always skips over intervals in exactly this way. Crosscutting, then, creates a unique set of temporal relations-order, ellipsis, simultaneity-which function for specific narrational ends.
Alternation of narrational point-of-view has a long history in literature and other arts, but crosscutting is often linked to specifically nineteenth-century theatrical and literary sources. Nicholas Vardac found 'cross-cut' scenes in nineteenth-century drama, which used dual box sets and area lighting to switch between lines of action. 25 Eisenstein traced Griffith's parallel montage through theatrical melodrama back to Dickens's novels. 26 The analogies with other arts emphasize the brevity of the scenes alternated and the simultaneity of the actions represented. Chapter 16 will show that both these aspects of crosscutting were common in American filmmaking long before 1917. But such analogies with other arts do not specify all the features of classical crosscutting.
Classical crosscutting traces out personal cause and effect, creates deadlines, and frees narration from restricting itself to a single character's point-of-view. We most commonly think of crosscutting as supporting a deadline-supremely, the last-minute rescue situation. But a silent film might employ crosscutting in a great many scenes-as exposition, as a reminder of characters' whereabouts, and especially as a way in which narration could control the viewer's hypothesis-framing. Crosscutting thus reveals narration to be omniscient (the narration knows that something important is happening in another line of action), but this omniscience, true to classical precept, is rendered as omnipresence.
In 1920, Loos and Emerson advised the screen-
-48
writer that two crosscut lines of action would help keep the audience interested. 27 Of the UnS silent films, 84 per cent use extensive passages of crosscutting. With the coming of sound, however, crosscutting became far less frequent. Of the UnS sound films, only 49 per cent use any crosscutting at all, and only 16 per cent use it as extensively as did silent films. The reasons are evident. Dialogue would not be cut as quickly as silent action, and crosscutting lines of dialogue (done in Europe by René Clair and Fritz Lang) probably seemed too narrationally intrusive for Hollywood film-making. 28 The abandonment of crosscutting thus became consonant with a greater reticence on the part of sound-film narration.
None the less, the principle behind crosscutting remained important for the sound film. As Chapter 23 will show, the rhythm of silent film editing found a functional equivalent in the sound film's rapid shifts from scene to scene. In *The Whole Town's Talking (1935), our hero's boss notices that he is late and begins to interrogate other employees. The scene switches to Jones at home, asleep; he wakes up, notices the time, and rushes off. We then see Jones arrive at work. Such shifts in locale could be motivated by sound links as well (music, radio or television broadcasts, phone conversations, etc.). In such ways, a rapid alternation of distinct scenes could stimulate crosscutting's characteristic play with time-consecutive order, ellipsis, and an overall sense of simultaneity. A discreet narration oversees time, making it subordinate to causality, while the spectator follows the causal thread.

The Classical Hollywood Cinema - Cap 3

The Classical Hollywood Cinema. Film Style and Mode of Production to 1960
David Bordwell, Janet Staiger and Kristin Thompson

3

Classical narration
A film's story does not simply shine forth; as viewers, we construct it on the basis of the plot, the material actually before us. The classical guidelines for this construction are those principles of causality and motivation already sketched out in Chapter 2. A film's plot usually makes those guidelines applicable by transmitting story information. This aspect of plot I shall call narration.
Hollywood's own discourse has sought to limit narration to the manipulation of the camera, as in John Cromwell's remark that, 'The most effective way of telling a story on the screen is to use the camera as the story-teller.' 1 And the classical film's narration itself encourages us to see it as presenting an apparently solid fictional world which has simply been filmed for our benefit. André Bazin describes the classical film as being like a photographed play; the story events seem to exist objectively, while the camera seems to do no more than give us the best view and emphasize the right things. 2 But narration can in fact draw upon any film technique as long as the technique can transmit story information. Conversations, figure position, facial expressions, and well-timed encounters between characters all function just as narrationally as do camera movements, cuts, or bursts of music.
From this standpoint, classical narration falls under the jurisdiction of all the types of motivation already surveyed. In a classical film, narration is motivated compositionally; it works to construct the story in specific ways. Narration may also be motivated generically, as when performers in a musical sing directly to the spectator or when a mystery film withholds some crucial story information. Narration is less often motivated 'realistically, ' although the voice-over commentary in semidocumentary fiction films might insist that the story action is based on fact. Artistically motivated narration is very rare in classical films and never occurs in a pure state. A non-classical director like Jean-Luc Godard can 'lay bare' a film's narrational principles, as does the beginning of Tout va bien (1972), in which anonymous voices play with alternative ways of opening the film, hiring cast and crew, and financing the film. But when a classical film wants to call attention to the 'palpability' of its narration, it must create a context that motivates baring the device by other means as well. For instance, in scene after scene of *The Man Who Laughs (1928), the narration conceals Gwynplaine's deformed mouth from us (by veils, strategically placed furniture, etc.). But in one scene, the narration lays bare this very pattern. During his stage act, Gwynplaine looks out at us and deliberately reveals his deformity; then a clown in his act slowly covers it again. The shot thus stages the act of revelation and concealment that has been central to the narration throughout. However, this baring of the device is partly motivated by realism (Gwynplaine is on stage, revealing his deformity to an audience in the fiction) and by causal necessity (for the story to proceed, a woman in the audience must see his mouth and take pity upon him). We encounter again the familiar multiple motivation of the classical text.
We could follow Hollywood's lead and simply label such carefully motivated narration 'invisible.' Hollywood's pride in concealed artistry implies that narration is imperceptible and unobtrusive. Editing must be seamless, camerawork 'subordinated to the fluid thought of the dramatic action.' 3 Some theorists have called the classical style transparent and illusionist, what Noël Burch has called 'the zero-degree style of filming.' 4 This is to say that classical technique is usually motivated compositionally. The chain of
-24-
cause and effect demands that we see a close-up of an important object or that we follow a character into a room.
'Invisible' may suffice as a rough description of how little most viewers notice technique, but it does not get us very far if we want to analyze how classical films work. Such concepts play down the constructed nature of the style; a transparent effect does not encourage us to probe beneath its smooth surface. The term is also imprecise. 'Invisibility' can refer to how much the narration tells us, upon what authority it knows or tells, or in what way it tells. A tangle of different problems of narration is packed into this 'invisibility.'
How then to characterize classical narration? Meir Sternberg has put forth a clear theory that will prove useful. 5 Sternberg suggests that narration (or the narrator) can be characterized along three spectra. 6 A narration is more or less self-conscious: that is, to a greater or lesser degree it displays its recognition that it is presenting information to an audience. 'Call me Ishmael' marks the narrator as quite self-conscious, as does a character's aside to the audience in an Elizabethan play. A novel which employs a diarist as narrator is far less self-conscious. Secondly, a narration is more or less knowledgeable. The omniscient speaker of Vanity Fair revels in his immense knowledge, while the correspondents in an epistolary novel know much less. As these examples suggest, the most common way of limiting a narrator's knowledge is by making a particular character the narrator. Thus the issue of knowledge involves point-of-view. Thirdly, a narration is more or less communicative. This term refers to how willing the narration is to share its knowledge. A diarist might know little but tell all, while an omniscient narrator like Henry Fielding's in Tom Jones may suppress a great deal of information. Some of Brecht's plays use projected titles which predict the outcome of a scene's action: this is less suppressive than a normal play's narration, which tends to minimize its own omniscience. 7
Sternberg's three scales can be summarized in a series of questions. How aware is the narration of addressing the audience? How much does the narration know? How willing is the narration to tell us what it knows?
Sternberg's categories help us analyze classical narration quite precisely. In the classical film, the narration is omniscient, but it lets that omniscience come forward more at some points than at others. These fluctuations are systematic. In the opening passages of the film, the narration is moderately self-conscious and overtly suppressive. As the film proceeds, the narration becomes less self-conscious and more communicative. The exceptions to these tendencies are also strictly codified. The end of the film may quickly reassert the narration's omniscience and self-consciousness.

The modest narration
Classical narration usually begins before the action does. True, the credits sequence can be seen as a realm of graphic play, an opening which is relatively 'open' to non-narrational elements. (Certainly it is in credits sequences that abstract cinema has had its most significant influence upon the classical style.) Yet the classical Hollywood film typically uses the credits sequence to initiate the film's narration. Even these forty to ninety seconds cannot be wasted. Furthermore, in these moments the narration is self-conscious to a high degree. Musical accompaniment already signals the presence of this narration, and often musical motifs in this overture will recur in the film proper. The title will most probably name or describe the main character (*Mickey [1918], *Gidget [1959], *King of the Rodeo [1928]) or indicate the nature of the action (*Going Highbrow [1935], *Impact [1949]). If not, the title can suggest the locale of the action (*Adventure Island [1947], *Wuthering Heights [1939]), a motif in the film (*Applause [1929], *Balalaika [1939]), or the time of the action (*The Night Holds Terror [1955]). The credits that list the cast may reinforce the title (e.g., *The King and the Chorus Girl [1937], starring Fernand Gravet and Joan Blondell), but they will certainly introduce the film's narrative hierarchy. Protagonist, secondary protagonist, opponents, and other major characters will be denoted by the order, size, and time onscreen of various actors' names. Some films strengthen this linkage by adding shots of the characters to the credits, in which the amount of the screen surface a character is allotted indicates the character's importance (fig 3.1).
-25-
(Compare the flattening effect of credits which make no distinction among major actors and walk-on parts, such as the 'democratic' credits of Jean-Marie Straub and Danièle Huillet's Not Reconciled [1964].) Even the studio logo, the MGM lion or the Paramount mountain, has been analyzed as a narrational transition. 8 The credits are thus highly self-conscious, explicitly addressed to the audience.
In the silent period, many films went no further than these cues, laying the credit sequence against black backgrounds or a standardized design (e.g., curtains, pillars, or picture frames). Some credits sequences, however, used 'art titles' whose designs depicted significant narrative elements. William S. Hart's *The Narrow Trail (1917), for instance, displays its credits against a painting of a stagecoach holdup. By the 1920s, such art titles were commonly used for exposition (see fig 3.2). Lettering could also indicate the period or setting of the story, a practice probably influenced by playbills and illustrated books: narration rendered as typography. In the 1920s, a credits sequence might appear over moving images (e.g., *Merry-Go-Round [1923]) or might be animated (e.g., *The Speed Spook [1924]). The sound cinema canonized this stylized 'narrativization' of the credits sequence, assigning it a range of functions.
The credits can anticipate a motif to appear in the story proper. In *Woman of the World (1925), the protagonist's scandalous tattoo is presented as an abstract design under the credits; in *The Black Hand (1950), a stiletto forms the background for the titles. Credits' imagery can also establish the space of the upcoming action, as do the snowy fir trees in *The Michigan Kid (1928) or the city view in *Casbah (1948). Credits often flaunt the narration's omniscience and tantalize us with glimpses of action to come. As early as *The Royal Pauper (1917), we find the credits summarizing the rags-to-riches story action by dissolving from a shot of the star, dressed as a poor girl, to a shot of her wearing expensive clothes. Thierry Kuntzel has shown how the opening credit sequence of The Most Dangerous Game (1932), a shot of a hand knocking at a door, stages an important gesture of the ensuing film and anticipates several motifs in the setting and action. The credits sequence of Bringing Up Baby (1938) presents stick-figure man, woman, and leopard engaged in actions that will reappear in the film; *Sweepstakes Winner (1939) employs the same strategy (see fig 3.3). As Kuntzel points out, such sequences are explicitly narrational: the unknown hand knocking at the door can only be the viewer's, giving an idealized representation of the viewer's entry into the film. 9 Such overt address to the spectator can also be seen in those still-life compositions of book pages or album leaves turned by unknown hands (e.g., *Penthouse [1933], *Easy to Look At [1945], *Play Girl [1941]). In the postwar period, direct address in credits sequences could also be accomplished through a voice-over narrator. In such ways, the credits sequence flaunts both the narration's omniscience and its ability to suppress whatever it likes.
Like credits, the early scenes of the action can reveal the narration quite boldly. Before 1925, the film might open with a symbolic prologue, mocked by Loos and Emerson as 'visionary scenes of Heaven or Hell, of the Fates weaving human lives in their web.' 10 (See, for example, fig 3.4, from The Devil's Bait [1917].) More often, silent films simply used expository titles to announce the salient features of the narration. In the sound era, other film techniques take on this role of foregrounding the narration. After the credits, *Partners in Crime (1928) reveals a city landscape and an inter-title, 'Gangsters and Gun War-A City Steeped in Crime' (see fig 3.5). Suddenly the title shatters as hands holding guns break through to fire directly at the audience (see figs 3.6 and 3.7). At the start of *Housewife (1934), the camera tracks with a milkman up to the front door and lingers on the front door as he leaves. There is a cut to the welcome mat, and the camera tracks in and tilts up to the doorbell and name card. The shots have treated the camera as if it were a guest strolling up to the house. *Easy to Look At (1945) opens with a voice-over narrator describing the heroine's arrival in the city: 'And thus New York's population is increased by one-and quite a number…' as a man on the street gawks at her. Such passages reveal the narration to be widely knowledgeable and highly aware of addressing an audience.
The narration can also exploit the opening moments to stress its ability to be more or less communicative. *The Case of the Lucky Legs (1935) opens with a flurry of women's legs striding
-26-
up a flight of steps (see fig 3.8) and then dissolves to a sign (see fig 3.9). Several pairs of legs are revealed (see fig 3.10). At the end of the scene, as a former contest winner tries to claim her prize, the swindler pushes her away (see fig 3.11) and the camera pans to an advertisement for the Lucky Legs contest (see fig 3.12). The image dissolves to a pair of legs stretched out (see fig 3.13) and pans to their owner, the latest bilked woman, sobbing. The gratuitous camera movement to the sign and the opening of the next scene provide overtly ironic commentary on the contest.
The explicit presence of the narration in these heavily expository beginnings is confirmed by the eventual emergence of the 'pre-credits sequence.' Here the film opens truly in medias res, with the credits presented only after an initial scene or two of story action. This practice began in the 1950s, possibly as a borrowing from television's technique of the 'teaser.' The effect of pre-credits action was to eliminate the credits as a distinct unit, sprinkling them through a short action sequence that conveyed minimal story information (e.g., the establishment of a locale or the connecting of two scenes by a trip). The postponement of the credits tacitly grants the narrational significance of whatever scenes open the film.
Yet once present in these opening passages, the narration quickly fades to the background. In the course of the opening scenes, the narration becomes less self-conscious, less omniscient, and more communicative. Very flagrant examples allow us to trace this fading process at work.
*The Caddy (1953) has a highly stylized credits sequence that signals the genre (comedy), repeats the principal motif (golf clubs, tees, tartan), and anticipates story events (the cartoon figures). (See fig 3.14.) The film's first shot reveals a theatre marquee which carries caricatures similar to those in the credits (see fig 3.15). The bandstand's design repeats the caricatures, linking the figures to the live protagonists we finally see (fig 3.16). In a sliding movement, the narration's cartoon images of Dean Martin and Jerry Lewis have become gradually replaced by the story's images of the characters themselves. A more complex example occurs in *The Canterville Ghost (1944). While a voice-over commentator tells of the Ghost's history, the image shows the relevant passage in a book, Famous Ghosts of England. There follows a flashback to 1634, which shows how the cowardly Simon was bricked up in a wall of the mansion. The camera tracks into a close-up of a birthmark on Simon's neck (see fig 3.17), which freezes into an illustration in the book as the voice-over commentary resumes (see fig 3.18). The page is turned as the narrator describes the castle today; the illustration of the castle (see fig 3.19) dissolves to the same image on film, into which the heroine Jessica rides (see fig 3.20). Action has replaced the non-diegetic voice, and we never see or hear the narration so evidently again.
The phasing out of the narrator is also visible in historical changes in the silent cinema's expository tactics. Before 1917, films commonly introduced characters in ways that called attention to the act of narration. An expository title would name and describe the character and attach the actor's name; then a shot might show the character striking a pose in a non-diegetic setting (e.g., a theater stage). After several characters were introduced this way, the fictional action would begin. After 1917, such signs of narration diminished. Characters would be introduced upon their first appearance in the action. Overt commentary in the titles ('Max, a Bully') would be replaced by images of the character enacting typical behavior (e.g., Max kicking a dog). 11
The role of expository inter-titles changed as well. Silent scenarists were aware that the expository title foregrounded narration. One writer compared the expository title to a Greek chorus, 'someone who is behind the scenes. They are in the secret of the play.' 12 Another critic was even more aware of the intrusion: 'The title may say no more than “Dawn” or “Night” or “Home”; but it clearly is the injected comment of an outsider who is assumed, by the author's own terms, to be absent.' 13 (This, he claimed, 'breaks the spell of complete absorption.') The presence of an unseen fictional narrator was also marked in expository titles by the use of the past tense, which became standard after 1916. After 1917, Hollywood film became less and less reliant upon expository inter-titles and more dependent upon dialogue titles. Between 1917 and 1921, one-fifth to one-third of a film's inter-titles would be expository; after 1921, expository titles con
-27
stituted less than a fifth of the total. In the later silent years, we find films with no expository titles at all. Placement and length changed too: after 1921, the early scenes of the film contain more and longer expository titles than do later scenes. The cultivation of the art title, the expository title enhanced by a pictorial design, further substituted image for language. Expositional tasks were shifted to character dialogue and action, not only across the period but within the individual film.
The judicious combination of expository titles, dialogue titles, and exemplary character action created a fairly knowledgeable and communicative narrator. Consider the opening scene of *Miss Lulu Bett (1921). The family assembles for dinner, and an expository title introduces each family member. The title is then followed by a character performing a typical action which confirms the title's description. After the narration identifies the youngest daughter, the images show her swiping food playfully. After the father is identified, he goes to the clock to check his watch. Once most of the family are introduced, another expository title introduces the elder daughter but adds the information that she wants to leave the family. This title is followed by a shot of her at the front gate, holding a boy's hand. Because the narration has already accurately characterized the other family members, we trust its information about the daughter's purely private desires-information which is in turn immediately confirmed by her action. The narration is omniscient and reliable. The smoothness of such narration was recognized in Europe in the silent era; a Parisian critic noted that the Hollywood film always begins with a long expository title explaining the film's theme, followed by the rapid introduction to and delineation of characters by means of titles and actions. The critic emphasized that Hollywood films avoided the gradual psychological revelation characteristic of Swedish and German films of the period. 14
What enables the narration to fade itself out so quickly? Any narrative film must inform the viewer of events that occured before the action which we see. The classical film confines itself almost completely to a sort of exposition described by Sternberg as concentrated and preliminary. 15 This means that the exposition is confined principally to the opening of the plot. In explaining how to write a screenplay, Emerson and Loos claim that the opening should 'explain briefly but clearly the essential facts which the audience must know in order to understand the story, ' preferably in one scene. 16 Such advice may seem commonplace, but we need to remember that this choice commits the Hollywood film to a slim range of narrational options. Scattered or delayed exposition has the power to alter the viewer's understanding of events; making the spectator wait to fill gaps of causality, character relations, and temporal events can increase curiosity and even create artistic motivation, baring the device of narration itself. But concentrated and preliminary narration helps the classical film to make the narration seem less omniscient and self-conscious.
Classical narration also steps to the background by starting in medias res. The exposition plunges us into an already-moving flow of cause and effect. As Loos and Emerson put it, the action must begin 'with the story itself and not with the history of the case which leads up to the story.' 17 When the characters thus assume the burden of exposition, the narration can seem to vanish.
*The Mad Martindales (1942) offers a simple case. After an expository title ('San Francisco 1900'), the film opens with a close-up of a cake, inscribed 'Happy Birthday Father.' The camera tracks back, and while a maid and butler decorate the cake they discuss household affairs. The camera follows the butler to the piano, where Evelyn, the elder daughter, sits. Evelyn and the butler converse. We then follow the butler to the study, past the younger daughter Cathy, who is sitting at the desk writing. The camera holds on her while the butler leaves. Bob, Cathy's friend, thrusts his head in the window, which gives her a chance to explain what she's writing (a feminist tract, surprisingly enough). The phone rings and Evelyn answers it. The caller is her boyfriend Peter, who proposes marriage to her. At this juncture, the girls' father arrives, having just bought a Poussin painting. While workmen uncrate the painting, the family discuss Cathy's graduation, Martindale's birthday, the news about Peter, etc. When the butler brings birthday champagne, Cathy raises the issue of unpaid bills; at this point, the lights go out, cut off by the utility company. As the scene ends, the family discovers that it is penniless and Cathy sorrow
-28
fully reveals her gift to her father-a wallet. You are right to think that this scene is overstuffed with information, but it is typical of Hollywood cinema's almost Scribean loading of exposition into a film's first scenes. By plunging in medias res with the first shot of 'Happy Birthday Father, ' the film lets the characters tell each other what we need to know.
Classical narration may reemerge more overtly in later portions of the film, but such reappearance will be intermittent and codified. In the silent cinema, the expository art title may include imagery that comments overtly on the action. Occasionally, the narration will reassert its omniscience by camera movement: the cliché example is the pan from the long shot of the stagecoach to the watching Indians on the ridge. In the sound film, an overlapping line of dialogue can link scenes in ways that call attention to the narration. Many of the examples of artistic motivation and 'baring the device' that I considered in the last chapter can now be seen as examples of self-conscious and flagrantly suppressive narration. Narrational intrusions may also be generically motivated: in a mystery film, framing only a portion of the criminal's body as the crime is committed, or in a historical film, making the narrator 'the voice of history.' 18 Whatever the genre, however, there is yet another moment that narration comes strongly forward in the classical film-during montage sequences.
Typically, the montage sequence compresses a considerable length of time or space, traces a large-scale event, or selects representative moments from a process. 19 Cliché instances are fluttering calendar leaves, brief images of a detective's search for witnesses, the rise of a singer given as bits of different performances, the accumulation of travel stickers on a trunk, or a flurry of newspaper headlines. Rudimentary montage sequences can be found in Hollywood films of the teens and early twenties. By 1927, montage sequences were very common, and they continue to be used in a variant form today.
From a historical perspective, the montage sequence is part of Hollywood's gradual reduction of overt narrational presence. Instead of a title saying 'They lowered the lifeboats, ' or 'While the jury was out, McGee waited in a cold sweat, ' the film can reveal glimpses of pertinent action. The montage sequence thus transposes conventions of prose narration into the cinema; Sartre cites Citizen Kane's montages as examples of the 'frequentative' tense (equivalent to writing 'He made his wife sing in every theater in America'). 20 Moreover, the montage sequence aims at continuity, linking the shots through non-diegetic music and smooth optical transitions (dissolves, wipes, superimpositions, occasionally cuts). Yet the montage sequence still makes narration come forward to a great degree. Extreme close-ups, canted angles, silhouettes, whip pans, and other obtrusive techniques differentiate this sort of segment from the orthodox scene. When newspapers swirl out of nowhere to flatten themselves obligingly for our inspection, or when hourglasses and calendar leaves whisk across the screen, we are addressed by a power that is free of normal narrative space and time. What keeps the montage sequence under control is its strict codification: it is, simply, the sequence which advances the story action in just this overt way. Flagrant as the montage sequence is, its rarity, its narrative function, and its narrowly conventional format assure its status as classical narration's most acceptable rhetorical flourish.

Causality, character, and point-of-view
After the concentrated, preliminary exposition and except for intrusions like montage sequences, the classical film reduces narration's prominence. Chapters 4 and 5 will show how this process shapes cinematic space and time. For now, I want only to indicate the general ways that classical Hollywood narration reveals self-consciousness, omniscience, and communicativeness.
After the opening portions of the classical film, the narration's self-consciousness is generally kept low, chiefly because character action and reaction convey the ongoing causal chain to us. It is here that the effect of an enclosed story world, Bazin's objectively existing play simply transmitted by the camera, is at its strongest. Many devices of nineteenth-century realist theater-exposition by character conversation, speeches and actions which motivate psychological developments, well-timed entrances and exits-all assure the homogeneity of the fictional world. This homogeneity has induced many theorists and most
-29
viewers to see the classical film as composed of a solid and integral diegetic world occasionally inflected by a narrational touch from the outside, as if our companion at a play were to tug our sleeve and point out a detail. We must, however, make the effort to see the film's diegetic world as itself constructed and, hence, ultimately just as narrational as the most obtrusive cut or voiceover commentary. Yet we need to recognize how important this apparently natural, actually covert narration is to the classical cinema. In what follows, I shall assume that this narration-through-character-interaction constitutes the most normal and least noticeable ploy of Hollywood narration.
The narration reinforces the homogeneity of the fictional world by means of a non-theatrical device: the use of public and impersonal sources of information that can be realistically or generically motivated within the film. The most common instrument is the newspaper. ROSEN FOUND GUILTY: the headline or article becomes an unquestioned surrogate for the narrator's presence. In many films of the 1930s, newspaper reporters become an expository chorus, initiating us into the action. Other public transmitters of information include radio, television, bulletin boards, posters, ticker tape, tour guides, and reference books (e.g., the Ghosts of England volume in *The Canterville Ghost). These impersonal sources of story information also prove invaluable in toning down the self-consciousness of montage sequences.
Classical narration is potentially omniscient, as credits and openings show and as Hollywood's own discourse generally acknowledges. A. Lindsley Lane, for example, refers to 'omniscient perception' as the basic law of film. In the bulk of the Hollywood film, this omniscience becomes overt occasionally but briefly, as when a camera angle or movement links characters who are unaware of each other. 21 The same omniscience becomes overt in the anticipatory qualities of narration-the character who enters a scene just before she or he is needed, the camera movement that accommodates a character's gesture just before it occurs, the unexpected cut to a doorbell just before a thumb presses it, the music that leads us to expect a prowler to jump out of the shrubbery. 'There is only one way to shoot a scene, ' Raoul Walsh claimed, 'and that's the way which shows the audience what's happening next.' 22
The most evident trace of the narration's omniscience is its omnipresence. The narration is unwilling to tell all, but it is willing to go anywhere. This is surely the basis of the tendency to collapse narration into camerawork: the camera can roam freely, crosscutting between locales or changing its position within a single room. 'The camera, ' writes Lane, 'stimulates, through correct choice of subject matter and setup, the sense within the percipient of “being at the most vital part of the experience-at the most advantageous point of perception” throughout the picture.' 23 Sometimes this ubiquity becomes only artistically motivated, as in those 'impossible' camera angles that view the action from within a fireplace or refrigerator. 24 Spatial omnipresence is, of course, justified by what story action occurs in any given place, and it is limited still further by specific schemata, as we shall see in Chapter 5. To avoid treating the camera as narrator, however, we should remember that what the camera does not show implies omnipresence negatively-the site of an action we will learn of only later, the whole figure of the mysterious intruder. The narration could show us all, but it refuses.
Classical narration admits itself to be spatially omnipresent, but it claims no comparable fluency in time. The narration will not move on its own into the past or the future. Once the action starts and marks a definite present, movements into the past are motivated through characters' memory. The flashback is not presented as an overt explanation on the narration's part; the narration simply presents what the character is recalling. Even more restrictive is classical narration's suppression of future events. No narration in any text can spill all the beans at once, but after the credits sequence, classical narration seldom overtly divulges anything about what will ensue. It is up to the characters to foreshadow events through dialogue and physical action. If this is the last job the crooks will pull, they must tell us, for the narration will not become more self-conscious in order to do so. If the love affair is to fail, the characters must intuit it: 'These things never happen twice' (*Interlude [1957]). At most, the narration can drop self-conscious hints, such as pointing out a significant detail that the
-30
characters have overlooked; e.g., the camera movement up to the 'Forgotten Anything?' sign on the hotel-room door in Touch of Evil (1957). More commonly, anticipatory motifs can be included if the shot is already motivated for another purpose. Near the end of *From Here to Eternity (1953), the attack on Pearl Harbor is anticipated when the camera pans to follow a character and reveals a calendar giving the date as December 6.
Classical narration thus delegates to character causality and genre conventions the bulk of the film's flow of information. When information must be suppressed, it is done through the characters. Characters can keep secrets from one another (and us). Confinement to a single point-of-view can also suppress story information. Genre conventions can cooperate, as the editors of Cahiers du cinéma point out in their analysis of Young Mr. Lincoln (1939). Here the narration must juggle three points of view so as to keep certain information from the spectator. Two brothers accused of murder each believe the other is guilty, while their mother also believes that one is guilty. When all three meet, it would be plausible for them to talk to one another and thus reveal each one's beliefs. But if this happened, the plot twist-that neither is guilty-would be given away prematurely. So the family's reunion is staged as a silent vigil the night before the trial's last day. This convention of courtroom dramas motivates withholding information from the audience. 25
Any narrative text must repeat important story information, and in the cinema, repetition takes on a special necessity; since the conditions of presentation mean that one cannot stop and go back, most films reiterate information again and again. The nature of that reiteration can, however, vary from film to film. 26 In a film by Godard or Eisenstein, the narration overtly repeats information that may not be repeated within the story. Sequences late in October (1928) and Weekend (1967) replay events that we have seen earlier in the film, and this repetition is not motivated by character memory. But a classical film assigns repetition to the characters. That is, the story action itself contains repetitions which the narration simply passes along. For example, after the credits for the film *Housewife (1934) have concluded, the opening scene shows the heroine harassed by her domestic duties. At the scene's close, a polltaker calls on her and asks her job; 'Oh…, ' she says, '…just a housewife.' 'Housewife, ' the polltaker repeats at the fade-out. In one scene of *The Whole Town's Talking (1935), we learn a man's profession the moment he enters the room; a group of police officials greet him in a chorus:
'Warden!'
'Warden, Chief!'
'Hello, Warden.'
'Hiya, Warden.'
Such repetition is not extensive-that would be as transgressive as no repetition at all. Optimally, a significant motif or informational bit should be shown or mentioned at three or four distinct moments, as in the warden chorus. Three is in fact a mystical number for Hollywood dramaturgy; an event becomes important if it is mentioned three times. The Hollywood slogan is to state every fact three times, once for the smart viewer, once for the average viewer, and once for slow Joe in the back row. 27 Leo McCarey recalls: 'Most gags were based on “the rule of three.” It became almost an unwritten rule.' 28 Irving Thalberg is reported to have said, 'I don't mean tell 'em three times in the same way. Maybe you tell 'em once in comedy, maybe you tell 'em once directly, maybe you tell 'em next time with a twist.' 29 For a rare instance of audacious repetition in the narration rather than the story, see fig 3.21.
Since classical narration communicates what it 'knows' by making characters haul the causal chain through the film, it might seem logical to assume that the classical film commonly restricts its knowledge to a single character's point-of-view. Logical, but wrong. If we take point-of-view to be an optical subjectivity, no classical film, not even the vaunted but misdescribed Lady in the Lake (1947), completely confines itself to what a character sees. If we regard a character's point-of-view as comprising what the character knows, we still find very few classical films that restrict themselves to this degree. The overwhelmingly common practice is to use the omnipresence of classical narration to move fluidly from one character to another.
The classical film typically contains a few subjective point-of-view shots (usually of printed
-31-
matter read by a character), but these are firmly anchored in an 'objective' frame of reference. Moreover, Hollywood's optical point-of-view cutting is seldom rigorously consistent. While in one shot a camera position will be marked as subjective, a few shots later the same viewpoint may be objective-often resulting in anomalies such as a character walking into his or her own field of vision (see figs 3.22 through 3.25). In a similar fashion, classical narration will confine itself to one character's limited knowledge, but this will then be played off against what other characters know. Clever narrational twists often depend upon restricting us to one character's point-of-view before revealing the total situation. Even flashbacks, which are initially motivated as limited, subjective point-of-view, seldom restrict themselves solely to what the character could have known. For such reasons, it is accurate to describe classical narration as fundamentally omniscient, even when particular spatial or temporal shifts are motivated by character subjectivity.
The Hollywood cinema quickly mastered shifts in point-of-view. As early as *Love and the Law (1919), one can find extensive sequences of optical point-of-view cutting (see figs 16.44 and 16.45). *The Michigan Kid (1928) begins with a montage of gold prospecting in Alaska and then moves our attention to a gambling hall. At one table sits Jim Rowen, identified by an inter-title as the owner of the hall. In talking to two customers, Jim reveals that he is selling out to go back to the States and rejoin the girl he left behind. As Jim packs to leave, he stares at his tattered picture of Rose. This triggers a flashback introducing Jim as a boy, playing with Rose and fighting off the delinquent Frank. The flashback ends and dissolves into Jim's optical viewpoint of Rose's picture. At this point, however, the film widens its narrational view. There is a cut to a customer in the gambling den. He looks at his watch before offering it as a stake. Thanks to another point-of-view shot, we see Rose's picture in his watch. Thus we know before Jim does that Frank has reentered his life. A bartender takes the watch to Jim, who appraises it; we are in suspense as to whether he will notice the picture. At first he does not, which increases the tension, but then he does. As he looks at the picture, the shot superimposes his memory image of Rose as a girl, then his newspaper picture of her. He asks the barkeep to bring Frank in. Using only two expository titles, the narration has presented the essential background of the story action and has fluently moved among various degrees of subjectivity. Beginning in medias res and letting the characters reveal exposition, the classical Hollywood film thus moves to subjectivity only occasionally-something possible for a narration endowed with omniscience.
The example from *The Michigan Kid shows that classical narration can exploit omnipresence to conceal information that individual characters possess. Occasionally the classical film flaunts such suppressive operations, opening up a gap between the narration's omniscient range of knowledge and its moderate communicativeness. Consider the opening of *Manhandled (1949), which shows a man sitting in a study. The framing carefully conceals his face. His wife and her lover return, but we see only their feet. After the lover leaves, the husband follows her upstairs, his face still offscreen. He approaches his wife and starts to strangle her. The sequence seems transgressive because the narration has overtly suppressed the faces of the killer and the lover. Yet at the end of the sequence, there is a dissolve and a voice says: 'At that point the dream always ends, doctor.' The overtness of the narration is justified retroactively as subjective. The greater emphasis placed upon 'psychoanalytic' explanations of causality in the 1940s created a trend toward such occasionally explicit narration. Similarly, play with point-of-view is a minor convention of the mystery film. Through Different Eyes (1929) and The Grand Central Mystery (1942) both use flashbacks to recount the same events from inconsistent points of view. The subjective film and the mystery film can thus make narration self-conscious and overtly suppressive, but only thanks to compositional and generic motivation. Consistently suppressive narration, such as that of Jean-Marie Straub and Danièle Huillet's Not Reconciled (1964) or Alain Resnais's Providence (1977), is unknown in the Hollywood paradigm.
Classical narration, then, plunges us in medias res and proceeds to reduce signs of its self-consciousness and omniscience. The narration accomplishes this reduction by means of spatial omnipresence, repetition of story information,
-32
minimal changes in temporal order, and plays between restricted and relatively unrestricted points of view. It is in the light of these aims that we must assess the power of that celebrated Hollywood 'continuity.' Because we see no gaps, we never question the narration, hence never question its source. When, in *Penthouse (1933), the scene shifts from a nightclub to a luxury yacht and the voice of the club's bandleader continues uninterrupted, now broadcast from a radio on board the yacht, we can recognize the narration's omnipresence but we are assured that no significant story action has been suppressed. At the end of a scene, a 'dialogue hook' anticipates the beginning of the next (e.g., 'Shall we go to lunch?'/long-shot of a cafe); such a tactic implies that the narration perfectly transmits the action. Crosscutting signals omnipresence and unrestricted point-of-view, while editing within the scene delegates to the characters the job of forwarding the story action. Chapters 4 and 5 will assess how narrational concerns have shaped classical patterns of space and time. At this point, it is worth looking briefly at one technique that is seldom considered a part of narration at all.

Music as destiny
From the start, musical accompaniment has provided the cinema's most overt continuity factor. In the silent cinema, piano or orchestral music ran along with the images, pointing them up and marking out how the audience should respond. Non-diegetic music was less pervasive in the early 1930s, but the rise of symphonic scoring in the work of Max Steiner, Erich Wolfgang Korngold, Ernest Newman, et al. reasserted classical cinema's interest in using music to flow continuously along with the action. Stravinsky's comparison of film music to wallpaper is apt, not only because it is so strongly decorative but because it fills in cracks and smoothes down rough textures. 30 Filmmakers have long recognized these functions. As early as 1911, a theater musician advised players not to stop a number abruptly when the scene changed. 31 Hollywood composers claimed that sudden stops and starts were avoidable by the process of imperceptibly fading the music up and down, the practice known in the trade as 'sneaking in and out.' 32
This continuous musical accompaniment functions as narration. It would be easy to show that film music strives to become as 'transparent' as any other technique-viz., not only the sneak-in but the neutrality of the compositional styles and the standardized uses to which they are put ('La Marseillaise' for shots of France, throbbing rhythms for chase scenes). Theodor Adorno and Hanns Eisler have heaped scorn upon Hollywood music as pleonastic and self-effacing; Brecht compared film music's 'invisibility' to the hypnotist's need to control the conditions of the trance. 33 Yet calling the music 'transparent' is as true but uninformative as calling the entire Hollywood style invisible. If music functions narrationally, how does it accomplish those tasks characteristic of classical narration?
The sources of Hollywood film music show its narrational bent very clearly. In eighteenth-century melodrama, background music was played to underscore dramatic points, sometimes even in alternation with lines of dialogue. American melodrama of the 1800s used sporadic vamping, but spectacle plays and pantomimes relied upon continuous musical accompaniment. 34 The most important influence upon Hollywood film scoring, however, was that of late nineteenth-century operatic and symphonic music, and Wagner was the crest of that influence. Wagner was a perfect model, since he exploited the narrational possibilities of music. Harmony, rhythm, and 'continuous melody' could correspond to the play's dramatic action, and leitmotifs could convey a character's thoughts, point up parallels between situations, even anticipate action or create irony. Adorno's monograph on Wagner even argues that the dream of the Gesamtkunstwerk anticipated the thoroughly rationalized artifact of the culture industry, as exemplified in the Hollywood film. 35
In the early teens, film trade journals solemnly supplied theater pianists with oversimplified accounts of Wagner's practice. One pianist explained: 'I attach a certain theme to each person in the picture and work them out, in whatever form the occasion may call for, not forgetting to use popular strains if necessary.' 36 When Carl Joseph Breil proudly claimed to be the first composer to write a score for a film, he said he used leitmotifs for the characters. 37 Silent film scores, usually pasted together out of standardized
-33
snatches of operas, orchestral music, and popular tunes, adhered to the crude leitmotif idea (see fig 12.16). Early synchronized-sound films with musical tracks continued the practice: when we see the Danube, we hear 'The Blue Danube' (The Wedding March 119281). With the post-1935 resurgence in film scoring, Wagner remained the model. Most of the major studio composers were trained in Europe and influenced by the sumptuous orchestration and long melodic lines characteristic of Viennese opera. 38 Max Steiner and Miklós Rózsa explicitly acknowledged Wagner's influence, as did Erich Wolfgang Korngold, who called a film 'a textless opera.' 39 Characters, places, situations-all were relentlessly assigned motifs, either original or borrowed. When motifs were not employed, certain passages functioned as a recitative to cue specific attitudes to the scene (e.g., comic music, suspense music). 40 Brecht complained that with such constantly present music, 'our actors are transformed into silent opera singers.' 41 But Sam Goldwyn gave the most terse advice: 'Write music like Wagner, only louder.' 42
Like the opera score, the classical film score enters into a system of narration, endowed with some degree of self-consciousness, a range of knowledge, and a degree of communicativeness. The use of non-diegetic music itself signals the narration's awareness of facing an audience, for the music exists solely for the spectator's benefit. The scale of the orchestral forces employed and the symphonic tradition itself create an impersonal wash of sound befitting the unspecific narrator of the classical film. 43 The score can also be said to be omniscient, what Parker Tyler has called 'a vocal apparatus of destiny.' 44 In the credits sequence, the music can lay out motifs to come, even tagging them to actors' names. During the film, music adheres to classical narration's rule of only allowing glimpses of its omniscience, as when the score anticipates the action by a few moments. In *Deep Valley (1947), for instance, just before the convict approaches the lovers, the music swiftly turns from pleasant to sinister. As George Antheil puts it, 'The characters in a film drama never know what is going to happen to them, but the music always knows.' 45
Most important, musical accompaniment is communicative only within the boundaries laid down by classical narration. Like the camera, music can be anywhere, and it can intuit the dramatic essence of the action. It remains, however, motivated by the story. When dialogue is present, the music must drop out or confine itself to a subdued coloristic background. 'If a scene is interspersed with silent spots, the orchestration is timed so closely that it is thicker during the silent shots. It must then be thinned down in a split second when dialogue comes in.' 46 Just as classical camerawork or editing becomes more overt when there is little dialogue, so the music comes into its own as an accompaniment for physical action. Here music becomes expressive according to certain conventions (static harmony for suspense or the macabre, chromaticism for tension, marked rhythm for chase scenes). 47 A 'sting' in the music can underline a significant line of dialogue very much in the manner of eighteenth-century melodrama.
Music can also reinforce point-of-view. It establishes time and place as easily as does an inter-title or a sign: 'Rule Britannia' over shots of London, eighteenth-century pastiche for the credits of *Monsieur Beaucaire (1946). In scoring Lust for Life (1956), Rózsa modeled his score upon Debussy in order to suggest Van Gogh's period. 48 To this 'unrestrictive' use of musical narration, Hollywood counterposes the possibility of subjective musical point-of-view. The music often expresses characters' mental states-agitated music for inner turmoil, ominous chords for tension, and the like. In The Jazz Singer (1927), we know Jakie is thinking of his mother when, as he sees her picture, we hear the 'Mammy' tune in the score. During the spate of subjective films of the 1940s, musical experiments increased (the theremin in Spellbound [1945], a playback reverberation in Murder, My Sweet [1944]). As one critic noted at the time, weird coloristic effects became more common because of 'the vogue for films dealing with amnesia, shock, suspense, neurosis, and kindred psychological and psychiatric themes. The music counterpart of the troubled mental states depicted in these films is a musical style which emphasizes vagueness and strangeness, especially in the realms of harmony and orchestration.' 49 By the mid-1930s, music could shift easily from unrestrictive to restrictive viewpoints, as when a character hums a tune to himself and then, as he steps outdoors, the orchestra takes it up. 50 Hollywood music could
-34
even create misleading narration, as in *Uncertain Glory (1944): when the prisoner Jean tells Bonet he wants to go to church to confess, the music is sentimental, but once Bonet lets him go, Jean flees and the music becomes flippant. The first musical passage is now revealed as having presented only Bonet's misconception about Jean's sincerity. Such practices, even such deceptions, are the logical consequence of making music-as-narration dependent upon character causality.
Since classical narration turns nearly all anticipations and recollections of story action over to the characters, music must not operate as a completely free-roaming narration. Here is one difference from Wagner's method, which did allow the music to flaunt its omniscience by ironic or prophetic uses of motifs. The Hollywood score, like the classical visual style, seldom includes overt recollections or far-flung anticipations of the action. The music confines itself to a moment-by-moment heightening of the story. Slight anticipations are permitted, but recollections of previous musical material must be motivated by a repetition of situation or by character memory. At the close of *Sunday Dinner for a Soldier (1944), Tessa's wave to Eric is accompanied by the ballroom music to which they had danced in an earlier scene. The classical text thus relies upon our forming strong associations upon a motifs first entry.
The narrational limits which the classical film puts upon music are dramatically illustrated in Hangover Square (1945). During the credits, a romantic piano concerto plays non-diegetically but does not conclude. Early in the film, when the composer George Bone goes to his apartment, his friend Barbara is playing the opening of his concerto, the same music we had heard over the credits. But Bone's version is also unfinished, and Barbara's father advises him to complete it. In the course of the action, Bone is plagued by murderous amnesiac spells triggered by discordant noises, which are rendered as subjective by means of chromatic and dissonant harmonies. Completing the concerto drives these from Bone's head, but in the film's climactic scene, when he plays the concerto at a soiree, he suffers another breakdown. Yet the performance continues, and the action of the last scene is accompanied throughout by Bone's concerto. Bone's romantic score wins out over the psychotic discordances, but only by becoming identical with the score of the film, the score that had been 'rehearsed' under the credits. The narration's power lies in the fact that Bone is allowed to score the last scene only by writing the score that the narration 'had in mind' all along. The narration's limits are revealed by its almost complete anticipation of Bone's concerto: the film cannot complete the piece before he does. Only the conclusion of the action-Bone finishing the performance alone in a burning building-brings the concerto and the film itself to a close. As 'The End' appears on the screen, the (non-diegetic) orchestra swallows the solo piano; now the narration can have the last word, and chord.

The reappearing narration
The finale of Hangover Square also illustrates the way in which the narration can reappear overtly but briefly at the film's very close. This close would minimally consist of a 'The End' title, usually against a background identical to that of the opening credits, and a non-diegetic musical flourish. Such devices buckle the film shut, making the 'narrator' simply a discreet curtailer, like the curtain that closes a play or 'The End' that concludes a novel. This narrational movement toward finality is laid bare in the credits of King Kong (1933). The opening credits are set against a triangular shape which steadily narrows as they proceed (see figs 3.26 and 3.27). Not until the end credit does the triangle diagram a complete closure (see fig 3.28). 51 After about 1970, it seems, films seldom exploited these narrational possibilities and instead dropped the 'The End' credit, shifted most of the opening credits to the final spot (as a signal of the end), and expanded the credits sequence to a Talmudic intricacy.
The narration can afford to be so modest at this point because the film has already informed the audience when it will end. Chapter 4 shows how deadlines work in this fashion. Characters also constantly look forward to closure. In *The Arkansas Traveler (1938), Traveler tells John: 'When this is all over, I want you to remember just one thing.' In the final moment of *Play Girl (1941), the heroine calls her maid to fetch the perfume she has worn for every flirtation: 'The
-35-
last time, Josie, the last time.' *Uncertain Glory (1944) ends with Jean about to sacrifice his life. Bonet: 'It's been a long road.' Jean: 'But it's come to the right ending.' The conditions for closure have also been non-diegetically anticipated by the narration. *The Shock Punch (1925) begins with expository titles that describe Dan Savage as a man who believes that life is a battle and the winner is one who 'can command the last reserve of physical power.' The next title continues: 'And as he wanted his son Ranny to be like that-to carry a final, deciding punch into every conflict-.' Needless to say, the film's action is resolved when Ranny flattens the man he is fighting. At the start of *The Black Hand (1950), a crawl title tells of Italian immigrants living in New York at the turn of the century. Most were good citizens, the narration explains, who fought the Black Hand and eventually purged their community of its influence. The title thus anticipates Gio's success in overthrowing the Mafia. At the film's close, a fireman mutters, 'Ah, these dagoes!' and the captain turns. 'I wonder where you think Americans come from.' His retort confirms the narration's initial estimate of the immigrants' civic virtues. In contrast, it is no trivial description of an avant-garde or modernist film to say that such films often do not let us know when they will stop. Films in these traditions deliberately exploit a sense of uncertainty about their boundaries, as when, in Last Year at Marienbad (1961), the narrator announces that The whole story has come to its end, ' but neglects to add that the film is only half over.
The work of classical narration may also peep out from the film's epilogue-a part of the final scene, or even a complete final scene, that shows the return of a stable narrative state. The screenwriter Frances Marion suggests ending the film as soon as possible after the action is resolved, but 'not before the expected rewards and penalties are meted out…. The final sequence should show the reaction of the protagonist when he has achieved his desires. Let the audience be satisfied that the future of the principals is settled.' 52 Emerson and Loos call this a short 'human interest' scene, an equivalent of 'And so they lived happily ever after.' 53 All the films in the UnS did include an epilogue, however brief; in two-thirds of them, the epilogue was a distinctly demarcated scene. A 1919 film *Love and the Law (1919), signalled its epilogue by a very self-conscious title: 'Patience, gentle audience, just one thing more.' Soon, however, no such cues were necessary and an epilogue could be included as a matter of course.
Epilogues will often tacitly refer back to the opening scene, proving the aptness of Raymond Bellour's remark that in the classical film the conclusion acknowledges itself as a result of the beginning. 54 *You for Me (1952) begins with Tony being peppered in the buttocks by a shotgun blast; a freeze frame catches him in a comic posture. The film ends with him sitting down on a knitting needle, accompanied by a freeze frame. *Sunday Dinner for a Soldier (1944) frames its story by the habitual action of the family waving to planes overhead; at the start, the planes are anonymous, but by the close, Tessa is in love with one pilot. The familiar here-we-go-again, or cyclical, epilogue is a variant of the same principle. The epilogue can even be quite self-conscious about its symmetry, as is the framing narration of *Impact (1944). The opening of the film corresponds to the opening of a dictionary by an anonymous hand, and the word 'impact' is enlarged. A voice-over commentary reads the somewhat improbable definition: 'Impact: The force with which two lives come together, sometimes for good, sometimes for evil.' At the end, the epilogue returns to the dictionary, but the definition has changed: 'Impact: The force with which two lives come together, sometimes for evil, sometimes for good.' The restoration of 'good' as the stable state creates an explicit balancing effect, as does shutting the book to announce the close of the film.
Most classical films use the story action to confirm our expectations of closure without further nudgings from the narration. But *Impact does show that during the last few seconds of the film, the narration can risk some self-consciousness. The familiar running gag, a motif repeated throughout the film to be capped in the final moments, reminds the audience to some degree of the arbitrariness of closure. Another self-conscious marking of the narration's perspective upon the story world is the camera that cranes back to a high angle upon a final tableau. Most overt is a finale like that of *Appointment for Love (1941), in which an elevator man turns from the couple and winks at the audience. As we would expect, such direct address is usually motivated
-36-
by genre (e.g., comedy) or realism (as in a frame story stressing the factual basis of the fiction).

The winding corridor
The belief that classical narration is invisible often accompanies an assumption that the spectator is passive. If the Hollywood film is a clear pane of glass, the audience can be visualized as a rapt onlooker. Again, Hollywood's own discourse has encouraged this. Concealment of artifice, technicians claim, makes watching the film like viewing reality. The camera becomes not only the storyteller but the viewer as well; the absent narrator is replaced by the 'ideal observer.' 55 Few theorists today would agree with Hollywood's equation of its style with natural perception, but contemporary accounts have still considered the spectator to be quite inactive. Most commonly, film theorists have employed concepts taken from perspective painting to explain the spectator's role. Yet terms like 'spectator placement, ' 'subject position, ' and other spatial metaphors break the film into a series of views targeted toward an inert perceiver. 56 In Chapter 5, I will consider 'perspective' as an account of the representation of classical space. For now, a metaphor involving both space and time will be useful. The spectator passes through the classical film as if moving through an architectural volume, remembering what she or he has already encountered, hazarding guesses about upcoming events, assembling images and sounds into a total shape. What, then, is the spectator's itinerary? Is it string-straight, or is it more like the baffling, 'crooked corridors' that Henry James prided himself upon designing? 57
The film begins. Concentrated, preliminary exposition that plunges us in medias res triggers strong first impressions, and these become the basis for our expectations across the entire film. Meir Sternberg calls this the 'primacy effect.' 58 He points out that in any narrative, the information provided first about a character or situation creates a fixed baseline against which later information is judged. As our earlier examples indicate, the classical cinema trades upon the primacy effect. Once the exposition has outlined a character's traits, the character should remain consistent. This means that actions must be unequivocal and significant. 59 The star system also encourages the creation of first impressions. 'The people who act in pictures are selected for their roles because of the precise character impressions that they convey to audiences. For instance, the moment you see Walter Pidgeon in a film you know immediately that he could not do a mean or petty thing.' 60 All of these factors cooperate to reinforce the primacy effect.
Many films open with dialogue that builds up an impression of the yet-to-be-introduced protagonist; when the character appears (played by an appropriate star, caught in a typical action), the impression is confirmed. In the first scene of *Speedy (1928), the young woman says that Speedy (Harold Lloyd) has a new job; her father comments that Speedy cannot keep any job because he is obsessed by baseball. Scene two begins with an expository title identifying the crucial game being played in Yankee Stadium, and shots of the game follow. Another expository title informs us that Speedy now works where he can phone the stadium. We then see a soda fountain, with Speedy as the soda jerk, going to the phone to learn the game's score. The rest of the scene confirms Pop's judgment of Speedy's character through gags showing Speedy carrying his baseball mania into his work. Dialogue title, expository title, character action, and star persona (Harold called himself 'Speedy' in The Freshman [1925]) all reinforce a single first impression.
The primacy effect is not confined to characterization, although first impressions are probably most firm in that realm. In some silent films, an unusually emphatic narration previews the essential theme and establishes the most coherent reading of what will follow. By extension, all the devices of 'planting' and foreshadowing motifs-objects, conditions, deadlines-gain their saliency from the primacy effect.
Once first impressions get erected, they are hard to knock down. Sternberg shows that we tend to take the first appearance of a motif as the 'true' one, which can withstand severe testing by contrary information. When, for instance, a character first presented as amiable later behaves grumpily, we are inclined to justify the grumpiness as a temporary deviation. 61 This tactic (again, reinforced by the star system) is a common way in which the classical film presents character change or development. In the opening
-37
of *The Miracle Woman (1931), Florence Fallon is so distraught by her father's death that she denounces his congregation as hypocrites and launches into a sermon on the need for kindness. An opportunistic promoter takes advantage of her fervor and talks her into getting revenge on people by becoming a phony faith healer. When we next see her, she behaves cynically. Because of first impressions, we see her cynical selfishness as a momentary aberration, caused by exceptional circumstances, and so we are not surprised when love recalls her to her father's ideals. The primacy effect helps explain why character change in the Hollywood film is not a drastic shift but a return to the path from which one has strayed.
First impressions in place, the spectator proceeds through the film. How does this process work? The narration creates gaps, holding back information and compelling the spectator to form hypotheses. Most minimally and generally, these hypotheses will pertain to what can happen next, but many other hypotheses might be elicited. The spectator may infer how much a character knows, or why a character acts this way, or what in the past the protagonist is trying to conceal. The viewer may also hypothesize about the narration itself: why am I being told this now? why is the key information being withheld? Sternberg sees every viewing hypothesis as having three properties. A hypothesis can be more or less probable. Some hypotheses are virtual certainties (e.g., that Bill will survive the flood in *Steamboat Bill Jr. [1928]). Other hypotheses are highly improbable (e.g., that Bill will not get the girl he loves). Most hypotheses fall somewhere in between. Hypotheses can also be more or less simultaneous; that is, sometimes we hold two or more hypotheses in balance at once, while at other moments one hypothesis simply gets replaced by another. If a man announces that he will get married, we hold simultaneous hypotheses (he will go through with it or he won't). But if a sworn bachelor suddenly shows up with a bride on his arm, the bachelor-hypothesis is simply replaced; the bachelor-hypothesis never competed with another possibility. Evidently, simultaneous hypotheses promote suspense and curiosity, while successive hypotheses promote surprise. Finally, a set of hypotheses can be more or less exclusive. Narration may force us to frame a few sharply distinguished hypotheses (in a chess game, there can only be win, lose, or draw), or the film may supply a range of overlapping and indistinct possibilities (setting out on a trip, one may undergo a wide variety of experiences). 62
The three scales of probability, simultaneity, and exclusivity take us a considerable way toward characterizing the activities of the classical spectator. Broadly speaking, Hollywood narration asks us to form hypotheses that are highly probable and sharply exclusive. Consider, as a naive example, *Roaring Timber (1937). In the first scene, a lumber-mill owner comes into a saloon looking for a new foreman. He tells the bartender he needs a tough guy for the job. Since we have already seen our protagonist, Jim, enter the bar, we form the hypothesis that the owner will ask him. The expectation is fairly probable, and there is no information to the contrary (no other man in the room is identified as a candidate). There is also a narrow range of alternatives (either the owner will ask Jim or he will not). Few hypotheses are as probable as this, but one of the indices of classical narration's reliability is that it seldom equivocates about the likeliest few hypotheses at any given moment. Similarly, the classical film sharply delimits the range of our expectations. The character's question is not 'What will I do with my life?' but 'Will I choose marriage or a career?' Even subtle cases operate by the same principles. *Beggars of Life (1928) begins with a wandering young man coming up to a farmhouse and finding a dead man inside. He then encounters a young woman who tells, in flashback, how the farmer tried to rape her and how she killed him. The alternative explanations (suicide, accident, homicide, etc.) narrow to a single one (self-defense), and this becomes steadily more probable as the woman's tale accounts for the details the young man had noticed. True, farcical forms of comedy permit almost anything to happen next, but there the improbability and open-endedness of permissable hypotheses are motivated as generic conventions, and we adjust our expectations accordingly. On the whole, classical narration creates probable and distinct hypotheses. Characters' goal orientation often reinforces and guides the direction these hypotheses will take. Incidentally, in *Roaring Timber, Jim accepts the foreman's job.
By threading together several probable and quite exclusive hypotheses, we participate in a
-38-
game of controlled expectation and likely confirmation. There is, however, more to the spectator's activity. Any fictional narration can call our attention to a gap or it can distract us from it. In a mystery film, for instance, the crucial clue may be indicated quite casually; the detective may notice it but we do not. If the narration thus distracts us, we do not form an appropriate hypothesis and the narration can then introduce new information. These successive hypotheses, as Sternberg calls them, create surprise. 63 Now it is characteristic of classical narration to use surprise very sparingly. Too many jolts would lead us to doubt the reliability of the narration, and the advantages of concentrated, preliminary, in medias res exposition would be lost. In our itinerary through the classical film, the banister cannot constantly collapse under our touch.
For this reason, classical narration usually calls our attention to gaps and allows us to set up simultaneous, competing hypotheses. The scenes from *Roaring Timber and *Beggars of Life afford clear instances, as does a sequence in *Interlude (1957). The heroine calls on the conductor Tonio Fischer; our knowledge of him has been identical with hers. While she waits for him, the narration takes us to another room, where Tonio is playing the piano for another woman. The scene raises questions about the woman's identity and Tonio's character traits, and these gaps encourage us to construct simultaneous alternatives to be tested in subsequent scenes.
Our hypothesis-forming activity can be thought of as a series of questions which the text impells us to ask. The questions can be posed literally, from one character to another, as in the beginning of *Monsieur Beaucaire (1946): 'Will there be war?' Or the questions can be more implicit. Roland Barthes speaks of this question-posing process as the 'hermeneutic code' and he shows how narratives have ways of delaying or recasting the question or equivocating about the answer. 64 The classical cinema always delays and may recast, but it seldom equivocates. At the start of *Play Girl (1941), we are uncertain whether Grace is a gold-digger or whether the title is ironic. But when the father of her current beau denounces her, not only does she not deny her scandalous past but she accepts a bribe to let the son go. The answer to our question, somewhat delayed, is unequivocal.
All of the foregoing instances illustrate another feature of the gaps that classical narration creates: they are filled. Sternberg distinguishes between permanent gaps, which the text never authoritatively lets us fill (e.g., lago's motives), and temporary gaps, which sooner or later we are able to fill. 65 It is a basic feature of classical narration to avoid permanent gaps. 'The perfect photoplay leaves no doubts, offers no explanations, starts nothing it cannot finish.' 66 The questions about Tonio in *Interlude are eventually answered. Concentrated preliminary exposition, causal motivation, the use of denouement and epilogue-all seek to assure that no holes remain in the film. This process of gap-filling helps create the continuity of impression upon which Hollywood prides itself. Each sequence, every line of dialogue, becomes a way of creating or developing or confirming a hypothesis; shot by shot, questions are posed and answered. Our progress through the film, as our first impressions are confirmed and our hypotheses focus toward certainty, resembles the graphic design in the titles of King Kong (figs 3.26-3.28): a pyramid narrowing to a point of intelligibility. One screenplay manual puts it well: 'In the beginning of the motion picture we don't know anything. During the course of the story, information is accumulated, until at the end we know everything.' 67
Again, one should not conclude that classical narration is naive or shallow, for subtle effects can be achieved within the admittedly constrained bounds of such narration. *Wine of Youth (1924) begins with three expository titles:
When our grandmothers were young, nice girls pretended to know nothing at all.
When our mothers were young, they admitted they knew a thing or two.
The girls of today pretend to know all there is to know.
There follow two parallel scenes. At a ball in 1870, a suitor proposes to a woman, and she accepts: 'There has never been a love as great as ours!' At another dance in 1897, a suitor proposes to the couple's daughter, and she too accepts, repeating the line her mother had uttered years before. The symmetry is quite exact: similar situations, same setting (a sofa in an alcove), even
-39-
the identical number of shots in each scene. At this point, the narration has established itself as highly reliable: the scenes have confirmed the titles' knowledge of women, and we have already formed strong first impressions about what the 'girls of today' will be like. (The word 'pretends' strongly suggests omniscience.) When the scene moves to the present, our impressions are confirmed. Jazz babies and lounge lizards are engaged in a wild party. Mary, the granddaughter and daughter of the other two women, refuses to marry her suitor. We form a hypothesis that this will not in the long run violate the pattern established in the first two scenes. Over the whole film we wait for Mary to reconcile herself to the decent young man who loves her. A harrowing family crisis demonstrates both the strains and the possibilities of marriage. Mary and her suitor are sitting on the sofa (the site of both previous courtships) and he proposes. She accepts: 'There has never been a love as great as ours!' It has been a long wait, but the narrational gap has finally been closed, and by an ironic repetition at that. The narration can even afford a twist-embracing, the couple tumble off the sofa-that lends a small surprise to the finale. Our hypotheses about the conclusion, established as very narrow and highly probable, are tested but finally validated, and in a way that also illustrates the recurrence of the Rule of Three.
There is one genre that may seem to run counter to all these claims about spectator activity in classical narration. The mystery film sometimes makes its narration quite overt: a shot of a shadowy figure or an anonymous hand makes the viewer quite aware of a self-conscious, omniscient, and suppressive narration. Similarly, the mystery film encourages the spectator to erect erroneous first impressions, confounds the viewer's most probable hypotheses, and stresses curiosity as much as suspense. (The mystery tale always depends upon highly retarded exposition, the true account coming to light only at the end.) The narration may even be revealed as retrospectively unreliable. Thus The Maltese Falcon (1941) offers an interesting contrast with *Wine of Youth. Not only does the narration abandon its initial adherence to Sam Spade's point-of-view by showing the killing of his partner Archer, but the narration also declines to show the killer (we see only a gloved hand). More important, the narration misleads us in an expository title at the very outset. Over a still-life of the Maltese falcon, the title recounts the statuette's origin and ends by remarking that its whereabouts remain a mystery 'to this day' (fig 3.29). When the characters find only a lead replica of the falcon, the opening title stands revealed as doubly misleading. The falcon in the still-life may be the phony, and the phrase 'to this day' which we might take as meaning 'until this story started, ' actually means 'even after the story concluded.' The opening title's equivocation is apparent only in retrospect. The same kind of misleading narration is at work in the beginning of * Manhandled, as I've already suggested (p. 32). A more drastic example, probably a limit-case, is Hitchcock's duplicitous flashback in the beginning of Stage Fright (1950).
The unreliable and overt narration of the mystery film remains, however, finally bound by classical precepts. First, the narration still depends chiefly upon suspense and forward momentum: the story is primarily that of an investigation, even if the goal happens to be the elucidation of a past event. Secondly, the mystery film relies completely upon cause and effect, since the mystery always revolves around missing links in the causal chain. Third, those links are always found, so even the gaps of the mystery film are temporary, not permanent. Most important, the mystery film's overt play of narration and hypothesis-forming is generically motivated. Since Poe and Doyle, the classical detective story has stressed the game of wits that the narrator proposes to the reader. In this genre, we want uncertainty, we expect both characters and narration to try to deceive us, and we therefore erect specific sorts of first impressions, cautious, provisional ones, based as much upon generic conventions as upon what we actually learn. We do not feel betrayed by the Falcon's opening title, since it is equivalent to the deceptive but 'fair' narrational manipulations in certain novels by Agatha Christie, John Dickson Carr, or Ellery Queen. The classical film thus can generically motivate an unreliable and overt narration.
The spectator moves through, or with, classical Hollywood narration by casting expectations in the form of hypotheses which the text shapes. Narration is fundamentally reliable, allowing hypotheses to be ranked in order of probability
-40-
and narrowed to a few distinct alternatives. Surprise and disorientation are secondary to suspense as to which alternatives will be confirmed. Curiosity about the past takes a minor role in relation to anticipation of future events. Gaps are continually and systematically opened and filled in, and no gap is permanent. Lest this process seem obvious or natural, recall such a film as Last Year at Marienbad (1961), which creates a fundamentally unreliable narration, a lack of redundancy, an open and relatively improbable set of hypotheses, a dependence upon surprise rather than suspense, a pervasive ambiguity about the past that makes the future impossible to anticipate, and many gaps left yawning at the film's close. This is of course an extreme example, but other narrative films contain non-classical narrative strategies. A film's narration could make the initial exposition less clear-cut, as does Godard's Sauve qui peut (la vie) (1980), or the narration could establish a firm primacy effect but then qualify or demolish it, as do films as different as Carl Theodor Dreyer's Day of Wrath (1943) and Resnais's Providence (1977). The Hollywood film does not lead us to invalid conclusions, as these films can; in the classical narrative, the corridor may be winding, but it is never crooked.