MATT HAUSKE

OZU, SOUND AND STYLE: A CINEMETRICAL ANALYSIS OF FOUR FILMS

The influence of Hollywood film style on Ozu is well documented.  Several scholars have demonstrated Ozu’s love for and extensive knowledge of contemporary Hollywood cinema.[1]  Kristin Thompson quotes an article by Ozu in which he says he could only remember having seen three Japanese films by the time he applied for a job at Shochiku in 1923: his knowledge of film was almost entirely Western and, specifically, American.[2]  David Bordwell identifies “three American mentors” who “had raised narrative unity to heights of finesse” that Ozu sought to emulate and that influenced his entire career: Charlie Chaplin, Ernst Lubitsch, and Harold Lloyd.[3]  He goes on to cite Chaplin’s A Woman of Paris (1923), Lubitsch’s The Marriage Circle (1924), and Lloyd’s The Freshman (1925) and Speedy (1928), along with the Lloyd persona in general, as major influences for Ozu, in terms of both specific films and Ozu’s overall structural approach to narrative and plot geometry.[4]  Bordwell also notes Ozu’s origins in nansensu, having apprenticed under Tadamoto Okubo in that genre, which also relates to American silent cinema, particularly the combination of body-based slapstick and social satire in the Hollywood comedies of Chaplin and Lubitsch, who used overt narration to create satiric social commentary.  Ozu drew his “zanier gags” from Lloyd, particularly in his silent college comedies.[5]  This comic impulse did not end when Ozu abandoned nansensu and college comedies and moved into the family or home drama, but, according to Bordwell, could be felt throughout his career in the consistently “ludic, open-ended narration” exhibited in all of his films.[6]

While Bordwell’s analysis of Ozu’s Hollywood influences helps to explain the often gag-like structure of his individual films, these influences can be examined at a more fundamental structural level.  Although Bordwell’s close analysis of metric editing in The Only Son and End of Summer does reveal Ozu’s attention to detail and desire to “create a subliminal norm” for the spectator and “instill a set of tacit expectations in the perceiver,” his emphasis is on the way Ozu uses the individual shot as the vehicle for cinematic communication, rather than on how he moves from shot to shot and how shots and sequences flow together.[7]  Bordwell writes,

these metric patterns, as the examples show, are intimately governed by what the shots show.  Ozu gives us time to see everything.  And by making images durationally equivalent he strengthens their identity as distinct pictorial units.  Indeed, his attention to the image made him subordinate everything to the shot as a composition.[8]

Bordwell’s argument is persuasive: Ozu does subordinate every constituent element of his shot to its overall composition and its place in the final editing scheme.  Though the conclusions he draws about Ozu’s style, illustrated by examples from The Only Son and End of Summer, are incisive and illuminating, we can do more to interrogate how and, possibly, why Ozu’s editing style did or did not change from film to film, genre to genre, or from his silent period to his sound period.  The latter will be the focus of this paper.

Before we determine what Ozu’s editing style was before he switched to sound, we must first identify a norm against which to define Ozu.  Hollywood style will serve as our relative norm, for two reasons.  First, other than Ozu’s films, there is no data on contemporary Japanese style: norms for Japanese films have not yet been determined, and because so few Japanese films have survived (particularly from the 1920s and 1930s) any norms that did exist may well be unrecoverable.  Second, Ozu was heavily influence by Hollywood, as were many of his contemporary Japanese directors.  Ozu was a member of the Film Science Research Group, a gathering of Japanese directors who would literally dissect American films for information on shot length and cutting patterns.[9]  Bordwell writes that in Japan, “scenarists were urged to study Hollywood scripts, and cinematographers counted and timed shots as they watched American films.”[10]  Ozu himself placed so much importance on editing rhythms that he had a special stopwatch built that registered both seconds and frames of film, allowing him to intricately control the duration of each shot when filming the scene and when doing the editing.[11]  Fortunately, we have a large (and growing) set of data from American and Hollywood films of the 1910s, 20s, 30s, and beyond with which to establish a general set of norms and trends against which Ozu’s style can be compared and defined.  Very generally, what is “typical” of Hollywood editing style?

Although the specific films of Chaplin, Lubitsch, and Lloyd cited by Bordwell have not yet been measured, we do have data on several Chaplin films from the teens and a Lubitsch film from the late 20s, and we can extrapolate from these relatively early works (as well as from other early Hollywood films) the patterns that would eventually become classical Hollywood style, of which we have measurements from several canonical examples from the 1930s and 1940s.  Chaplin’s His New Job (1915) provides one example.  The film consists mainly of a series of gags set in a talent agency and on a film set.  According to the notes and the graph of the measurement, the editing tends to follow the logic of each constituent gag rather than being organized around a larger narrative that spans the entire film.  Figure 1 shows the graph of His New Job:

Figure 1

Figure 1: His New Job (Charlie Chaplin, 1915).
Yellow = pictorial image. Red = dialogue intertitle. Orange = expository intertitle.  Measured by Torey Liepa.

http://www.cinemetrics.lv/movie.php?movie_ID=253

A few notes about how to read a CineMetrics graph are in order. 

1)  Each shot of the film is represented by a vertical bar starting at the top and descending according to the length of the shot.  The longer the duration of the shot, the lower the bar descends.  The x-axis represents the duration of the film.  In this film of about 30 minutes there are a total of 175 shots, for an average shot length (ASL) of 10.3 seconds.  The shortest shot is 0.6 seconds, and the longest is 102.9 seconds.

2)  This measurement includes both images and intertitles.  Intertitles cannot be neglected when measuring a silent film, as they are an integral component of a film’s editing meter and combine with the images to create the film’s visual rhythm.  Ozu was undoubtedly aware of this, as were his “American mentors.”[12]  His New Job was measured for three different kinds of shots, as represented by the three different colors of vertical bars: Yellow represents a pictorial image (actors, sets, etc.); orange represents an expository intertitle; and red represents a dialogue intertitle.[13]  This is a measurement scheme developed by Torey Liepa, who measured this film.  His New Job features 152 pictorial shots, 9 expository intertitles, and 14 dialogue intertitles.  More information on these statistics can be found on the website.

3)  The red, snaking line moving across the middle of the graph is the film’s polynomial trendline, which indicates the general tendency of the speed of the editing throughout the course of the film.  The trendline can display varying degrees of sensitivity to the general editing tendencies of the film, ranging from 1 (least sensitive) to 12 (most sensitive).  The trendline of the graph for His New Job has a degree of 6, which tends to be the most common trendline degree used because it strikes a nice balance between overall trends and specific ASL variations.[14] At a degree of 1, the trendline for any film has no curve, but is rather a straight line that can do one of three things: ascend (indicating a general increase in editing speed), descend (indicating a decrease in editing speed), or remain flat (indicating that the editing speed stays relatively constant throughout the duration of the film).  In Figure 1, the trendline shows a significant degree of variance over the course of the film.  In His New Job, the trendline at 1 moves down from 9 seconds at the beginning to about 11.5 seconds at the end, indicating that as the film proceeds, the editing speed tends to decrease (alternatively, we can say the average shot length increases).

4)  In Figure 1, the graph is slightly truncated.  Several bars extend beyond the bottom of the graph because their duration is greater than 30 seconds.  The entire graph can be seen on the website if the vertical resolution is set to 5 pixels/sec and the height is set to 750 pixels.  Several other adjustments can be made on the website, including altering the degree of the trendline and, most important for Ozu’s silent films and other films with shot totals that approach and exceed 1000, the “Step” can be changed from “Auto” to “1”, which will display a more complete and legible record of each individual shot in the film.

Although the editing of His New Job is not necessarily organized around an overarching narrative logic, the trendline is typical of the classical Hollywood style in one very important respect: the radical variance in cutting speed over the course of the film.[15]  The film starts a bit slow in terms of editing speed: the ASL at about 2 minutes into the film is about 12 seconds per shot.  The editing then accelerates radically, as a flurry of shorter shots decrease the ASL so that about 8.5 minutes into the film the ASL is close to 7 seconds per shot.  The editing then slows down as the film progresses, dropping down below 12 seconds at the 20-minute mark, before rebounding slightly for what is presumably a climactic finale, and ending with an ASL of about 10.6 seconds.  At only three or four points during the course of the film does the film’s actual ASL match its overall ASL (10.3 sec).  The rest of the time the film’s editing deviates radically from the average, ranging from 7 to 12 seconds.  Furthermore, the standard deviation of the shot durations, which we will call the film’s cutting swing, is 12 seconds (twenty percent higher than the ASL), meaning that a majority of the film’s shots fall somewhere between 0 and 22 seconds in length.[16]  Cutting swing will prove an important factor in analyzing Ozu’s films, especially in comparison with his American mentors.

Though Chaplin’s His New Job, made in 1915, is a rather early film to be using in a comparison with Ozu, who made his first film in 1927, it is, unfortunately, the latest silent Chaplin film that has yet been measured using CineMetrics.[17]  Even so, other more canonical American silent films do tend to follow the pattern set by His New Job, which features a wide range of shot lengths and a dynamic cutting profile that tends to either increase or decrease through the course of the film.  For instance, D.W. Griffith’s The Birth of a Nation (1915) and Intolerance (1916) feature marked decreases in ASL as they progress, even though they feature relatively small cutting swings.  The Birth of a Nation begins at around 9 seconds per shot and ends up at 5, and Intolerance starts at about 7.5 and increases to 4.5.  As Griffith’s style is considered prototypical for what would become classical Hollywood style, his films provide paradigmatic examples of general Hollywood tendencies that can be compared to Ozu’s style.  Later silent Hollywood comedies, such as Buster Keaton’s 1927 The General, may provide a better data set against which to compare Ozu’s silent comedies.  Also currently available on the CineMetrics website for comparison with Ozu is Ernst Lubitsch’s The Love Parade (1929) (Figure 2).  Although this is a sound film, made five years after the influential The Marriage Circle, it does provide a marked contrast to Ozu’s later sound films, particularly when viewed with a higher, more sensitive trendline.  The marked changes in editing tempo over the course of the film are underscored by the film’s high cutting swing: with an overall ASL of 13.3 seconds (475 shots over 105 minutes 18 seconds), the cutting swing is 18.4, resulting in a wide range of durations for the majority of the film’s shots (from 0 to 31.7 seconds).

Figure 2

Figure 2. The Love Parade (Ernst Lubitsch, 1929).
Trendline 8.  Measured by Charles O’Brien.
http://www.cinemetrics.lv/movie.php?movie_ID=197

Hollywood style tends to feature radical changes in ASL over the course of a single film, which includes a wide range of shot durations.[18]  But, despite the influence Hollywood films unquestionably had on Ozu, the cinemetric profiles of his films do not follow a characteristically ‘Hollywood’ cinemetric pattern.  Ozu’s films have remarkably flat trendlines at nearly all degrees of sensitivity, and this tendency is evident in both his silent and his sound films.  Because Bordwell has so thoroughly examined the relation of Ozu’s découpage style with that of Hollywood, my focus will not be on the differences between Ozu’s style and what effect the transition from silent to sound filmmaking had on his editing style.  While further cinemetrical analysis could be done on the relation between Ozu and his favorite Hollywood films that would round out or perhaps even contradict some of Bordwell’s findings, my goal here was to establish Hollywood editing rhythms as a norm from which Ozu deviated, similar to Bordwell’s own tactic.  Hollywood style will come back later as part of a speculation as to the effect (or intended effect) of Ozu’s idiosyncratic editing.

The Transition to Sound and the Effect on ASL

The adoption of sound in the cinema is perhaps the most radical technological change in the medium’s stylistic history.  Average shot length in American films immediately doubled because of the transition to sound, for a number of very practical reasons, primarily because it became much more difficult to edit both a visual and a sound track.[19]  Scenes were shot with multiple cameras simultaneously from a variety of angles so that some cutting within the scene could still be done, but the final cut of the image track had to correspond exactly with the single sound recording that had been made.  It was more economical to shoot long takes of dialogue than continue with fast editing and shot/ reverse shot style featuring one line of dialogue per shot, a style that had been conventionalized during the silent era.  A side effect of this was the rise of the musical, which could feature long takes of the film star singing.  This is borne out by the measurement of The Love Parade, in which Charles O’Brien distinguishes between ‘story’ shots and ‘song’ shots: the ASL of the ‘song’ shots (20.8) is almost twice that of the ‘story’ shots (11.2).  Another dialogue-related reason for the increase in ASL was the absence of intertitles.  Even without shot/ reverse shot, dialogue scenes that were shot with two or more characters in one frame no longer had to be interrupted by intertitles after each actor said their line.  Scenes could be played out between multiple actors for a longer time without cutting away to a reaction shot or an intertitle.

The doubling of ASL as a result of the transition to sound was not restricted to the American cinema but appears to have been a universal phenomenon.[20]  Ozu was not immune to the effect.  The ASL of An Inn in Tokyo (1935, 4.5 sec ASL), Ozu’s last silent film, is half that of The Only Son (1936, 9.1 sec ASL), his first sound film.[21]  By this time in America, the impediments to increased cutting speed in sound film seem to have been relatively overcome as a result of further technological innovations in sound recording and editing, though ASLs of sound films still did not match those of the silents.[22]  Nevertheless, Ozu’s first foray into sound film features relatively slow cutting, and his films do not speed up again until the 1950s, by the end of which he adopts an almost uniform ASL of 7 seconds.[23]  Throughout the late 1930s and 1940s, Ozu’s cutting rate remained over 9, with a few exceptions.[24]  Bordwell mentions Ozu’s increased ASL in his early sound period, but he does not go far enough in explaining how Ozu went from consistently fast editing to inordinately slow editing (14.8 seconds for There Was a Father (1942)).[25]  A more thorough analysis of Ozu’s films than is possible in this essay might shed light on the reasons behind this, and measurements of more films from this period would allow us to determine the speed of Ozu’s films relative to those of his contemporaries.

One possible explanation relates to the differences between An Inn in Tokyo and The Only Son, and deals with the absence of intertitles, which clearly affected Ozu’s films and all films in the silent era.  Ozu’s intertitles carry special significance for two reasons: his genre of choice and their relation to the creation of a national cinema.  Working in family dramas for most of his career, his films tend to be heavy with dialogue rather than physical action, requiring more intertitles than a more action-oriented genre like slapstick would.  At the same time, the Shochiku studio’s decision to enter the international market by making films that were universally legible lent itself to Ozu’s Westernized style of filmmaking.  The dialogue-heavy genre and the emergence of a national cinema were also conspiring factors that allowed the studios to finally break free of the control of the benshi.[26]  Ozu’s legible shot constructions and editing techniques, combined with the large number of intertitles that spelled out what the characters were saying, allowed audiences to experience his films without the need of a benshi to interpret the depicted events.  Although a significant amount of data does not yet exist on average ratios of intertitles to pictorial shots, this ratio in the films that have been measured for this project does seem significantly high: intertitles account for 288 of 1040 shots (28%) in An Inn in Tokyo and 260 of 1084 shots (24%).  It would be interesting to see how this percentage changes in a film like Dragnet Girl (1933), which is more action-oriented and comes before Ozu had completely moved into the family drama stage of his career. Indeed, Bordwell notes a marked increase in the number of intertitles in his films from 1932-33, from 12% of shots to 27%.[27] 

Though Ozu’s ASL doubled upon the conversion to sound in accordance with the universal trend, the statistical profile of his first sound film does not appear to have undergone a significant change.  Statistically, The Only Son is not radically different from Ozu’s other films: the cutting swing of 7.3 is less than the ASL, indicating a relatively restricted range of shot durations, according to the “Greek Cross” idea.  The film’s trendline, on the other hand, is noticeably more dynamic than the other measured Ozu films, cresting early at around 9 minutes, then dipping before slightly rebounding near the end, at about the time Tomibo gets injured by the horse (Figure 4). 

Figure 4

Figure 4. The Only Son (1936). Trendline 6.
http://www.cinemetrics.lv/movie.php?movie_ID=375.

Of the four Ozu films measured here, The Only Son is clearly the most dynamic in terms of editing, but this dynamization of editing does not appear to have been Ozu’s ultimate intention, if the flat trendline of the later, more mature Floating Weeds is any indication (Figure 8).  This dynamic trendline could also be the result of the film being Ozu’s first true foray into sound.  He said of The Only Son, “Because I couldn’t get rid of the mood and style of the silent movies at all, I became quite bewildered.  In spite of my understanding perfectly well that everything is different in a talkie, this movie had the style of a silent.”[28]  Bordwell connects Ozu’s bewilderment mainly to the film’s visual style in its use of “bold depth effects,” “the play with layers of focus in backgrounds,” and “the use of staggered staging.”  In terms of editing he only mentions “the way cutaways break up conventional ABA point-of-view editing,” but perhaps the remnants of silent style affected Ozu’s editing in this film on a more fundamental level.  With more thorough measurements that take into account the absence of intertitles and the consequent absence of the effect on the film’s overall rhythm they might create, we could more concretely determine the reason behind the abnormality of the editing in The Only Son.

Figure 5

Figure 5. An Inn in Tokyo (1935).  The Imaginary Feast. 
Pink = pictorial shot.  Blue = intertitle. DVD 19:08-22:35.
http://www.cinemetrics.lv/movie.php?movie_ID=406

If we compare two similar scenes from An Inn in Tokyo and The Only Son, we can see how the absence of intertitles helped cause the increase in ASL and, perhaps, how it dynamized the cinemetrical profile.  In An Inn in Tokyo, Kihachi and his sons have an imaginary feast in a barren field (19:08-22:34; Figure 5).  The segment lasts 3 minutes 26 seconds, divided among 35 pictorial shots and 19 intertitles.  Beginning with shot 21 (after the fifth intertitle), there is a long exchange between Kihachi and Zenko, who pretends serves his father sake and a meal of noodles.  The action is filmed in a two-shot with Kihachi on the left facing Zenko on the right; both are seated on the ground.  This exchange lasts 1 minute 36 seconds, divided by 22 cuts, alternating between pictorial shots and intertitles in a 1:1 ratio.  With the intertitles, the scene fits nicely in Ozu’s overall editing rhythm.  The trendline of the sequence is very even, with a cutting swing of 3; the trendline of the overall film is also very even, and has a cutting swing of 3.2.  But with such a small number of shots in an individual scene, the trendline is not our focus here.  Rather, I am interested in the number of intertitles that break up the feast sequence.  Of the nineteen intertitles, eleven of them are inserted into the single shot in which Kihachi and Zenko mime a meal.  During this exchange the camera does not move at all.  Although we cannot tell for certain whether the exchange we see is comprised of a combination of several takes of the same scene or the entire scene was performed all at once, the effect is that the scene is fluid, that the camera does not move and that the action as it is depicted in shot 21 continues through shot 43.  Without the intertitles, the shot would be longer than a minute, over twice as long as the longest shot in the entire film.  Although a single instance of such a shot would not make a significant statistical or stylistic difference in a film composed of 1040 shots, multiple instances of similar intertitle interruptions would drastically change the look and rhythm of the film.  Though a thorough measure has not been done, if we imagine the number of medium shots of speakers interrupted by intertitles that are immediately resumed after the intertitle disappears, a common practice in silent cinema, it seems obvious that a major shift in editing style and rhythm would result from the removal of these intertitles.

Figure 6

Figure 6. The Only Son - Visit with Okubo.
DVD 30:40-34:54.
http://www.cinemetrics.lv/movie.php?movie_ID=408

A similar dining scene occurs in The Only Son, when Ryosuke takes his mother, Otsune, to visit Okubo, his old teacher (30:40-34:54, Figure 6).  This sequence of 23 shots features three (Shots 3, 9, and 19) that are each longer than the longest shot in An Inn in Tokyo (25.3 sec).  The first two shots of the sequence provide one of the false transitions discussed by Bordwell in his analysis of Ozu’s use of intermediate spaces in this film.[29]  The third shot is the first that contains characters, and it is the longest of the sequence.  Otsune, Ryosuke and Okubo sit on the floor, much like Kihachi and his son did in An Inn in Tokyo.  The shot lasts about 32 seconds, during which they take turns talking.  Because the film is in sound and intertitles are no longer used, the scene can play out in much longer takes featuring larger chunks of dialogue.  With sound, there is no longer a need to cut in to a medium shot of a speaker, then to an intertitle, then either to the speaker or the listener.  The two other lengthy takes in the sequence are also shot from this camera position: the first features Okubo’s son crying for money, and the second depicts the introduction of his wife and baby.  Again, what’s interesting here is not the scene’s dynamic profile as represented by the trendline, but rather what the raw data tell us about Ozu’s preferred method of staging his scenes.  The increased shot length in his sound films may reflect a tendency already present in his silent films to stage scenes in longer takes with large amounts of dialogue.  This might help explain why his films had such a large ASL during much of the 1940s.  On the other hand, in his complex use of intertitles and his extremely legible, Americanized style, Ozu already showed a tendency to be ahead of the curve in the silent era.  His silent films, such as Dragnet Girl and A Story of Floating Weeds, include shots of characters reacting to off-screen sounds, playfully indicating Ozu’s awareness of sound technology even as he was refusing to use it.[30]  By staging scenes in longer takes, as he does in the example from An Inn in Tokyo, Ozu may be preparing for the necessary change in style that will come with his eventual adoption of sound, or he may be playfully indicating his awareness of stylistic changes that had already taken place in American and Japanese film as a result of their adoption of sound.

The Ukigusa Story

Several examples of pictorial shots interrupted by intertitles can be found in A Story of Floating Weeds (1934), the film that immediately precedes An Inn in Tokyo in Ozu’s oeuvre. However, A Story of Floating Weeds presents a different opportunity for comparison between the silent and sound periods of Ozu’s oeuvre, in that he remade the film in 1959 in color and sound as Floating Weeds.  Separated by 25 years in the life and career of the director, the differences between these two films are too vast to do justice to them here.  However, for the purposes of this paper, we might begin to trace out some similarities and differences between these two films using the CineMetrics tool.  The graph of A Story of Floating Weeds shows the same characteristically Ozuian regularity of ASL over the course of the film (Figure 7).  At trendline 1, the film exhibits a slight increase in ASL from the beginning (around 5) to the end (about 4.1), but for the most part the editing speed remains around the overall ASL of 4.6 throughout the film, only dipping below 3 toward the end.  This is reflected in the cutting swing of 3.5, indicating that most of the shots in the film are between 1.1 and 8.1 seconds.  The macrostructural symmetry of Ozu’s narrative discussed by Bordwell can also be seen in the graph of the entire film.[31]  At Step 1, trendline 6, there are two slight humps at the beginning and end of the film, with a smaller hump exactly in the middle, at around 44-45 minutes.  This symmetrical pattern is also evident in the graphs for An Inn in Tokyo and Floating Weeds.  (Ozu’s “bewilderment” with true sound recording and the absence of intertitles may account for the asymmetry of the graph of The Only Son.)  The episodic or modular nature of Ozu’s films can also be seen in the overall trendlines, such as trendline 12 for A Story of Floating Weeds.[32]

Figure 7

Figure 7. A Story of Floating Weeds.
Pink = pictorial shot. Blue = intertitle. Step: 1, Height: 300 pix, Trendline: 6.
http://www.cinemetrics.lv/movie.php?movie_ID=349.

Although Ozu made significant changes to the Ukigusa story between 1934 and 1959 on a nearly scene-by-scene basis, the basic outline remains the same, and the family relationship between the mother, father and son is the core of the story.  While a full comparison of the two versions of Floating Weeds is impossible here, we can again examine matching scenes to look for differences in Ozu’s style.  For instance, the climax at the end of each film, in which the mother reveals the father’s identity to their son, provides such an example.  Despite the overall flatness of Ozu’s cinemetric curves, both versions of Floating Weeds tend to crest around this moment in the narrative: that is, the editing speed of both films reaches a high point at the moment of confrontation and revelation.  Before closely analyzing the climactic confrontation scenes, a few words should be said about the graphs of the overall films (Figures 7 and 8).  When we initially look at the graphs using the same trendline degree (6), it is immediately clear that the line for Floating Weeds appears more dynamic (‘curvier’) than that of A Story of Floating Weeds.  This could result from a few factors.  First, we might assume that Ozu’s editing became more dynamic (however subtly) as he got older, and so his films had a more variable range of shot lengths used to make his narratives more expressive or more exciting.  However, the raw data from the 1959 film still reveal an ASL that is greater than the cutting swing, indicating a relatively controlled and restricted range of shot durations.  This statistic is a bit misleading, as the potential range of shot durations within one standard deviation of the mean in Floating Weeds (1.3 to 13.5) is far larger than that of A Story of Floating Weeds (1.1 to 8.1), but the graphs nevertheless bear out this tendency toward regularity and evenness.  The trendline of Floating Weeds at degree 6 can be further flattened by adjusting the ‘Step’ function to 1 in order to show more of the shots in the graph.  For Floating Weeds, making this adjustment results in a trendline that is more characteristically Ozuian and flat, indicating that perhaps his cutting style had not changed that drastically from silent to sound, or even for the 25 years between his Ukigusa films.  The precipitous dip in cutting speed at the end of the film, from nearly 6 seconds per shot to about 11.5 seconds per shot in about 12 minutes of screen time is a bit puzzling, but can at least partially be explained by the 3 shots longer than 30 seconds (along with one longer than 20) within the last 10 minutes of the film.  One of these lasts 50 seconds, and depicts Sumiko’s devoted, agonized attempt to light Komajuro’s cigarette.

Figure 8

Figure 8. Floating Weeds.
Step: 1, Height: 300 pix, Trendline: 6.
http://www.cinemetrics.lv/movie.php?movie_ID=369

 

Figure 9

Figure 9. A Story of Floating Weeds. Dissolution of the family.
Pink = pictorial shot. Blue = intertitle. Step: 1, Height: 200 pix, Trendline: 6. DVD 1:08:28-1:22:23. http://www.cinemetrics.lv/movie.php?movie_ID=402.

At the level of individual scenes, however, a cinemetrical analysis might provide a different picture.  Using the climactic confrontation between mother, father and son as an example, a drastically different editing style can be seen.  The scene in A Story of Floating Weeds displays a very flat trendline and has an overall ASL of 3.7 with a cutting swing of 2.0 (Figure 9).  Although the measurement differentiates between pictorial shots and intertitles, and the graph can be displayed with or without either type of shot, it is difficult to determine just what effect the intertitles have on the graph and the scene’s rhythm.  Most likely they will tend to flatten the graph because the cutting swing of the intertitles is probably lower than that of the pictorial shots, but these data are not yet available through the CineMetrics software.[33]  Even by removing the intertitles from the graph, the trendline remains noticeably flat.  Undoubtedly the same phenomenon occurring in An Inn in Tokyo, that of pictorial shots interrupted by intertitles, making longer shots into shorter ones, is happening here.  The sections of heavy dialogue in the scene (on the graph from about 5 minutes to 7, then from about 8 to 11:30) contain several examples of this phenomenon.  Shinkichi and Otsune both have speeches that span several intertitles, between which Ozu will often cut back to the original speaker, so a shot of Shinkichi speaking will cut to an intertitle (“I have no father like him”), then back to Shinkichi speaking, then to a second intertitle (“My father was a civil servant.  He’s dead.”).  The presence of sound would change the scene not only by removing the intertitles and allowing Shinkichi to deliver his speech in an unbroken take, but also might lengthen the shot by causing Ozu not to cut away to the reactions of other characters in the middle of a speech.  During Otsune’s line about Kihachi sending money back to provide for Shinkichi’s education, Ozu cuts not only between Otsune and intertitles, but also to Shinkichi.  That is, shot-by-shot, Otsune’s speech looks like this:

1.      (ms) Otsune speaking

2.      Intertitle of Otsune’s words

3.      (ms) Shinkichi

4.      Intertitle of Otsune’s words

5.      (ms) Shinkichi

6.      (ms) Otsune speaking

7.      Intertitle of Otsune’s words

8.      (ms) Otsune speaking

9.      Intertitle of Otsune’s words

10.    (ms) Shinkichi

11.    (ms) Otsune speaking

12.    Intertitle of Otsune’s words

13.    (ms) Shinkichi

14.    Intertitle of Otsune’s words

15.    (ms) Shinkichi

Bordwell would undoubtedly point out the stanza-like pattern of this group of shots: shots 1-5 mirror shots 11-15, with shots 6-10 slightly modifying the rhyme scheme by placing Otsune in the central shot instead of Shinkichi.  However, looking at this scene in comparison to a sound film, it is clear that this juxtaposition of intertitles between medium shots of speaker and listener has the effect of overlapping dialogue.  However, in his sound period, Ozu rarely (if ever) used overlapping dialogue.  How, then, would he stage this dramatic speech in sound? 

This question is of course a trap, particularly when we ask it about someone like Ozu, whose entire goal, as Bordwell argues, is to play with the viewer’s expectations, which are based on conventional cinematic style.  In 1959, equipped with the tool of sound, we might expect a director to retain or even expand Oyoshi’s speech, giving the actress a scene that could demonstrate her individual skill as an artist.  However, as we know, Ozu was not an actor’s director.[34]  According to Bordwell, Ozu’s films emphasize “the primacy of the pictorial and the urge to homogenize all shots, to make them equivalent visual units,” at the expense of narrative complexity.[35]  With this in mind, there ironically seems to be a greater emphasis on dialogue in the silent version of the Ukigusa story than in the sound.  Whereas Otsune delivers a long, impassioned explanation and defense of Kihachi’s abandonment of her and Shinkichi, in the scene of the dissolution of the family in Floating Weeds, Komajuro (the Kihachi character) cuts Oyoshi off midway through her speech and allows Kiyoshi to shout at him and storm off (Figure 10).  In the 1934 version, the intertitles during the speeches are given at least equal weight as the pictorial shots, which tend to be medium frontal shots of the speaker.  The cinemetric graphs show that the longest intertitles are located in this confrontation scene, some lasting as long as 7 seconds.  The increased duration of intertitles in the scene is also clear: in the overall film, the ASL for intertitles is 3.1, while in this scene it is 3.4.  The emphasis on dialogue is also borne out in terms of quantity: in the film, 260 of 1084 shots (24%) are intertitles.  In this scene, 61 of 223 shots (27%) are intertitles. 

Figure 10

Figure 10. Floating Weeds. Dissolution of the family.
Trendline 6. DVD 1:41:18-1:53:44. http://www.cinemetrics.lv/movie.php?movie_ID=405.

No doubt, dialogue is also extremely important in the analogous scene in Floating Weeds.  The film and genre are still the same, and we expect a family drama to climax with more dialogue and emotional expressions than physical action.[36]  But in the confrontation scene in Floating Weeds, dynamic staging takes on the role of overt narration that Ozu’s intertitles played in the 1934 version.  Oyoshi’s appeal to her son takes place in a long shot with Komajuro seated on the left, Oyoshi standing on the right, and Kiyoshi standing exactly in the center, his back to the camera and his head cut off by the top edge of the frame, emphasizing his height and literally causing him to “stand out” in contrast to his parents.  In this version, the staging expresses the characters’ relative mental states more than the dialogue: Komajuro sits dejectedly, no longer attempting to conceal his age and impotence, which have just been exposed by his own son’s manhandling of him; Oyoshi stands, her body slightly bent, fearful that she will lose both her companion and her son; and Kiyoshi faces away from the camera, his head out of the camera’s range, not even allowing the audience to attempt to penetrate the vacant stare he wears through most of the film.  The eloquence of the mother’s speech in the 1934 version, in which Ozu could use intertitles to exercise his overt narration, is removed from the scene for the sound version, its overt narrative expressivity replaced by intricate staging that, if anything, represents Kayo’s point of view as silent witness to the proceedings.

Ozu and Play

For Bordwell, Ozu’s style represents a ludic engagement with the Hollywood style he admired.  Ozu playfully subverts and toys with paradigmatic Hollywood devices and rules, especially the 180-degree rule and rules governing camera placement and movement, shot patterns and construction, the use of intertitles, and nearly everything having to do with classical Hollywood filmmaking practices. Bordwell writes that, “In Ozu, the principle of a ludic narration pervades the entire film.  His narrational asides are only the most evident outcroppings of a systematic narrational playfulness.”[37]  This reading of Ozu’s work as a playful engagement with conventional Hollywood filmmaking and storytelling practice can be extended to the cinemetrical level.

Briefly, I want to propose another explanation of the effect of the regularity in Ozu’s editing, the rarity of his deviation from a small range of shot durations.  Borrowing from Bordwell’s identification of a ludic principle operating in Ozu’s films, the director’s regular editing also allows for the creation of playful narrative surprises and shocks by subverting the “subliminal norms” and “tacit expectations” established by the films’ repetitive meter. [38]  These surprises arise naturally within the narrative in two ways: (1) significant or extraordinary events can occur within a regularly timed shot, or (2) a shot of above-average length (beyond the cutting swing) can be inserted among a series of regularly timed shots.  The Only Son contains a wonderful example of the latter technique.  The longest shot depicts a corner of Ryosuke’s home showing a baby bottle on the floor and a Japanese poster on the wall, with no human figures in the image.  This 57-second shot occurs after the late-night scolding of Ryosuke by his mother.  The shot’s duration is important both statistically and artistically.  Statistically, it stands out more than any other shot in the film, running over 6 standard deviations longer than the ASL and over 25% longer than the next longest shot.[39]  Artistically, the shot is an image of pure time: though at first nothing appears to be happening in the static shot, the viewer slowly becomes aware of chirping birds on the soundtrack and a gradual brightening of the room.  The shot depicts the break of day at an accelerated, almost supernatural pace, using a single minute to show an event that occurs over several.  Ironically, even though the event depicted is accelerated by the duration of the shot, at the same time the shot makes the audience impatient and curious as to what is happening.  Because of the steadiness of the editing and the emotional significance of the prior scene, the shot seems even longer than it is, thereby making it even more meaningful.  Once the viewer realizes that daybreak is being depicted, Ozu’s sense of playfulness and gamesmanship become evident.

Some examples of the other technique, showing significant or extraordinary events in normally timed shots, can be found in the Ukigusa films.  The first scenes in which Kihachi beats Otoki (1934) and Komajuro beats Kayo (1959) appear shocking not only because we have not seen him behave this way before, but also because the steady rhythm of Ozu’s editing tends to lull the audience into a comfortable register of spectatorship.[40]  Though we might expect moments of violence to be accompanied by fast editing and brief shots, Ozu subverts our expectations in this regard, as Bordwell aptly points out: “There will be no sudden accelerations, no flurry of rapid shots; the calm pace of the cutting will match the unfolding syuzhet events.”[41]  Very often, these two conditions will coincide: a significant event will be shown in a longer than average shot.  The attacks on Otoki/ Kayo and Otaka/ Sumiko in the theater are timed more or less regularly, but the altercation with the son at the end in both versions occurs in a longer than average shot and, of course, marks perhaps the most significant moment in the narrative.  When Kihachi/ Kiyoshi steps in to defend Otoki/ Kayo from his father’s beatings, the shot that depicts the struggle in which the son pushes the father to the ground is longer than average in both versions.  In A Story of Floating Weeds the shot lasts about 13.4 seconds, and is the third of the three longest takes in the climactic scene.[42]  In the 1959 version, the father/ son struggle is longer than average for the scene, but not as long relative to the average as the shot in the 1934 version.[43]  Instead, the shots of Kiyoshi bringing Kayo to his mother for approval (~23.3 seconds) and of Komajuro apologizing to Kayo and asking her to look after Kiyoshi, thereby entrusting his future to her (~21 seconds), are weighted more heavily in terms of shot duration.  Whether they outweigh the shot of Kiyoshi’s manhandling of Komajuro in terms of narrative significance is more a matter of interpretation. 

When more options for displaying and reading the statistics derived from CineMetrics become available, and when more films are measured, we will be able to draw more concrete conclusions from the data about the effect of the editing on the film experience and the changes Ozu made between silent and sound versions of the same story or similar scenes.  Other statistical data in the films can also be measured, including number of characters in the frame, character movement, camera movement, music, and lines or words of dialogue. [44]  CineMetrics will remain a tool for opening up discussions about different aspects of individual films and groups of films, their construction and reception, the historical habits of directors and editors (and perhaps screenwriters, set designers and actors), and questions of generic or national norms at given moments in history.

The addition of sound to Ozu’s filmmaking practice clearly had tangible effects on the way his films looked (and sounded), but it is remarkable that he was still able to find ways to maintain the regular rhythm and create the subliminal norms that characterized his desired editing style.  Despite the massive increase in shot length as a result of the transition to sound, Ozu was able to maintain the regular meter of his films for a number of reasons, including the continuation of his work in the ‘home drama’ genre throughout this period.  In addition, toward the end of his silent period his films seem already to be tending toward longer takes in the style of sound film.  However, among these technical and generic reasons should not be forgotten the playful rigor and single-mindedness of this unique and idiosyncratic artist.


[1] Donald Richie, Ozu: His Life and Films (Berkeley: University of California Press, 1974); Kristin Thompson, “Notes on the Spatial System of Ozu’s Early Films,” Wide Angle, Vol. 1, No. 4 (1977) p. 8-17; David Bordwell, Ozu and the Poetics of Cinema (Princeton, NJ: Princeton University Press, 1988) pp. 19-21.

[2] Thompson, p. 9.

[3] Bordwell, Ozu, pp. 151-152, 155-159.

[4] Ibid.

[5] Ibid, p. 155.

[6] Ibid, p. 158.

[7] Ibid, p. 75-76.

[8] Ibid, p. 76.

[9] This information came from Michael Raine.  I have not been able to find a direct citation of it, partly because much of the information on this period in Ozu’s life has not been translated from Japanese.

[10] Bordwell, p. 19.

[11] Ibid, p. 75.

[12] Ibid, p. 67.

[13] Toggle the “Color Code?” tab below the graph to ‘Yes’ and click Redraw to see the colors on the website.

[14] Change the “degree of the trendline” tab to 6 and click Redraw to see what I mean.

[15] Using CineMetrics, Yuri Tsivian measured City Lights, A Dog’s Life, and The Kid with an eye toward exploring Chaplin’s practice of building his films around laughs of varying intensity.

[16] The ratio of ASL to cutting swing is emerging as a topic of interest on the CineMetrics site.  Very generally, when this ratio is greater than 1, the film’s editing can be said to be more restricted than when it is less than one, which would indicate a wider variety of shot durations and a more dynamic overall editing rhythm, as we see in most Hollywood films.  Tsivian is calling the rare equivalence of ASL and cutting swing “The Greek Cross,” and has discussions of the phenomenon in the comments at http://www.CineMetrics.lv/movie.php?movie_ID=358 and http://www.CineMetrics.lv/movie.php?movie_ID=369.

[17] NB: Since the time of writing, later silent Chaplin films have been added, including The Kid, The Gold Rush, and A Dog’s Life.

[18] See also the graph for Stagecoach for a paradigmatic example of classical Hollywood editing patterns, with the trendline set to 6.

[19] See Barry Salt, Film Style and Technology (London: Starword, 1983) pp. 231-232.

[20] Ibid.  See also Charles O’Brien, Cinema’s Conversion to Sound: Technology and Film Style in France and the U.S. (Bloomington: Indiana University Press, 2005).

[21] Most of Bordwell’s figures vary slightly from mine for as-yet undetermined reasons.  His method is to count the shots and divide by the film’s length.  CineMetrics is basically the same method, with the addition of software to measure each shot individually.  Our disparity between An Inn in Tokyo and The Only Son is not very significant, but for The Story of Floating Weeds the difference is quite marked.  Unless otherwise noted, all the results used in the paper come from my measurements and others available on the CineMetrics website.

[22] Salt, p. 282.

[23] Bordwell, p. 377.

[24] Ibid.

[25] Ibid.

[26] See Aaron Gerow, “One Print in the Age of Mechanical Reproduction: Film Industry and Culture in 1910s Japan,” Screening the Past (1 Nov 2000),
http://www.latrobe.edu.au/screeningthepast/firstrelease/fr1100/agfr11e.htm.

[27] Bordwell, p. 67.

[28] Quoted in Bordwell, p. 271.

[29] Bordwell, pp. 271-272.

[30] Ibid, p. 67.

[31] Ibid, 56-57.

[32] Bordwell, Ozu, p. 63, draws on Richie, Ozu, pp. 15-16, for the modular nature of Ozu’s films, especially his home dramas.

[33] Statistics like standard deviation and maximum shot length of different shot types for films measured in Advanced Mode have since been added to the website.  In the case of this scene in A Story of Floating Weeds, this hypothesis is borne out by the data: the cutting swing of the pictorial shots is 2.1; that of the intertitles is two-thirds of that, 1.4.

[34] See Bordwell, pp. 83-85.

[35] Ibid, p. 98.

[36] “The action’s dramatic register is often low” in a family drama. Ibid, p. 55.

[37] Ibid, p. 66.

[38] Ibid, 75-76.

[39] To see these shots on the graph, go to the website and change the ‘Height’ to 750 pixels.

[40] Bordwell, p. 359, discusses the rhythmic nature of the beating of Kayo and Sumiko in Floating Weeds.

[41] Ibid, p. 75.

[42] See graph and comments for a more detailed discussion of each long shot in this sequence.

[43] The shot is either 19 seconds or about 13.2 seconds, but unfortunately my notes are not clear which shot it is on the graph.  I’m pretty sure the shot is one of the two under ‘6:23’.

[44] Charley Leary has measured several John Cassavettes films with these variables in mind, including number of figures in the frame (Shadows), camera movement (Love Streams), and content of the soundtrack (Husbands).