BARRY SALT

HOW THEY CUT DIALOGUE SCENES

None of the available books about film editing give any principles for cutting dialogue scenes in movies, nor do interviews with practising editors give any real guidance on this point. The important thing about dialogue scenes is that they occur in all films, whereas car chases and musical numbers and so on do not, so cutting dialogue must be a basic part of editing technique. The variables that seem to be significant can be shown by using some cutting points in John Ford’s The Grapes of Wrath (1940). Obviously the most common type of cut from one shot to the next inside a dialogue scene is that made at the same point in time on the picture and sound tracks, somewhere inside the gap between one speaker finishing speaking, and another replying. This can be illustrated in the following diagram, based on what one sees on the computer screen when editing.

I measure this sort of cut by counting the number of frames from the end of the last speech sound in the outgoing shot to the picture cut (length 'A' above), and the number of frames from the picture cut to the first speech sound (length 'B' above) in the incoming shot. An alternative description of this measurement is that the A length runs from the beginning of the pause between the voices of the two speakers to the picture cut, and the B length runs from the picture cut to the end of the pause, at the point when the second speaker begins to reply.

The other sorts of cuts are those made at different points in time on the picture track and the sound track. The first class of these is like that below:

In this cut, the sound of the voice of the person seen in the outgoing shot continues under the picture of the person who is eventually going to reply in the incoming shot, until that person replies. Again, I measure the exact placement of this cut in the picture with respect to the speeches by the lengths A and B from the cut in the picture to the end (or start) of the sounds, though in this case A is given a negative value. This sort of cut is nowadays called an 'L-edit', or 'L-cut', after the outline shape of an 'L' that can be made out in the layout of the shapes of the parts in the diagram above.

The converse edit is illustrated below, which has the sound of the speaker who is about to be shown full face in the next shot being first heard under the end of the outgoing shot. This is called a 'J-edit' or 'J-cut'.

            Both these latter types of dialogue cut are nowadays referred to as 'split edits'. These terms have only been used in the last decade or so in film editing, after being taken over from videotape editing in television, where they are much older, dating back to the early computer-controlled linear videotape editing systems. However, these types of cut were made in sound motion pictures long before that, but they were described as 'overlapping the sound'. That description had a certain degree of ambiguity, as it could be confused with laying one sound track over another sound track, often referred to as 'overlapping dialogue'. As far as I can tell, film editors have no specific name for what I call the 'A-length', probably because it is in general very short. And the only name given to the 'B-length' of which I am aware is by Donn Cambern, in his interview with Gabrielle Oldham in First Cut: Conversations With Film Editors (Cal U Press, 1992), where he calls it 'the lead' (p.205). I use this term interchangeably with my term, the 'B-length', in what follows.

            The three situations in editing dialogue detailed above cover most of the cutting that occurs in such scenes, with the exception of a cut to and then from a reaction shot of someone in the scene who is not speaking, but just listening. I take a count of the number of such reaction shots separately. There still remains the extremely rare situation where the relation of the speeches to the cuts is too complex to be covered by my method, for instance when a number of people are speaking at once, but I have just ignored these.

            So the way I investigated this matter was to start at the beginning of the film, and record the A and B lengths for each dialogue cut down the length of the film, until I had covered them all, or reached 200 dialogue cuts, whichever happened sooner. Most films that I have analysed contain less than 200 dialogue cuts, but there are a very few with a lot of talking and a lot of shots, such as Mr. Smith Goes to Washington, where I only got half way through before reaching 200 dialogue cuts. I made graphs for each film showing the numbers of cuts with different A-lengths and different B-lengths, like the ones below for The Grapes of Wrath.

In the first of these graphs, the height of the bars represent the percentage of cuts with any particular A-length between zero and 20 frames. (The negative frame lengths represent L-edits.) So you can see that for this film 35 percent of the dialogue cuts were made within one frame after the last sound of the end of the speech, before cutting to a shot of the person who is going to reply. You can also see that for most of the cuts the A-length is less than 6 frames long, and indeed after removing the four L-edits from consideration, no less than 75% of the A-lengths are less than 6 frames long. Most of the other 29 American films made from 1936 to 1999 that I looked at also have A-lengths less than 6 frames, as you can see in the table below.

            In the second graph, you can see that the B-lengths or leads are much more spread out over the range of possibilities. In fact for The Grapes of Wrath  no less than 23% of the leads are greater than 20 frames, and so do not appear on the graph above. Incidentally, these measurements are in film frames, not video frames, so 24 frames is always equal to one second throughout this article.           

            Here is a summary of my investigations for 30 American films made between 1936 and 1999. Added to them, there are two British films from 1959 and 1962 inserted in the middle of the sequence. The figures are for the percentages of L-edits and J-edits in the dialogue scenes, the average pause length in frames for all the dialogue cuts, the percentage of reaction shots amongst all the shots in the dialogue scenes, and the percentage of A-lengths less than 6 frames long amongst the dialogue cuts.  

Film

Year

Editor

% of

L-edits

% of

J-edits

Average Pause (in frames)

% of

Reaction

shots

% A-lengths

less than

6 frames

Mr. Deeds Goes to Town

1936

Gene Havlick

9

6

13

9

66

Each Dawn I Die

1939

Thomas Richards

10

2

12

10

66

Another Thin Man

1939

Frederick Y. Smith

26

12

11

15

70

Love Affair

1939

G. Hively & E. Dmytryk

10

4

17

20

54

Mysterious Mr. Wong

1939

Jack Ogilvie

2

11

13

38

32

Mr. Smith Goes to Washington

1939

Al Clark & G. Havlick

10

7

6

21

83

Harlem Rides the Range

1939

?

24

7

14

29

68

Destry Rides Again

1939

Milton Carruth

8

3

16

13

63

Little Princess, The

1939

Louis Loeffler

10

4

15

5

62

Old Maid, The

1939

George Amy

11

4

16

8

83

At the Circus

1939

William H. Terhune

17

4

10

7

64

Rose of Washington Square

1939

Louis Loeffler

14

6

14

8

71

Invisible Killer, The

1939

Holbrook N. Todd

14

8

12

20

66

Grapes of Wrath, The

1940

Robert Simpson

4

2

21

18

71

Casablanca

1942

Owen Marks

12

4

12

6

59

 

 

 

 

 

Jesse James

1957

Robert Simpson

30

3

20

16

74

Best of Everything, The

1959

Robert Simpson

10

4

26

9

56

Compulsion

1959

William Reynolds

16

8

23

21

55

Wild River

1960

William Reynolds

15

4

25

55

57

Pillow Talk

1959

Milton Carruth

18

5

22

26

63

Imitation of Life

1959

Milton Carruth

22

11

17

19

57

Ben-Hur

1959

Ralph Winters

14

5

37

30

36

Gidget

1959

William A. Lyon

12

5

24

14

36

They Came to Cordura

1959

William A. Lyon

9

8

24

19

37

Last Train from Gun Hill

1959

Warren Low

3

1

25

10

33

 

 

 

 

 

Too Many Crooks

1959

Bill Lewthwaite

64

5

14

21

62

On the Beat

1962

Bill Lewthwaite

49

2

15

26

72

Stakeout

1987

Tom Rolf

37

4

14

32

68

 

 

 

 

 

Three to Tango

1999

Stephen Semel

11

12

13

45

64

Talented Mr. Ripley, The

1999

Walter Murch

10

10

17

39

54

Angela's Ashes

1999

Gerry Hambling

16

9

20

57

68

Life

1999

Jeffrey Wolf

11

10

7

36

80

            From the figures for percentages of A-lengths less than 6 frames given in the last column you can see the basis for my conclusion that films editors mostly cut to the next speaker just after the first speaker has finished. The exceptions are an extremely cheap B-movie from 1939 (Mysterious Mr. Wong), and four films from 1959. So the data shows that cutting to the next speaker very shortly after the first speaker has finished is indeed the norm in dialogue editing, but that there is some variation from this for one reason or another. This also means that the B-length (the 'lead') is usually almost as long as the actual pause between the two speeches on either side of the cut. Nevertheless, there is also a variation from one editor to another in this quantity, with a few editors cutting more loosely to the beginning of the pause between speakers. The obvious instances of this are William A. Lyon, Warren Low, and Jack Ogilvie.

            Another important conclusion is that the use of J-edits is not very common, while L-edits are more frequently used. I believe the reason for this is that L-edits show the listener's reaction to what the speaker in the first shot is saying towards the end of his or her speech. This is the point by which the listener has most fully absorbed the significance of what is being said to them.  They are effective if they show the listening character’s facial reaction to what the speaker is saying, but only if the listener’s reaction has real dramatic significance at that moment. The listener's reaction will of course also appear in the lead (or B-length) of the next shot before they speak, but using an L-edit enables the pause to be shorter, and hence speed the story on. The reverse arrangement using a J-edit is less likely to show any dramatically significant facial expression, and so it is less used. A J-edit can also create initial confusion about who is speaking if more than two people are present in a scene.

          The amount of L-edits used varies appreciably between editors, and this can be seen as a difference of editing style between editors. There may be a small general change over time in the use of L-edits, if the four 1999 films studied are truly representative of recent editing practice, as they use a less of this technique than all the earlier films. The highest amount of L-edits in my results is recorded for the two British comedies, Too Many Crooks (1959) and On the Beat (1962), edited by Bill Lewthwaite. In Too Many Crooks, no less than 64% of the dialogue cuts are L-edits. And my subjective impression is that this works just fine, not to mention the fact that both films went down well with the British public when they were released. I discussed L-edits with Bill Lewthwaite a dozen years ago when he was briefly in charge of the editing department at the London Film School, at the end of his long career of editing feature films, but it was quite difficult to get much out of him on the subject. Finally he grudgingly conceded that overlapping forwards (L-edits) was more a British thing than an American thing. Since Gerry Hambling, who edited Angela's Ashes, is also English, but uses a normal amount of L-edits, I am inclined to think that an extreme fondness for L-edits might be more of a Bill Lewthwaite thing. Though this subject could do with more research.

Pause Length

            The average pause length, which is the absolute number of frames between the end of one speech, and the beginning of the next speech from the first speaker in the next shot, also varies from film to film, but as you can see for the figures for the editors who cut more than one film of those included in my table, there is a strong tendency for a particular editor to tend to use a particular length. So this is another possible indicator for editing style. Care is needed here, as the typical pause length for American films has obviously changed over time. In 1939-40 you can see from the figures above that it was around 14 frames, while by 1957-60 it had substantially increased to around 24 frames. This then decreased back to around 14 frames over the forty years between 1959 and 1999. The reason for dialogue editing getting slacker between 1939 and 1959, not only with an increase in pause length, but also in the way that A-lengths increased, is a mystery to me, as the cutting rate (ASL) had already returned to 1939 values by 1959, after the long takes of the 'forties departed.

            On the other hand, the decrease in the average pause length over the last forty or so years since 1959 is surely due to the increase in the cutting rate over the same period. This increase in the number of shots in a feature film is best measured in terms of the Average Shot Length (ASL), a quantity I introduced into film analysis forty years ago. For American films the typical Average Shot Length was about 8 seconds in 1959, while in 1999 it had decreased to about 5 seconds. (A more detailed analysis of the way cutting rates have changed over time and space can be found in the third edition of my book; Film Style and Technology: History and Analysis.)

Reaction Shots

            The number of reaction shots has noticeably increased in recent times. This increase in the number of reaction shots is also fairly certainly due to the general increase in cutting rate over the last 60 years. The number of reaction shots that an editor can put into a dialogue scene is limited by the amount of material the director has shot showing the listeners in a scene, which is part of the general coverage of the action – that is, repeated shots of the same action from different camera positions. Traditionally, many directors, such as John Ford and Alfred Hitchcock, prided themselves on 'cutting in the camera', and only shooting the shots that they knew they needed for the final edited film. Effectively, they were not shooting any coverage. In such cases, the editor was limited to the reaction shots the director wanted. But in recent decades the amount of coverage shot for ordinary feature films has vastly increased, so as to get more angles on the scenes all the way through, and hence the editor is in a position to use many more reaction shots. The amount of reaction shots used is also influenced to some extent by the nature of the story of the film. When characters in a film make long speeches to an audience of one kind or another, the automatic reaction of editors is to cut away from them to the faces of people in their audience from time to time. This is the case, for instance, in Wild River.

Personal Editing Styles

            Over and above the general trends, the figures I have quoted in the table show some individual variation between editors in the way they cut dialogue scenes. Apart from the low number of L-edits and J-edits used by William A. Lyon and Warren Low that has already been mentioned, there is the very high number of L-edits in Stakeout. These sorts of extreme things do not occur without intention, and in one of the few statements describing an editor's personal style that I have come across, Tom Rolf says, in his interview with Gabrielle Oldham in First Cut: Conversations With Film Editors (page 126), 'I like to overlap a lot.' He then says, 'Another of my minor laws, I never let an actor start his dialogue off-stage. He should start on-stage, and then segue to whoever else is reacting to it.' I think he is speaking figuratively here, and actually means “out of shot”, rather than “off-stage”. These statements exactly agree with the objective evidence I have got for Stakeout, with the overlaps almost entirely in the forward direction (i.e. L-edits), and with very few J-edits used. 

            On the other hand, the exceptionally short length of the pauses in Mr. Smith Goes To Washington is pretty certainly due to the director, Frank Capra, insisting that the editor speed the film up as much as possible, to go with the way Capra always kept the action going fast inside the shot.

            The opposite effect can be seen in the editing of three films by M. Night Shyamalan, The Sixth Sense, Unbreakable, and Signs. When I first saw Signs, I noticed the very long pauses in the dialogue across the cuts, which generally occur in the middle of these long pauses, so I checked it, and also Shyamalan's two  previous films, for this quantity. As you can see in the table below, this average pause is in the region of one and a half seconds, about twice as great as the typical figure for other films from around the same date. But all three of these Shyamalan films have different editors. So here we have a prominent stylistic feature of the editing which has clearly been insisted upon by the director, and is not due to the editor. As a control test, to show that this is not just a generic feature of this sort of spooky suspense film, I also quote the results for What Lies Beneath (2000), which has similar content elements to The Sixth Sense, with the protagonist experiencing ghostly visitations that are treated by psychiatric sessions. As you can see, What Lies Beneath has an average pause length of 23 frames, which is closer to the normal figure.

            When I first saw The Sixth Sense, I did not particularly notice the long pauses between speeches, but I noticed them straight away while watching Signs. I would say that my initial subjective impressions correspond to the dramatic grip of the story of the first film, and the relative lack of the same dramatic grip in Signs. In The Sixth Sense the problems of the main characters, and what they are trying to achieve, connect together with one another and with the story convincingly, whereas Signs is really just a standard unearthly monster film, with the themes of the redemption of the failed priest and the failed athlete not really connected to the basic plot throughout the story. And this is why we do not see more of the long pause in editing, for it can interfere with the film’s hold on the audience, unless the script is particularly gripping dramatically. The only earlier film that I have studied which has a similarly long average pause is Ben-Hur, but films about the life of Jesus have been slow and awe-struck since the very beginning of cinema.

Film

Year

Editor

% of

L-edits

% of

 J-edits

Average Pause (in frames)

% of

Reaction Shots

% of A-lengths less than

6 frames

The Sixth Sense

1999

Andrew Mondshein

15

12

44

55

18

Unbreakable

2000

Dylan Tichenor

12

12

43

24

23

Signs

2002

Barbara Tulliver

2

5

37

15

13

What Lies Beneath

2000

Arthur Schmidt Jr.

24

13

23

37

35

 

 

 

 

 

Return of the Secaucus Seven

1979

John Sayles

8

41

8

12

45

Lianna

1983

John Sayles

15

33

12

10

38

Baby It's You

1983

Sonya Polonsky

21

9

32

13

57

Brother From Another Planet

1984

John Sayles

7

54

19

34

14

The above table also includes my results for four films directed by John Sayles. This is because Warren Buckland pointed out to me a peculiar feature of the dialogue editing of John Sayles' films, which becomes apparent when you analyse the cuts in them. When Sayles directed the first of his own films, The Return of the Secaucus Seven, after writing Piranha for Roger Corman, he also edited it, apparently without having any previous experience or training in the craft of editing. The result is the extremely high proportion of J-edits (41%) in the dialogue scenes, and since many of these scenes involve several actors, this creates repeated confusion about who is speaking until they are identified by a cut to the speaker. Sayles has since eased up slightly on his use of J-edits, but for his films that have had studio funding, namely Baby, It's You, Matewan, and Eight Men Out, he has had to use a professional editor. In those cases the editor has used a minimal amount of J-edits.

Murch Blinking

            My sample of films from 1999 includes The Talented Mr. Ripley, which was edited by Walter Murch. Murch is well known for developing his own personal theories about editing, which centre on the human physiological feature of eye blinking. His theory has three aspects. He believes that his choice of the best point to end any shot is when he, Walter Murch, blinks while examining the uncut shot, secondly that the film audience blinks in unison at each cut in the film that they are watching, and thirdly that the best point to cut a shot is near where the actor in the shot blinks. These ideas are expounded in pages 57 to 72 of his book In the Blink of an Eye: A Perspective on Film Editing (Silman-James Press, 2nd Revised edition, (2001)). If you look at the table above, you can see that for the editing variables that I am examining, The Talented Mr. Ripley is very close indeed to Three to Tango, a comedy made the same year, so Murch’s ideas are not making him cut any differently to Stephen Semel, the editor of the latter.    As to how people actually react to film cuts, the best experimental evidence is in the article Edit Blindness: The relationship between attention and global change blindness in dynamic scenes by Tim J. Smith and John M. Henderson in The Journal of Eye Movement Research 2(2):6, 1-17. This research shows that insofar as there is any synchronism of viewer blinking with cuts in a movie, it is weak and very limited, and it is not exact, but only approximate. As far as Walter Murch’s other idea about cutting near the point when an actor blinks in the shot, I observe that he does not specify whether this should be before, after, or during the actor blink, which makes searching for the postulated effect a little difficult. Personally, I think that the general tendency would be for editors to cut before any blink by an actor, as including a blink at the end of a shot gives an impression of psychological weakness to the actor’s performance. Incidentally, most good film actors consciously control their blinking at crucial moments in a close shot.

Summing Up

The variables I use to describe dialogue editing style do show some consistent variation between editors, and so they can be used to identify the existence of personal styles for different film editors, whether those editors recognize it or not. Also, if you look carefully at the figures for pause length in the first big table above, you can see that the average pause length increased from the 1939 films to the 1959 films. (The average of the average pause lengths for 1939 is 13.6 seconds, while the average for the 1959 films is 24.3 seconds.) At the same time, the percentage of a-lengths less than six frames went down from an average of 66% for the 1939 films to 50% for the 1959 films. To put it another way, the editing of dialogue got slacker over the intervening 20 years from 1939 to 1959. Since 1959, the cutting rate of American films has increased vastly, so that the usual Average Shot Length has decreased from about 8 seconds in 1959 to a bit above 4 seconds nowadays. Detailed figures on the change in cutting rates over film history can be found in my Film Style and Technology: History and Analysis. If the four films from 1999 listed in the table are anything to go by, the usual pause length in dialogue cutting has decreased again, to about 14 seconds, so we have got back to the tightness of dialogue cutting usual in the late ‘thirties.

I think the simple and crude recipe for adequate dialogue cutting is: cut the shot immediately after the last sound of the speech in the outgoing shot, and give a lead of about 8 frames to the incoming shot. Use the occasional reaction shot in the appropriate places, and other exigencies, such as the occasional necessity, even in a dialogue scene, for a cut on action will stop this approach being too mechanical. Nevertheless, the essential expressive function of the cut with relation to the dramatic content of the scene comes from variations in the length of the pause, and hence variations in the length of the lead. So you have to recognize the dramatic context when choosing the pause length, and sometimes a lengthening or shortening is appropriate. It is precisely in this point that the real art comes into the craft of editing dialogue. Doing it this way worked for Robert Simpson, and he got two Oscars and several award nominations doing it. If you want to really push the movie along, while still allowing appropriate reaction time for the respondent, use more L-edits like Bill Lewthwaite, instead of cutting just after the end of the sound.

I add that my personal belief, based on working on the cuts of lots of student films, is that altering a cut (any sort of cut, not just dialogue cuts) by one frame either way does not make any difference to the quality of the cut, but two frames either way can.

                                                                                                            Barry Salt, 2012