Statistical Style Analysis
The basic idea behind my methods of statistical style analysis is that the form of films noticeably differ from one to another, and that the variables used to study this should be based on the concepts that film-makers actually use.
Average Shot Length
The simplest to get is the Average Shot Length (ASL). This is the length of the film in seconds divided by the number of shots in it. I only quote one decimal place, except for films with an ASL under 2 seconds, for which I quote two decimal places. For older films I only take the length and count the shots from after the last title card in the front titles, which is usually the director's credit. For modern films that have a pre-credit sequence, I add in the length and number of shots in that as well. I do not include the length and shots in a credit sequence that has titles superimposed on action. In silent films, and indeed sound films, I count intertitles as shots.
For silent films, I measure the ASL at what seems to me to be the correct speed of projection, on the basis of the naturalness of the movement of the actors. That is, at the speed the film would have been shot at when it was made. This is admittedly a tricky point, and there is something to be said for quoting a value as though it was being projected at sound speed, because this value is more definite. But on the other hand, this produces the misleading impression in people who do not understand the situation, that the cutting is faster, in comparison to that of sound movies, than it really was when the film was made. And for movies that were cranked at 16 frames per second, like Battleship Potemkin, that gives an ASL 33% less than it is at the correct speed.
I count the number of reverse angle cuts (RA), which are defined as changes of camera direction of more than 90 degrees, and I express it as a percentage of the total number of shots in the film.
Another variable I collect is the number of cuts to or from a Point of View (POV) shot in the two films. POV shots are those taken from the position of one of the participants in the scene, or at any rate from a position on their line of sight. these are expressed as a percentage of the total number of shot transitions in the film, which equals the number of shots in the film.
Insert shots (INS) are shots of things, and do not include an actor's face. I count a very distant shot of a street scene, in which no principal actors are included, as an insert. They are also expressed as a percentage of the total number of shots.
Although there is a small amount of disagreement about precisely what shot scale corresponds to each descriptive term, it is sufficient for the purposes of analysis to define carefully what one means by each category, and then stick to it. I am using categories of Scale of Shot like those commonly used in the film industry from the nineteen‑forties, as follows: Big Close Up (BCU) shows head only, Close Up (CU) shows head and shoulders, Medium Close Up (MCU) includes body from the waist up, Medium Shot (MS) includes from just below the hip to above the head of upright actors, Medium Long Shot (MLS) shows the body from the knee upwards, Long Shot (LS) shows at least the full height of the body, and Very Long Shot (VLS) shows the actor small in the frame. In recent decades in film and television the vaguer term ‘Wide Shot’ has come to replace the various kinds of Long Shot described above, but I am keeping to the more finely graded terminology used when I first became involved with film-making nearly fifty years ago. This shows up characteristic distinctions between many films from earlier periods of cinema, and even today, certain important distinctions would vanish if one used only a three level scale.
For comparative purposes, so as to give an obvious relative measure of the likelihood of a director choosing a particular scale of shot in a film, I have taken the actual number of shots of each scale in a film, and then normalized the number to correspond to the number that would have occurred if the film was made up of 500 shots. If you want percentages, you can divide the figures by 5.
I handle the objective treatment of camera movement by tabulating the numbers of shots with the different kinds of camera movement in each film, and again normalizing to the number that would be expected if the film in question contained 500 shots. The categories I use are Panning, Tilting, Panning and Tilting simultaneously, Tracking both without and with panning movements, movement involving the use of a camera crane, and Zooming. I have divided this last category into Zooming straight in or out, and Zooming with panning and/or tilting in my treatment of television shows. Only panning or tilting movements of more than 10 degrees are counted, as small movements to keep the actors well framed as they change their position slightly are made automatically by camera operators, and in general need no special thought about their relation to the director's ideas of staging. The same applies to small movements of a foot or so in the position of the rolling camera pedestal or dolly during the shot, and also of the height of the camera. I do not distinguish the different methods of supporting the camera, so that hand-held tracking and Steadicam tracking go in together with the traditional tracking with the camera on a dolly, or rolling pedestal in the case of TV.
Analysing the complete film is obviously much preferable, but you will not be too far out in general with a forty minute sample. Contrary to received ideas, the overall cutting rate for the second half most films is either the same speed, or not much faster than that of their first halves, though there are exceptions.
Ideally, the analysis should be done by recording the complete characteristics of each shot (scale, movement, length, etc.) in succession down the length of the film. This permits the most complete analysis of all the possible interrelationships between the variables. But although I initially tried this thirty odd years ago, I found that it took about three times longer than the method I have since used, so I have only used it for a few special cases since. My present method collects each quantity sequentially over the length of the film.
Up until recent years, I have always worked with prints of the films I was analysing, and indeed almost exclusively with 35 mm. prints, and I worked with them on Steenbecks and other flat-bed editing machines, which is obviously ideal for complete accuracy. But now I have turned to working from DVDs and VHS tapes. I feed these into a non-linear editing system (NLE), in fact Adobe Premiere on an ordinary PC (though a cheaper NLE would do just as well), and while they were being digitized in real time, I record the camera moves from the window in the digitizing programme screen. For the experienced analyst, this is just possible to do in real time, even for the fastest cut films. Then I go more slowly through the film in the NLE programme, recording the Scale of Shot, which usually requires some stopping and starting and going back, particularly for the films with very short ASLs. I also record the Inserts on this pass. Two more passes are necessary to get the numbers of reverse angles and POV shots. If I have a VHS tape, and one can always make one from a DVD, I usually do these last things on a VHS recorder with a jog-shuttle control, as I can usually manage recording these last two quantities at high speed for most films. Alternatively, it is possible to do the complete analytical process entirely on a VHS recorder with a jog-shuttle control, as I have done when analysing television programmes in the past. For my analytical procedure the standard control system for DVD players is very awkward to use when trying to work directly with the DVD disc.
There are important cautions to be made about the analytical process when working from tape recordings or DVDs of films. The first of these relates purely to the use of recordings made for the PAL television system. These are initially created from film prints that were shot at 24 frames per second when the original film were made, but are always transferred to the consumer medium at 25 frames per second. This means that their running time when played on PAL system devices is shortened by 4% of the original running time. This means that a correction factor has to be applied to the ASL has to be applied by multiplying it by a factor of 25/24. I have applied this correction in the above results. No correction is necessary for NTSC recordings. More important is the question of Scale of Shot determination from video and DVD copies of films. For old Academy screen ratio films, both 16 mm. copies and, even more so, video copies are cropped in all around the frame on transfer to a greater extent than the screen masking when they are shown in the cinema, or on a Steenbeck. The effect of this on the Scale of Shot is fairly slight, as it shifts a very small proportion of the CUs into the BCU category, and an even smaller proportion of the more distant Shot Scales into the next closer category. Since all American feature films made since 1954 are intended to be masked to widescreen on projection, or are shot in one of the anamorphic ‘Scope systems, or in a wide film system, the difficulty does not exist in quite this form for wide screen films. The problem is that films made since then which are shot “flat”, i.e. with spherical lenses on the camera, may have the full Academy image, which was invariably recorded on the negative for American films, transferred to video, and not masked in to the widescreen proportions that were intended to be seen in the cinema. Despite the fact that DVD transfers are virtually always given the correct masking, and there is an increasing trend to releasing VHS copies properly masked in to wide screen, this problem has received a new boost from the shooting of many films in Super 35. In this process, the camera exposes what is called the “full” aperture in the gate of the camera, which is equivalent to the old silent period aperture. This image on the original film is masked in to widescreen or even to ‘Scope proportions by optical printing when making the release prints of the film.
Where possible, I used DVD copies when analysing the sample, and I also check with the VHS copy of the same film where possible. This check showed that in the case of Deep Blue Sea, which was shot in Super 35, the VHS copy had been taken from the full frame, and the DVD copy, like the cinema prints, was taken from the middle of the original frame in ‘Scope proportions. This meant that for any shot much more could be seen of the scene vertically in the VHS frame than could be seen in the DVD copy. That is, if I had analysed the VHS copy, I would have found that the film was shot from much further back than it really was, with respect to the intended cinema release framing. Another difficulty that can occur with films shots in Panavision, or other ‘Scope systems, (as opposed to merely being filmed with a Panavision camera with ordinary spherical lenses), is that full frame VHS copies can be made by “scanning and panning” the 1:2.35 ‘Scope frame. A pan made across the ‘Scope frame during the video transfer will show almost the true height of the frame, so creating no more of a problem than a video copy of an old Academy ratio film, but a scanning cut from one end of the ‘Scope frame to the other, which sometimes happens, introduces an apparent extra cut into the film which wasn’t there before. If there are a substantial number of these, this will affect the ASL slightly. Fortunately, the expert eye can detect most of them, but even I find I have missed some scanning cuts on re-examining a film. But to repeat, as long as you stick to DVD copies most of these difficulties can be avoided.