IMDB link: http://www.imdb.com/title/tt0019760
Submitted by Adelheid Heftberger, Yuri Tsivian, Barbara Wurm on 2008-06-12
Adelheid Heftberger, Yuri Tsivian, Barbara Wurm's comment:
This submission is the first in a sequence. We (the three names above) plan to go on submitting timing data for Man with a Movie Camera and segments thereof in both simple and various advanced modes, and discuss the results in comment boxes as we go. This look-and-talk sequence is part of a project "Digital Formalism: The Vienna Vertov Collection" (www.digitalformalism.org) made possible through a cooperation of three institutions: the Department for Theatre, Film and Media Studies (Vienna University), the ?sterreichisches Filmmuseum, and the Interactive Media Systems Group (Vienna University of Technology).
This measurement is a frame-by-frame count. The count is based on a 35 mm print preserved in the Vienna collection (provenance: Gosfilmofond of Russia) digitized and annotated using Anvil software. Adelheid Heftberger was in charge of annotating this digital copy and locating 5 splits between the film’s 6 reels (not yet identified in this submission). The result was proofread and checked by Heftberger and Edith Schlemmer against the 35mm print on a Moviola (Arri) viewing table and by Heftberger and Tsivian against a shot list compiled in the 1980s by Vlada Petric and Roberta Reeder using a 16 mm print from the Harvard Film Archive.
To translate frame numbers into seconds we proceeded from an assumption that the correct projection speed for this film was 24 fps. This assumption (which to some may sound pretty unorthodox) was based on two kinds of evidence. One is Kevin Brownlow’s detailed investigation into projection practices in the late 1920s which those interested will find at http://www.cinemaweb.com/silentfilm/bookshelf/18_kb_2.htm. The other is a document from the Vertov archive – a music sheet complete with time markers used by an orchestra at the film’s opening in 1929 (see in: Yuri Tsivian, "Dziga Vertov's Frozen Music: Cue Sheets and a Music Scenario for The Man with the Movie Camera," Griffithiana, No 54, October 1995, pp. 92-121) from which it follows it was then shown at 24.
Author: Yuri Tsivian Date: 2008-06-18
This begins our sequence of comments with Barbara, Heidi and me cast as main players.
Since I feel at home with Cinemetrics, and Heidi and Barbara have merely arrived – before we start talking about MWMC (the whole point of the three us being here) let me show them around.
Now – raise your eyes for me from the text you are currently reading, leap the graph, and go all the way up till you hit the line which says “Average shot length.” Very good.
Average shot length (or ASL) is an index widely used in film style studies. It tells us something about film cutting rates.
To visualize what cutting rates are and why they are important to know about I suggest you do the following crazy thing. Plug up your ears and put a veil in front of your eyes – thick enough not to see objects around you but thin enough for you to perceive impacts of light.
Having done so, turn on your TV or, better, visit a movie theater near you.
Will you still be able to guess whether the film that’s being shown is a Vertov movie or a movie made by Tarkovsky or the scene you came in on features a car chase or a romantic gondola ride across Venice?
If you will it must be because the flicker told you. Cutting rates matter, and so does ASL. ASL is a flicker index. The shorter the ASL number, the faster the film.
Now – unveil your eyes, Barbara and Heidi, and look at the ASL value for the film we have submitted. It is 2.3 seconds. Is it good or bad? What does it tell us about MWMC?
We won’t be able to make this number speak unless we compare it with the corresponding number for other films.
How to do this? Easy. Shift your eyes to the left and look for the “Cinemetrics database” hot-spot. Click. Good.
What you see is an alphabetically ordered list of movies submitted to Cinemetrics by many different people in more than 2 years of its existence. The upper row (in red) is clickable. Look to the right and you’ll easily find a button that will allow us to sort films by their ASL.
Click on it, and wait till the system rearranges the list of films.
Let me tell you something while we wait. There are several hundred films sitting there, submitted by some hundred people for a hundred different reasons. So – the database is what statisticians might call an unbiased pool of samples. This means that the variety of films included in this database more or less corresponds to the variety of all films ever made. In this the Cinemetrics database is somewhat similar to the Noah’s Arc.
(Let’s have a drink to the old drunk Noah one day, Heidi. Barbara will have to wait till her baby boy is born, then join us too).
OK. By now the database has sorted itself out according to its films’ cutting rates. The fastest cut films (films with record-low ASLs) come first, the slower one are far down.
Look for our Man with a Movie Camera. Here it is. Amazing. Within its category (=whole films, not fragments, feature films, not shorts) it comes as the third or fourth (tied with October made around the same time by Vertov’s contemporary, compatriot and competitor Eisenstein).
Outrun only by special effects champions Moulin Rouge and Dark City.
Now that we know that MWMC is one of the fastest films ever cut, we can click on its title and go back to our data file about this film.
What do we find under the “Average shot length” line?
Aha – Length of film. This is clear. The only thing we need to keep in mind is that the actual length of a silent movies depended on the speed with which it was projected. Our choice (argued in the “Author’s comment” line) was 24 frames per second.
Go one line down. – “Number of shots.”
This point asks for a comment from Heidi or Barbara or both. There are several shot lists in existence and the number of shots varies in each. Sauzier counted 1,712, Croft-Rose 1,716, Petric 1,682, and Alifragkis/Penz, 1, 695 (see pages 224).
In addition, some time ago our reliable contributor Radomir D. Kokes submitted MWMC using our dedicated tool Chelovek s kino-apparatom: (12) ASL 2.6 – and his number of shots is 1,511 (this is understandable – the film is too fast for anyone’s clicking finger to register all its cuts).
Why this discrepancy between shot lists?
Different prints? Hardly. There are only 3 versions of MWMC in existence: a Gosfilmofond version (predominantly used by Western scholars), the Krasnogorsk version (the fullest: full frame + reel numbers shown before each reel) and the Netherlands version (with the baby-birth scene censored out – Barbara please copy). The difference boils down to just a few shots.
Difference in what is treated as a “shot?” Likely – do we count black flicker-frames and titles as “shots”?
Still – the number of is a moot “?” to resolve. Meanwhile let us declare our list the “definitive one.” Let others worry.
As for us – we now go one line down the register of MWMC statistics.
“Minimum SL” means: the shortest shot in the film.
Hmm – the shortest shot in MWMC = 0 seconds.
Not even 0.1. How can it be?
I feel Heidi is already impatient to address this mathematical conundrum. Let us leave this to her, OK, Heidi?
“Maximum SL” is 22.6 seconds. The longest shot in MWMC.
Which one could it be?
To establish this, look at the graph below.
RULE NUMBER ONE before you start working with the cinemetrics graphs:
see if the word Step found under the graph is highlighted in red. If it is (as here) it means that the graph has been compressed to fit the page. To see every single shot of MWMC you need to reset Step to 1 and hit “Redraw” (you must hit Redraw after any action you want the graph to respond).
Do it now.
As you can see, the graph has grown in length.
You can now look for the longest shot in the film.
Why do need want to do this? For no reason at all. I am just showing off the page. I want you to know what you can and what you cannot expect Cinemetrics to do.
So suppose you need to know which shot in the film stays longer than others on the screen.
You look at the white “icicles” hanging down from the “roof” of the graph.
Each of these icicles represents a shot.
The longer the icicle the longer the shot.
Some icicles – like in the middle of the film – have as good as melted away.
Some are still quite long.
One is so long that it hits the floor of the graph.
How long is it, actually? No problem at all. Change Height to 300 and hit “redraw.”
The graph has grown again, now in height.
We can see the longest icicle’s tip now.
Strain your eyes to see how it measures against the scale given on the left wall of the graph (or on the right one, if need be). The words running alongside it bottom up say “Shot length in seconds.” They tell the truth.
Now -- the longest icicle in longer than 22. This is our shot.
What remains for Heidi to do is to place it on the timeline of the film. The upper side of the graph is calibrated in mites and seconds. Our longest shot is to be found around 12 minutes into the film.
Now Heidi can go to back her Anvil machine and tell us what the shot is about, perhaps why it is so unusually long (unusually for MWMC, of course), and even copy/past a still of it for us to remember.
But don’t hurry away to do this yet.
There are still a few lines above the graph to comment upon.
One of them says “Range” – the longest shot minus the shortest one. In our case the range between the shortest and the longest shots is 22.6 seconds – for the shortest one equals 0.
The next line – “Standard deviation” – is a more flexible and sophisticated way of assessing variations between the long and the short. Like the ASL, it is a mean index. It shows how widely shot lengths are dispersed.
If the Average shot length tells us about a film’s cutting rate, Standard deviation tells us about its “cutting swing.”
The way films are edited is largely characterized by these two variables: cutting rate and cutting swing.
To visualize what this parameters are imagine an abstract film in which all even shots contain empty transparent frames, while all odd shots ate pitch-black.
As I said earlier on, films’ ASL, or cutting rate, is, in a sense a flicker coefficient. The shorter the ASL, the stronger the flicker.
Now, there may be 2 kinds of flickering style. Imagine that all the shots in your film, white or black, are of equal length. Regardless of the rate with which black and white alternate such a flicker-generator would be deadly monotonous. All the teens would be out of the discotheque in a minute.
This may be the reason why I don’t know of a single film with a zero standard deviation, not even 2001: A Space Odyssey. There are directors like Ozu who are not interested in experimenting with their cutting swings and there are directors like Orson Welles who are – in their films fast cut sequences live side by side with extremely long takes. We will need to see how Vertov fits in this continuum.
But this is for a later conversation.
Later on we will also need to look at the red trendline that crosses the graph and play with its degrees.
We will talk more about the MWMC specifics later on.
What I did today was a short tour around the workspace we’ll need to be in and its machine-tools.
The last thing I want you to look at is the last line above the graph: “Statistics type: simple.”
This means that of all the riches that make editing interestingly diverse we have so far submitted naked shot length numbers. There is nothing that tells us whether the longest shot in the film, for instance, is a close-up or a medium shot, a high angle shot or not, a live shot or an object animation.
We can do this on a later stage – and we will. We will resubmit data with additional parameters added. But it is important that we first explore Man with A Movie Camera naked, study its basic anatomy and only then dress it in various garbs – remember, Barbara, our favorite game when we were kids: having a naked paper-cut doll and dressing her in a variety of colored dresses. I don’t know about you, but I found it useful to study the figure thoroughly before I tried the first dress on.
I think I have pretty much exhausted you patience – if you are still there at all. In your comments you do not need to be as verbose as I was today. I certainly will not.
This long beginning is what editors call an “establishing shot” – a shot that sets up the scene (or a conversation) – so it can afford to be somewhat longer than the rest of them. Let’s decide use some cutting rate and cutting swing variations in or triple talk.
What do you think the trendline tells us about MWMC? I think this might be a good subject for the next conversation.
I am also curious about that longest shot and about the possible reasons in the discrepancies between our “total shot count” and other lists.
Author: Adelheid Heftberger Date: 2008-06-24
Duty first: Here we go answering requests
Being adressed personally a couple of times (and impatient for the promised drinks "to Noah"), let me answer as far as the annotation in Anvil is concerned: Miminum Shot Length 0, a mystery? Actually this is due to a technical "problem". The data created in Anvil (anvil-file) has to be exported via MatLab into Excel. We annotate in this way: The first frame of the next shot is the last shot if the previous shot. If we now export these data into MatLab, we substract always 1. And when a shot is 1 Frame long, this leads to 0 Frames then. So, 0 is due to a technical process. Of course the Minimum Shot Length is 1, so it is - depending on speed and in Anvil we go at 18fps, for example 0.18 s. And in this way you also have to consider the Maximum Shot Length.
Ok, so you are curious, which shot is the longest one: Let's have a look. First of all, I can sort the Excel-File, because there I have Start Frame and End Frame, make a subtraction, then I get the Frame-Numbers, which I sort descending. That tells me, that the longest shot has 542 Frames (plus 1) and it is to be found at Start Frame 18631. You see the advantage of being able to navigate through the film using actual Frame Numbers. Result: It is the shot, where several buses are leaving the garage, one after the other. And in my Anvil it is at 17.15 min.
What I would like to see in this graph is types of shots, but that will be added later, is that right Yuri? Because in Silent Films the lengh of Intertitles might be a little deceiving. One of the longest shots in MWMC is an intertitle (fourth longest and fifth longest according to my Excel-list). In MWMC we KNOW that there is only Intertitles in the beginning of the film, but for the others?
What I am especially interested in (Yuri, remember our presentation in January) is a possible segmentation into "semantic units" and how they correspond with cutting rate. My vague theory is, that at the end of "segments" (we will have to think of a name) the cutting rate speeds up and the cinematrics graph is a wonderful visualization for a thought like this. I will have to go back to this soon and thoroughly.
One remark in the cutting swing: Is there a way in Cinemetrics (and at this point, Yuri could you tell me a little about the Advanced Mode?) to visualize it? I would love to see that. I am quite familiar with standard deviations, so a graph including it, possibly also including a time line would be great. Leads us to the question of montage/editing principles.
Total shot count and other lists, definitely. I will work on that, because I am deeply convinced that it is necessary to sit down and do the "boring" homework of comparing empiric data.
Yuri, Barbara, any more questions? Was I clear enough?
Author: Barbara Wurm Date: 2008-06-29
Hello, this is big B. with little b. still inside (feet & fists occasionally extending my bodily boarders like the longest shot icicles in our graph transcend the framing of the diagram (if we work with the wrong combination of step-vertical resolution-height).
So much has been said in your two statements already, that I won’t add any information or bring up new topics for now, even if my main concern is the graph as such, the hidden and obvious laws of representation applied, its underlying principles - like:
Why are they stalactites and not stalagmites? Just a convention? We usually look from left to right and from bottom down? But this convention apparently applies to images; in diagrams we usually have the x-axis at the bottom. – Or another one:
b) What would change, if on the x-axis we had a frame- or time-based scale, not one based on shots / shot length (since approx. 1 cm in the default representation mode – step 3, vertical resolution 10 pixels/sec - can represent 5min 42sec as in the first marked step, the one between 0:00 and 5:42, BUT at the same time also only 9 seconds - as in the 9th step - the one between 41:26 and 41:35). – Or another one:
c) Wouldn’t it be great, if the graph worked interactive? So that in order to find out, which shot the longest icicle refers to, one wouldn’t have to guess where approximately in the film it can be found, and wouldn’t need to go back to the film again - but this is just a technical remark and I am sure, it was one of the first things anyway that came to your mind when you set up the tool. Heidi explained the advantages of the frame-by-frame scrolling possibility in Anvil. It’s an exact way of automatically establishing (and detecting) frame no. / shot no. / AND the corresponding frame (film still).
My 4 topics & questions for now:
1) (to Yuri): STANDARD DEVIATION
Despite your exceptional tour-guide qualities, I still haven’t understood the STANDARD DEVIATION or SWING thing. What exactly does it mean? How can I try to imagine it (other than visualizing a virtual abstract black or white flicker film, which I find hard to do)? Is it an "average" value to "illustrate" HOW OFTEN (in our case: every 2,5 sec in average) there is a change between short and long shots? Does it usually also include the “radicalness” or “roughness” (Semyon Timoshenko called it “stepen’ rezkosti”) of changes of shot scales (from ECU to ELS; etc)?
Several questions relate to this:
- the statistical one: what exactly is represented by this average value? ASL I understand. One ASL compared to another gives me information about the speed qualities of a director, an editor, a certain period, a certain genre, whatever. Accordingly we measure in seconds. But Standard deviation? How can seconds be units for the “swing” – isn’t it more like a differential value?
- the mathematical one: A film with zero deviation is one with absolutely the same shot length for all shots; “monotonous”, as you said. But which reasonable number or figure would represent the opposite cutting swing? If zero is the minimum, what would the maximum be? Can we grasp it? Is it a liminal value? If you compare one cutting swing rate with another: What would it tell us? A. is more diversified than B. (since the tool doesn’t offer red alias clickable features here – does that mean that the standard deviation value is generally considered less important than the ASL?)
- the aesthetical one I: Does it have an actual influence on perception? How? (the higher the value, the more entertaining a movie?). Isn’t it rather a combination of the knowledge of a certain value/rate and the perception of a film? In a Jim Benning film I know that I am in a Benning film, so I know what to expect – most probably all shots are of equal length. But neither does this lead to my perception of the film as “boring” (rather the opposite here: people can get pretty excited about these kind of experiments), nor would I feel more entertained, if one shot were in fact longer than the other. I wouldn’t even notice. But Benning tells me. And if I didn’t know, now Cinemetrics would tell me (I get that much).
- the aesthetical one II: The interesting thing is that to me it seems we can only interpret standard deviation in combination with the ASL. The flicker effect relates to high speed cutting rates only. But what about the Tarkovskian way of editing? Isn’t standard deviation a value that highly depends on other rates and values? I assume that there is a strong interdependency between ASL and Standard deviation. But which one, I don’t know. And I am not sure if one could generalize it. I think in perception we notice this swing rate rather within a range of average values, neither with extreme values close to zero nor with extreme values close to infinity.
I join Heidi here, who wants to know more about the visualization possibilities for the cutting swing.
2) (to Heidi): MINIMUM SL (the “zero conundrum”)
Does your explanation of how we count shots in ANVIL, and how we exported the data to an excel file in order to ease and accelerate the submission of MWMC to the Cinemetrics tool, and how this came along with a small but important mathematical error (since 0 and 1 show rather different mathematical behavior, as we all know) mean that in order to find out the exact number of frames per shot, we now always have to add 1 – at least, if the data comes from Anvil or other frame-by-frame based tools?
And: What is the background of annotating the way you describe it? (Why is the last frame of a shot treated / counted as if it were the first frame of the following shot)?
3) MAXIMUM SL
In Cinemetrics (Step 3) it is around 18.3 seconds long.
Only in Cinemetrics (Step 1) it is 22.6 seconds long. Step 1 is the truth, we heard.
In Cinemetrics (Step 10) it is not even the same shot that seems to be the longest (for whichever reason).
Doesn’t – if there needs to be a compression to fit the graph on the page – also the actual scale need to be compressed then? Is it really impossible to find out about the actual shot length in seconds in a different compression step? Why is there still a scale saying “shot length in seconds”, which leads you to the assumption that (if in Step 3), the shot is 18.3 seconds long, and (if in Step 10), it is shorter than another one occurring at around min. 48 or 49?
3.1.2. Projection Speed
Heidi went back to the frame based counting: MSL is 543 frames long.
As you have guessed by now, I am not much of a mathematical brain. But I am trying hard and I noticed a little error in Heidi’s explanation of the minimum shot length – 1 frame is not “0.18 sec”, but:
An assumed projection speed of 24 fps means that each frame runs for 0,041666 (period.) seconds.
An assumed projection speed of 18 fps means that each frame runs for 0,055555 (period.) seconds.
Let’s multiply each speed rate by the number of frames (543) just to make sure we are talking about the same MSL:
For Cinemetrics we get: 22,625 seconds
For Anvil / DigiForm we also get: 22,625 seconds
(That’s not very surprising, since both data is based on the Anvil protocol).
How come Yuri expects the longest shot at around 12 min. – and Heidi detects it at 17min:15sec.?
Again the answer is: projection speed.
The MSL in Cinemetrics occurs at approx. 720 sec.
The MSL in Anvil / DigiForm occurs at exactly 1035 sec.
Is there a way to find out how - by taking into account the according projection speed (not by straining eyes) – we can find out EXACTLY at what minute the maximum shot takes place WITHOUT Heidi’s Anvil information about the fact that it is Start Frame 18631 we are looking for?
This is what I presume: only if we DO know the exact frame number of the start frame of the longest shot we can transfer it onto a time based scale:
In Anvil / DigiForm the following arithmetic applies: 1/18 x 18631 = 1035, 055555555 (period) = 17min. 15 se
Same arithmetics for Cinemetrics (1/24 x 18631): 776,291666666 (period) == 12 min. 56 sec.
If my presumption is right, then Heidi’s statement needs to be emphasized: "You see the advantage of being able to navigate through the film using actual Frame Numbers." – It is not only about practicality, it seems to be about exactness, too.
4) Discrepancy of total shot count:
- Is “Sauzier”: Sauzier, Betrand: An Interpretation of TMWMC. Studies in Visual Communication 11.4 (1985), 34-53 or is it his dissertation?
- Who and when was Croft-Rose?
- We DO need to find out, I think, a) which prints were used; and b) which counting techniques were applied.
- But this won’t keep us from clarifying HOW to treat a) BLACK FRAMES, and b) INTERTITLES.
Heidi / Yuri: are there certain rules for each method / Anvil and Cinemetrics?
Shows the development and tendency of shot length over the whole film. For MWMC the shots generally tend to get shorter, the closer we are to the end of the movie.
6) We hereby promise to be shorter next time. B&b
Author: Yuri Tsivian: how we proceed Date: 2008-07-08
Spent most of my time on Monday observing my one-year old grandson Gabriel walk and fall all by himself. To rise after he falls the first thing he does is to shoot up his ass, unbend his knees and, as his hands are still touching the ground, use them to walk his torso towards his legs; then, he unbends his spine, and is erect again. Don't even try to use this method, adults!
Back to my laptop found the above comments coming from Barbara and Heidi. I will address some of their points piecemeal (rather than in one longish comment) announcing the topic in the "author" line. I will mainly address the points that take our conversation one step forward -- if I skip this or that point it will be not because I find it unimportant, but because it may lead our conversation astray.
So -- see my next comment below.
Author: Yuri to Barbara: why shot lists differ Date: 2008-07-08
In my opening comment I suggested Barbara might want to look into why different Man with a Movie Camera shot lists made by different people at different times differ in their resulting number of shots.
Here what she says -- with my replies inserted:
BARBARA: Discrepancy of total shot count:
- Is “Sauzier”: Sauzier, Betrand: An Interpretation of TMWMC. Studies in Visual Communication 11.4 (1985), 34-53 or is it his dissertation?
YURI: the former.
- Who and when was Croft-Rose?
YURI: Stephen Crofts / Olivia Rose: "An Essay Towards Man with a Movie Camera." In: Screen 18:1 (1977), pp.9-60.
BARBARA: We DO need to find out, I think, a) which prints were used; and b) which counting techniques were applied.
YURI: right; but as we know the prints known in the West in the 70s and 80s were all identical; so the disrepancy must have been due to different counting methods plus human error. Heidi and I found many a human error (including simple math) as each of went through the best and most detailed list made in the 1980s by Vlada Petric and Roberta Reeder (the yet unpublished typescript of which was given to me by Vlada Petric).
BARBARA: - But this won’t keep us from clarifying HOW to treat a) BLACK FRAMES, and b) INTERTITLES.
YURI: Heidi's Anvil treats each black frame as a shot. This seems to be the best solution, for Vertov intentionally uses them to create "black flicker" effects -- as he does in the railroad sequence, for instance. She also treats intertitles as shots, as is the habit with all Cinemetrics users.
Author: Y to B: from where the icicles grow Date: 2008-07-08
In my opening comments explaining how to read Cinemetrics graphs I used the word “icicles” to refer to white lines that represent shots of different lengths. Barbara had a question about this.
BARBARA: Why are they stalactites and not stalagmites? Just a convention?
YURI: Let me specify what it is exactly you are asking. In the world statistics, we are more used to quantitative values being arranged bottom up, not top down. For instance, if your graph represents the daily fluctuation of the euro against the dollar, a European statistician will more likely arrange the euro values along the bottom of the table to show off proudly how strong this currency grows each new day. And, sad as the down-looking trendline may look, American Wall Street news would give the dollar the bottom place, for it is the fate of $$ that interest us more in Chicago.
Likewise, if we were interested in the SLOWNESS of film editing it would make sense to align shot lengths along the bottom of the graph, as indeed these are represented in a similar graph generation for his purpose by Stavros Alifragkis who was using Excel statistics tools.
This is the graph produced by Stavros.
As you can see, the peaks in his graph are the longer shots, a reverse image of the corresponding graph measured in Cinemetrics (above). This is what Barbara has called “stalagmites.”
The reason why we use “stalactites” instead is because we are interested in SPEED rather than slowness, in CUTTING RATES rather than shot lengths per se. That our red trendline (of which more later) looks up as the film progresses shows there is an ACCELERATION of cutting.
Author: Y to H&B about the 0 mystery Date: 2008-07-09
In my opening comment I suggested that Heidi may have a comment on a seemingly paradoxical fact that the data list found above the graph yields “zero” for the Minimum shot length. If the length of an object equals zero – is this still an object or, as common sense tells us, there is no object at all? We have had two responses to this, one coming from Heidi, another, from Barbara.
HEIDI: Minimum Shot Length 0, a mystery? Of course the Minimum Shot Length is 1, so it is - depending on speed and in Anvil we go at 18fps, for example 0.18 s. And in this way you also have to consider the Maximum Shot Length.
BARBARA: I noticed a little error in Heidi’s explanation of the minimum shot length – 1 frame is not “0.18 sec”, but: if an assumed projection speed is 24 fps it means that each frame runs for 0,041666 (period.) seconds. And at an assumed projection speed of 18 fps each frame runs for 0,055555 (period.) seconds.
YURI: Barbara is right, Heidi, dear. What you meant by “Shot Length is 1” is that every shot, even the shortest one, should be counted as one shot – but the phrase is unfortunate, for Cinemetrics measures shot lengths in terms of seconds, and in many cases it is <1. Nor is your statement that “depending on speed and in Anvil we go at 18fps, for example 0.18 s.” is fully correct. 1/18 is not the same as 0.18”, it is 0.05”. So – this is why Cinemetrics assigns the zero-value to one-frame shots. When the system encounters a shot which is that much smaller than a decimal of a second it rounds it down to zero. I think we can live with this “0” figure above the graph as long as we keep in mind that what it actually says is that the Minimum shot length is close to zero.
Author: Y to H&B about Swings and things Date: 2008-07-10
In my opening comment I refer to two important variables of film editing. One of them, we recall, is called “cutting rate,” the other, “cutting swing.”
In terms of statistics, the “cutting rate” (the tempo with which this or that film is cut) is linked to the ASL – the lower the average shot length, the higher the rate. This point I assume was well explained – it raised no eyebrows either on Barbara’s or on Heidi’s face.
I had less luck with the “cutting swing” however – and even less with the statistical notion of “standard deviation” which as I claimed was linked to this elusive variable. I guessed this from the feedback I got from Barbara and Heidi.
BARBARA (gestures violently):
I still haven’t understood the STANDARD DEVIATION or SWING thing. What exactly does it mean? How can I try to imagine it (other than visualizing a virtual abstract black or white flicker film, which I find hard to do)? Is it an "average" value to "illustrate" HOW OFTEN (in our case: every 2,5 sec in average) there is a change between short and long shots?
YURI: Let me try another way of explaining what the term “cutting swing” refers to.
First of all, what is a “swing”?
Imagine a pendulum that swings like this:
Let us say that at its full swing our pendulum goes through 5 positions: 1-2-3-4-5, and 5-4-3-2-1 and so on. Very good.
Let us now assume that this pendulum has a "limiter" of sorts that allows us to reduce its swing by half. When the limiter is on the pendulum works in what can be described as half-swing: 2-3-4, and 4-3-2, and back again. Still with me, Barbie dear?
If you are, you will probably agree that if the position of the pendulum across a period of time is merely 3-3-3 the pendulum can be said not to be moving at all. Amen.
As in clocks, so in films. Suppose that a demented film editor comes to your cutting room and says: you know what, I have looked at many films and have calculated the optimal ASL for a perfect romantic comedy. It is 3 seconds. And I’ll see to it that all the shots in your romantic comedy are exactly 3 seconds long.
What happens then to our pendulum, dear Barbara? You’ve said it – the pendulum is dead calm. it's on 3-3-3. It shows no swing – it hangs. Hardly a good recipe for a romantic comedy or, for all I know, for anything remotely romantic.
Now – some directors love playing with extremes. Orson Welles or Jean-Luc Godard, for instance, will typically use very short and very long shots in one and the same film. To go on with the pendulum metaphor, these people will say: I want my pendulum to be always at full swing. So I will go back and forth between 1 second and 5 second shots.
Others might say: I want my film be fast (or slow) throughout. In terms of the pendulum it is like swinging 2-3-4, 4-3-2. Never go as far as 5 or 1. It’s a sound stylistic choice. Some very good directors – like Ozu – preferred to narrow the swing of the cutting pendulum for their own rhythmic considerations. We have a very good case study of such a strategy submitted to Cinemetrics by Matt Hauske, the University of Chicago student.
Now – the swinging pendulum is, of course, too crude a metaphor to account for subtle temporal variations in editing within the duration of a film. This is why when speaking of films’ cutting swing we in Cinemetrics prefer to use a more flexible mathematical notion “standard deviation.” Not a math person myself, I found an easy and helpful explanation of the standard deviation concept for those who might not know math by Robert Niles. Niles speaks about food, not about film, but as we all gain and lose weight from time to time you’ll find his examples easy to extrapolate to shot lengths.
HEIDI (her brow furrowed with thought): One remark on the cutting swing: Is there a way in Cinemetrics to visualize it? I would love to see that. I am quite familiar with standard deviations, so a graph including it, possibly also including a time line would be great. Leads us to the question of montage/editing principles.
YURI: there is that famous “bell chart” Niles talks about to represent standard deviation graphically, but it is of little use for Cinemetrics for, as you just said, in our case the important thing is to see how exactly shot length fluctuate across a film’s timeline. Let’s put aside the standard deviation figure for a moment and talk again about the cutting swing. One way of visualizing the cutting swing is to play with the trendline -- changing what is known to be its “degree.”
In order to play with the trendline do it you’ll need to come with me all the way up this page to where the graph for Man with a Movie Camera is sitting. It is too far from the text you are reading to be able to follow what I say and to push the buttons below the graph. Therefore -- stay here. I'll do the changes for you using the helpful "Insert Cinemetrics graph into the comment" option. Later you may want to try out graph buttons yourselves.
Normally before we start working with graphs we make sure we change “Step” to 1 to see the film in full rather than adjusted to the size of the page, but as we are now looking at the tool, not at the film, there is no need to do this change now.
What we’ll need to be changing are numbers 1 to 12 in the box opposite the phrase “Degree of the trendline.”
But before we do any changes, the graph looks like this:
For now, the default degree is 1, and the red trendline is straight. It climbs from above 4 to 1 seconds which means that overall as the film unfolds its ASL decreases and the cutting rate accelerates. But then, does it happen as steadily as the trendline seems to indicate?
To find this out we’ll need to start changing its degree. Any degree will tell us a slightly different story about Vertov’s editing strategy, but, to make things faster, let me change degree 1 to degree 6 and hit “Redraw.” The result will look like this:
You can see that the red trendline now looks like a wave. Roughly, the curves of this wave correspond to the ups and downs of the black-and-white background which, as we recall, is a diagram of different shot lengths. The background is rocky, the trendline is smooth, therefore this process is known as “data smoothing.”
It is by the profile of the wave that we can glimpse the film’s cutting swing. In our case the profile is relatively low, but in a film like Godard’s A Bout de Souffle (1960) in which the Standard deviation is twice the size of ASL the trendline at the same Degree 6 cam look quite stormy. This is how Godard's graph looks:
Before I finish this comment (bedtime for Barbara, beertime for Heidi), let me return to Man with a Movie Camera graph and direct your attention to another thing. At Degree 6 of the trendline we do not see anything specific about the left-hand and right-hand borders of the line. Experience shows that for most of the filmmakers the beginning and the end of the film are the spots which they tend to edit differently that the middle. We know too little about the laws of editing yet to offer a sensible explanation of why this is so, but we know enough to be paying attention to these two spots -- the beginning and the end.
Now, if you change the Degree of the trendline to its upper limit – 12, the line will become sensitive enough to detect fainter impulses at the beginning and at the end of Man with a Movie Camera.
I'll do it for you.
See? There are two distinct protuberance at the beginning and the end, somewhat similar and symmetrical in form.
I suggest that these and other curves and curvatures of MWMC become the subject of our next conversation.
Author: Yuri about the future plan Date: 2008-07-10
The above comment concludes the first part of our conversation a propos Man with a Movie Camera in the eye of Cinemetrics. As palnned, this first part was less about Man with a Movie Camera and more about Cinemetrics and its tools. We now go to part two which will be about what the Cinemetrics tools tell us about Man with a Movie Camera.
I suggest we start it off with asking why. Why does the straight trendline at Degree 1 climb and not descend (in the statistics lingo they say: shows a positive rather that negative slope). Why are the fluctuations of the trendline at other Degrees exactly what they are? Why certain sections of the film (like around 30, 40, 50 and 60 minutes into the film, as the upper scale of the graph shows us) are cut super-fast while at other times the editing tempo is more laid-back? Is there a pattern behind this periodicity?
Heidi, you told me you had a degree in engineering and had worked in a chemical lab before switching to studying films. If so I am sure you remember a picture of that bearded Russian called Mendeleev. Barbara as a Russianist must have heard about him too. He became famous for observing recurrent ('periodic") trends in what appeared to others to be a chaotic distribution of elements; he aslo explained this periodicity using what we in humaities might call "semantic" fetures inherent in these elements.
So if Heidi and Barbara go all the way up to the MWMC graph, play with its trendline degrees and scrutinize its "icicles" (don't forgret to set Step at 1) I am sure they'll come up with a set of ideas and hypothesis about the internal strucuture and periodicity of Vertov's film. If you use comment boxes to share you thoughts about this we will be able use Gunars' help to go on and anatomize (or "atomize", as Mendeleev might say) MWMC further.
Author: B to Y: Re: from where the icicles grow Date: 2008-07-20
This is very interesting & after a while sounds even logical ;). Amazing how a simple decision about the way values are depicted (like mirroring) can change semantics. On the other hand, if we look at statistical film-analysis literature and the graphs used there, the one produced by Stavros is not at all an exception but rather common (I am referring to a collection of different visualisations of formal aspects of films, which our colleague Anton Fuxjäger has put together going through common text books).
All I am trying to say: the shifts you describe, 1) from "interest in slowness" to "interest in rapidity", and 2) from "interest in shot length per se" to "interest in cutting rates and acceleration" need to be emphasized and considered, when we talk about the epistemological value of visualisations. They maybe "sound" less important than they actually are!!
But they are mean shifts, too, because in acknowledging them, we actually admit that we are using tools that are not simply "objective", but rather carry with them already certain suggestions or assumptions we have, before we start measuring and analysing. Is this a (masochistic) way to undermine oneself or rather: the so called objective method? - or is this rather what people like Lorrain Daston and Peter Galison are arguing: that "objectivity" in research has always been and still is being "undermined" not only by subjective factors (like the researcher), but also by the tools and instruments of "representation".
However, this would lead to yet another epistemological inquiry and I promised to myself that I`d try to keep pace with your empirical or rather inductive drive (let`s just keep these things in mind, I kind of like to ponder over them). DV's the MWMC then, soon ...
Author: B to Y&H: why shot lists differ Date: 2008-07-20
My encouragement to lay bare our devices was mainly addressed towards Heidi, since from what I`ve heard so far about her/our "actual frame-by-frame" annotating method, black frames are slightly more complicated than you describe it, Yuri. Of course, whereever intentional black frames (meaning: a certain number or positioning of them) can be found, they matter - Petric has indicated the flicker-creating effect in the railroad sequence in his book, I think. Apparently it can be regarded as something like the deconstruction of Roget's famous experimental discovery of the stroboscopic effect: it actually disrupts the seemingly continuous perception (based on the phi-e.) by introducing bars (the black frames) - which in this very railroad scene produces the perception of the movement of the train as not only "tik-taked" (ta-tá-ta) but also as backwards. Right? We should try this: Watch it at 24fps to 18fps. See if speed matters. Has anyone done this? Does the train move up or down - forward or backward?
But: there are other black frames (single ones, which, I think, are not being counted in our system, right?). Is there a reason for their existence? Is the problem of the "typical fading in / fading out black frames" we have to deal with a problem that occurs when a print is being copied again and again (a problem of transmission); or is it a "contemporary" problem, a problem of shooting, camera lenses, irises and photosensitivity - which would also have an effect on the editing/cutting principles of such finnicky people like Vertov & Svilova.
However: I do accept that these questions need to be put aside since on a statistical ("average") level they hardly matter, I guess. DV's the MWMC then, soon ...
Author: Yuri to Barbara on data representation Date: 2008-07-20
You are absolutely right. The philosophy of statistical graphs is high for high, low for low. If you are interested in the growth of a foetus related to the pregnancy month (present company excluded) a correct graph will always lead uphill left to right. Conversely, the graph showing the likelyhood of you getting pregnant as you age will lead downhill; but if you want to know how fast your chances of having sex without risking pregnancy grow with age the sane graph will look up, not down. All depends on the word we use.
Our key word is "speed" (and not "fuel expenditure," for instance).
Author: Barbara to Yuri on "ustanovka" Date: 2008-07-20
Hej - I am desperately trying to understand the trendline thing and you are distracting me with your "ustanovka" on (my) pregnancy. In order to contribute at least one interesting question to the trendline, the key word - for someone like me - unfortunately will always be fuel expenditure and not speed. What a funny idea: to conquer (vertov's) speed by (our own) speed. So: What should I do: watch MWMC again first and then take an estranged look at the graph, or study the graph plus all relevant comments - yours, others on the discussion board & elsewhere (which is what i started doing a while ago)?? Now: there is a question.
Author: Barbara: Below zero Date: 2008-07-20
May I ask: Why is the trendline in degree "12" below zero? - Is it because of the "exaggeration" factor in degree 12?
Author: Barbara about trendlines in general Date: 2008-07-20
Sorry about de-speeding the process again. But if I want to contribute at least one useful hypothesis on the curve "I see", I need to know what I see ... Or: Do I just have to get familiar with the fact that the red line doesn't visualize the actual range - nor does it represent the actual deviation factor? I have tried to follow Gunar's explanations about "Range and degree of trendline", and also your comments about "tides and ebbes", but to be honest, I don't think, I fully understood. Is it important to understand, in order to "read" the curve? - I have the feeling that yes. And that is has something to do with my (untackled) question: "But Standard deviation? How can seconds be units for the “swing” – isn’t it more like a differential value?" (Ok, I learnt: not "differential values", but "polynominal lines"). Anyway. It's again the tool, I am trying to understand, Vertov has to wait. -- Let me just add another simple question about the Godard graph instead: When the graph reaches a certain "peak" at around 27:00 - does that mean that generally the shots tend to get longer (and shorter afterwards, until around 60:00) - or does it mean that at 27:00 Godard's playful experiments with shot length variation reaches its peak and that after an hour or so his need for variation is exhausted? And one other question - concerning the pendulum: will a constant change between 1 sec and 5 sec shots lead to a higher variation value than, a) a constant change between 2 sec and 3 sec; or b) a number of non-constant changes (without any patterning) like 1 - 5 - 4 - 2 - 1 - 1 - 1 etc. etc etc. .... I am pretty confused. But at least honest.
Author: Date: 2008-07-21
Barbara, your questions make total sense. And there all the time we need to come to terms about tools and terms before we use them to cut up our main dish, Dziga Vertov. So, no rush.
Let me first clarify my previous comment. I used two arbitrary examples of statistical graphs (it could be temperature graphs, fuel efficiency graphs, anything -- they are all alike) to address your question why the bar chart which Cinemetrics uses to graphically represent successive shot lengths has its bars attached to the upper side of the graph rather than to the bottom as is habitual in statistics. My reply was – because as a rule statisticians attach to the bottom what they are more interested in. In cinemetrics studies we less interested in shot lengths per se tan we are in the frequency with which they succeed each other. Hence the frequency of cuts between shots is inversely proportional to the length of these shots we just turned the picture upside down. (Disregard my previous remark about “speed” – I was referring to car speed versus fuel economy, but then realized that the analogy would be more confusing than helpful.)
You ask how to read the red curve. Rule # 1: ALWAYS READ IT AGAINST THE BAR CHART (the family of white hanging “icicles”) each bar of which represents a shot in all its length. When you learn to do this you will have no questions left about what the red trendline represents.
For example: if you now go all the way up to the graph for Man with a Movie Camera and adjust the trendline to Degree 12 (the maximum) you will easily notice that whenever its film shots (=white “icicles”) grow shorter (e.g. 3 minutes, 58 and 65 minutes into the film) the red line tends to form a higher wave; conversely, whenever shots become longer there is cleavage between the waves. (I am sure you have not forgotten to set “Step” to 1 as I taught you last week? Good girl!)
This also answers your Godard question, I think. You asked it about the trendline behavior at Degree 6 in the graph for Godard’s A Bout de Souffle (1960) which I pasted in one of my comments posted earlier on as an example of a rather high cutting swing. You wondered whether the cleavage observed around 27th minute into the film perhaps meant more than a simple fact that Godard’s shots become longer around this time. The answer is no, nein, niet. It points to this and to this alone. The trendline is not a value – it’s a simple pointer.
This leads us to your question about standard deviation and why it is expressed in seconds. Well, distinct from the trendline, standard deviation is a value. Now I hope a car ride metaphor will be more helpful than it was in my previous comment. Imagine that you drive from point A to point B on a road as empty as you find them in the Death Valley desert; and imagine you drive the same distance in Austria through all the towns and villages and Alps and sheep it is famously full of. The average speed is the same, but as you drive across the desert you don’t need to change it too often, while with all these Austrian speed limitations you’ll be forced now to drive slowly, now real fast.
Get what I am driving at? Yes, there is a mechanism to distinguish between the two rides in statistical terms. This will be about how much ON THE AVERAGE the Austrian driver DEVIATES from the AVERAGE SPEED, and how little the American driver does. It is only logical that the results are given in terms of speed – on the average the Austrian deviates from 40 km/h by, say 20 km/h, and the other guy only by 5. (I hope I get this right – I know there is more math under this than I can handle).
Author: Barb: 1 question, several observation Date: 2008-07-24
QUESTION: What B wants to know in more detail.
In your comment on Radomir D. Kokes' submission of DV's MWMC you say the following: "We learn for instance that its cutting rate (ASL) and cutting swing (St Dev) are more or less the same (the "Greek cross" ratio)." -- But. What exactly do we learn from that?
(That is to say: even if - according to Wiki's explanation of standard deviation - it is a "useful property of standard deviation [...] that, unlike variance, it is expressed in the same units as the data." - I don't believe that it makes sense to draw any sort of "ratio" referring to the two bars of a cross. I think the visualization of this kind of ratio should be parallel, not rectangular, no? ... you see, I am still trying to understand what kind of "relation" we can speak of, between the average value and the average standard deviation)
OBSERVATIONS: What B sees (trendline, MWMC, degree 12).
1. DV doesn't seem much of a swinger. no real mountains, hills only.
2. with twp significant exceptions: beginning and end = strong framing, dramatic explosion in the beginning, smoother (but hyperdramatic in terms of real number / ASL) ending
3. the below zero 'value' (which is not a value, I know) is still alarming.
4. overall only 4 "peaks", in 2 out of 4 cases (the 2 "inner" ones) rather plateaus than peaks, anyway. A bit like the coming up stages at the Tour de France (the ones after Alpe d'Huez - regardless to Austrian speed limiations by the way an Austrian named Kohl (cabbage) was actually not doing too bad lately, he is still wearing the shirt with the red dots, don't know if there are any aquivalents in sailing though...).
5. this is disregarding - so far - the fact, that there ARE little peaky-peaks in between, significantly exposed during the rather numerous super-short-passages:
- the second drastic period of super-short shots (somewhere between 8:00 and 9:30)
- the third one (around 11:30)
- the fourth (between approx. 29:30 and 30:20)
- the fifth (between 36:30 and 37:00 - interrupted by obviously only one comparatively long shot)
- the sixth, super minimal ASL sequence (at around 41:00 - 41:40 - NB : if my min:sec estimation is not all too wrong, most supershort-phases go on for about 1-2 minutes)
- five less significant ones inbetween,
- a seventh (between 58:40 and 59:00 - completed by longer-shot interruptions)
- an eight (between 62:40 and 63:20 - which is interesting, because it shows some kind of rhythmical short-long-short-long-patterning)
- a ninth (between 63:50 and 64:50, interrupted by one longer shot and a little longer-shot-3head-family)
6. What do we learn from the observation you made - that the framing peaks are "somewhat similar and symmetrical in form"?
7. Another interesting aspect seems to be the appearance of the two inner hills, as I called them (see point 4). They also seem somewhat similar in form. But what seems even more important is the fact that if we look at the film as a whole unite, they are not symmetrically positioned, but rather tend to occur towards the end of the film (the first appearing right in the middle of the film, the second one about ten minutes before the end peak). Even if we still haven't had a closer look at the parts/reels - I don't even know, where exactly they start - this "trend" reminds me of Timoshenko's graph of a "normal film" - without going into detail here about what exactly he refers to by using the term "udarnye mesta" (stressed positions) - the little circles.
Semen Timoshenko: The Art of Cinema and the Montage of Film. Leningrad 1926.
Intermezzo: FORGIVE ME: ONE REMARK - coming back to what I named question "b)" in my first entry (2008-06-29) and to an already closed discussion: the "Length of the film representation"-question, as you call it on the discussion board: Has Gunars ever tried to draw a graph, where the x-axis is based on an "evenly calibrated time flow"? I really wonder what our graph would look like.
starts at approx. 7.8 ASL (vs. 4.0 - as indicated in degree 1)
ends with a first peak between 3:00 and 4:00, with an ASL around 2 seconds only; that's a pretty huge rise, I think = significant acceleration.
a smoother decline again, deceleration, down to nearly ASL 4.7 somewhere between 12:00 and 16:00 -- NB: this was also the phase that included the maximum shot length: 22.6 sec. long, appearing at 12:56, as I calculated in accordance with the frame number, Heidi fed us with --
phase 3 merges into phase 2:
a long, pretty steady, even even trendline, with a very smooth rise at around 16:00
constant, overall smooth rise - "ignoring a barrage of very short shots (as around 30 minutes into the film)", as Yuri called it, until it reaches a real plateau
the plateau ... which actually strikes me nuts, when I take into account rule nr. 1 "ALWAYS READ THE TRENDLINE AGAINST THE BAR CHART": this barrage (which is, of course, not really going on for such a long time - only around 40 sec. ... taking up around 8 percent of the overall film length represented...) --- but imagine the plateau: 40 seconds of super-minimal-frame-shots (ASL approx. 1.3).
another smooth decline within this barrage - which scene is it by the way? ... hand-movements?; moving over a pretty swell short-long-short-long-shot-structured organpipe-configuration; going down to ASL 3.0
a very small (lower) plateau at around 51:00 is the starting point for the next smooth rise ...
reaching an absolute top plateau in terms of ASL (1 sec) ... going on for about 4 seconds altogether, so there must be plenty of 1- or 2-frame shots here.
another smooth decline starting at around 59:00, going down to ASL 2 again (which is still below the ASL of the film in total (2.3);) at around 63:30.
last ascent = a pretty sharp acceleration taking more than 3 minutes, travelling a bit in the below-zero-zone (patterned quite neatly), and ending - with a slightly reconciling touch - with two "longer" shots - three seconds of pure contemplation.
You see. I still haven't tackled any questions you rose in (y)our future plan... (the "Why"-serious). Let's see, if at some point our (at least: your) explanations finally will become "absurdly easy".
Author: Yuri to Barbara abd Heidi Date: 2008-07-24
Thanks, Barbara, this mapping and thinking is very useful. Now that we know in detail about the peaks and plateaus and valleys in the graph we can begin to think of how to interpret them.
I hope you do not mind if I suspend your questions about crosses and axis for a while as leading this conversation into interesting side-alleys which we may as well postpone visiting in order to move ahead.
Moving ahead is moving towards Vertov and the interpretation of his film.
To pave the way towards this I want to ask (both of you and myself) the following question.
What does the Average Shot Length tell us? What makes some film directors cut their films faster and others cut their films slowly?
There can be two answers to this, long and short. The short answer is: nothing. Nothing MAKES you take this or that road. It always will be your free stylistic choice.
But one can also say: of course, but then, even when you have opted for your cutting style to be fast or slow, your shot lengths still vary within the duration of a film. What is it that makes some shots longer and some shots shorter – and this, speaking about one and the same film? Another way of asking this is: what is it that makes the film’s pace accelerate at some sections and slow it down at other section?
To answer this we’ll need to think about STORYTELLING CONVENTIONS. That wonderful little diagram which Barbara cut out from the Russian 1926 book on editing by Timoshenko has this caption over it: “The line for a normal film will look like this.” What he meant by the “normal” film was: the film that follows most of the storytelling conventions largely shared by filmmakers, novelist and playwrights alike. He might have also added: conventions that shape the expectations most film viewers share as they come to the movies to watch films.
Can one pinpoint and name at least some of storytelling conventions that can be said to influence the length of this or that shot?
After having looked on a weekly basis for more than 2 years at most of the graphs so far submitted to Cinemetrics I think I can.
I’ll try to now. Tell me if this makes sense. We’ll then try to figure out which of these conventions works for MWMC.
I suggest that we first single out three categories of storytelling conventions that tend to bear on internal fluctuation of cutting rates within what Timoshenko calls a “normal” film:
1) Event-driven ones.
2) Story-flow related.
3) Conventional editing patterns.
Let me explain what I mean by each.
1) As you may have noticed some events in our daily life appear to have an in-built kinetic program of sorts dictated by what they are about. When a person is alive it normally moves; the dead don’t. It may be because of this undeniable fact that most of the funerals I have happened to attend were slow-moving affairs, as were marches and violin pieces played at them. Likewise, I do not recall too many films in which a funeral procession would be the liveliest piece of editing in the film – other than in Clair’s L’Entr’acte, but then, it is exactly because it goes by contraries that this film is called a Dada movie. Take two Griffith movies I once submitted: Isn't Life Wonderful: (6) ASL 6.9 and America: (6) ASL 4.8 In my comments under these films you will find explanations like: slow-cut death scenes; slow-cut romance scene. This is what I’ve proposed to call EVENT-DRIVEN CUTTING RATES (or ED-factor). Try to find a car chase or a show-down that would be cut as slowly as a funeral afterwards.
2) Storytelling is not an uninterrupted string of words or events, at least good storytelling isn’t. As Aristotle tells us in his Poetics, a story is always a cycle, for every story has the beginning, the middle, and the end. This is, of course, the roughest and the most obvious thing you can tell about a story, but we cannot afford dismissing it as a truism, for this is one of the conventions film editors reckon with. You may find the credits sequence attached to the beginning or (less frequently) to an end of a movie, of you can have a brief prologue preceding them, but I do not recall many cases in which the credits appeared right in the middle of the film. Or, less obviously, but as importantly, your typical (Timoshenko’s “normal”) narrative or drama will likely begin calmly – then something happens – then someone counteracts – complications – impasse – climax – resolution – and, relax, slowdown again. Believe it or not, this narrative up-and-down translates to shot-lengths, and, consequently, to the cutting rates. Sorry for bringing up the ABC of filmmaking, but because one of editing’s functions is to articulate stories visually STORY-FLOW RELATED CUTTING RATES (or SF-factor) matter more than we tend to think.
There are more types of sub-cycles the big story cycle consists of than I can possibly mention in one comment, but one of them is silent-film related and is therefore important to our case. In the silent era, films came in (roughly) 10-15 minute “reels” and if a movie theater was not equipped with two projectors (as all theatres are nowadays) there were breaks every 10 or 15 minutes for the projectionist to load the new reel into the projector. For those technical breaks not to interfere with the flow and the tension of a film filmmakers observed the following rule: make a reel a quasi-independent unit with its own little climax, and end each reel so that each technical break coincides with a narrative pause. (These sub-climaxes is exactly what Timoshenko marked using single circles in his 1926 diagram – each comes shortly before reel-breaks marked on the horizontal axis). To see whether Vertov’s editing counted with this rule we’ll need to submit to cinemetrics each reel separately – in order to see whether or not out trendline curves follow a certain common pattern.
3) There exist stable editing patterns which “normal” filmmakers use in certain scenes. In dialogue scenes this is the reverse angles editing, for instance, there is crosscutting in chase or rescue sequences, or ‘montage sequences” to cover briefly a larger span of time. Each of these comes with its own kinetic characteristics in the trail. This is what I called CONVENTIONAL EDITING PATTERNS (or CE-factor).
Tired? Me too. What I suggest are the following 3 tasks.
Look at the Man with a Movie Camera graph with an eye to the ED-factor. Are there things in this film – slower cutting rates, faster cutting rates – that can be explained away as “event-driven”?
Are there perhaps also spots whose cutting rates are CE-factor dictated?
And we’ll ask Gunars to submit the film reel by reel to figure out if reels mattered to Vertov. If Heidi is ready with the 5-part mapping of the film it is time to provide this too. This way we’ll about the SF-factor significance in Man with a Movie Camera.
Author: Heidi (reappearing) to Yuri and Barbara Date: 2008-07-25
Admiring your detailled discussions on graphs and standard deviation (and only admiring without contributing I have to admit) - but this is my moment to come back into the picture. Also because for further discussions we need more hard data. As Yuri said, we will begin by submitting the reels. That is not too difficult if you have the editing table at hand, it also makes you aware of the handwritten numbers on the film material itself. The film, as we see, is of course already devided into little parts, which represent the filmed sequences, am I right here dear film historians? Also adding these numbers into the big super list I am preparing right now (to rule the world with it afterwards) would be very nice. It definitely helps to find the reels, if they are (which is the case in the filmmuseum in vienna) always in two on one big reel. Anyway, reels - done. I will give you here also the frame numbers in addition to the submission on the graph by Gunars:
Reel 1: Startframe
Reel 2: Startframe 15794)
Reel 3: Startframe 32605)
Reel 4: Startframe 46863)
Reel 5: Startframe 60953)
Reel 6: Startframe 77503)
I will further provide Petrics segmentation, is that what you meant by "5 parts" Yuri? I remember us all sitting under that big tent in Bologna and developing a plan for the next submissions, which is a segmentation on one hand and later on we will think out POV and patterns. More about that later, when the work is done (and there is a lot of work to do).
As far as the SF is concerned: Yes, let's have a look at the reels - to me that is the most interesting thing to begin with and later break the reels into smaller parts, focusing more on the connection of smaller episodes. Let me just add, that I am also looking for reels in the other Vertov movies and I am convinced that Vertovs editing counted with this rule. It actually helps to think this way (narrative units), especially when there are no cue marks. My favourite occupation at the moment
I will go on working and reading. And, by the way, thank you Yuri for bringing up ABC of filmmaking!
Author: B to Y&H and Gunars Date: 2008-07-26
Before we dig into the single reels, is it possible to get some of the charts we have (say … trend line degrees 1, 6, and 12) with marked reel boarders?
I recalculated Heidi's exact frame figures into our 24fps system:
Reel 1: 00:00
Reel 2: 10:58
Reel 3: 22:39
Reel 4: 32:33
Reel 5: 42:20
Reel 6: 53:49
In order to illustrate what I mean, I have applied the reel boarders manually (and therefore at pretty much estimated, maybe even arbitrary positions). This is just so we'll have a picture of the whole film as one unit incl. reel partitions. I think that already with this we can start to see if any of the phases described (and/or how they relate to the "normal" pattern in Timoshenko – which I copied just underneath this "manually" separated chart) make sense in terms of story-flow-related cutting rates (SF). E.g. it is pretty striking how the "barrage of very short shots (as around 30 minutes into the film)" as a climax (ignored by the trendline) and the one at around 40 min. into the film appear just where they are supposed to appear (Timosh., reel/part 3 and 4); in reel 4 and 6 the trend line(s) show incredible similarities (eg. two peaks at pretty much the "expected" position in reel 6!).
3 questions here.
@ Gunars: Is there a possibility – as I already tried to find out in my previous comment - to draw a MWMC-graph with the x-axis based on an "evenly calibrated time flow – like in Timoshenko?
@ Heidi: What do you mean by "The film, as we see, is of course already divided into little parts, which represent the filmed sequences"?
@ Yuri – concerning the factors and our methodical approach - 1) Event-driven (ED), 2) Story-flow related (SF), and 3) Conventional Editing patterns (CE) – just in order to stick to the rules (deduce from the graph first without trying to get back to the evidence of one's own memory of the film or to actually watch it again): I guess one cannot talk about ED and CE, if one has no clue about a certain film, right? As I am a perfect studying object (no memory of what so ever, especially not the order of appearance of sequences related to reels/parts) … I shall try to stick to the graph(s) only so far. You made me believe already that there is some of Vertov in it. This way we'll approach the THING from 3 different preconditioned positions: Y – with the MWMC structure & a lot of curves & a lot of applied film history/theory in this mind/brain/memory; H – with her extensive & detailed knowledge of every little frame of MWMC, after having done the annotation and having redone it; B – the tabula rasa, the naive spectator/observer. This should make a pretty good experimental setting (in terms of AGAP – the "average graph analysing person"). ;) B.
Author: B Date: 2008-07-26
Here is a short overview of ASL of the 6 reels-parts.
reel 1: 3.3
reel 2: 4.5
reel 3: 2.6
reel 4: 1.5
reel 5: 3.0
reel 6: 1.4
Author: Gunars to B Date: 2008-07-28
The quickest way I could think of is using photoshop to squeeze short ASL reels and stretching the longer ones:
Author: Yuri to Barbara & Gunars Date: 2008-07-28
The collation above is instructive and useful.
Let me recap what Barbara suggests and what Gunars has done at her request.
Earlier on in our conversation Barbara has adduced a graph found in an old Soviet book by film director Sergei Timoshenko called Film Art and Film Editing (1926). The graph comes from a final section of Timoshenko’s book called “Rhythm and Punches.”
What is a “punch”?
According to Timoshenko, the “punch” (udarnoe mesto) is a sequence whose impact on the film viewer is intended to be stronger than is the impact of the rest of the film’s scenes.
Speaking of “punches”, Timoshenko has in mind two inter-related things. First, our interest is always higher when something vital is at stake: will the boy save the girl; will the girl kiss the boy – etc. Secondly, the punch-spot must differ from the rest of the spots in terms of rhythm. In action films punches are edited fast, so as to keep the viewer’s attention on edge; conversely, in what Timoshenko’s calls “psychological punches” a punch could be a long-held shot – as that chilling close-up of Conrad Veidt’s face as a waking-up somnambulant in Das Cabinet des Dr Caligari : (6) ASL 11.
Punches, Timoshenko goes on, must be few and they must be carefully placed.
For him, the correct placement of punches is so crucial a point that he even conjured up a punch-diagram of an imagined “normal film” which Barbara has copied and pasted for us to compare it to the cinemetrics diagram for Man with a Movie Camera.
In this graph we see 8 punches, scaled according to their impact power.
Stronger punches are shown as double circles; less strong ones, as single circles. The line that joins the 8 circles represents the non-punch, average, ordinary sequences in the film.
Now, the horizontal axis of Timoshenko’s graph is evenly calibrated 1 to 6. These are the 6 reels his imaginary film consists of. As we recall, as old movies were shown in old film theatres, film projectionists used to make technical breaks after each reel, to load the new reel in the projector.
As we can see, the minor punches (single circles) in Timoshenko’s graph fall in the middle of each reel – mostly, closer to the end. These are reel punches, thanks to which we perceive each reel as a cycle in its own right, with its own beginning, culmination and closure.
The strong, double-circled punches form the impact-structure of the film as a whole.
To bring across Timoshenko’s punch philosophy, let me translate one passage from his book, exactly above the graph.
This is what Timoshenko writes.
“The punches must be used sparingly; there number must be minimized to the most necessary ones: one at the beginning of the film (the one that sends the “look!” signal to the viewer), one at the climax point (the main thing!), and one at the denouement at the end of the film (the end). In addition, less strong, but still salient rhythmical changes can happen in each separate reel of the film, and the impact intensity of those must grow stronger towards the final part of the film. The further the film unfolds, the stronger must be the punched delivered to the viewers’ perception of it.” (p. 69).
What Gunars has done in the above collation of graphs was to scale down Man with a Movie Camera as a whole for its reel breaks (which Barbara highlighted with yellow bars) to match the reel breaks in Timoshenko’s imagined “normal” film.
Now we can see how MWMC (one of the least “normal” films ever made!) stands against the “norm” as Timoshenko conceives it as far as its editing profile is concerned.
Interestingly, if we look at the trendline, it conforms to Timoshenko’s rules pretty well. We must not let ourselves be hypnotized by these similarities, but similarities these are. There is a punch at reel 1, reels 2-3-4 go uphill (as Timoshenko teaches us they should), reel 5 makes a step down to prepare us from two grand punched in reel 6.
Again: talking about similarities is only good as long as we go on from there to talk about distinctions, but this a good point to start, dear Barbara, thank you for bringing old Timoshenko to out table.
Author: Yuri to Heidi: more ABC on prints Date: 2008-07-29
Tomorrow is the day.
There are 3 archival generations of MWMC in existence. Tomorrow, I was told, I will get access to what I suspect to be the oldest, the most authentic and least seen print generation of Man with a Movie Camera. I is preserved at the TV studio archive in Riga.
Let me explain briefly what the 3 print generations differ in.
PRINT GENERATION 1.
These are prints that come down to us in their original storage and exhibition format.
What is “storage and exhibition format” and why do we need to know how this or that motion picture used to be stored and exhibited 80 years ago?
In those days , regardless of their length, films used to be stored in metal cans that contained circa 300 meters of 35 mm wide film each. This is how they were filmed, this is how they were stored, and this is how they were shipped to the exhibitor. Now, because – let me say this again – in those days silent films were frequently shown using a single projector, any movie longer than 10-15 minutes long would have to be shown with a technical break between reels.
So what? Why bother about this nowadays? Why all this fuss about the original format?
As we know from Timoshenko (see quote above) and many other sources, for this break not to work against the story unity filmmakers worked out a strategy to match the story flow to reel breaks. Responsible filmmakers thought about their stories in terms of 300-meter legs. More often than not each reel would start with the title “Act N” and end with the title “End of act N.” Vertov swore, as you recall, not to use words in films, so instead of “Act N” in MWMC reels began with a beautiful silvery numeral, 1 to 6, rising against the velvety black background. And, to end a reel, more often than not Vertov would use an emblematic closing shot – like a close-up of a camera lens with its iris closing.
This is how the original, first-generation print of MWMC was stored and exhibited.
PRINT GENERATION 2.
In later times, with practically all movie theaters equipped with two projectors, reel-to-reel breaks were eliminated, so in those brave new times it seemed there was no need in all those “Act N” and “End of act N” titles. So in those brave new times archivists who preserved silent films and distributors who rereleased them decided they would better get rid of the old-fashioned reel-break markers thinking they only interfered with the smooth story flow of the film. Alas! It never occur to them (as it does to us) that the division into "Acts" was PART of the story flow, that this was how silent films BREATHED in their silent days.
So, it was then that all the reel numbers apart from number 1, numbers 2-6, each of them rising like a silvery moon on the velvety background, were amputated from MWMC – as by those barbaric barbers of doctors who would cut your appendix telling you it you have no use for it anyways.
PRINT GENERATION 3.
The 35 mm film which you know, dear Heidi, the one preserved at the Austrian Film Archive in Vienna is the third generation copy -- the one that comes in 3 reels instead of 6.
As you know most archives nowadays prefer to store films in double-reel cans. So, when they acquire an old 6-reel movie, for instance, the first thing they would do is to open the original smaller cans, pull out the film reels, and, having cut off the unnecessary leaders, splice odd and even reels into pairs: 1+2, 3+4, ... x+y, and then put each pair into a larger can. This is being done because modern projectors can hadnle 600-meter reels and also for the economy of storage space, so the x+y, odd+even reel marriage is nothing else than a marriage of convenience.
All this is something you know only too well, Heidi dear, for you, guided by Frau Schlemmer (God give her many long and happy years to do this for all Vertov’s films) spent hours undoing the harm and divorcing artificially coupled reels.
Why did you need to do so? Because in our mutual work for Cinemetrics we wanted to know about the original, first-generation story flow of MWMC. So – you divorced the unhappy reels, Gunars submitted them to cinemetrics – and, are we happy now? Are reels 2, 3, 4, 5 and 6 perfectly happy now without their strange bedfellows?
No they are not, because THEY HAVE NO NAMES. Their names -- their cute silvery numbers – are missing. And each number has a certain shot lengths, remember!
So the work you have done – good work, much needed work – was merely the UNDOING OF PRINT GENERATION 3 TO PRINT GENERATION 2.
THE 1rst GENERATION RESCUE PLAN. Many years ago I saw that Riga print and as I recall the numbers were in. I called the archive last night and they said I was welcome to see the print.
Keep fingers crossed, Heidi and all. If my memory is correct I’ll find the original reel breaks, count how many frames the reel numbers take to rise and come back to you with new info to insert into your Advil. Then we’ll resubmit. Tomorrow is the day. We may catch the elusive first-generation print.
Author: Date: 2008-07-30
Back from the TV archive editing table.
Found the print preserved there to be the first-generation print for Man with a Movie Camera.
And yes! As I predicted, there are silvery numbers 1 to 6 rising at the beginning of each reel, and not only that. At the end of each reel the same silvery number is falling back – and out of sight.
So, my memory has not failed me this time: something is wrong about reel-by-reel graphs submitted to Cinemetrics earlier on. The reel break between 1 and 2 was established wrongly; we need to double-check the other reel breaks, too.
What I did was to create a reel-to-reel break record which Heidi will now look up in order to double-check the break-down she’d established using the Generation 3 print preserved at the Austrian film museum.
Here is this record.
REEL 1 ENDS with the Venetian-blinds/eye-blinking crosscutting sequence. The last shot of the reel is a camera lens shown in a close up: we see its iris diaphragm closing. Then the silvery number “1” is seen falling (90 frames).
REEL 2 STARTS with the silvery number “2” rising. This is how it looks on the picture I made today on that editing table in Riga I used to inspect the print.
The number “2” takes 54 frames to rise. Then, the same shot of the camera lens in a close up which had closed reel 1 appears again, only now its iris opens, not closes.
REEL 2 ENDS with shots of faces frozen on Svilova’s editing table; the last shot of the reel is on her shelf with rolls of film stored. Then the silvery number “2” is seen falling (60 frames).
REEL 3 STARTS with the silvery number “3” rising (50 frames) followed by: shot of Svilova’s shelf with rolls of film stored and a shots of a frozen face on Svilova’s editing table. REEL 3 ENDS with the shot in which we see the man with his movie camera filming firefighters riding the fire engine. Then the silvery number “3” is seen falling (34 frames).
REEL 4 STARTS WITHOUT THE SILVERY NUMBER “4” RISING. Most probably, the shot with the “4” number is simply missing, for the “4” does appear at the end of the reel to announce its end.
The first shot of reel 4 shows a camera lens in a close up which we see rising from the bottom side of the frame till it looks directly at the viewer.
REEL 4 ENDS with the exact opposite movement of the same lens, now lowering its gaze below the bottom side of the frame. Then the silvery number “4” is seen falling (41 frames).
REEL 5 STARTS with the silvery number “5” rising (40 frames).
Follow the shots of machine-tools slowing down and stopping.
REEL 5 ENDS with shots of carrousel and, as the last shot, the cameraman on a motorcycle approaching the viewer with his camera pointed at us. Then the silvery number “5” is seen falling (51 frames).
REEL 6 STARTS with the silvery number “6” rising (39 frames). Then we see a wave apparently caused by a motorboat from which this shot is filmed, tree in the wave, etc. REEL 6 ends with an eye superimposed on the camera lens (69 frames) and the end title in Ukrainian (Kinec).
Author: Gunars Civjans Date: 2008-08-01
I have added the reel number shots to the reel by reel submission. They are marked with Advanced mode color red, or pale green in reel 4, where we assume the falling number is missing. These additions (taken from Yuri's above record of the Riga "Generation 1" print and added to Heidi's digitized data of the Vienna "Generation 3" print) are too small to change the basic statistics, so Barbara's ASL overview posted on 2008-07-26 remains valid.
Keep in mind: if you wish to see how the graph looks without the Riga numbers, simply uncheck the box under the "number" category -- and you are back to the Vienna print.
Author: Heidi Date: 2008-08-01
I have - using Yuris comments - checked again the print in Vienna and corrected the previous reel breaks.
Here are the final numbers:
Reel 1: Startframe
Reel 2: Startframe 17164
Reel 3: Startframe 32605
Reel 4: Startframe 46863
Reel 5: Startframe 60953
Reel 6: Startframe 78191
I had to correct Reel 2 and 6. Now I hear you ask, why is that so difficult and why is it not possible to see on the print, where the reel breaks are. The answer is: Going for splices in the print and even using cue marks as clues is very deceiving, as we see. Obviously the rising numbers for the acts have been cut out and new cue marks have been set in some cases. Not all the splices in the print are "real", some are copied and therefore difficult to notice as reel breaks. But now we have a final result, which also makes more sense, if we think of SF.
It might be helpful to see the start and end frame of each reel, here we go:
Reel 1: (lense closing)
Reel 2: (first shot lense opening)
Reel 4: (first objective rising, the second going down)
Reel 5: (motion towards the camers from the cameraman on motorcycle)
This is the first segmentation into reels.
Author: Yuri to Barbara abd Heidi Date: 2008-08-02
Let me recap in a few words the few things we have established about the story flow of Man with a Movie Camera this week – something never brought up about this film before.
Now we know that MWMC was originally edited with an eye to being shown in 6 legs with 5 breaks in between, not in one fell swoop as has been earlier assumed;
We know that in order for the viewer not to experience these breaks as so many abrupt and unwitting interruptions Vertov has bracketed the reels with corresponding numbers rising at the beginning of a reel and falling back at its end;
And we know that alongside these kinetic-numeric brackets there appear to be other visual means meant to say to the viewer that each consecutive reel is both a whole in itself and a part of a larger whole. We can tell this by looking at Heidi’s picture-record of the very first shot each reel begins with and of the last shot which closes it.
Are there perhaps fluctuations in the film's cutting tempo that would be matched to the openings and the closures of each reel? To find this out we’ll need to take a look at reel graphs, of course. But before we go there, let us think a little more about the falling and rising numbers and about the 12 shots Heidi has pasted above.
Let me pose 2 questions I want us to think about, one about the numbers, and the other about the pictures.
1) This first one may seem embarrassingly plain, but it is useful to ask for it helps us to realize how many things we tend to take for granted. It did not take Vertov a lot of thinking to do I am sure in order to decide that rising numbers should be spliced to the beginning of each reel, while the falling ones should be used to signal the reel end. Neither do we perceive this as a strange thing to do. Still, it makes sense to ask: why did he do this the way he did? Why not the other way round, for instance: number “3” falls and reel 3 begins; reel 3 ends and number “3” stands up?
2) The other question is about the above pictures, and is related to their function. It would be an easy question to answer if each reel started and ended with the same emblematic shot, like the lens iris opening and closing, for instance. Then, these would be – like those numbers – mere reel dividers. Instead, Vertov makes us part of a consistent, but by no means uniform, game of repetitions, mirror symmetries, recurrent motifs and visual matches. Can we do something to crack this game – to find a trend in the way Vertov selects what he think is the best frame to end and to begin this or that reel?
Author: Heidi to Yuri Date: 2008-08-04
I will comment on three things.
1. I will add some more precise information on start of reel 1.
2. Try to find an answer to Yuris first -
3. - and second question.
1. Start Reel 1:
Important: The frames are taken from the start-shot and end-shot of the reel. It is not the very first one or the very last one. And: Let's not mix start of reel and start of act, which would be marked by the rising and faling numbers.
What I add now is the first shot of reel one AFTER the credits. Reel 1 da capo:
And here is a frame taken from the first shot after NUMBER 1 rising (FIRST ACT) and end of reel:
and finally here is the famous NUMBER 1 rising: (in motion)
Again, what I submitted in my last comment was the actual first frames (CREDITS), now I made it more precise with the frist shot AFTER THE CREDITS and finally the first shot AFTER NUMBER 1 RISING.
Author: Heidi to Yuri and Barbara Date: 2008-08-04
Thoughts on Question Number 1:
"Why did he do this the way he did? Why not the other way round, for instance: number “3” falls and reel 3 begins; reel 3 ends and number “3” stands up?"
Hard to think of an answer to that question, because I am so ready to draw quick conclusions like "For an audience that is the normal way of percieving a changement of acts". Simply because it seems so natural to start a new "sentence", a new argument like that in communication, like rising a flag, like Vertov again in the movie rising the objective and lower it again - see beginning and end of reel 4 in MWMC. It also reminds me of Vertovs famous illustration about his interval-theory, where he speaks about his work to consist of phrases, rises, peaks and falls.
Maybe he experimented a lot with this special issue? Let's look at his film STRIDE, SOVIET! from 1926, also animated to indicate motion:
middle part filling up with "white"
Can there be a better answer? Curious to hear Barbara and Yuri.
Author: Heidi to Barbara and Yuri Date: 2008-08-04
Thoughts on Question Number 2:
"To find a trend in the way Vertov selects what he think is the best frame to end and to begin this or that reel?"
First of all: How sure can we be sure that the shots we have now are the ones which Vertov intended? I might need to go back to his writings and drawings maybe. So every conclusion is fragile, or is it not? Well, let's try and see how far we can build hypothesis and it makes sense.
But where do we begin with our analysis? With the first shot after number one rising? or with the first shot after the credits? I chose the first shot after the credits for the moment, please contradict if necessary!
I will make some first suggestions:
a) All starts and ends include references to filmmaking, be it the cameraman himself, the camers lense, objective or editing room. BUT: not beginning of reel 5 and 6. Here we have a machine detail and the sea. Any ides?
b) I will describe the ends and beginnings of the reels a little, maybe that helps.
Reel 1 ends with blinds and eyes opening and closing, intercut very fast. Then lense closing.
Reel 2 starts with lense opening lense and a long low angle travelling through the streets and ends with Freeze frames in the editing room.
Reel 3 begins with same scenery Freeze frames coming to life and ends with sequence of firefighters and ambulance intercut, we see the cameraman standing in the car with the firefighters, shooting.
Reel 4 begins with objective rising and a low angle travelling and ends with a big crescendo (very fast cutting rate, flicker effect of Kaufman and machines) and objective falling.
Reel 5 begins with machines being switched off and ends with a long intercut of motorcycles and carousel, then the camerman is moving towards the camera, pointing at us with HIS camera.
Reel 6 begins with the sea, trees in the wind and ends in the known way with fast intercut, until the lense is closing while we still can see the eye in a Mutiple Exposure.
Again: Curious to hear Barbara and Yuri!
Author: Yuri to Heidi Date: 2008-08-05
Heidi, thanks for your clear comments and extra pictures; I feel quite prepared to comment on both, but let's wait for Barbara to join in; while Barbara is thinking -- let me do one thing: update the 12-still reel-break record you've submitted on 2008-08-01 according to your useful suggestion: omit the credits and intro titles and look at images only. Then the whole thing will look like this:
Reel 1 begins with the Man with/on camera and ends with the iris diaphragm closing:
Reel 2 begins with the same iris diaphragm opening and ends with film rolls stored on the editing table:
Reel 3 begins with rolls of films stored on the shelf in the editing room and ends with the cameraman (this frame is too dark to see him clearly) filming firefighters.
Reel 4 begins with a camera lens rising from the bottom side of the frame and ends with the same done in the opposite direction.
Reel 5 begins with machine-tools slowing down and stopping and ends with the cameraman on a motorcycle approaching.
Reel 6 begins with a wake-wave from a motorboat and end with a widening eye superimposed on a lens whose diaphragm is closing:
Author: Yuri on Question One Date: 2008-08-05
OK, let us not wait for Barbara's input on Question One this time. Here is the question, your reply to it and my reply to yours.
YURI asked: Why did Vertov decide that rising numbers should be spliced to the beginning of each reel, while the falling ones should be used to signal the reel end? Why did Vertov do this the way he did? Why not the other way round, for instance: number “3” falls and reel 3 begins; reel 3 ends and number “3” stands up?
HEIDI replied: For an audience that is the normal way of perceiving a change of acts. Simply because it seems so natural to start a new "sentence", a new argument like that in communication, like rising a flag, like Vertov again in the movie rising the objective and lower it again - see beginning and end of reel 4 in MWMC or the black reel numbers that become white in his film STRIDE, SOVIET! from 1926.
YURI: I fully agree with the point that Vertov did it this way because he knew this was more “normal” and “natural” (your words) to announce the beginning of a reel by making its number rise, and its end by making it fall. I also support the parallel you have drawn between the rising and falling reel numbers on one hand, and the camera lens rising its eye at the beginning of reel 4 and lowering it before the reel ends on the other. And I jumped up of joy from my chair when I saw your example from STRIDE, SOVIET! in which reel numbers that open reels are filled with white. You are right: in all those cases we can say – it looks more “natural” to announce something that begins with rising, filling or brightening and something that ends the other way round.
But is it ALL we can say? I do not think we should rest our case there. Because it is exactly here that things become really interesting and relevant for Vertov’s film. If we agree that this was a natural thing to do, the next natural question is – why does it seem so natural to you, to me, to Vertov?
Don’t shrug your shoulders, Heidi. This question is not as academic as it may sound. The same can be asked about MAN WITH A MOVIE CAMERA as a whole, or about the genre of documentary “city symphonies” in general. Why does MWMC begin with the shots of city sleeping, then of a young lady waking up -- and why has Vertov chosen to take us through the city’s daily cycle? Why does BERLIN, THE SYMPHONY OF THE GREAT CITY show us this town’s life from dawn to dusk?
One way of explaining the “natural” is to turn to nature. There are some things all human beings share. Two of them are relevant to our subjects. We, human, are diurnal animals and we are bipedal. “Diurnal” means that we all more or less wake up in the morning and go to bed after sunset. So does the sun, by the way. “Bipedal” means that when we go to bed we become horizontal, and we become vertical (“get up,” “stehen auf,” “rise”) when we begin our day. When we are babies, we crawl, when we are corpses, we fall. See where I am heading?
Are there cognitive scientists or Gestalt-psychologists already that have proven experimentally that human mind more readily associates rising and brightening objects with the beginning, and falling and darkening with closure than the other way round? Hello there!
If not, they must; but meanwhile, let us consider Vertov’s films enough of a proof. Vertov is human, so his number “1” in MAN WITH A MOVIE CAMERA is rising like a bright numeric sun against the pitch-black darkness; and this is also why his number “I” is STRIDE, SOVIET! grows white, not black. If Vertov were a bat I bet he would do it the other way round.
One last remark on Question One.
Count how many films you have seen that start with an alarm-clock going off. A rather standard beginning for your run-of-the-mill film, isn’t it? Let me tell you more: philologists tell us that “morning” and “evening” is one of the most ancient and most common ways to start and end a poem, a story, a drama – something they call a "topos”. Latin poets used this a lot, and even “Book 2” of Homer’s Odyssey starts with this passage: “Now when the child of morning, rosy-fingered Dawn, appeared, Telemachus rose and dressed himself.” And do you remember the last line? Of course: “Thus, then, the ship sped on her way through the watches of the night from dark till dawn.”
And you say: Vertov.
Blow out your candle, Heidi, dear, and when you get up tomorrow read Robert Browning’s drama in verse Pippa Passes – what are Pippa’s opening and closing monologs addressed to? And after you close the book find and see D.W. Griffith’s 1909 screen version of Pippa Passes – the first film I am aware of in film history to use the dawn-to-dusk story structure and lighting schemes .
Author: Yuri to Heidi Date: 2008-08-08
We now move on to Question 2: what do 12 reel-dividing images (posted above) tell us about Vertov's handling of the story flow of the film?
HEIDI: How sure can we be that the shots we have now are the ones which Vertov intended? I might need to go back to his writings and drawings maybe. So every conclusion is fragile, or is it not?
YURI: It is and it is not. It may be as fragile as every film’s celluloid base is, for one cannot rely on a single print, particularly if this print is a distribution positive. These are as a rule quite worn and torn. On the other hand, our conclusions need not be fragile at all – not if there are several versions of the film with only a slight difference between them – as is the case with MWMC. The definitive version of a film is usually considered the one about which it is known that the filmmaker who made it was present at its showing. We are lucky – for we know that the Dutch print of MWMC was shown at the Film Liga when Vertov was in Holland and was present at the showing. And as I recall it does not differ much from the Vienna/Gosfilmofond print – with the exception of the baby-birth scene which the Dutch censors considered too graphic.
HEIDI: But where do we begin with our analysis? With the first shot after number one rising? Or with the first shot after the credits? I chose the first shot after the credits for the moment, please contradict if necessary!
YURI: I agree with your choice. Let us assume we have spoken enough about the numbers rising/falling, have established their connection with cognitive preferences of the human brain as well as the fact that they chime in with the day-cycle story format which Vertov has chosen for his film as a whole; we also observed, we recall, that the latter, in its turn, chimes in with the time-honored topos in literature (Homer, Browning) and film (Griffith, Ruttmann). So, we now move on to the inner “reel brackets,” the ones whose still we can see on the frame-record in the comment of 2008-08-05.
HEIDI: All starts and ends include references to filmmaking, be it the cameraman himself, the camera lens, or editing room. BUT: not the beginning of reel 5 and 6. Here we have a machine detail and the sea. Any ideas?
YURI: Only a general one, which I import from poetry studies (not irrelevant here – for Vertov did write poetry and, to think of them, the end/beginning frames we are discussing do look like “visual rhymes,” don’t they?). I am speaking about the rule of intermittent consistency. What does this rule say about verses? Metric verses are typically consistently regular – they have regular stresses separated by regular numbers of syllables, and, in addition, their lines often end with rhymed words.
Now, those people in poetry studies who, like us at Cinemetrics, are not averse to count have established that poets avoid being consistent throughout -- for, paradoxically, if a poem is fully consistent with a chosen pattern of regularities, it becomes predictable, and its very consistency loses its impact. To avoid this, poets break their own rules from time to time misplacing an accent of adding an extra syllable, for instance. My theory is that the shots that begin reels 5 and 6 are such consistency breakers.
Now – look again at the consistency pattern in the 12 shots which sandwich the 6 reels. They are all about consistency forming and un-forming. 10 shots out of 12, as you rightly observed, are thematically linked to filmmaking, 8 to filming, 2 to editing. We can also say that in all the 8 “filming” shots we see a camera lens looking in our direction. But, importantly, none of the 8 filming shots is an exact repetition of any other one. They vary is terms of shot scales and in terms of movement.
Let us take a quick look at some of Vertov’s “rhymes.”
Reel 1, shot A rhymes with reel 1, shot B as lens to lens, with the shot scale being different.
2A is the same shot as 1B with the iris blades moving in opposite directions. The subtle joke these two shots play on us is that these are the proverbial “iris-in” and “iris-out,” only glimpsed from the other side of the fence.
5A and 5B turn the camera lens into a semblance of the sun rising from the horizon in A and setting in B. Intended, perhaps, to repeat, in a nutshell, the day-cycle structure of the film as a whole .
6B combines two opposite movements: as the eye widens, the iris narrows. The climax and closure in the same shot.
Author: Yuri to Barbara abd Heidi Date: 2008-08-09
We now turn to reel graphs. Here they are, the 6 reels of MWMC, all shown at trendline Degree 1. Red-cloured bars marking the end and the beginning of reels stand for falling and rising reel numbers found in the Riga print.
Author: Yuri to Barbara and Heidi Date: 2008-08-09
Let us not be distracted by the fact that the graphs do not look uniform in scale. The program is set up this way. As Gunars explained in one of the topic at the Discussion board, CineMetrics counts the "length of the film" in terms of the number of shots this film consists of rather than of evenly calibrated time flow, hense different x-axis length. In addition the reels vary slightly in terms of footage.
As we look at the 6 trendlins alone, is there a tendency we can speak of?
Author: Heidi to Yuri Date: 2008-08-13
Six trendlines, all of them "faling", but one - Reel 2.
Let's have a look, what is happening in the movie in reel 2: The cameramen is travelling through the streets (low anlge), the city is waking up, machines are set to work, traffic starts, people go to the market, the cameraman is shooting from a "camera-car", following another carriage and it ends with freeze frames in the editing room. What we have here is long shots, episodes, a "narrative structure" to descibe city life, introducing places and people. And we could presume that the audience percieved this part as a quite "normal" film, without being interrupted a lot by Vertov playing his fast-cutting-game.
If I now have a closer look and estimate the downtilt of the graphs, hopefully getting on the right path, we get (in a rough estimation) the following numbers:
Reel 1: 4 to 2,5 (Diff 1,5)
Reel 2: 4 to 5 (Diff 1)
Reel 3: 4 to 1,5 (Diff 2,5)
Reel 4: 3 to 0 (Diff 3)
Reel 5: 4 to 2 (Diff 2)
Reel 6: 2 to 0,5 (Diff 1,5)
I repeat the pattern: 1,5 : 1 : 2,5 : 3: 2: 1,5
Reel 4 has the biggest difference: 3. And if I guess right, Yuri will has some analogies with greek drama, right? Let me consult Wikipedia first, where I find: Prologue, Paraodos, Stasimon, Epeisodion, Exodos.
To be continued.
Author: Yuri to Heidi Date: 2008-08-13
This is what I found worth thinking about, too: 5 trendlines out of 6 have a "positive slope" (the official term used in statistics for lines that go up left to right) and only one has a negative one. The only thing I want to correct, Heidi, is an obvious slip of the tongue on your part: you said the prevailing 5 were "falling" (or had "downtilts") whereas one, according to you, was climbing. In reality, the other way round -- an Austrian mountain girl like you should know better about slopes!
And I agree: once we have established that there is a 83.3% pronounced acceleration trend (5 reels accelerate, 1 reel slows down) the next question to ask is: what is it that could have caused the exception? Exceptions do often tell you more about rules than the rules themselves might. I believe it was Frances Bacon (correct me if I am wrong) who advocated the value of "absence" for science: if you want to study the qualities present in the air, start with studying the vacuum. Put a mouse in a retort, pump out all air and see if the mouse will live. Induction method.
So -- you are right when you say: "Let's have a look, what is happening in the movie in reel 2." But what is it? What weighs the trendline down at the end of reel 2? You suggest that here perhaps Vertov's editing is more EVENT-DRIVEN (the ED-factor, remember?) than it is elsewhere in the film -- for indeed, for the most part of MWMC Vertov makes the life follow the tempo of cutting rather than letting the cutting follow the tempo of life.
You are right: reel 2 is somewhat different in this sense. But this explanation only accounts for the fact that the cutting rate of reel 2 lingers somewhere between 4 and 5 seconds (which is slower than the overall ASL), but does not explain why the trend looks down instead of looking up as the rest of them do...
Let's have another look, OK? What happens at the end of reel 2 that makes it a maverick among the rest of Vertov's optimistic reels?
Author: Heidi to Yuri Date: 2008-08-18
Tyrolean mountain girls obviously lose their senses concerning slopes when they are living in the "shire" for too long. Could also be that I was too impatient because I had to retype everything after my whole comment being lost somehow when posting it. You told me before it could happen.
YURI: But this explanation only accounts for the fact that the cutting rate of reel 2 lingers somewhere between 4 and 5 seconds (which is slower than the overall ASL), but does not explain why the trend looks down instead of looking up as the rest of them do...
Time to remember (and look up) what you explained about the trendlines in cinemetrics in an earlier comment. And then lets have this another look as Yuri suggested. I changed the Degree of the trendline to its upper limit – 12, in order to make the line more sensitive.
Compare it for example to Reel 4 (where I said the positive slope to be the steepest), also with degree set to 12:
Author: Heidi - continuing comment Date: 2008-08-18
The trendline in Reel 2 never goes below zero and in the second half not even below 4. So from the second half on the shots must be quite similarly long and not interrupted too much by shorter ones. Here the camerman is standing on a car, cranking his camera and following another carriage. Intercut with railway images, wheels and trains. And at the very end the frozen images, literally freezing time in the movie, being a connection to the editing room. Altogether we have 9 shots with various images, people and streets, apart from the first image (horse running) all around 100 frames long. We can also see that in the graph.
Forseeing your next question (what about the short cuts at the beginning?): Not so easy to say with absolute certainity, but it should be the machines starting to work and the cameraman climbing the chimney. And because I like climbing and it also makes the discussion more visual, I post this sequence in screenshots. Enjoy:
Author: Yuri to Heidi Date: 2008-08-20
I have something to add to the following line:
HEIDI: At the very end some frozen images, literally freezing time in the movie, form a transition to the editing room sequence. Altogether we have 9 shots with various frozen images, people and streets, each around 100 frames long.
YURI: Good point, Heidi, this is what I think, too: it must be that sudden stillness at the end that makes Reel 2 unique among the film’s reels. And you were also right in pointing out that the merry and lively labor shots at the beginning of the reel must have been responsible for the faster cutting rate at the beginning of the reel.
Don’t you think, however, that now as you and I are of the same mind regarding these two basic facts about Reel 2 you and I can, hand in hand, make a brave step into the unknown and try to formulate a tentative rule that might more or less apply to editing in general?
Let’s try to.
I hope you won’t disagree, dear Heidi, if I say that the observation you just made about the labor shots being cut faster and the frozen shots being cut slower fits pretty well within the EVENT-DRIVEN CUTTING RATES (or ED-factor) of the three categories of storytelling conventions I have suggested earlier on in my comment posted on 2008-07-24.
- if we look at the amount of within-the-shots agitation in the section showing people and machines in full-speed action and then connect this to the between-shots agitation of cutting;
- if we look at the lack of any within-the-shots agitation in the section showing how these activities come to a sudden halt and then connect this to the slow-down of cutting;
- if we recall that Vertov’s own music score for MWMC instructs that all music stops as soon as the frozen shots appear on the screen in order to create the impression of all temporality suddenly gone (see in: Yuri Tsivian, "Dziga Vertov's Frozen Music: Cue Sheets and a Music Scenario for The Man with the Movie Camera," Griffithiana, No 54, October 1995, pp. 92-121);
- then we are reasonably entitled to formulate what can be called the ED-RULE that film editing tends to obey.
Here it is:
When we observe a direct correlation between the tempo of movement within shots and the frequency of cuts between them chances are we are dealing with the ED-RULE of cinemetrics at work. In its more general form, the ED-RULE links cutting rates to the dynamic characteristics of the event this or that sequence relates to. If we look at the upwards thrust of the 1-st degree trendline for MWMC as a whole, for instance, we can say that, among other factors, this acceleration is explained by the fact that in the beginning, the city sleeps, and is very much awake in the end.
The ED-RULE is only one of a number of rules responsible for fluctuations of cutting rates within the duration of a film; in addition, as any rule it allows for exceptions. But it is important to keep in mind when we analyze editing.
N'est-ce pas, Heidi, dear?
Author: Heidi to Yuri Date: 2008-08-22
Very intersting Yuri, I took some time to read and think and try to really understand what you mean. And I was looking around for people to discuss it with. So here is a contribution of Matteo - remember our guide in Bologna to find a real italian restaurant? - who is not only a gourmet, but also a very spirited film scholar. I showed him your ED-rule and this is what he said:
MATTEO: This ED-rule is very linked to the a narrative concept of film very familiar to western film (and here the word WESTERN comes not by chance, since is very easy to think about the dilatation of time before the gunshot in Leone´s movies), starting with early american movies, in particular of cuorse with the parallel editing of Griffith (also a guy very keen in western genre). It is definitely connected to a very precise concept of time in film, which could be defined linear. What is interesting is that the debate about the concept of time in cinema was already very advanced at that time, and I keep thinking to Jean Epstein, Jean Vigo and to the extremists of the french Impressionism, and their obsession for water. And here it comes the point in Vertov, that, beyond any possible doubt was against certain capitalistic and commercial perverted practices in the cinema of that time. Could be that in MWMC he uses the ED rule (if is possible to say that it can be used), but moreover he can show how it is possible to mock it, to get rid of it, to be free of certain "formalized" standards of narrative cinema. This could be the reason of the series of freezed scenes, depicting fast action movement. Maybe he wants to show that the rule is reversible, or better that a it is possible to define dynamics through editing?
HEIDI: And of course to connect to the editing process and filmmaking, which was pointed out very often. And I remember somebody pointing out (unfortunately i don't recall who) that images frozen on the screen also meant danger, the film stuck in the projector, being exposed to the lamp and therefore easily set on fire. So stopping the film in this way meant not only surprise for the audience and more attention, but a potentially dangerous situation.
How often do these frozen frames appear anyway? - and I mean real frozen one, because Vertov also uses unmoved objects, like showing the film strip and in my opinion this has the same function. This happens at the end of reel 2, as we already said, and later on in two occasions and interestingly in sports sequences. We have the hurdlers and the horse racing. Some images to stimulate your memory and keep in mind that before and after the frozen frames we have already slow motion:
(Somewhere in the middle of Reel 5)
So the sport scenes are very much connected with slow motion, maybe to focus on the movement, make a "motion study" like Muybridge did or even Vertov himself with his often-quoted jump from the grotto, in order to capture not only motions but also e-motions. No coincidence maybe that there is always a horse involved. Why is that by the way and why not a cat? Because we consider a horse also more sports equipment than an animal? Or because his movement is considered so elegant? Or it is simply a trained animal and would run on demand and also stop on demand?
But let's go back to the RULE. I think I understood and it makes sense. My question is: Is there a way of visualizing it? One suggestion would be to use two diagrams and overlap them. Let's say we have the usual cinemetrics graph (shots and shotlength) and another one with a very rough measurement of object motion within the shots. And yes, I did that in Anvil, but more focusing on the DIRECTION, not intensity of the object motion within the shot. If we had for example categories like "slow, average and fast" (rough I said) then I could try to annotate every shot in that way. And we would be able to prove the rule. Am I thinking correctly? This could be a task for later.
Before we make a step further to the other rules, I mutter into my beard: Still I am a bit confused with the distinction of events that are slow per se (like a funeral) and the slow cutting VERSUS actions that are fast (like horse racing, sports), making it slow with filmic tecniques, and slow cutting. Vertov uses both. Please help me out here whoever, Matteo? Yuri?
MATTEO: Well, if I understood correctly the essence of the rule (but this is definitely something that Yuri could make clearer), the dynamics of the actions are, somehow, orienting the shot rate of a determinate sequence: let´s make an example and say that a fight scene in Spiderman 2 has an average amount of 50 shots per minute (quite high), while the romantic interlude between the hero and his kitty has only an average amount of 15 SPM (if a sequence like this could possibly last more than 30 seconds in a film like that). But, my point is that, although this is undeniably true for all (or almost) the main stream/narrative movies, this rule becomes "playable", a sort of toy in the hands of some film-makers, in particular the master cat of formalism. In my modest opinion Vertov literally plays with time and action, decomposing and analyzing (to be a bit greek) the essence of the action, just like Muybridge and Marey did with their scientific experiments. Just I would say that Vertov is more a political anatomist. He is doing this thing with great irony, and also with the intent of revolting (maybe revolutioning?) the common narrative practice of cinema. Is there a better example for showing a rule than giving to the audience its mechanics?
Author: Yuri to Heidi and Matteo Date: 2008-08-27
Thanks for bringing in Matteo, Heidi. We can cast him as a body double for Barbara whose doctor, as she wrote to me, advised her to stay away from any contacts with Cinemetrics until her baby is born – just in case. Welcome aboard, Matteo!
Before I respond to you and Heidi, let me explain to you in a few words what it is she and I are busy doing here – what is our epistemological mission, as it were.
What we are doing may at first look like having a discussion, but in fact it is not a discussion in the usual sense of the word. Discussions is a time-honored tradition in philosophy (remember Plato, Matteo, and Socrates, and the symposia they had in good old times?) which in more modern times has taken the form of blogging. "To discuss" is the prevailing mode of dealing with things in humanities. In classrooms and at round tables we discuss films, we interpret Vertov, we form opinions, we exchange them and we listen (or not) to others. In the discussion mode, we don’t really need any other tools than language and memory. Discussion means knowledge through others, knowledge by communication.
This is not what we do here on Cinemetrics. Discussing films is not the only way of dealing with them. Another way of gaining knowledge about this or that object is to communicate with the object itself, by direct observation, by measuring and testing. This is what is usually understood by "studying." It is true that in this sense the word “study” is more often used in exact sciences than it is in humanities, but if you decide to stay here for a while you will soon discover that even in films studies some measuring and testing cannot do much harm. I know Heidi shares this view – as you may know she has a degree in chemical engineering and has worked in a chemical lab. If you feel comfortable in an environment like this you’ll easily fit in.
So, what do we do here other than measure shots? We talk films, but we don't discuss them. When people discuss films they operate with opinions; when they study them they operate with hypotheses. This is what we do on Cinemetrics.
Let me now step back and bring you up to date as to how far Heidi and I have got so far in our hypothesis-building.
In the comments I posted in July-August I proposed 2 hypotheses about film editing. These hypotheses are based on what I managed to observe looking at the Cinemetrics data for more than 2 years of its existence.
Let us call these observations “premises.”
PREMISE 1. It is an established fact that at certain points each given film is cut faster or slower than its average shot length (ASL) figure indicates.
PREMISE 2. There is little doubt that these are not random fluctuations. That the film’s cutting becomes faster at one point or slower at another must be caused by certain factors.
QUESTION: What are these factors?
HYPOTHESIS 1: As I suggested in my comment of 2008-07-24 there must be three kinds of factors at work:
1) Event-driven (ED) factors.
2) Story-flow (SF) related factors.
3) Conventional editing (CE) related factors.
OK, let’s move on.
In my comments as of 2008-08-20 I proposed to take a closer look at the first group of the three kinds of factors mentioned above. What I did was to form a specific, ED-factor related hypothesis. This second hypothesis is based on the following premises.
PREMISE 1: Because all life as we know it manifests itself in form of movement, every observable event can be mapped onto a grid of dynamic properties. On a map like this running will not overlap with sleeping; likewise, funeral corteges on this map will occupy a position much different from stock exchange activities. Etc. Any objections?
PREMISE 2: As the medium of cinema (as distinct, for instance, from the arrested medium of painting) renders movement by means of movement, it stands to reason that some dynamic properties that go with certain events correlate with what Eisenstein has termed the dynamics of film form, in our case, cutting rates.
HYPOTHESIS 2: There must be a rule, let’s call it the “ED-rule of editing,” in compliance with which editors tend to cut fast shots faster.
As all rules, particularly those found in art, the ED-rule is not binding; if it indeed exists, it exists as a norm one can either observe or violate. Violations never vitiate the norm, they reinforce it.
HEIDI (interrupts): Still I am a bit confused with the distinction of events that are slow per se (like a funeral) and the slow cutting VERSUS actions that are fast (like horse racing, sports), making it slow with filmic techniques, and slow cutting.
YURI: In Vertov’s case, this does not seem too hard to sort out. We can agree to disregard the “subject matter” of the shot (be it a horse race or a track-and-field race) and go by whether there is some real, physical movement in the shot.
But your question reveals a certain weakness in the ED-rule hypothesis that deserves more thought than I have given it. We cannot always go by the speed of the actual shot.
Take two examples. Imagine a landscape sequence introducing a Belgian village in which the film’s action is going to take place. It’s full of windmills whose wings turn merrily. Still, this will hardly impact the cutting rates, will it?
Example two. As we glimpse from some Cinemetrics submissions, classical showdown scenes in Western films (like Bud Boetticher’s Ride Lonesome: (6) ASL 7.8) tend to have higher cutting rates; but there is not much movement in these scenes other than the shot when the two Western heroes draw their guns; before this happens they just look at each other intensely, and this intensity is reinforces by faster reverse angles cutting (remember Sergio Leone?). But again: here we may be dealing with the CE (Conventional editing) factor.
The main thing, Heidi, is to go case by case, rather than trying to find a formula that fits all cases. We are talking rules, not laws. So, to test the ED-rule hypothesis it may make sense to start with asking: is there really a correlation of sorts between the speed of movement within the shot and the frequency of cut?
HEIDI: My question is: Is there a way of visualizing it? One suggestion would be to use two diagrams and overlap them. Let's say we have the usual cinemetrics graph (shots and shot length) and another one with a very rough measurement of object motion within the shots. And yes, I did that in Anvil, but more focusing on the DIRECTION, not intensity of the object motion within the shot. If we had for example categories like "slow, average and fast" then I could try to annotate every shot in that way. And we would be able to prove the rule. Am I thinking correctly?
YURI: Correctly? You must be kidding. A brilliant thought. All the more so because if you rate internal shot speed as "slow, average and fast" (don’t forget “still” and “frozen”!) we can talk of STATISTICAL CORRELATION (in statistics, correlation only works for meaningful quantifiable data, and your rating scales provide a good ground for such a quantification). If you do this in Anvil, Heidi, and if it turns out that the ED-factor is statistically significant, it will become known as the “Heftberger correlation,” I promise.
MATTEO: This ED-rule is linked to the narrative concept of film, and is very familiar to western film (and here the word WESTERN comes not by chance, since is very easy to think about the dilatation of time before the gunshot in Leone´s movies), starting with early American movies, in particular of course with the parallel editing of Griffith. It is definitely connected to a very precise concept of time in film, which could be defined as linear. And here it comes the point in Vertov, who beyond any possible doubt was against certain capitalistic and commercial perverted practices in the cinema of that time. Could be that in MWMC he uses the ED rule (if is possible to show that it can be used), but moreover he can show how it is possible to mock it, to get rid of it, to be free of certain "formalized" standards of narrative cinema.
YURI: It may be a little too early to ask this question, Matteo, for we don’t even know if the ED-rule exists and if so then to what extent it applies to Vertov. However, if I change my hat of a cinemetrician to that of a Vertov expert, I think I can add something to what you said. You are perfectly right in pointing out that as a radical left-wing Soviet filmmaker Vertov declared that all Western cinema was the opium for the people and only his Kino-Eye was capable of telling hte truth. This is what we learn from all his writings and manifestos. It seems immensely plausible to go on from there and to say that, by denying the cinema of the West, Vertov also denied its stylistic and technical achievements, but it is exactly here, Matteo, that we must watch our step. If you open Vertov’s very first manifestos you will find out that, much like Kuleshov and other young Soviet filmmakers of his time, in his philippics against bourgeois cinema Vertov made one caveat: “American montage.” This is the achievement, he wrote (praising Griffith, by the way), which we the Soviets must appropriate and perfect. He never said anything against fast cutting – he admired it and strived to out-Griffith Griffith in his Kino-Pravdas.
On the other hand, Matteo, you are right in saying that Vertov’s editing exercises are more self-reflexive that Griffith’s, that he plays and experiments with cutting rather than uses it as a means to an end. MWMC is all about self-reflexivity. I fully agree.
But – I promised myself never to lapse into a seminar-like “discussion mode” when I work in the Cinemetrics environment, so I shut up. Here, we do not interpret Vertov – we interpret data about Vertov. A gigantic difference!
Author: Heidi to Yuri and Matteo Date: 2008-09-05
YURI: The main thing, Heidi, is to go case by case, rather than trying to find a formula that fits all cases. We are talking rules, not laws. So, to test the ED-rule hypothesis it may make sense to start with asking: is there really a correlation of sorts between the speed of movement within the shot and the frequency of cut?
HEIDI: I fully agree and I will do that. As Cinemetrics can add 8 different categories (corresponding to 8 different colours), here are the ones worked out in the meantime:
1. No natural movement
2. Frozen images
3. Slow Motion (also slowing down and stopping)
4. Slow (naturally)
7. Accelerated Motion
8. Irrelevant (Titles, One-Frame-Shots, Black Frames)
I am curious to see, whether there will be a correlation or not, but first let me sit down and do it. As Yuri pointed out, how human beings perceive movement (fast or slow) should be quite commonly agreed on. Maybe you want to add something here, while I try to find time for my task?
Author: Yuri to Heidi Date: 2008-09-05
Excellent plan. And, in addition to the colors, we'll then have numeric data for each of the 8 categories which will allow us to establish if there is a correlation between the in-shot speed and the ASL, and if so, the size of the correlation.
A problem you might encounter is what to do with mobile framing -- camera movements and pans. I think this should also be subsumed under "movement" -- even if the lanscape the camera pans across is still. That the camera ride with the lens looking up at high-rise tops at the beginning of Reel 2 -- it's clearly slow; conversely, the pan-stop-pan shot at the beginning of Reel 6 (from roof to the film thater marquee to trees) is deliberately fast.
Author: Heidi to Yuri and Gunars Date: 2008-09-24
I am finished with my first attempt to "measure" motion shot-by-shot and annotate the categories we agreed on. Gunars gets the list this minute and now I am curious if something statistically relevant will come out of it.
Let me just add a few words how I proceeded: I did it completely from my own perception, sometimes asking people around to give a statement and the funny thing is: mostly we chose the same motion-type. So much for a kind of "universal human" perception of speed, at least in the western world in 2008. Interesting also to watch myself in the process, seeing that I developed - after some time - my own schedule, measuring the motions also WITHIN the film, remembering similar shots before and trying to have an overall homogeneus level. I had no problems with fast camera movements and pans, they did not distract me, but added naturally to the annotation. A little bit deceiving was rational thinking, not so much with human movement, but if for example a machine might usually have another "normal" speed or not and then judging the motion in the film now, which means including information of a world outside the film. I tried to do it only looking at the images.
So let's bring it on and analyse a bit more!
Author: Heidi to everybody Date: 2008-10-17
From here on our talk about possible correlations between shots length and motion within the shot (aka ED-rule-hypothesis) continues under:
(or go to the Cinemetrics Database and choose: Man with A Movie Camera - Motion Type)
Author: Date: 2008-12-17
HEIDI: Back again from working on „MWMC Motion Types“ and as we somehow established the ED-rule I would say, let’s go back to the two other Storytelling conventions and see how far we get. Refreshing our memory:
YURI: 2) Storytelling is not an uninterrupted string of words or events, at least good storytelling isn’t. As Aristotle tells us in his Poetics, a story is always a cycle, for every story has the beginning, the middle, and the end. This is, of course, the roughest and the most obvious thing you can tell about a story, but we cannot afford dismissing it as a truism, for this is one of the conventions film editors reckon with. You may find the credits sequence attached to the beginning or (less frequently) to an end of a movie, of you can have a brief prologue preceding them, but I do not recall many cases in which the credits appeared right in the middle of the film. Or, less obviously, but as importantly, your typical (Timoshenko’s “normal”) narrative or drama will likely begin calmly – then something happens – then someone counteracts – complications – impasse – climax – resolution – and, relax, slowdown again. Believe it or not, this narrative up-and-down translates to shot-lengths, and, consequently, to the cutting rates. Sorry for bringing up the ABC of filmmaking, but because one of editing’s functions is to articulate stories visually STORY-FLOW RELATED CUTTING RATES (or SF-factor) matter more than we tend to think.
There are more types of sub-cycles the big story cycle consists of than I can possibly mention in one comment, but one of them is silent-film related and is therefore important to our case. In the silent era, films came in (roughly) 10-15 minute “reels” and if a movie theater was not equipped with two projectors (as all theatres are nowadays) there were breaks every 10 or 15 minutes for the projectionist to load the new reel into the projector. For those technical breaks not to interfere with the flow and the tension of a film filmmakers observed the following rule: make a reel a quasi-independent unit with its own little climax, and end each reel so that each technical break coincides with a narrative pause. (These sub-climaxes is exactly what Timoshenko marked using single circles in his 1926 diagram – each comes shortly before reel-breaks marked on the horizontal axis). To see whether Vertov’s editing counted with this rule we’ll need to submit to cinemetrics each reel separately – in order to see whether or not out trendline curves follow a certain common pattern.
HEIDI: Ok, we divided the movie into six reels and submitted them. We also have the correct reel beginnings and endings and we started discussing something like a poetic structure. So let’s go back to that and see if we can squeeze out more of our graphs. As the work of Lev Manovich demonstrates, visualization is the first step, it is absolutely necessary to get on with an INTERPRETATION.
YURI: 3) There exist stable editing patterns which “normal” filmmakers use in certain scenes. In dialogue scenes this is the reverse angles editing, for instance, there is crosscutting in chase or rescue sequences, or ‘montage sequences” to cover briefly a larger span of time. Each of these comes with its own kinetic characteristics in the trail. This is what I called CONVENTIONAL EDITING PATTERNS (or CE-factor).
HEIDI: As we saw in our work with the Motion Types, this category is very much interconnected with the ED-factor:
YURI: As you may recall, Heidi, I suspected from the outset that we would have problems with the first of our categories, the Non-Motion shots. The thing is, non-motions shots are not only about no motion. They take part in too many different games, serve too many masters, and are used for too many purposes to be able to tell us much about the speed-driven correlation. Take, for instance, the shot-reverse-shot editing pattern, like when a young lady practices sharp-shooting. We see her, aiming, and then the target, her, the target, her, the target etc -- and then she pulls the trigger and the target is set in motion. This is a fast-cut sequence, it builds up tension, yet technically it largely consists of non-motion shots (much like when two motionless cowboys are having a showdown in a Sergio Leone Western). There is more than one case like this in MWMC, as the one with the mannequins "looking" at empty streets. The looking-looked-at pattern, in order for it to cohere, must be a fast one, and it is, yet technically it consists of still shots. What we have here is an effect of interference: the Conventional Editing Factor interferes with the Event-Driven Factor and the purity of our statistical data is therefore compromised.
HEIDI: Can we discuss this a bit more? Extracting “Conventional Editing Patterns”, starting with this one?
So, Yuri, are you in?
Author: Heidi to Yuri Date: 2008-12-17
There is something else I wanted to add: If Vertov himself speaks of "frazy", being a smaller part of "chasty" (reels), where would we find them here? Will we be able to see the phrases and even more the peaks in the phrases, if we analyse reel by reel? How far can we abstract questions like these from the "narrative", finding out these things by looking at the shot-length only? I tend to think we can also go by ASL for a while, see where it leads us to. And Yuri, what I have in mind is of course the famous diagram by Vertov that you explained last january at the conference. We have been working on Odinnacatyj so far, trying to extract phrases, but only by looking at the visual content, which makes the concept very vague and hard to defend, because questions come up immediately, like "why do you draw the boarder here and not elsewhere"? So I would like to see if there exist different kinds of phrases and if we are able to detect them in the graph. Which is a whole new topic, but maybe connected enough to bring it on here. Cosa ne pensi?
Author: Yuri to Heidi Date: 2008-12-30
Of course, I am game, Heidi, dear. The only thing is that I am just back to Riga with my nose running after Moscow's library chases, and I am leaving Riga for Chicago on Jan 1 early in the morning.
So briefly: the "phrasing" graph from Vertov's 1922 manifesto you mentioned in your comment looks like this:
The important thing is, as we can see, that Vertov sees "wholes" as arches: the big arch for the "work", the smaller parts, under the big unmbrella of the "whole," too, take the form of arches.
Vertov's diagram is not about cutting rates alone, but cutting rates do follow the same principle -- rise-and-fall constitutes a unit, both the work as a a unit and its sub-units.
If you look at Man with a Movie Camera and its cinemetrics graphs you are likely to discover the same principle, with variations: slower segments parts are used to articulate the flow of the film.
We will certainly get to the Story-Flow factor in MWMC later on (next year, that is), but let us first ask why is it that slower parts are used to signal the end of a segment (or of the whole).
Does is have to do with a certain tradition is music? You are Austrian, so this is your call.
Or does it have to do with dynamic universals of human conditions the likes of which we discussed earlier on?
Have you ever heard the BBC Greenwhich Time Signal, or "pips"? Are they all of the same length? Listen to this.
Or if you have Vista Windows you get a series of pop-up boxes each time your Outlook checks you mail. If there are more that one new message, how does the system signal -- this one is the last? Of course -- by keeping it a bit longer than the rest of them.
Author: Heidi Date: 2009-01-04
YURI: But let us first ask why is it that slower parts are used to signal the end of a segment (or of the whole). Does is have to do with a certain tradition in music?
HEIDI: The answer depends on if we talk about the whole film or the reels separately. If we look at each reel and try to make a connection with for example a symphonic scheme (which would be in the classical form 4 “parts”: slow – slow – medium slow – fast), then there is no obvious formal relation. On the whole film – well, at least it ends with “fast”. Is there a special “structure”, a special form of composition, you had in mind? Challenging a poor clarinet playing Austrian to big talks about classical music?
Or did you refer to certain practices in music like “ritardando”? From my experiences with playing in an orchestra when I was young I can remember that it was very easy to make people clap their hands (for example to the Radetzkymarsch) faster and faster, but as soon as you slowed down speed, everybody fell out of rhythm immediately. Why is that so? Why is it more troublesome to find a common understanding of how to slow down than to together speed up? That goes for marching, walking, dancing, making music etc.
Maybe the answer is connected to your second question:
YURI: Or does it have to do with dynamic universals of human conditions the likes of which we discussed earlier on?
HEIDI: Or can physics provide an answer or at least a possible path to an answer? I am thinking about the “Newton’s laws of motion”. I quote:
Lex I: Corpus omne perseverare in statu suo quiescendi vel movendi uniformiter in directum, nisi quatenus a viribus impressis cogitur statum illum mutare. Every body perseveres in its state of being at rest or of moving uniformly straight forward, except insofar as it is compelled to change its state by force impressed. 
What I am trying to say, that somehow humans perceive retardation more, because it causes problems. That is pure observation here. But I thought it might still be helpful to think about why Vertov chose this practice at the end of reels.
One more remark on the “human condition”: People usually walk and move at a certain “normal” speed, they slow down when they are very tired or sick. Or very relaxed, but at a second look we easily can tell the difference. But I think we DO take this second look if somebody is moving slower than the rest. Think about a very common image used in contemporary film: A mass of people filmed bird’s eye view under cranked (moving very fast) and in the middle a single person motionless or moving slowly. Do you have an image of what I am trying to describe?
All speed-effect are noticed, but maybe “slow” more than “fast”? And I know I am talking probably more about in-shot-motion now, but that also could go for shot length.
I raised more questions than answers, pick the sentences that make sense to you and let’s move from there. Brainstorming and collecting thoughts, this is what we do first, right?