There are a few useful things one learns playing with the cinemetrics database sort-by buttons. If your click on “Date” you’ll learn that this website is already one year old. Gunars Civjans set it up in November, 2005, with one or two test movies he and I submitted. On December 18 Casper Tybjerg of Denmark submitted two short by Dreyer. For us this was the first sign cinemetrics might interest someone other than ourselves.
If you click on “Submitted by” you will see that in the course of this year cinemetrics has attracted 22 outside contributors. As many of them not only submit their films but also look up and comment on other submissions, to call these people “outside contributors” is a mere technical description. It may be my fantasy, but I perceive people contributing to cinemetrics as a team. Or, better (since Gunars and I know more about sailing than we do about baseball) as a crew. There is no land in view yet, but there is a sense we are under sail.
If you click on “Film title” and have enough patience to count them you will count 294 different titles. But then, this is not the only thing we learn looking at the cinemetrics database.
Click on “Year” and you’ll see that the space of time it covers is 1902-2006 – if not the entire span of the history of film, then the entire history of editing.
I also found it instructive to sort the database by “ASL” – to find out about the shortest and the longest average shot length yet registered. It turns out that our ASLs range from an incredible 0.9 seconds to an incredible 46.
Is there a trend to be found in the distribution of longer and shorter ASL data across the database? Even a rough estimate shows there is. It is not by chance that the slowest film of those posted on cinemetrics so far was made in 1902 and that the fastest was made in 2000. No one knows better than our crew how complex the question of editing is and how many factors affect the cutting tempo of films, but if we are asked “do films become faster in the course of film history, yes or no?” cinemetrics says yes. True, we have not yet inspected every film ever made, but 294 film selected by 22 people for different purposes is a decent unbiased sample.
Let me show something using a thumb-rule method. Click on the “ASL” sort-by button and compare the upper section of the resulting list with the lower one. Take, let us say, 12 fastest and 12 slowest films submitted so far and compare their years. The picture I thus got is this (I exclude film fragments and the baseball game of course):
SLOW END (ordered by year): an ASL of 12 films made in 1902, 1909, 1909, 1909, 1911, 1911, 1912, 1913, 1929, 1915, 1966 and 1974 ranges between 46 and 19.7 seconds.
FAST END: (ordered by year): an ASL of 12 films made in 1966, 1968, 1971, 1976, 1999, 2000, 2002, 2003, 2003, 2003, 2004, 2004 and 2006 ranges between 0.9 and 3.3 seconds.
Another way of showing this is to sort the database by year. This will show you that the fastest film made between 1902 and 1909 has an ASL of 15.8 seconds while the slowest one made between 2000 and 2006 has an ASL of 10.01 seconds. In other words, the fastest runner in the beginning-of-the-past-century group is 5.7 seconds behind the slowest one of the twenty-first century.
I am not betting here and I am anything but a speed fetishist. I am aware of the fact that the slowest film in the modern group happens to be Before Sunset – a remarkable, one-off movies perhaps of the strength of its slowness. Nor am I implying that in the course of time filmmakers “learned” how to cut faster. But the obvious fact that the history of cinema is not the history of sports, and that in the middle of it (as we learn from Barry Slat and Charles O’Brien) there is significant slow-down of cutting – all of this does not warrant us to look away from the fact that cutting becomes faster. The harder the fact to explain the more useful it is for the student of film.
If we sort the database by “Mode” we’ll find out that roughly half of the submitted titles have been timed in the advanced mode. This allowed us to see a few interesting patterns which would have been much harder without the customizable buttons the cinemetrics tool allows us to use.
Here are a few things we learned about film cutting by way of inspecting films in the advanced mode this year.
As Torey Liepa has shown on multiple examples from the 1910s, intertitles in this period are invariably shorter than shots. This may vary from case to case, from country to country and from one filmmaker to another. For instance, as Kian Bergstrom has shown, in Das Cabinet des Dr. Caligari the title length is close to that of images, and in Nosferatu the average length of written texts even exceeds the ASL of images. Still, Nosferatu (with its maps, cryptic letters and other inserts) appears to be an exception (another exception – intentional, this time, is Greenaway’s Dear Phone dissected by Kian Bergstrom) – as a rule, in silent films words take less time than action, which squares well with Matt Hauske’s advanced mode timing of some of the silent films by Ozu.
As Charles O’Brien has shown scenes with dialogue and singing in the early talkies are invariably longer than action-only shots. This discovery stands in an interesting relationship with Liepa’s and, like his, deserves further thinking. Put together, the “O’Brien law” and “Liepa law” will help film historians to account for the slow-down in cutting in the early sound period.
As John C has shown a correlation exists between the tightness of the shot and the length of the take – the closer the shot the shorter its ASL. Charley Leary’s results also confirm this. It appears that the correlation does not apply to long and very longs shots. Too few shot-scale related measurements have been to say how strong and stable this correlation is. My sense is that it changes over film history, if so it may be very instrumental it tracing the history of film style. It may also turn out that the scale/ASL correlation changes depending on which shot scale is dominant in this particular film. If this is the case there must be a correlation between the ASL of the given scale and the percentage of time this particular scale takes of the total length of the film.
Charley Leary has shown us various unexpected goals for which the advance mode button can be programmed. It can be used to establish the ratio of interior and exterior shots (see also Stagecoach analyzed by Matt Hauske), to connect the length of a take with the number of “bodies” in the shot, with the type of camera movements, and even with music beats. And Torey Liepa has even measured the back-and-forth between the baseball TV show and the commercial breaks. If he is hired by a TV station efficiency expert, cinemetrics disclaims any responsibility for sport-show revenues. And if we look at the films submitted by Matt Hauske we’ll see that his approach is to break films into sequences and look at what factors define editing in each. v Now, if you click on the Discussion board you’ll find 16 discussion topics – and, chance are, one or two inscrutable messages that accost Cinemetrics with mysterious regularity. Extraterrestrials sending their advice? Then we apologize for terminating them.