It was Barry Salt who focused attention on the value of Average Shot Lengths as a descriptive/ analytical tool. I believe he first proposed this measuring device in his article, "Statistical Style Analysis of Motion Pictures," Film Quarterly 28, 1 [Fall 1974]. In this and later work, Salt used ASL as a way of tracking individual directors' styles, as well as trends within film history generally. On the latter dimension, Salt averaged his ASLs for a given period (eg, the 1920s, the 1930s, etc.) to come up with a Mean Average Shot Length for the period. He then compared this with the MASL of other periods.
In my work and in collaboration with Kristin Thompson, I have taken a slightly different tack, arguing that instead of a single numerical value we ought to look for preferred ranges of average shot lengths. In my book, THE WAY HOLLYWOOD TELLS IT, I try to show that the acceleration of cutting in recent decades can be seen as moving from an ASL range of 8-11 seconds before 1960 and towards a range of 4-6 seconds in recent years.
To measure ASL, I simply record the length of the film (in meters, feet, or seconds) and count the number of shots. I then divide the latter into the former. This is straightforward arithmetic, but our primary-school classes taught us that averages are not the whole story. A film with many short shots and a few very long ones can yield an Average Shot Length that fails to reflect the variations in the data. What we have lacked is an efficient way to measure other central- tendency phenomena: the median (the middle item) and the mode (the most frequently occurring value). In addition, we have needed ways to capture the overall editing profile of a film, the serial organization of shots in scenes or sequences or "acts" or whatever part we want to study.
The software developed by Gunars and Yuri will provide a precise tool for these purposes. Yuri's discussion of Intolerance at our Madison conference on film aesthetics showed how the CineMetrics program could capture significant variations in mean, median, and mode in each of the film's epochs. CineMetrics can also disclose finer- grained patterns in a film's editing scheme, allowing us to see at a glance regularities and variations at several levels (the sequence, the section, the entire film).
Of course such quantitative tools need to be supplemented constantly by qualitative ones--the researcher's understanding of narrative context and of the convergence of other stylistic devices (staging, lighting, camerawork, sound design). Nevertheless, by developing precise tools we often discover fresh uses for them, and we can't predict what new insights CineMetrics will generate in the hands of imaginative researchers.
A footnote: While Kristin was working on her recent book on Lubitsch, we were delighted to discover that a German writer proposed an early version of ASL analysis. In 1926 Georg Otto Stindt compared the number of shots per reel in German films and in US films and commented that American films tended to increase the ASL as the film approached its climax, while German films didn't. Stindt also noted that the ASL of German films had fallen sharply between 1921 and 1923. Kristin gives more details in HERR LUBITSCH GOES TO HOLLYWOOD: GERMAN AND AMERICAN FILM AFTER WORLD WAR I (Amsterdam: Amsterdam University Press, 2005), p. 117. As usual, history holds many surprises.