As I understand it, the nub of the matter is that the increase in the autoregression indices, AR and mAR, and in the slope of the 1/f measure in Fourier decomposition of shot lengths, results from having more strings of roughly equal length shots in a film.

These strings of roughly equal length shots occur mainly in action scenes, but also in scenes edited to music, either in dramatic films, or in actual musicals. (Anchors Aweigh and Popeye are actually musicals, not simple comedies.) This can be seen by comparing the shot length records for two of Cutting et al.’s films, The 39 Steps: (6) ASL 8.6 and Sunset Blvd.: (6) ASL 14.9. on the Cinemetrics database. If you use the Moving Average function at 20 shots, you can see that Sunset Blvd. has no flat regions with approximately equal shot lengths, whereas The 39 Steps has several. The 39 Steps has a high mAR etc., while Sunset Blvd. has a low mAR according to Cutting, DeLong & Nothelfer.

The trend towards increasing these indices comes from having more of the running time of an action film devoted to action scenes. There may also be a trend towards editing action scenes to a pre-existing music track, either a dummy one, or the actual piece of pop music that is going to be used in the final sound track, which also helps the effect.

Editors have more rushes to deal with for two reasons. More shots are taken so that an increased cutting rate is possible for the film, and as part of that, more “coverage” is shot. (“Coverage” is more shots of the same action taken from different angles.) Many of the Hollywood directors of the classical period prided themselves on “cutting in the camera” – i.e. only taking the shots they knew would be needed in the final cut. One does not hear that sort of boast any longer.

So this current situation does give editors a bit more freedom about the lengths of shot they use. But not in ordinary dramas, where the length of the lines of dialogue strongly influence the shot lengths.

Insofar as I was talking about a “floor effect” in my previous comments, I was referring to the fact that if you have a distribution like that for Derailed, pretty well all the shots have to be of the length 1, 2, or 3 seconds, when measured to the nearest second. And at that accuracy, there are likely to be more strings of shots of nearly the same length. (There is obviously an actual floor in film shot lengths, since no shot can be shorter than one frame. However, we are still some way short of that in ordinary commercial films, though not in avant-garde films.)

So a minor query is about the measured accuracy of the shot lengths actually used in AEHF. I can’t find a statement about this in the piece. Is everything measured and worked to the frame, or to one tenth of a second, or just to the nearest second?

Another thing about Cutting et al.’s ideas that still bother me is the question, “Why are films mostly still falling short of the maximum mAR and 1/f slope after 90 years of the use of standard film construction, if the postulated psychological effect is so powerful?”

The new suggestion that “long waves” (in the Fourier transform?) measure the location of “Acts” in the film structure seems dubious to me, since I don’t believe that film scripts are actually constructed in terms of “Acts”. I think most films script-writers don’t believe in “Act” divisions either. The act concept is indeed used by many film producers and directors in discussing scripts, and of course by non-script writers writing books about how to write film scripts. Anyway, this can be tested by using the Fourier transform method to see whether it gives locations for act boundaries that coincide with those postulated in a particular film by someone who believes in them, such as Kristin Thompson.

Barry Salt, 2010