Statistical Analysis of Television Style: What Can Numbers Tell Us about TV Editing?
Jeremy Butler, The University of Alabama
Ostensibly, my essay is on the “Statistical Analysis of Television Style,” but it also offers some general thoughts on how media scholarship is changing in the digital era. And so it seems particularly appropriate to take advantage of the online, digital embellishment of the print publication of Cinema Journal in the relatively new “Afterthoughts and Postscripts” (A&P) feature. My first (after)thought is to use A&P to present my article’s illustrations in color and at full resolution (see below). I have already done this on my own hacked-together companion Website, but it’s best to have these materials on SCMS’s official online space—making them more easily accessible and, one hopes, more permanently accessible. Color, high-resolution imagery, as well as audio and video, have been slow to come to SCMS publications, but with the addition of the A&P posts and SCMS’s newly launched online journal of videographic essays ([in]Transition) we may be witnessing an evolution in media studies. However, the future of such efforts remains unclear. Will they supplant the conventional academic essay in a print-originating journal? (And how much longer can journals sustain their print origins?) Or will they be seen as a mere embellishment of those essays? It seems likely that some middle ground will emerge—a location where the logical rigor of a conventional essay can benefit from the analytical power and flexibility of online, videographic essays.
My own engagement with digital humanities grew out of a desire to more accurately measure and analyze the rhythm of editing in television and film—hence my subtitle, “What Can Numbers Tell Us about TV Editing?” It was based on data I collected from Shot Logger, a PHP/MySQL site I initially coded in my “spare time” in 2007. As the years went by and the amount of data grew in Shot Logger, I came to the realization that there may be some statistical value to it; but as someone untrained in statistics, I had to call on several colleagues to explain its statistical worth to me. Without their collaboration, this essay could never have been written. I still have only a basic understanding of statistics, but my experience with Shot Logger has piqued my interest in using statistics and other software-based tools to describe and analyze phenomena that are too big or too small to be comprehended through traditional humanist methods. I thought that my next step would be to experiment with the tools for visual analysis developed by Lev Manovich and the Software Studies Initiative, which I dabbled in during the time just before I submitted this manuscript to Cinema Journal. However, my digital-humanities interests have taken me in a different direction.
It occurred to me that the software I developed for Shot Logger to measure the time between cuts could be modified for analyses of other time-based aspects of television and film. And this realization aligned with a new book project on the sitcom I have undertaken. Most TV scholars—myself included—disdain the sitcom laugh track, but it is undeniably a foundational component of the genre ever since I Love Lucy and radio sitcoms before it. Best I can tell, there has been very little rigorous study of the laugh track itself. We all know intuitively that some shows have more hyperactive laugh tracks than others, but how can this be described accurately? Then it came to me. Shot Logger could be rejiggered to measure the time in between laughs and count their frequency. And so during the past few weeks I have been forking (as programmers say) Shot Logger into Laugh Logger. The results are still extremely preliminary as I have only logged laughs in ten episodes from four programs, but I’ve already learned that Bewitched, M*A*S*H and How I Met Your Mother—four programs with very different laugh tracks—share the same median time between laughs, which I’ve dubbed the “median laugh length” (MLL). In all of the episodes from those programs, the MLL is 2 seconds. Seinfeld, which revolutionized sitcoms in a variety of ways, has a significantly longer time between laughs (an MLL of approximately 10 seconds). Obviously, it’s too soon to tell where this research may lead me, but I’m encouraged that a tool I developed for shot measurement might prove productive in other research. I will report my progress (or lack thereof) on LaughLogger.org in the near future.
Please click on the following images to enlarge them.
Figure 1: Lignes de Temps facilitates detailed, shot-by-shot analysis of films -- allowing annotations organized around a time (Institute of Research and Innovation of the Centre Pompidou, 2006).
Figure 2: Cinemetrics crowdsources the collection of editing data by distributing its own shot-measurement software and soliciting individuals to contribute data on films and TV programs (Cinemetrics, 2005).
Figure 3: EDIT2000 converts the edit decision lists generated by digital video editors into statistical charts and other formats, such as this chart showing the lengths of fourteen shots in a short sequence and calculating their average shot length and media shot length (EDIT2000, 2008).
Figure 4: Videana attempts to automate the analysis of shot lengths and the visual aspects of individual shots. Among other data, it generates a shot list, with thumbnail images, that identifies when each shot begins and how long it lasts (Media Upheavals, University of Siegen, Germany, 2007).
Figure 5: Shot Logger processes frame captures to automate shot-length measurement in films and TV programs (Jeremy Butler, 2007).
Figure 6: A highly simplified flowchart of television's production and reception illuminates how statistical correlations can be used to understand correlations among these processes.
Figure 7: The scatter chart contrasts the average shot length (squares) and media shot length (diamonds) of US narrative TV programs for each year from 1951 to 2011. Dashed and solid trend lines indicate the rate at which ASL and MSL have been getting shorter and converging.