By Susie Allen
Library of Congress image by Eadweard Muybridge (1878)

Films are many things: They are works of art, historical documents, and commercial products. But for scholars like film historian Yuri Tsivian, they are also a treasure trove of data.

The length of a shot, how close the camera is to an actor, the way it moves—Tsivian believes that statistical analysis of these quantitative elements can offer new insights into cinema’s past and future.

Breaking a film into a set of quantitative measures is a bit like diagramming a sentence. “It anatomizes a movie,” explains Tsivian, the William Colvin Professor of Art History, Cinema and Media Studies, Slavic Languages and Literature, and the College.

In 2005, Tsivian founded Cinemetrics, a website devoted to the quantitative study of film. It gives researchers access to open-source film measurement software and publishes the data for scholars around the world to use in their own research.

The software is designed to capture data related to film editing—average shot length, median shot length, and number of shots, among others—that Tsivian and others believe shed light on a filmmaker’s style and artistic choices. We know intuitively that comedies usually feel “faster” than dramas, and older movies feel “slower” than new ones. Cinemetrics allows users to test the hunch.

The site has become an important online gathering place for scholars interested in quantitative film analysis, who regularly submit new films to the database. The flexible nature of the software and the site helped it grow from a tool for studying film editing to a global forum on experimental methods in cinema studies.

In the last decade, the Cinemetrics site has yielded some impressive statistics of its own: a database of some 15,000 films, a New York Times article that used the software to capture data on the relationship between gender and screen time, dozens of journal articles and online papers, and most recently, a conference at UChicago that brought together the international community of cinemetrics scholars.

The rise of Cinemetrics

Quantitative measures of film go back to the earliest days of cinema. In the medium’s infancy, filmmakers and producers struggled to find an editing tempo that would feel comfortable to audiences more accustomed to watching live theater. Some early filmmakers measured shots with stopwatches and even counted frames when editing.

In the 1970s, film historian Barry Salt first proposed that studying average shot length—how it changes within a director’s oeuvre, or how it changes over time across many films—might be useful in studying the evolution of film style. But Salt’s pioneering work was limited by the laborious, time-consuming exercise of measuring films in the pre-digital age, when every shot was timed by hand.

The “intellectual germ” of Tsivian’s interest in statistical film analysis was an essay on D.W. Griffith’s Intolerance. The 1916 film jumps between four stories set in different times and places: ancient Babylonia, Biblical Judea, Renaissance France, and modern America. Tsivian suspected that Griffith’s editing tempo would be different in each era, with the modern story cut more quickly than the Babylonian story.

Tsivian set to work measuring the film, and discovered that there was indeed a significant difference in cutting rate among the four stories. “The older the time, the slower the tempo,” he says. Tsivian doesn’t think the change in editing tempo was deliberate on Griffith’s part, but rather “a gut feeling” that modernity should feel faster than ancient times.

Inspired by his experiment, Tsivian asked computer scientist and statistician Gunars Civjans to develop software that would make it easier to measure films. At Civjans’ recommendation, Tsivian made the software available online—a decision that gave rise to the active community of scholars using Cinemetrics.

Cinemetrics across borders

With support from the Neubauer Collegium for Culture and Society, Tsivian has hosted several collaborators for extended visits to the University of Chicago. The Neubauer Collegium also supported a recent conference, “Cinemetrics Across Borders,” which brought together many of the most active Cinemetrics users—some of whom had corresponded for nearly a decade before meeting in person. Tsivian sees Cinemetrics as “a reverse engineering” of the work of a film editor. He knew he wanted to hear from a practitioner—someone who “has this knowledge of editing in the fingers.”

He invited Sandra Adair, who edited the Oscar-nominated 2014 film Boyhood. In her keynote address, Adair explained how she approached the task of editing a film that unfolded over 12 years.

Adair echoed Tsivian’s belief that an editor’s work is largely unconscious: “What guides me is my gut, my instinct, and my taste,” she said. “I am rarely thinking about why I’m making the decisions I’m making.”

Yet even Adair could see the value in Cinemetrics. In a master class for undergraduates in the Cinema and Media Studies program, she encouraged young filmmakers to use the Cinemetrics tool to study their own films and hone their instinct for editing tempo.

Other presenters at the conference focused on detailed studies of particular filmmakers or works, using Cinemetrics to track the changing style of Finnish director Aki Kaurismäki or to explain how editing creates suspense in the American reality show “America’s Next Top Model.”

James E. Cutting, a psychologist from Cornell University, looked at factors like motion, cutting rate, and color contrast of films over time—all of which correlate with increased attention and all of which, according to his research, have increased in popular movies over time. “It’s clear that film has evolved to match our perceptual and cognitive abilities,” Cutting argued in his lecture.

Studying a medium like film through the lens of statistics may seem needlessly dispassionate to some. For Tsivian, statistical film analysis is one of many ways to think about cinema, and it may not be useful to every film scholar. Cinemetrics, he argues, is simply an addition to the analytical toolbox. “There is no ‘either/or,’” he says.

But statistics and art aren’t always as far apart as one might imagine. As part of his research on film and perception, Cutting and his team began developing visual representations of various films, with each frame represented by a vertical raster: The results resembled abstract art.

Cutting paused at the image. “This conference is mostly humanists,” he observed, “but they too know data can be beautiful.”

Originally published on May 26, 2015.