The evolution of pace in popular movies

Abstract

Movies have changed dramatically over the last 100 years. Several of these changes in popular English-language filmmaking practice are reflected in patterns of film style as distributed over the length of movies. In particular, arrangements of shot durations, motion, and luminance have altered and come to reflect aspects of the narrative form. Narrative form, on the other hand, appears to have been relatively unchanged over that time and is often characterized as having four more or less equal duration parts, sometimes called acts - setup, complication, development, and climax. The altered patterns in film style found here affect a movie's pace: increasing shot durations and decreasing motion in the setup, darkening across the complication and development followed by brightening across the climax, decreasing shot durations and increasing motion during the first part of the climax followed by increasing shot durations and decreasing motion at the end of the climax. Decreasing shot durations mean more cuts; more cuts mean potentially more saccades that drive attention; more motion also captures attention; and brighter and darker images are associated with positive and negative emotions. Coupled with narrative form, all of these may serve to increase the engagement of the movie viewer.

Keywords: Attention; Emotion; Evolution; Film style; Movies; Narrative; Pace; Popular culture.

Fig. 1

The movie-length patterns of shot duration (in seconds, left panel), of the extent of motion (middle panel), and of relative luminance (right panel) across 180 popular English-language movies released from 1930 to 2015. The lengths of the movies were normalized so that each measure was assessed across 100 equal-duration bins, and those values were normalized within each movie before averaging across them. Shot duration (converted back to seconds), flicker motion (converted back to mean correlations of eight-bit pixel values across frames), and luminance (converted back to the mean pixel value per frame) are three roughly orthogonal measures of film style (Cutting, 2016). Polynomial fits are shown as darker lines surrounded by white areas that correspond to the 95% confidence interval of that polynomial. Lighter colored regions denote the 95% confidence intervals on the data, given that polynomial. The data of each panel are also separated by vertical lines that suggest four narrative divisions (see Thompson, 1999): the setup (roughly bins 1–25), the complication (bins 26–50), the development (bins 51–75), and the climax (bins 76–100). Also noted are a prologue (initial and variable number of bins of the setup) and an epilogue (a variable number of bins at the end of the climax), optional units that do not appear in every film. A take is another name for a shot. The term long shot is reserved for a wide-angle shot, not a long-duration shot

Fig. 2

Variations in mean shot duration (in seconds) across the duration-normalized films of three different eras of popular movies. The extents of the vertical axes show equal variance in the data across the three panels. The statistical lines and regions of the middle and right panels are as they were in Fig. 1; there is no reliable pattern in the left panel. Vertical lines in all panels separate the bins into the four narrative sections, as in Fig. 1

Fig. 3

The mean flicker motion variation across the duration-normalized movies in three different eras of popular movies. The extents of the vertical axes show equal variances in the three datasets. Again, the statistical lines and regions of the middle and right panels are as they were in Fig. 1; there is no robust pattern in the left panel, but the purple line shows a reliable polynomial pattern that fails cross-validation. Vertical lines in all panels separate the bins into four narrative sections

Fig. 4

The mean luminance variation across the duration-normalized movies in three different eras of popular movies. The extents of the vertical axes show equal variance in the three datasets. Again, the statistical line and regions of the left panel are as they were in Fig. 1; here, there is no robust pattern in the left or middle panel, but the purple lines show statistically reliable patterns that failed cross-validation. Vertical lines in all panels separate the bins into four narrative sections

Fig. 5

In the 30 silent movies of this sample, there are no reliable patterns of shot duration, motion, or luminance that survive cross-validation, regardless of whether the shots of intertitles are included in the analysis. However, the pattern of shot durations and luminance with intertitles and motion without intertitles showed statistically reliable polynomial fits, but these failed cross-validation