Fractal Structure of Event Segmentation: Lessons From Reel and Real Events

Abstract

We investigated fractality in the segmentation of cinematic and live events. Four movies from distinct genres were first analyzed and found to be fractal, albeit in different regions of the fractal range. In Experiment 1, viewers parsed the films' events, attending to short or long timescales, using a button box to indicate break points. The temporal structure of film viewing was found to be fractal regardless of the scale of attention and to reflect the differing fractal editing structure of the films. Experiment 2 found event segmentation to be fractal in a natural setting, a basketball game, as well. An additional parallel between reel and real events was uncovered: fractal indices were at the higher end of the range for both the film genre intended to be anxiogenic and the more suspenseful portion of the basketball game. We argue that the fractal structure of perception reflects the fractal structure of events whether the events are in film or in nature. We speculate that fractal structure has implications for understanding subjective experience and, ultimately, the correspondence between films and the world.

Notes

¹Here and elsewhere we present the average interval (i.e., the average interedit interval or the average intertap interval) rather than the average number of edits or taps. Because all of the time series are the same length (2 hr) the average number of taps can be derived by dividing 7,200 (120 min × 60 s) by the average interval.

²It is worth noting that the play-by-play time series provided by ESPN is not an objective measure. It is also the product of a perceiver who notes the time that specific types of events occurred (e.g., shots, rebounds, fouls, free throws, turnovers, time-outs). Moreover, the game clock is stopped for some of these events and not for others and there is no indication of how long these stoppages are.