But there's another way to think about it, which is in terms of attention. Tim Smith, Daniel Levin, and James Cutting (2012) have recently presented results on film editing in what they call the Hollywood style, also called continuity editing. They argue that this kind of editing as it has developed in the last hundred years has come close to the ways in which people direct their attention to events and actions in everyday life. If we follow their argument, we can see how viewers of movies made in the Hollywood style seem to see something like reality, or at least what they would have seen had they been there in the world of the story.
In ordinary visual perception we sample the world from a sequence of fixations. At each fixation only data in the visual field from an area about the size of a nickel or a five eurocents piece held at arm's length are seen in high acuity. About 50% of the optic nerve and 50% of the visual system is given over to processing input from these areas of high acuity. Between fixations the eyes change position in a movement called a saccade, during which no new information is registered from the retina. One reason we change our fixation is to pay attention to a movement. To do this, we make a saccade to the point of movement. Film editors insert cuts in the same manner, for instance when a character moves a hand or turns his or her head. These are called "match-action" cuts. Cuts of this kind therefore correspond to when there is a natural cut by means of a saccade. Smith et al. have performed an experiment in which they had participants register the occurrence of cuts in a movie. People were good at noticing cuts at the end of an action, when the act was completed; about 90% of such cuts were noticed. By contrast only about 70% of match-action cuts were noticed by viewers, and the authors argue that this lesser rate of recognition is because match-action cuts correspond to attention switches during saccades.
Smith et al also found similarities in people's distribution of attention to scenes in which actions began, continued, and ended both in ordinary life and in movies. In other words, actions are natural units in our perception of the world, and movie-makers follow this same pattern. When watching a movie, distinctive fMRI activation occurred at the end of an action, whereas when viewpoint simply changed within a scene, this same kind of fMRI activation did not occur.
It seems that we parse the social world into agents and actions. This may not be surprising, but Smith and his colleagues have shown that the idea scales down to fairly small actions such as a glance, picking up an object, a slight change of expression.
The image is from Smith et al.'s paper. It shows how the eye fixations of 19 viewers are closely grouped within a shot seen during a movie clip. Yellow spots are for viewers who saw the clip with audio, and pink spots are of viewers who saw the clip without audio.