Abstract
Filmmakers often rely upon actors' gaze direction to create a compelling sense of space. We tested whether this “sense of space” leads observers to form correct representations of implied object locations, and whether correct localization is more difficult when a canonical, space-defining gaze event is disrupted. Participants viewed edited films with four shots. In the canonical films, the first shot showed two actors sitting at a table. The second shot showed one actor lifting an object and looking at the other actor. The third showed the second actor appearing to shift his/her gaze to follow the object being placed in one of four locations on the table. The final shot was a close-up showing the object being placed on the table. None of the shots identified the target location except the gaze-shift. In the non-canonical version of the film, the third and fourth shots were swapped. Participants viewed 8–16 of these sequences (half canonical and half non-canonical). Immediately after each sequence, participants were asked which location the object was placed in. At the end of the experiment, participants were asked if they had noticed the different shot orderings. In each of two experiments, participants who were unaware of the different orderings performed more poorly on the non-canonical sequences than participants who were aware of the orderings, while participants who were aware of the different orderings performed better on the non-canonical films. We ague that participants can successfully infer the locations of objects based on gaze, but that they approach these sequences in one of two modes: a default mode in which they are minimally aware of moment-to-moment event structure, and assume typical event sequencing, and a visually focused mode in which viewers attend to fine event structure, allowing them to account for atypical events.