One perspective or two? Viewpoint dependency in visual events

Abstract

Viewpoint dependency in dynamic events is still an open question. Movies present a unique case of complex visual stimuli where consecutive shots are filmed from multiple viewpoints. In the present study, we have examined whether people remember viewpoint-specific information in movie-like visual scenes. We have used naturalistic activities which involved two actors where a) the sequence is presented from one or both actors' viewpoints and b) the individual actions were in a natural order or they were scrambled. The results indicated that memory for individual shots decreased when the sequence was presented from both actors' perspectives. Also, people were mostly unaware of the order manipulation, and reordering the individual actions did not lead to a decrease in memory performance. The results favor the film-form model, which suggests that the spatial relations in movie scenes are represented by taking the first shot of a scene as the basis and incorporating the views in subsequent shots accordingly. This argues for an economical encoding in visual events, which favors narrative continuity over spatial relations.