Perceptual segmentation of natural events: Theory, methods, and applications

Humans organize the continuous ebb and flow of everyday experience into temporal units, or events. A Sunday afternoon might be broken up into shopping for groceries, doing laundry, and calling home. In turn, doing laundry might be broken up into sorting clothes, loading the machine, adding detergent, and so forth. A growing body of research shows that this organization results, at least in part, from cognitive processes that segment ongoing experience into meaningful events as part of normal perception (e.g., Newtson, 1973). This has important consequences across a range of cognitive functions including attention, comprehension, and working and episodic memory (e.g., Boltz, 1992; Ezzyat and Davachi, 2011; Nobre, 2001; Schwan and Garsoffky, 2004; Zacks, Speer, Vettel, and Jacoby, 2006). The primary goal of this chapter is to provide a tutorial on techniques for studying event segmentation. We will illustrate these techniques using applications to clinical neuroscience, interface design, and film. We hope to show that research on applied perception can benefit from an understanding of how we perceive temporal structure. Event Segmentation Theory Event segmentation theory (Zacks et al., 2007) proposes a computational and neurophysiological account of how a continuous activity - such as grocery shopping - is broken down into discrete subevents (e.g., getting a cart, choosing grocery items, standing in line, paying for groceries). Here, we restrict discussion to the information-processing level. (The relevant neural correlates are discussed in Zacks et al., 2007.) According to event segmentation theory, information related to ongoing activity is held active in an event model, which is a representation of what is currently happening similar to situation models constructed during reading (e.g., Zwaan and Radvansky, 1998). Event models are maintained in working memory. They represent information that remains fairly stable throughout an event and integrate information across multiple modalities.