Representing spatial shifts in event processing

Representing spatial shifts in event processing Glenn Patrick Williams (g.w.williams@dundee.ac.uk) Anuenue Kukona (a.b.bakerkukona@dundee.ac.uk) Yuki Kamide ( y.kamide@dundee.ac.uk ) School of Psychology, University of Dundee, DD1 4HN, UK associated with these locations. However, this effect was subsequently shown to be due to the number of ‘event boundaries’ encountered, rather than the distance travelled. When controlling for distance, but manipulating the number of event boundaries – specifically, doorways – crossed, accessibility for information decreased (Radvansky & Copeland, 2006; Rinck, Hahnel, Bower, & Glowalla, 1997), suggesting that “walking through doorways causes forgetting” (Radvansky & Copeland: p.1154). Abstract Two experiments explored the effects of changes in distance and location on the accessibility of event-related information during language comprehension. In Experiment 1, listeners viewed visual scenes depicting a location containing several objects, while they listened to narratives describing an agent either staying in that initial location, or moving to a new one (either close or far away), and then thinking about one of the depicted objects. We found that eye movements to these objects were modulated (reduced) by changes in location, rather than distance. In Experiment 2, listeners viewed scenes depicting two rooms, while they listened to narratives describing an object moving either between the rooms, or within one room. When the object was mentioned following the event, we found fewer eye movements to it when the movement occurred between rooms. We discuss these results in relation to the Event Horizon model. Keywords: Event cognition; Mental models; Situation models; Spatial processing; Motion events; Visual world paradigm. Introduction When understanding a narrative, we must track information along a number of (often) changing dimensions. For example, what is being referred to, and where is it? Language comprehenders are assumed to activate a set of mental representations that contain information needed to establish a coherent situation (or mental) model of the events described in a narrative (Glenberg, Meyer, & Lindem, 1987). But what is maintained in a situation model, and how does the structure of a situation model affect comprehension? Location representation A series of experiments by Bower and colleagues (Morrow, Greenspan, & Bower, 1987; Rinck & Bower, 1995, 2000) aimed to address whether spatial information is retained in a situation model. They found that language comprehenders often took the perspective of an agent, and focused on information associated with the agent’s current location, while suppressing information associated with other locations. Critically, these experiments reported a spatial gradient of accessibility (Rinck & Bower, 1995), in which the further the agent travelled from previous locations, the harder it was for comprehenders to retrieve information Competing representations Why does crossing an event boundary affect accessibility? The Event Horizon model (Radvansky, 2012) claims that information is structured around, and segmented according to, ‘event boundaries’. If a boundary is crossed, information either side of it is segmented into separate events. This account also argues that (i) one event will be more activated or more in focus than others in comprehenders’ working memory; and (ii) when objects are contained in multiple events, competition can occur (Radvansky & Zacks, 2011). As such, this model predicts a cost to switching focus from one event to another, and for accessing a single representation of an object that is represented in two events. Current experiments Although the current evidence suggests that situation models are structured around event boundaries, and not distance, much of this evidence depends on explicit memory-based tasks, which often take place long after the processing of critical linguistic input. Indeed, both distance and event boundaries have been shown to influence the accessibility of information under certain task conditions (Rinck & Denis, 2004). Here, we addressed two issues: in the absence of an overt task, (i) are situation models organised around distance or event boundaries (Experiment 1); and (ii) do event-based effects stem from event switching costs, or memory load costs (i.e., of maintaining multiple events; Experiment 2)? We used the visual-world eye-tracking paradigm (e.g., Tanenhaus, Spivey-Knowlton, Eberhard, & Sedivy, 1995) to explore these issues.