Spatial Registry Model : Towards a Grounded Account for Executive Attention

Spatial Registry Model : Towards a Grounded Account for Executive Attention Nader Noori † (nnoori@usc.edu), Laurent Itti †,‡ (itti@usc.edu) †Computer Science Department, University of Southern California Los Angeles, CA 90089 USA ‡Department of Psychology, University of Southern California Los Angeles, CA 90089 USA Abstract Mental tasks that feature algorithmic processing with symbolic items are shown to rely on brain regions known for visual- spatial functions. Yet, exactly how these functions may help execution of amodal tasks remains an open question. Here we propose a hypothesis for manipulation of items in work- ing memory, which relies on registering items in a spatially- organized short-term memory store. Switching executive at- tention to items that need processing may then be embodied through shifting spatial attention towards those registry loca- tions. We studied gaze shifts of human subjects during mem- ory tasks as a proxy for shifts in spatial attention. Analysis of gaze shifts during sorting random sequences of five deci- mal digits indicates that sorting in memory elicits gaze shifts that correlate with sorting procedure. Our proposal estab- lishes a functional relationship between general-purpose pro- duction mechanisms that support algorithmic memory tasks with amodal items, and modal systems for perception and ac- tion. Keywords: Cognitive Architecture; Algorithmic Informa- tion Processing; Visuospatial Attention; Executive Attention; Working Memory; Spatial Registry Model. Introduction Cognition by means of amodal symbolic concepts in an algo- rithmic manner is a unique trait of the human species. Identi- fying the relationship between this newly emerged symbolic machinery and evolutionarily older systems for perception and action is a fundamental question about the nature of hu- man cognition that has motivated numerous studies during recent decades. The focus of a host of these studies has been grounding representation of symbolic concepts in perception and action (Barsalou, 2008; Dehaene & Cohen, 2007; Wood, Willmes, Nuerk, & Fischer, 2008). However, this mode of human cognition is equally depen- dent on a general purpose machinery that can support robust execution of algorithms. Such a machinery should provide basic functions such as temporary maintenance of relevant items and applying selective processing to the maintained items. For example think of an instance of a mental subtrac- tion problem (let’s say 412 - 78) which normally needs to be done in a controlled and algorithmic way rather than an au- tomatic way or direct recall from the long term memory. In this case a feasible algorithm for the mental operation should fulfill the constraints of this machinery such as capacity limi- tations in working memory. Given the limited capacity of the working memory, one might argue that an efficient algorithm for mental subtraction might be different from formal algo- rithms for subtracting Arabic decimal numbers (which usu- ally do not take the notion of working memory into account). These operations usually require maintenance of several sym- bols each one subject to a specific process in each stage. Recent behavioral studies conducted with concurrent tasks paradigm have identified that memory tasks involved with symbolic concepts that feature memory manipulation, inter- fere with visual processing to a larger extent as compared to passive maintaining of similar items (Aky¨urek, Hommel, & Jolicœur, 2007; Fougnie & Marois, 2007; Han & Kim, 2004; Peterson, Beck, & Wong, 2008; Spinks, Zhang, Fox, Gao, & Tan, 2004). These findings suggest that visual perception and parts of this machinery in charge of memory manipulation draw on common resources crucial for both processes. Meanwhile among all regions that are actively involved in executive memory tasks, recently the role of a parietal re- gion with strong visual-spatial characteristics has been high- lighted. More specifically in a patient study, (Koenigs, Bar- bey, Postle, & Grafman, 2009) showed that damage to the su- perior parietal lobule (SPL) is reliably associated with deficits on tests involving the manipulation and rearrangement of in- formation in working memory. This brain region is known for its contributions to a variety of visuospatial functions such as saccadic eye movements(Quintana & Fuster, 1993), visu- ospatial attention (Kanwisher & Wojciulik, 2000; Griffin & Nobre, 2003), visuospatial short-term memory (D’Esposito et al., 1998) and visuomotor functions (Ferraina, Battaglia- Mayer, Genovesio, Archambault, & Caminiti, 2009). Significance of Koenigs et al. ’s findings is partly related to showing that dependency of the symbolic machinery to per- ceptual system is not limited to representation and is also re- lated to the processes that support execution of dynamic or active memory tasks. Tasks featuring memory manipulation are also known to be attention-demanding and thus some researchers have sug- gested that involvement of SPL in these tasks is related to focusing executive attention (Osaka, Komori, Morishita, & Osaka, 2007). Yet existing theoretical frameworks have not elucidated how visual-spatial characteristics of this region are exploited for either memory manipulation or focusing atten- tion. To provide a theoretical account for involvement of vi- suospatial systems in memory manipulation we propose the Spatial Registry Model (SPM). Spatial Registry Model Our proposal assumes a functional role for brain regions with visual-spatial encoding features in registering memory items in a spatially-organized short-term memory. We assume that an item in working memory may register with a correspond-