Project Summary  Real-­world scenes contain far more information that we can perceive and comprehend at any given moment. A  key mechanism for making real-­world scene perception tractable is visual attention—the mechanism of  preferentially processing only part of the scene at any given time. What we attend to in a scene determines  what we see, understand, and remember. Attention is guided by both the visual properties of the scene itself  and by our knowledge about similar scenes and the world in general. How knowledge is used to guide  attention through a meaningful scene remains largely unknown. The central idea behind this proposal is to  address this fundamental scientific question by focusing on two critical aspects of scene knowledge:  knowledge about where a given object is likely to appear in a scene, and knowledge about which regions and  objects in a scene are meaningful and informative. The studies aim to determine how spatial and meaning  constraints are used to guide attention in scenes.   This proposed research is innovative in combining high-­resolution eyetracking with novel experimental  paradigms for manipulating and measuring knowledge-­based constraints. First, a new fusion of spatial learning  methods with eyetracking is used to study the influence of spatial knowledge on attentional guidance. Second,  new quantitative scene-­rating and information-­theoretic metrics are used to index meaning in scenes, providing  a new theoretical approach to scene meaning and new empirical tools for investigating meaning. Third, real-­ time scene manipulation based on the viewer’s eye movements is combined with manipulations of spatial and  meaning constraints to investigate how quickly knowledge about a scene becomes available to guide attention.  The project is significant in challenging current models to explain the role of knowledge in guiding attention in  scenes. The experiments are designed to advance the field regardless of the outcome, and will provide rich  and theoretically constraining results that may have a transformative effect on current theory. In addition, the  proposed research has important translational implications because deficits in attention and perception are  suffered by many psychiatric and neurological populations. By understanding how knowledge influences the  guidance of attention in real scenes, the proposed studies can ultimately lead to the development of targeted  rehabilitation strategies for the real world that better capitalize on both disrupted and spared functions.

Project Narrative  This research investigates how our knowledge of the natural environment helps us guide attention in real-world scenes. The studies focus on knowledge concerning where a given object is likely to appear in a given scene, and knowledge concerning which regions are likely to be informative in a given scene. The experiments combine high-resolution eyetracking, tight experimental control of spatial knowledge, novel application of information-theoretic methods to quantify information value in scenes (surprisal and entropy), and new gaze-contingent display-change methods for exploring the time-­course of knowledge use. Predictions from cognitive and image-based attentional guidance models are contrasted. Findings will increase our theoretical understanding of attention and perception as applied to meaningful real-world scenes, have the potential to assist in the identification and of individuals with attentional and perceptual deficits, and can ultimately lead to the development of targeted rehabilitation strategies for the real world.