Attentional Selection In Natural Scenes

Abstract

The number of objects present in our visual field at any point in time exceeds the processing capacities of our visual system, resulting in inter-object competition. This competition can be resolved through the process of selective visual attention. This process allows us to prioritize locations, features, objects, or entire object categories that are relevant for ongoing behavior, while at the same time ignoring those that are currently irrelevant.

The work presented here aimed to further our understanding of the neural mechanisms and behavioral processes involved in the selection of relevant categorical information from our natural visual environment, or, real-world visual search. The framework that has guided this research posits that real-world visual search is mediated by the formation of category-specific search templates. These search templates are implemented in object-selective cortex where representations are biased towards the task-relevant object category. The work presented here extends this framework in several important ways. First, in addition to being biased towards task-relevant information, representations in object-selective cortex are biased away from previously but no longer relevant object categories, suggesting that outdated attentional templates are actively suppressed during real-world visual search (Chapter 1). Second, category-specific search templates are implemented by control signals that arise from regions within the frontoparietal attention network which are known to carry high-level object representations (Chapter 2). Third, following the implementation of search templates, attention is efficiently and automatically guided towards a wide range of task-relevant information (Chapter 3). Fourth, attentional search templates are informed by semantic long-term memory, such that attention is automatically drawn towards objects that are predictive of the presence of the target category (Chapter 3). Finally, category-specific search templates may be supported by spatially-specific templates that guide attention towards probable target locations (Chapter 4).