Keywords
visual attention, inhibition, feedback
Abstract
The featural and spatial specificity of visual representations broadly decrease along the ventral visual stream. The selection of behaviorally relevant information, or attention, must therefore establish spatial correspondence across the visual hierarchy while maintaining behavioral guidance to relevant visual features. Moreover, selection is often accompanied by an inhibitory surround as well as selective inhibition of distractor items. Considering these key functions, we describe a biologically realistic computational theory of visual selection and inhibition through feedforward and feedback signals along the ventral visual stream, complemented by feature-agnostic spatial competition in the pulvinar nuclei. Our model simulates signature visual search behaviors and the N2pc and Pd, which are EEG correlates of attention.
Start Date
15-5-2024 4:00 PM
End Date
15-5-2024 5:00 PM
Location
University Park, PA
figure in high resolution
modvis_acknowledgements_references.pdf (170 kB)
funding acknowledgements and full reference list
Included in
Feature Integration and Spatial Localization for Attention across the Visual Hierarchy
University Park, PA
The featural and spatial specificity of visual representations broadly decrease along the ventral visual stream. The selection of behaviorally relevant information, or attention, must therefore establish spatial correspondence across the visual hierarchy while maintaining behavioral guidance to relevant visual features. Moreover, selection is often accompanied by an inhibitory surround as well as selective inhibition of distractor items. Considering these key functions, we describe a biologically realistic computational theory of visual selection and inhibition through feedforward and feedback signals along the ventral visual stream, complemented by feature-agnostic spatial competition in the pulvinar nuclei. Our model simulates signature visual search behaviors and the N2pc and Pd, which are EEG correlates of attention.