Accurate Saccade Guidance without the N2pc Component: A Computational Explanation with RAGNAROC-PLUS (RAG+)
Research Poster Social & Behavioral Sciences 2025 Graduate ExhibitionPresentation by Joyce Tam
Exhibition Number 37
Abstract
The N2pc is a neural correlate of covert attention often equated with pre-saccadic guidance. However, observations of accurate saccades without a preceding N2pc challenge this idea. Here, we present a computational model based on structural and functional aspects of the ventral visual stream and the ventral-medial pulvinar . RAG+ differentiates between feedforward (posterior to anterior) attention, which is rapid but spatially coarse-grained, and the optionally invoked feedback attention, which improves the spatial resolution of attention and converges attended locations across feature-selective maps. In RAG+, the N2pc results from feedback, such that its occurrence is not mandatory, but instead associated with the need to pinpoint specific locations. Within the same set of mechanisms, RAG+ successfully simulates accurate saccades without pre-saccadic N2pc and other standard benchmarks regarding the N2pc and search behaviors. We present RAG+ as a formal theory of visual attention encompassing the intricate interactions between covert and overt attention.
Importance
Understanding how we decide on what and where to attend is now more important than ever in this age of information explosion. We build a novel model to account for the integrative use of eye movements and covert attention in visual attentional orienting. A computational model of this kind can produce simulations and predictions that allow us to test scientific theories as well as practical applicability in areas such as visual design and visual assistant technologies.