Decoding Visual Spatial Attention Control

Sreenivasan Meyyappan,Abhijit Rajan,Qiang Yang,George R Mangun,Mingzhou Ding
DOI: https://doi.org/10.1101/2023.08.05.552084
2024-11-05
Abstract:In visual spatial attention control, top-down signals are thought to originate in the dorsal attention network and propagate to the visual cortex to shift baseline activity and bias sensory processing. How such top-down influences are distributed among different areas in the visual hierarchy remains a matter of debate. It also remains unclear whether these changes in baseline activity lead to improved performance. We addressed these issues by analyzing attention-related fMRI baseline activity during the anticipatory period of a cued spatial attention task from two datasets and found the following results: First, univariate analysis revealed that covert attention significantly enhanced baseline activity in higher-order visual areas contralateral to the attended visual hemifield, whereas the effects in lower-order visual areas (e.g., V1) were weaker and inconsistent. Second, in contrast, multivariate pattern analysis (MVPA) revealed significant above-chance decoding of attend-left versus attend-right conditions in all visual cortical areas, with lower-order visual areas actually having higher decoding accuracy values than higher-order visual areas. Third, decoding accuracy, not the magnitude of differential univariate activation, predicted the stimulus discrimination performance. Fourth, the MVPA findings were replicated across two experimental conditions where the direction of spatial attention was driven either by external instructions (instructed attention) or by purely internal decisions of the participants (choice attention). These findings provide new information on the spatial extent of top-down attentional biasing in visual cortex and how it affects perception and behavior.
Neuroscience
What problem does this paper attempt to address?