Decoding the Locus of Covert Visuospatial Attention from EEG Signals

Area of Science:

• Cognitive Neuroscience
• Neuroscience
• Human Computer Interaction

Background:

• Event-related potential (ERP) components can indicate left or right visual field attention.
• Precise spatial localization of attention using EEG during central fixation remains unclear.

Purpose of the Study:

• To determine if EEG signals can track visual spatial attention with greater spatial precision.
• To evaluate the decoding accuracy of attention location using machine learning.

Main Methods:

• Modified Posner cueing task with endogenous cues.
• Machine learning classification (multi-class SVM) with leave-one-out cross-validation.
• Analysis of EEG signals from occipital and parietal regions for 200 ms presentation sequences.

Main Results:

• ERP features predicted attention location with 57% accuracy (p < .001), significantly above chance (25%).
• The mean prediction error was 0.55 degrees of visual angle.
• EEG successfully predicted attention allocation (vs. non-allocation) with 79% accuracy (p < .001).

Conclusions:

• EEG signals contain information for precise visuospatial attention decoding.
• Findings support the potential for EEG-based tracking of attention focus.
• Future research can explore advanced applications in brain-computer interfaces and cognitive studies.