Filling in the Gaps: Spatiotemporal Dynamics of Alpha Oscillations in Visual (Non)Illusory Perception
This project explores how the brain constructs visual experience when information is incomplete; specifically, how we come to “see” shapes that are not physically present. Using EEG, I investigated how neural activity unfolds over time as participants perceive classic visual illusions, such as Kanizsa figures, compared to closely matched control stimuli. The study focuses on alpha-band brain oscillations, which have been proposed to support top-down inference during perception, and explores how these signals relate to both perceptual decisions and confidence.
This work was conducted at the Conscious Brain Lab of the University of Amsterdam, under the supervision of Dr. Simon van Gaal and Dr. Jan Willem de Gee.
Abstract
What happens in the brain when we see things that are not even physically present? Alpha-band oscillations (8–12 Hz) have been proposed as neural substrates of top-down prediction in visual perception, particularly in the context of illusory contour processing. However, the spatiotemporal dynamics of alpha power and its relationship to behavior during visual inference remain unclear.
In this EEG study, 28 participants performed two visual discrimination tasks (i.e. detecting either illusory Kanizsa figures or rotated triangle configurations) while also reporting confidence on each trial. Time-frequency analyses dissociated pre- and post-stimulus alpha-band dynamics across frontal, parietal, and occipital regions. Contrary to hypotheses, I observed no condition-specific modulation of alpha power before or after stimulus onset. Instead, alpha exhibited a general U-shaped post-stimulus trajectory as well as anticipatory suppression across conditions, consistent with domain-general sensory engagement rather than predictive feedback.
Exploratory spectrograms revealed a transient occipital alpha suppression (~200–300 ms) during illusory trials, possibly suggesting early perceptual integration. Additionally, alpha power predicted confidence in a region- and condition-specific manner, showing a double dissociation between Illusion and Rotation trials. These findings challenge the view that alpha power directly encodes top-down predictions and instead suggest a broader, context-sensitive role in perceptual readiness and metacognitive evaluation.
Together, the results highlight the need for more temporally precise and feature-specific methods to characterize the neural mechanisms of perceptual inference.
Keywords: visual perception, illusory contours, alpha oscillations, EEG, predictive coding.