Contextual biases and information processing in visual perception

Su, Yannan (2024): Contextual biases and information processing in visual perception. Dissertation, LMU München: Graduate School of Systemic Neurosciences (GSN)

Abstract

Human visual perception relies on context, as the brain integrates visual sensory evidence with contextual information to form a visual percept. When physical visual features are perceived differently from what they actually are, visual biases can occur. Extensive research has been conducted to investigate contextual information processing for different types of contextual biases and distinct visual features. The current thesis aimed to investigate biases resulting from natural and instructive contexts, with a particular focus on two visual features: color and orientation. Two specific questions are addressed in this thesis: (1) whether perceptual bias reflects information from natural context in color vision, and (2) how a reference acts as a context that biases the perception of orientation or color features. The thesis consists of three studies presented as three manuscripts, each addressing a different aspect of contextual biases in visual perception. The first study tackles the first question by measuring perceptual biases in discriminating between noisy hue ensembles. The results show systematic biases with zero-crossings near a non-cardinal, blue-yellow color space axis. A Bayesian observer model further reveals a prior for natural daylights underlying these perceptual biases. The second and third studies focus on the second question by investigating the typical phenomenon of contextual biases: reference repulsion. The second study shows that reference repulsion can occur in a late, decision-related stage of visual processing, where explicit and implicit processes might differ. The third study demonstrates the repulsion effects of hue reference in color vision, along with striking non-uniformities of the effects across colors. Both studies explain the repulsion biases with an encoding-decoding model, suggesting different visual features might share context-dependent reweighting of sensory representations. Overall, this thesis provides new evidence of contextual biases for various visual features and offers insights into the visual processing of contextual information.

• Item Type: Theses (Dissertation, LMU Munich)
• Keywords: visual perception, visual information processing, contextual biases, perceptual biases, hue perception, orientation perception, Bayesian perception
• Subjects: 500 Natural sciences and mathematics; 500 Natural sciences and mathematics > 570 Life sciences
• Faculties: Graduate School of Systemic Neurosciences (GSN)
• Language: English
• Date of oral examination: 21. March 2024
• 1. Referee: Wachtler-Kulla, Thomas
• Signature of the printed copy: 0001/UMC 30502