The Role of Top-Down Processes in Three-Dimensional Shape from Shading

Abstract

The visual system relies on several depth cues to create three-dimensional (3D) perception from two-dimensional (2D) retinal input. One such cue is shading, whereby 3D shape can be inferred from the shading pattern of an object in a process termed ‘shape from shading’. Previous research on shape from shading has described the process as occurring in a bottomup manner, with processing for shape from shading being completed early and without the need for attention. However, previous research has largely ignored the influence of top-down processes in shape from shading, such as the role of top-down (guided) attention. This is crucial, as recent research not only implicates higher-level cortical areas as being involved in shape from shading, but current theories of shape from shading have begun to discuss a potential role of top-down attention in the processing of explicit 3D shape from shading. Chapter 1 provides a review of previous research on shape from shading, with a focus on contrasting early research, which argues that processing for shape from shading occurs early and automatically, with more recent research which suggests that the extraction of explicit 3D shape requires effortful processing that occurs after early automatic processing. Chapter 2 explores the notion that top-down attention is required for later neural processing related to explicit 3D shape from shading. We asked participants to either passively view or attend to shaded objects which could appear as either convex, concave or 2D, whilst we recorded event-related potentials as an electrophysiological measure of the time-course of neural processing related to shape from shading. We found early processing related to shape from shading regardless of attentional condition, although this activity became right lateralised when participants attended to the stimuli. Critically, we found that later processing for shape from shading was reliant on the deployment of top-down attention. In Chapter 3, we measure the N2pc component, a known marker of visual attention, to further examine later neural processing that is dependent on top-down attention to identify 3D shape from shading. Interestingly, neural activity for this effect was found only over the left hemisphere, supporting claims that the ventral visual pathway in the left hemisphere is specialised in processing 3D shape. Chapter 4 moves to an investigation of whether the assumption that light comes from above an object (the ‘light from above prior’) provides a top-down influence on the perception of shape from shading even whilst conflicting information is present regarding the position of light sources in a visual scene. We find that observers judged 3D shaded objects as if light appeared from above, even whilst a light source was placed underneath the object. However, observers were able to adapt to horizontal illumination directions, supporting the notion that the left bias in the assumed light source direction is adaptable to environmental cues. Despite this, short-term associations between an object and an explicit light source position had no influence on participant judgments when the light source was removed. A conclusion to the thesis is provided in Chapter 5, which highlights the value of our findings in understanding how the visual system extracts 3D shape from shading cues. Specifically, we argue that early processing for ‘depth from shading’ is automatic, whilst later processing for shape from shading requires top-down attention. We also consider differences in how innate and learnt assumptions about the world guide our perception of shape from shading, and our findings are placed in the wider context of 3D perception in real-world settings.

Date of Award	30 Jan 2026
Original language	English
Supervisor	Ayelet Sapir (Supervisor), Debbie Mills (Supervisor) & Paloma Marí-Beffa (Supervisor)