Seeing the Structure of Objects

This dissertation is about our visual perception of objects and their geometrical properties. I offer an account of visual shape perception, and then apply this account in developing a theory of how vision secures reference to objects. Chapter 1 provides an overview of the issues to be addressed. Chapters 2 and 3 concern our perception of shape. Specifically, chapter 2 argues that shape perception is layered: We perceive objects as having multiple shape properties, and these properties have varying degrees of abstraction. This picture contrasts sharply with certain views of shape representation in the philosophical and psychological literature, which I label metric views. Metric views claim, roughly, that vision only explicitly represents certain metric properties of objects, such as location, length, distance, and angle. Chapter 3 argues that visual shape perception is mereologically structured: Roughly, we perceive an object’s decomposition into parts, the intrinsic shapes of its parts, and the locations of the joints between parts. I argue that this forms the basis of a type of perceptual constancy—structure constancy. Moreover, I argue that this approach embodies a radical departure from views on which the visual experience of spatial properties is wholly viewer-centered. Chapters 4 and 5 concern object perception. Chapter 4 considers the problem of how a visual representation secures reference to an external object. I argue that the two leading approaches to this problem (which I call the pure causal view and the location-based view) face serious difficulties. I then argue that part-based visual shape representation plays a crucial role in the mechanism of visual reference-fixing. Chapter 5 addresses the question of what counts as an object for visual perception. More specifically, what types of things does vision pick out and track over time? On one recently popular view, visual processes of selection and tracking are specifically tuned to a class of entities called Spelke-objects. I argue that this view is problematic, primarily because it places excessively strong constraints on the geometry and topology of visual objects. I then defend a different account on which visual objects are (roughly) those things that satisfy traditional perceptual organization criteria.Ph.D.Includes bibliographical referencesby Edwin James Green Jr