Abstract Visual representations express distinctively visual content. Such content takes the form of a kind of space where objects and properties are assigned locations in relation to a viewpoint. Many have conceived of visual space as a metric three-dimensional volume, analogous to physical space. Yet this assumption, I argue, over-constrains visual content, excluding the ubiquitous phenomenon of indeterminate depth perception. In this paper, I propose that visual contents are view spaces: two-dimensional directional arrays of objects, properties and relations. View spaces prioritize visual direction as a core aspect of structure, while demoting depth to a variable feature like colour or shape. This proposal accommodates depth indeterminacy while preserving distinctive visual structure, and it aligns with the use of feature maps in vision science and computer vision. I will argue that this architectural differentiation of direction and depth is supported by a wide range of evidence from philosophy and psychology.
Gabriel Greenberg (Sat,) studied this question.