The Constructive Nature of Color Vision and Its Neural Basis

Abstract

Our visual world is made up of colored surfaces. The color of a surface is physically determined by its reflectance, i.e., how much energy it reflects as a function of wavelength. Reflected light, however, provides only ambiguous information about the color of a surface as it depends on the spectral properties of both the surface and the illumination. Despite the confounding effects of illumination on the reflected light, the visual system is remarkably good at inferring the reflectance of a surface, enabling observers to perceive surface colors as stable across illumination changes. This capacity of the visual system is called color constancy and it highlights that color vision is a constructive process. The research presented here investigates the neural basis of some of the most relevant aspects of the constructive nature of human color vision using machine learning algorithms and functional neuroimaging. The experiments demonstrate that color-related prior knowledge influences neural signals already in the earliest area of visual processing in the cortex, area V1, whereas in object imagery, perceived color shared neural representations with the color of the imagined objects in human V4. A direct test for illumination-invariant surface color representation showed that neural coding in V1 as well as a region anterior to human V4 was robust against illumination changes. In sum, the present research shows how different aspects of the constructive nature of color vision can be mapped to different regions in the ventral visual pathway.