Abstract
Even though color vision is commonly defined as the ability of an organism to distinguish objects based on the wavelengths of light they reflect, color research has mainly focused on the representation of colors independent of the problem of distinguishing objects. Thus the mechanism by which color contributes to object recognition remains unclear, as little is known about how color and object information are co-represented in the part of the brain responsible for object recognition: in primates, inferotemporal (IT) cortex. The recent discovery of "color patches" in macaque IT cortex makes this problem experimentally tractable. Here we recorded neurons in three color patches, middle color patch CLC, and two anterior color patches ALC and AMC, while presenting images of objects systematically varied in hue. We found that all three patches contain high concentrations of hue-selective cells, and carry significant information about both hue and object identity. We found two clear transformations across the three patches. The first transformation, from CLC to ALC, reduces information about object identity. The second transformation, from ALC to AMC, mainly affects representation of hue: color space is represented in a dramatically distorted way in AMC, with over-representation of yellow and red, the natural colors of mammal faces and bodies; furthermore, AMC develops an expanded representation of primate faces, displaying hue-invariant representation of monkey identity. Our findings suggest that IT cortex uses three distinct computational strategies to represent colored objects: multiplexing hue and object shape across all objects (CLC), extracting hue largely invariant to shape (ALC and AMC), and multiplexing hue and object shape specifically for ecologically important objects (AMC). Overall, our study reveals the neural architecture for representing colored objects in IT cortex, and sheds light on the general organizational principles of IT cortex.
Meeting abstract presented at VSS 2017