Abstract
An isolated patch of light has a hue closely linked to its chromaticity but a patch does not invoke some fundamental processes that contribute to color perception in natural viewing. A key difference between light seen in isolation versus within a natural scene is that only the latter has spatial structure (i.e., distinct regions of different chromaticities). That spatial structure reveals basic neural processes of color perception. First, a surrounding pattern composed of two different chromaticities can cause a profound induced color shift (and a striking illusion), even though a uniform surround at either chromaticity in the pattern induces only a small and unimpressive shift. This can be accounted for by cortical receptive-field organization. Second, spatial structure raises the question of binding color to form (the binding problem is trivial with a single isolated patch because there is only one form to which a color representation can be bound). Obvious and sustained misbinding of color to form is perceived with binocular rivalry between two equiluminant gratings of different colors and orientations (e.g., a vertical red/white grating to one eye and a horizontal blue/white grating to the other eye). The rivalrous forms alternate (horizontal or vertical) but the grating often contains both colors (thus a horizontal or vertical red/blue grating, another remarkable illusion). Though the form in one eye is suppressed, the color from that eye is neither suppressed nor mixed with the color from the other eye. Instead, the chromatic neural representation from the form-suppressed eye is expressed in a non-retinotopic location of the dominant form. This implies color and form are represented separately in visual cortex, and that perception of colored objects depends on a neural binding process; hue and form are not inseparably represented, as sometimes proposed, by neurons jointly selective for chromaticity, spatial frequency and orientation.
Supported by NIH grant EY-04802.