Item

Abstract

INTRO Human ventral occipitotemporal (VOT) cortex is critical for normal color perception. Damage in this region causes disruption of color appearance. Also, superimposing an isoluminant chromatic pattern on an achromatic pattern produces strong responses in VOT (PET and fMRI). Human color responses in VOT have often been compared with single-unit measurements in macaque despite large differences in the experimental protocols used in the two species. Here we report measurements of color responses to the same stimuli in both species. Our experiments are designed to clarify the nature of color computations in human VOT and evaluate homologies between the cortical color responses in the two species.
METHODS Stimuli were 12x12 checkerboards spanning 20º of visual angle. Individual patch chromaticities were drawn at random from one of three distributions in cone contrast space:a The L+M+S axis (achromatic luminance contrast),b The L-M axis (equiluminant color contrast) or c The plane defined by the axes in a and b. Maximum luminance and color contrasts were set at 6 and 1.5 respectively to equalize activation in V1. Block-design fMRI scans were of two types: 'C' (a color exchange) and 'L' (a novel luminance exchange). In 'C', blocks of stimuli drawn from condition a alternated with blocks from condition c. In 'L', stimuli from b alternated with c. BOLD signal modulations were localized with respect to retinotopic areas.
RESULTS In humans, both color ('C') and luminance ('L') exchanges produced powerful activations in VOT. Responses to 'L' extended onto the dorsal surface, but responses to 'C' did not. In macaque, the largest 'C' activations were in V2 and V4.
CONCLUSION No regions responded uniquely to the color exchange ('C').We present a model of these color responses in human and macaque that begins with the V1 opponent color response and is followed by a non-linearity in or prior to VOT.