Abstract
In the early stages of color processing, color opponent neurons combine the signals from the various cones linearly. Their selectivity in color space is broad, and explains behavioral thresholds for the detection of uniformly colored discs. Is selectivity sharper for targets with more complex spatial properties, that are probably detected at later stages of the visual pathways?
We measured the chromatic selectivity of mechanisms responsible for the detection of static, circular Glass patterns (CGPs). In a display with a fixed number of dot pairs, we measured the proportion of pairs contributing to the CGP necessary for its reliable detection. To isolate the early stages of CGP processing, we varied the color of the dots within a pair. Human observers' thresholds were lowest when both dots had the same color, and increased in proportion with the color difference between the dots. The selectivity of the underlying mechanisms is broad, consistent with the notion that they combine their inputs linearly. To isolate late stages of CGP processing, we used uniformly colored CGPs embedded in static noise. We varied the color of the noise relative to that of the CGP. As reported previously (Cardinal and Kiper, 2000), detection was easiest when noise and CGP had opposite colors, and hardest when they had the same color. The selectivity of the mechanisms involved in this task is also consistent with a linear combination of their inputs. We compared the selectivity of CGP mechanisms to those underlying color categorization, measured with a single hue-scaling procedure. Relative to those involved in CGP detection, the mechanisms involved in color categorization are narrowly tuned. Our results suggest that the human visual system relies primarily on broadly tuned mechanisms for the detection of colored objects.