Abstract
A central issue in color vision is the nature and number of mechanisms that process signals from the cones. Some groups, including ours, have selected stimuli and analyzed data represented in cone contrast space, and found results consistent with a limited set (e.g., 6) of (unipolar) mechanisms. Other groups, working in cone excitation space, have found evidence for a larger number of ("higher order") mechanisms. Hansen and Gegenfurtner (2010 VSS) argue that this empirical discrepancy is due to a poor choice of color angles by researchers using cone contrast space. When they used chromatic masking noise with a color angle near the corner of the detection contour in the (ͧ0;L/L, ͧ0;M/M) plane of cone contrast space, Hansen and Gegenfurtner found masking that was highly selective for that noise color direction. Their interpretation was that there are multiple color mechanisms, including some tuned to color directions near the noise direction. We have replicated Hansen and Gegenfurtner’s result, using randomly flickering colored lines (42°/222° and 48°/228° in the (ͧ0;L/L, ͧ0;M/M) plane) superimposed on 1 cpd Gabor tests. As they reported, masking was highly selective, with thresholds elevated most at the noise angle, even compared to angles only 3° away. However, we show here that selective masking can occur with no more than six linear color mechanisms when the mechanisms are not orthogonal. It is not necessary to postulate multiple mechanisms, and certainly not necessary to have mechanisms tuned to the noise direction, to account for selective masking: even very simple models can generate this apparently-complex behavior. Whether the simplest of these models can account for all the published results on chromatic detection is an open question, but a selective masking result alone is not sufficient to reject models with a limited number of mechanisms.
Meeting abstract presented at VSS 2013