Abstract
We used a noise masking paradigm to examine the interaction between color-selective symmetry detection mechanisms in the visual system. We used a 2AFC paradigm in which a random dot noise mask was presented in both intervals. One interval contained a target, while the other, a random dot control. The target consisted of a red and a green symmetric pattern with the same (both were 45° or −45°) or orthogonal (one 45°and the other −45°) orientation. The observers were to determine which interval contained the symmetric target. We measured the target density threshold at various noise densities. For all conditions, the target density threshold increased with noise density with a slope 0.96 on log-log coordinates. The threshold for the same-orientation condition was lower than that for the orthogonal condition at all noise densities. We fit our data with a divisive inhibition model for symmetry pattern detection (Chen & Tyler, 2010 PLOS One), in which the response of a symmetry detector is the excitation of a linear symmetry operator raised to a power and then divided by the divisive inhibition from all relevant symmetry operators. The best fit showed that the mutual inhibition between symmetry detectors in the same-orientation condition was only 13% of that in the orthogonal condition. Hence, instead of a strong same-orientation inhibition commonly observed in experiments using Gabor patches, it is actually easier for the visual system to integrate symmetric patterns of the same symmetric axis.