Binocular combination of contrast signals.

We studied the detectability of dichoptically presented vertical grating patterns that varied in the ratio of the contrasts presented to the two eyes. The resulting threshold data fall on a binocular summation contour well described by a power summation equation with an exponent near 2. We studied the effect of adding one-dimensional visual noise, either correlated or uncorrelated between the eyes, to the grating patterns. The addition of uncorrelated noise elevated thresholds uniformly for all interocular ratios, while correlated noise elevated thresholds for stimuli whose ratios were near 1 more than thresholds for other stimuli. We also examined the effects of monocular adaptation to a high-contrast grating on the form of the summation contour. Such adaptation elevates threshold in a manner that varies continuously with the interocular contrast ratio of the test targets, and increases the amount of binocular summation. Each of several current models can explain some of our results, but no one of them seems capable of accounting for all three sets of data. We therefore develop a new multiple-channel model, the distribution model, which postulates a family of linear binocular channels that vary in their sensitivities to the two monocular inputs. This model can account for our data and those of others concerning binocular summation, masking, adaptation and interocular transfer. We conclude that there exists a system of ocular dominance channels in the human visual system.