Cite as:
Sperling G, Ding J, 2003, "A neurally based computational theory of binocular combination" Perception 32 ECVP Abstract Supplement

A neurally based computational theory of binocular combination

G Sperling, J Ding

When the left and right eyes view similar but slightly different stimuli, how are they combined to form the perceived 'cyclopean' image? To obtain the basic data for such a theory, observers viewed parallel, horizontal sine-wave gratings that differed in phase and in contrast between the two eyes. The sum of two such sines is itself a sine. The phase of the perceived sine is used to compute the relative contribution of each eye to the cyclopean image. Our neurally based gain-control model proposes that each eye not only inhibits the other eye in proportion to its own contrast energy, but also inhibits the other eye's reciprocal inhibition. The cyclopean image is the sum of the surviving signals from the two eyes. This theory has the robust (and correct) property that, when the two eyes have identical inputs, the cyclopean image is the same as when one eye alone receives input. It also makes excellent predictions of all the cases in which the images in the two eyes differ in relative phase and in contrast. It makes the prediction, which was verified, that adding random noise to the sine-wave in one eye will increase its relative contribution because the more-stimulated eye then exerts more gain control over its competitor. The gain-controlling property of added band-limited noise is used to determine the bandwidth of the competitive gain-control mechanism. The elaborated theory makes pixel-by-pixel predictions of the content of the cyclopean image, and accounts for the results of many classical experiments in binocular combination.

[Supported by Air Force Office of Scientific Research, Human Information Processing Program.]

These web-based abstracts are provided for ease of seaching and access, but certain aspects (such as as mathematics) may not appear in their optimum form. For the final published version of this abstract, please see
ECVP 2003 Abstract Supplement (complete) size: 1712 Kb