Allan Dobbins

Binocular Vision and Ocular Dominance Columns: A Functional Theory

No theory has emerged for cortical columns that satisfactorily accounts for the bewildering array of data in different sensory systems and species (Horton & Adams, 2005). A particularly vexing case involves ocular dominance columns, which are found in carnivora and primates, but not all primates. Yet primates without ocular dominance columns appear to have normal stereopsis. This seems distinctly problematic for the view that ocular dominance columns have a special functional role. Here we develop the proposal that ocular dominance columns do endow their possessors with unique capabilities. In particular, ocular dominance columns provide the substrate for two functionally distinct classes of binocular neurons: correlation cells and anti-correlation cells (roughly corresponding to Tuned and Near/Far cells respectively in the terminology of G. Poggio and colleagues). Correlation cells produce nonlinear binocular facilitation that depends on precise phase/position differences in the two eyes. In contrast, anti-correlation cells receive excitation through one eye and disparity-dependent inhibition via the envelope of a set of inputs through the other eye.
Anti-correlation cells conserve eye-of-origin information and this enables them to represent the asymmetries in differential binocular occlusion or da Vinci stereopsis. Just as significantly, anti-correlation cells provide a graded binocular inhibition that generates the Near/Far error signals for vergence eye movements. (Predictions arising from this shall be discussed in some detail.) We argue that these ocularly asymmetric, anti-correlation cells are the
primary purpose of the ocular dominance architecture, underlying the operation of a primitive binocular system that permits the experience-dependent bootstrapping of the fine correlation-based system. A prediction of this theory is that primates that don’t have ocular dominance columns lack a complement of anti-correlation neurons and hence depend on other behaviors/cues to develop their correlation-based stereo system. Consistent with this idea is the observation that juvenile monkeys of species without ocular dominance columns are known to frequently make a variety of head movements, apparently employing motion parallax in the absence of the primitive binocular segmentation system.