An ecological analysis of binocular vision.

Abstract

Underlying the classic binocularity problems of singleness and three-dimensionality is a theory that the stimulus for binocular vision constitutes two two-dimensional images and metric differences between parts of those images. This characterization of the stimulus is criticized here and in its stead an ecologically-based characterization is presented wherein a binocular transformational invariant is shown to specify absolute (body-scaled) size, shape, and distance. The transformation is characterized as a rotation and its specificity to distance assumes a constant interocular distance and either homogeneously textured or extended surfaces. Four experiments demonstrate perceivers' abilities to detect this information and report (verbally or by reaching) the absolute distances of surfaces in stereograms. A fifth experiment revealed that accurate performance did not depend on oculomotor information. The assumptions of texture extent and distribution and constant interocular distance and their possible violations were discussed. A sixth experiment demonstrated that violations of interocular distance are absorbed by surface shapes. The existence and detection of a binocular rotation dissolves the putative problems of singleness and stereopsis, indicates that the importance of having two frontal eyes is for perception of absolute distance, and reformulates the problems for an algorithmic (physiological) theory of vision.