At some early stage in the visual system the retinal images coming from the right and left eyes are combined into what Julesz has called the 'cyclopean' image CI, and the two separate images from which CI is formed are left side or repressed if not fully or partially absorbed into CI. But as what stage of the visual process does this inter-ocular combination occur? Does it precede or follow the formation of all early, or of some early, or of some late, percepts? This is related to the question of whether the disappearance of the individual monocular percepts when a binocular image has been formed involves their active repression by some competing binocular percept, or is a consequence of their never having been formed at all, since the cylopean image is formed first and becomes the sole basis for formation of all other percepts? These questions will be explored in the present paper.
In our first figure, Cafe Wall images of opposite slopes have been embedded in separate anaglyphic planes. Viewed normally the intensities sum and an upward slope is seen. Viewed anaglyphically through the red (resp. blue) filter alone, a downward (resp. upward) slope is seen. When the figure is viewed anaglyphically using both eyes, one sees an image that at first seems chaotic but then resolves itself into a series of alternating horizontal bands, consisting of yellow and black bands in a neutral depth plane, with black and white bands between them, either strongly forward of the yellow and black bands, or far behind them (this percept is bistable). These bands tend to slope mildly upward, but may slope momentarily downward when viewed immediately after a brief glimpse of the downward-sloping monocular red-filtered image. (Blink the eye behind the blue filter). In the anaglyphic view the red bricks (in the red-cyan bands) appear black and the cyan brings appear more white than cyan.
Figure 1. Progressive Cafe Wall figures of opposite slope embedded in separate anaglyphic planes
The behavior of these four percepts cn be explained as follows. The normal view and each of the filtered monocular views contain ordinary progressive glass patterns and behave accordingly. The image seen in the normal view reflects the fact that the cyan bricks are considerably brighter than the red. The binocular anaglyphic percept is shaped by the strong preference of the binocular system to combine dark regions with dark and light with light when possible. This causes the red-filtered image of the cyan bricks (seen as black) to fuse with the blue-filtered image of the red bricks (also seen as black), by a one-brick shift either to the left or right. For this reason, these rows appear in the anaglyphic view as either strongly raised in 3-D or strongly lowered. Since the evidence for both of these percepts is equally strong, either of these two percept can emerge, in a bistable way. Independently of whether binocular fusion of the red-cyan rows is accomplished by a left or a right shift, the cylopean dark bricks seem prone to be seen in the red-brick position, which forms a cyclopean 'progressive Cafe' pattern, and causes a cyclopean upward-slant-to-the right-percept to emerge. This perception ends to dominate irrespective of whether the cyan-red rows are seen advanced or recessed in 3-D.
Figure 2. Standard Cafe Wall figures of opposing slopes embedded in separate anaglyphic planes
Figure 3. A progressive and a standard Cafe Wall figure embedded in separate anaglyphic planes
In figure 4 a large circle has been cut out of the red and green planes, leaving the illusory upward slope of the red-pane Cafe figure visible in the normal view. In the anaglyphic view, a large black circle is present to the blue-filtered eye, but the downward-sloping Cafe pattern continues normally in the red-iltered view.
Figure 4. Effect of removing an area from one of two distinct images presented binocularly
Our next figure shows the curious tendency of images seen only by a single eye to shift position. Viewed normally we see a clear phi motion, of a single bright dot on the left and of a group of four bright dots on the right.
Figure 5. Binocular phi motion
Figure 5b. Displacement of unmatched binocular image elements
Figure 6. Glass figure perception cannot be formed from dot-groups presented to separate eyes
Figure 7. Progressive phi motion cannot be formed from dot-groups presented to separate eyes
Figure 8 shows that Glass 'swirl' perception is weakened or destroyed if its separate dot groups are lifted into distinct anaglyphic planes. The animation alternates between two states, in both of which the two groups of dots forming a glass figure are represented by pairs of dots. In bothe o these sates a clear Glass swirl perceptis present if the image is viewed either normally
The perception generated by this figure argues that Glass swirl perception follows stereo perception during visual processing.
Figure 9. Horizontal and vertical line groups presented to separate eyes