Foundations of Cyclopean Perception (1971) Bela Julesz, University of Chicago Press

Chapter 2 – Foundations of Cyclopean Perception

2.2 Separation of Peripheral and Central Processes

We still lack a physiological psychology of human visual perception. One of the greatest obstacles to its attainment is our inability even to guess the neural levels where certain perceptual phenomena might occur. Any hint that might elucidate the peripheral or cortical origin of a psychological phenomenon has great heuristic value. After all, there are at least four synaptic junctions in the visual pathways from the receptors to the first stage of the visual cortex, and at each synaptic level each neuron can be connected to thousands of others. If it were possible to bypass three, four, six, or even more synaptic levels and portray and display the desired information at some cortical level, many neurophysiological and psychological problems could be simplified.

Page 15

At this point the reader might ask an important question about cyclopean stimulation. In the examples mentioned above, no single stimulus contained the global information. This information was provided only by the correlation between two (or more) stimuli. On the other hand, one might regard the single stimulus that physically contains the desired information, but in a form that is beyond the extraction capabilities of the peripheral processes, as a cyclopean stimulus. What, then, is our criterion for a cyclopean stimulus? Is it enough to define it by saying that the peripheral system cannot cope with the desired information, or do we insist that this information should not even physically exist at peripheral levels?

Page 17

2.4 Monocular, Binocular, and Cyclopean Stimulation

It must be emphasized that cyclopean stimulation by random-dot stereograms is neither monocular nor binocular stimulation. Since real-life objects appear to be the same whether viewed monocularly or binocularly, cyclopean stimulation is a new condition in which the monocular and binocular information is separated. The only condition under which cyclopean stimulation is approximated by nature is in animal camouflage. Here the animal develops a pigmentation that blends with the texture of the background. Very often monocularly camouflaged are seen as separated in depth when viewed stereoscopically.

At this point let us briefly review some facts about binocular vision. If the left and right retinal images are identical or similar in the sense that they are the projections of a three-dimensional object from slightly different views, the fusion of the two images into a single binocular view will inevitably take place. If the disparity between the retinal projections increases, as in the case for very near objects, the images are seen as double images, but still in depth. This is the range of patent stereopsis. For random-dot stereograms patent stereopsis probably does not exist. Certainly prior to fusion no double image can occur, since before fusion there is no global image. After fusion one can try to increase disparity by pulling retinally stabilized images apart…

If the images on the retinae are not similar or if the disparity increases above the limit of patent stereopsis, binocular rivalry is experienced. Under this condition parts of one image dominate parts of the other image, and they alternate dominance in time.

Page 23

Even the Zöllner-anorthoscopic-illusion may not be cyclopean, it raises an important point: the visual system may assemble a set of partial views that fall on the same retinal area over a period of time into a single percept.

Page 25  

Chapter 5 – Binocular Depth Perception

4. Random-dot stereograms do not contain monocular familiarity depth cues. In the absence of these complex cues binocular depth perception can be studied in its purest form.

Page 142

Whether matter and energy behave like waves or particles depends on the experimental conditions used. Nowadays no one would ask which of the two structures is “true”; it is accepted that both are and are different “projections” of a hyperstructure. It is time that psychologists also realize that the most complex organ in the known universe – the brain – should show different structures as a function of the input stimulation.

Page 143

During the course of evolution, binocular depth perception emerged from binocular vision after some favourable bodily changes occurred. Originally binocular vision served to provide panoramic vision by evaluating the two separate views that were cast on the retinae, which point sideways. The development of head movements with the accompanying complex movements of eye-coordination gave rise to physiological mechanisms that could convey certain depth information. Such a mechanism is the accommodation of the eye-lenses., and according to Walls (1942) it evolved forst and most universally among the animal kingdom, yet failed to develop into a depth sensing organ. Indeed, accommodation is a very insensitive depth cue, since the evaluation of blur is dependant on the contour properties of the stimulus; furthermore, in bright daylight the reduced pupil diameter results in increased depth of field. On the other hand, monocular parallax as a result of head movements led to the development of a very sensitive depth mechanism. Two successive images on one eye are two projections of an object from different positions. For monocular parallax the displacement between two successive eye positions is a vector, and the exact head and eye positions have to be known to evaluate depth. For animals that capture food with their heads monocular parallax led to satisfactory depth evaluation.

Page 143-144

The recession of a protruding snout or beak in some animals who capture food by paws, hands, or pouncing, with the accompanying complex vergence movements of the eyes permitted a coarse registration between the two monocular fields. A finer registration between overlapping areas of correspondence was achieved by a neural mechanism that evolved and yielded a new sensation of stereoscopic depth.

Indeed, with stereopsis, spatial localization of objects is perceived vividly as an independent sensation (similar to the sensation of color and brightness), and as such, helps to form an internal model of the outside world.  

Although there are several other depth cues (both monocular and binocular), stereopsis is the strongest depth cue. The monocular depth cues (as we will discuss later) require very complex image processing and can only be utilized under favourable circumstances. The binocular depth cues besides stereopsis are vergence and correlative accommodation (differential focusing of the two eyes).

Stereopsis is based on the geometrical fact that two-dimensional projections of a three-dimensional object on the left and right retinae differ in their horizontal positions. This horizontal shift between corresponding points in the two retinal images is called retinal disparity, or simply disparity. The terms binocular parallax or parallax shift are also used. Wheatstone’s discovery of the stereoscope (1838) first demonstrated that disparity gives rise to stereopsis. That disparity alone is adequate for stereopsis was demonstrated by Dove (1841), using exposures too brief for any convergence movement of the eyes to be initiated.

Stereopsis itself gives us the experience of relative depth only. It enables us to rank-order the nearness-farness of objects within a region of space around a fixation point. If we fixate on some point in space – that is, its retinal projections coincide with the centers of the foveae – then the projections from all features lying on the circle through this point will, by definition, have zero horizontal disparity. This is called the Vieth-Müller circle. Projections from a point lying inside this circle will have what is termed convergent (or uncrossed) disparity.

Page 144

 

In order to obtain retinal disparity, however, the stereopsis mechanism has to establish correspondence with those parts of the images that are projections of the same object on the left and right retinae. Strangely enough this basic problem of stereopsis was completely ignored until the introduction of random-dot stereograms.

The demonstration that binocular disparity alone can give powerful depth sensations re-establishes the importance of stereopsis as an important depth cue.

Even if binocular disparity were a weak depth cue, it is a powerful asset. Since time immemorial animals developed camouflage and easily blended with the environmental background. It is possible to hide rather successfully this way when the predator has only monocular vision. In order to see a camouflaged moth on a tree branch requires very complex form recognition capabilities. However, as random-dot stereopsis demonstrates, even under monocular camouflage, the hidden objects jump out in depth when stereoscopically fused. What is more, this object separation does not necessitate any familiarity with the stimulus and therefore stereopsis could evolve at a relatively early stage in the processing chain of vision. 

Knowledge of the anatomical and physiological bases for stereopsis goes back to work by Descartes, but it was Newton who first proposed a partial decussation of the optic nerve fibres. The optic tract of one side receives information from the same half-fields of both eyes. This decussation of the optic tract occurs at the chiasma, and the corresponding neurones from the two retinae project to two nearby layers in the LGN. There is apparently little interaction in the LGN between the two visual pathways (Bishop, et al. 1959), whereas the majority of neurones in the striate cortex can be binocularly driven (Hubel and Wiesel 1962).

There is still no physiological evidence for the next hierarchical level, which resolves ambiguities for complex stereograms and will be termed global stereopsis. Evidence for global stereopsis comes from psychological experiments with complex stereograms; the next step in physiology is the elucidation of this mechanism. Probably this global stereopsis mechanism will be found only in visually sophisticated animals and certainly in the monkey.

Page 146

…the psychological fact that we can fixate at a point lying on the midline and perceive points closer or farther along the midline in depth indicates some neural interaction between areas in the two retinae that fall on opposite half-fields. This observation, although suggestive…

page 147