Story of Psychology - Morton Hunt [306]
Julesz had barely begun vision research when he came up with the idea that made him instantly famous in psychological circles. He had been surprised to find, in reading about stereoscopic depth perception, general acceptance of stereopsis as the result of the brain’s matching cues to form and depth in each eye’s image. This was thought to lead to fusion of the images and depth perception. Julesz, who had had some experience in Hungary as a radar engineer, felt sure that this was wrong.
After all, in order to break camouflage in aerial reconnaissance, one would view aerial images (taken from two somewhat different positions) through a stereoscope, and the camouflaged target would jump out in vivid depth. Of course, in real life, there is no ideal camouflage, and after a stereoscopic viewing one can detect with a single eye a few faint cues that might discriminate a target from its surroundings. So I used one of the first big computers, an IBM704 that had just arrived at Bell Labs, to create ideally camouflaged stereoscopic images.65
These consisted of randomly created patterns of black and white dots, as in this pair:
FIGURE 34
When these patterns are stereoscopically merged, the center floats upward.
There are no cues to depth in these two patterns when each is looked at alone. But although they are largely identical, a small square area in the center has been slightly shifted to one side by the computer so that when each image is seen by one eye and the patterns merged, that area produces a binocular disparity—and seems to float above the rest of the background. (To see this remarkable effect, hold a 4″×6″ card or a sheet of paper vertically in front of and perpendicular to the page so that each eye sees only one image. Focus on one corner of the pattern, and in a little while the two images will migrate toward each other and fuse. At that point the center square will appear to hover an inch or so above the page.)
The random-dot stereogram is far more than an amusing trick. It proves that stereoscopic vision does not depend on cues in each retinal image to create the experience of three-dimensionality, and that, on the contrary, the brain fuses the meaningless images and thereby reveals the hidden cues to three-dimensionality. This is not a cognitive process, not a matter of learning to interpret cues to depth, but an innate neurological process taking place in a particular layer of the visual cortex. That is where a highly organized mass of interacting cells performs a correlation of the dots in the patterns, yielding fusion and the perception of the three-dimensional effect.66 (Stereopsis is not the only way we achieve depth perception. Julesz’s work does not rule out others, including those which involve learning.)
Julesz is proud that his discovery led Hubel and Wiesel and others to turn their attention from form perception to the investigation of binocular vision, but modestly adds:
I never regarded my role of introducing random-dot stereograms into psychology as a great intellectual achievement, despite its many consequences for brain research. It was just a lucky coincidence, a clash between two cultures, an association between two foreign languages (that of the psychologist and the engineer) in the head of a bilingual.67
Yet another theory about depth perception was proposed several decades ago—one that was neither specifically neural nor specifically cognitive. Not that its proponent tactfully combined the two; on the contrary, he virtually ignored the neural theory and dismissed the cognitive theories as unnecessary and based on wrong assumptions.
Only a thoroughgoing maverick would reject a century’s worth of depth-perception research and claim to have found a totally different and correct approach. Only a true nonconformist would assert that we perceive depth neither by neural detection nor inference from cues but “directly” and automatically. Only a brash individualist would present a radical epistemology in which the physics of light is said to give us an accurate, literal