Johnny: Dad, why does the elephant have such a big trunk?
Father : I haven’t the foggiest idea, son.
Johnny: And why does the giraffe have such a long neck?
Father: Only God knows, son.
Johnny: Dad, why does the lion roar so loud?
Father: I don’t know, son.
Mother: Johnny, why do you pester your tired father with so many silly questions?
Father: That’s all right, my dear, let the child ask — that’s the way he learns!
That old joke is the preface to a scientific autobiography of sorts by the brilliant experimental psychologist, Bela Julesz,and learning was one of many topics on which he was innovative, to say the least. If you recongnize the name, it’s likely because Julesz is widely credited with inventing the computer generated random dot stereogram (RDS) in 1959. Leave it to Julesz to cast Dialogues on Perception as an exchange between “A” the author, and “B” his alter ego. The reader is assumed to take part in the dialogues by agreeing with “A” or “B”, or perhaps by having a third opinion, orthogonal to the ones expressed, and deriving a novel idea. With this literary device Julesz actually shapes the book as metaphor for the RDS itself: the interplay of an “A’ and a “B” pair that give rise to a uniquely binocular third dimension.
In a prior blog post I wrote about Julesz’s other seminal book, Foundations of Cyclopean Perception. While Dialogues is decidedly more literary, it is chock full of important scientific insights. Here I want to extract what Julesz writes on p. 62 to show that the RDS itself is a form of optical illuision, or at the very least a binocular virtual reality. Julesz makes the following statement: “There are many demonstrations that illustrate that stereopsis of RDSs (to which I often refer as either ‘global stereopsis’ or ‘cyclopean perception’) must be an early process, based on some correlation-like process prior to object (form) recogntion. Here, I only take one example from my monograph (Julesz 1971) that shows that optical illusions must occur after binocular combination of information after several synaptic processing stages in the retina, the lateral geniculate nucleus, and the cortex. This technique of process localization without a scalpel, which I called ‘psychoanantomy’, enables us to trace the information flow in the visual system by portraying visual information not with the usual luminance gradients, but with binocular disparity gradients and using RDSs.”
The image above is taken from a chapter on motion and depth in Foundations of Vision by Brian Wandell. Juelsz notes that since the left and right images are devoid of all monocular cues, including shapes and contours, binocular perception of the RDS is constrained by the maximum binocular disparity falling within the limit of Panum’s fusional area. With a dynamic RDS, each subsequent frame contains uncorrelated random dots (but similar binocularly correlated areas), adn these uncorrelated frames erase or mask the previous frames. Therefore, Julesz notes, in addition to the fact that RDSs are devoid of all familiarity cues, quick masking prevents the top-down processes from penetrating consciousness in time to influence the bottom-up processing of stereopsis.
This reinforces the point that rather than the RDS being the holy grail of binocular vision, it is more suitable to think of it as one tool in the investigation of binocular function. It is a powerful tool, no doubt, but it has been hijacked to mean something that Julesz never intended, which is to say that those who lack appreciation of RDS don’t have a high level of binocular vision. Indeed, real world stereopsis and its conftribution to judgements about visual space are so much richer and more informed by familiar monocular and binocular cues that are purposely absent in Julesz RDSs. In this sense, the RDS is an impoverished illusion of real world stereoscopic vision.
We addressed this in our six part series on the dual nature of stereopsis, and Sue Barry addresses it beautifully in her book Fixing My Gaze as well as in her blogs, particuilarly when she addresses how one learns to palpate visual space through binocular viewpoints. The RDS remains a very useful tool in optometric vision therapy, and we use it regularly in our practice, though in the world of iPad Apps and exciting graphics it sometimes becomes challenging to motivate kids to hang in there to completion.
By the way, added to the annals of successful people undaunted by rejection, it’s mind boggling to realize that Julesz’s first paper on random dot stereograms was rejected when submitted to the Journal of the Optical Society of America, which is how it came to be published in the Bell Labs Technical Journal. You can learn more about this giant in binocular vision by reading the tributes to him by Papathomas and Siegel.