Learning-like phenomena in stereopsis

Abstract

WHEN confronted with Julesz random-dot stereograms¹ for the first time, naive observers often need as long as a minute or more to see a global figure in depth. With repeated exposure to the stereograms this perception time is progressively reduced until they begin to see depth almost instantaneously². This acceleration of the stereoscopic process over successive trials is a remarkable example of perceptual learning^2,3, and unlike many other kinds of perceptual learning it seems to be a purely visual effect. Further, stereopsis is a basic visual function for which much of the processing may occur as early as the prestriate cortex⁴. It is important to note that learning-like changes can affect such a process, and knowledge of elementary visual function may be helpful in determining just what is learned. Here I explore the position specificity of stereo-learning, and find that in most observers a significant proportion of the learning that is acquired in one retinal area does not transfer to neighbouring areas. One question raised by this finding is whether such position specificity necessarily implies localisation of the learning process to specific anatomical areas in the brain.