The eyes are responsible for transducing light, but the brain must interpret the retinal images before a person can see the properties of the objects around them. To do this the brain extracts signals (such as motion, color, and binocular disparities) from the retinal images and then uses these signals as ''cues'' to construct visual percepts that accurately represent the local environment. For example, binocular disparities are signals that are used by the brain as a cue for depth, which can be demonstrated by looking at a stereogram. A question of long standing interest is how the brain knows which cues to use during the construction of a given percept. In other words, how does the brain learn to use cues appropriately during perception?
With support of the National Science Foundation, Dr. Backus and his colleagues are conducting experiments to clarify an important aspect of this problem, namely, how the brain decides to start utilizing a new cue during perception. Recent work in the Backus lab confirmed that under certain conditions the brain''s visual system can be trained to use new cues. This training was achieved by means of classical (Pavlovian) conditioning procedures in which a new signal (such as a motion direction) was paired with depth cues that were already trusted by the brain (such as binocular disparity). The experiments will use simulated 3D stimuli to measure the rate at which new cues are learned under a variety of conditions. In addition to advancing our basic understanding of perceptual learning, this work may lead to the development of new techniques for training human perception and to better computer vision systems that can improve themselves through learning.
