Experience-Dependent Perceptual Categorization in a Behaving Real-World Device

Abstract

The ability of organisms to categorize objects depends on their sensory experience in an environment. We studied the role of frequency and temporal order of stimuli in perceptual categorization on Darwin VI, a neuronal model interfaced with a behaving real-world device. The model consisted of several distinct biologically based networks, representing areas of the central nervous system. Darwin VI was trained to perform a conditioning task, using objects differing in their visual patterns as well as taste. Initially, behaviors were triggered by taste. After conditioning, responses were trig gered by pattern vision, and the simulated visual system showed specific and invariant activity patterns for different categories of visual stimuli. We found that visual system activity devoted to an infrequently sampled stimulus class, during early exploration, resulted in a reduction of activity that remained even if the stimulus was sampled at normal rates later. In contrast, frequently sampled stimulus classes resulted in more visual system activity than normal. Presenting exemplars of only two out of three stimulus classes to a conditioned Darwin VI resulted in a significant loss of visual system neurons capable of responding to the deprived stimulus. These studies point to the importance of early sensory experience for the development of perceptual categories.