9601287 BHARUCHA In perception, some objects or features seem more stable than others. For example, edges that are slightly oblique are perceived as aberrations of perfectly horizontal or vertical edges, as when we attempt to straighten a painting that appears to be hanging askew. Against a background of oblique lines of the same orientation, however, a horizontal or vertical line is perceived as unstable. Perceptual stability and instability are thus context dependent. Stable elements serve as cognitive reference points, and unstable elements are perceived in relation to (are perceptually anchored to) their closest stable elements. This research will explore mechanisms underlying the perceptual anchoring of auditory events in familiar auditory contexts. In music that utilizes culturally familiar patterns, for example, some tones are heard as more unstable, unfamiliar or dissonant than others. Unstable tones seem to demand resolution to (are anchored by) stable tones that are pitch neighbors. This research will develop neural net models to account for how learned auditory patterns are heard in terms of stable and unstable elements, and how unstable elements are anchored by stable neighbors. When a tone is heard, a mechanism called auditory selective attention focuses processing in a narrow band of frequency-tuned neural units around the frequency of the tone. The neural units that are most active within that band then capture auditory selective attention. Thus when an unstable tone is heard, the stable tones that are closest in frequency capture auditory selective attention, resulting in the anchoring of the unstable tone. This research will include the detailed development and computer simulation of the neural modeling outlined here. It will also include a set of perceptual experiments that test the model that will vary the frequency distance of unstable tones to their stable neighbors. The work should lead to a better understanding of the perce ption of unfamiliar events in familiar perceptual environments, with implications for the design of artificial perceptual recognition systems. ***
