When experiencing an ambiguous sensory stimulus (e.g., the faces-vase image), subjects may report random alternations (time scale, seconds) between the possible interpretations. I will describe dynamical systems models for neuronal populations that compete for dominance through mutual inhibition, influenced by slow adaptation and noise. In highly idealized formulations network units are percept specific without direct representation of stimulus features. Our behavioral experiments and modeling involve perception of ambiguous auditory stimuli. The models incorporate feature specificity, tonotopically organized inputs and receptive fields, so that perceptual selectivity is emergent rather than builtin. Our model addresses the effects of selective attention, distractor and deviant sounds as well as the transient, so-called build-up, phase of sound source segregation as when entering a cocktail party.
Sponsored by the NYU-ECNU Institute of Brain and Cognitive Science at NYU Shanghai