Noise-induced alternations in attractor models of perceptual bistability

Ruben Moreno-Bote

New York University, Center for Neural Science, New York, USA

When a stimulus supports two distinct interpretations, perception alternates in an irregular manner between them. I will present demos of perceptually bistable visual displays and review salient statistical properties of their alternation dynamics. As we watch these ambiguous stimuli, one interesting question arises: what causes the perceptual switches? Most previous modeling work assumes that switches are due to firing rate adaptation or synaptic depression, while noise plays a secondary role. I develop a spiking network with attractor dynamics in which alternations are induced by noise, and are absent without it (Moreno-Bote et al., J Neurophysiol 98: 1125-1139, 2007.) The model requires a precise balance between noise and adaptation in order to produce realistic distributions of dominance durations. This balance is shown to be a general property that does not depend on the details of the neuronal architecture. The model accounts for salient properties of bistable perceptual phenomena, most notably Levelt¢s propositions. I will finally describe statistical properties of perceptual bistability recently discovered in our lab (maximum alternation rate at equidominance and the scalar property of duration distributions) that can also be explained by the model.

Back