BMBF Bernstein Group "Components of Cognition: from small Networks to Flexible Rules" Bistable perception: neural representations at the bifurcation

Multi-stable perception is not the "memoryless" process as which it was long regarded. Recent psychophysical studies reveal both stabilizing and destabilizing effects of perceptual history: the recent experience of a particular percept makes it both more likely (in the short run) and less likely (in the long run) to experience the same percept again. The destabilizing effect presumably reflects some form of adaptation (either neuronal or synaptic). Surprisingly, the destabilizing effect does not seem to actually cause perceptual reversals. Rather, reversals appear to be noise-driven. Cumulative measures of the destabilizing effect reveal only a limited control over reversals. When the cumulative measures of competing percepts are balanced, perception enters transitional states in which neither percept dominates. A competitive network of biophysically realistic neurons replicates this behaviour only when it resides just next to the bifurcation between an adaptation driven regime (limit cycle) and a noise driven regime (metastability). There are a number of ecological and functional reasons why perceptual representations should be established at this bifurcation point.