Professor of Psychiatry and Computational Neuroscience
Charité – Universitätsmedizin Berlin
Perceptual inference is the process by which current beliefs are used to infer the probable causes of the incoming sensory signals. When these sensory signals are perceptually ambiguous, inference may result in spontaneous alterations between two or more perceptual states, a phenomenon called multistable perception. The neural mechanisms of the underlying inferential process have remained controversial. Whereas some authors argue that multistable perception is governed by local processes in sensory cortices, others have proposed a role for higher-level frontoparietal brain regions in driving perceptual inference. Here, I will propose an account of multistable perception that can reconcile these apparently contradictory views within the computational framework of predictive coding. I will also present results from computational modeling in a Bayesian framework and model-based fMRI that support the proposed account. Finally, I will outline how altered predictive coding may explain abnormal inference in psychotic states and present empirical behavioral and neuroimaging work that used multistable perception to probe the role of predictive feedback signaling in psychosis.