We depart from the notion that current diagnostic criteria for schizophrenia are based on psychopathological descriptions dating back to the mid-20th century; laboratory markers are missing. We approach this issue by employing machine learning algorithms for diagnostic single-subject classification using well-established, schizophrenia-sensitive neurophysiological and neuropsychological parameters in collaboration with Dr. Florin Popescu (Fraunhofer Institute for Open Communication Systems, Berlin).

Predictive coding refers to a hierarchically organized basic cognitive model, where mismatches resulting from comparisons of fed-forward sensory data with top-down signals elicit cortical prediction errors that are then used for model adjustment. Current biocomputational models of cognition in schizophrenia increasingly incorporate cortical and subcortical prediction error deficits to explain a wide range of symptoms including delusions, hallucinations, or passivity experiences.

Important aspects of social interactions include sending and responding to verbal and non-verbal signals. The face processing and the mirror neuron systems offer distinct routes for studying social cognition via sensory and motor cognition pathways by using, e.g., within-modality repetition suppression and cross-modal adaptation, respectively. We seek to explore imitative and contextual motor cognition domains in schizophrenia in a collaboration project with Dr. Roman Liepelt (Department of Psychology, University Münster) and Dr. Christiane Montag (Department of Psychiatry, Charité Berlin).

Andres Neuhaus, MD