O8.8. THE NEURAL DYNAMICS OF BELIEF FORMATION: IMPAIRMENTS SPECIFIC TO THE SCHIZOPHRENIA SPECTRUM AND FEATURES THAT ALIGN ON THE PSYCHOSIS CONTINUUM

Recent theories in computational psychiatry have proposed that unusual perceptual experiences and delusional beliefs in psychotic disorders may emerge as a consequence of aberrant inference and disruptions in prediction error updating. The current study investigates anomalies in belief formation and updating that are specific to the schizophrenia spectrum. Furthermore, we examine psychosis as a continuum, which extends into the non-clinical population, and investigate anomalies that align on the continuum of psychosis. We characterised the underlying neural dynamics of impaired belief formation, or ‘regularity learning’ in stable and volatile contexts. 66 participants (22 inpatients with a schizophrenia spectrum disorder (SZS), 22 non-psychotic inpatients (NP), and 22 non-clinical controls (NC)) completed an auditory oddball task with volatility manipulated. The psychosis continuum in our sample was measured using the Community Assessment of Psychic Experience Questionnaire. We recorded prediction error responses with electroencephalography and behaviorally measured regularity learning errors to inferences on the probability of sounds presented within stable and volatile contexts. Furthermore, we explored how the neural dynamics relate to psychotic experiences using Dynamic Causal Modelling. Attenuated prediction error responses were specifically observed in the SZS but not in NP group, with reductions in mismatch negativity (MMN) in stable, and P300 in volatile contexts. Attenuation in P300 was related to poorer regularity learning and psychotic experiences across the whole sample. Whole-brain source reconstruction showed decreased activity for P300 prediction errors in SZS compared to NC in the medial frontal gyrus, insula and primary auditory cortex (A1); and for SZS compared to NP in the right superior temporal gyrus (STG). Further, we explored the effective connectivity differences between these regions, which also capture the know auditory prediction error network architecture. Dynamic Causal Modelling with Parametric Empirical Bayes analysis revealed stronger left A1 intrinsic connectivity in the SZS compared to the NP group, which was also related to better regularity learning. Furthermore, people that experienced more hallucinations and general psychotic-like symptoms had decreased intrinsic connectivity in the right IFG, which was also found to be related to poorer regularity learning. The findings provide evidence that reductions in auditory prediction error signaling is specific to the schizophrenia spectrum, related to attenuation in MMN and decreased activity in superior temporal gyrus. However, the results also support the notion of the psychosis continuum, with psychotic experiences across the three groups being related to poorer regularity learning and decreases in P300, suggesting a weaker prediction model of sensory probabilities. Weaker right IFG intrinsic connectivity may underlie impaired prediction formation, in patients experiencing hallucinations, and people within the general population who also experience psychotic-like symptoms. We suggest that increased intrinsic left A1 connectivity may play a compensatory role in SZS compared to other patient groups during regularity learning.