Cortical Disinhibition, Attractor Dynamics, and Belief Updating in Schizophrenia

Abstract

Genetic and pharmacological evidence implicates N-methyl-D-aspartate receptor (NMDAR) dysfunction in the pathophysiology of schizophrenia. Dysfunction of this key receptor - if localised to inhibitory interneurons - could cause a net disinhibition of cortex and increase in `noise'. These effects can be computationally modelled in a variety of ways: by reducing the precision in Bayesian models of behaviour, by estimating neuronal excitability changes in schizophrenia from evoked responses, or - as described in detail here - by modelling abnormal belief updating in a probabilistic inference task. Features of belief updating in schizophrenia include greater updating to unexpected evidence, lower updating to consistent evidence, and greater stochasticity in responding. All of these features can be explained by a loss of stability of `attractor states' in cortex and the representations they encode. Indeed, a hierarchical Bayesian model of belief updating indicates that subjects with schizophrenia have a consistently increased `belief instability' parameter. This instability could be a direct result of cortical disinhibition: this hypothesis should be explored in future studies.