Understanding semantic impairments in schizophrenia from a predictive coding perspective

Schizophrenia is characterized by profound semantic impairments that manifest as disrupted language and thought. We provide empirical support for the hypothesis that predictive coding forms a unifying framework for understanding these deficits by reinforcing theoretical ideas with quantitative neuroimaging evidence. According to predictive coding theory, the brain continuously generates predictions about incoming information, and prediction errors drive model updates when expectations diverge from sensory input. This review synthesizes evidence from cognitive neuroscience, computational psychiatry, and neurolinguistics to demonstrate how aberrant prediction error signaling disrupts hierarchical semantic processing in schizophrenia. Behavioral studies have revealed atypical semantic processing in priming and fluency tasks. Electrophysiological studies have shown altered neural responses to semantic incongruence, particularly reduced N400 effects. Furthermore, we have used voxel-wise modeling, graph theory, and topological analysis to demonstrate fundamentally disorganized semantic networks in schizophrenia, characterized by reduced small-worldness, excessive homogenization, and diminished representational variability. These converging findings are consistent with a neurocomputational account wherein semantic deficits reflect disrupted predictive mechanisms. This theoretical framework suggests that miscalibrated precision weighting of prediction errors leads to either over-activation of irrelevant semantic associations or impoverished semantic processing. This perspective offers insights into schizophrenia pathophysiology and guidance for targeted interventions to restore predictive coding function.