Aberrant predictive learning along the positive schizotypy - autistic traits continuum: evidence from ambiguous social information processing

Abstract

Deficits in social information processing have been observed in both schizophrenia spectrum disorders (SSD) and autism spectrum disorder (ASD), though the underlying mechanisms may differ. From a predictive coding perspective, such deficits are thought to arise from an overreliance on prior expectations in SSD, whereas individuals with ASD may exhibit difficulties in forming or using such expectations. However, very few studies have investigated the behavioral markers underlying social predictive learning along the ASD-SSD continuum. Using a novel cue–outcome associative learning task, this study examined how prior expectations influence the perception of ambiguous social information. A total of 121 healthy participants (aged 17–25) completed the task, along with self-report measures of positive schizotypal (Schizotypal Personality Questionnaire, SPQ; Community Assessment of Psychic Experiences, CAPE) and autistic traits (Autism Spectrum Quotient, AQ). Computational modeling using the Hierarchical Gaussian Filter (HGF) and correlational analyses revealed that higher levels of positive schizotypal traits were associated with greater reliance on prior beliefs and reduced flexibility in updating prediction errors. In contrast, expected inverse associations for autistic traits were not consistently observed. These results support the hyper-prior hypothesis of schizophrenia and highlight aberrant predictive mechanisms in positive schizotypy. The predictive coding framework might be useful for differentiating between SSD- and ASD-related social cognitive difficulties, with implications for targeted intervention strategies.