bioRxiv

Cognitive-Perceptual Style on the Autism-Schizotypy Continuum Shapes Surprisal Sensitivity

Preprint: cross-sectional study of 340 peoplePopulations
Follow

Abstract

Abstract How the brain weighs prior knowledge against incoming evidence varies systematically across individuals – and this variation may lie at the heart of an Autism-Schizotypy Continuum (ASC) in the healthy population. In a large lifespan sample (N = 340; age 18–82), we used the Schizotypal Personality Questionnaire and a refined Autism Spectrum Quotient model to position individuals along this continuum, then asked how their placement predicts sensitivity to lexical surprisal during self-paced reading. First, perceptual style is not fixed: older adults showed reliably less schizotypal profiles, making the ASC a developmentally dynamic dimension. Second, ASC position interacted with age to modulate surprisal sensitivity across 160,000+ word-level reading times: autism-like profiles yielded stronger disruption by unexpected words, an effect that grew across the lifespan. We interpret this as reflecting a tightening of predictions formed from linguistic context – the less schizotypy-like the profile, and the older the reader, the more strongly incoming evidence is weighted against expectation. These effects remained robust under cross-validation. Our findings establish the Autism-Schizotypy Continuum as a dynamic, lifespan-sensitive framework for understanding how individuals differ in their responsiveness to incoming linguistic evidence. Significance Statement Individual differences in perception are often framed as a trade-off between prior knowledge and sensory input, yet how these differences evolve with age and shape real-time language processing remains unclear. In a large adult lifespan sample, we show that cognitive-perceptual style shifts systematically with age and predicts sensitivity to lexical surprisal during reading. Our findings recast the Autism-Schizotypy Continuum as indexing sensitivity to incoming evidence and highlight the need for lifespan-sensitive, individualized models of predictive processing.