Distinct Biases Shape Perceptual Inference Dynamics Along the Autism-Psychosis Spectra

Predictive processing suggests that variations in precision-weighted perceptual inference can track individuals across the autism- and psychosis spectra. Psychosis is linked to a relative under-weighting of low-level priors, while autism may arise from a tendency to overestimate environmental volatility. This study empirically test computational accounts of the autism- and psychosis spectra.In an online experiment (N=150), we recorded participants’ orientation judgements of sequential Gabor stimuli alongside their autism- and psychosis-like traits. Perceptual decision-making was modelled as a Bayesian time-series, with trial-wise responses resulting from a mixture of current- and prior information on short- and long timescales. Through a series of mechanistic models, we explored how sub-clinical scores modulated the usage of priors from different timescales. In model comparisons, the winning model suggests a reduced weighting of stable stimulus features with increasing autism-like traits. Furthermore, heightened psychosis proneness was associated with a more variable relationship between stimuli and percepts.Our results align with computational accounts of autism and psychosis as disorders of precision with idiosyncratic tuning tendencies. We advocate for an integrated view on perceptual disturbances in mental disorders, considering the continuous expression of autism and psychosis.