Disentangling the effects of Anxious, Autistic and Psychotic Traits on Perceptual Inference

The brain combines sensory information and prior information, taking into account uncertainty, to perceive the world. This inference process approaches optimality in humans but with inter-individual differences associated with psychological traits. Previous results on these differences are in fact highly heterogeneous and even contradictory. We highlight experimental, modeling, and analysis choices that may contribute to this heterogeneity. We propose a set of tasks utilizing explicit and implicit priors, combined with computational modeling, to isolate the decision and learning stages of perceptual inference. Using a multidimensional approach, we characterized differences in perceptual inference associated with anxious, autistic and psychotic traits in two large samples from the general population. Our findings reveal that anxious, autistic, and psychotic traits form three distinct, yet correlated, dimensions. More anxious traits were associated with enhanced performance and greater reliance on sensory information at the decision stage. Autistic traits were not associated with any difference in perceptual inference; results for psychotic traits were inconsistent across the two samples. Results are partly different when using unidimensional analyses. Together, these results stress the importance of a multidimensional approach that takes anxious traits into account to characterize inter-individual differences in perceptual inference.