A Systematic Review and Meta-analysis of Empirical Evidence for the Simple Bayesian Model of Autism.

The Bayesian framework conceptualizes human perception as a process of probabilistic inference, where the brain integrates prior expectations with incoming sensory evidence to construct a mental model of the world. Within this framework, several distinct theories-collectively termed the "simple Bayesian model"-suggest that perceptual atypicalities in autism stem from an imbalance between the precision of prior beliefs and sensory input. This study presents a systematic review and the first meta-analysis to evaluate empirical evidence for the simple Bayesian model. We synthesized 24 effect sizes from 23 eligible studies using a random-effects model to test its core predictions: that autistic individuals exhibit universally "broader" priors and/or heightened sensory precision compared to non-autistic controls. We found a significant, small-to-moderate overall effect in the predicted direction (Hedge's g = 0.37). However, heterogeneity across studies was large and significant and was not explained by any of the examined moderators: prior type (structural vs. contextual), stimulus type (social vs. nonsocial), task setting (implicit vs. explicit), cognitive domain (higher-level cognition vs. perception), or participant characteristics. Given the significant unexplained heterogeneity, our findings offer only limited support for a universal "simple Bayesian model" of autism. We conclude that future research should move beyond the simple Bayesian model to investigate more sophisticated, hierarchical Bayesian accounts of autism.