A fundamental function of sensory systems is to segment sensory inputs into distinct perceptual objects (termed perceptual grouping and segmentation, PGS). Atypical PGS is a robust characteristic of autism spectrum disorder (ASD), with evidence ranging from enhanced visuospatial abilities (e.g. Wechsler’s Block Design test, visual target detection) to reduced susceptibility to illusions in geometric displays (e.g. Kanizsa’s triangles, Ebbinghaus illusion). This suggests that the typical influence of spatial and temporal context on PGS may be altered in ASD, leading to altered perception.

Classical descriptive theories, such as Weak Central Coherence and Enhanced Perceptual Function, captured some phenomenology of atypical contextual processing and PGS, but ignored that typical contextual influences often follow the rules of Bayesian inference. On the other hand, Bayesian accounts of ASD are in vogue but are often used informally, leading to conflicting results. Furthermore, research on atypical perception either uses impoverished, synthetic stimuli or lacks a well-controlled characterization of visual processes when natural stimuli are used. This makes it difficult to relate findings on atypical perception to real-life perceptual experience.

To address those barriers, this interdisciplinary collaboration will utilize a transformative paradigm to study visual contextual influences in ASD participants without intellectual impairment and typically developed controls with matched age, sex and IQ. Coen-Cagli and colleagues will obtain multimodal measurements during PGS of natural scenes using their innovative design¹, and will test their Bayesian theory of dynamic perceptual inference. The team expects the results to reveal how the time course of PGS is influenced by natural sensory inputs and priors in ASD compared to typically developed controls, and to identify neural hallmarks specific to the different components of atypical Bayesian inference. Coen-Cagli will also explore the potential for developing objective, model-based measures of autism phenotype and early biomarkers of underlying neuropathology.

1. Vacher J. et al. PLoS Comput. Biolo 19, e1011483 (2023) PubMed