Abstract
Inspired by recent predictive processing models of autism, we investigated if individuals with autism developed implicit expectations and typical surprise responses to stimuli engaging low-level visual processing and targeting the early visual cortex. Specifically, we presented sine-wave gratings at two visual field locations known to evoke C1 event related potential (ERP) components and, unbeknownst to the participant, created an orientation/location contingency (i.e., some orientations were presented 80% of the time at a particular location). We collected high-density EEG responses, behavioral orientation adjustment responses and confidence measures as a probe of metacognition. We tested 54 adult participants divided into two groups (autism vs. control). Our results replicated the work of Jabar et al. (2017) showing that high probability orientations are associated to lower C1 amplitudes and smaller errors in the adjustment task. For group comparisons, only the control group showed significant differences between high and low probability trials at both EEG and behavioral measures. This suggests that individuals with ASD were unable to learn implicit expectations behaviorally and neutrally, at the level of the early visual cortex (i.e., sharpening the responses to frequent stimuli or responding differently to expected and unexpected stimuli). This suggests a deficit in perceptual learning and/or feature-based attention at the level of the early visual cortex for individuals with autism, which we interpret as altered learning of contextual priors in light of current predictive coding models of autism. Interestingly, these results run counter findings of intact implicit learning in several other tasks, as well as intact (or even improved) low-level processing in autism. However, the combination of implicit learning and low-level perception (as well as the measurement of the expectation-based modulation of the early C1 component) is novel to the field, and may explain the divergence. More extensive analyses of the metacognition measures will also be discussed.