Early Sensory Prediction Errors Are Less Modulated By Global Context in Autism Spectrum Disorder

Background: Recent theories of autism spectrum disorder (ASD) attempt to explain both social and sensory symptoms of ASD by using the predictive coding framework. Central to these domain general accounts is the hypothesis that individuals with ASD are less flexible in modulating local prediction errors as a function of global top-down expectations. Though some studies already provided preliminary support for these recent theories, this main hypothesis remains to be tested.

Objectives: As a direct test of this central hypothesis, we used electroencephalography (EEG) to investigate whether local prediction error processing was less modulated by global context in ASD.

Methods: A group of 24 adults diagnosed with ASD was compared to a gender-, age- and IQ-matched group of 24 neurotypical adults on a well-validated hierarchical predictive coding paradigm. In this auditory oddball task, participants listened to short sequences of either five identical sounds or four identical sounds and a fifth deviant sound. The latter condition is known to generate the mismatch negativity (MMN) component, believed to reflect early sensory prediction error processing. Crucially, we manipulated the relative frequency of deviant sound sequences across blocks, as previous studies have shown that in blocks with frequent deviant sound sequences, top-down expectations seem to attenuate the MMN. We predicted that this modulation by global context would be less pronounced in the ASD group.

Results: Both groups showed a MMN that was modulated by global context. However, this effect was significantly smaller in the ASD group as compared to the neurotypical group. In contrast, the P3b, as an electroencephalographic marker of conscious expectation processes, did not differ across groups.

Conclusions: Our results demonstrate that the MMN was less modulated by global context in the ASD compared to the neurotypical group. These findings confirm the central hypothesis of contemporary predictive coding accounts of ASD, indicating that individuals with ASD are less flexible in modulating their low-level prediction errors according to more global contexts. Therefore, this study provides an important piece of evidence for these domain general theories of ASD.