Human iEEG recordings during audiovisual speech violations support and extend predictive routing models

Human intracranial EEG recordings during audiovisual speech tasks provide evidence for and expand predictive routing models of predictive processing in the human brain. Rather than solely relying on sensory input to reconstruct the world, this theory suggests the brain uses internal predictions guided by neuronal oscillations. The emerging predictive routing model posits that gamma and alpha-beta rhythms dynamically interact to implement this mechanism.

Specifically, predictions enhance alpha-beta oscillations while suppressing gamma activity during predicted input processing. While this model has been supported by animal studies, its validity within the human brain has not been fully investigated. This study addresses this gap by analyzing iEEG recordings from seven presurgical epilepsy patients with active recordings in superior temporal and prefrontal/premotor brain regions.

The researchers measured power in theta, alpha-beta, and high gamma bands across three conditions: simple auditory deviance, multisensory integration, and an audiovisual incongruity condition causing the McGurk illusion. Results showed increased high gamma and decreased alpha-beta activity in the superior temporal cortex during auditory deviance and the McGurk condition.

In contrast, frontal areas showed similar alpha-beta decreases but lacked the high gamma prediction error signal. Instead, theta band activity revealed a strong prediction error signal in the McGurk condition, aligning with theories on theta's role in reorganizing cortical networks. This research supports and extends existing predictive routing models, demonstrating their applicability to human-specific forms of predictive processing.