Item

Abstract

Adaptive interaction with our rapidly and continuously evolving environment relies on predictions that capture the embedded statistical regularities. One neural structure associated with prediction is the cerebellum, traditionally in spatiotemporal motor control, but recently also in sensory prediction and attention, especially in the time domain. However, in addition to where and when, the brain also predicts what, a well-established example being semantic prediction in language. Yet, the role of the cerebellum in feature prediction has been controversial. Here, we addressed this issue by investigating the neurophysiological signatures of semantic prediction in patients with cerebellar damage compared to healthy controls. We used a semantic prediction paradigm during EEG recording. The word-by-word presentation of a sentence induced a specific prediction about the most likely final word to follow. The final word of the sentence could be either congruent or incongruent with this prediction, eliciting a violation response known as the N400 potential. Impaired feature-based predictions in cerebellar patients should therefore be reflected in an alteration or even abolition of this EEG response. We found that behaviorally cerebellar patients were still able to detect semantic violations. However, there were clear differences between patients and controls in these physiological signals, underscoring the importance of the cerebellum for these predictions. Most importantly, the N400 response was not abolished in cerebellar patients, but rather its latency was delayed. These findings may explain previous conflicting results and reveal a novel neurofunctional role of the cerebellum in feature prediction in the language domain.