Dentate nucleus activity reflects prediction errors during language processing

Poster Session B - Sunday, April 14, 2024, 8:00 – 10:00 am EDT, Sheraton Hall ABC

Darlene Floden¹ (flodend@ccf.org), Anshul Srivastava¹, Olivia Hogue¹, Slobodin Noah¹, Kenneth Baker¹, Andre Machado¹, Raghavan Gopalakrishnan¹; ¹Neurological Institute, Cleveland Clinic, Cleveland, OH USA

The cerebellum uses internal models to predict and facilitate sensorimotor, and possibly cognitive, outcomes. We recorded activity from the dentate nucleus and scalp during a semantic priming task to test the hypothesis that cerebellar activity is modulated by predictive context during language processing. Participants were eight patients with middle cerebral artery strokes enrolled in a clinical trial of dentate stimulation to treat motor impairment. Participants viewed 440 prime-probe word pairs in a LOW-expectancy condition (72% of pairs were unrelated words to deter predictive processing) followed by a HIGH-expectancy condition (73% of pairs were related words to encourage predictive processing). Eighty word-symbol pairs served as low-frequency (15% probability), non-semantic control stimuli. Scalp EEG time-locked to probe words revealed the typical N400 semantic potential at centro-parietal sites to words but not symbols. The N400 was largest to unrelated probe words in the HIGH condition, consistent with prior studies. Debiased weighted phase lag index analysis and cluster analysis revealed alpha band phase coherence between the dentate and frontal and parietal regions. Granger causality analysis revealed greater bidirectional signaling in the HIGH condition when predictions were violated (unrelated/symbols) relative to the LOW condition; cerebellar-to-parietal signaling was greater early in the trial (up to 400ms) while parietal-to-cerebellar signaling was greater during the N400 window. Dentate-to-frontal signaling was present between 500-700ms when predictions were violated (unrelated/symbols) in the HIGH condition. These findings suggest that the cerebellum is selectively engaged when predictions regarding upcoming semantic content are violated but does not respond to semantic content, per se.

Topic Area: LANGUAGE: Semantic