Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Role of the Cerebellum in Semantic Analogical Reasoning
Poster Session C, Thursday, October 1, 10:45 am - 12:45 pm, Wangari Maathai
This poster is part of the Sandbox Series.
Carolyn N. Irving1, Amanda LeBel1, Cong Du1, Amanda Chirino-Perez1, Theo Wilson2, Richard Ivry1; 1University of California, Berkeley, 2University of Virginia
Despite extensive investigation, how the cerebellum supports language remains unclear. Neuroimaging studies find strong cerebellar activation during a variety of linguistic tasks and have revealed semantic representations across the cerebellar cortex, similar to that observed in the cerebral cortex. However, patients with adult-onset cerebellar degeneration (CD) show only subtle problems on language tasks, with the most pronounced impairments observed on semantic and phonemic fluency tasks. Taken as a whole, this body of research has resulted in largely descriptive accounts of cerebellar involvement in language. In the present study, we use a generative analogy task with a goal to characterize cerebellar contributions to semantic processing. Participants were presented with stimuli structured as 'A is to B as C is to ____' and required to generate a response to complete the analogy. We characterized the stimuli and the participants’ responses using a semantic embedding space (GloVe), allowing us to test multiple hypotheses about the role of the cerebellum in semantic processing. For example, we quantified the relational (i.e., A:B) and mapping (i.e., A:C) distances as a measure of semantic manipulation (e.g., moving between conceptual relationships). We also quantified the cosine distance between the participant’s response and target word (i.e., D) as a measure of semantic error magnitude. Preliminary results (CD N=20; Controls N=10) suggest that the cerebellum is not involved in semantic processing per se. While both CD and control participants show a decrease in accuracy as the mapping distance of the analogies (A:C) increases, there is no Group x Distance interaction. In terms of semantic error, both CD and control participants show a similar error distance, suggesting that their responses are semantically similar. However, the CD group shows a larger decrease in accuracy compared to controls when a trial has higher surprisal. This pattern indicates that greater demands on executive functioning impacts the CD group more than controls. An additional analysis categorizing the types of errors produced by each participant group suggests that the type of errors made by the CD group are more consistent with cognitive control deficits than with impaired semantic processing, relative to controls. For example, the CD group is more likely to make reasoning errors, applying the wrong rule (e.g.: “wealthy is to rich as calm is to” “strong”) than lexical errors in which they apply the correct rule but produce an incorrect word (e.g.: “desk is to furniture as toothbrush is to” “bathroom stuff”). Together, these pilot data suggest that the cerebellum might support the cognitive flexibility required for semantic processing, but not linguistic processing selectively.
Topic Areas: Meaning: Lexical Semantics, Control, Selection, and Executive Processes