Poster Presentation

Search Abstracts | Symposia | Slide Sessions | Poster Sessions

Altered semantic representations in the cerebellum and cerebrum of autistic individuals

Poster Session A, Friday, September 12, 11:00 am - 12:30 pm, Field House

Amanda LeBel¹, Bangjie Wang², Ashley Blanco², Matteo Visconti di Oleggio Castello¹, Richard Ivry¹, Jack Gallant¹, Anila D'Mello^2,3; ¹University of California Berkeley, ²University of Texas at Dallas, ³University of Texas Southwestern Medical Center

Individuals with autism spectrum disorder (ASD) process lexical semantic information differently than neurotypical people. However, the brain representations associated with lexical-semantic processing in ASD remains unclear. Previous work has shown that crus I/II of the cerebellum is one of the most common brain regions to show atypical structural development in ASD populations and that these structural abnormalities are correlated with early language delays. Further, neuroimaging studies have shown that crus I/II of the cerebellum represent social semantic information in neurotypical populations. These two lines of evidence suggest that structural and functional differences in crus I/II of the cerebellum may account for behavioral language differences in ASD. Previous studies of lexical semantic representation in ASD have been limited in two important ways. The first is that conventional approaches have often under-sampled the semantic space by comparing how autistic participants differ from neurotypicals on only a limited number of words or categories. Second, traditional approaches have often failed to consider alternate brain regions besides the neocortex such as the cerebellum. The exclusion of the cerebellum biases our understanding of representational changes in ASD towards the neocortex. To address the limitations in previous studies, we combined natural language stimuli with encoding modeling techniques to precisely map the semantic system in the cortex and cerebellum. To map the semantic system, BOLD fMRI data was collected while participants (ASD =12; neurotypical = 12) listened to naturalistic language stories. Each participant listened to 11 different ~10-minute stories, with a total of 2 hours of fMRI data per participant. To model language representations in each participant, we extracted a series of feature spaces from the narrative stories to capture different types of language information across the language processing hierarchy. These feature spaces captured information that ranged from low-level auditory information to high level semantic and pragmatic information. We used regularized regression to model the relationship between the feature spaces and the BOLD response for each voxel and evaluated the model on a held out test set. We compared model weights between ASD participants and neurotypicals to understand differences in language representations between the two groups. Low-level language representations associated with auditory and phonological processing were similar between ASD and neurotypical participants in the cortex and cerebellum. In contrast, we observed large differences in semantic representation between the ASD group compared to the neurotypical group. Semantic representations were much more sparse in ASD participants than in neurotypical participants. This effect was strongest in crus I/II of the cerebellum, although differences are also evident in the frontal lobe of the neocortex. Representations related to social processing and theory of mind were significantly underrepresented in both the cerebellum and cortex in ASD participants compared to neurotypical participants. These findings show that the semantic representations are altered in ASD particularly in the cerebellum and frontal lobe of the neocortex. This underscores the potential importance of the cerebellum for studies of ASD.

Topic Areas: Disorders: Developmental, Meaning: Lexical Semantics