Structural and Functional Network Predictors of Language and Communication in Autistic Youth

Abstract

Language ability in autism spectrum disorder (ASD) varies widely across individuals, yet the neural basis of this heterogeneity remains poorly understood. Most prior work has focused on classical cortical language regions, leaving open the question of how subcortical and cerebellar contributors shape language outcomes. This thesis examined whether structural and functional features of the cerebellar-thalamo-cortical language network predict language ability and social-communication symptom severity in autistic youth. Using structural and resting-state functional MRI data from the Autism Brain Imaging Data Exchange (ABIDE), I fit robust linear models relating regional volumes, functional connectivity, and volumetric interactions to standardized language assessments (CELF), clinician-rated communication symptoms (ADOS), and caregiver-reported communicative history (ADI-R) in 373 ASD participants aged 7-19 years. Four associations survived Benjamini-Hochberg correction within their respective test families. Resting-state functional connectivity from right Crus II to right inferior frontal gyrus negatively predicted CELF receptive language scores, with stronger coupling associated with weaker language ability– a pattern consistent with reduced network differentiation in autism. Bilateral medial geniculate nucleus (MGN) volume predicted CELF expressive and core language scores, identifying the auditory thalamic relay as a structural correlate of language ability. These results are consistent with a view of language in autism that is shaped by distributed cortical-subcortical-cerebellar circuits and individual subcortical structures, rather than by classical cortical language regions alone.