Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Intact engagement of speech production regions during implicit linguistic learning in autistic adults
Poster Session D, Saturday, September 13, 5:00 - 6:30 pm, Field House
Anna Ciriello1, Amanda O'Brien2,3, Anqi Hu1, John Gabrieli2, Zhenghan Qi1; 1Northeastern, 2Massachusetts Institute of Technology, 3Harvard University
Background: Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by social, cognitive, and communicative challenges, with many individuals also showing motor abnormalities (Maffei et al., 2024). A recent review on oromotor deficits in autism (Maffei et al., 2023) found strong correlations between speech motor and language development in autism. This was partially supported by fMRI findings in autistic children with impaired phonological skills. These children exhibited atypical engagement of the speech production network during a nonword repetition task (O’Brien et al., 2023). The current fMRI study focuses on the role of the speech production network during an artificial language learning task. Our prior research suggests a strong relationship between artificial language learning abilities and natural language learning skills in autistic individuals (Hu et al., 2024). We expect typical engagement of these brain regions in autistic individuals who achieved typical-like language skills. Methods: 30 neurotypical adults (NT) and 19 autistic adults (ASD) participated in this study. We identified thirteen subject-specific speech production functional regions of interest (fROIs) from a speech production localizer task (Picture Naming > Picture Viewing; O’Brien, 2025) using the group-constrained subject-specific (GCSS) method (Fedorenko et al., 2010). We examined how these fROIs were activated during the artificial language learning task across two conditions: Structured (i.e., sequences including embedded triplet patterns) and Random (i.e., sequences with randomly ordered syllables). A subset of participants (ASD = 14; NT = 25) also completed a similar artificial language learning task outside the scanner. Participants were asked to detect a target syllable when listening to a continuous stream of syllables with embedded triplets. They subsequently completed a two-alternative forced-choice (2AFC) task for their recognition of the triplets. Results: In both groups, we saw an overall effect of condition (NT: p < .0001; ASD: p < .05) with no group by condition interaction, confirming an intact engagement of the speech production network in autistic individuals. Both groups also showed similar regions sensitive to embedded linguistic patterns: left STG (NT: p = .0209; ASD p = .0145), right planum temporale (NT: p = .0015; ASD: p = .0065), and right STG (NT: p < .0001; ASD: p = .0075). The right prefrontal cortex is engaged during statistical learning only for the NT group (p = .0418). Compared to NT, the ASD group showed reduced activation across both conditions in the right prefrontal cortex (p = .045) and left STG (p = .0096). There was no three-way interaction between group, condition, and parcel. In participants who completed the out-of-scanner learning task, we observed similar above-chance 2AFC performance between groups but a steeper learning slope in response time during familiarization in ASD than TD (p = 0.01). Conclusion: The convergence of behavioral and neural evidence suggests that the speech production system is similarly engaged in linguistic statistical learning in autistic and neurotypical adults, reflecting overlapping mechanisms. Future research will examine whether the engagement of the speech production network during learning varies with individuals’ language skills and learning outcomes.
Topic Areas: Language Production, Disorders: Developmental