Search Abstracts | Symposia | Slide Sessions | Poster Sessions
The extended language network: A large-scale characterization of language-responsive regions beyond the core fronto-temporal network
Poster Session B, Friday, October 25, 10:00 - 11:30 am, Great Hall 3 and 4
Agata Wolna1,2, Benjamin Lipkin1, Evelina Fedorenko1; 1Massachusetts Institute of Technology, 2Jagiellonian University, Krakow, Poland
Research on the neural basis of language has focused on the core left frontal and temporal brain areas and their right-hemisphere homotopic areas. However, brain imaging, intracranial recordings, and patient studies suggest that other areas—cortical, subcortical, and cerebellar—also contribute to linguistic processing. Here, using a large fMRI dataset (n=706), we identify and validate a comprehensive set of brain regions consistently activated during language processing. All participants performed a language ‘localizer’ task – an extensively validated paradigm which contrasts reading sentences with reading nonwords (Fedorenko et al., 2010). Importantly, this paradigm effectively isolates language processing from both lower-level perceptual and motor processes and from higher-level knowledge and reasoning (for review, see Fedorenko et al., 2024). To search for language-responsive areas outside of the core frontal and temporal areas, we used two approaches. First, we used a group-constrained subject-specific (GSS) approach (Fedorenko et al., 2010; Julian et al., 2012) to search across the brain for areas within which most participants show responses to the language localizer contrast. This approach produces a set of ‘parcels’ that can be used as masks to identify subject-specific functional regions of interest (fROIs). And second, we used pre-existing anatomical and multimodal brain atlases: the Glasser atlas (Glasser, et al., 2016), the Desikan-Kiliany-Tourville atlas (Klein, Tourville, 2012), and the Harvard-Oxford Cortical and Subcortical atlases (Desikan et al., 2006). In both approaches, subject-specific fROIs were defined within each GSS-based parcel or atlas-based anatomical area as the 10% of the voxels showing the strongest language response. Subsequently, we characterized all the language-responsive fROIs by examining their response to (i) the reading language localizer task (reading sentences vs. nonwords; n = 706; here, we used across-runs cross-validation to ensure independence between the data used to define the fROIs vs. to examine their responses), (ii) an auditory language localizer task (listening to intact vs. acoustically degraded passages; n=116; Malik-Moraleda, Ayyash et al., 2022); and (iii) a spatial working memory task (remembering more vs. fewer locations within a grid; n=422), commonly used as a localizer contrast for the domain-general Multiple Demand network (e.g., Assem et al., 2020). Our analyses replicate previous findings showing robust engagement of a left-lateralized fronto-temporal network in language processing. Critically, we identified a set of additional language-responsive areas, including areas in the medial superior frontal cortex, in the precuneus, and on the ventral temporal surface, as well as several right-lateralized cerebellar regions. The analysis of subcortical regions revealed responses to language within the hippocampus, amygdala, caudate, and thalamus. Many of these areas show responses during both reading and listening to language and selectivity for language relative to demanding non-linguistic tasks. The GSS and atlas-based analyses yield largely convergent results. This work lays the foundation for systematic characterization of the language-responsive areas beyond the core network. Understanding the contributions of these additional language-responsive brain areas can help paint a more complete picture of language processing, including how the core language areas may interact with non-language-selective systems underlying perception, motor control, and cognition.
Topic Areas: Methods,