Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Data-driven individual functional parcellations of IFG reveal entangled systems for language, speech, and working memory
Poster Session A, Wednesday, September 30, 11:00 am - 1:00 pm, Wangari Maathai
Rebecca Belisle1, Emily Stephen1, Tyler Perrachione1; 1Boston University
Myriad functional descriptions of IFG have been proposed based largely on group-averaged functional data; classical disciplines in linguistics like syntax, semantics, and phonology; and structural boundaries based on macroanatomical landmarks like pars opercularis, triangularis, and orbitalis. However, a common scheme for describing the functional organization of individual brains may not necessarily respect the disciplinary or anatomical divisions we have invented to describe language or the brain (Blank & Fedorenko, 2020; Hickok et al., 2023; Marr, 1982). Here, we used data-driven clustering to parcellate IFG in individual brains based on common patterns of vertex-wise activity to multiple fMRI tasks spanning working memory, receptive language, and speech production. Six fMRI tasks (passive language listening, nonword discrimination, nonword repetition, real-word repetition, visuospatial sequence working memory, and auditory digit span) and structural MRI data (T1-, T2-, and diffusion-weighted volumes) were obtained from N=24 native English-speaking young adults. We ran k-means clustering using the pattern of fMRI responses to these tasks across all subjects’ cortical surface vertices within either left or right IFG. Each vertex was assigned to a cluster based on proximity in the 6-dimensional functional feature space. The optimal number of clusters was selected holistically using multiple metrics, including sum of squared errors and silhouette scores. We then identified structural features (probabilistic tractography, intracortical microstructure, and cortical morphometry) unique to each functional region. This procedure revealed a “receptive language cluster” (exclusively responsive to the language task), a “multiple demand task cluster”, (exclusively responsive to digit span and spatial working memory), and a “speech cluster” (engaged in tasks with overt vocal production) bilaterally, as well as a LH “supramodal cluster” (responsive to all tasks). The receptive language cluster spanned the more gyral part of the entire IFG, had distinctly increased intracortical myelin, and had greater connectivity with subcortical areas, especially pulvinar, consistent with recent evidence suggesting cortical–subcortical connections shape higher-level audition and language (Lohse et al., 2026; Woolnough et al., 2026). The LH supramodal cluster consisted of smaller nodes adjacent to the language cluster and was distinguished by diverse thalamic connectivity (e.g., MGN, LGN, pulvinar) and connectivity to premotor- and sequencing-associated regions (e.g., preSMA, SMA, and dorsal motor cortex). Interestingly, the more sulcal speech cluster was distinguished by connectivity to motor planning (e.g., caudate, preSMA), but not motor execution areas like precentral gyrus. The speech region was also characterized by connectivity with areas implicated in higher-level cognitive processes like laterodorsal thalamus (Perry & Mitchell, 2019), insula, planum temporale, and temporal pole, underscoring the IFG speech area’s potentially more abstract role, such as serving as a speech sound map (Kearney & Guenther, 2019). Together, these results suggest that (i) data-driven parcellation of IFG can reveal novel, coherent functional organization across individuals; (ii) LH IFG may have multiple dissociable linguistic contributions reflected in regions that are specifically responsive to receptive language, speech production, and a “supramodal” array of functions that includes both language and non-language tasks; and (iii) IFG–subcortical connectivity may be the key to understanding IFG’s functional organization and unique contributions to language processing.
Topic Areas: Language Production, Computational Approaches