Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Poster Slams
The heterogeneous engagement of the language network during statistical learning
Poster A62 in Poster Session A, Thursday, October 6, 10:15 am - 12:00 pm EDT, Millennium Hall
Julie Schneider1,2, Terri Scott3, Jennifer Legault2, Zhenghan Qi2,4; 1Louisiana State University, 2University of Delaware, 3University of California San Francisco, 4Northeastern University
Recent behavioral research has demonstrated a reciprocal relationship between prior language experiences and performance during statistical learning (SL). Despite the significant overlap between SL and language in their richness of regularities, it remains unknown whether the neural network involved in language processing is similarly engaged in SL. The current study probes whether individuals recruit the same brain regions equally across language processing and SL tasks. Twenty-two adults completed an auditory SL fMRI task (Schneider et al., 2020) and an auditory language localizer fMRI task (Scott et al., 2017). Group-level univariate analyses revealed neural engagement during SL in regions often associated with language processing. To directly test whether engagement during SL occurred in language-specific regions across individuals, we constructed a set of functional regions of interest (fROI) from the language localizer task by contrasting intact and degraded conditions (similar to Fedorenko et al., 2010). From this parcellation of the language localizer maps, eleven fROIs emerged: two in the left superior temporal gyrus – posterior and anterior, left middle temporal gyrus, left temporal pole, left precentral gyrus, the left inferior frontal gyrus pars opercularis, two in the right superior temporal gyrus – posterior and anterior, right middle temporal gyrus, right temporal pole and right precentral gyrus. Within this subject-specific language network, the bilateral superior and middle temporal gyri were activated during the SL task, showing greater neural activity in response to processing structured versus random syllable sequences. However, due to inter-subject heterogeneity in brain activation during SL, there was no significant conjunction in these same regions across learners. Moreover, each participant’s patterns of neural activation were not correlated across the two tasks. Using a search-light multi-voxel pattern similarity analysis approach in the whole brain, we identified a brain region outside of the classic language network, the left supramarginal gyrus (LSMG), that showed significant correlated patterns across the two tasks for 60% of learners. Further investigation of this correlation revealed that, for the 60% of participants who showed cross-task correlation in the LSMG, the LSMG was only actively recruited for the auditory SL task, but not for the language task. Our approaches, relying on rigorous functional localization techniques, confirmed the involvement of language regions during SL. We identified that the bilateral MTG and the left STG, parts of the frontotemporal core-language network (Fedorenko & Thompson-Schill, 2014, Price 2010), are sensitive to embedded regularities in a stream of meaningless syllables. These findings are consistent with previous reports on the activation of temporal cortices in various auditory SL paradigms (McNealy, Mazziotta, & Dapretto, 2006; 2011; Karuza et al, 2013; Plante et al., 2015; Cunillera et al., 2009). In addition to these important confirmatory results, our study contributes to the recent emerging literature on individual differences of SL (e.g., Erickson et al., 2016; Siegelman et al., 2015; 2017), by providing neural evidence for heterogeneous learning patterns across individuals.
Topic Areas: Speech Perception, Meaning: Lexical Semantics