Auditory cortex morphology is related to overlaps between phonological inventories of multilinguals' languages
Olga Kepinska1, Josue Dalboni da Rocha2, Carola Tuerk3, Alexis Hervais-Adelman4, Cathy Price5, David Green5, Narly Golestani1,6; 1University of Vienna, 2St. Jude Children's Research Hospital, 3University of Montreal, 4University of Zurich, 5University College London, 6University of Geneva
Heschl's gyrus (HG), the cortical structure housing early auditory cortex, exhibits large individual variation in shape and size. Although formed in-utero, HG's volume has been previously associated both with putatively genetically driven factors (speech sounds learning abilities and musical aptitude), and environmental variables including bilingual language experience. One possibility is that distinct influences (environmental versus genetic) are reflected by different measures describing the structure of HG. Indeed, a recent large-scale genome-wide association meta-analysis suggested that cortical surface area is influenced by genetics and cortical thickness reflects environmentally driven neuroplasticity (Grasby et al., 2020). In this study, we leveraged a unique sample of 136 participants exposed to between 1 to 6 languages (2.65 languages on average) and asked whether the variability in morphology of their HG (cortical thickness in particular) reflected the variability in their language experience. We further explored whether typological distances between multilinguals' languages were associated with the neural signatures of multilingualism in the auditory cortex. Specifically, we investigated whether neuroanatomical indices describing HG were related to cross-linguistic phonological information: segmental, feature-level, or counts of phonological classes. To describe the language background of our participants, we expressed Age of Onset(s) of Acquisition (AoA) of different languages in a continuous quantitative measure using Shannon’s entropy equation. Subsequently, the PHOIBLE database and open-source software (Dediu & Moisik, 2016) were used to construct three measures of typological distance between the languages spoken by our participants: (1) overlaps in sets of segments belonging to each language; (2) overlaps in distinctive articulatory features describing the segments of each language (e.g., "short", "long"); and (3) similarity in counts of phonological classes that share certain features (e.g., "front rounded vowels", "clicks"). Next, the summed phonological distances between all language pairs for each participant were weighted by the AoA information for each of the participants' languages, resulting in three different indices of language experience accounting for typological relations between languages. We processed the T1 structural MRI data with FreeSurfer’s brain structural pipeline, and segmented HG using an automated toolbox (TASH, Dalboni da Rocha et al., 2020). Our dependent variables were TASH-derived measures (volume, surface area and thickness) for the extracted HG labels. For each gyrus and each measure, we first fit linear models assessing the relationships with multilingual language experience but ignoring typological relations between languages, controlling for age, gender and total brain volume or mean thickness. Out of all investigated cortical measures, only average thickness of the second HG (bilaterally) proved to be related to participants' language experience. Next, based on this result, we performed a model comparison procedure which showed that the language experience index including cross-linguistic segmental-level information explained the most variance in average thickness values of the second HG (both left and right). The direction of this effect was negative, showing that the more extensive and varied one's language experience, the thinner the cortex of their second HG. We hypothesize that this finding might reflect experience-driven pruning and neural efficiency, which would need to be tested in further longitudinal studies of language acquisition.
Topic Areas: Multilingualism, Phonology and Phonological Working Memory