Illustration of the sample’s language experience. Each bar represents a single participant’s overall language experience; the height of the stacked bars within each bar represents the AoA index for individual languages (the taller the bar, the earlier in life a given language was acquired). The color of each stacked bar refers to the number of phonemes in each language’s phonological inventory. For reference, English phonological inventory has 40 phonemes (Stanford Phonology Archive, 2019a). Prior to plotting, data was sorted by the overall language experience based on a sum of AoA index for participants’ individual languages; consequently, data of participants with most diverse language experience can be found on the left-hand side of the figure, and the right-hand side includes data from monolinguals (i.e., exposed only to one language).

Overview of the performed analyses.

Number of participants, their demographic and language experience characteristics displaying different overall shapes of the TTG (i.e., total number of identified gyri in the left and right hemisphere). Last column lists whole sample’s descriptive statistics.

Left and right second transverse temporal gyri and language experience. Multiple regression model parameters (parameter estimates and standard errors, in brackets; p-values are listed according to the coding presented underneath the table) for the average cortical thickness values of the second TTG, as predicted by the four language experience indices: (1) the cumulative language experience measure not accounting for typology, and cumulative language experience weighted by overlaps between languages at the level of (2) acoustic/articulatory features, (3) phonemes, and (4) counts of phonological classes. Last two rows present model comparison results (additional variance explained and BF10 values). NB. all models including typological information were compared against the ‘No typology’ model.

Multilingual language experience and thickness of the second TTG. Average thickness of the second TTG in the left and right hemisphere were negatively related to the multilingual language experience index weighted by their phoneme-level phonological distances. Plots show residuals, controlling for age, sex and mean hemispheric thickness.

Thickness of the right first TTG and PT in participants with a single TTG in the right hemisphere. Average thickness of HG in the right hemisphere was positively related to the amount of multilingual experience, irrespective of typological relations between languages (left panel). The average thickness of the right PT was not related to language experience (right panel). Plots show residuals, controlling for age, sex, and mean hemispheric thickness.

Multilingual language experience and thickness of the second TTG in an independent sample of participants. (A) Average thickness of the left second TTG was not significantly related to the language experience index; average thickness of the right second TTG was significantly related to the language experience indices accounting for phoneme-level phonological overlaps between multilinguals’ languages (B) and feature level information (C). The model including phonological feature level information presented in panel (C) had the best fit to the average thickness data of the right second TTG.

Thickness of left and right second TTGs and language experience in an independent sample of participants. Multiple regression model parameters (parameter estimates and standard errors, in brackets; p-values are listed according to the coding presented underneath the table) for the average cortical thickness of the second TTG, as predicted by the four language experience indices: (1) the cumulative language experience measure not accounting for typology, and cumulative language experience weighted by overlaps between languages at the level of (2) features, (3) phonemes, and (4) counts of phonological classes. The last two rows present model comparison results (additional variance explained and BF10 values). NB. all models including typological information were compared against the ‘No typology’ model.

Illustration of the replication sample’s language experience. As in Figure 1, each bar here represents a single participant’s overall language experience; the height of the stacked bars within each bar represents the AoA index for individual languages (the taller the bar, the earlier in life a given language was acquired). The color of each stacked bar refers to the number of phonemes in each language’s phonological inventory.

Similarity matrices of typological distances between all languages represented in the study (N = 36) based on: (1) distances in distinctive acoustic and articulatory features describing the phonemes of each language (e.g., “short”, “long”); (2) distances in sets of phonemes belonging to each language; and (3) distances based on counts of phonological classes that share certain features (e.g., “consonants”, “front rounded vowels”, “clicks”). Data for individual languages were collected from the PHOIBLE database (Moran et al., 2019) and open-source software (Dediu & Moisik, 2016). The figure was generated in R, with the package pheatmap (Kolde, 2019), version 1.0.12.

Auditory ROIs used in the analysis. The ROIs are overlaid on an inflated surface in the native space of one of the participants.

Results of linear mixed models (parameter estimates and standard errors, in brackets; p-values are listed according to the coding presented underneath the table) testing the effect of language experience on the structure (volume, area, and average thickness) of the auditory regions: planum polare, Heschl’s gyrus, Heschl’s sulcus, planum temporale, anterior and posterior superior temporal gyrus, and anterior, middle, and posterior superior temporal sulcus. Anterior STG was used as the reference level.

Results of the linear mixed models testing the effect of Language Experience on the structure (volume, area and average thickness) of the gyri in the superior temporal region: first, second and third TTG. Anterior TTG was used as the reference level.

Results of a whole-brain vertex-wise analysis, aimed at establishing relations between the language experience index and whole-brain cortical thickness. Overlaid on the inflated surface of the fsaverage template brain is the thresholded at p < .0001 (uncorrected) significance map from the conducted F-test showing a negative relationship between cortical thickness in the highlighted region and the degree of multilingual language experience.

Cumulative phoneme inventory and thickness of the second TTG. Average thickness of the second TTG in the left and right hemisphere in relation to the number of unique phonemes each participant was exposed to across all their languages (the plotted values are residuals controlled for age, sex, mean hemispheric thickness and the language experience index irrespective of typology).

left and right second transverse temporal gyri and cumulative phoneme inventory. Multiple regression model parameters (parameter estimates and standard errors, in brackets; p-values are listed according to the coding presented underneath the table) for the average cortical thickness of the second TTG (left and right), as predicted by the “cumulative phoneme inventory” index. For comparison, models with the cumulative language experience measure not accounting for typology, and cumulative language experience weighted by overlaps between languages at the level of phonemes are also reported. Last two rows present model comparison results (additional variance explained and BF10 and BF01 values are also reported).

Right superior temporal plane (Heschl’s gyrus and planum temporale) and language experience in participants with one TTG. Multiple regression model parameters (parameter estimates and standard errors, in brackets; p-values are listed according to the coding presented underneath the table) for the average cortical thickness of the right Heschl’s gyrus, and the right planum temporale, as predicted by the four language experience indices: (1) the cumulative language experience measure not accounting for typology, and cumulative language experience weighted by overlaps between languages at the level of (2) phonemes, (3) acoustic/articulatory features and (4) counts of phonological classes. Last two rows present model comparison results (additional variance explained and BF10 values); NB. all models including typological information were compared against the ‘No typology’ model.

Multilingual proficiency and thickness of the second TTG. Average thickness of the second TTG in the left and right hemisphere as a function of language proficiency in all languages of each participant (top panel), only their proficient languages (middle panel) and only their non-proficient languages (bottom panel). Plots show residuals, controlling for age, sex and mean hemispheric thickness.