Language access differentially alters functional connectivity during emotion perception across cultures

The study investigates whether access to emotion category words and individuals’ cultural background jointly shape the neural mechanisms of emotion perception. Whereas basic emotion theories posit reflexive recognition of universal facial expressions, constructionist accounts argue that perceivers use conceptual knowledge—shaped by prior experience and culture—to make meaning of facial movements, with language serving as a key mechanism to access this knowledge. The authors hypothesized that priming English emotion labels would alter functional connectivity during perception of anger and disgust, with stronger effects for Chinese participants who may have less chronically accessible English emotion concept knowledge than White American participants. The work addresses how culture and language interact to influence the brain’s functional networks underlying emotion perception.

Prior behavioral work shows that language facilitates categorical perception and speeds judgments of emotional faces, while removing access to emotion words impairs accuracy and increases response times. Meta-analytic neuroimaging findings indicate that the presence of emotion words increases activation in the left inferior frontal gyrus (semantic retrieval), whereas their absence increases amygdala/parahippocampal activity, consistent with greater ambiguity and contextual processing. Cross-cultural research demonstrates substantial variation in emotion semantics across languages and differences in the production and perception of facial expressions, including category–face mappings (e.g., disgust) and neural responses to emotional faces across cultures. These literatures motivate the prediction that emotion word accessibility and cultural background interact to influence functional connectivity among regions supporting semantic processing, visual perception, and social cognition during emotion perception.

Participants: Fifty-one consented; 45 participated; final N=36 after exclusions (motion >2 mm: 4; experimental errors: 4; early termination: 1). Groups: 15 Chinese (mean age 20.4±2.2; 9 female) and 21 White American (mean age 22.3±3.4; 10 female). No significant age differences by sex or cultural group. White Americans were US-born/raised by monolingual English-speaking non-Hispanic White American parents, with no residence outside the US. Chinese participants were born/raised in mainland China (excluding Hong Kong, Macau, Taiwan) by monolingual Mandarin-speaking parents, had not resided outside their provinces before moving to the US as adults, had lived in the US <20 months on average (M=18.4±15.1 months), and all achieved high TOEFL proficiency (M=107.6; range 102–112). Practice task: Two practice runs outside the scanner used the label "Sadness" or control text "XXXXXX" with sadness faces to avoid priming anger/disgust categories before scanning. Design: Four 3-minute fMRI runs in a 2 (Face Expression: Anger vs Disgust) × 2 (Prime: Emotion Label vs Control) block design. At the start of each run, participants viewed either an emotion label ("Anger" or "Disgust") or control text ("XXXXXX") for 2000 ms; labels appeared only once per run to keep priming subtle. Participants then passively viewed 40 faces per run (each 2000 ms) portraying the run-congruent expression (anger in Label Anger and Control Anger; disgust in Label Disgust and Control Disgust). Interstimulus intervals were jittered fixation crosses (2000–8000 ms). Runs were counterbalanced; faces were randomly ordered within runs. Stimuli: Practice stimuli (sadness) were from the NimStim set. Main task stimuli were from the IASLab Face set: 10 White female actors each posing anger and disgust (closed-mouth), ensuring similar muscular activation across categories and high inter-rater agreement. MRI acquisition: 3T Siemens PRISMA. Structural T1: 0.8 mm isotropic; TR=2400 ms; TE=2.22 ms; 208 slices; matrix 320×320; FOV=256 mm. Structural T2-weighted matched-bandwidth: slice thickness 3 mm; 38 slices; TR=5700 ms; TE=65 ms; matrix 192×192; FOV=230 mm; voxel 1.2×1.2×3.0 mm3. Functional EPI: 37 slices; slice thickness 3 mm; TR=2000 ms; TE=25 ms; matrix 92×92; FOV=230 mm; voxel 2.5×2.5×3.0 mm3. Preprocessing: SPM12 with slice-time correction, realignment, normalization to MNI (2 mm), artifact scrubbing (ART), nuisance regression (white matter, CSF, linear trends, 12 motion parameters and derivatives), band-pass filtering (0.008–0.09 Hz), and spatial smoothing (8 mm FWHM). Functional connectivity analysis: Generalized psychophysiological interaction (gPPI) in CONN (v18b). Seeds: left IFG (parcellated via Schaefer atlas using peak coordinates from Brooks et al., 2017) and bilateral amygdala (6 mm spheres via MarsBaR centered at meta-analytic peaks from Lindquist et al., 2012). Targets: 70 ROIs from Harvard-Oxford and AAL atlases spanning semantic, social, and emotion perception networks (e.g., superior/middle/inferior frontal gyri, temporal poles, superior/middle temporal gyri, parietal regions, angular/supramarginal gyri, medial prefrontal cortex, anterior cingulate, anterior insula, precuneus, parahippocampal, lingual, fusiform, cerebellar lobules including vermis). First-level gPPI modeled psychological terms (task conditions convolved with HRF), physiological seed time series, and their interaction. Second-level analyses included cultural group as a covariate with contrasts Chinese > White American for Disgust Label > Disgust Control and Anger Label > Anger Control. Multiple comparisons were controlled using FDR (seed-level; two-tailed p<0.05) with Gaussian Random Field Theory–based parametric statistics.

• No significant functional connectivity differences for Anger Label vs Control for either cultural group; no group differences for anger conditions. - Disgust condition: Chinese participants showed decreased functional connectivity between left IFG and regions implicated in visual perception (bilateral lingual gyri), semantic representation (left anterior middle temporal gyrus), and social cognition (cerebellar vermis IX) when primed with the label "disgust" compared to control text; this effect was not observed in White American participants. Statistical evidence: omnibus seed-level effect for left IFG in Disgust Label > Control for Chinese > White Americans, F(8.27)=2.58, p=0.031 (two-sided FDR seed-level corrected p<0.05). Follow-up t-tests (Chinese): left aMTG t(34)=-3.53; left lingual gyrus t(34)=-4.23; right lingual gyrus t(34)=-4.81; vermis IX t(34)=-3.33 (all p<0.05, FDR-corrected). - No significant amygdala connectivity differences between conditions or groups.

The findings support constructionist accounts positing that language-mediated access to emotion concepts shapes neural processing of emotional faces and that this influence varies by cultural background. For Chinese participants, priming the English label "disgust" reduced IFG connectivity with regions supporting visual processing (lingual gyri), semantic representation (aMTG), and mentalizing (vermis IX), suggesting that providing a culturally specific emotion word may lessen the need for elaborative meaning-making when interpreting facial behaviors posed by White American actors. The absence of effects in White American participants is consistent with more chronically accessible English emotion concepts in that group. The category-specific effect for disgust aligns with cross-cultural work indicating that disgust is less distinctive in Chinese versus Western European contexts and may be less central historically in Chinese cultural texts, potentially increasing reliance on labels to access host-culture concept knowledge. Overall, results indicate that both culture and language jointly modulate functional connectivity during emotion perception, advancing understanding of how conceptual knowledge organizes brain networks in socioemotional cognition.

This study provides preliminary evidence that neural representations of emotional facial expressions depend on perceivers’ conceptual knowledge, specifically access to emotion words, and on cultural background. Priming an English emotion label altered functional connectivity during disgust perception in Chinese but not White American participants, highlighting how language and culture interact to shape emotion perception. These insights have implications for socioemotional processing in multicultural contexts. Future work should replicate with larger samples, include active tasks and same-race stimuli, test labels in participants’ primary languages, and assess effects of acculturation and familiarity with host-culture facial expressions.

• Design: Priming was explicit but brief and presented only once per run; control text used non-words (XXXXXX) rather than semantic controls, which was chosen to parallel prior meta-analytic contrasts but leaves open alternative interpretations. Passive viewing precluded verification of explicit categorization but avoided task-demand confounds. - Sample: Modest N (N=36) relative to neuroimaging standards may reduce power, particularly for whole-brain analyses; some functional activation analyses yielded null results, which should not be overinterpreted. - Cultural sampling: Two-culture comparison (Chinese vs White American) limits generalizability; Chinese participants were recent immigrants living in the US, potentially differing from China-residing peers in acculturation or personality. - Stimuli and language: Only White female faces were used; labels were in English only and not tailored to participants’ primary language; no same-race stimuli were included for each group. - Future needs: Larger, more diverse samples; inclusion of continuous cultural measures; manipulation of acculturation and familiarity; inclusion of active tasks; testing labels in primary vs secondary language and addition of same-race/other-race conditions.