A shared acoustic core and culture-specific emotion words: a transfer-learning analysis of cross-cultural music emotion

Introduction Cross-cultural differences in the perception of music and emotion are well documented, yet their acoustic origins remain unclear. Do these variations stem from the physical-acoustic properties of the music itself, or from the culture-specific emotion-word systems that listeners use to describe it? Methods We analyzed a public dataset in which listeners from five national and language samples rated the same 1,180 songs (15-s excerpts). We integrated these ratings with a 41-dimensional interpretable musicological acoustic fingerprint and used a transfer-learning protocol to quantify how the mapping from acoustics to emotion generalizes across samples. Results Within each sample, ridge regression predicted emotion ratings ( R 2 = 0.14–0.20; ρ = 0.37–0.45), and random forest and gradient boosting did not outperform ridge. Because each sample’s 20 words were independently elicited and rank-ordered, position-wise pairing is semantically arbitrary: it produced a signed ρ of 0.057 but a mean |ρ| of 0.312, with 44% of correlations negative. Semantic alignment raised signed ρ to 0.221 (mean |ρ| = 0.344; 24% negative). A per-dimension sign calibration raised the signed mean to 0.307, essentially the mean |ρ|, showing that the apparent gain reflects orientation rather than new rank information. Divergence was associated with sparser, darker, more static acoustics, but acoustics explained only ~8% of variance. The aligned transfer generalized to unseen songs (ρ = 0.210 vs. 0.055 for position-wise pairing), though it remained about half the within-sample level. Discussion These findings support a partially shared acoustic core wrapped in culture-specific emotion words. Semantic alignment improves both signed transfer and directional consistency, but a quarter of aligned dimensions remain negatively related. Language is an important, modifiable component, not the sole or principal cause of cultural variation.

Authors

Institutions

Publication Details

Journal
Frontiers in Psychology
Published
2026-11-18
DOI
https://doi.org/10.3389/fpsyg.2026.1988378
Primary Topic
Neuroscience and Music Perception
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A shared acoustic core and culture-specific emotion words: a transfer-learning analysis of cross-cultural music emotion

Hongzhe Zhang
Frontiers in Psychology
Neuroscience and Music Perception
article

A shared acoustic core and culture-specific emotion words: a transfer-learning analysis of cross-cultural music emotion

Hongzhe Zhang
article en

Abstract

Introduction Cross-cultural differences in the perception of music and emotion are well documented, yet their acoustic origins remain unclear. Do these variations stem from the physical-acoustic properties of the music itself, or from the culture-specific emotion-word systems that listeners use to describe it? Methods We analyzed a public dataset in which listeners from five national and language samples rated the same 1,180 songs (15-s excerpts). We integrated these ratings with a 41-dimensional interpretable musicological acoustic fingerprint and used a transfer-learning protocol to quantify how the mapping from acoustics to emotion generalizes across samples. Results Within each sample, ridge regression predicted emotion ratings ( R 2 = 0.14–0.20; ρ = 0.37–0.45), and random forest and gradient boosting did not outperform ridge. Because each sample’s 20 words were independently elicited and rank-ordered, position-wise pairing is semantically arbitrary: it produced a signed ρ of 0.057 but a mean |ρ| of 0.312, with 44% of correlations negative. Semantic alignment raised signed ρ to 0.221 (mean |ρ| = 0.344; 24% negative). A per-dimension sign calibration raised the signed mean to 0.307, essentially the mean |ρ|, showing that the apparent gain reflects orientation rather than new rank information. Divergence was associated with sparser, darker, more static acoustics, but acoustics explained only ~8% of variance. The aligned transfer generalized to unseen songs (ρ = 0.210 vs. 0.055 for position-wise pairing), though it remained about half the within-sample level. Discussion These findings support a partially shared acoustic core wrapped in culture-specific emotion words. Semantic alignment improves both signed transfer and directional consistency, but a quarter of aligned dimensions remain negatively related. Language is an important, modifiable component, not the sole or principal cause of cultural variation.

Frontiers in PsychologyVol. 17
University of Auckland (NZ), Auckland University of Technology (NZ)
Neuroscience and Music Perception
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon