Fiscal Year
FY 2025
April 2025 – March 2026

Principal Investigator
Hyeonjeong Jeong
Professor, Graduate School of International Cultural Studies

Co-Investigators
Hyoungsun Kim
Graduate School of International Cultural Studies
Motoaki Sugiura
Professor, Institute of Development, Aging and Cancer

Research Keywords
Embodied metaphors ; Interoception ; Emotion ; fMRI ; MVPA

1. Research Overview

Embodied theories propose that understanding language involves the partial reinstatement of bodily and sensorimotor states associated with described action or experience. In principle, this framework extends to emotional experience and its underlying bodily states, including those arising from the viscera. Although emotional experiences are fundamentally embodied, with precise bodily feelings finding nuanced expressions in language, neural evidence for interoceptive reinstatement through language remains scarce compared to exteroceptive and action-related simulation.

In Study 1, we investigated whether embodied language expressions of emotional experience differ at the neural level from abstract emotion labels, using functional magnetic resonance imaging. We found that embodied expressions (e.g., I felt my heart being squashed) engage brain areas known to represent bodily states, namely the anterior insula and anterior cingulate cortex. Importantly, this pattern was not found with abstract labels of emotion (e.g., I felt devastated), despite being preceded by identical emotional contexts. This suggests embodied language can convey embodied aspects of emotional experience beyond abstract meaning, based on the reinstatement of those bodily states.

In Study 2, we investigate the information content of bodily representations reinstated through language. Given that reinstated states should preserve the information of their bodily origin, this study tests whether the information structure decoded from direct cardiac and gastric interoception generalizes to embodied language expressions. The neural activation pattern from each trial is mapped as a feature vector in an n-dimensional vector space (where n equals the number of voxels). A linear support vector classifier is trained to distinguish cardiac from gastric interoception trials by finding the optimal hyperplane. Critically, the same classifier is then applied to data from the language task, testing whether the learned boundary generalizes to distinguish heart- from stomach-associated embodied expressions. Successful generalization would demonstrate that language-induced neural patterns carry the same source-specific bodily representations as direct interoception. This study is currently ongoing, and preliminary findings will be made available by September 2026.

2. Significance of the Research and Future Prospects

This project asks whether and how language encodes and transmits body-grounded aspects of emotional experience, beyond the level of abstract category or intensity, at the level of specific bodily states decodable from neural activity patterns. This interdisciplinary project holds implications across Language Sciences, Phenomenology, and Affective Neuroscience, with the potential to elucidate how language can communicate the felt aspects of human experience, beyond those driven by information from the outside world.

The current project focuses on one dimension of bodily state information: source location, or where in the body the feeling originates. This is a tractable and theoretically motivated starting point, but it is not the only phenomenologically distinctive dimension along which embodied emotional experiences differ. Two experiences of fear can both originate in the chest yet feel very different—one a sudden clench, the other a slow, but overwhelming chill—differentiated by their dynamic texture or qualities of felt experience. A planned extension will address this dimension directly, focusing on metaphorical expressions that reflect how a feeling moves, builds, and unfolds over time and their underlying neural representations.

Another promising direction is to extend this framework to large-scale cross-linguistic and cross-cultural investigations. Research on embodied language has been conducted predominantly in English, despite substantial cross-cultural variation in how embodied emotions are conceptualized and expressed through language. Conducting parallel neuroimaging studies across typologically and culturally diverse language communities would allow a more complete understanding of how universal properties of bodily states and culturally learned conceptualizations jointly shape emotional experience.

3. Conclusion

What kind of knowledge do embodied expressions such as 怒りがこみ上げる (“anger wells up”) encode? Using functional magnetic resonance imaging (fMRI), this project investigates the content of communication via embodied expressions of emotional experience at the neural level. Although language cannot fully convey rich subjective experience in its entirety, it provides a shared symbolic system through which private experiences can be approximated, communicated, and understood across individuals. We combine traditional univariate analyses with machine learning-based classification of multivariate patterns in the brain to examine whether and how nuanced expressions in everyday language can convey bodily states that constitute embodied experience of emotion.