| Abstract |
| The question of whether core affective dimensions are represented in a modality-general or modality-specific manner remains a central issue in affective science. Most prior research examined vision and audition, leaving unclear whether other modalities, such as gustation, share similar affective structures. Using behavioral ratings from Park and Kim (2024), we applied a Bayesian hierarchical multivariate model to examine emotional responses to gustatory and auditory stimuli. We defined theoretically motivated contrasts for valence and arousal and assessed alignment between estimated category effects and these contrasts through posterior projections, cosine similarity, and Mahalanobis distances. Across analyses, valence exhibited strong alignment between modalities, with minimal cross-modal differences, supporting a modality-general representation. Conversely, arousal showed weak and divergent alignments: taste displayed weak positive alignment, whereas music exhibited negative alignment, accompanied by greater cross-modal separability. Posterior predictive checks indicated that the model adequately captured the observed condition means. These results align with previous neuroimaging, psychophysiological, and behavioral evidence suggesting that valence is encoded in abstract, supramodal representations, while arousal is more sensitive to modality-specific processes. Methodologically, this study demonstrates the value of Bayesian multivariate modeling in quantifying uncertainty and testing theoretically driven contrasts in multidimensional affective data. |
|
|
| Key Words |
| Bayesian Analysis, Valence, Arousal, Modality-General, Modality-Specific, Posterior Predictive Check |
|
|
 |
|
