Extraction of Physical Phonemic Features Encoding Emotional Conceptual Information

This study investigates the extent to which conceptual information is encoded by the physical articulatory features in long-term memory. We used onomatopoeic words representing crying as prototypical expressions of sadness, which were phonologically decomposed into the physical articulatory features of phonemes. A total of 997 participants rated the emotional congruency of these onomatopoeic words. Using Elastic Net regression, we identified a subset of articulatory features that predicted rating of sadness. Of the 87 candidates, 8 features, including 5 articulatory features, were retained in the best-performing predictive model (R² = 0.454). These findings provide empirical evidence that physical articulatory features constituting phonemes finely convey psychological emotional content.