Sound-Meaning Relations in Japanese Tanka

:speech_balloon: Speaker : Jan Auracher

:classical_building: Affiliation : National University of Singapore

Title : Sound-Meaning Relations in Japanese Tanka

Abstract : In this study, the relationship between formant frequency and emotional tone was tested. To this end, 147 participants assessed each between 21 and 25 out of 75 pseudo-randomly selected Japanese Tanka on items comprising the three dimensions Evaluation, Potency, and Activity. Results show a significant correlation between formant dispersion and the dimensions of Potency and Activity. Stanzas with a high formant dispersion were more often assessed as expressing weakness and calmness than stanzas with low formant dispersion. These results also cooperate previous findings with German poetry. This suggests that sound-meaning relations in poetry are universal (i.e., language independent).

:newspaper: Long abstract
Research on the relation between sound and meaning in language (i.e., on sound iconicity) has reported substantial evidence for implicit associations between articulatory-acoustic characteristics of phonemes and emotions (for reviews see [1,2]). A specific branch of studies within the realm of sound iconicity has investigated the relationship between the content of texts and the frequency of occurrence of particular phonemes in these texts. However, when compared to other areas of research on sound iconicity, results from these studies provide a fairly ambiguous picture [3]. It has been proposed, though, that these inconsistencies can partly be explained by the research methods applied [4]. In this study the phonetic properties of a text were operationalized by its acoustic characteristics rather than by the relative occurrence of distinct phonemes. Building on results from recent research on sound-meaning relations [5-7], it was hypothesized that a text’s average formant dispersion, i.e., the average distance between first and second formant across all vowels of a text, is a significant predictor for the text’s emotional tone as assessed by its readers.
Method
147 undergraduate Japanese students assessed the emotional tone of 75 Tanka. Tanka are a traditional form of Japanese poetry consisting of 31 syllables (or morae) arranged in a characteristic 5-7-5-7-7 pattern. Randomly selected from a corpus of 3094 poems, the Taka were separated into three categories: Two categories with extreme values for formant dispersion that were either above or below the general mean ± standard deviation, and a neutral category that had a formant dispersion within the range of the general mean ± standard deviation. In what follows, the three categories will be referred to as bright, dark, and neutral for Tanka with an extremely high, an extremely low, or a neutral formant dispersion, respectively. Each category contained 25 Tanka. Every participant assessed pseudo-randomly selected 20 Tanka on six bi-polar items comprising the three-dimensional EPA-model (Evaluation, Potency, Activity) [8]. To analyze the data, Linear-Mixed-Models per EPA-dimension were conducted, with the stanza’s Category as a fixed factor and Participants as random factor.

Results
The results generally support the hypothesis. The average ratings per Tanka category were significantly different for the dimensions of Potency and Activity. Moreover, the results also confirm previous findings from a study of sound-meaning relationships in German poetry [9]. On the other hand, the relationship between the evaluation dimension and average formant frequency is not significant, and the regression indicates a negative correlation, which is inconsistent with the results reported for German poetry.
A closer look at the results per item underlines the strong correlation between formant dispersion and the dimension Potency, which has already been reported in many studies (for reviews see [6, 10-11]). On the other hand, the results for Evaluation and Activity show significance only for one item per dimension. Moreover, the results of the dark-bright item indicate that the Tanka with a relatively small averaged formant dispersion were perceived as brighter than Tanka with a normal or wide formant dispersion. This is even more surprising since studies have repeatedly reported that vowels with a small formant dispersion are perceived as dark sounding and vice versa [12-13]. Thus, further studies will be necessary to see if this result is specific only to Japanese.

Conclusion
Our data clearly suggest that phonetic features of Tanka can be used as an indicator of the emotional tone of the textual content. The fact that similar results have been reported for German poetry suggests that at least some sound-meaning relationships are universal in poetry. However, whether similar effects can be found in other languages or language families and for other genres requires further investigation.

References
1.Hinton, L. et al. (1994). In J. Nichols, J. et al. (Eds.), Sound symbolism (pp. 1–14). Cambridge: Cambridge University Press.
2. Schmidtke, D. S. et al. (2014). Frontiers in Psychology, 5, 80. Frontiers | Phonological iconicity | Psychology
3. Tsur, R., & Gafni, C. (2019). Methodological issues in the study of phonetic symbolism. Scientific Study of Literature, 9(2), 195-229. Methodological issues in the study of phonetic symbolism | John Benjamins
4. Auracher, J. (2019). Scientific Study of Literature, 9(2), 230-239. Commentary on Tsur and Gafni | John Benjamins
5. Auracher, J. (2017). PloS One, 12(11), e0187196. Sound iconicity of abstract concepts: Place of articulation is implicitly associated with abstract concepts of size and social dominance
6. Auracher, J. et al. (2019). PloS One, 14(5), e0216930. Contiguity-based sound iconicity: The meaning of words resonates with phonetic properties of their immediate verbal contexts
7. Hoshi, H. et al. (2019). i-Perception, 10(4), 2041669519861981. https://doi.org/10.1177/2041669519861981
8. Osgood, C. E. et al. (1978). The measurement of meaning. Urbana-Champaign: University of Illinois Press.
9. Auracher, J. et al. (2020). Cognitive Science, 44(10), e12906. https://doi.org/10.1111/cogs.12906
10. Nuckolls, J. B. (1999). Annual Review of Anthropology, 28(1), 225–252. https://doi.org/10.1146/annurev.anthro.28.1.225
11. Tsur, R. (2006). Journal of Pragmatics, 38(6), 905–924. Redirecting
12. Newman, S. S. (1933). The American Journal of Psychology, 45(1), 53–75. Further Experiments in Phonetic Symbolism on JSTOR.
13. Cuskley, C. et al. (2019). Behavior Research Methods, 51(4), 1651–1675. Cross-modal associations and synesthesia: Categorical perception and structure in vowel–color mappings in a large online sample | SpringerLink