An acoustic and articulatory study on variation of high vowel devoicing across prosodic contexts and speakers in Korean

Abstract

The present study investigates how high vowel devoicing in Korean is phonetically manifested as a function of prosodic boundary and prosodic prominence. In order to explore the nature of devoicing and the degree of devoicing in relation to prosodic structure, devoicing proportion was calculated by dividing the acoustic duration of an onset consonant by that of an entire syllable in the first syllable of CV.CV target words. In addition, the tongue height was measured for the same set of target syllables in various prosodic positions, using EMA (Electromagnetic Articulography), and the correlation between the acoustic devoicing proportion and the articulatory tongue height was examined in order to explore whether there is a direct biomechanical influence of tongue height on the degree of devoicing. The results of acoustic data showed that devoicing proportion was gradually distributed (from 22 to 100%), showing phonetic nature of devoicing process rather than a categorical phonological rule application. For the completely devoiced tokens which constituted 20% of the devoiced tokens, a greater number of tokens were found in the prosodically weak position (i.e., phrase-medial or unfocused tokens). The prosodic influence on devoicing is in line with previous findings on other coarticulatory processes, supporting the idea that devoicing is phonetic coarticulation caused by gesture overlap in Korean. The prosodic effects on the degree of devoicing, however, differed across speakers: Some speakers showed less devoicing in the prosodically strong positions (i.e., domain-initial or focused positions), while others showed more devoicing in the same positions. The speaker variation lends further support to the phonetic status of devoicing in Korean and also imply the presence of speaker-specific phonetic grammar. In addition, the correlation between the acoustic devoicing proportion and the articulatory tongue height revealed only a subtle correlation, indicating that tongue height did not directly influence devoicing. Prosodic modulation on the tongue height also showed speaker variation. Heightened tongue position in a prosodically strong condition did not result in more devoicing, suggesting that higher tongue position does not necessarily provoke more devoicing due to their biomechanical relationship. Instead, higher tongue position was often matched with less devoicing (i.e., more voicing) under focus, which is in line with the enhancement of both features [+voice], [+high] of the vowel. These results may show that the high tongue position did not directly cause devoicing, implying that prosodic modulation on the glottis appears to follow speakers’ intention rather than following biomechanical causality from the tongue height, although that there are speaker variations.