An Acoustic Study of the Japanese Short and Long Vowel Distinction

Abstract

Abstract The present study concerns the vowel length distinction between Japanese short and long vowels. Previous studies (Han 1962, Hirata 2004, Kozasa 2005) concluded that short and long vowels differ from each other in terms of duration and pitch fall; that is, long vowels are about 2.4 times longer than short vowels, and long vowels have a pitch fall within a word while short vowels do not. Most studies confirmed the differences between short and long vowels when they are accented. However, none of the previous studies has investigated the differences between short and long vowels using the same test words when a vowel is unaccented; that is, when a pitch fall is absent. The present study concerned the environment where there is no pitch fall within the target vowel: the compound noun context; where HL pitch accent becomes HH, and H becomes L. An acoustic study was conducted in order to investigate the difference between short and long vowels in Japanese in two different contexts, the accented context, and the unaccented context, in which the target word is compounded with the suffix [jo:] in order to neutralize the pitch accent High (H) to Low (L) and HL to LL. The target vowels in the accented context are labeled H for short and HL for long vowels, and the ones in the unaccented context are labeled L for short and LL for long vowels. A wordlist reading task was used for the production study. Measurements were made for pitch onset (F0 value at onset of target vowel), pitch offset (F0 value at offset of target vowel), pitch fall (onset- offset), vowel duration, F1, F2, F3. Analyses showed that F0 neutralization in the unaccented context was incomplete as measured by F0 onset and F0 offset. However, pitch neutralization from H to L as measured by pitch fall was observed. In addition, long vowels were longer than short vowels, and longer in the unaccented context than in the accented context. Pitch fall was significantly larger in the accented context and statistically nonexistent in the unaccented context. Although accented vowels were significantly longer than unaccented vowels, the ratio of short and long vowels was 1:2.4 and 1:3.2 in accented and unaccented contexts, respectively. The difference between short and long vowels was larger in the unaccented context. In terms of vowel quality, F1, F2 and F3 all three significantly differed between short and long vowels in both the accented and unaccented contexts. F1 was significantly higher in the accented context than in the unaccented context and the difference between short and long vowels was significantly larger in the unaccented context. F2 was significantly lower in the accented context than in the unaccented context and the difference between short and long vowels was significantly larger in the unaccented context. The F3 difference between short and long vowels was significant in both accented and unaccented contexts. This difference, however, was larger in the accented than the unaccented context unlike F1 and F2. When vowels were unaccented, the speakers showed larger differences between short and long vowels in F1 and F2, presumably due to the lack of pitch fall.