An Emerging Gender Difference in Japanese Vowel Devoicing 179Figure 6.2 shows a devoiced token of the same word. There are no for-
mants, periodic wave peak, pitch track, nor intensity movement where the
vowel /u/ should be. There is a small intensity rise where /ts/ is released, but
nothing else. This is the devoiced /u/.
There were cases in which it is hard to determine voicing. In this study,
if three of the preceding criteria were met, the vowel was coded as devoiced.
For example, if there were no periodic wave, no clear formants, no intensity
(and no audible voicing), but there was a pitch track in the vowel’s position, I
nevertheless regarded the vowel as devoiced.
2.2 Factors that affect vowel devoicing
There are many factors that affect the likelihood of devoicing; linguistic fac-
tors include the preceding and following consonant identity, pitch accent of
the target vowel, the existence of potentially devoiceable vowels before and/
or after the target vowel, and the morpheme boundary type. The preced-
ing and following consonant identity is important because vowel devoicing
occurs between voiceless consonants, and previous studies show that there
is a difference in their effect on vowel devoicing depending on the type
Figure 6.2 /see+katsu#hi/ (‘living expenses’) with a devoiced vowel.