An Emerging Gender Difference in Japanese Vowel Devoicing 181of the standard (males) and in atypical situations (informal), it is necessary
to look more closely at the social factors that might affect Japanese vowel
devoicing, including age, sex, and speech style.
- Data
3.1 Data collection
The data used in this study came from 21 men and 21 women, all of whom
grew up in the Tokyo area. Most of them were also born there and have at least
one parent who grew up in Tokyo. There are three age groups: 15 were in the
younger age group, 14 were in the middle group, and 13 respondents made up
the older group. Originally, I also included three social classes: 14 participants
were considered working class, 16 were classi¿ ed as lower middle class, and
12 respondents made up the upper middle class group. However, this distinc-
tion did not show a statistical signi¿ cance.
Data were collected through sociolinguistic interviews, which were recorded
on tape. An interview consisted of three stages. First, I asked the respondent
some demographic questions—where they were born and raised, where their
parents were born and raised, their occupation, their parents’ occupation, where
they live now, what they like to do for leisure, and so on. Then they were asked
to read a word list and a reading passage. The word list consisted of 90 words,
phrases, or short sentences, containing all the possible combinations of the pre-
ceding and following voiceless consonants, including a pause, for both high
vowels. The reading passage used as many words from the word list as pos-
sible without becoming too long (1.5 pages). The entire interview session was
recorded, and they ranged from 20 minutes to 2.5 hours in length.
3.2 Data analysis
The recorded data were transferred and digitized using the sound analysis pro-
gram Praat in order to determine the voicing of the vowels. After the acoustic
analysis was done, all the necessary data was input in an Excel spreadsheet,
and a statistical analysis was done using Goldvarb (a logistic regression pro-
gram) to ¿ nd out signi¿ cant factors and the relative signi¿ cance of the values
within each factor. Goldvarb is capable of dealing with the very small num-
bers in some cells that may arise in the study of conversational data. The total
number of tokens (vowels) used for the statistical analyses is over 30,000.