For the purpose of examining regional typology of dialect usage, we analyzed data from a nationwide survey of language consciousness targeting Japanese residents aged 16 and over. We adopted latent class analysis in order to derive the individual-level probabilistic clustering of language usage. Based on the results we tried to identify the regional correspondence to these clusterings (latent classes) and scrutinized the relationship between demographic characteristics of respondents and these classes. The analysis yielded five latent classes, which can be conceived as follows: "Class 1 Active Dialect Speaker", "Class 2 Common Language Speaker", "Class 3 Moderate Code Switcher", "Class 4 Active Code Switcher", and "Class 5 Indecisive Respondent". Among the demographic characteristics, hometown, occupation, educational background, and age all had a statistically significant effect on class membership (with effect size running from largest to smallest in this order), while gender and size of city of residence did not. The regional distribution of respondents' hometowns showed that certain classes are typical for respondents who grew up in specific regions: "Class 1 Kinki, Chugoku and Shikoku Area", "Class 2 Tokyo Metropolitan Area and Hokkaido", "Class 3 Northern Kanto, Koshin-etsu, Hokuriku, and Tokai Area", "Class 4 Okinawa, Kyushu, Tohoku and Chugoku Area", and "Class 5 Hokkaido". Occupation also showed a specific relationship to some of the classes. With regard to age, we observed a non-linear relationship both in the analysis of the total sample and in the analyses by regions. Generally speaking, each age-effect curve has two or more change points, most of which are observed between the ages of 35 and 40 and at around 60. Based on these results, we discuss the implications of regional typology and the relationship between the results of our study and previous studies.
話者分類に基づく地域類型化の試み : 全国方言意識調査データを用いた潜在クラス分析による検討
papers0307.pdf
(1.1MB)
運用指針、オープンアクセス方針、等
