A corpus-based study of geminate devoicing in Japanese: linguistic factors

要旨

Abstract Ever since Nishimura, 2003 , Nishimura, 2006 pointed out that voiced geminates can optionally devoice in Japanese when they co-occur with another voiced obstruent (e.g. /baddo/ → [batto] ‘bad’), the pattern has been analyzed within a number of theoretical frameworks and studied in several experimental studies. However, there are only a few studies on actual production data. Moreover, most of the previous studies have generally assumed that this pattern is a linguistically monolithic phenomenon—all OCP-violating geminates would have equal probability of devoicing. By studying the Corpus of Spontaneous Japanese ( Kokuritsu-Kokugo-Kenkyuujo, 2008 ), we show that many linguistic factors affect the probability of the devoicing of voiced geminates: in addition to OCP (voice), we find effects of the location of the trigger with respect to the target, the number of triggers, place of articulation, and lexical usage frequency. All of these observed patterns accord well with phonetic considerations and/or cross-linguistic tendencies. We conclude that geminate devoicing in Japanese phonology is not a linguistically monolithic phenomenon, because the probability of devoicing is affected by several linguistic factors. We suggest that future analyses of this phenomenon should take into account the factors that are identified in this project.