LingLunch : Georgia Zellou

Georgia Zellou (UC Davis)
The role of coarticulatory variation in speaker recognition

Much prior work reports that there is a large amount of cross-speaker heterogeneity in patterns of produced anticipatory nasal coarticulation within a single speech community (Yu, 2019; Beddor et al., 2018; Zellou, 2017). How do listeners accommodate talker-specific variation in coarticulation? We propose that listeners encode talker-specific coarticulatory information. More specifically, we hypothesize that since increased coarticulation provides linguistically-relevant information (e.g., Beddor et al., 2013) there will be variation in listeners? ability to use coarticulatory patterns across speakers -- voices in which there is greater extents of coarticulatory overlap will be easier to recognize than voices in which there is less coarticulatory information. In order to test this, the present study applies the traditional methods of a talker recognition study (e.g., Mullennix et al., 2011) to explore whether cross-talker voice recall varies based on differences in produced coarticulatory vowel nasality in American English.