The Impact of Palatalization Rate on Phonological Variants: A Corpus-based Study of Word-final /t/ in American English

This study investigates whether individual speakers' varying palatalization rates in the environment where word-final /t/ is followed by words beginning with /j/ affect the distribution of /t/ variants in other environments in American English. Based on data from the Buckeye Corpus, we identified and classified word-final /t/ variants across four different environments, such as word-final /t/ after vowels followed by word-initial vowels, liquids, nasals, and obstruents, respectively. A total of 21,627 tokens were analyzed, and variants of /t/ were categorized as palatalized, glottalized, flapped, deleted, or unchanged. We applied Multinomial Logistic Regression to explore correlations between palatalization rates and the distribution of /t/ variants in the four environments. Results revealed that palatalization rates had minimal impact on flapping except when preceding liquids, where a proportional correlation was observed. Unchanged /t/ exhibited a proportional correlation with palatalization rates in most environments, but showed a weak inverse relationship when followed by liquids. Deletion consistently correlated proportionally with palatalization across all environments, while glottalization demonstrated a significant inverse correlation. Ultimately, as palatalization rates increase, deletion and unchanged /t/ become more favored, while glottalization is less likely to occur.