Abstract:
Recently, it has been shown that segments are phonetically strengthened at the beginning of prosodic units [Jun (1993) (for Korean); Fougeron and Keating (1995) (for English); Fougeron (1996) (for French)]. This study investigates whether the syntactically defined Tone Sandhi Group (TSG) in Taiwanese is part of the prosodic hierarchy based on the pattern of phonetic strengthening. VOT and closure duration of word initial /p[sup h]/ are compared in five prosodic conditions, varying in the position within the intonational phrase (IP) and TSG. Preliminary results show that VOT and closure duration of /p[sup h]/ are longer IP-initially than IP-medially. Furthermore, VOT, but not closure duration, is consistently longer TSG-initially. Interestingly, both VOT and closure are shorter TSG-finally: Final weakening marks the TSG boundary, just as initial strengthening marks the IP boundary, suggesting that TSG is a prosodic unit. However, incorporating TSG into the prosodic hierarchy is problematic for the strict layer hypothesis, which states that a lower-level prosodic unit is exhaustively contained within a higher level prosodic unit. An alternative is to assume that phonetics may mark both nonprosodic (TSG) and prosodic (IP) boundaries. [Work supported by NSF.]