Coronal Stop Deletion (CSD) is believed to be constrained by the phonological environment and the word’s morphological make-up (i.e., monomorphemic, semi-weak past tense, regular past tense) (Guy 1991; Hazen 2011). In this talk, we present results from an analysis of CSD in the speech of four speakers of Hawai'i English. In addition to using a traditional approach (i.e., auditory analysis of CSD in word-final position), we also analyzed CSD in non-final position (e.g., followed by plural -s) and conducted acoustic phonetic analysis of /t/ in tokens where it was realized. The results demonstrate highly significant effects of environment. Additionally, there is no effect of morpheme type; instead, the results provide evidence that deletion is favored in words with a greater number of morphemes, regardless of whether the coronal stop is found in the stem or the affix. Additionally, the acoustic analysis suggests that the effect of following environment is acoustically gradient: tokens followed by a voiceless, non-alveolar consonant (i.e., the environment that most strongly disfavors deletion in our data) are most likely to be realized with complete occlusion in the vocal tract. Implications for models of speech production are discussed.

Hawai‘i is home to many languages, including an English-lexified creole, known locally as Pidgin. Despite much work on the development of Pidgin, little is known about the acoustic properties of Pidgin consonants. This talk presents results from an analysis of word-initial plosives produced by eight speakers of Pidgin across two age groups, investigating what factors influence VOT and closure duration in the plosives. The data were analyzed using linear regression models with by-speaker random intercepts. Results demonstrate that VOT in word-initial plosives tends to be shorter in Pidgin than in English, with mean duration values nearly half those reported for English in Yao (2007). The effects of social and linguistic factors, however, tend to be similar (e.g., shorter VOT for older speakers, VOT for /k/ is longer than for /t/ or /p/). Results also indicate that VOT is significantly shorter both for words of Hawaiian origin (β =-0.01, t(289)=-2.0, and p<0.05) and—in a separate model fit to a subset of the data with only non-Hawaiian words—highly frequent words (β=-0.01, t(216)=-4.2, and p<0.001). In contrast, no factors tested significantly predicted closure duration. Taken together, results demonstrate that a combination of linguistic, social, and probabilistic factors influence VOT in Pidgin.