This paper explores deliberative and automatic processes in phonetic imitation across geographic dialects. A group of 136 American English speakers shadowed an Australian model talker and were instructed to either imitate her, avoid imitating her, or neither. Vowel quality of BATH, TRAP, KIT, and PRICE was analyzed for change in formant-difference values (F2 - F1) from baseline to shadowed token, as was rhoticity in the word class NEAR (F3 - F2). Participants showed a high degree of control, greatly increasing imitation when asked to do so and eliminating it entirely when asked to avoid imitation. Alongside deliberative control, sociolinguistic meaning also emerged as crucial. The two most well-known dialect features, BATH and NEAR, showed the highest degree of imitation in the Imitate condition, and NEAR was the only feature imitated when no instruction regarding imitation was given. Further, participants who made explicit mention of either BATH or NEAR in the post-task questionnaire showed increased imitation on the feature in both the Imitate and the No Instruction conditions. Overall, the results suggest that participants have a high degree of control over imitation in the shadowing context but that sociolinguistic meaning also plays a strong role.
