Contrasts, topic shifts, and prosodic variation in conversational data
Fragmentation between ´formal´/quantitative and ´functional´/qualitative approaches is a well-known problem in linguistic research. The current study combines a quantitative approach (drawn from experimental phonetics/phonology, EP) and a qualitative approach (drawn from Conversation Analysis, CA) to investigate prosodic variation in conversational speech. The notion of ´contrast´, where two stretches of speech are partly the same, partly different, and compared as regards the differences (Mann & Thompson 1992) has been investigated in both EP and CA approaches. Similarly, the global organization of discourse around topics has been relevant for both approaches, particularly as regards variation in prosodic features such as pitch and speech rate. Contrastive structures can be used as a tool to effect topic change in conversation, providing an entry point to the investigation of conversational data using both qualitative and quantitative methods. CA principles for identifying stretches of speech that have the same function were used to identify a collection of contrasts from recordings of conversations. Then statistical analyses following the principles of EP research were applied to the set of contrasts. Finally, CA analysis was applied to explain findings and address exceptional cases and outliers. This combined methodology is shown to be sensitive to similar kinds of prosodic variation as have been identified in lab-based EP studies, while more flexible in that it can motivate sources of variation in individual stretches of conversation.