Durational Patterns of Recurrent Multiword Combinations in Mandarin Spontaneous Speech Production
This study examined the prosodic realization of recurrent multiword combinations (RMC) in Mandarin spontaneous speech production and asked (a) whether speakers produce RMCs differently compared to novel combinations, and (b) how the RMC durational patterns are connected to its distributional properties and constituent structures. RMCs were first defined based on their distributional criteria in a large representative corpus, and a subset of these RMCs used in a phone-aligned spontaneous speech corpus were identified for the analyses of the relationship between their duration on one end and their distributional statistics (RMC frequency and lexical associations in two directions) and constituent structures (projected constituent level and boundary) on the other. The results suggest that Mandarin speakers are sensitive to the multifaceted multiword distributional properties, which are mediated by the constituent structures of RMCs. We discuss how these distinct durational patterns contribute to our understanding of the pragmatic and interactional role of multiword units in language processing and development.