Prosodic boundaries in German: Final lengthening in spontaneous speech.

Abstract

Most theories of prosodic structure postulate at least two phrasal categories above the word level, a minor and a major one. One correlate of phrasal boundary marking is lengthening on the right edge of a phrase. To gain a theory neutral understanding of the nature of prosodic boundaries, a Gaussian mixed model (GMM) was applied to durational data, estimating the underlying clusters of a continuous distribution. Spontaneous speech data were collected with standardized interviews from 19 adolescent speakers of a multi-ethnolect (Kiezdeutsch, Hood German) spoken in Berlin, orthographically transcribed and added to a database that allowed for searches of the particle “so”(so or like) in different conditions. Acoustic durations of /z/ and /o/ of so in phrase final position (as determined by orthographical markings or by following pauses) were labeled and z-transformed per speaker. The results show that a model including two clusters accounts best for the data. Nevertheless, there is an overlap between the two distributions supporting the gradient nature of boundaries. Further analyses with more tokens of so including word level junctures are currently under way. The impact of our data-driven results on theoretical models of prosodic boundaries is discussed. [Supported by a grant from BMBF.]