In most of the cases spontaneaously uttered units of speech (e.g. in face-to-face dialogues) contain performance phenomena like repairs, breaking offs, omissions and others that motivate a restructuring procedure which allows storage or further processing of the input. In our view, this restructuring procedure is based on the segmentation of the input into a set of functional (i.e. meaningful) units. These functional units should be cleared from redundant parts of the source text, non-overlapping in most of the cases, ordered hierarchically according to the importance of the content, consist of a comparable size. Segmentation above word boundary is no problem for written texts. Amtrup and Jekat 1995, p.298:Written language offers conventionalized methods for segmentation, e.g. splitting into sentences, phrases or - based on structural analysis - constituents. As for spoken texts, functional units seem to be an important basis for processing but there are no criteria for segmentation which are as evident as to written language. Nevertheless, the restructuring of the spoken input has to be carried out very fast especially in the case of interpretation when more than one language is processed. Recordings of dialogue interpreting give insight to the output of the restructuring procedure. Our hypothesis is that prosodic cues are very important for speech segmentation and help to identify meaningful functional units. In this paper we will demonstrate some effects of the restructuring procedure and draw up evidence for the importance of prosodic cues within the process of restructuring.