The research described in this thesis focuses on how speakers make use of prosodic aspects (such as accent lending and boundary marking pitch movements and pauses) to realize the structure of spoken discourse. Structure in this respect refers to discourse boundaries of different depths on the one hand, and to important information at the word level on the other. Furthermore, it was investigated how listeners make use of these prosodic cues to detect the structure of spoken discourse, again in terms of boundaries such as sentences and paragraphs, and informative words. To this end, Van Donzel first develops a text-based framework to analyze the structure of spontaneously spoken discourse. The application of this framework to the verbatim transcriptions of spoken discourse then provided a detailed analysis in terms of discourse boundaries and important information. The combination of i. the actual prosodic realization by the speakers, and ii. the structure perceived by the listeners, provided useful information about what prosodic means are used in the realization and perception of the structure of spoken discourse. The results of the present study show that speakers make use of boundary tones and/or pauses to mark discourse boundaries, dependent on the depth of the boundary. Pauses are important for listeners to decide where boundaries occur in the discourse. To mark important information at the word level, speakers mainly make use of pitch accents. Information that is new to the discourse is realized with a pitch accent more often than information that adds little to the content. Pitch accents are also indicative for listeners to perceive important information. This study is of interest to experimental phoneticians, as well as to researchers in the field of discourse studies and pragmatics.