The contributions to this volume represent the state of the art of computational analysis and modeling of spontaneous speech. Many research laboratories around the world are now working on methods of processing natural speech, so that machines can interpret what people say, rather than making people adapt their speech to a style that machines can understand. The book is in four main sections. The first gives the theoretical background and discusses how spontaneous speech differs from the laboratory-speech that has been the focus of earlier analyses. The next part focuses on prosody and the structure of the spoken message, and the third discusses the generation and modeling of prosody for speech synthesis. The concluding part considers the use of prosodic information in automatic speech recognition. Each part begins with an overview of the field; other chapters are based on a workshop on computational approaches to processing spontaneous speech and represent the current state of research.