On the controlled elicitation of spontaneous speech

This paper presents an elicitation method that allows the controlled analysis of properties of spontaneous speech. The method is characterized by the fact that a speaker is constrained by the experimeter with respect to what he will say, while keeping his speech output spontaneous. As an example of the potential of this technique, a pilot study is presented in which such spontaneous discourse was used in prosodic research. The purpose of this simple experiment was to find out whether the endings of large-scale units in spontaneous discourse have particular melodic correlates that are communicatively relevant.