Some observations on different strategies for the timing of fundamental frequency events

The acoustic manifestations of the prosodic features are heavily influenced by the segmental compositions of the utterances. In order to study the acoustic correlates of the prosodic features it is necessary to diminish the influence of the segments. In investigations about the fundamental frequency contour utterances built up of only sonorants are often used and thereby the influence of constrictions in the vocal tract is avoided or at least diminished. In order to arrive at a detailed model of the fundamental frequency for e.g. speech synthesis it is necessary to exactly know what is happening in different segmental environments. This information can also shed light on the underlying mechanisms of the timing of the prosodic patterns. In the present study we are examining the effects which voiceless consonants have upon the fundamental frequency contour not only in words carrying the tonal word accents but also words carrying in sentence accent and terminal juncture. Three different hypotheses are considered the truncation, the timing adjustment and the rate adjustment hypothesis. The results show that the truncation hypothesis seems to be of limited validity at least for a detailed description of what is happening in the fundamental frequency contour in Swedish.