Using a replication task to study prosodic highlighting

Résumé En

It is hypothesized that prosodic highlighting (the accenting of a constituent for reasons other than rhythmic, i. e. focus marking or expressive emphasis) can be best studied by asking subjects to act out the script of a conversation, either reading it aloud or performing it from memory. Four subjects “replicated” a previously recorded and transcribed spontaneous conversation in French, thus allowing for a three-way comparison between spontaneous, read and interpreted speech using the exact same text. A group of prosody experts annotated the occurrences of prosodic highlighting in each recording. The results confirm the hypothesis on one count but not on the others. The frequency of occurrence of prosodic highlighting is, as expected, highest in interpretation, followed by reading. However, mean F0 and mean syllabic duration of the annotated words do not follow the same gradation. On the phonological side, there are no differences in the distribution of prosodic contours present on the annotated words, as well as a few other features.