Relationships Between Speech Timing and Perceived Hostility in a French Corpus of Political Debates

This study investigates the relationship between perceived hostility and speech timing features within extracts from Montreuil’s City Council sessions in 2013, marked by a tense political context at this time. A dataset of 118 speech extracts from the mayor (Dominique Voynet) and four of her political opponents during the City Council has been analyzed through the combination of perception tests and speech timing phenomena, estimated from classical timing-related measurements and custom metrics. We also develop a methodological framework for the phonetic analysis of nonscripted speech: a double perceptive evaluation of the original dataset (22 participants) allowed us to measure the difference of hostility perceived (dHost) between the original audio extracts and their read transcriptions, and the five speakers produced the same utterances in a controlled reading task to make the direct comparison with original extracts possible. Correlations between dHost and speech timing features differences between each original utterance and its control counterpart show that perceived hostility is mainly influenced by local deviations to the expected accentuation pattern in French combined with the insertion of silent pauses. Moreover, a finer-grained analysis of rhythmic features reveals different strategies amongst speakers, especially regarding the realization of interpausal speech rate variation and final syllables lengthening.