Variation in speech tempo and its relationship to prosodic boundary occurrence in two speech genres

The present study focuses on two problems connected with speech tempo. First, earlier research has been prevalently concerned with central tendencies while variation was mostly perceived as an auxiliary result. We believe, however, that information about data dispersion is essential for proper modelling and experiment design in the field of temporal structure of speech. Therefore, the present study provides reference values for some of the tempo metrics of variation that pertain to (a) between-genre differences, (b) within-genre differences, (c) inter-speaker differences, and (d) intra-speaker differences. Second, we tested the claim that faster tempi lead to fewer prosodic breaks in spoken texts. This claim had been supported by studies where a respondent was asked to produce the same text at various rates. We, on the other hand, pose a question of the number of prosodic breaks in speakers who are fast or slow inherently. The material used in the study represents two genres: poetry reciting and news reading, and we obtained recordings from 24 speakers in each genre. Apart from providing the quantifications, the outcomes suggest, for example, that the predisposition of individual speakers to produce fast or slow tempi differs between the two genres. The fastest speakers in news reading were not necessarily the fastest in poetry reciting. This result points at specific behaviour in different situations and invites caution concerning the idea of hard-wired speaking stereotypes in individuals. Also, the correlation between speakers’ rates and the number of phrases they produced was significant only in news reading, not in poetry reciting. This result was corroborated by greater variation in prosodic boundary placement in news reading. In addition, the results offer an insight into the relationship between articulation rate and speech rate, together with the comparison of measurements in syllables per second and phones per second. The latter can be of interest since Czech (the language of the material) belongs to languages with a complex syllabic structure.