Speaking Rate Variation and the Interaction between Hierarchical Rhythmic Levels

This paper presents linear models built for the purpose of investigating tempo-dependent interaction between two components of Polish speech rhythm, namely Rhythmic Prominence Intervals (RPI) and syllables, within the framework of coupled oscillator model ([1], [2], [3]). The relative coupling strength parameter (r) calculated from regression coefficients shows increasing strength of the lower-level syllabic oscillator with increasing tempo and, conversely, greater dominance of the RPI oscillator at slower speaking rates. These findings indicate the coexistence of different timing strategies or patterns within one language and add to the evidence against the traditional, categorical division of languages into syllableand stress-timed (e.g., [4]). Altogether, the results of our models indicate a strong linear relationship between syllable count in RPI and its duration across different speaking rates (very slow – very fast) and reveal some interesting properties of the rhythmic/timing structure of Polish, e.g. greater flexibility of tempo deceleration comparing to other languages. The models account for between 85% and 94% of the variability in RPI duration (based on adj. R) when speaker is added as a random factor, or when the laboratory measured speaking rate (in syll./sec.) is used instead of intended speaking rate (tempo class).