Perceptually informed quantification of speech rhythm in pairwise variability indices.

Two previous experiments demonstrated that f0 and duration are interdependent in the perception of rhythmic groups in speech and sentence rhythmicality, and that the relative weighting of tonal and durational cues depends on listeners' native language. The listeners were native speakers of Swiss German, Swiss French, or Metropolitan French (i.e. from France). The experiment reported here investigates a means of applying this perceptual finding to production data from these three languages, to make a rhythm metric, the Pairwise Variability Index (PVI), perceptually informed. The relative weighting that an appropriate duration and an appropriate f0 contributed to listeners' rhythmicality judgements is calculated, and these language-specific weighting values are incorporated into combined durational- tonal PVIs, to quantify rhythm in the three languages. The results demonstrate that Swiss German and Swiss/Metropolitan French are distinct according to classic durational PVIs, but more similar according to PVIs which are acoustically multidimensional and language-specifically weighted. It is concluded that rhythm produced by speakers, when quantified to account for the acoustic multidimensionality and language-specificity of rhythm perceived by listeners, may be less cross-linguistically divergent than durational rhythm metrics suggest. An evaluation of these language-specifically weighted PVIs concludes that if rhythm metrics remain in use, they should link rhythm perception with rhythm production.