A comparison of multiple speech tempo measures: Inter-correlations and discriminating power

Studies that quantify speech tempo on acoustic grounds typically use one of various rate measures. Explicit comparisons of the distributions generated by these measures are rare, although they help assess the robustness of generalisations across studies; moreover, for forensic purposes it is valuable to compare measures in terms of their discriminating power. We compare five common rate measures ― canonical and surface syllable and phone rates, and CV segment rate ― calculated over fluent stretches of spontaneous speech produced by 30 English speakers. We report deletion rates and correlations between the five measures and assess discriminating powers using likelihood ratios. Results suggest that in a sizeable English corpus with normal deletion rates, these five rates are closely inter-correlated and have similar discriminating powers; therefore, for common analytical purposes the choice between these measures is unlikely to substantially affect outcomes.