Empirical models of tone, rhythm and intonation for the analysis of speech prosody.

This contribution is a reflection on criteria for an empirical framework for the investigation of speech prosody across different languages and, as far as possible, across different theoretical points of view. It is proposed to define different units, called the Tonal Unit [τ] and the Rhythm Unit [ρ], as the domains of short-term planning of pitch and timing, respectively, as well as a larger Intonation Unit [IU] for longer-term prosodic characteristics like changes of register and tempo. This makes it possible to compare models and their implementation. A comparison of Jassem's model of rhythm with that of Halliday on a 5.5 hour database of spoken English (Aix-Marsec) largely confirmed several of Jassem's predictions. The same database is now being used to investigate the short term planning of pitch within the Tonal Unit.