Towards a unified predictive model of natural language rhythm

This paper presents a natural language rhythm model, conceived so as to comply with the basic epistemological requirements of: explicitness, predictivity, unification. The last requirement refers to the fact that, over and above terminological convenience, the competing rhythmical types (such as the traditional constrast ‘syllable- vs. stresstiming’) should be regarded as the two extremes of a continuum, rather than radically alternative types. The proposed model is based on two independent levels: level-I (phonotactic) and level-II (sentential). At each level, languages may be characterized as more or less ‘controlling’ or ‘compensating’, along a continuum that ideally goes from a maximum of rigidity to a maximum of flexibility. Most importantly, languages may present competing tendencies at the two levels. This possibly accounts for the often elusive character of the rhythmic tendencies of the individual languages. As far as level-I is concerned, a new algorithm (Control/Compensation Index) is proposed in order to check the rhythmical inclination of the languages. As for level-II, the O’Dell and Nieminen algorithm is exploited. Although provisional, the results obtained demonstrate that it is possible to base research on speech rhythm on entirely predictive models, allowing for direct falsifiability.