This paper attempts to translate two phonological models of prosodic structure into quantitative predictions which can be empirically tested on a large corpus of spoken English. Specifically the Abercrombie/Halliday model of the stressfoot is compared to the Jassem model of (narrow) rhythm unit and anacrusis. The data analysed was a five and a half hour corpus of spoken English (Aix-Marsec). Preliminary results from this analysis suggest that the Jassem model is in nearly all cases superior to the Abercrombie/Halliday model, i.e. that it is the narrow rhythm unit and not the foot which is the essential component of the rhythm of British English. The data suggest furthermore that there is no specific lengthening for stressed syllables.