The article sets up a twofold modelling of verbal communication: i.e. syntactic computing, and textual computing. Syntactic computing consists in producing/interpreting the elementary units which give the message its basic structure. Contrary to commonly received ideas, it is argued that those elementary units follow the ‘sentence’ pattern, assuming that a sentence may include not only the core (propositional structure) but also some peripheric elements. Both speaker and hearer try (successfully or not) to build (/reconstruct) well-formed sentences, in a step by step (‘sequence’ by ‘sequence’) process. Textual computing deals with the immediate integration of processed sequences (which lose their previous autonomy) into a higher level of textual representation. The analysis of the oral text (interview of Anita Musso) shows that the sentence level is easily recognisable (through unequivocal cores, and peripheric elements identifiable as such) and clearly plays a basic structuring role, which no alternative unit is likely to play. Textual units of a higher level than the sentence level are much more difficult to determine and to describe, for at least three reasons: (a) sequences may provide heterogeneous materials; (b) embedding sequences in order to build a higher representation is a continuous process, which leads to complex hierarchical structures; (c) speakers often follow a non-consistent expository path (as is the case in the text under study). Intonation has been deliberately left aside: the aim was to extend a purely segmental analysis as far as possible towards a complete account of the text.