(Avec L. Buscail) "When text creates its own context: Personal and demonstrative pronouns, accenting and contrast in English"

International audienceUsing the ToBI system of annotation of prosodic events, this paper argues, contra Kameyama (1999), that it is not always possible or revealing to determine the preferred referent of a contrastively-stressed 3rd person pronoun or demonstrative in terms of the referential preferences of its unstressed counterpart. Instead, we claim that the presence or absence of contrastive pitch accents on demonstratives or personal pronouns may realise two main types of information structure within the indexical clause (respectively, "argument focus" and "predicate focus": Lambrecht, 1994). Each of these realisations is closely bound up with the search induced via the contrastive or weak accenting for a relevant coherence relation, both to establish the intended referent and to integrate the two discourse units at issue (cf. Hobbs, 1990; Kehler, 2004; Cornish 2009). A new discourse referent or a new perspective on an existing one is then introduced into the discourse model under construction, on occasion "abductively" (cf. Hobbs, 2004); this is achieved in terms of the contexts that their establishment both presupposes and creates, together with a (defeasible) conversational implicature (de Hoop, 2004). The natural interactive spoken data we present make it possible to establish the crucial role played by various aspects of context in the establishment both of the contrasting entity (in certain cases) and, more generally, of the alternative entity or set of entities which is/are being contrasted