Defended on 15-11-2012, in Nice.
This dissertation consists in the study of modals and semi-modals in a corpus of American presidential debates (1960-2008). Our aim is to characterize their use with respect to issues of power and influence. On the one hand, the theoretical framework borrows from various fields of linguistic analysis, namely enunciation, pragmatics and discourse analysis; the latter helps define notions such as persuasive communication, manipulation and influence strategies. Besides, we suggest that the modals and some semi-modals can activate dialogic echoes and serve as evidential markers. On the other hand, the study relies on a computer-based discourse analysis methodology, namely logometry. Considered as a valuable heuristic approach, it runs statistical measures on linguistic data – be it tokens, lemmas or parts of speech – so as to provide the analyst with frequency lists and cooccurrence networks, among other functionalities. Logometry also offers a bottom-up as well as a top-down approach to corpus scrutiny. The results help identify three enunciative strategies depending on the level of commitment of the utterer with his epistemic or non-epistemic judgements. Also, they determine the existence of several types of discourses that are considered in the light of various issues: speech practice evolution, ideological positioning, rhetorical identity of political parties, candidates’ campaign strategies. As far as the modals and semi-modals CAN, MUST, HAVE TO, HAVE GOT TO, NEED TO and OUGHT TO are concerned, a close examination of their contexts of use shows that they reflect different strategies of presentation of the self, of others, of the audience and of the factual context.
View online : Thèses.fr