Discourse segmentation and discourse markers in spoken French

In this talk I will give an overview of my work on spoken discourse segmentation and on discourse markers in order to explore further research avenues about what is so specific about the discourse level, i.e. to make progress in investigating why discourse is a crucial notion for understanding human communication (Sanders & Spooren, 2007). On the basis of my (mainly) functionally oriented corpus-based work on DMs (both onomasiological and semasiological) (see e.g. Bolly et al., in press; Degand, Cornillie & Pietrandrea, 2013; Degand & Fagard, 2011; Pander Maat & Degand, 2001) I would like to develop the idea of “the paradox of discourse markers”: On the level of the sentence (syntax and semantics), DMs are optional, they are not even considered as a (morpho-)syntactic category; on the pragmatic level however, they are communicatively obligatory (in the sense of Diewald 2011). It follows that the study of discourse markers should learn us more about the underlying cognitive and functional principles of human communication, i.e. as indices of fundamental cognitive processing during (spoken) language production. Uncovering cognitive principles of (spoken) language production also underlies my plans regarding further work in the area of discourse segmentation. Discourse segmentation in units is a crucial process in order to understand discourse production and comprehension. We developed a method for segmenting spoken discourse in Basic Discourse Units (BDUs), based on the interaction between syntactic units (dependency clauses) and prosodic units (major intonation units). BDUs result from the coincidence of syntactic and prosodic boundaries, corresponding to distinct but complementary linguistic encodings. This mapping gives rise to different types of discourse units (congruent, syntax-bound, intonation-bound, regulatory) (Degand & Simon 2009a, Simon & Degand, 2011). Thus, our claim is that the prosody-syntax interface gives rise to a distinctive discursive level of analysis contributing to the unfolding (linear) discourse, e.g. in the form of different discursive strategies (Degand & Simon, 2009b; Martin, Degand, & Simon 2014). The BDU segmentation has been applied to a corpus of spoken French, LOCAS-F (Degand, Martin & Simon, 2014; Degand & Simon, in prep.) comprising 14 different speech situations (political debate, interview, spontaneous conversation, conference, …). The data is now available to explore whether these BDUs have cognitive validity in production and/or comprehension