Within the large panel of theoretical and empirical approaches to discourse, the bulk of research on discourse markers and discourse relations targets functional classification, striving (and so far, failing) to achieve some consensus on categories and taxonomies meeting the requirements of exhaustivity, cognitive plausibility and operationality when applied to authentic data, either written or spoken (e.g. Prasad et al. 2008; Rehbein et al. 2016; [Author]). By contrast, the role of syntactic considerations in the definition and disambiguation of discourse markers is often neglected (see Danlos 2013; Webber et al. 2016 for a few recent exceptions). In particular, most authors refer to the “clause” as the minimal unit to enter discourse-level relations (as opposed to syntactic, sentence-level relations), assuming that such a concept is stable when it actually covers several borderline cases (e.g. verbless clauses, relative clauses). Furthermore, the status of subordinating conjunctions within the category is also questionable given that some of their uses are syntactically integrated in the dependency structure and therefore do not meet the optionality and weak-clause association criteria (Blühdorn 2008; [Author], [Author]). In this paper, we propose to address this gap and investigate the syntax-discourse interface, focusing in particular on the relation between position, degree of syntactic integration and function of discourse markers. In doing so, we aim at identifying borderline uses of discourse markers which should potentially be excluded from the category, although they are traditionally included in definitions where syntactic considerations are overlooked. Concretely, we will take grammatical category (e.g. coordinating and subordinating conjunction, adverb, etc.), syntactic position (e.g. initial, medial) and co-occurrence with pauses as evidence for different functional categories as defined and applied by [Author] to a comparable corpus of spoken English and French fully annotated for discourse markers and their features. The degree of association between some of these features and specific functions of discourse markers will be statistically tested and interpreted as a cue to their relative centrality in the category, taking high frequency of use as a measure of prototypicality. The specific behavior of ideational/semantic/objective relations (e.g. a causal relation between two facts) and that of subordinating conjunctions (e.g. because, although) observed in the corpus suggest a continuum between syntax and discourse which questions the systematic integration of these borderline elements in the category and functional spectrum of discourse markers. Our corpus-based results also draw attention to the specificities of spoken syntax (for instance interrupted or incomplete utterances, subclauses without a governing verb) and their impact on the identification and functional annotation of discourse markers.
Degand, L., & Crible, L. (2017). Revisiting the role of syntax in discourse marker annotation. 50th Annual Meeting of the Societas Linguistica Europaea, Zurich, Switzerland. https://hdl.handle.net/2078.5/241742