On describing polysemous discourse markers. What does translation add to the picture?* Liesbeth Degand Université catholique de Louvain 1. Introduction Imagine the following piece of conversation taking place between a mother and her youngest daughter: - Alors Maman, tu viens? - Je dois encore travailler un peu … - Mais alors tu ne verras pas le début du film. Tu fais quoi encore ? - J’écris un article. Une de mes collègues va avoir 60 ans, alors on lui offre un livre. - Et ça va être bien ? - … This piece of pseudo-conversation contains a number of different uses of the French discourse marker alors (‘then/so/well’). Translating the three occurrences of alors into Dutch, for instance, would probably result in three different translations namely zeg, dan, dus, respectively. I would like to put forward here that alors is a polysemous discourse marker in Modern French, and that the use of translation data can be used as an additional means to disentangle the different related meanings of this marker. The paper is further structured as follows. Section 2 tackles the issue of describing the polysemy of discourse markers, section 3 briefly reviews the use of translation data as a heuristics to describe the semantics of discourse markers. Sections 4 to 6 describe the semantics of alors and present a study of its use in translations from French to Dutch and from Dutch to French. The article ends with a number of concluding remarks. 2. The polysemy of Discourse markers Discourse markers are polysemous linguistic items. Straightforward as this assertion might seem, it is not. Actually, when talking about discourse markers (henceforth DMs), nothing seems straightforward. First, there is the issue of deciding what constitutes the class of DMs. Despite the explosion of empirical and theoretical research on DMs in the last 20 years, a univocal answer to that question is still lacking (see, e.g., the seminal work by Schiffrin 1987, Jucker & Ziv 1998, Lenk 1998, Hansen 1998, Andersen & Fretheim 2000, Fischer 2000 and 2006, Aijmer 2002, Aijmer & Simon-Vandenbergen 2006). Exactly 10 years ago, Schourup (1999: 228) noted that “[d]espite the quantity of research in this area (…) no consensus has emerged regarding fundamental issues of terminology and classification.” Agreement is indeed poor but on the observation that DMs build a formally and functionally highly complex category. There are nevertheless a number of syntactic and semantic properties of DMs the research community agrees on. According to Schourup (1999), there is agreement on at least (!) three properties: (i) connectivity, (ii)