Aboutness topic, discourse topic and the structure of discourse

The paper addresses the relation between several dimensions along which discourse has been assumed to be structured – topical structure, hierarchical structure, QUD-structure and thematic structure – and points at previously undescribed mismatches between those. BACKGROUND ASSUMPTIONS: As discourse progresses, the aboutness topic of a sentence (Reinhart, 1981; Roberts, 2011; Krifka, 2007) may remain aboutness topic of subsequent sentences in discourse, or the aboutness topic may change (Givon, 1983). This relation between the aboutness topics of subsequent sentences in discourse, which I call the topical structure of discourse, constitutes one dimension along which discourse may be structured. Discourse may also be structured along a hierarchy of (explicit or implicit) questions under discussion (QUDs) which individual sentences and sequences of sentences in discourse can be seen to answer (von Stutterheim, 1994; van Kuppevelt, 1995; Roberts, 1996). According to (Roberts, 1996), achieving the goal of all discourse, which is the attempt to answer the global QUD ”What is the way things are?”, involves developing sub-goals addressed in terms of answering sub-QUDs. The resulting discourse has a hierarchical QUD-structure where each sentence addresses its own, local QUD, and sequences of sentences may answer a joint global QUD. A third dimension is what I call the thematic structure of discourse: a sequence of sentences may exhibit certain thematic continuity in terms of a common discourse topic (van Dijk, 1976; Asher, 1993; van Kuppevelt, 1995). Finally, discourse may be structured into a hierarchy of discourse units (DUs), where a superordinate sentence/DU may dominate one or several subordinated sentences/DUs. This hierarchical structure of discourse is governed by two types of discourse relations between sentences/DUs, coordinating and subordinating discourse relations (Asher and Vieu, 2005).1 The relation between these structures has been discussed before, but I am not aware of a model that takes all these dimensions into consideration. In (van Kuppevelt, 1995; Roberts, 2011), QUD-structure corresponds to thematic structure, since the global QUD of a DU corresponds to the discourse topic of that DU. (Frey, 2005) shows that thematic continuity should be distinguished from topical continuity. Finally, it has been commonly assumed that the QUDanalysis of discourse structure is compatible with the analysis in terms of discourse relations, since the latter can be characterized in terms of implicit questions that relate a sentence to preceding sentences in discourse cf. e.g. (Kehler, 2012). CLAIMS: I. Thematic structure corresponds to QUD-structure, but there is a mismatch between thematic/QUD-structure and hierarchical structure; II. Topical and thematic structure do not coincide, but thematic structure is sensitive to topical structure, in a way hierarchical structure isn’t. EVIDENCE I: The model of (van Kuppevelt, 1995) which I employ assumes that topicality in terms of a hierarchy of topic-comment structure is a basic organizing principle of discourse structure. A question-based notion of topic-commment structure is used for both individual sentences and larger DUs. A question Q determines a (discourse) topic T defined as a set of possible values (objects, places, times, reasons) of the ”topic term” of the question. One of these values is selected by answer A. The topic term corresponds to background in the focus-background distinction (cf. e.g. (Krifka, 2007)), rather than to aboutness topic, and represents a (contextually given or evoked) indeterminacy that needs further specification. The comment C is provided by A. If