Contrast in a QUD-based information-structure model

In this paper I look at contrast within the model of information-structure annotation proposed by Riester, Brunetti, and De Kuthy (2018), Riester (2019), Brunetti, De Kuthy, and Riester (2021), which is based on the notion of “Question under Discussion” or QUD (Roberts [1996] 2012, among many others). The model assumes that every utterance in a discourse is preceded by an implicit question, and proposes four principles that constrain the formulation of such questions. One principle, “Parallelism”, accounts for two or more utterances answering the same QUD. Such a discourse configuration provides the ground for contrast (Umbach 2004, 2005; Repp 2016); specifically, what is called “Simple” Parallelism accounts for occurrences of contrastive focus, while “Complex” Parallelism is relevant when two alternative sets are evoked, namely with contrastive topics. I assume that contrast is accounted for in terms of contrastive discourse relations and make the working hypothesis that contrastive relations always co-occur with Parallelism. Partially following Repp (2016), I assume four contrastive relations, specifically: similar, oppose, corr(rection) and concession; and describe what contextual and semantic restrictions make these relations different. By analyzing naturalistic data from spoken and written interviews in Italian and French, whose utterances were annotated for their QUDs and information structure and for their contrastive relations, I show how the differences among these relations can be partially accounted for in terms of their QUD structure. I also look at the interplay between contrastive relations and Simple and Complex Parallelism. I eventually show and discuss cases where contrastive relations and Parallelism do not co-occur. This mostly happens when a concession relation holds between the discourse segments.