Towards a new typology of the referential information structure of specificational it-clefts

Abstract This paper investigates the referential information structure (IS) of full and reduced specificational it-clefts, which interacts as a distinct layer with their relational IS. Drawing on spontaneous spoken data from the London-Lund Corpus, this study examines how discourse-new and discourse-given information, i.e., referential IS, is distributed over the clefted noun phrase (NP) and cleft-relative clause. To assess the discourse-familiarity of nominal referents and open propositions, I develop an analytical model hinging on the predictability of information in accordance with the prospective dynamic of spoken language. The findings show that the distribution of given and new information is more diverse than described in existing typologies. In particular, it is revealed that the hitherto overlooked pattern in which both value and variable are discourse-given and in which the specification relation may be new or given is in fact the most common one. The findings also show that the choice between full and reduced it-clefts as two basic options is only partly motivated by the discourse-familiarity of the variable. The analysis of the prosodically coded relational IS reveals that multiple prosodic patterns can be mapped onto each category of clefts, thus demonstrating that the conflation of the referential and relational IS of it-clefts is untenable.