Background Deletion: The Syntax of Clausal Ellipsis in Hindi/Urdu

Abstract

This thesis explores the syntax of subsententials in Hindi/Urdu (H/U), with an empirical focus on sluicing and fragment answers. Such phenomena represent instances where all clausal material is unpronounced, except for a (wh-)phrase (e.g., A: Who did John see? B: Mary.). With evidence from various connectivity effects, I argue that a non-structuralist analysis of such configurations, which does not assume the presence of tacit material, is untenable. Rather, subsententials in this language warrant a structuralist solution, according to which the surfacing remnant base-generates in an underlying clause whose morpho-syntax is reduced at PF. Most structuralist proposals of clausal ellipsis, including scant proposals of sluicing in H/U, assume PF-deletion targets a syntactic constituent (TP), forcing the remnant to escape the ellipsis site it is born in (e.g., A: Who did John see? B: Mary <John saw t>., where material in angled brackets = PF-deletion). I argue against this conventional view, given the conceptual and empirical issues it raises (chiefly, exceptional movement). I alternatively suggest that non-pronunciation freely and maximally affects morpho-syntactic material surrounding the remnant, allowing it to remain in situ (e.g., A: Who did John see? B: <John saw> Mary.), as has been argued elsewhere (Morgan 1973; Kimura 2010, 2013; Bruening 2015; Abe 2015; Ott and Struckmeier 2018; Sato et al. 2018; Griffiths 2019; among others). Although PF-deletion is not sensitive to syntactic constituency from this perspective, morpho-syntactic material that does PF-delete must constitute the propositional background of the elided clause to ensure recoverability of unspoken content, reducing clausal ellipsis in H/U to background deletion. I then propose to extend this analysis to correlativization in H/U, wherein a left-peripheral relative clause is associated with a correlate in the host clause (HC) (e.g., The girl who is standing, she is tall). Correlatives (CRs) in this language bear discordant characteristics: on the one hand, CRs display properties that point to their connectivity with the HC (e.g., reconstruction effects) and, on the other, features that point to their extra-sentential status (e.g., the lack of a gap in the HC and prosodic separation of CR–HC). Following recent work on Romance/Germanic left-dislocation (Ott 2014, 2015), which presents a similar paradox, I propose that CRs are not syntactically dependent on their HC; conversely, they are relatives that are part of a separate root clause which is juxtaposed in discourse with the HC and is reduced at PF under identity with it (e.g., The girl who is standing <is tall>. She is tall.). Connectivity effects of CRs are argued to be ellipsis-mediated, given that such effects are likewise observed in sluicing and fragment answers. In reconciling the contradictory properties of CRs, this novel proposal has an explanatory edge over existing accounts that assume syntactic integration of CR–HC and that rely on either movement or base-generation.