Fragments, truncated clefts and island-sensitivity

Research output: Contribution to conferencePaperpeer-review

26 Downloads (Pure)


Perhaps the most persistent puzzle in the literature on ellipsis is the question of why ellipsis ameliorates island violations in certain cases, but not in others. I will focus on the question of why fragment answers ameliorate islands in new information focus contexts, but not in contrastive contexts (e.g., Griffiths & Lipták 2014), and on why the same pattern is found with truncated it-clefts (e.g., Barros et al. 2014): (1) a. A: Abby speaks the same Balkan language that one of the boys speaks. B: Yeah, CHARLIE. / Yeah, it’s CHARLIE. b. A: Does Abby speak the same Balkan language that BEN speaks? B: *No, CHARLIE. / No, it’s CHARLIE. I will argue that fragments are derived by a rule Pronounce Focus (PronFoc), which simply sends the maximal F-marked constituent in a CP to PF, to the exclusion of the remainder of that CP (see also Abe 2015). This accounts directly for the lack of island-sensitivity in (1aB). If the ellipsis implemented by PronFoc is ‘identified’ in terms of presupposition (the ellipsis site is ‘identified’ if the elliptical clause presupposes the antecedent clause; e.g., Merchant 2001, Abe 2015), contrastive fragments must be subject to a distinct identification condition, as (1bB) does not presuppose (1bA). I claim that the identification condition for contrastive fragments instead makes reference to their ‘focus-background’ structure. Arguably, contrastive foci differ from new information foci in that they must be associated with a ‘background’ that must be linguistically identified in some way (e.g., Neeleman & van de Koot 2008). In the case of in situ contrastive foci, the background can normally be marked prosodically: for example, languages such as Russian have distinct intonational contours for new information and contrastive foci (Krylova & Khavronina 1984). Alternatively, the background can be marked via Aˊ-movement of the focus (e.g., Neeleman & van de Koot 2008). If PronFoc applies, however, the prosodic option is lost, leaving only the Aˊ-movement option. This accounts for the island-sensitivity of (1bB). Further evidence that movement of the remnant is not necessary to derive fragments comes from the ‘utterance-final effect’ (UFE) observed by Barros et al. (2014): the fact that contrastive fragments lose their island-sensitivity if the base position of the focus is clause-final: (2) A: Did they leave because you offended MARY? B: (?)No, SARAH. As I will show, the UFE is arguably related to the ‘scope extension’ property of only- and not-marked DPs when they appear in clause-final position (e.g., Taglicht 1984, Kayne 1998). As Błaszczak & Gärtner (2005) show, deriving scope extension through movement of the focus is problematic; rather, the relevant condition appears to be that the scope must form a ‘linearly and prosodically continuous string’, a requirement which also holds of the UFE. The parallel extends to Italian, in which preverbal subjects lack scope extension, while postverbal subjects permit it (Longobardi 1992); the same pattern is seen with the UFE (Ludovico Franco p.c.). As Barros et al. (2014) show, truncated it-clefts (TCs) generally pattern with fragments in terms of island-sensitivity. In fact, they reduce the island-insensitivity of new information fragments to the island-insensitivity of TCs, claiming that the latter can be used in the ellipsis site as a way of ‘evading’ an island violation. As for why new information TCs are themselves island-insensitive, they suggest that the TC in this case can lack the ‘cleft clause’ of full clefts altogether (a ‘bare TC’). In order to avoid overgeneration in contrastive fragments, they suggest that contrastive TCs must be derived using the ‘raising’ structure proposed by Reeve (2012), in which the focus moves out of the cleft clause prior to ellipsis of this clause; this ensures that contrastive TCs will be island-sensitive, and thus that they cannot rescue the equivalent contrastive fragments. This explanation is problematic, however, as the forcing of a raising structure is purely a stipulation designed to ensure island-sensitivity. In fact, it follows under the present analysis: because a contrastive focus must be associated with a marked background, the cleft clause must be present in order for BM to be able to apply. Because movement is independently required in clefts (e.g., Reeve 2012), the island-sensitivity of contrastive TCs follows. Because of this independently required movement, the analysis also predicts that contrastive TCs will differ from contrastive fragments in not exhibiting the UFE, which I show to be generally correct: (41) A: Won’t Ben leave the party if you bring CIGARETTES? B: No, ALCOHOL. B´: #No, it’s ALCOHOL. These findings provide further support for the claim, made particularly convincingly in Barros et al. (2014), that ellipsis never directly repairs island violations.
Original languageEnglish
Number of pages10
Publication statusPublished - 2016
EventEllipsis Across Borders Conference 2016 - University of Sarajevo, Sarajevo, Bosnia and Herzegovina
Duration: 20 Jun 201621 Jun 2016


ConferenceEllipsis Across Borders Conference 2016
Country/TerritoryBosnia and Herzegovina
Internet address


  • focus context
  • Pronounce Focus
  • clefts
  • island violations


Dive into the research topics of 'Fragments, truncated clefts and island-sensitivity'. Together they form a unique fingerprint.

Cite this