Volume 33, Issue 4 - Special Issue: Ellipsis: licensing, structure and identity, Issue Editors: Evelyn Gandón-Chapela and Javier Pérez-Guerra

Following an introduction to ellipsis and its consequences both for communication and for the theory of grammar, this paper addresses a number of issues which have been debated in the linguistic literature on ellipsis, namely structure, identity and licensing. This study brings to the readers’ attention the effects which such conditions have for the study of ellipsis in different languages as well as their constraints within the generative framework. As case studies on the conditions and the characterisation of the linguistic strategy under scrutiny, this paper also summarises the main approaches to ellipsis in the studies embodied in this special issue.

The study of ellipsis, being a mismatch between form and meaning, has already proven to have consequences for our understanding of language in general, as it has helped us gain insights in other domains of the grammar. This paper focuses on one of these domains, namely the notion of Spell-Out and the theory of phases that has been developed within the Minimalist Program (Chomsky 1995, The minimalist program . Cambridge, MA: MIT Press, Chomsky 2001, Derivation by phase. In Michael Kenstowicz (ed.), Ken Hale: A life in language , 1–52. Cambridge, MA: MIT Press, and Chomsky 2005, On phases. Ms. MIT). Several authors have been tempted to tie ellipsis to Phase Theory, as ellipsis would be non-pronunciation at PF instead of pronunciation. In other words, ellipsis is the flip coin of Spell-Out, and the two differ only at PF. Although attractive, this proposal will be pointed out to run into empirical problems with regards to extraction possibilities. The data suggest that the difference between ellipsis and non-ellipsis is not simply decided at PF, but in the syntax already. At the same time, however, this paper aims to maintain the intuition behind the link between ellipsis and phases. It explores the chunks of structure that are targeted by both operations as well as their triggers.

Brazilian Portuguese (BP) is known to license anaphoric null objects (ANO), that is, null objects with a linguistic antecedent. It also licenses VP ellipsis (VPE), with auxiliaries, modals and main verbs, the latter a case of V-stranding VPE (V-VPE), the one with which we will be concerned. Although ANOs and V-VPE may have identical surface strings in BP, we propose that they do not have the same structure. We argue that ANOs in BP are cases of DP ellipsis, and they present four properties that have been associated with (VP) ellipsis in the literature: a) availability of strict/sloppy readings (Ross, 1967. Constraints on variables in syntax . Cambridge, MA: MIT dissertation); b) structural parallelism between the antecedent and the elided string (Fiengo and May 1994. Indices and identity . Cambridge, MA: MIT Press); c) formal licensing by a functional category immediately c-commanding the elided string (Lobeck 1995. Ellipsis: Functional heads, licensing, and identification . Oxford: Oxford University Press); and d) disjunctive E-type readings (Simons 1996. Disjunction and anaphora. Proceedings of Semantics and Linguistic Theory 6. 245–260; Simons 2001. On issues in the semantics and pragmatics of disjunction . Ithaca, NY: Cornell University dissertation; Sakamoto 2013. Beyond sloppy identity of elliptic arguments: An interim report. Paper presented at the LingLunch, University of Connecticut. November 5). We then present evidence for an analysis of the ANO as a base-generated empty phrase marker that will be reconstructed in LF. We also propose that ANOs and VPE in BP are licensed by a lexicalized aspectual head, as a consequence of the loss of generalized verb movement in the language.

The analysis of gapping still leaves many questions open. At least three aspects remain controversial: whether it involves unpronounced structure, whether clausal or non-clausal constituents are conjoined, and whether all gapped sentences are must be analyzed uniformly. On the basis of French gapped constructions such as Marie n’est jamais allée à Rome ni Jean à Pékin ‘Mary never went to Rome nor John to Beijing’, the paper argues that gapping is not an homogeneous phenomenon. It first shows that in such gapped clauses involving ni, a strict Negative Polarity Item, a negation low inside the first TP can license the ‘negative coordinator’ ni, a fact that a TP-deletion analysis of gapping or a fragment coordination one can capture. Conversely, it shows that a vP-coordination analysis correctly predicts the properties of ni-gapped clauses. However, such an analysis cannot extend to gapped clauses conjoined by double ni : the second conjunct, this time, must be a full clause or a fragment.

Recent work on ellipsis has promoted the hypothesis that syntax plays a limited role in the computation of the required isomorphism conditions between elliptical clauses and their antecedents. For example, Merchant (2008, An asymmetry in voice mismatches in VP-ellipsis and pseudogapping. Linguistic Inquiry 39. 169–179; 2010, Vocal and argumentative oppositions. Paper presented at the Workshop on Morphological Voice and Its Grammatical Interfaces: Theoretical Modeling and Psycholinguistic Validation. University of Vienna, June 25–27; 2013, Voice and ellipsis. Linguistic Inquiry 44(1). 77–108) shows the relevance of the syntactic heads related to voice and argument structure. Focusing on sluicing, Chung (2013, Syntactic identity in sluicing: How much, and why. Linguistic Inquiry 44(1). 1–44) adds Case-valuing heads to the list and Takita (2013, Strengthening the role of case in ellipsis. Paper presented at the Identity in Ellipsis Conference, Leiden University, September 20–21) argues that Case is the only necessary syntactic information. This article focuses on sluicing and provides further support for the role of syntax in ellipsis isomorphism conditions. However, it shows that Case cannot be a distinct primitive component in isomorphism computation since grammatical sentences will be wrongly predicted to be ungrammatical and vice versa. The analysis of cross-linguistic data provides evidence that a syntactic requirement, labeled the Remnant Condition, is indeed necessary. This condition is based on the general notion of unrecoverable information and on the proposed concept of Id(entity)-source, a restricted version of the sluiced clause, the derivation of which has to comply with all the required (syntactic and non-syntactic) identity conditions and, crucially, must be sensitive to the morphosyntactic information present both in the antecedent clause and in the remnant. The goal of the Remnant Condition is to ensure the correct syntactic integration of the remnant phrase in the Id-source. Sluicing facts related to Case, voice and argument structure, merging and sprouting are captured by a combination of the Remnant Condition, general principles of grammar and non-sluicing-related syntactic principles internal to the different languages.

This paper investigates the use of ellipsis diagnostics to determine whether deletion has occurred in as -parentheticals. Superficially, as -parentheticals look as though they contain verb phrase ellipsis, but previous work argues that they do not. A number of diagnostics that have been used to distinguish ellipsis from null pro-forms are discussed in some detail. These diagnostics show that the deletion process in as -parentheticals is distinct from canonical verb phrase ellipsis. Instead, it shares properties with comparative deletion. Thus, I argue that once a broader range of facts is considered, we need to include some manner of deletion in the derivation as -parentheticals.

Romanian aspectual complement ellipsis reveals the fact competing selection frames of verbs filter out potentially well-formed instances of deletion, this situation revealing yet another constraint on ellipsis with verbal licensers. Situations of this sort were dealt with in previous generative models by assuming “global derivational constraints” no longer available in the current minimalist architecture of the grammar. After delimiting the domain of ellipsis with verbal licensers in Romanian (the passive auxiliary be , the modal and aspectual verbs in biclausal non-restructuring configurations allow the deletion of their complement), we turn to the behaviour of Romanian aspectual verbs. Romanian aspectuals have a double selection frame: (i) raising verbs occurring in biclausal configurations, and (ii) full transitive verbs with a direct object denoting a hidden eventive component. A superficial application of the standard ellipsis diagnostics may lead one to reject the idea that aspectual verbs are licensers of ellipsis as direct object extraction and antecedent-contained deletion with a direct object pivot are blocked. This is, however, due to the fact that these verbs have a transitive selection frame. When other types of constituents from the ellipsis site are extracted or included in antecedent-contained deletion configurations (e. g., indirect objects, subcategorized PPs), the construction is well-formed, this testifying to a fully articulated ellipsis site. In other words, the transitive selection frame blocks an otherwise potentially well-formed instance of ellipsis. Apparent cases of Null Complement Anaphora thus appear to be hidden situations of “undiagnosable ellipsis”.

This paper offers an analysis of NP ellipsis with adjectival remnants in Polish and Hungarian based on an investigation of the featural composition of heads contained within the extended nominal and adjectival projections, as well as configurations resulting from an application of Agree and the Principle of Chain Reduction. Elliptical NPs are analysed as headed by categorising n heads lacking a complement. Whereas in Polish NP-internal agreement results in such heads being defective goals, treated as lower copies at the SM interface, and hence being unpronounced, in Hungarian, where most NP-internal modifiers do not agree with the head nouns, the content of the n head (i. e. {[Num: β], [Case: γ]} on the current assumptions) is realised on the final NP-internal modifier inside elliptical NPs. The difference between Polish and Hungarian with respect to the (lack of the) effect of NP-internal agreement on the elliptical patterns follows from the differences in the features of functional categories and the timing of Spell-Out/Transfer. The discussion suggests that the distinction between NP and DP languages should be taken into consideration in analyses of ellipsis in the nominal domain.