Verb-stranding verb phrase ellipsis Nicholas LaCara · University of Toronto ÕóìÕ · Õä October óþÕß
1 Overview
• Vu§f-«±§Zo verb phrase ellipsis (êê£u) is a phenomenonwhere all of the See also my lecture notes from contents of a verb phrase with the exception of the verb does not appear. 11 September 2017.
• A (main) verb is present, but all verb-phrase-internal material disappears. Underlined verbs are stranded. Italicized verbs are antecedents. (Õ) Quando a Ana pôs os óculos na mesa, a Maria também When the Ana put.£«± the glasses on.the table, the Maria too pôs . put.£«± ‘When Ana put the glasses on the table, the Maria did too.’ Portuguese (Cyrino and Matos 2002:(14a)) • is phenomenon can be shown to obey the many constraints on verb phrase This gives vvpe the appearance non-constituent ellipsis (ê£u) in general. of being a deletion (see below). • However, the verb (seemingly the most critical part of a verb phrase) somehow escapes deletion.
• e standard view has been to assume that êê£u involves verb movement to an I'll uses strikethrough when I inectional position out of a ê£u site: needtoshowthecontentsofa deleted phrase; otherwise, I willuse toshowthe (ó) [TP a Maria também pôs [VP tVþ os oculos na mesa] ] (hypothesized) position of an elided phrase.
• In order for a language to have êê£u, two things are necessary:
i. e language must be shown to have ê£u. ii. e language must have Vþ to Iþ movement. This is why English lacks vvpe. Thoughithas vpe, it lacks verb • One of the stranger things about this phenomenon is what has come to be known movement. as the êu§fZ ou±±í §u¤¶§uu±.
– Verbal material extracted from the ellipsis site by head movement must match the material in the antecedent. – Since the identity requirement on ellipsis is thought to be calculated at LF/Semantics, this suggests that the extracted material behaves as if it is still in the ê£u site at LF.
• Goldberg (óþþ¢) originally takes this to mean that verbs must obligatorily recon- struct at LF for the purposes of calculating ellipsis identity.
– It’s not totally clear why this should happen. Phrasal movement need not reconstruct in this way.
Õ Nicholas LaCara · Verb-stranding verb phrase ellipsis 2
– If verb movement does not happen in the narrow syntax, as Chomsky Goldberg (2005:127) raises this (óþþÕ) proposes, and is only a PF phenomenon, then this might help us possibility, though she does not explicitly pursue it. understand this behavior.
• However, as Gribanova (óþÕì) shows us, the verbal identity requirement is not absolute, and mismatches are allowed to occur.
Today Õ. êê£u vs. object drop
ó. Accounting for the verbal identity constraint
2 Verb-stranding vpe
• Verb-stranding ê£u occurs in a number of languages that are not directly related. Ithasbeenarguedtoexistin other languages, including • ere is strong evidence it exists in the following languages: Finnish (Holmberg 2001), Japanese (Otani and Whitman 1991), Serbo-Croation (Boeckx (ì) Dúirt siad go dtiocfadh siad, ach ni tháinig ariamh . and Stjepanović 2001),and say.£«± they h come.ho they, but u come.£«± ever. Welsh (Rouveret 2012). ‘ey said they would come, but they never did.’ Irish (McCloskey 2011) (¦) A Ana não leva o computador para as aulas, porque the Ana not bring.£§u«.ì« the computer to the classes, because os amigos também não levam . the friends too not bring.£§u«.ì£ ‘Ana does not bring her computer to the classes because her friends do Portuguese not bring either.’ (Cyrino and Matos 2002) (¢) Q: Šošana yoda’at še-hizmant et aba šela la-mesiba? Shoshana know that-invite.£«± Zhh father of.her to.the-party ‘(Does) Shoshana know that you invited her father to the party?’ A: Lo, hi hit’alpa lifney še-hizmanti . no, she faint.£«± before that-invite.£«±.Õ« ‘No, she fainted before I did.’ Hebrew (Goldberg 2005) (ä) Q: Ty poznakomil Mašu s Petej? you. introduce.«. Maša.Zhh with Peter.«±§ ‘Did you introduce Masha to Peter?’ A: Konečno poznakomil ! of-course introduce.«. ‘Of course I introduced (Masha to Peter)!’ Russian (Gribanova 2013) (ß) Mama a-li-wek-a ki-kombe meza-ni na m-toto mother Õ«¶-£«±-put-ê ß-cup Étable-h and Õ-child a-li-wek-a pia. Õ«¶-£«±-put-ê too ‘e mother put the cup on the table, and the child did too.’ Swahili (Goldberg 2005) Nicholas LaCara · Verb-stranding verb phrase ellipsis 3
• If we maintain the usual assumption that ellipsis deletes a constituent, then any apparent case of non-constituent deletion must involve movement out of the ellipsis site.
• Since êê£u appears to delete everything in the verb phrase up to but excluding This is the centralhypothesis vpe the verb, it stands to reason that the verb must move out of the ellipsis site. underlying verb-stranding .
– Fortunately, all of these languages have independently been claimed to That can be V0-to-T0, 0 0 have verb-movement. at is, in non-elliptical clauses, there is evidence V -to-Asp , etc. for Vþ-to-Iþ movement of some sort.
2.1 Object drop • As with any good syntactic problem, there is a complication: Many languages that appear to have êê£u also have fuh± o§£. is leads to a great deal of Ornullobjects,ortopicdrop... syntactic ambiguity between the constructions:
() O João leu esse livro, e a Ana também leu . the João read that book and the Ana too read.
a. aAnatambémleu [tVþ esse livro]. (ê£u) =‘Ana also did .’ b. aAnatambémleu ∅. (null object) = ‘Ana also read it.’
• Fortunately, distinguishing these two phenomena is possible.
– Since object drop in most of these languages behaves like an A′-dependency This is sometimes called topic drop (following Huang ÕÉó), only ê£u can occur in islands. for this reason. – Object drop can only be responsible for the loss of a single internal argu- ment and cannot occur in islands. When more than one argument disap- pears, it is typically the work of ê£u. – Finally, there is a general requirement that a verb stranded by ê£u match the verb in the antecedent; no such requirement seems to hold of object drop.
• Unfortunately, these three requirements are all interrelated, and nding clear Not because it is hard to create, examples that isolates each of them in action is hard to come by. but because many authors do not always give clear, full • e following chart summarizes the properties of each phenomenon. paradigms.
(É) e standard view: This breakdownis widely reported in the literature on êê£u Null Obj. Hebrew, Irish, Portuguese, and Russian (Cyrino and Matos Island sensitive No Yes 2002; Doron 1990; Goldberg Verbal identity Yes No 2005; Gribanova 2013; Targets Everything Objects McCloskey 1991). Nicholas LaCara · Verb-stranding verb phrase ellipsis 4
2.2 Island sensitivity • An observation going back at least to Raposo’s (ÕÉä) discussion of European Portuguese is that null objects do not appear in islands, but êê£u can.
• Doron (ÕÉÉþ:Õþ–Õì) shows the same for Hebrew.
(Õþ) Q: saragt et ha- sveder ha-ze? you-knit Zhh the sweater this? ‘Did you knit this sweater?’ A: lo, aval ha- baxura Se- sarga natna li oto be- matana. Inthisminimalpair,wesee no, but the girl who knit gave to-me it for present thatwhenthestrandedverbis inanisland,theverbroots ‘No, but the girl who did gave it to me for a present.’ cannot be different. The A: *lo, aval ha- baxura Se- kanta natna li oto be- matana. mismatch would be possible if object drop were available in no, but the girl who bought gave to-me it for present this condition. ‘No, but the girl who bought *(it) gave to me for a present.’
′ • Following Huang (ÕÉó), null objects are thought to be an A -construction: Top- icalization of null operator.
• Importantly, while null objects target a single element, êê£u must delete every- Some other VP-internal thing in the VP. material might be droppable in some cases, tough (Costa and • It is typically assumed, therefore, that when multiple VP-internal elements are Duarte 2001; Santos 2009) missing that it cannot be due to null objects.
2.3 The verbal identity requirement • With island sensitivity in mind, we can start looking for other properties that distinguish the two phenomena.
• Crucially, êê£u exhibits the êu§fZ ou±±í §u¤¶§uu± (êo).
• is is a lexical identity requirement. Verb roots and other material extracted You'll notice in other examples from ellipsis sites must morphologically match the verbs in the antecedent. that inflectional morphology ontheverbneednotmatch. (ÕÕ) Portuguese: Cyrino and Matos 2002:(14) a. Quando a Ana pôs os óculos na mesa, a Maria when the Ana put.£«±.ì« the glasses on.the table, the Maria também pôs . also put.£«±.ì« ‘When Ana put the glasses on the table, Maria did too.’ b. *Quando a Ana colocou os óculos na mesa, a Maria Note that the two verbs in when the Ana put.£«±.ì« the glasses on.the table, the Maria these examples appear to meanthesamething.Thisis também pôs . not simply a semantic meaning also put.£«±.ì« requirement – the verbs must ‘When Ana put the glasses on the table, Maria did too.’ actuallymatch.This is takento be a problem for semantic identity constraints on ellipsis. Nicholas LaCara · Verb-stranding verb phrase ellipsis 5
(Õó) Hebrew: Goldberg 2005:160, (1) Q: (Ha’im) Miryam hevi’a et Dvora la-xanut? ¤ Miryam brought Zhh Dvora to.the-store ‘Did Miryam bring Dvora to the store?’ A: Ken, hi hevi’a. yes, she brought. ‘Yes, she brought [Dvora to the store].’ A: *Ken, hi lakxa. yes, she took ‘Yes, she took [Dvora to the store].’ • Null objects do not impose such a requirement: (Õì) Ela tirou o anel do dedo e guardou ∅ no cofre. She took.o the ring from.the nger and put it in.the safe ‘She took o the ring from her nger and put (it) in the safe.’ Portuguese (Cyrino and Matos 2002:(15a)) (Õ¦) Sara raxca et kol ha-calaxot ve-xilka ∅ la-’orexim. Sara washed Zhh all the-plates and-distributed to.the-guests ‘Sara washed all (of) the plates and distributed (them) to the guests.’ Hebrew (Goldberg 2005:50, (41)) 2.4 Complications • It is possible to drop material other than objects in a lot of circumstances, and the sensitivity of object drop to islands is not always robust (at least in Portuguese; Costa and Duarte óþþÕ). • Moreover, it is distinctly possible that some verbs take null complement anaphora nca is familiar from English: as a complement rather than null pronouns (Santos óþþÉ), a possibility that is John couldn't jump rope, but he tried ∅. oen overlooked. • e verbal identity requirement, while apparently robust in Irish and Hebrew, The conditions under which has been reported to be violable in Russian (Gribanova óþÕì) and European Por- this happens inRussianare limited, though. It is unclear tuguese (Santos óþþÉ). is means, at best, that verbal identity can only be used whatthe constraints are in as a one-way implication (if verbal identity, then ellipsis). Portuguese, if any.
3 Verb movement out of ellipsis sites
• Lotus Goldberg’s (óþþ¢) dissertation is a long-form argument in favor of the view that êê£u is generated by moving a verb out of an ellipsis site, a view rst clearly articulated by McCloskey (ÕÉÉÕ) for Irish.
þ þ þ (Õ¢) [IP I +V [vP V XP YP ZP ] ]
• is stands in contrast to the idea that verbs could be base-generated or merged directly in inectional positions just in case ellipsis occurs, as was proposed by Doron (ÕÉÉþ). e antecedent (including a verb-movement trace) would then be copied in at LF: Nicholas LaCara · Verb-stranding verb phrase ellipsis 6
þ LF þ (Õä) [IP [Iþ V ] [vP ∅ ] ] Ð→ [IP [Iþ V ] [vP tV ...]]
• ere are two main arguments:
i. Generating verbs outside of an ellipsis site would require some strange se- Idiscussedthison11 mantics for verbs. September. ii. ere is no clear way to account for the verbal identity requirement if verbs are generated/rst merged outside the ellipsis site.
3.1 Generating verbs in I0 is a semantic headache • Recall that in order to generate verbs outside the verb phrase, Goldberg (óþþ¢) argues that we would need VP pro-forms of several dierent times, one for each possible verbal argument structure.
• If verbs can be merged directly in Iþ, they must compose with VP pro-form somehow. Since verbs come with various semantic types, verb-stranding lan- guages would need a plethora of VP pro-forms to accommodate semantic com- position.
– A pro-VP for a transitive verb would be of type ⟨⟨e, ⟨e, t⟩⟩, ⟨e, t⟩⟩. is would compose with the verb in Iþ (presumably type ⟨e, ⟨e, t⟩⟩) and then the subject (type ⟨e⟩). – A ditransitive would need a pro-form of type ⟨⟨e, ⟨e, ⟨e, t⟩⟩⟩, ⟨e, t⟩⟩.Atran- sitive taking a clausal complement would need as still dierent type. – It is not clear that any such “VP-minus-the-V” pronouns are attested cross- linguistically.
• A verb movement account, however, side-steps this issue straightforwardly. No special null pro-forms are necessary. Verb stranding simply falls out from verb movement and (PF) deletion.
3.2 The verbal identity requirement again • Second, if verb movement leaves traces, as Doron (ÕÉÉþ) assumes, it is not fully Verbs of different types might clear why the verbal identity should hold. leave traces of different types of course. What matters here is • On an LF-copying analysis, a verb should be able to bind another trace/copy of thatevenverbsofthesame semantic type obey the verbal verb movement assuming that the trace is of the correct type. identity requirement. • So, for instance, in (Õþ), repeated here, assuming that ‘knit’ and ‘buy’ are both This requires some form of transitive verbs of type ⟨e, t⟩ and leave a trace of the same type, the stranded non-movement chain formationtooccuratLF;see verb kanta ‘bought’ should be able to bind the trace of sarga ‘knit’ that has been Chung et al. 1995. copied at LF:
(Õþ) Q: saragt et [vP tsaragt ha- sveder ha-ze]? you-knit Zhh the sweater this? ‘Did you knit this sweater?’ Nicholas LaCara · Verb-stranding verb phrase ellipsis 7
A: *lo, aval ha- baxura [CP Se- kanta [vP ∅]] natna li oto. no, but the girl who bought gave to-me it ‘No, but the girl who bought *(it) gave it to me.’ LF (Õß) [IP [Iþ kanta] [vP ∅ ] ] Ð→ [IP [Iþ kanta] [vP tsaragt ha-sveder ha-ze]]
• How would one capture this under an LF copying analysis? Could therebe some sort of lexical identity requirement imposed on traces/copies?
• If such a requirement were to exist, we would have to be very careful in how we state it. LF Copying analyses need a requirement weaker than strict lexical identity, but stronger than just argument structural identity.
– It could not be construed over something like argument structure, since verbs with the same argument structure are not interchangeable:
(Õ) Q: Yicxak nišek et Aviva? Yitzchak kissed Zhh Aviva ‘Did Yitzchak kiss Aviva?’ A: *Lo, hu ìfu. no, he embraced ‘No, he hugged [Aviva].’
– It could not be strict morphological identity either because inectional morphology need not match:
(ÕÉ) Q: Tazmini et Dvora la-mesiba? invite.¶±.ó.« Zhh Dvora to.the-party ‘Will you invite Dvora to the party?’ A: Kvar hizmanti. already invited.Õ« ‘I already invited [Dvora to the party].’
• Rather, we might want to say that the êo is a property of the traces and that they must be bound by the elements that le them.
– us,atracelebyVþ must be bound by the same (lexical) Vþ that le the trace, and a trace of vþ must be bound by the same vþ that le the trace.
• e real problem with this is that phrasal material is not subject to a matching requirement.
(óþ) ‘Mikael’ can bind copies of ‘Joey’: PF: Joey was accepted to the conference, and Mikael will be too.
LF: Joey was [vP Joey [VP accepted Joey to the conference]], and Mikael will be [vP Joey [VP accepted Joey to the conference]] too. ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ LF copy Nicholas LaCara · Verb-stranding verb phrase ellipsis 8
• is is the bizarre fact: e matching requirement does not hold of phrasal ma- terial, only head-moved material.
• We would have to stipulate that traces of head movement are subject to an iden- tity requirement that the traces of phrasal movement are not.
3.3 Accounting for verbal identity 3.3.1 Reconstruction • Goldberg’s takes the position that verbs move to Iþ as normal and that ê£u deletes vP at PF.
• To account for the verbal identity requirement, she proposes that verbs must Alternatively, obligatorily obligatorily reconstruct into their base position. interpret the lower copy.
þ þ (óÕ) a. PF:[V +I [VP tV ...]] þ þ b. LF: [ +I [VP V ...]]
• is will cause the verb to be interpreted in the elided verb phrase at LF. Thisisnotcompatiblewiththe notion of semantic identity • e verbal identity requirement should then fall out from a strict LF identity introduced by Merchant (2001) requirement on ellipsis.
• us, the verbal identity requirement falls out the same mechanisms as the (gen- eral) identity requirement on ellipsis.
• Languages vary based on where verbs are pronounced but not where they are interpreted.
3.3.2 PF head movement • As Goldberg (óþþ¢:Õþ–Õó) discusses, however, this observation is compatible Schoorlemmer and with the view that head movement is not a syntactic phenomenon at all. Temmerman (2012) pursuethis proposalinmore detail. • If heads move only at PF (or, at least, do not move in the narrow syntax), they will be in their base positions at LF, and reconstruction is rendered unnecessary.
þ þ þ (óó) a. PF:[V +I [VP V ...]]
þ þ b. LF: [ I [VP V ...]]
• is has the same consequences as above, except we don’t need to posit obliga- tory reconstruction.
• To the extent that there is independent evidence against narrow syntactic move- ment of heads, this may therefore be a more appealing analysis. Nicholas LaCara · Verb-stranding verb phrase ellipsis 9
4 Lingering issues
• As mentioned above the robustness of the verbal identity requirement has been challenged in recent years.
– Gribanova (óþÕì) and Santos (óþþÉ) have shown that it appears to be vio- lable in Russian and European Portuguese, respectively. – Understanding what makes it violable is important for understanding what it is.
• If wh-elements reconstruct, they should aect the identity requirement.
References
Boeckx, Cedric, and Sandra Stjepanović. óþþÕ. Head-ing toward PF. Linguistic Inquiry ìó:즢–좢. Chomsky, Noam. óþþÕ. Derivation by Phase. In Ken Hale: A Life in Language, ed. Michael Ken- stowicz, Õ–¢ó. Cambridge, Mass: MIT Press. Chung, Sandra, William Ladusaw, and James McCloskey. ÕÉÉ¢. Sluicing and Logical Form. Nat- ural Language Semantics ì:óìÉ–óó. Costa, João, and Inês Duarte. óþþÕ. Objectos nulos em debate. In Razões e emoção. Miscelânea de estudos para Maria Helena Mateus, ed. Inês Duarte and I Castro. Lisbon: Imprensa Nacional- Casa da Moeda. Cyrino, Sonia M. L., and Gabriela Matos. óþþó. VP ellipsis in European and Brazilian Portuguese – a comparative analysis. Journal of Portuguese Linguistics Õ:Õßß–ÕÉ¢. Doron, Edit. ÕÉÉþ. V Movement and VP-Ellipsis. Ms., e Hebrew University of Jerusalem. Published as Doron ÕÉÉÉ. Doron, Edit. ÕÉÉÉ. V Movement and VP-Ellipsis. In Fragments: Studies in Ellipsis and Gapping, ed. S. Lappin and E. Benmamoun, Õó¦–Õ¦þ. Oxford: Oxford University Press. Goldberg, Lotus. óþþ¢. Verb-Stranding VP Ellipsis: A Cross-Linguistic Study. Doctoral Disserta- tion, McGill, Monteal, QC. Gribanova, Vera. óþÕì. Verb-stranding verb phrase ellipsis and the structureof the Russian verbal complex. Natural Language and Linguistic eory ìÕ:ÉÕ–Õìä. Holmberg, Anders. óþþÕ. e syntax of yes and no in Finnish. Studia Linguistica ¢¢:Õ¦Õ–Õߢ. Huang, Cheng-Teh James. ÕÉó. Logical Relations in Chinese and the eory of Grammar. Doc- toral Dissertation, Massachusetts Institute of Technology, Cambridge, Mass. McCloskey, James. ÕÉÉÕ. Clause Structure, Ellipsis and Proper Government in Irish. Lingua ¢:ó¢É–ìþó. McCloskey, James. óþÕÕ. e Shape of Irish Clauses. In Formal Approaches to Celtic Linguistics, ed. Andrew Carnie, Õ¦ì–Õß. Newcastle upon Tyne: Cambridge Scholars Publishing. Merchant, Jason. óþþÕ. e Syntax of Silence: Sluicing, Islands, and the eory of Ellipsis. Oxford Studies in eoretical Linguistics. Oxford University Press. Otani, Kazuyo, and John Whitman. ÕÉÉÕ. V-Raising and VP-Ellipsis. Linguistic Inquiry óó:즢– ì¢. Raposo, Eduardo. ÕÉä. On the Null Object in European Portuguese. In Studies in Romance Linguistics, ed. Osvaldo Jaeggli andCarmen Silva-Corvalán, Publications in Language Sciences ó¦. Dordrecht: Foris Publications. Rouveret, Alain. óþÕó. VP ellipsis, phases and the syntax of morphology. Natural Language and Linguistic eory ìþ:Éß–Éäì. Santos, Ana Lúcia. óþþÉ. Minimal Answers: Ellipsis, Syntax and Discourse in the Acquisition of European Portuguese. John Benjamins. Schoorlemmer, Erik, and Tanja Temmerman. óþÕó. Head Movement as a PF-Phenomenon: Evi- dence from Identity under Ellipsis. In Proceedings of the óÉth West Coast Conference on Formal Linguistics, ed. Jaehoon Choi, E. Alan Hogue, Jerey Punske, Deniz Tat, Jessamyn Schertz, and Alex Trueman. Somerville, Mass.: Cascadilla Proceedings Project.