Available online at www.sciencedirect.com

ScienceDirect

Licensing NPIs and licensing silence: Have/be yet to inEnglish

www.elsevier.com/locate/linguaLingua 148 (2014) 213--239

Stephanie Harves a,*, Neil Myler b

aDepartment of Linguistics, New York University, United StatesbDepartment of Linguistics, Boston University, United States

Received 5 April 2013; received in revised form 11 April 2014; accepted 22 May 2014Available online

Abstract

This paper discusses the syntax of the have/be yet to construction in English, as in John has/is yet to eat dinner. As pointed out byKelly (2008), this construction raises a number of questions. How is the NPI yet licensed? Why is have interpreted as a perfect auxiliaryverb, in spite of the fact that it appears to take an infinitival complement, rather than a perfect participle? What accounts for the apparentfree alternation between have and be? We argue that have in the have yet to construction is, for many speakers, perfect have, whichselects for a silent raising predicate that has negative implicative semantics. This predicate is responsible for licensing the NPI yet. Wefurther show that the apparent free alternation between have yet to and be yet to is illusory. The category of the silent predicate can beshown to be different in each case in a way that is to be expected given independent c-selectional properties of have and be in English.© 2014 Elsevier B.V. All rights reserved.

Keywords: NPIs; Yet; Negation; Have/be alternations

1. The puzzles

Kelly (2008) points out several syntactic and semantic puzzles in regard to the constructions in (1), compared with theirparaphrases in (2).

(1)

* CorrE-m

http://dx0024-38

a.

espondail add

.doi.org41/© 2

John has yet to eat dinner.

b. John is yet to eat dinner.

(2)

a. John hasn’t eaten dinner yet. b. John didn’t eat dinner yet.

The first puzzle regards the presence of yet in the sentences in (1). How is yet, a Negative Polarity Item (NPI), licensed?The paraphrases in (2) contain negation, so could it be that the sentences in (1) contain an instance of silent sententialnegation?

Second, we appear to have a clash between the syntax and the semantic interpretation of have here. That is, why dowe see have to plus an infinitival complement in (1a) as opposed to have plus a perfect participle as in (2a)? Have toappears in English when the interpretation of have is modal, rather than the aspectual perfect, as in (3a). Could it be that

ing author.resses: stephanie.harves@nyu.edu (S. Harves), myler@bu.edu (N. Myler).

/10.1016/j.lingua.2014.05.012014 Elsevier B.V. All rights reserved.

S. Harves, N. Myler / Lingua 148 (2014) 213--239214

the have in this construction is, in fact, modal have? One can raise a similar question for be, since be to in English also hasa modal interpretation (3b).

(3)

1 Follomoved

2 An aWe add

a.

wing Kelemennonymress th

John has to leave by 5:00.

ayne (2004 et seq.), all CAPS will bts, where the head of the chain isous reviewer asks why yet cannot sis issue along with other potential

(� John needs to leave by 5:00.)

b. John is to leave by 5:00. (� John is supposed to leave by 5:00.)

Third, in light of the grammaticality of (1a), and its interpretation in (2a), why is the sentence in (4) ungrammatical?

(4)

*John has yet eaten dinner.

Finally, what are the similarities and/or differences between (1a) and (1b)? Do have and be alternate freely, or are theredifferences underlying the syntax of these constructions?

Kelly (2008) suggests that yet conveys negative perfect aspect in the constructions in (1), although it is not clear howexactly it comes to take on this meaning on its own.Wewill argue that yet does not, in fact, convey negative perfect aspecton its own. Rather, we will adhere to the standard assumption that yet is an NPI and argue that this NPI is licensed by asilent perfect participle FAILED, which has negative implicative semantics. Specifically, we will argue that the sentence in(1a) has the derivation shown in (5).1

(5)

[TD$INLINE]

John has yet FAILED [TP <John> to eat dinner <yet>].

The linear placement of yet (which is atypical, as we will show) arises via movement. We suggest that yet raises into thespecifier of its licensor, FAILED, which is in turn rendered silent by the presence of the NPI yet in its specifier. Our proposalis thus that the have/be yet to construction arises from a process that, while highly restricted in English, is robustly attestedin other languages. In particular, this relationship between the movement of an NPI and the silence of its licensor is a well-known feature of Ibero-Romance languages, as we discuss below. Further, the proposal has the following consequences:(i) the have/be yet to construction is biclausal and involves raising; (ii) the construction is downward-entailing, licensing theNPI yet in the embedded clause; (iii) there is no sentential negation present in the syntactic derivation of this construction.Below we present empirical arguments in favor of each of these consequences.2

The analysis presented in this paper follows a line of research in contemporary syntax whereby syntactic movement isargued to have an effect on the pronunciation or silence of an element. The silence of a given lexical item can be argued tofollow from Koopman’s (1996) Generalized Doubly-Filled Comp Filter, whereby only a Head or its Specifier may bepronounced if both are filled. This kind of approach has been adopted to account for a number of silent elements in avariety of syntactic constructions (e.g. in Collins, 2007; Kayne, 2005, 2010; Nchare and Terzi, 2014; Wood, 2013). Moregenerally, our analysis falls in line with a number of proposals that account for various mysterious morphosyntacticproperties cross-linguistically by arguing for the presence of a silent predicate. Proposals have been made for silentMEANT in English (Kayne, 2012/to appear), silent GO in Germanic and Slovenian (van Riemsdijk, 2002a; Marusič andZaucer, 2005), silent particle verbs in Dutch (van Riemsdijk, 2002b, 2012), silent FEEL-LIKE in Slovenian (Marusič andZaucer, 2006), and silent HAVE in English (Ross, 1976; McCawley, 1979; Larson et al., 1997; Harves and Kayne, 2012;Wood, 2013). The existence of semantically contentful silent elements in the syntax should come as no surprise, given thearchitecture of the grammar. Semantics is an interpretive system which assembles the meanings of sentences from themeanings of their parts and the way they are put together- in other words, semantics is a compositional interpretation ofthe output of syntax. Similarly, the PF component constructs a phonological representation from the output of syntax. Thepossibility that a terminal node might not have any associated phonology is quite expected once we recognize that PFand syntax are distinct systems -- just as there are elements of phonology that are ignored by syntactic computation(e.g., features like [� voice]), so we expect there to be syntactic elements that are ignored by phonology. Hence, becausesemantics and PF do not interact directly, it would require a stipulation to prevent situations in which a given terminal nodehas a meaning but no associated phonology (which is all that a silent element amounts to).

The remainder of this paper is structured as follows. In Section 2, we discuss the licensing environments of yet as anNPI. In Section 3, we consider the hypothesis that sentences with have/be yet to contain an occurrence of silent sentential

e used to indicate SILENT elements. We will use <angled brackets> to indicate silent copies ofpronounced and the tail is not.imply be argued to be licensed by a negative implicature, along the lines of Linebarger (1987).challenges in Section 4.4.

S. Harves, N. Myler / Lingua 148 (2014) 213--239 215

negation, in light of their paraphrases in (2) above. In Section 4 wemove on to the specifics of the syntax of have yet to andconsider the syntax of be yet to in Section 5. We conclude in Section 6.

2. The licensing of Yet

Since the influential work of Ladusaw (1979), it has been argued that NPIs are licensed in the scope of a DownwardEntailing (DE) operator. However, not all NPIs are licensed in the same downward entailing environments. Moreover, aspointed out by Linebarger (1980, 1987), some NPIs are licensed in some non-DE contexts. We begin by summarizing thecontextswhere yet is (and is not) licensed. This allowsus to narrowdown the spaceof hypotheseswith respect towhat couldbe licensing yet in the have/be yet to construction. We then move briefly to a discussion of distinctions between strong andweakNPIs, where strongNPIs are argued to additionally require an anti-additive context. It will then be shown that the have/be yet to construction is anti-additive, which we will ultimately argue follows from the silent licensor of yet, namely FAILED.

2.1. Environments that license yet

Levinson (2008) provides a comprehensive list of NPI-licensing environments where yet is found. These environmentsare shown in (6--14). In all cases, we provide at least one example using yet and at least one example with some other NPI(like any or ever), thus confirming that these are indeed NPI-licensing environments.

(6)

Negation a. Mary hasn’t been to Paris yet. b. Mary hasn’t ever been to Paris.

(7)

Scope of few and rarely a. Few tourists are here yet. b. Few Americans have ever been to Spain. c. That has rarely happened yet, but it’s going to happen a lot more.3

d.

Americans rarely visit any African countries.

(8)

Negative implicative verbs: prevent, refuse, fail, forget a. They have refused to admit it yet. b. The cold temperatures have prevented me from doing any planting yet. c. John has failed to buy any fresh fruit.

(9)

too Adj to Verb a. He is too young to understand this yet. b. He is too cynical to ever take things at face value.

(10)

without VERB-ing a. He has just ruined his life, without knowing it yet. b. Mary has crossed the border without presenting any documentation.

(11)

doubt a. I doubt the lamp is fixed yet. b. John doubts that Mary has ever been to Paris.

(12)

Yes-No Questions a. Have you finished your dinner yet? b. Have you ever been to Paris?

(13)

Comparative Clauses a. This is better [than anything yet invented]. b. That linguist is smarter [than anyone I’ve ever met].

(14)

Superlatives a. This is by far the best book I have yet purchased in the field of Web Design. b. This is the best film that I have ever seen.

S. Harves, N. Myler / Lingua 148 (2014) 213--239216

In addition to these contexts, we can also add the only DP, as pointed out to us by an external reviewer, who attributes thisobservation to Hoeksema (1986). This context patterns with the superlatives in (14).

(15)

3 An ansome cro

4 The cfurther di

5 As potime. We

6 See G(2005, 20weak NP

The Only DP

a.

*

onymoss-speontextsscussiointed othankiannak11) forIs, rath

This is the only paper to have been published on this construction yet.

b. This is the only paper to have ever been published on this construction.

2.2. Downward entailing environments that do not license yet

However, as pointed out by Levinson (2008), not all NPI-licensors are capable of licensing yet. As shown in (16--19)below, a number of environments that license other weak NPIs do not license yet.

(16)

Antecedents of Conditionals4

a.

If you’ve ever been to Paris, then you’ve undoubtedly seen the Eiffel Tower. b. If John has any money left, he’ll blow it on booze and smokes. c. *If you’ve been to Paris yet, then you’ve undoubtedly seen the Eiffel Tower.

(17)

Restrictor of universals a. [Everyone [who has ever seen the movie]] has liked it. b. *[Everyone [who has seen the movie yet]] has liked it.

(18)

Complements of emotive factives a. I regretted that I ever went to Dallas. b. I regretted that my car had broken down yet. c. I am surprised that he has ever spoken to her. b. *I am surprised that he has spoken to her yet.

(19)

The second argument of only (subject to dialectal variation)5

a.

Only John has ever arrived on time. b. %Only John has arrived yet. c. Only Lucy has any money left. d. %Only Lucy has finished her HW yet.

In sum, we have the following list of options available to us when considering a potential (silent) NPI-licensor for yet in thehave/be yet to construction.

(20)

yet-licensors a. Silent Negation

us reviewer finds (7c) rather deviant, judgingaker variation as to whether yet can be licin (16--17) were noted by Hoeksema (2008n.ut by Atlas (1993, 1996),Only John is in factan external reviewer for drawing this to ouidou (1997, 1998, 2006, 2010) for argumearguments that the presuppositions of stroer than strong NPIs, we will not throw our

e.

it as??ensed) in his

not dowr attentnts thang NPIhat into

Silent too Adj to Verb

b. Silent few or rarely f. Silent without Verb-ing c. Silent negative implicative verb g. Silent Comparative d. Silent doubt h. Silent Superlative/The only

2.3. Strong vs. weak NPIs and anti-additivity

Having considered the set of potential licensors for yet, we now briefly review one key difference between two types ofNPIs, strong NPIs and weak NPIs. As discussed by Zwarts (1998) and van der Wouden (1997), strong NPIs require anti-additivity as a licensing requirement, in addition to a downward entailing context.6 Zwarts provides the following definitionfor anti-additivity.

. It seems that, just as with the second argument of only in (19) below, there isin the scope of rarely.discussion of the licensing of NPI modals. See Iatridou and Zeijlstra (2012) for

nward entailing. (19a) is false or truth-valueless if nobody has ever arrived onion.t nonveridicality or antiveridicality is the key to NPI-licensing. See Gajewskilicensors must also be taken into account. Since yet passes diagnostics forthe ring around the debate concerning strong NPI-licensing here.

(21) Let B and B* be two Boolean algebras. A function f from B to B* is said to be

S. Harves, N. Myler / Lingua 148 (2014) 213--239 217

anti-additive iff for each two elements X and Y of the algebra B:

f(X [ Y) = f(X) \ f(Y). (Zwarts, 1998: 222)

We can isolate differences in the strength of NPIs by considering their behavior in anti-additive and non-anti-additive DEcontexts. For example, negation and negative existential DPs such as not a single DP are downward entailing, anti-additive operators that license strong NPIs such as until X, in years, lift a finger, and utter a sound. Quantificational DPssuch as not every DP are DE but not anti-additive. The examples in (22--25) show that strong NPIs require anti-additivity inaddition to DE in order to be licensed.

(22)

until a. John didn’t arrive until five. b. Not a single student/No student arrived until five. c. Not every student arrived until five. (Gajewski, 2005: 47) *

(23)

in years a. John hasn’t visited in years. b. Not a single student/No student has visited in years. c. Not every student has visited in years. (Gajewski, 2005: 47) *

(24)

lift a finger a. John hasn’t lifted a finger around here all week. b. Not a single student has lifted a finger around here all week. c. *Not every student has lifted a finger around here all week.

(25)

utter a sound a. John didn’t utter a sound. b. Not a single student uttered a sound. c. *Not every student uttered a sound.

In contrast to strong NPIs, weak NPIs such as any are licensed under quantificational not every DP, as shown in (26). Andas we see in (27), yet patterns with the weak NPI any, rather than with the strong NPIs in (22--25). That is, yet is notrestricted to anti-additive contexts.

(26)

any a. John didn’t say anything. b. Not a single student/No student said anything. c. Not everyone said anything.

(27)

yet a. John hasn’t visited Paris yet. b. Not a single student/No student has visited Paris yet. c. Not everyone has visited Paris yet.

Returning to the have/be yet to construction, if we hope to determine a potential licensor for yet, we must first determinewhether we are dealing with a context that licenses both strong and weak NPIs or only weak NPIs. The examples in (28)show strong NPIs such as lift a finger and utter a sound are perfectly natural under have yet to (the judgments can bereplicated for be yet to, for speakers who allow that construction).

(28)

a. John has yet to visit any cities in Europe. b. John has yet to ever visit me. c. John has yet to lift a finger around here. d. John has yet to utter a sound.

S. Harves, N. Myler / Lingua 148 (2014) 213--239218

The examples in (29--30) confirm that have/be yet to is both DE and anti-additive. In (29) we see the application of the DeMorgan law that isolates downward entailing contexts (i.e., the left-to-right application of the definition in 21). In (30), wesee the application of the definition in (21) from right-to-left, which identifies anti-additivity.7

(29)

7 Givenare know

(i) *Joh(ii) *Joh

Howeverlicensingand until

(iii) a.b.c.d.

Presumamonths/ydue to a sadditionaincompat

(iv) a.b.c.d.e.

We thank

a.

that thn to be

n hasn had

, we wostrongfive on

JohnJohn*John*John

bly theears imimilar cl complible wit

*JohJohJoh*Joh*Joh

two a

John has yet to visit Paris or Berlin.

is construction is anti-additive, the unacceptability of (i) and (ii) may seemstrong NPIs.

yet to visit Paris in years/in months.yet to visit Bill until five.

uld argue that these judgments do not threaten the conclusion that have yeNPIs). Instead, the ungrammaticality of (i) and (ii) reduces to an incompatibthe other, as shown in (iii).

hasn’t visited yet.hasn’t visited in years/months.hasn’t visited yet in years/months.hasn’t visited in years/months yet.

incompatibility in (i) follows from the fact that yet either entails or strongly iplicates that John has visited at some point in the past, yielding a contradicontradiction.Until five in (iv.a) implicates that John did, indeed, arrive at somication arises here.Yet requires perfect aspect and is incompatible with theh the perfect (iv.d). It is therefore doubly unsurprising that these two exp

n didn’t arrive yet.n hasn’t arrived yet.n didn’t arrive until five.n hasn’t arrived until five.n hasn’t/didn’t arrive until five yet.

nonymous reviewers for encouraging us to discuss these facts here.

)

surp

t to isility b

mpliction. Te po

simpressi

(DE)

b. John has yet to visit Paris and John has yet to visit Berlin.

(30)

a. John has yet to visit Paris and John has yet to visit Berlin. ) (AA) b. John has yet to visit Paris or Berlin.

Having established that have/be yet to is anti-additive, we now turn to a consideration of what, at first, seems to be a likelycandidate for a silent yet-licensor, namely, sentential negation.

3. Testing for silent negation

Recall the examples that we started with in (1), along with their paraphrases in (2), repeated here as (31) and(32).

(31)

a. John has yet to eat dinner. b. John is yet to eat dinner.

(32)

a. John hasn’t eaten dinner yet. b. John didn’t eat dinner yet.

Since the paraphrases in (32) naturally contain sentential negation, it is appealing to consider the hypothesis that silentsentential negation is indeed responsible for licensing yet in (31). We will argue, however, that adopting this hypothesiswould be a mistake.

A number of diagnostics have been used since Klima (1964) for determining the presence or absence of sententialnegation in a clause.

rising at first, since in years/months and until five

anti-additive (and thus expected to be capable ofetween yet on the one hand and in years/months

ates that John hasn’t visited in the past, while inhe incompatibility of yet and until five is arguablyint, while yet in (iv.b) implies that he didn’t. But anle past tense (iv.a).Until five, on the other hand, isons cannot co-occur in (ii) above.

S. Harves, N. Myler / Lingua 148 (2014) 213--239 219

(33)

8 Notestandard‘‘John ha

9 We th

Klima (1964) Tests

Sentential negation exists in a clause if: a.

*

*

*

*

*

*

that thistag ques yet toank Sa

It takes a positive rather than a negative tag question.

b. It can be continued with a conjunct headed by neither rather than so. c. It can be continued with a phrase that begins with not even. d. It can be continued with a conjunct that contains either rather than too.

A simple illustration of how these tests are used in negated versus non-negated sentences is given in (34--35).

(34)

a. John didn’t attend Mary’s lecture last week, did he/*didn’t he? b. John didn’t attend Mary’s lecture last week, and neither did Bill/*so did Bill. c. John didn’t attend Mary’s lecture last week, not even for a minute. d. John didn’t attend Mary’s lecture last week, and Bill didn’t either/*too.

(35)

a. John attended Mary’s lecture last week, *did he/didn’t he? b. John attended Mary’s lecture last week, and *neither did Bill/so did Bill. c. John attended Mary’s lecture last week, *not even for a minute. d. John attended Mary’s lecture last week, and Bill did, too/*either.

As (34) shows, sentences with sentential negation pass all of the Klima-tests, while the sentences in (35) show thatsentences which lack sentential negation fail these tests. An additional (post-Klima) test for negation comes from negativeparentheticals, as shown in (36).

(36)

a. John didn’t, I don’t think, attend Mary’s lecture last week. b. *John attended, I don’t think, Mary’s lecture last week.

With these tests in mind, let us now return to the construction under consideration here and apply these diagnostics tosentences with have/be yet to.

(37)

a. John has yet to eat dinner, has he?8

b.

John has yet to eat dinner, hasn’t he/doesn’t he? c. John has yet to eat dinner, and neither has Mary. d. John has yet to eat dinner, and so has/does Mary. e. John has yet to eat dinner, not even once. f. John has yet to eat dinner, and Bill hasn’t, either. g.

*John has yet to eat dinner, I think/*I don’t think.

(38)

a. John is yet to eat dinner, is he? b. John is yet to eat dinner, isn’t he? c. John is yet to eat dinner, and neither is Mary. d. John is yet to eat dinner, and so is Mary. e. John is yet to eat dinner, not even once. f. John is yet to eat dinner, and Bill isn’t, either. g.

*John is yet to eat dinner, I think/*I don’t think.

Application of the Klima tests as well as the test using negative parentheticals in (37--38) suggests that there is nosentential negation present in the have/be yet to construction. One might wonder, however, whether Klima’s tests applyonly to overt instances of negation, rather than instances of SILENT negation. If this is indeed the case, then one needs tocheck the Klima tests in caseswhere it can be argued that sentential negation is syntactically present but silent.9 Two suchcases will be considered here -- pas deletion in French and no deletion in Spanish and Catalan.

sentence is grammatical under a particular interpretation in some dialects of English, but with a reading that is different from astion interpretation. It means something like, ‘‘Aha! John has yet to eat dinner. Intriguing!’’ It does not have the interpretation,eat dinner, right?’’ Similar comments apply to example (38a).lvador Mascarenhas for bringing this issue to our attention.

S. Harves, N. Myler / Lingua 148 (2014) 213--239220

3.1. French pas-deletion10

A handful of verbs in French allow sentential negation (pas) to be deleted in a clause while maintaining a negativeinterpretation, suggesting that negation is still syntactically present. Such verbs are cesser ‘to cease,’ pouvoir ‘can,’ oser‘to dare,’ and savoir ‘to know.’11

(39)

10 See C11 We th12 Althodegree quinclude M13 FrencAlthough

(i) a.

b.

An at

(ii) a.

b.

See Sailoellipsis.

a.

ollinsank Vaugh, asantifiearquesh lacksFrench

OlivOliv‘Oliv

OlivOliv‘Oliv

tempt t

*OlivOliv‘Oliv

*OlivOliv‘Oli

r (2012

Olivier

and Postalentine Han anonyr (the revie(2003), Stag-quesdoes hav

ier n’aier PRT-haier hasn’t

ier aier hasier has go

o create t

ier n’aier PRT-hier hasn’t

ier aier hasvier has g

, to appea

ne

l (201acquamouswer cchwations oe tag-

pas nogone

visitévisitedne to

ag-que

paas nogone

visitévisiteone to

r) for

cessait

*

4) for morerd, Olivier Previewer noites Alrengarzschild (20f the depequestions,

s visitét visitedto Paris, h

Paris, nParis P

Paris, has

stions of t

s visitét visitedto Paris, h

Paris, ad Paris hParis, has

arguments

(pas)

on Frenauluis,tes, it haand Ke08) andndent tysuch as

Paris,Paris,as he?’

’est-ce pRT-is-it non’t he?’

he Engli

Paris,Paris,as he?’

-t-ilas-PRT-hn’t he?’

that dep

de

ch pasand Pies somennedy (Gajew

pe founn’est-c

n’est-cePRT-is-it

as?t

sh type

a-t-ilhas-PRT

/n-ae /PRT

endent

manger.

-deletion anrre Pica fortimes been2014), who sski (2008)).d in Englishe pas, they c

pas?not

results in u

/n-a-t-i-he /PRT-ha

-t-il pas?-has-PRT he

tag-question

Olivier

PRT ceased (not) of/to eat ‘Olivier didn’t stop eating.’

b.

Je ne sais (pas) ce qu’est l’amour. I PRT know (not) that what-is the-love ‘I don’t know what love is.’

That pas in the French ne pas construction is inherently negative, as opposed to ne, is supported by two well-known facts.First, ne has so-called ‘‘expletive’’ negative uses in certain contexts. That is, ne occurs in certain environments which are notobviously negative, somethingwhich isnot trueofpas.Onesuchenvironment is in comparatives (example fromRich, 2009).12

(40)

Luc en veut plus que Thierry n’ en a. Luc of.it wants more than Thierry PRT of.it has ‘Luc wants more of it than Thierry has.’

Second, in colloquial French ne is often dropped, maintaining a negative interpretation, while the same cannot be said forpas, with the exception of the handful of predicates mentioned above (examples in 39).

(41)

Olivier (ne) mange (pas) de chocolat. Olivier (PRT) eats (not) of chocolate *‘Olivier doesn’t eat chocolate.’

Since negation can be silent in (39), we might ask whether the Klima tests for negation can be applied to such sentences.Before applying any tests to the sentences in (39), however, we must first see whether Klima’s tests are applicable toFrench at all. As the sentences in (42--43) show, both the neither test and the not even X tests can be used in French.13

d Neg-deletion more broadly.the French judgments reported in this section.suggested for comparatives in English that they contain a covert negativetate that the idea goes back to Ross (1968), as a recent example. Others

and can therefore not be used as a reliable test for sentential negation.an be used after both positive and negative utterances, as shown in (i).

ngrammaticality, as the following examples show.

l pas?s-PRT he not

not

s are only available in languages that independently have canonical VP

(42) a. Olivier n’est pas allé à Paris. Marie non plus.

S. Harves, N. Myler / Lingua 148 (2014) 213--239 221

*

*

*

*

*

Olivier

PRT-is not gone to Paris. Marie neither. ‘Olivier hasn’t gone to Paris. Neither has Marie.’

b.

Olivier est allé à Paris. Marie non plus. Olivier is gone to Paris. Marie neither. ‘*Olivier has gone to Paris. Neither has Marie.’

c.

Olivier est allé à Paris. Marie aussi. Olivier is gone to Paris. Marie too/also. ‘Olivier has gone to Paris. Neither has Marie.’

(43)

a. Olivier n’a pas mangé son gâteau, même pas une miette. Olivier PRT-has not eaten his cake even not a crumb ‘Olivier hasn’t eaten his cake, not even a crumb.’

b.

Olivier a mangé son gâteau, même pas une miette. Olivier has eaten his cake, even not a crumb ‘*Olivier has eaten his cake, not even a crumb.’

If we apply these tests to sentences that allow pas-deletion, we see that the tests yield grammatical results.

(44)

a. Je ne sais ce qu’est l’amour. Olivier non plus. I PRT know that what-is the-love. Olivier neither ‘I don’t know what love is. Neither does Olivier.’

b.

Olivier ne cessera de jouer à la Playstation, même pas pour une seconde. Olivier PRT will-cease of/to play at the Playstation, even not for one second ‘Olivier won’t stop playing Playstation, not even for a second.’

This suggests that pas is still syntactically active, although silent. These facts lend independent support to our claim abovethat the have/be yet to construction in English does not contain an instance of silent sentential negation, since sentenceswith have/be yet to fail the Klima tests. If the construction had an occurrence of silent negation, we would expect suchsentences to pattern with instances of pas deletion in French, contrary to fact.

3.2. Ibero-Romance no-deletion

Further examples of silent sentential negation come from languages such as Spanish and Catalan. Spanish andCatalan are both Negative Concord languages, where n-words are licensed by an instance of overt sentential negation(no). However, when these n-words are preverbal, in Spanish, sentential negation is obligatorily dropped, as shown in(45b-c). In Catalan, sentential negation is optionally dropped, as shown in (46b-c).

(45)

Spanish (Zagona, 2002: 197--200) a. (No) vino nadie.

NEG

came nobody ‘Nobody came.’

b.

Nadie vino. Nobody came ‘Nobody came.’

c.

Nadie no vino. Nobody NEG came ‘Nobody came.’

(46)

Catalan (Espinal, 2000: 559) a. (No) ha vist ningú.

NEG

has seen nobody ‘S/he has not seen anybody.’

S. Harves, N. Myler / Lingua 148 (2014) 213--239222

14 We nthis idea.however,compatib15 Like F16 We th

b.

*

ote herSee, fohas ale withrench,ank Pa

Ningú

e that not ar example,rgued agaiour approaSpanish laola Ceped

(no)

ll analysHerburgnst thisch in thcks depa for jud

ha

es of thier (2001positione senseendentgments

vist

s construct), who argu, arguing ththat both atag-questioon these s

res.

nobody (NEG) has seen anything ‘Nobody has seen anything.’

c.

A NINGU (no) ha vist. P nobody (NEG) has seen ‘S/he has seen nobody.’

If we assume, as has been argued in the literature (seeHaegeman and Zanuttini, 1991; Uribe-Extebarría, 1994;and Zagona,2002, among others) that n-words in these languages are licensed by sentential negation, then we would expect theselanguages to passKlima’s tests for negation, evenwhenno is silent, as in (45b) and (46b).14WeuseSpanish as our test casehere, since deletion of no is obligatory in these contexts. We first show that the Klima tests are applicable to Spanish.15

As with French above, both the neither test (47) and the not even X test (48) can be used as diagnostics for thepresence or absence of negation in a preceding clause.16

(47)

a. Juan no vino nunca, María

ion res that tpprons,ente

tampoco

ely on the preat preverbahe n-wordsaches rely oso we cannonces.

/*tambien.

Juan NEG came never Maria neither /*also ‘Juan never came, neither did Mary.’

b.

Juan vino, María *tampoco /tambien. Juan came, Maria *neither /also ‘Juan came, Maria did, too.’

(48)

a. Juan no vino nunca, ni siquiera una

sel n-hernt u

vez.

Juan NEG came never not even one time ‘Juan never came, not even once.’

b.

Juan vino, ni siquiera una vez. Juan came, not even one time ‘*Juan came, not even once.’

As the sentences in (49) show, cases of no-deletion in Spanish pass these Klima tests.

(49)

Klima tests in no-deletion contexts a. Juan nunca vino, María tampoco /*tambien.

nce owordse area silese th

Juan

never came, Maria neither /*also ‘Juan never came, neither did Maria.’

b.

Juan nunca vino, ni siquiera una vez. Juan never came, not even one time ‘Juan never came, not even once.’

Hence, Spanish offers further evidence in support of the claim that Klima’s tests for negation hold not only of overtnegation but of silent negation as well. Since the have/be yet to construction does not pass any of the Klima tests, weconclude that this construction does not include an instance of sentential negation.

We are now left with the following set of options for a silent yet-licensor.

(50)

Potential yet-licensors a. Silent Negation e. Silent too Adj to Verb b. Silent few or rarely f. Silent without Verb-ing c. Silent negative implicative verb g. Silent Comparative d. Silent doubt h. Silent Superlative

f silent sentential negation. Some authors have argued againstare best treated as Negative Quantifiers. Zeijlstra (2004, 2008),licensed by a silent negative operator. Zeijlstra’s analysis is

nt NPI-licensor.em as a test for the presence or absence of negation.

S. Harves, N. Myler / Lingua 148 (2014) 213--239 223

In order to determine which of these options is viable, we now turn to a detailed discussion of the syntax of have/be yet to,beginning with have yet to.

4. The syntax of have yet to

One of the first questions that arises in examining the have yet to construction is, what kind of have are we dealingwith here? Is this Perfect auxiliary have? Modal have? Possessive have? Light verb have? As we have seen from theparaphrases of sentences with have yet to above, auxiliary Perfect (AUX) havematches the semantic interpretation ofthe construction but not the overt syntax. Modal have, on the other hand, matches the overt syntax, since it takes aninfinitival-to complement, but it does not appear to match the semantics. That is, have yet to does not appear to havethe deontic modality associated with it that modal have to does. This can be shown by considering the sentencesin (51).

(51)

17 An anhave yet

(i) ??

We do nofinds??Jomovemen18 We nproperties19 We nEnglish dthose speI-languaginvolved

a.

*

*

onymoto cons

John h

t find thhn hast of yeote tha‘‘Neg

ote herialectsakers wes of ais comp

John has yet to visit Paris.

b. John has to visit Paris. c. John has yet to visit Paris, although he doesn’t have to visit Paris.17

That (51a) does not entail (51b) is shown by (51c), where asserting the negation of (51b) does not give rise to acontradiction with respect to (51a). Thus, semantically speaking, Modal have does not seem to be an appropriatecandidate for this construction. In order to probe this further, we now turn to the syntax of have yet to.

4.1. Syntactic diagnostics: the NICE properties of have yet to

Given the semantic argument against Modal have being present in the have yet to construction, one should ideally beable to identify syntactic properties of the construction that support the semantic argument. One way of probing the syntaxof have here is to consider its so-called NICE properties (Huddleston, 1976).18

As is by now well known, Perfect AUX have differs from both modal have and possessive have with respect toNegation, Inversion, Contraction, and Ellipsis. The sentences in (52--54) show the following: (i) only AUX have precedesnegation and rejects do-support (a-b examples); (ii) only AUX have undergoes Inversion in questions (c-d examples); (iii)only AUX have Contracts with subjects (e examples); and (iv) only AUX have allows for Ellipsis of all lexical materialfollowing have, i.e., VP-ellipsis (f examples).19

(52)

a. I haven’t eaten lunch.

us reviewer notes that, for him/her, (51c) contrasts with (i),truction.

as failed to visit Paris yet, though he doesn’t have to vis

e same contrast. We suggest that the deviance of (i) in thefailed to visit Paris yet independently rather deviant. Thist into the specifier of FAILED (discussed below) is obligat the original NICE properties were defined as ‘‘Negatioation, Inversion, Contraction, and Ellipsis.’’ ‘‘Code’’ for Hue that we are reporting our own judgments as well as thowho accept the starred sentences in (54), although we rehose grammar reflects the judgments reported here.Wemll native speakers of English with respect to whether halex and subtle, and in some ways poorly understood (se

(AUX have)

b. I don’t have eaten lunch. c. Have I eaten lunch? d. Do I have eaten lunch? e. I’ve eaten lunch. f. I have eaten lunch, and Mary has/*does, too.

and asks why this would be if there were indeed a silent FAILED in the

it Paris.

judgment of the reviewer is related to the fact that the same reviewerfact in turn would be explained if, in the grammar of the reviewer, thetory.n, Inversion, Code, and Emphatic Affirmation.’’ Instead, we use theddleston (1976) is equivalent to what we today call VP-Ellipsis.se of our informants. We have not consulted with native speakers ofcognize that such dialects exist. Our analysis is intended to apply toake no attempt to present a unified analysis that would account for theve has auxiliary-like or lexical-verb-like behavior, since the variatione also footnote 21).

*

S. Harves, N. Myler / Lingua 148 (2014) 213--239224

(53)

20 The fafine-grainsame confinds thatsupport. Pmust leav

a.

*

*

*

*

ct thated appnectioncontracresume for f

I haven’t to eat lunch.

contraction is the ‘‘easiest’’ auxiliary-like property for thisroach to the nature of the NICE properties than the traditio, it is interesting to note that Thoms (2012), looking at Britistion is the most widely accepted such property, with fewerably, then, there is a deeper reason for the relative ease ofuture research.

(Modal have)

b. I don’t have to eat lunch. c. Have I to eat lunch? d. Do I have to eat lunch? e. I’ve to eat lunch. f. I have to eat lunch, and Mary *has/does, too.

*

(54)

a. I haven’t a new car. (Poss have) b. I don’t have a new car c. Have I a new car? d. Do I have a new car? e. I’ve a new car. f. I have a new car, and Mary *has/does, too.

Having laid out a number of syntactic diagnostics for distinguishing between three variants of have, we now return to thehave yet to construction. As the data in (55) show, the facts are not crystal clear. A survey of 11 native speakers of Englishyields the following results (we return presently to what is meant by the notation ‘‘%/*’’ next to the negation cases).

(55)

NICE Properties with have yet to a. %/* John hasn’t yet to win the hearts of his classmates.

have to have has intriguing impnal distinction between auxiliah dialects in which possessivespeakers also allowing inversiocontraction in the have yet to co

(Negation)

b. %/* John doesn’t have yet to win the hearts of his classmates. c. % Has John yet to win the hearts of his classmates? (Inversion) d. % Does John have yet to win the hearts of his classmates? e. I’ve yet to win the hearts of my classmates. (Contraction) f. % John has yet to win the hearts of his classmates, and Bill has, too. (Ellipsis) g. % John has yet to win the hearts of his classmates, and Bill does, too.

The % sign here conceals the nature of the variation in play, which is worth breaking down in more detail. Broadlyspeaking, there are four types of speaker. For one group, have in this construction patterns consistently with auxiliaryhave. For another group, have consistently patterns with lexical have (with the exception of allowing contraction, which isavailable for everyone).20 A third group of speakers, which includes both of the present authors, consists of speakers whoseem to permit both types of have. Finally, there are speakers that accept simple declarative cases with or withoutcontraction, but find more complex cases involving T-to-C movement or VP-ellipsis ungrammatical. We suspect that thesespeakers do not have this construction as a productive part of their I-language, instead having only a passive knowledge ofthe construction fromhearing other speakers use it. Their passive knowledge of the construction thus allows them to rate thesimple declarative cases as fully acceptable, but their intuitions collapse when presented with more complex cases.

We now address the issue of the judgments on negation. The notation %/* in (55) reflects the fact that speakers rejectan occurrence of overt negation in this construction out of the blue, regardless of whether these speakers tend to treathave in this construction as AUX have or modal/possessive have (i.e., whether they allow do-support or not). It seems thatin order for sentential negation to be grammatical in the have yet to construction, it must be interpreted as the negation ofdenial, and not standard sentential negation. When provided with a context like (56), speakers accept sentential negationin accordance with whether they treat have as auxiliary, lexical, or both.

(56)

Speaker A: John has yet to win the hearts of his classmates.

Speaker B:

What do you mean? John hasn’t yet to win the hearts of his classmates. He already has!

Speaker B’:

What do you mean? John doesn’t have yet to win the hearts of his classmates. He already did!

lications, since it suggests that a morery and lexical verb is necessary. In thehave has auxiliary-like properties, alson, negation and VP-ellipsis without do-nstruction, the elucidation of which we

S. Harves, N. Myler / Lingua 148 (2014) 213--239 225

We return to this issue below, showing that our analysis in terms of the presence of a silent negative implicative predicateFAILED accounts for this effect as a subcase of yet’s behavior when forced to take scope under both a sentential negationand a negative implicative verb simultaneously.

To summarize this part of the discussion, it must be accepted that there is a great deal of variation in how differentidiolects of English represent the have yet to construction syntactically. For some speakers, it is clearly main verb have,although even these speakers permit contraction of have in the have yet to construction (this is unusual for main verb havein American English; although see footnote 21). For others, it is clearly auxiliary have. Still others arguably do not have theconstruction as a productive part of their I-language, but have some passive knowledge of its simple cases. The amount ofvariation here is striking, and it is legitimate to ask why there should be so much variation in how this construction isrepresented by native English speakers.

Our proposal belowwill be that the have yet to construction always contains a silent element heading the complement ofhave. This idea proves to contain the seed of the answer to the question as to why speakers vary so much in how theyrepresent the construction. On our proposal, English speakers acquiring the have yet to construction are forced to come tosomeconclusionabout the categoryof the silent elementwhichheadshave’s complement. In generatinga hypothesis aboutthe category of this silent head, the learner will be constrained by her knowledge of have’s complementation possibilities inEnglish generally.Of course,have turns out to havea range of possible complements, includingDP (formain verbhave) anda participial verb phrase (for auxiliary have). In consequence, the learner’s hypothesis space is narrowed down by, but notfully determinedby, her knowledgeofhave’s selectional properties. Facedwith this indeterminacybetween lexical verbhaveand perfect auxiliary have, the learner makes a guess one way or the other, or hedges her bets by assigning two possiblestructures to the construction- one which employs lexical have, and one which employs auxiliary have.

In the following sections, we will concentrate on how to derive the version of this construction in which have is treatedas an auxiliary. It nonetheless behooves us to sketch how the version of the construction which employs lexical have isrepresented, and we will make a brief suggestion concerning this issue.

4.2. FAILED as the silent licensor of yet

So far we have concluded that have in the have yet to construction is AUX have for a number of speakers (although notall), and that silent sentential negation is not responsible for licensing yet. Concentrating on speakers who have AUX havein this construction, we are now in a position to choose among our remaining alternatives for a silent licensor of yet.

(57)

21 One mEnglish. Isomewhathat there

Potential yet-licensors

a.

ight qt is certt weakeis such

Silent Negation

uestion whether the silent element preseainly a logical possibility that the elemenr analysis in which the silent predicate haa counterpart, even while pointing out po

e.

nt in tht in ques no ovtential p

Silent too Adj to Verb

b. Silent few or rarely f. Silent without Verb-ing c. Silent negative implicative verb g. Silent Comparative d. Silent doubt h. Silent Superlative

Since perfect AUX have requires a perfect participle, we are left with the options in (c) and (d). We can quickly rule outdoubt as a contender on both syntactic and semantic grounds. The sentence in (58a), for example, is not paraphrasableby the sentence in (58b). It is both syntactically ill-formed and semantically inappropriate as a potential paraphrase.

(58)

a. John has yet to visit Paris. b. *John has doubted to visit Paris yet.

There is no presupposition, entailment, or implicature that John has any doubts about not visiting Paris in (58a). He can bequite certain that he has in fact not been to Paris yet. Furthermore, the selectional properties of doubt appear to rule it outas a candidate, since doubt does not select for infinitival complements, but rather, finite complements, as in (59).

(59)

John doubts that he visited Paris.

We are therefore left with negative implicative verbs such as prevent, refuse, fail, forget as potentials for the silent perfectparticiple in this construction.21Prevent can immediately be ruled out on syntactic grounds, since it does not take infinitivalcomplements.

is construction need be assumed to be synonymous with an overt lexical item instion has no overt counterpart, and it is therefore always possible to retreat to aert counterpart in English. However, we prefer to focus on the stronger hypothesisroblems for it as we go. We thank two anonymous reviewers for raising this point.

(60) *John has prevented to visit Paris.

S. Harves, N. Myler / Lingua 148 (2014) 213--239226

Refuse and forget, while taking infinitival complements, as shown in (61b), do not appear to be semantically appropriateas silent licensors here, since the have yet to construction does not entail or presuppose that the subject has eitherrefused or forgotten anything.

(61)

a. John has yet to visit Paris. ≠ b. John has refused/forgotten to visit Paris.

Fail, on the other hand, appears to be both syntactically and semantically appropriate as a paraphrase for this construction(62b).

(62)

a. John has yet to visit Paris. ) b. John has failed to visit Paris yet.

We know that fail is downward-entailing and that it licenses NPIs in its infinitival complement. For instance, if John hasfailed to ever visit France (63a), then it follows that he has failed to ever visit Paris (63b).

(63)

a. John has failed to ever visit France. ) b. John has failed to ever visit Paris.

If silent FAILED is indeed syntactically present in the have yet to construction, then we expect such sentences to bedownward entailing as well, which is indeed the case, as shown in (64).

(64)

a. John has yet to visit France. ) b. John has yet to visit Paris.

Furthermore, recall the discussion of anti-additivity above. We showed that sentences with have yet to are anti-additive(repeated here as 65--66).

(65)

a. John has yet to visit Paris or Berlin. ) b. John has yet to visit Paris and John has yet to visit Berlin.

(66)

a. John has yet to visit Paris and John has yet to visit Berlin. ) b. John has yet to visit Paris or Berlin.

We would therefore expect that sentences with overt fail should yield anti-additive contexts as well. Indeed, as shown in(67--68), this prediction is borne out.

(67)

a. John has failed to visit Paris or Berlin. ) b. John has failed to visit Paris and John has failed to visit Berlin.

(68)

a. John has failed to visit Paris and John has failed to visit Berlin. ) b. John has failed to visit Paris or Berlin.

An additional prediction made by our proposal is that sentences with have yet to will pass diagnostics for raisingpredicates, since fail behaves as a raising predicate. The sentences below show that fail allows for idiomatic readings withdiscontinuous idioms (69a) and also allows for expletive subjects (69b-c), two properties shared by raising predicates butnot control predicates.

(69)

a. The shit has failed to hit the fan. b. There has failed to be a comprehensive discussion of this topic in the literature. c. It has failed to snow all weekend.

S. Harves, N. Myler / Lingua 148 (2014) 213--239 227

As the examples in (70) show, sentences with have yet to pass these same diagnostic tests, which suggests that the silentpredicate is indeed a raising verb, not a control verb.

(70)

22 We th23 An anyet whenhowever,

a.

ank Chonymoit comthat th

The shit has yet to hit the fan.

b. There has yet to be a comprehensive discussion of this topic in the literature. c. It has yet to snow all weekend.

Another piece of independent evidence in favor of a silent predicate FAILED in the have yet to construction, as opposed tosilent sentential negation or some other NPI licensor, comes from quite-modification of yet, at least in the grammars ofsome speakers.22 Certain NPI-licensing environments allow for the existence of either yet or quite yet. Several of thesecontexts are shown in (71).

(71)

a. John hasn’t arrived (quite) yet. b. He is too young to understand this (quite) yet. c. I doubt the lamp is fixed (quite) yet. d. Mary has crossed the border without realizing it (quite) yet.

However, not all contexts that license yet licensequite yet.While yes-noquestionssuchas (72a) and superlativesas in (72b)license yet, quite yet is disallowed. And as (73) shows, quite yet is disallowed from occurring in the have yet to construction.

(72)

a. Has John arrived (*quite) yet? b. This is the best film that has been shown (*quite) yet.

(73)

a. John has (*quite) yet to eat his dinner. b. John has (*quite) yet to visit Paris.

If one of the NPI licensors in (71) were the silent element responsible for licensing yet in (72), then the ungrammaticality ofquite yet would be mysterious. As the examples in (74) show, quite yet is not licensed in the presence of overt fail. Hence,we do not expect it to be licensed when FAIL is silent, as in (73).

(74)

a. John has failed to visit Paris (*quite) yet.23

b.

Mary has failed to write her grandmother (*quite) yet.

As a final argument for our analysis in terms of FAIL, note that it yields an explanation for the difficulty of using sententialnegation in the have/be yet to construction. In order to derive this construction on our approach, yet must merge in theembedded clause and raise into the Specifier of FAILED (the details of this are laid out in the next subsection). In anegative context, this will mean merging yet in the scope of both Negation and the negative implicative verb. It turns outthat yet is ungrammatical with such scope.

(75)

a. John has failed to arrive yet (at his final destination). b. John hasn’t failed to arrive yet (*at his final destination).

Example (75b) has grammatical readings in which yet is interpreted as modifying the VP failed to arrive (at his finaldestination). Thesematrix readingshave thescopen’t > yet > failed. Thereare twosuch readings, paraphrasableas follows:

(76)

a. It is not yet the case that John has officially failed to arrive (i.e., we can’t yet discount the possibilitythat he will arrive eventually).

b.

There has not yet been an event in which John failed to arrive (i.e., John has always managed toarrive in the past).

(75b) is ungrammatical if yet is interpretedwith theVParrive (at his final destination) unless the sentence is interpreted asa direct rejection of something someone has just said. Turning to have yet to, we find it acts like (75b) on the embedded

ris Collins for bringing quite-modification to our attention.us reviewer finds these examples fine, and thus does not have the claimed parallelism between have yet to and have failed to . . .

es to quite modification. It follows that the argument from quite yet cannot be used for the grammar of this speaker. We note,e other tests employed in this section do go through for the reviewer, so that our overall conclusions are not threatened.

S. Harves, N. Myler / Lingua 148 (2014) 213--239228

interpretation of yet. The example in (77) is grammatical only as an explicit denial of something that someone hasjust said.

(77)

24 By asatisfy thThis obliin existe

(i) I(ii) T25 We ninterpretfor his drdoesn’t mfailed to

John {hasn’t / doesn’t have} yet to arrive at his final destination.

This follows from our approach, because in order to license silent FAILED, yet must be merged in the embedded clause,

where it takes scope, and then raised into FAILED’s specifier (for more details on this derivation, see the next section).Merging yet straight into the matrix clause will not allow FAILED to be silent, and so the absence of matrix interpretations foryet in (77) isexplained.Mergingyetstraight into thematrix clause insteadyieldsoneof the following,dependinguponhow yetis linearized.

(78)

s

gn

dh

ai

p

a.

sumptie Neg Catory rtial con

idn’t seere areote hertion, naving exange tass hi

John hasn’t failed to arrive (at his final destination) yet.

b. John hasn’t yet failed to arrive (at his final destination).

In other words, yet cannot appear in the scope of both negation and a negative implicative verb at the same time except on

an explicit denial reading. Since, on our account, deriving a have yet to construction requires first-merging yet in theembedded clause and interpreting it there,24 yetwill be interpreted under the scope of both negation and FAILED, and wecorrectly predict that have yet to should have an explicit denial reading under negation.

Before moving on to the syntactic derivation of sentences with have yet to, let us briefly summarize the results of thissection.We have argued that yet in the have yet to construction is licensed by the silent negative implicative verb FAIL. Wesaw a number of arguments in favor of such an analysis, based on the shared behavior of sentences with overt fail vs.silent FAIL (i.e., have yet to). We summarize the shared properties of these constructions in (79).

(79)

Shared properties of sentences with have yet to and sentences with overt fail a. Both constructions are downward entailing and anti-additive (63--68). b. Both constructions pass diagnostics for raising predicates (69--70). c. Neither construction allows yet to be modified by quite (73--74). d. Neither construction allows yet to be interpreted in the embedded clause under negation (75--77).

One issue that we have yet to address in full is the position of yet in sentences with have yet to compared to sentences

where fail is overt. We turn to this issue now and use the differences in placement/pronunciation of yet as the key tounderstanding the licensing of silent FAIL.

4.3. The syntactic derivation of have yet to

Thus far, we have only briefly alluded to the position of yet in the sentences under investigation here. Since we arearguing that sentences with have yet to vs. those with overt fail share an underlying structure, it is perhaps surprising thatyet is pronounced in two different places, depending on the pronunciation or silence of fail. Consider the position of yet inthe examples in (80).

(80)

a. John has yet to eat dinner. b. John has yet to visit Paris. c. John has failed to eat dinner yet. d. John has failed to visit Paris yet.

When failed is overt, yet is obligatorily pronounced sentence-finally. That is, yet cannot be pronounced adjacent to have

when failed is overt, as shown in (81).25

on, yetmay not be interpreted as scoping above FAILED even after moving into the latter’s specifier. This follows if movement toriterion of Haegeman and Zanuttini (1991), the spirit of whose proposal we follow in the next section, reconstructs obligatorily.

econstruction must be assumed in any case, in order to explain why NPIs (which pattern like indefinites -- cf. their grammaticalitystructions, as in (ii)) do not have readings corresponding to the existential quantifier having scope over the sentential negation:

e any boys. (: > 9; *9 > :)n’t any boys here.e that some speakers allow for yet to occur in this position in certain cases. However, when it does, it seems to have a differentmely, one that is more akin to nevertheless. For example, imagine the following scenario. John has been studying and practicingam for months, in hopes of getting his driver’s license. The day of his driving exam comes, but in spite of all his practicing, he stillo pass the exam. In this context, some speakers allow the following: In spite of all his preparation, John has yet (= nevertheless)s exam.

S. Harves, N. Myler / Lingua 148 (2014) 213--239 229

(81)

a.

*

*

*

*

*

John has yet failed to eat dinner.

b.

**John has yet failed to visit Paris.

One crucial fact about fail as an NPI licensor is that this predicate only licenses NPIs within its infinitival complement. It

does not license NPIs in its own clause.

(82)

a. John has ever failed to visit France. b. John has failed to ever visit France. c. John has failed any exam. d. John hasn’t failed any exam. e. John has failed to pass any exam.

In (82a-b) we see that fail cannot license the NPI ever in the matrix clause, although it can license it within its infinitival

complement. Similarly, if fail occurs in its transitive guise, as in (82c), it cannot license an NPI in the direct object positionunless sentential negation occurs as well (82d). That fail is incapable of licensing NPIs within its own clause suggests thatyet in the have yet to construction Externally Merges in the infinitival clause and then Moves to/Internally Merges in thematrix clause. We propose that yet raises to the Specifier of FAIL in the have yet to construction, and that this raising isprecisely what accounts for the silence of FAIL. The derivation we propose is given in (83).

(83)

[TD$INLINE]

TPT’

PerfP

vP/VP

V’

TP

T’

vP

<John> visit Paris <yet>

Tto

<John>FAILED

Vyet

AdvP

<has>Perf+ T

has

DP

John

If yet raises to the Specifier of FAILED, then yet is pronounced in the Specifier position, and the head FAILED is silent. If yetremains in situ, then failed is pronounced in the matrix clause. This type of analysis is reminiscent of certain analysesproposed for constructions involving NPIs discussed earlier, namely, Neg-deletion/no-deletion in Romance. Recall theSpanish and Catalan facts presented in Section 4.3 above.

(84)

Spanish (Zagona, 2002: 197--200) a. (No) vino nadie.

NEG

came nobody ‘Nobody came.’

b.

Nadie vino. Nobody came ‘Nobody came.’

c.

Nadie no vino. Nobody NEG came ‘Nobody came.’

(85)

Catalan (Espinal, 2000: 559) a. (No) ha vist ningú.

NEG

has seen nobody ‘S/he has not seen anybody.’

b.

Ningú (no) ha vist res.

S. Harves, N. Myler / Lingua 148 (2014) 213--239230

26 We wThese el27 One qovergene

(i) *Joh

Any syntasilent preAnother epredicateappeal toattracts toone tech

*

ant to bementsuestionrate po

n has

ctic anpositionxampleonly sisomeits spe

nically

nobody

e careful nhave verythat two essibilities?

ever to finis

alysis that rTO raisesis Marusič

lent in the pnotion of secifier (hereimplements

(NEG)

ot to dradifferenxternal rTo be m

h his di

elies onthe queand Zauresencelection hyet), in othis, it i

has

w too ct propeevieweore sp

nner.

a particstioncer’sof a reere. Order tos a qu

seen

lose a crties, asrs raiseecific, w

ular SILas to wh(2006) aflexive cne coulcaptureestion fo

anything

‘Nobody has seen anything.’

c.

A NINGU (no) ha vist. P nobody (NEG) has seen ‘S/he has seen nobody.’

In both Spanish and Catalan, when an NPI/n-word raises past sentential negation, the Neg-head is silent, obligatorily so inSpanish and optionally in Catalan. One can argue, following Haegeman and Zanuttini (1991) and Zagona (2002), thatsatisfaction of something akin to the Neg-Criterion results in silence of the Neg head.

(86)

The Neg Criterion (Haegem

omparison between FAILED and negation on the one hanis well known.is, why is it only the NPI yet that raises/participates in thihy aren’t sentences such as (i) grammatical?

ENT lexical item is faced with this type of question. For exy it is only the lexical item home that licenses the silencnalysis of silent FEEL-LIKE in various Slavic languages (blitic in these languages? And why is it only the predicate Fd posit some matching feature X on both the licensing hethe fact that only particular lexical items participate in varior all analyses that allow for silent lexical elements in th

an and Zanuttini, 1991: 244)

a. Each Neg X0 must be in a Spec-Head relation with a Negative operator; b. Each Negative operator must be in a Spec-Head relation with a Neg X0.

The silence of the Neg-head could be argued to derive from Koopman’s (1996) Generalized Doubly-Filled Comp Filter,defined in Koopman and Szabolcsi (2000) as follows (see also Collins, 2007 and Nchare and Terzi, 2014).

(87)

Generalized Doubly-Filled Comp Filter (Koopman & Szabolcsi 2000: 4) No projection has both an overt specifier and an overt head at the end of the derivation.

Weargue here that the silence of FAILED in the have yet to construction is due to the same principle governing the silenceof negation with fronted n-words in Spanish and Catalan. It is unlikely to be a mere coincidence that in both cases we aredealing with the silence or pronunciation of an NPI licensing head.26 If, on the other hand, such an analysis turns out not tobe correct for n-words in Spanish and Catalan, then have yet to is more akin to other constructions where the silence of ahead is argued to follow from the Generalized Doubly-Filled Comp Filter as the result of movement into that head’sspecifier (as in the references mentioned above).27

To conclude this section, we return to a puzzle raised at the outset of this paper, namely, why is the sentence in (88a)ungrammatical, given the paraphrases of sentences with have yet to?

(88)

a. John has yet eaten dinner. b. John has yet to eat dinner. c. John hasn’t eaten dinner yet.

At this point our answer should be clear. The sentence in (88a) is ungrammatical because there is no silent negativeimplicative verb FAILED available in this sentence. The perfect participle eaten is filling the verbal head where FAILEDwould occur. Moreover, there is no silent sentential negation here.

Before proceeding, we make a brief suggestion regarding how the version of have yet to that contains lexical have is tobe accounted for. For those speakers who allow a version of have yet to involving lexical/possessive have, we mightsuggest that the silent predicate involved is the derived nominal form of a negative implicative verb, with have acting in itslight verb use. The overt counterpart of the relevant derived nominal might, in fact, be failure, since some speakers acceptovert failure in a light verb construction along the lines of (89).

d and yet and n-words on the other.

s construction? Doesn’t our analysis

ample, Collins’ (2007) analysis of thee of the preposition here in English.ut especially Slovenian). Why is thisEEL-LIKE? It seems that we need toad (here FAIL) and the phrase that itus constructions. Regardless of howe syntax.

S. Harves, N. Myler / Lingua 148 (2014) 213--239 231

(89)

28 We notherwisthis spea29 An aone mighThe pointhe haverepresenposition

%John has had a failure to do his homework yet this year.

More investigation is needed to discover to what extent the availability of (89) in a given I-language correlates with theavailability of the lexical/possessive have version of have yet to. If the correlation does not go through for all speakers, wewill have to conclude a different derived nominal than failure is involved. All that our thesis requires is that the derivednominal in question be negative implicative in nature.

4.4. Challenges and alternatives

Beatrice Santorini has proposed to us (pers. comm.) a very interesting counteranalysis of the have yet to constructionwhichat first sightalso seemsable to capture thedatahere.Shepoints out thatyet’s statusasanNPIemergedover time,andthat historically yet had non-NPI uses. Furthermore, some such non-NPI uses remain in modern English in a restricted way,often (but not always) with an archaic flavor. Example (90a) is an example from EarlyModern English drawn from the Penn-Helsinki Parsed Corpus of Early Modern English; (90b) is a modern attestation from the script of a video game in which theEnglish dialog is often deliberately archaic; (90c) represents cases involvingmodals which do not sound archaic to our ears.

(90)

e

n

t

u

a.

ote herapparker.onymot wondeis wellyet to cts this csually

I have yet many thinges to saye vnto you (TYNDNEW-E1-P1,16,1J.326)

b. You are young yet, Raziel- you still retain many of your vampiric weaknesses.

(from Legacy of Kain: Soul Reaver)

c. He might/could yet arrive.

Santorini’s suggestion is thus that the have yet to construction involves two sorts of archaized syntax. First, she proposesthat this version of yet is representative of the older, pre-NPI stage; second, the unusual word-order arises because have(which she proposes is themodalhave inJohnhas to leave) undergoesV-to-Tmovement, just as it did inearlierEnglish (andstill does in some British varieties). The existence of speakers who reject instances of the construction in interrogatives, VPellipsis cases etc. can then be explained as a case of their not having full control over this archaic syntax.

There are four reasons why we do not adopt this analysis. The first is that there are speakers who robustly allow do-support in this construction in interrogatives, under negation, and in VP-ellipsis. This is totally unexpected if the word orderwith have preceding yet is produced by V-to-T movement, for then we would expect do-support to be ruled out entirely.The second is that cases involving modals like (90c) do not pattern with have yet to as regards negation. While it ispossible to negate have yet towith an explicit denial interpretation, it seems to us that this is not possible at all in cases like(90c). Hence, (91) is completely impossible for the present authors and many other speakers we have consulted, even ifused to directly contradict someone who has uttered (90c).28

(91)

*Wait a minute, he {mightn’t / couldn’t} yet arrive! (We already know he can’t make it.)

Third, in relying on the have yet to construction involving modal have, Santorini’s suggestion falls afoul of the semanticevidence we adduced earlier to show that the have yet to construction does not entail the existence of an obligation, asmodal have does. Recall, for instance, the fact that (92) (repeated here from (51c)) does not yield a contradiction.

(92)

John has yet to visit Paris, although he doesn’t have to visit Paris.

Fourth, as an anonymous reviewer points out to us, themodal have to construction generally rules out interveningmaterialbetween have and to. The very form of the have yet to construction sets it apart in this respect.29

e that an anonymous reviewer dissents from this judgment, finding this example grammatical. However, the same reviewerently agrees with the judgments in this section. Therefore, the other three arguments presented in this section still go through for

us reviewer points out that, given that there seem to be many ways of representing the have yet to construction cross-dialectally,r whether there will be some speakers who represent the have yet to construction in the manner suggested by Santorini after all.-taken, however, we note that we have not been able to find any speakers for whom the existence of an obligation is entailed byonstruction in the same way as it is for the modal have construction. This gives rise to a further question: if, in fact, no speakeronstruction using modal have, why would this be? We suggest that the presence of yet in between have and infinitive to, aimpossible with adverbs in the modal have construction (see (93)), might prevent learners from entertaining this hypothesis.

(93) I have (*urgently/*definitely) to finish this review.

S. Harves, N. Myler / Lingua 148 (2014) 213--239232

30 We t31 -sja

A second potential counteranalysis might accept that the have yet to construction involves the modern English NPIversion of yet, but attempt to explain how yet is licensed in some way that does not involve the postulation of a silentlicensor. Such an approach is suggested to us by an anonymous reviewer, who points to Linebarger’s (1987) argumentsthat NPIs can be licensed by a negative implicature, and suggests wemight construe such a negative implicature from thehave yet to construction. The question then becomes where this negative implicature might come from in the absence of asilent negative implicative predicate. The reviewer offers the following as a way of deriving such an implicature from themeaning of the have yet to construction: ‘‘there is an obligation at some time t0, there is still an obligation at t1 > t0, so whathad to be done to remove the obligation did not take place between t0 and t1, so the speaker/subject failed to do so, at leastbefore t1’’. Notice, though, that this suggested implicature hinges on the have in have yet to being modal have- otherwise,there is no source for the ‘‘there is an obligation at time t000 part of the meaning for have yet to that the reviewer suggests,and therefore no way to get the chain of implicational reasoning required by this suggestion off the ground. We havealready shown while discussing (92) and (93) that the have in this construction cannot possibly be modal have, and weconclude that the reviewer’s suggestion cannot be defended, even if Linebarger’s (1987) arguments are accepted, unlesssome other plausible source for the putative negative implicature is suggested. While we cannot claim in this paper tohave ruled out all imaginable sources for such an implicature, nor does it seem obvious to us what the source of thisimplicature would be given that it cannot possibly be modal have, and we will leave it as an open challenge for those whowould counteranalyse the have yet to construction in these terms to develop such an account.

Wewould furthermore point out that it is far from self-evident that the idea of licensing by negative implicature is correct.For one, Kadmon and Landman (1993) point out numerous problems with this notion, most of which stem from issues ofovergeneration -- it is almost always possible to conjure up a negative implicature for a sentence, and it is not clear whetherthe sorts of negative implicatures that ‘‘count’’ for NPI licensing can be characterized in a general enough way to avoid theprediction that NPIs should be licensed everywhere.

A different challenge to our analysis comes from the pragmatics of the constructions we are seeking to relate.While ouranalysis manages to capture all of the syntactic similarities between sentences containing overt instances of fail and silentinstances of FAIL, there is one issue regarding the interpretation of sentences with have yet to that is unexpected underour analysis.30 Consider the sentences in (94).

(94)

his

a.

ank outhe R

He is gravely ill, to be sure, but he has {yet/?failed} to die.

b. The bomb has {yet/failed} to go off. c. Luckily, the air conditioning has {yet/?failed} to break down.

As an anonymous reviewer points out, the sentences in (94a) and (94c) are perfectly natural if yet is pronounced, ratherthan fail. When fail is overt, the sentences become degraded. Sentence (94b), on the other hand, is most naturally utteredwith overt fail if we presuppose that it would be a good thing for the bomb to explode -- in other words, it would be strangefor the members of a bomb-disposal squad to pronounce (94b) with fail, but quite normal for the members of a terroristgroup to do so. The generalization appears to be that overt fail induces a weak implicature that the eventuality denoted bythe infinitival complement is a desirable outcome, whereas have yet to is more neutral. The question for our analysis iswhether this pragmatic difference renders it implausible to assume that the silent predicate in the have yet to constructionis an instantiation of fail.

While a truth-conditional difference between have yet to and have failed to would certainly compel us to abandon thisanalysis, it is not clear that such a pragmatic, non-truth-conditional difference need be taken as discomfirming. The questionturns onwhether pragmatic elements ofmeaning are computed purely on the basis of the output of syntax, asweassume forsemantics proper, or whether specific details of the phonology of a construction can also play a role in computing pragmaticmeanings (in awaywewouldwant to exclude for semantics itself). The issue is complex, andwe cannot do justice to it here,but it doesnot seem implausible that certainmore ‘‘surface-y’’properties of anutterance canplay into pragmatic reasoning. Ifso, then the silent and overt formsof a single syntactic element couldwell turn out to havedifferent pragmatic effects, and thiscould be what it is at issue in the have yet to construction with respect to silent vs. overt failed.

Such a state of affairs is, in fact, suggested by Marusič and Zaucer (2006) in their discussion of the impersonal ‘feel-like’ construction in Slavic. They discuss slight pragmatic differences between the silent and overt forms of the verb FEEL-LIKE (which is overtly realized as a verb equivalent to English ‘want’) in various Slavic languages. By way of illustration,consider examples of this construction from Russian in (95).31

r reviewers for raising this issue, and we particularly thank one reviewer for providing us with the examples given in (94) here.ussian equivalent of se/si in Romance languages.

(95) a. Mne ne xočetsja rabotat’. (Russian)

S. Harves, N. Myler / Lingua 148 (2014) 213--239 233

32 We th33 We noone Ame

ank Mte thatrican d

aria Gousknot all natialect and

ova, Sive speone B

onia Kasyanenkakers of Englishritish dialect of En

o, and Inna Lappear to haglish.

Me.DAT

NEG want.3SG.sja work.INF ‘I don’t feel like working.’

b.

Mne ne rabotaetsja. Me.DAT NEG work.3SG.sja ‘I can’t seem to work.’

(96)

a. Mne ne xočetsja spat’. Me.DAT NEG want.3SG.sja sleep.INF ‘I’m not sleepy.’

b.

Mne ne spitsja. Me.DAT NEG sleep.3SG.sja ‘I can’t seem to sleep.’

In Russian, the sentences in (95a) and (96a) have a reading that highlights a ‘physiological drive,’ to use Marusič andZaucer’s (2006: 1151) term. The ‘feel-like’ translation is meant to capture this intuition. The (b) sentences, on the otherhand, do not share this reading.32 These sentences have a reading that is meant to be captured by the translations here.The speaker might want to work or sleep, but something external is preventing them from doing so. Marusič and Zaucer(2006: 1150) conclude from this that while the overt vs. covert/silent predicates in these constructions are syntacticallyparallel, they are only near-synonyms.

Whether we should conclude, as they have, that there is a silent predicate in the (b) sentences above that has no directovert counterpart in the language, or whether we should argue that the (non-)pronunciation of a predicate can have animpact on pragmatics, is an open question. If the covert counterpart of fail in English is best treated as a negativeimplicative verb that is only a near-synonym with overt fail, the observations presented in (94) can be accounted for.

Having addressed some challenges and alternatives to our analysis of have yet to, we nowmove on to a considerationof the be yet to construction.

5. The syntax of be yet to

Having presented our analysis of the have yet to construction, we now turn our attention to be yet to. On the surface,there appears to be free variability with respect to the choice of be or have here. More specifically, it is not obvious thatthere are differences in interpretation with respect to the choice of be or have in this construction, at least for thosespeakers who have both variants as part of their grammar.33

(97)

a. John is/has yet to visit Paris. b. John is/has yet to finish his homework.

Our proposal involving a silent perfect participle FAILED for have yet to cannot obviously be extended here, given theungrammaticality of the sentence in (98).

(98)

*John is failed to visit Paris yet.

That (98) is ungrammatical is not surprising.Be is not an option for a Perfect auxiliary verb in English.Be is grammatical asa progressive auxiliary verb, however, but as (99) shows, a progressive participle failing is incompatible with yet.

(99)

*John is failing to visit Paris yet.

Yet requires perfect aspect and is incompatible with the progressive, even in the presence of negation (100c). Thus, asilent progressive predicate will not be a suitable licensor here.

ivitz for the Russian judgments here.ve the be yet to construction, although the authors of this paper do. We represent

(100) a. John is always failing to eat dinner.

S. Harves, N. Myler / Lingua 148 (2014) 213--239234

34 Again,

*

*

this se

b.

John is always failing to eat dinner yet. c.

**John isn’t always failing to eat dinner yet.

In order to determine what licenses yet in the be yet to construction, wemust return to the same questions asked above forhave yet to. What syntactic and semantic properties are characteristic of this construction? Like have yet to, the be yet toconstruction is also anti-additive.

(101)

a. John is yet to visit Paris or Berlin.

ntence is unacceptable under the intended reading, i.e., John is yet

)

b. John is yet to visit Paris and John is yet to visit Berlin.

(102)

a. John is yet to visit Paris and John is yet to visit Berlin. ) b. John is yet to visit Paris or Berlin.

In addition, as with have yet to, be yet to fails to pass the Klima (1964) tests for negation, ruling out silent sententialnegation as a potential licensor.

(103)

a. John is yet to eat dinner, is he?34

b.

John is yet to eat dinner, isn’t he? c. John is yet to eat dinner, and neither is Mary. d. John is yet to eat dinner, and so is Mary. e. John is yet to eat his cake, not even a crumb. f.

*John is yet to eat dinner, I think/*I don’t think.

Given the options for potential complements to be in English, we have ruled out a silent progressive participle as thelicensor for yet in this construction. We have not yet ruled out a silent nominal complement, a PP complement, anadjectival complement, or a passive predicate. In the next section wewill argue that the silent licensor of yet in the be yet toconstruction must be a silent adjectival passive predicate.

5.1. Evidence for a SILENT adjectival passive predicate

When be occurs in English, it allows for a number of different types of complements. In addition to occurring withprogressive participles, as mentioned above, be also takes DP, PP, and AdjP complements, as shown in (104).

(104)

a. John is a student. Nominal complement/predicate b. John is in the garden. Prepositional Phrase c. John is brilliant. Adjectival complement/predicate d. John was kissed by Mary. Verbal Passive Participle e. John is unknown. Adjectival Passive Participle

In order to narrow down the list of possibilities at our disposal, we consider the behavior of be yet to in two differentsyntactic configurations -- sentences involving though-movement and specificational pseudoclefts.

We begin by considering though-movement, first pointed out by Ross (1967/1986), because the behavior of be yet tovs. have yet to interestingly differs in this construction. Though-movement involves the fronting of a predicate when asentence is introduced by though.

(105)

a. [Afraid of Bill] though he is, John is brave most of the time. b. [Proud of his children] though he is, John still wants them to do better. c. [Handsome] though I think he is, I still don’t like him.

As the following examples illustrate, only be yet to may occur in this construction. Have yet to may not.

(106)

a. [Yet to score] though he is/*has, Messi is still Barcelona’s best player. b. [Yet to visit Paris] though she is/*has, Mary knows exactly where the Louvre is. c. [Yet to pass his driving exam] though he is/*has, I still consider John a great driver.

to eat dinner, right?

S. Harves, N. Myler / Lingua 148 (2014) 213--239 235

These facts suggest that the silent licensor of yet indeed differs in the be yet to and have yet to constructions. As shown in(105) above, though-movement is allowed to occur with adjectival predicates. It does not appear to be allowed with perfectparticiples, however, as shown in (107), which lends independent evidence to our claim in Section 4 that the have in haveyet to is Perfect AUX have.

(107)

35 While sincluding thproductive(An anonymnot allow ajudgments36 We notsubstance37 Specifi

(i) a.b.

We do not

a.

*

****

ome spe prespart ofous re

ny movon thee that tis is yecationa

WhatWhat

consid

[Beaten the eggs thoroughly] though she has, Mary’s soufflé is still going to fall.35

b.

**[Finished dinner] though he has, John is still hungry.

A number of other XPs may also undergo though-movement, however. In other words, this construction is not limited toadjectival predicates. DPs and verbal and adjectival passive participles may also front (108). Therefore, this constructionalone is not enough to diagnose the syntactic category of the silent licensor of yet in sentences with be yet to.

(108)

a. [A brilliant linguist] though he is, John can’t speak any language other than English. b. [Brutally murdered] though he eventually was, Caesar was once a popular emperor. c. [Unbeaten] though they still are, Barcelona has a number of tough matches ahead.

This leads us to consider the behavior of sentences with be yet to in another construction, namely, specificationalpseudoclefts. As the sentences in (109) show, be yet to does not occur in such constructions.36

(109)

a. What Messi is is yet to score. b. What John is is yet to visit Paris. c.

**What Obama is is yet to close Guantanamo.

As pointedoutbyHiggins (1973), suchpseudocleftsoccurwithavarietyof syntactic categoriesas the ‘‘pivot.’’Thesentencesin (110) show that the types of pseudoclefts we are concerned with may occur with NP/DP, AP, and PP predicates.37

(110)

a. What John is is a murderer. b. What John is is ridiculous. c. What John is is in trouble.

If we consider the behavior of this construction with other types of predicative complements that be selects for, we see thatparticipial predicates and passives of all types are disallowed.

(111)

a. What John is is eating cake.

eakers discern no contrast between have and be wient authors, for whom the contrast mentioned in thetheir grammar have the reverse preference: disallowviewer for whom be yet to is a frozen form does haveement rules to affect subparts of the be yet to conssentences in (106) cannot reveal anything about thhis generalization applies only to specificational pst to be determined.l pseudoclefts in general can occur with VPs and C

he then did was cut his finger.proves that you are wrong is that they weren’t ev

er such examples here, since it seems unlikely tha

b.

What John was was brutally murdered. c. What the door is is opened. d. What Barcelona was was unbeaten. (Ok on a predicational reading.)

Since be yet to is also disallowed from occurring in this construction, as shown in (109), we have now narrowed down therange of possibilities for a silent licensor of yet even further. It must not be the case that a DP, a non-passive adjectivalpredicate, or a PP is the culprit. We are left with progressive and passive participles as possibilities.

th respect to though-movement (accepting both), there are also speakers,text is robust. As far as we know, no speakers who have be yet to as aing though-movement of the complement of be but allowing it with have.the reverse preference for cases like (106), but this is because s/he does

truction. Given this broader constraint on the reviewer’s grammar, his/here though-movability of be’s complement vis-à-vis have’s).eudoclefts; predicational pseudoclefts are apparently fine, e.g. What this

Ps as well, as shown by Higgins.

en there. (Higgins, 1973: 12)

t they could be involved in the be yet to construction.

S. Harves, N. Myler / Lingua 148 (2014) 213--239236

We argued above that a silent progressive predicate is not a contender, given that yet does not occur withprogressives, even in the presence of another NPI licensor such as negation (recall the examples in 100). This leaves uswith passive predicates. In order to decide between a silent verbal or adjectival passive participle, one needs to check thebehavior of be yet to against diagnostics used for distinguishing between verbal and adjectival participles.

Three diagnostics that are used for distinguishing between verbal/eventive and adjectival passives are (i) thepossibility of modification with an agentive by-phrase, (ii) the possibility of embedding under a raising predicate such asseems or remains, and (iii) the interpretation of temporal adverbials such as at 10:00.38 The sentences in (112--114) showthat by-phrases are disfavored in adjectival passives but perfect with verbal ones, while only adjectival passives can beembedded under raising verbs. The interpretation of the examples in (114) is such that an eventive reading is onlypossible in (114a). That is, a discovery was made at 10:00 am. In (114b), on the other hand, no event of ‘‘undiscovery’’took place at 10:00 am.

(112)

38 See W39 In dialeBarker (1940 This exBelgium w

a.

*

asow (cts of B92), deampleere bea

The treasure was discovered by some pirates.

b. *The treasure was undiscovered by some pirates.

(113)

a. The treasure remained discovered by some pirates. b.

*The treasure remained undiscovered.

(114)

a. The solution to the problem was discovered at 10:00 am. b. The solution to the problem remained undiscovered at 10:00 am.

Applying these tests to sentences with be yet to suggests that the silent passive predicate responsible for licensing yet inthis construction is indeed an adjectival predicate. By-phrases are disallowed in this construction (115a), but theconstruction allows for embedding under raising verbs such as seem and remain (115b). Moreover, modification by atemporal PP such as at 10:00/at the 60 minute mark does not yield an eventive interpretation (115c).

(115)

a. Barcelona is yet to win by Messi. b. Barcelona remains/seems yet to be beaten this season.39

c.

At 10:00/At the 60 minute mark, Messi was yet to score.

An additional piece of evidence in favor of a silent adjectival predicate comes from coordination. Although exceptions doexist (Bayer, 1996), a generalization about coordination is that only syntactic categories of the same type coordinate (thisis what Williams, 1978 calls the ‘‘Law of the Coordination of Likes’’). As the examples in (116) illustrate, coordination of beyet to clauses is possible with adjectival predicates, but not with verbal passives.

(116)

a. Messi is exhausted and yet to score. b. Last I checked, the car remained scratched and yet to be sold. c. Last I checked, Belgium was beaten by Croatia and yet to qualify for the World Cup.40 *

5.2. (Problematic) possibilities for a SILENT AP predicate

Having determined that the silent licensor of yet in this construction is a silent adjectival passive predicate, we are nowleft with deciding upon the actual predicate itself. We assume that the structure underlying this construction is the structureshown in (117). This structure is identical to the structure given for have yet to above, modulo the presence of be vs. haveand the presence of an AP predicate vs. a perfect verbal participle. We assume that yetMerges in the embedded clauseand moves to the Specifier of the adjectival passive predicate, licensing its silence.

1977) for one of the first comprehensive discussions for distinguishing between verbal and adjectival passives.ritish English we get plural agreement with the collective nouns, i.e., Barcelona remain/seem yet to be beaten this season. Seen Dikken (2001), and Smith (2012) for much discussion.is ungrammatical for British speakers who accept the sentence cited in the previous footnote, even if we alter the agreement toten. . ..

S. Harves, N. Myler / Lingua 148 (2014) 213--239 237

(117)

[TD$INLINE]

TP

T’

vP

AP

A’

TP

T’

vP

DP

John v+Tis

<is>

AdvP

yet A???-ED/EN

<John>

Tto

<John> visit Paris<yet>

This silent adjectival predicate must have negative implicative semantics, in order to license yet as an NPI. At this point,however, we have not yet been able to find an overt counterpart of such a predicate in English. There is one prima faciepossible candidate which can be dismissed. We discuss it in some detail in the following paragraphs, since futureresearchers might benefit from seeing that it is a dead end.

Kayne (2009, 2012) examines a seemingly related construction in English, namely, the modal be to construction, andposits a silent adjectival predicate EXPECTED orMEANT (or some subcomponent thereof) to account for the semanticsassociated with this construction. Examples of modal be to are given in (118), with Kayne’s (2012) proposal for thesentence in (118a) shown in (119).

(118)

a. You are to return home before midnight. b. Messi is to sign his contract before a new coach is named. c. You are to turn your assignments in by 5:00 pm on Tuesday.

(119)

You are MEANT FOR <you> to return home by midnight. (Kayne, 2012: 7)

One argument in favor of silentMEANT rather than EXPECTED is the fact that overtmeant does not license by-phrases,while overt expected does (120). The be to construction does not license by-phrases, either (121).

(120)

a. You were meant to return home by midnight by your entire family. (Kayne, 2012: 8) b. You were expected to return home by midnight by your entire family.

*

(121)

*You were to return by midnight by your entire family.

While Kayne’s (2012) proposal for silentMEANT seems promising for the modal be to construction, it does not carry overto the be yet to construction. While be yet to also disallows by-phrases, like overt meant, overt meant does not licenseNPIs, which discounts it as a covert counterpart to the construction considered here. Likewise, overt expected does notlicense NPIs, either.

(122)

a. John is meant to [give a speech yet/ever give a speech]. b.

**John is expected to [give a speech yet/ever give a speech].

An additional difference between be yet to and be to relates to finiteness. As noted by Pullum andWilson (1977), the be toconstruction appears to be limited to finite clauses, which is not true of be yet to, as shown in (123). We therefore appear tobe dealing with different syntactic constructions in each case. We leave it to future research to determine exactly what thesilent adjectival passive predicate corresponds to overtly in English, if anything.

(123)

a. John seems to be to do his homework. b.

*John seems to be yet to do his homework.

S. Harves, N. Myler / Lingua 148 (2014) 213--239238

6. Conclusions

We began this paper by pointing out a number of seemingly idiosyncratic properties of the have/be yet to construction:(i) the irregular placement andNPI behavior of yet, (ii) the apparent mismatch of the construction’s surface infinitival syntaxwith its semantic interpretation, and (iii) the apparent otherwise unattested free alternation between have and be (forEnglish speakers who allow both have yet to and be yet to). Our proposal that these constructions involve a negativeimplicative predicate which is rendered silent by the raising of yet into its specifier (with yet in turn licensed as anNPI by thenegative implicative predicate) yields an instant solution to problems (i) and (ii). Moreover our solution is motivated by theexistence of dependencies between movement and silence cross-linguistically, since there are many phenomena inwhich the silence of a given head is conditioned by the movement of a phrase into its specifier position. Hence, whatseemed like a bizarre and idiosyncratic property of English actually turns out to be nothing more than a highly restrictedreflection of a well-attested phenomenon permitted by UG. Furthermore, we were able to show that the category of thenegative implicative predicate involved in our solution to (i) and (ii) is different between the have yet to construction (whereit is a past participle or a derived nominal depending on dialect) and the be yet to construction (where it is an adjectivalpassive participle). The problem posed by (iii) is thus dissolved.

Acknowledgments

Many thanks to the audiences at PLC and the NYU Syntax/Semantics Brown Bag Forum for comments andsuggestions on the work presented here, with special thanks to Chris Collins, Paul Postal, and Beatrice Santorini. Wewould also like to thank the anonymous Lingua reviewers whose comments led to considerable improvements in thispaper.

References

Alrenga, P., Kennedy, C., 2014. No More shall we part: quantifiers in English comparatives. Nat. Lang. Semant. 22, 1--53.Atlas, J.D., 1993. The Importance of being ‘only’: testing the neo-Gricean versus neo-entailment paradigms. J. Semant. 10, 301--318.Atlas, J.D., 1996. ‘Only’ noun phrases, pseudo-negative generalized quantifiers, negative polarity items, and monotonicity. J. Semant. 13, 265--

332.Barker, C., 1992. Group rerms in English: representing groups as atoms. J. Semant. 9, 69--93.Bayer, S., 1996. The coordination of unlike categories. Language 72, 579--616.Collins, C., 2007. Home sweet home. In: Levinson, L., Savescu-Ciucivara, O. (Eds.), New York University Working Papers in Linguistics, vol. 1.

Available online at http://linguistics.as.nyu.edu/object/linguistics.grad.nyuwpl.Collins, C., Postal, P., 2014. Classical NEG Raising, An Essay on the Syntax of Negation. MIT Press, Cambridge, MA.Den Dikken, M., 2001. Pluringulars, pronouns and quirky agreement. Linguist. Rev. 18 (1), 19--41.Espinal, M. Teresa, 2000. On the semantic status of n-words in Spanish and Catalan. Lingua 110 (8), 557--580.Gajewski, J., 2005. Neg-Raising: Presupposition and Polarity (PhD dissertation). MIT.Gajewski, Jon., 2008. More on quantifiers in comparative clauses. In: Friedman, T., Ito, S. (Eds.), Proceedings of Semantics and Linguistic

Theory, 18. CLC Publications, Ithaca, NY, pp. 340--357.Gajewski, J., 2011. Licensing Strong NPIs. Nat. Lang. Semant. 19, 109--148.Giannakidou, A., 1997. The Landscape of Polarity Items (PhD dissertation). University of Groningen.Giannakidou, A., 1998. Polarity Sensitivity as (non) Veridical Dependency. John Benjamins, Amsterdam and Philadelphia.Giannakidou, A., 2006. Only, emotive factives, and the dual nature of polarity dependency. Language 82 (3), 575--603.Giannakidou, A., 2010. The dynamics of change in Dutch enig: from nonveridicality to strong negative polarity. Nat. Lang. Linguist.Theory 28,

861--875.Haegeman, L., Zanuttini, R., 1991. Negative heads and the neg-criterion. Linguist. Rev. 8, 233--251.Harves, S., Kayne, R.S., 2012. Having need and needing have. Linguist. Inq. 43 (1), 120--132.Herburger, E., 2001. The negative concord puzzle revisited. Nat. Lang. Semant. 9, 289--333.Higgins, F.R., 1973. The Pseudo-Cleft Construction in English (PhD dissertation). MIT.Hoeksema, J., 1986. Monotonie en superlatieven. In: Hoppenbrouwers, C., et al. (Eds.), Proeven van Taalwelenschap. TABU, Groningen,

pp. 38--49.Hoeksema, J., 2008. Distributieprofielen van negatief-polaire uitdrukkingen: een vergelijking van het Nederlands. Engels en Duits. Tabu 37,

111--195.Huddleston, R., 1976. Some theoretical issues in the description of the English verb. Lingua 40, 331--383.Iatridou, S., Zeijlstra, H., 2012. Negation, Polarity, and Deontic Modals. Lingbuzz/001431.Kadmon, N., Landman, F., 1993. Any. Linguist. Philos. 16 (4), 353--422.Kayne, R.S., 2004. Here and there. In: Leclère, C., Laporte, E., Piot, M., Silberztein, M. (Eds.), Lexique, Syntaxe, et Lexique-Grammaire/Syntax,

Lexis, and Lexicon Grammar: Papers in Honour of Maurice Gross. John Benjamins Publishing Co., Amsterdam and Philadelphia, pp. 253--275.

Kayne, R.S., 2005. Movement and Silence. Oxford University Press, New York, NY.Kayne, R.S., 2009. Why is syntax antisymmetric? In: Talk Presented at University College London.Kayne, R.S., 2010. Comparisons and Contrasts. Oxford University Press, New York, NY.

S. Harves, N. Myler / Lingua 148 (2014) 213--239 239

Kayne, R.S., 2012. Comparative syntax and English is to. Linguist. Anal. 39 (submitted for publication).Kelly, J.R., 2008. Yet as a negative perfect marker in English. Snippets 18, 8--9.Klima, E., 1964. Negation in English. In: Fodor, J., Katz, J. (Eds.), The Structure of Language. Prentice-Hall, Englewood Cliffs, NJ, pp. 246--323.Koopman, H., 1996. The Spec-head configuration. In: Garrett, E., Lee, F. (Eds.), Syntax at Sunset, UCLA Working Papers in Syntax and

Semantics, 1. pp. 37--64.Koopman, H., Szabolcsi, A., 2000. Verbal Complexes. MIT Press, Cambridge, MA.Ladusaw, W., 1979. Polarity as Inherent Scope Relations (PhD dissertation). University of Texas, Austin.Larson, R., den Dikken, M., Ludlow, P., 1997. Intensional Transitive Verbs and Abstract Clausal Complementation, Ms.Levinson, D., 2008. Licensing of Negative Polarity Particles Yet, Anymore, Either, and Neither: Combining DownwardMonotonicity and Assertivity

(PhD dissertation). Stanford University.Linebarger, M., 1980. The Grammar of Negative Polarity (PhD dissertation). MIT.Linebarger, M., 1987. Negative polarity and grammatical representation. Linguist. Philos. 10, 325--387.Marques, R., 2003. Licensing and interpretation of N-words in comparative clauses. In: Weisgerber, M. (Ed.), Proceedings of Sinn and Bedeutung

7. Universität Konstanz, pp. 199--212.Marusič, F., Zaucer, R., 2005. On phonologically null verbs. GO and beyond. In: Blaho, S., Vicente, L., Schoorlemmer, E. (Eds.), Proceedings of

ConSOLE XII, pp. 231--248.Marusič, F., Zaucer, R., 2006. On the intensional feel-like construction in Slovenian: a case of a phonologically null verb. Nat. Lang. Linguist.

Theory 24 (4), 1093--1159.McCawley, J., 1979. On identifying the remains of deceased clauses. In: McCawley, J.D. (Ed.), Adverbs, Vowels, and Other Objects of Wonder.

University of Chicago Press, Chicago, IL, pp. 84--95.Nchare, A.L., Arhonto, T., 2014. Licensing silent structure: the spatial prepositions of Shupamem. Nat. Lang. Linguist. Theory 32, 673--710.Pullum, G., Wilson, D., 1977. Autonomous syntax and the analysis of auxiliaries. Language 53 (4), 741--788.Rich, P., 2009.What is n’t doing there? French expletive negation in comparative clauses. In: Poster presentation given at the 84th annual meeting

of the Linguistic Society of America, Baltimore MD.Ross, J.R., 1967. Constraints on Variables in Syntax (PhD dissertation). MIT (Later published as Ross. (1986)).Ross, J.R., 1968. A proposed rule of tree-pruning. In: David, A., Reibel, Sanford, A., Schane, (Eds.), Modern Studies in English: Readings in

Transformational Grammar, 288--299. Prentice-Hall, Englewood Clis, NJ.Ross, J.R., 1976. To Have Have and Not to Have Have. In: Jazayery, M.A., et al. (Eds.), Linguistic and Literary Studies in Honor of Archibald A.

Hill. Peter de Ridder Press, Lisse, pp. 265--270.Ross, J.R., 1986. Infinite Syntax! Ablex Publishing Company, Norwood, NJ.Sailor, C., 2012. Tag Questions and the Typology of VP Ellipsis, Ms. UCLA.Sailor, C., to appear. VP ellipsis in tag questions: a typological approach. In: Proceedings of the 46th Annual Meeting of the Chicago Linguistics

Society.Schwarzschild, R., 2008. The semantics of comparatives and other degree constructions. Lang. Linguist. Compass 2, 308--331.Smith, Peter W., 2012. Collective (dis)agreement: on a 3/4 pattern of British English collective NPs. In: Talk given at ConSOLE XX, Lepizig, p.

2012.Thoms, G., 2012. Ellipsis, polarity and the variety of verb movement in Scandinavian. In: Talk given at the 27th Comparative Germanic Syntax

Workshop, Yale University.Uribe-Extebarría, M., 1994. Interface Licensing Conditions on Negative Polarity Items: A Theory of Polarity and Tense Interactions

(PhD dissertation). UCONN, Storrs.van der Wouden, T., 1997. Negative Contexts: Collocation, Polarity and Multiple Negation. Routledge, London.van Riemsdijk, H.C., 2002a. The unbearable lightness of going. the projection parameter as a pure parameter governing the distribution of elliptic

motion verbs in germanic. J. Comp. Germ. Linguist. 5, 143--196.van Riemsdijk, H.C., 2002b. The morphology of nothingness. In: Haraguchi, S. (Ed.), Proceedings of the LP, Conference at Meikai University,

Prague, Carl’s University.van Riemsdijk, H.C., 2012. The absent, the silent and the audible: Some thoughts on the morphology of silent verbs. In: Torrego, E. (Ed.), Of

Grammar, Words, and Verses. In Honor of Carlos Piera. John Benjamins, Amsterdam, pp. 19--39.Wasow, T., 1977. Transformations and the Lexicon. In: Peter, W., Culicover, T., Wasow, A., Akmajian, (Eds.), Formal Syntax.Williams, E., 1978. Across-the-board rule application. Linguist. Inq. 9 (1), 31--43.Wood, J., 2013. Parasitic participles in the syntax of verbal rather. Lingua 137, 59--87.Zagona, K., 2002. The Syntax of Spanish. Cambridge University Press, Cambridge.Zeijlstra, H.H., 2004. Sentential Negation and Negative Concord (PhD dissertation). University of Amsterdam, LOT publications, Utrecht.Zeijlstra, H.H., 2008. Negative Concord is Syntactic Agreement, Lingbuzz/000645.Zwarts, F., 1998. Three types of polarity. In: Hamm, F., Hinrichs, E.W. (Eds.), Plurality and Quantification. Kluwer, Dordrecht, pp. 177--238.