• Tidak ada hasil yang ditemukan

2016. Negation and Negative Dependencies

N/A
N/A
Protected

Academic year: 2018

Membagikan "2016. Negation and Negative Dependencies"

Copied!
22
0
0

Teks penuh

(1)

I N

A D V A

N

C

E

Negation and Negative

Dependencies

Hedde Zeijlstra

Seminar f ¨ur Englische Philologie, Georg-August-Universit¨at G ¨ottingen, D-37073 G ¨ottingen, Germany; email: hzeijls@uni-goettingen.de

Annu. Rev. Linguist. 2016. 2:17.1–17.22 TheAnnual Review of Linguisticsis online at linguist.annualreviews.org

This article’s doi:

10.1146/annurev-linguist-030514-125126 Copyright c2016 by Annual Reviews. All rights reserved

Keywords

negation, negative concord, Negative Polarity Items, negative indefinites, split scope

Abstract

Languages may vary greatly in the way they express negation. Most lan-guages exploit specifically designated negative markers, such as Englishnot. Many languages may also use negative indefinites (such as Englishnobodyor

nothing) to express negation. The behavior of these negative indefinites is

(2)

1. INTRODUCTION

A universal property of natural language is that every language is able to express negation; in other words, every language has some device at its disposal to reverse the truth value of the propositional content of a sentence. However, languages may differ significantly as to how they express this negation. For instance, languages may use different categorial elements in order to express negation; English, for example, apart from having the two negative markersnotand -n’tmay also use so-called negative indefinites (nobody,nothing,never,. . .), as is illustrated in examples 1–2:

(1a) John did not leave (1b) John didn’t leave

(2a) Nobody left

(2b) Mary never went there (2c) I’d never do that

Apart from this, languages can also differ in terms of the number of manifestations of negatively marked elements. In some languages, like English, a semantic negation is realized by only one single negatively marked element, but in other languages multiple negative elements can yield a single negative meaning. In such languages, the presence of one particular negative element often depends on another one. For instance, French has two negative markers, a preverbal negative markerne, which generally attaches to the finite verb, and a negative adverbpas. Whereaspas can render a sentence negative by itself,necannot do so, and always requires the presence of an additional negative element, such aspasorrien‘nobody’:

(3a) Jean (ne) mange *(pas)

Jean neg eats neg ‘Jean doesn’t eat’

(3b) Jean (ne) mange *(rien)

Jean neg eats n-thing ‘Jean doesn’t eat anything’

In this sense, the behavior ofneis similar to that of so-called Negative Polarity Items (NPIs). NPIs, such as Englishever, are elements whose distribution is limited to a number of contexts, which in some sense (discussed in Section 4, below) count as negative:

(4a) Suzanne didn’t ever leave (4b) Suzanne hardly ever left (4c) *Suzanne ever left

(3)

negative element. Italiannessuno(n-person), which in preverbal position can induce a semantic negation but in postverbal position cannot (and requires an additional negative element, like the negative markernon, to be present), is a good example:

(5a) Nessuno ha telefonato

n-person has called ‘Nobody called’

(5b) *Ha telefonato nessuno

Has called n-body Int.: ‘Nobody called’

(5c) Non ha telefonato nessuno

neg has called n-body ‘Nobody called’

The phenomenon where two elements that by themselves, in certain positions, can render a sentence negative together yield one negation is called Negative Concord (NC). Scholars studying the semantics of NC focus primarily on the meaning of sentences such as example 5c: Why does this sentence mean ‘Nobody called’ and not ‘It is not the case that nobody called’? But equally (or perhaps even more) important is the question as to why such n-words yield (in Italian, at least, in postverbal position) a negative dependency: Why must they be accompanied by a preverbal negative element, like the negative markernon?

In this article, I review what I consider the most important issues in the domain of negation: NC and negative polarity. In particular, I focus on the following questions: (a) Why are certain elements able to induce a semantic negation, whereas others only encode a negative dependency, and why can some elements do both? And (b) what determines the crosslinguistic variation that can be attested with respect to the behavior and distribution of negation and negative dependencies? This article is organized as follows. In Section 2, I introduce the various ways that languages can express negation, distinguish languages that exhibit NC from languages that lack it, and discuss some crucial differences between different types of NC languages. In Section 3, I further discuss the semantic behavior of negative indefinites in languages that lack NC. Then, in Section 4, I focus on NPIs, showing that their internal variation is much larger than generally thought, which may have strong repercussions for the ways that such elements must be analyzed.

2. NEGATION, NEGATIVE CONCORD, AND NEGATIVE QUANTIFIERS

As indicated in Section 1, a way to express sentential negation other than by means of negative markers is by using negative indefinites. This expression strategy exists in a number of languages. In Dutch, for instance, sentential negation can be expressed by including a negative indefinite, in either preverbal or postverbal position. The same applies to English, as the translations of the following examples show:

(6a) Niemand ziet hem

n-body sees him ‘Nobody sees him’

(6b) Hij ziet niemand

(4)

These facts strongly suggest that negative indefinites in these languages are (semantically active) negative quantifiers. Further evidence for an analysis of such negative indefinites in terms of nega-tive quantifiers comes from the fact that, when combined with a neganega-tive marker, a Dutch neganega-tive indefinite yields a so-called double-negation (DN) reading, where each morphosyntactically nega-tive element corresponds to a semantic negation. The same applies to sentences with two neganega-tive indefinites:

(7a) Jan belt niet niemand

Jan calls neg n-body

DN: ‘Jan doesn’t call nobody’=‘Jan calls somebody’

(7b) Niemand belt niemand

n-body calls n-body

DN: ‘Nobody calls nobody’=‘Everybody calls somebody’

At the same time, in many other languages, a clause-internal combination of two elements that can independently induce a semantic negation yields an NC reading, which contains only one semantic negation. This has been illustrated for Italian in example 5 in Section 1. Even though bothnon‘not’ andnessuno‘nobody’ receive a negative interpretation (shown in examples 8aand 8b), jointly they do not yield two semantic negations (example 8c):

(8a) Gianni non ha telefonato

Gianni neg has called ‘Gianni didn’t call’

(8b) Nessuno ha telefonato

n-body has called ‘Nobody called’

(8c) Non ha telefonato nessuno

neg has called n-body ‘Nobody called’

At first sight, the reading of example 8aseems to violate the principle of compositionality (Frege 1892, Janssen 1997). Why is it that such readings do not contain two semantic negations?

However, this essentially semantic question is not the only question to be addressed. Languages also differ crosslinguistically with respect to whether they exhibit NC or not. Moreover, there are two kinds of NC languages: strict NC and nonstrict NC languages (after Giannakidou 2000). In a nonstrict NC language, such as Italian, preverbal n-words cannot precede the negative marker, but postverbal n-words require an additional preverbal negative element:

(9a) Ieri nessuno (*non) ha telefonato

Yesterday n-body neg has called ‘Yesterday nobody called’

(9b) Ieri *(non) ha telefonato nessuno

Yesterday neg has called n-body ‘Yesterday nobody called’

(5)

(10a) Dnes nikdo *(ne-)vol´a Today n-body neg-calls ‘Today nobody calls’

(10b) Dnes *(ne-)vola nikdo Today neg-calls n-body ‘Today nobody calls’

Even though Czech and Italian differ in this respect (whether preverbal n-words can establish an NC relation with the additional negative marker), they behave similarly with respect to the obligatoriness of NC: When the negative marker can be there, it must be there. The negative markers in examples 9band 10 may not simply be removed. In other languages this is not the case. In West Flemish, for instance, NC is optional (Haegeman 1995, Haegeman & Zanuttini 1996):

(11) . . .da Val`ere niemand (nie) ken . . . that Val`ere n-body neg knows ‘. . . that Val`ere doesn’t know anybody’

Therefore, two questions have to be addressed:

How can the compositionality problem be explained?

How can the range of variation that languages crosslinguistically exhibit with respect to the distribution and behavior of NC be explained?

Most accounts of NC focus on the first question and either discuss only a subclass of NC languages or leave the second question aside. However, given that, as discussed below, several different approaches may be able to address the first question, each with its own successes and shortcomings, the second question might be a way to evaluate these approaches: To what extent are the existing approaches able to capture the crosslinguistic variation attested? It is impossible to discuss all analyses of NC in this review. Therefore, I describe only two exponents of what I take to be the major approaches to NC: the Negative Quantifier Approach and the NPI Approach. Below, I first briefly introduce these approaches and then evaluate how they can account for the attested crosslinguistic variation.

2.1. The Negative Quantifier Approach

The Negative Quantifier Approach takes the fact that n-words can induce a semantic negation to heart and takes all n-words to be negative quantifiers; some kind of absorption mechanism then accounts for why two n-words (or an n-word and a negative marker) are interpreted as if there were only a single negation (Zanuttini 1991, Haegeman 1995, Haegeman & Zanuttini 1996, De Swart & Sag 2002, Watanabe 2004). Crosslinguistic variation with respect to the semantic behavior and distribution of NC then either reduces to crosslinguistic variation with respect to the availability of such semantic absorption mechanisms or should be taken to be independent from what semantically underlies NC.

I focus here on the proposal by De Swart & Sag (2002), which is based on ideas proposed by Zanuttini (1991, 1997, 2001), Haegeman (1995) and Haegeman & Zanuttini (1996). Under this approach, NC is similar to the so-called pair-list readings of multiplewh-questions. Take, for instance, example 12:

(6)

The most salient reading of this sentence is not so much for which person is it the case that this person bought which thing, but rather what pairs<x,y> there are, such that personxbought thingy. An answer then could be:

(13) John bought apples and Mary pears

Thus, example 12 comes about with a reading where a singlewh-operator applies to pairs of variables. Such a reading is different from one that consists of a pair ofwh-operators, each binding a single variable.

For De Swart & Sag (2002), the exact same mechanism applies to sentences with multiple negative quantifiers. For instance, the French sentence in example 14aalso has two readings: one where every negative quantifier binds a single variable and one where a single quantifier binds a pair of variables.

(14a) Personne (n’)aime personne n-body neg loves n-body

(14b) DN: No one is such that they love no one ¬∃x¬∃ylove(x,y)

‘Nobody loves nobody’

(14c) NC: No pair of people is such that one loves the other ¬∃<x,y>love(x,y)

‘No one loves anyone’

Reading 14bamounts to a DN reading, whereas reading 14cis the NC reading. Thus, if the mech-anism responsible for the creation of the pair-list readings in multiplewh-questions (standardly referred to as quantifier resumption after May 1985) also applies to multiple negative indefinites, then it is predicted that every sentence containing two n-words must have two readings as well: a DN reading and an NC reading. As De Swart and Sag show, this ambiguity for sentences like example 14ais indeed attested among speakers of French.

The main advantage of this proposal is that the availability of NC readings does not have to be independently accounted for but rather comes for free, once it is assumed that quantifier resump-tion applies to all kinds of quantifiers, including negative quantifiers. At the same time, the strength of this proposal is also its weakness. The prediction that every sentence containing two or more negative quantifiers is always ambiguous between an NC reading and a DN reading is too strong. Although French reflects this kind of ambiguity, similar constructions in most other languages are clearly unambiguous. Therefore, a question arises as to why languages display crosslinguistic variation in this respect. For De Swart & Sag (2002, p. 390), this is “really a question about the relation between language system and language use.” In principle, both interpretations are always available, and for these authors language usage determines which reading surfaces in the end.

However, the approach does not make any claims about the grammatical requirements behind NC. For instance, why is NC obligatory in the large majority of NC languages? Why couldn’t a sole n-word, for instance, in preverbal position render a sentence negative by itself? How does the distinction between strict and nonstrict NC languages come about?

(7)

in strict NC languages. Additional constraints might require that negation at least should appear once in preverbal position (NegFirst, dating back to Horn 1989). The interplay between the above-mentioned constraints and NegFirst could then yield nonstrict NC languages. Irrespective of the exact choice of constraints and their internal rankings, crucial for De Swart is that the syntax and semantics behind NC act on different planes.

A question that remains open under this approach, though, is why n-words may sometimes establish NC relations with elements that do not appear to be semantically negative. This is, for instance, the case in complement clauses of verbs expressing doubt or fear, prepositions such as

without, or comparatives, as the following examples from another nonstrict NC language, Spanish

(taken from Herburger 2001), illustrate:

(15a) Dudo que vayan a encontar nada

Doubt.1SG that will.3PL.SBJ that PRT find n-thing ‘I doubt they will find anything’

(15b) Juan ha llegado m´as tarde que nunca Juan has arrived more late than ever ‘Juan has arrived later than ever’

For De Swart and Sag, all these elements must be analyzed as containing a true negation, something that they explicitly show for Frenchsans‘without.’ Ifwithoutunderlyingly means something like

not with, then thisnotmay undergo resumption again. Still, the question remains as to why some

elements that are not antiadditive may participate in NC relations and others cannot, and why languages differ with respect to this matter. Slavic languages, for instance, are more restrictive in this sense.

2.2. The Negative Polarity Item Approach

The existence of sentences such as examples 15aand 15b, however, suggests that n-words are different from plain negative quantifiers. In fact, these contexts (expressions of doubt or compar-atives) in which the n-words appear in examples 15aand 15bare all so-called downward entailing (DE) contexts: contexts in which NPIs are typically licensed (also see Section 4.1). In this sense, n-words are similar to NPIs, such as Englishanyorever, that may appear in exactly those contexts where an n-word does not seem to bring in a semantic negation of its own. This similarity is shown in examples 16ac(the first two of which are translations of examples 15aand 15b), whereanyand everare fine.

(16a) I doubt they will find anything (16b) Juan has arrived later than ever (16c) No one loves anyone

It is tempting to solve the problem regarding examples 15aand 15bby assuming that n-words form a particular class of NPIs. The problem for such an analysis, though, is that n-words that are outside NPI-licensing contexts may not exhibit any NPI-like behavior. In such contexts, they behave instead like negative quantifiers. This is, for instance, the case with fragment answers, illustrated in examples 17acfor Spanish (taken from Herburger 2001):

(8)

(17b) Q: A qui´en viste? A: A nadie!/*a un alma

To whom saw-2SG? To n-body/to a soul (NPI)

‘Who did you see?’ ‘Nobody/*anybody’

(17c) Me caso contigo o con nadie

Me marry.1S with.you or with n-body ‘I marry you or nobody’

This problem has been addressed by Ladusaw (1992), who concludes that the hypothesis that all n-words are NPIs can be entertained only as long as some mechanism is available that can license these NPIs in absence of an overt negation. In other words, n-words that induce a semantic negation are nothing but NPIs, like the English any-terms, that are licensed by some covert negative operator in absence of a covert negation.

Thus, the two following questions arise: (a) What is the exact mechanism that ensures that only n-words in non-DE contexts may trigger the presence of an abstract negative operator? And (b) why do n-words have this self-licensing property, whereas all other known NPIs do not? Only if these questions can be satisfactorily answered can Ladusaw’s hypothesis be successfully pursued. Ladusaw (1992) does not provide a full-fledged analysis of how this self-licensing mechanism can be implemented, but others have followed up on this idea. One such account, proposed by Giannakidou (2000), argues that, at least in strict NC languages, like Greek, the difference lies in the quantificational status of n-words (universal quantifiers outscoping their licenser) as opposed to plain NPIs (existential quantifiers or indefinites scoping under their licenser). Another proposal, put forward by Zeijlstra (2004, 2008) and pursued by Haegeman & Lohndal (2010), states that n-words are not plain NPIs but rather indefinites that, in contrast to NPIs, must stand in a syntactic Agree relation with a negative operator.

Zeijlstra proposes that n-words are plain indefinites that carry some uninterpretable negative feature [uNEG] that is to be checked against a higher, semantically negative element that carries an interpretable formal negative feature [iNEG]. NC, in his view, is then nothing but an instance of syntactic agreement. Because the Agree system, after Chomsky (1995, 2001), that Zeijlstra adopts allows for Multiple Agree (Ura 1996, Hiraiwa 2001), multiple n-words can be checked against a single negative operator.

Zeijlstra further assumes that, just as in other cases of syntactic agreement, the element carrying the interpretable feature may be phonologically null. A good parallel is null subjects. Whereas in some languages finite verbs do not agree with their subjects and every subject must be pronounced, other languages allow their finite verbs to agree with their subjects and allow the actual subject to be a phonologically null element. Null subjecthood and NC, for Zeijlstra, are two sides of the same coin, as illustrated in the two sentences below for Italian. In examples 18aand 18b, some element (the n-word and the finite verb, respectively) is equipped with a feature that requires some other, possibly null element, to check it (the abstract negative operator and abstractpro, respectively):

(18a) OpNEG[iNEG]nessuno[uNEG]telefona (18b) Pro[i3SG]telefona[u3SG]

(9)

NC results from a similar treatment of negative markers. Given that in strict NC languages n-words may precede the negative marker, negative markers in those languages should be taken to carry a [uNEG] feature as well; by contrast, the semantically active negative operator is always phonologically null. In nonstrict NC languages, the negative marker may appear only to the left of n-words (in an NC construction), and it always corresponds to the locus of the interpretation of semantic negation. Therefore, in those languages the negative marker must carry [iNEG] as well. The relevant structures for preverbal and postverbal n-words for Italian and Czech are as follows:

(19a) OpNEG[iNEG]nessuno[uNEG]telefona (19b) Non[iNEG]telefona nessuno[uNEG]

(20a) OpNEG[iNEG]nikdo[uNEG]nevol´a[uNEG] (20b) OpNEG[iNEG]nevol´a[uNEG]nikdo[uNEG]

2.3. Two Predictions

The Negative Quantifier Approach and the NPI Approach appear to differ in two respects. First, they differ with respect to the semantic status of n-words: Are they semantically negative or not? Second, and slightly less often observed, the two approaches differ with respect to the question whether the syntax and semantics of NC are distinct—in other words, does the syntactic behavior of n-words (such as their optional or obligatory licensing requirement by the negative marker) follow from independent mechanisms, or does the semantic behavior follow as a result of the underlying syntactic mechanism? At the same time, both approaches, as represented by the two analyses discussed above, make two clear predictions that are similar, and that are generally also made by other analyses of NC: (a) In DN languages, every negative indefinite is a negative quantifier; and (b) even if n-words are NPIs, they should form a different class of NPIs than plain NPIs do, namely NPIs that cannot be subject to any self-licensing mechanism. In Sections 3 and 4, I discuss the validity of these two predictions by introducing some challenges for them.

3. NEGATIVE QUANTIFIERS AND SPLIT-SCOPE EFFECTS

The first prediction that the previous section ended with was that in DN languages every negative indefinite is a negative quantifier. For the Negative Quantifier Approach, this is because in every language a negative indefinite is a negative quantifier; for the NPI approach, this is to distinguish NPIs from n-words and to account for the fact that in DN languages every negatively marked element induces a semantic negation.

In this section, I introduce a different phenomenon that casts doubt on the view that negative indefinites in DN languages are negative quantifiers. Take examples 21 and 22, from German and Dutch, respectively (discussed in Rullmann 1995 and Penka 2007, 2010, among others):

Du musst keine Krawatte anziehen

You must no tie wear

(10)

Ze mogen geen verpleegkundige ontslaan

They may no nurse fire

(22a) ‘There is no nurse who they are allowed to fire’ ¬>>may (22b) ‘They are allowed not to fire a nurse’ may>¬>∃ (22c) ‘They are not allowed to fire a nurse’ ¬>may>∃

In both examples, three readings are available: one reading where the entire negative indefinite takes wide scope with respect to the modal verb (¬>∃>must/may), one (slightly less available) reading where the entire negative indefinite takes narrow scope (must/may>¬>∃), and a so-called split-scope reading where the negative part of the negative indefinite outscopes the modal but where the indefinite part still scopes under the modal (¬>must/may>∃). Whereas the first and second readings could simply reduce to the wide or narrow scope of the negative quantifier, the third reading cannot do so.

Independent evidence for the existence of these readings, showing that they are not simply entailed readings, can be found in example 23. Germanbrauchen, being an NPI, must scope under negation. At the same time, when appearing in an existential construction with expletivees‘there,’ an indefinite embedded under a modal can only take narrow scope. Consequently, both the wide-scope and narrow-wide-scope readings of the modal in examples 23band 23c, respectively, are ruled out. Still, the sentence is grammatical with a split-scope reading. This finding provides independent evidence for the existence of separate split-scope readings (Penka 2010).

Es braucht kein Arzt anwesend zu sein

There needs no physician present to be

(23a) ‘It is not required that there be a physician present’ ¬>need>∃ (23b) *‘There is no physician who is required to be present’ ¬>∃>need (23c) *‘It is required that there be no physician present’ need>¬>∃

Split-scope effects do not appear only in combinations with modals and negative indefinites. Object-intensional verbs also invoke split-scope readings (although the narrow-scope reading is independently ruled out; see Zimmermann 1993), as is shown in examples 24 and 25 for German and Dutch, respectively:

Perikles schuldet Socrates kein Pferd

Perikles owes Socrates no horse

(24a) ‘Perikles is not obliged to give Socrates a horse’ ¬>owe>∃ (24b) ‘There is no horse that Perikles is obliged to give to Socrates’ ¬>∃>owe (24c) *‘Perikles is obliged not to give Socrates a horse’ owe>¬>∃

Hans zoekt geen eenhoorn

Hans seeks no unicorn

(11)

Finally, as De Swart (2000) has shown, scope splitting is not restricted to negative indefinites but may apply to all kinds of DE DPs, such asfew, as shown below (note that the third reading is unavailable because Dutchhoeven‘need’ is an NPI):

Ze hoeven weinig verpleegkundigen te ontslaan

They need few nurses to fire

(26a) ‘They are required to fire few nurses’ ¬>need>∃ (26b) ‘There are few nurses who they need to fire’ ¬>>need (26c) *‘They need to fire few nurses’ need>¬>∃

Hence, a paradox arises. Given the discussion in Section 2, the conclusion that negative indefinites in DN languages are negative quantifiers seems unavoidable (both approaches converge on this conclusion). But such negative indefinites do not manifest the exact behavior one might expect if they are indeed negative quantifiers. Either one is forced to assume that these negative indefinites are indeed negative quantifiers but that their deviant behavior (i.e., their ability to give rise to split-scope readings) follows from something else, or one has to give up the idea that they are negative quantifiers. Under the latter perspective, either there is a third type of negative element (alongside negative quantifiers and n-words), or negative indefinites in DN languages are actually n-words that independently cannot give rise to NC readings. This final possibility would mean that it is not NC languages but rather DN languages that are the odd man out. I discuss the three positions (namely that negative indefinites in DN languages are negative quantifiers, n-words, or something else) in more detail below.

3.1. The Negative Quantifier Approach

Geurts (1999) argues that negative indefinites, at least in DN languages, are negative quantifiers, though these quantifiers do not necessarily quantify over individuals but may also quantify over kinds in the sense of Carlson (1977). This is illustrated in example 27 from German (taken from Geurts 1999):

(27) Ich suche keine Putzfrau

I seek no cleaning lady ‘I do not look for a cleaning lady’

For Geurts, both the narrow- and wide-scope readings of example 27 involve quantification over individuals. The wide-scope reading, for instance, says that there are no cleaning ladies such that I am looking for them. Geurts argues, however, that the sentence may also have a reading where

cleaning ladyrefers to a kind. In this case, the sentence would mean that there is no kind-element

cleaning ladysuch that I seek this kind or, in other words, that ‘I am not a cleaning-lady-seeker.’

That reading is equivalent to the split-scope reading: If I am not a cleaning-lady-seeker, that means that it is not the case that I am seeking a cleaning lady.

If this line of reasoning is correct, no additional assumptions have to be made in order to account for the existence of split-scope readings, and the null hypothesis that all negative indefinites in DN languages should be taken to be negative quantifiers can be maintained. However, several scholars, most notably Penka (2007, 2010), have pointed out several problems for this approach.

(12)

where the negative indefinite contains a numeral (‘two cars’), and so cannot be taken to denote a kind in the first place. However, other analyses under this approach may overcome this problem; for example, Abels & Marti (2010) argue that a treatment of negative indefinites in DN languages as quantifiers over so-called choice functions (instead of kinds) can yield the split-scope readings in examples 28 and 29:

(28) Ich suche keinen Student, der gestern in Arnims Vorlesung war

I seek no student who yesterday in Arnim’s lecture was ‘I do not look for a student who attended Arnim’s lecture yesterday’

(29) Wir m ¨ussen keine zwei Autos haben

We must no two cars have

‘We don’t need to have two cars’

A potentially more problematic argument is that Geurts’s approach ultimately has to allude to some sort of lexical ambiguity in order to enable negative quantifiers to quantify over both individuals and kinds. This requirement renders the Negative Quantifier Approach weaker, as it effectively introduces a third type of negative indefinite (one that quantifies over kinds), even though its original rationale was to reduce all types of negative indefinites to one. This argument does not only play a role in Geurts’s particular analysis of split-scope readings. De Swart (2000), who proposes a modification of Geurts’s account in which split-scope readings are not the result of quantification over kinds but over properties, postulates two different lexical entries of negative indefinites. However, Abels & Marti (2010) postulate only one lexical entry for negative indefinites (and other quantifiers) as quantifiers over choice functions.

3.2. The N-Word Approach

To circumvent the problems that the Negative Quantifier Approach faces in accounting for the split-scope readings of negative indefinites in DN languages, while not increasing the taxonomy of negative indefinites, Penka (2007, 2010) argues that all negative indefinites, in both NC and DN languages, are semantically nonnegative. Penka draws a parallel between NC and split-scope readings, and argues that the same process underlies both phenomena. Adopting Zeijlstra’s ver-sion of the NPI Approach to NC (where n-words are taken to be semantically nonnegative, and carry an uninterpretable negative feature [uNEG] that needs to be checked against a potentially phonologically abstract negative operator), Penka argues that in DN languages, like Dutch and German, the same process is going on, the only difference being that multiple Agree between one negation operator and multiple n-words is not allowed. In these languages, a negative operator can check off only one negatively marked indefinite. Thus, every negative indefinite is semantically nonnegative and carries a [uNEG] feature, and it needs to have its feature checked against an abstract negative operatorOpNEG[iNEG]. If two negative indefinites show up in the sentence, each negative indefinite must be licensed by a separateOpNEG[iNEG].

(13)

(30) . . .dass du OpNEGkeine Krawatte anziehen musst

. . . that you no tie wear must

(31) dass du [ OpNEG[iNEG][IP[VP[DPkeine[uNEG]Krawatte] anziehen] musst]] ¬>must>∃

(32) dass du [IP[DPOpNEG[iNEG]keine[uNEG]Krawatte] [IP[VPanziehen] musst]]¬>∃>must

(33) dass du [IP[VP[DPOpNEG[iNEG]keine[uNEG]Krawatte] anziehen] musst] must>¬>∃

The advantage of Penka’s account is that it adopts an independently motivated mechanism to account for split-scope readings, and Penka does not have to make any particular claims about the behavior of negative indefinites in DN languages. However, this analysis also faces some problems. First, Penka (2010) takes every language to exhibit formal negative features, along the lines of Zeijlstra. However, Zeijlstra’s analysis is not directly compatible with this extension. In his approach, the difference between NC and DN languages is that negative indefinites in DN languages have no formal negative features at their disposal (Zeijlstra 2004, 2008); for Zeijlstra, formal negative features can be acquired only in NC languages. Moreover, it is not clear how the crosslinguistic variation with respect to the possibility of (not) being subject to multiple Agree can be lexically encoded.

Another problem for Penka may be that her analysis does not straightforwardly extend to VO languages, even though VO languages, like English, may also exhibit split-scope effects, albeit to a lesser extent. Iatridou & Sichel (2011) report various examples, such as example 34, which for many speakers of English yields a split-scope reading. For these authors, example 34 can mean that it is not the case that you must do some homework today:

(34) You have to do no homework today

Iatridou and Sichel provide an independent account of why not every speaker accepts the split-scope readings of example 34, suggesting that English is in principle a language that exhibits split-scope effects too, and that split-scope readings are not restricted to OV languages. However, there is no position PF-adjacent tono homework, from where the abstract operator can outscope the modal. Thus, without further modification, the English facts would be problematic for Penka’s approach.

3.3. The Decomposition Approach

An alternative way of accounting for split-scope effects is lexical decomposition, as has been proposed by Jacobs (1980) and Rullmann (1995). This approach states that a negative indefinite, such as Germankeinor Dutchgeen, underlyingly consists of a plain negative marker and a plain indefinite. Only under PF adjacency can the two be jointly spelled out as a single morphological word, following a rule like the following (for German and for Dutch).

(35) nicht+einkein

nicht+eengeen

(14)

(36) Ze mogen geen verpleegkundige ontslaan

They may no nurse fire

(37) [Zij mogeni[niet een verpleegster [ontslaan ti]]] ¬>∃>may

(38) [Zij mogen [niet een verpleegster ontslaan] ti]] may>¬>∃

(39) [Zij mogeniniet [een verpleegster [ontslaan ti]]] ¬>may>∃

This Lexical Decomposition Approach, in its various forms, circumvents the semantic problems of split-scope effects by taking the spelling out of negative indefinites to be a PF phenomenon.

A problem for such analyses is that they, again, apply only to OV languages, namely languages where a VP-external negative marker may appear to the direct left of a VP-internal object. In languages like English, however, the surface position of a negative indefinite is always lower than the position of the negative marker. Take, for instance, example 34, repeated here:

(40) You have to do no homework today

Even though this sentence may have a split-scope reading, there is no way that this split-scope reading could have been derived from PF adjacency of the negative marker and the indefinite: The position of the negative marker must be structurally higher than the modal, whereas the modal must always appear to the left of its complement.

To solve this problem, Zeijlstra (2011) proposes a mixture of the Negative Quantifier Approach and the Lexical Decomposition Approach. In short, he posits that the negation and the indefinite merge into a negative quantifier that can further undergo quantifier raising. This process leads to two copies of the [NEG]+[INDEF] treelet. Consequently, one of these copies may be spelled out at PF as a negative indefinite along the lines of Jacobs (1980), Rullmann (1995), and Penka, whereas at the level of LF, nothing forbids the negation from being interpreted in the higher copy and the indefinite in the lower copy. The underlying LF of example 40 is then as follows, whereas the lower copy gets jointly spelled out asno homeworkat PF:

(41) You[[[NEG] [INDEF]] homework]have to do[[[NEG] [INDEF]]homework]

On one hand, under Zeijlstra’s (2011) account, the Lexical Decomposition Approach can be extended to all languages, rather than only VO languages. On the other hand, this account crucially relies on the availability of negative quantifiers to undergo quantifier raising. This assumption is not uncontroversial. von Fintel & Iatridou (2003) have shown that negative indefinites cannot scope over other quantifiers, as the following example demonstrates for English:

(42) Everybody touched no dessert ∀>¬∃;∗¬∃>∀

Thus, Zeijlstra’s account can be successful only if some other mechanism is responsible for ruling out the possibility of the negative indefinite to outscope the universal quantifier in sentences such as example 42.

(15)

4. NEGATIVE POLARITY

The first prediction of all approaches for NC, namely that negative indefinites in DN languages are negative quantifiers, is not without problems. Let us now look at the second prediction, namely that n-words should be different from plain NPIs (i.e., NPIs that, unlike n-words, cannot induce any negation of their own). First, let us look at some general properties of plain NPIs.

The best-known examples of plain NPIs are formed by the Englishany-terms, although many more exist (e.g., Englishyet,need,either, orlift a finger):

(43a) We *(didn’t) readanybooks (43b) I have*(n’t) been thereyet (43c) Ineed*(n’t) do that

(43d) I *(didn’t) read the book, and John *(didn’t)either (43e) Nobody/*somebodylifted a finger

NPI-hood, however, is not restricted to English. Most if not all languages seem to display NPIs (see Haspelmath 1997 for a nonexhaustive list), and many languages exhibit a typology of NPIs, often at least as rich as that of English.

NPIs have received wide attention from scholars of syntax, semantics, and pragmatics, and they have constituted a fruitful and popular research area over the past 30 years. As Ladusaw (1996) points out in a seminal overview article, the study of the behavior of NPIs has been dominated by several research questions, of which I discuss two: the licenser question and the licensee question. The licenser question aims to determine what counts as a proper NPI licensing context. The licensee question seeks to learn why certain elements are allowed to occur only in particular contexts and what distinguishes them from polarity-insensitive elements.

4.1. The Licenser Question

As examples 43aeshow, NPIs are licensed in negative contexts, but NPIs are not restricted to contexts that are negative. Examples 44adshow that NPIs can also be licensed in various other contexts, such as restrictors of universal quantifiers, questions, and expressions likeat mostorhardly:

(44a) Every student who knowsanythingabout linguistics will join the event (44b) Do you wantanycookies?

(44c) At most three students didanyhomework (44d) Mary hardly likesanycookies

What, then, is the general property responsible for licensing NPIs? The first proposal, still one of the most important and influential, that tries to reduce all NPI licensing contexts to a single semantic property (briefly introduced in Section 1.2) is Ladusaw’s (1979) proposal, based on Fauconnier (1975), that all NPI licensers are DE. DE is defined as follows (adapted from van der Wouden 1994):

(45) δis downward entailing iff∀X∀Y(X⊆Y)→([[δ]](Y)⊆[[δ]](X)).

(16)

(46a) Mary is wearing a red shirt → Mary is wearing a shirt (46b) Mary is wearing a shirt Mary is wearing a red shirt

In DE contexts, the entailment relations are reversed. This is shown for the negative contexts in examples 47aand 47band for some nonnegative contexts in examples 48aand 48b, where the only valid inferences are now from a set to its subsets.

(47a) Nobody is wearing a red shirt Nobody is wearing a shirt Nobody is wearing a shirt → Nobody is wearing a red shirt (47b) John is not wearing a red shirt John is not wearing a shirt

John is not wearing a shirt → John is not wearing a red shirt

(48a) Every student went to bed → Every linguistics student went to bed (48b) At most three students left → At most three students left early

Although this proposal is considered a milestone in the study of NPIs, it faces several serious problems. It turns out that not every NPI can be licensed in every DE context, and other NPIs can be licensed outside DE contexts.

With respect to the first problem, some NPIs are subject to different licensing conditions than others. For instance, whereas the English any-terms seem to be acceptable in all DE contexts, the Dutch counterpart toany,ook maar, is ruled out in DE contexts likeniet iedereen‘not everybody’:

(49a) Nobody/Not everybody ateanything

(49b) Niemand/*Niet iedereen heeft ook maar iets gegeten Nobody/Not everybody has PRT PRT something eaten ‘Nobody/Not everybody ate anything’

van der Wouden (1994), elaborating on Zwarts (1995), argues that DE should be considered one layer of a negative hierarchy, where the true negation (not) forms the highest layer, followed by so-called antiadditive elements (nobody,nothing,no), followed by the next layer, DE-ness. Formally, a functionf is antiadditive ifff(A∨B)⇔(f(A)∧f(B)). For example,no studentis antiadditive,

asno students drinks or smokesis truth-conditionally equivalent tono student drinks and no student

smokes.Not everyis not antiadditive, asnot everybody drinks and not everybody smokesdoes not entail

thatnot everybody drinks or smokes. In a situation where only John doesn’t drink and only Mary

doesn’t smoke, it is indeed the case that not everybody drinks and not everybody smokes, but it is true that everybody drinks or smokes.

NPIs, then, differ with respect to which layer of negativity is qualified to license them. English anyis licensed in DE contexts (and thus in all negative contexts); others are licensed only in antiadditive contexts (such as Dutchook maar); and some NPIs can be licensed only by the sentential negative marker. An example of the last category is the Dutch idiomvoor de poes, as inzij is(

niet)

voor de poes(‘she is not for the cat,’ meaning ‘she’s pretty tough’; see van der Wouden 1994). NPIs

that can be licensed by all DE contexts are referred to as weak NPIs; NPIs that can be licensed only by antiadditive contexts are known as strong NPIs; NPIs that can co-occur only with the negative marker, like the Dutch idiom, are called superstrong NPIs.

(17)

(2008), among others, show that Dutch NPIhoevencannot occur in the first argument of a universal quantifier, which is DE but not antiadditive, even though it can occur in other nonantiadditive DE contexts such asweinig‘few’:

(50a) *Iedereen die hoeft te vertrekken, moet nu opstaan Everybody who needs to leave must now get up ‘Everybody who needs to leave, must get up now’

(50b) Weinig mensen hoeven te vertrekken

Few people need to leave

‘Few people need to leave’

For this reason, Lin et al. (2015) argue that NPIs likehoevenform another class of NPIs, in between strong and weak NPIs.

Finally, Giannakidou (1997, 1999, 2000, 2011), among others, shows that whereas DE-ness is not always a sufficient condition for NPI licensing, it is not always a necessary condition for it either. For instance, Chineseshenme‘any’ can be licensed under modals or futurate contexts (Lin et al. 2014):

(51a) Yuehan keneng mai le shenme shu

John maybe buy PRF any books

‘Maybe John has bought some book(s)’

(51b) Yuehan mingtian yao qu mai shenme shu

John tomorrow will go buy any book

‘John is going to buy some book tomorrow’

Apparently, DE-ness does not seem to be the weakest type of NPI-licensing context. There-fore, Giannakidou proposes, following Zwarts (1995), to further extend the hierarchy of negative contexts by another layer of negativity, namely nonveridicality:

(52) A propositional operatorFis nonveridical ifFpdoes not entail or presuppose thatp is true in some individual’s epistemic model (after Giannakidou 1997, 1999, 2010).

To clarify this definition,maybein sentence 53ais a nonveridical operator, whereasunfortunately in sentence 53bis veridical because a speaker uttering sentence 53adoes not take the sentence

John is illto be necessarily true, whereas a speaker uttering sentence 53bdoes do so.

(53a) Maybe John is ill (53b) Unfortunately John is ill

NPIs likeshenmeappear to be licensed by all nonveridical contexts, thus constituting another type of NPIs, called superweak NPIs.

(18)

4.2. The Licensee Question

Perhaps even more important than what licenses an NPI is what property an NPI has such that it can occur only in a particular type of context. This question has dominated the study of NPI licensing over the past 20 years.

Two types of approaches have been formulated to address this question. For some scholars, NPI-hood reduces to some semantic and/or pragmatic requirement that NPIs can be felicitously uttered only in negative contexts of some sort (DE, antiadditive, or nonveridical). For others, the answer lies in syntax or the lexicon; in other words, NPIs are accompanied by some syntactic or lexical requirement that forces them appear in negative environments only.

4.2.1. Semantic/pragmatic approaches. The first major contribution toward the first approach is the widening+strengthening account by Kadmon & Landman (1993). These authors’ account consists of two steps. First, they propose that NPI indefinites, such as Englishany-terms, seman-tically differ from plain indefinites in that NPIs are domain wideners. Such domain-widening indefinites extend the domain of reference beyond the contextual restrictions that plain indefi-nites are subject to. Take sentences 54aand 54b, which contain Kadmon and Landman’s original examples:

(54a) I don’t have potatoes (54b) I don’t have any potatoes

Whereas example 54aentails that in a particular domain the speaker doesn’t have potatoes, example 54bsuggests that the speaker doesn’t have even a single old potato in some corner of the kitchen. The second step in Kadmon and Landman’s line of reasoning is that sentences containing NPIs likeanymust be stronger than sentences containing a plain indefinite. Example 54bis stronger than example 54a: The set of situations where the former is true is a clear subset of the set of situations where the latter is true, so the latter entails the former. The strengthening requirement is thus met. However, the fact that example 54bis stronger than example 54ais due to the presence of the negative marker: Given that negation is DE, removal of the negation in examples 54aand 54b would reverse the entailment relation. Therefore, without the presence of the negation, a sentence like example 54bwould actually be weaker than the sentence withoutany. Uttering sentence 54b without the negation would thus violate the pragmatic strengthening condition. For Kadmon & Landman (1993), this is exactly what rules out sentences containing unlicensed NPIs.

The idea that NPIs come along with widening and strengthening effects, which mean that they can be felicitously uttered only in DE contexts, has been expressed in various ways. Krifka (1991, 1995), for instance, argues that the strengthening condition follows as an implicature, as sentences with a weak reading generally come along with an implicature that the stronger reading is ruled out. In this respect, he focuses on elements denoting minimal amounts and explains that those elements in particular are prone to becoming NPIs. To ensure that NPIs are always subject to a strengthening requirement, Chierchia (2006, 2013) proposes that NPIs are domain wideners that carry an additional, syntactic feature that requires NPIs to appear under the direct scope of an abstract strengthening operator (or exhaustification operator, in his terms), which states that any stronger scalar alternatives of the sentence containing the NPI are false. It then follows that such strengthened or exhaustified domain wideners yield a semantic contradiction unless they appear in DE contexts. If this contradiction arises, then NPIs are ruled out in all non-DE contexts.

(19)

infelicitousness and semantic contradictions are generally not judged as being ungrammatical, as shown in examples 55aand 55bbelow (taken from Giannakidou 2011):

(55a) The king of France is my brother

(55b) John is born in New York, and he is not born in New York

However, the judgments on unlicensed NPIs are much stronger: Speakers generally feel them to be ungrammatical. Chierchia (2013), following ideas by Gajewski (2002, 2005), makes a distinction between grammatical and lexical contradictions. The contradictions in examples 55aand 55bare lexical, as they follow from the particular choice of lexical items (twiceNew York, for instance), and the contradiction disappears when one of the two instances ofNew Yorkis replaced byParis. However, in a sentence containingany, no lexical replacement can save the sentence. For Chierchia, only grammatical contradictions give rise to ungrammaticality judgments.

Giannakidou also argues that it is problematic that the Kadmon–Landman types of analyses apply only to indefinite NPIs. Although most NPIs are indefinites, not all of them are. For instance, NPIs likeeitherorneedare not.

The third problem for Giannakidou is that widening+ strengthening approaches do not naturally extend to cases where NPIs are licensed in nonveridical, non-DE contexts, such as modals. To solve this final problem, Giannakidou (2011) argues that even within the domain of indefinite NPIs a distinction must be drawn between NPI indefinites that appear in DE contexts only (and could potentially be analyzed as having their NPI property derived from their domain-widening effects) and those that do not. For her, the latter type of NPIs is lexically deficient for referentiality. She assumes that NPIs like Chineseshenme‘any,’ which are acceptable not only in DE contexts but also in all kinds of other nonveridical contexts, can be uttered felicitously only when they do not have to refer to some entity in the real world. Therefore, these elements are expected not to appear in veridical contexts. However, this approach predicts that, in principle, every such NPI is acceptable in all nonveridical contexts, which in turn is too weak a restriction on most NPIs.

4.2.2. Syntactic/lexical approaches. Although currently many scholars assume that sentences containing unlicensed NPIs are ill-formed for pragmatic and/or semantic reasons, others have argued that such sentences are ungrammatical as a result of some syntactic or lexical constraint. The tradition that takes NPIs to have a syntactic or lexical requirement that they be licensed by a (semi)negative operator goes back to Klima (1964), and this approach has been presented in more modern frameworks by Progovac (1993, 1994), who takes NPI licensing to be some special instance of syntactic binding; by Laka (1990); and most recently by Postal (2000, 2004) and Collins & Postal (2014).

Postal (2000, 2004), followed by Szabolcsi (2004) and, in a slightly different way, by Collins & Postal (2014), revives Klima’s theory and claims that NPIs, such as Englishany, underlyingly carry a negation, suggesting the following syntactic representation ofany:

(56) any: [DNEG [SOME]]

(20)

instance of syntactic feature checking or lexical incorporation, all kinds of elements should be able to act as NPIs, whereas the distribution of most if not all NPIs seems to be restricted semantically. Another problem is that it is sometimes hard to recognize the incorporated negation in an NPI-licensing context. Whereas one could legitimately assume that, in an NPI-NPI-licensing context,as fewis represented underlyingly as [NOT MANY], this is much harder to see for NPI-licensing contexts, such aseveryorif.

4.2.3. Concluding remarks. So far, we have observed that, even without the presence of n-words, NPIs do not seem to form a homogeneous class. We have also observed that various explanations for NPI-hood also face several problems. In addition, it should follow from every explanation for NPI-hood why n-words behave differently than other NPIs.

At the same time, the suggestion that NPI-hood is not a homogeneous class may also provide a solution to the problems noted in connection to the licensee question. It may very well be that various NPIs are NPIs for different reasons, as already suggested by Giannakidou (2011). For instance, whereas weak NPIs, which are almost always indefinites, may indeed be NPIs for semantic or pragmatic reasons (along the lines of Kadmon & Landman 1993; Krifka 1991, 1995; and Chierchia 2006, 2013), other NPIs need not always be indefinites. Lin et al. (2015), for instance, argue that Postal’s (2000, 2004) approach actually works for those NPIs that, like Englishneed, have a more limited distribution (between strong and weak NPIs). And superweak NPIs like Chinese

shenme, in contrast to weak NPIs, can readily be accounted for by Giannakidou’s approach.

Under that pluriform view, n-words would, again, be another type of NPIs, such as negative quantifiers that undergo quantifier resumption or indefinites that carry an uninterpretable negative feature that has to be checked by a negative operator in the syntax. The variety of NPI-hood then naturally guarantees that n-words may be some kinds of NPI while still being different from other, plain NPIs, which renders the second prediction stated at the end of the discussion on NC and n-words.

DISCLOSURE STATEMENT

The author is not aware of any affiliations, memberships, funding, or financial holdings that might be perceived as affecting the objectivity of this review.

ACKNOWLEDGMENTS

This article is based on earlier work on negation and negative polarity; in addition to the articles referred to in the text, this article is based on an earlier overview and handbook articles (Zeijlstra 2007, 2013, 2015) and on discussions with many people in this subfield, most notably Chris Collins, Gennaro Chierchia, Anastasia Giannakidou, Sabine Iatridou, Jing Lin, and Doris Penka. All errors are my own.

LITERATURE CITED

Abels K, Marti L. 2010. A unified approach to split scope.Nat. Lang. Semant.18:435–47

Carlson G. 1977.Reference to kinds in English. PhD thesis, Dep. Linguist., Univ. Mass., Amherst. 311 pp. Chierchia G. 2006. Broaden your views: implicatures of domain widening and the logicality of language.

Linguist. Inq.37:535–90

(21)

Chomsky N. 2001. Derivation by phase. InKen Hale: A Life in Language, ed. M Kenstovicz, pp. 1–54. Cambridge, MA: MIT Press

Collins C, Postal P. 2014.Classical NEG Raising. Cambridge, MA: MIT Press

De Swart H. 2000. Scope ambiguities with negative quantifiers.Stud. Linguist. Philos.72:109–32 De Swart H. 2010.Expression and Interpretation of Negation: An OT Typology. Dordrecht, Neth.: Springer De Swart H, Sag I. 2002. Negative concord in Romance.Linguist. Philos.25:373–417

Fauconnier G. 1975. Polarity and the scale principle.Chicago Linguist. Soc.11:188–99 Frege G. 1892. ¨Uber Sinn und Bedeutung.Z. Philos. Philos. Krit.100:25–50 Gajewski J. 2002.L-analyticity in natural language. Work. pap., Dep. Linguist., MIT

Gajewski J. 2005.Neg-raising: polarity and presupposition. PhD thesis, Dep. Linguist., MIT, Cambridge, MA. 184 pp.

Geurts B. 1999. On no.J. Semant.13:67–86

Giannakidou A. 1997.The landscape of polarity items. PhD thesis, Dep. Linguist., Univ. Groningen, Groningen, Neth. 230 pp.

Giannakidou A. 1998.Polarity Sensitivity as(Non)veridical Dependency. Amsterdam/Philadelphia: Benjamins Giannakidou A. 1999. Affective dependencies.Linguist. Philos.22:367–421

Giannakidou A. 2000. Negative. . .concord?Nat. Lang. Linguist. Theory18:457–523

Giannakidou A. 2011. Negative polarity and positive polarity: licensing, variation, and compositionality. In

The Handbook of Natural Language Meaning, ed. K von Heusinger, C Maienborn, P Portner, pp. 1660–712. Berlin: de Gruyter. 2nd ed.

Haegeman L. 1995.The Syntax of Negation. Cambridge, UK: Cambridge Univ. Press

Haegeman L, Lohndal T. 2010. Negative concord and (multiple) agree: a case study of West Flemish.Linguist. Inq.41:181–211

Haegeman L, Zanuttini R. 1996. Negative concord in West Flemish. InParameters and Functional Heads. Essays in Comparative Syntax, ed. A Belletti, L Rizzi, pp. 117–79. Oxford, UK: Oxford Univ. Press

Haspelmath M. 1997.Indefinite Pronouns.Oxford, UK: Oxford Univ. Press Herburger E. 2001. The negative concord puzzle revisited.J. Semant.9:289–333

Hiraiwa K. 2001. Multiple agreement and the defective intervention effect. InProceedings of the MIT–Harvard Joint Conference(HUMIT 2000), ed. O Matushansky, pp. 67–80. Cambridge, MA: MIT Work. Pap. Linguist.

Hoeksema J. 2008. Distributieprofielen van negatief-polaire uitdrukkingen: een vergelijking van het Nederlands, Engels en Duits.Tabu37:111–95

Horn LR. 1989.A Natural History of Negation. Chicago: Univ. Chicago Press

Iatridou S, Sichel I. 2011. Negative DPs, A-movement, and scope diminishment.Linguist. Inq.42:595–629 Jacobs J. 1980. Lexical decomposition in Montague grammar.Theor. Linguist.7:121–36

Janssen T. 1997. Compositionality (with an Appendix by B. Partee). InHandbook for Logic and Language, ed. J Van Benthem, A Ter Meulen, pp. 417–73. Cambridge, MA: MIT Press

Kadmon N, Landman F. 1993. ‘Any.’Linguist. Philos.16:353–422

Klima E. 1964. Negation in English. InThe Structure of Language, ed. JA Fodor, JJ Katz, pp. 246–323. Englewood Cliffs, NJ: Prentice Hall

Krifka M. 1991. Some remarks on polarity items. InSemantic Universals and Universal Semantics, ed. D Zaefferer, pp. 150–89. Berlin/New York: Foris

Krifka M. 1995. The semantics and pragmatics of polarity items in assertion.Linguist. Anal.15:209–57 Ladusaw WA. 1979.Polarity sensitivity as inherent scope relations. PhD thesis, Dep. Linguist., Univ. Tex., Austin.

215 pp.

Ladusaw WA. 1992. Expressing negation. InProceedings of the 2nd Semantics and Linguistic Theory Conference

(SALT 2), ed. C Barker, D Dowty, pp. 237–59. Ithaca, NY: Cornell Univ.

Ladusaw WA. 1996. Negation and polarity items. InThe Handbook of Contemporary Semantic Theory, ed. S Lappin, pp. 321–41. Oxford, UK: Blackwell

Laka MI. 1990.Negation in syntax: on the nature of functional categories and projections. PhD thesis, Dep. Linguist., MIT, Cambridge, MA. 270 pp.

(22)

Lin J, Weerman F, Zeijlstra H. 2015. Emerging NPIs: the acquisition of Dutchhoeven(‘need’).Linguist. Rev.

32:333–74

May R. 1985.Logical Form. Cambridge, MA: MIT Press

Penka D. 2007.Negative indefinites. PhD thesis, Eberhard-Karls-Univ. T ¨ubingen, T ¨ubingen, Ger.

Penka D. 2010.Oxford Studies in Theoretical Linguistics, vol. 32:Negative Indefinites. Oxford, UK: Oxford Univ. Press

Postal P. 2000.An introduction to the grammar of Squat. Work. pap., Dep. Linguist., NYU

Postal P. 2004. The structure of one type of American English vulgar minimizer. InSkeptical Linguistic Essays, ed. P Postal, pp. 159–72. New York: Oxford Univ. Press

Progovac L. 1993.A binding approach to polarity sensitivity. PhD thesis, Dep. Linguist., Univ. South. Calif., Pasadena. 850 pp.

Progovac L. 1994.Positive and Negative Polarity: A Binding Approach. Cambridge, UK: Cambridge Univ. Press Rullmann H. 1995. Geen eenheid.Tabu25:194–97

Szabolcsi A. 2004. Positive polarity—negative polarity.Nat. Lang. Linguist. Theory22:409–52

Ura H. 1996.Multiple feature-checking: a theory of grammatical function splitting. PhD thesis, Dep. Linguist., MIT, Cambridge, MA. 477 pp.

van der Wouden T. 1994.Negative contexts. PhD thesis, Dutch Dep., Univ. Groningen, Groningen, Neth. 243 pp.

von Fintel K, Iatridou S. 2003. Epistemic containment.Linguist. Inq.34:173–98 Watanabe A. 2004. The genesis of negative concord.Linguist. Inq.35:559–612

Zanuttini R. 1991.Syntactic properties of sentential negation: a comparative study of Romance languages.PhD thesis, Univ. Pa., Philadelphia

Zanuttini R. 1997. Negation and Clausal Structure: A Comparative Study of Romance Languages. New York/Oxford, UK: Oxford Univ. Press

Zanuttini R. 2001. Sentential negation. InThe Handbook of Contemporary Syntactic Theory, ed. M Baltin, C Collins, pp. 511–35. New York: Blackwell

Zeijlstra H. 2004.Sentential negation and negative concord. PhD thesis, Univ. Amsterdam. 315 pp.

Zeijlstra H. 2007. Negation in natural language: on the form and meaning of negative elements.Lang. Linguist. Compass1:498–518

Zeijlstra H. 2008. On the syntactic flexibility of formal features. InThe Limits of Syntactic Variation, ed. T Biberauer, pp. 143–73. Amsterdam: Benjamins

Zeijlstra H. 2011. On the syntactically complex status of negative indefinites.J. Comp. Ger. Linguist.14:111–38 Zeijlstra H. 2013. Negation and polarity. InThe Cambridge Handbook of Generative Syntax, ed. M den Dikken,

pp. 793–826. Cambridge, UK: Cambridge Univ. Press

Zeijlstra H. 2015. The morpho-syntactic realization of sentential negation. InThe International Syntax Hand-book, ed. A Alexiadou, T Kiss, pp. 275–310. Berlin: de Gruyter

Referensi

Dokumen terkait

Dari hasil pengelompokan Indeks Kerentanan Pantai diperoleh bahwa Wilayah yang sangat rentan terhadap kenaikan muka laut adalah sepanjang 2 km yaitu Desa Cikoneng,

[r]

This work was aimed to study of carbaryl and carbofuran analysis from fortified cabbage (Brassica oleracea) by liquid-liquid extraction, clean-up using SPE, followed by HPLC

Berdasarkan proses Pengadaan Langsung yang dilaksanakan untuk Pengadaan Penyedia Pekerjaan Konstruksi Pemeliharaan Rutin/ Berkala Gedung Kantor pada Dinas Kesehatan

Purposive Sampling dalam arti unit analisis (pengunjung wisata pemancingan, pengusaha kolam pemancingan, dan unsur pemerintah daerah) dipilih secara sengaja dengan

Judul Penelitian : Meningkatkan Prestasi Belajar PKN Pokok Bahasan Rumah Sehat pada Siswa Kelas V Semester II SDN Sumurgung II dengan Bimbingan

bahasa sebagai penggunaan dalam kajian bahasa. Keduanya saling melengkapi dan menyempurnakan untuk menemukan ideologi tersembunyi dari penutur. Linguistik dalam

Sedangkan nilai rata-rata ROA dari 40 laporan keuangan yang digunakan adalah sebesar 0,00878 dengan standar deviasi sebesar 0,010975 dimana ROA tertinggi