CHAPTER 1 INTRODUCTION
1.3 Outline of the Thesis
In chapter 2, I will discuss properties and previous analyses of licensing contexts of NPIs and the limits of these hypotheses. The semantic constraints and the syntactic patterns of the construction will be delved into. Existential wh-words with their own distinctive properties require different analyses of their own distributions.
In chapter 3, I will revise Cheng’s (1994) analysis of wh-words and make it suitable to wh-words in TSM. The licensing of wh-words by negation is not as
9
straightforward as that of typical NPIs. The types of modality/aspect should be put into consideration. Structural requirements as well as long-distance licensing will also be examined. Minimality comes into play when interrogative and existential licensers co-exist. Finally, I also deal with the licensing of the NPI jīm-hô ‘any’.
In chapter 4, I will sum up this study by giving a chart of the licensing contexts of the wh-words in TSM.
10
Chapter 2 Literature Review
In this chapter, I will discuss the properties and previous analyses of NPIs in English from the perspective of both semantics and syntax in section 2.1. Then I will present previous analyses of existential wh-words in Mandarin Chinese in section 2.2 and the analysis of the wh-words in TSM in section 2.3. Finally, a brief summary will be given in section 2.4.
2.1 The Licensing of Negative Polarity Items in English
Below I will delve into licensing contexts including negation, downward-entailing triggers, and nonveridicality as well as their limits and then Progovac’s binding approach will be probed into as well.
The well-known Fauconnier-Ladusaw (1975) hypothesis claims downward-entailing environments license most English NPIs.
Given a close look at those data shown in NPI items, we find that downward-entailing environments seem to be a necessary condition, as the example (1) shows:
11
(Portner 2005)
(1) a. I don’t have a dog. entails b. I don’t have a Keeshond.
As Portner (2005) points out, the fact that negation not creates entailments from properties to sub-properties can be described by saying that what is involved is downward entailing.
However, there are some counterexamples to this classic downward-entailing analysis in English. For instance, NPIs can appear in the scope of the generalized quantifier ‘only DP’, which is by no means downward-entailing trigger, as illustrated below (Xie 2007):
(2) Only John ate any fruit.
(3) apple ⊆ fruit
(4) Only John ate fruits. ! 2 (5) Only John ate apples.
Here ‘Only John ate fruits’ does not entail ‘Only John ate apples.’ John could eat something else like bananas, grapes, etc.
Von Fintel (1999) also provides some counterexamples. One of them is as follows:
2 The symbol ‘! ’ means not entail.
12
(Von Fintel 1999) (6) Only John ever ate any kale for breakfast.
(7) Only John ate vegetables for breakfast. ! Only John ate kale for breakfast.
On the other hand, Giannakidou (2002) proposes that NPIs appear in environments:
(8) where the propositions of the sentences are either negated or not asserted.
Giannakidou (1998, 2002) further proposes that the licensing of NPIs is
contingent upon whether the environment in which the NPI appears is nonveridical or not. Given the wide range of nondownward entailing environments, the notion of nonveridicality has gradually replaced Fauconnier-Ladusaw Hypothesis. The polarity licensers form the hierarchy as follows (Ernst 2008):
(9)
Nonveridicality theory evaluates a speaker’s commitment and the truth of the proposition Q. The licensers in the hierarchy above do not preserve truth value and thus the environments are nonveridical. The empirical data shows that
downward-13
entailing is not workable for some sentences. Instead, nonveridicality is a better one to account for the licensing of NPIs. The licensing contexts range from typical negation to quantifiers, downward-entailing, and finally to nonveridical sentences like
questions and conditionals. The following examples attest to this claim:
(Portner 2005) (10) a. Does Shelby ever bite?
b. Does Noah have any money?
(11) a. If Shelby ever bites you, I’ll put him up for adoption.
b. If Noah has any money, he can buy some candy.
Despite the fact that the nonveridicality theory seems to cover a wider licensing context for NPIs, it remains a purely semantic one and still needs further revision in terms of locality conditions. Progovac (1993) proposes an interactive model combining a modified version of Ladusaw’s (1980) downward entailment (DE) theory and the binding approach advocated in Progovac (1988). The inadequacies of the DE have been clarified in previous sections. The binding approach, however, takes into consideration the locality conditions. Progovac (1988) proposes that NPIs are governed by essentially the same principles as that of reflexives and pronouns.
That is, they are subject to the Binding Theory of Chomsky (1981). The relevant principles are as follows:
(Progovac 1993) (12) Principle A: An NPI must be bound in its governing category.
(13) Principle B: A PPI must be free in its governing category.
14
An NPI needs to be bound in its governing category. Their binders are negation and Op in Comp. Both of them are in A'-positions. If a NPI can’t find its binder in its own governing category, it raises in LF and thus satisfies their locality requirement in LF so long as it bears quantificational properties.
Chinese NPIs are well-known to be bound by superordinate negation, while the NPIs of other languages including English are assumed to be not (Wang 1996). The following examples (Progovac 1993), however, prove that some NPIs in English are licensed long distance:
(14) I am not saying that [IP John saw anyone].
(15) Peter denied Op that [IP John saw anyone].
(16) Op Did [IP John see anyone]?
(17) *I am not saying that [IP John arrived until 7 o’clock].
Among all the examples above, only the NPI until in example (17) is not licensed long distance. Progovac thus claims that ‘being quantificational in nature NPIs any/ever can raise in LF, thus satisfying their locality requirement in LF (cf.
Progovac (1988).’ And then adjoining to the local IP would suffice to bring the NPI within the same IP domain with Op in [Spec, CP] and raising to the Spec of Comp to be licensed by superordinate negation in the matrix Infl. (Progovac 1993). Note that there are always quantificational NPIs and being quantificational makes raising possible. Until, being an NPI, is not able to raise in LF because it lacks
quantificational quality.
Further exploration of superordinate licensing will be included in chapter 3.
15
2.2 The Licensing of Existential Wh-words in Mandarin Chinese
So far the theoretical frameworks mentioned above have mainly dealt with any in English. In this section, I will review the previous analysis of existential wh-word in Mandarin Chinese, including Li (1992) in 2.2.1, Cheng (1994) in 2.2.2 and Lin (1996) in 2.2.3 respectively.
2.2.1 Li’s (1992) Analysis
Li (1992) has dealt with indefinite wh-words in Mandarin Chinese. In her
analysis, wh-words can have a non-interrogative indefinite interpretation as well as an interrogative one. She also argues that wh-words are variables that need binders and that yes-no question markers, negation markers, and other negative-like operators are binders of indefinite wh-words and give them non-interrogative readings
(quantificational readings). On the other hand, if a element is bound by a wh-question operator, it is interpreted as interrogative.
In addition to semantic constraints, an indefinite wh-word in Mandarin Chinese is subject to syntactic constraints as well.
Semantically, the licensing context of an indefinite wh-word in Mandarin Chinese is affected by the truth value of the proposition containing the indefinite wh.
The licensing contexts of an indefinite wh-word include negation in (18) - (19), yes-no questions in (20), conditionals in (21), A-yes-not-A questions in (22), complements of non-factive verbs in (23), and sentences that express uncertainty in (24).
16
Negation
(18) a. Tā bù xǐhuān shéi.
(s)he NEG like who
‘He does not like anyone.
b. *Tā xǐhuān shéi.
(s)he like who ‘He likes anyone.’
(19) a. Tā bù wéi shéi gōngzuò.
he NEG for who work ‘He does not work for anyone.’
b. *Tā wéi shéi gōngzuò.
he for who work ‘He works for anyone’
Yes-No Questions
(20) a. Yǒu shéi xǐhuān nǐ ma?
have who like you Qu
‘Does someone like you?’
b. Měiguó hé Yīngguó yǒu shénme bùtóng ma?
America and England have what difference Qu
‘Is there any difference between America and England?’
17
Conditionals
(21) a. Rúguǒ yǒu shéi tǎoyàn tā, nǐ jiù gàosù wǒ.
if have who hate him you then tell me ‘If anyone hates him, then (you) tell me.’
b. Rúguǒ tā tǎoyàn shéi, jiù gàosù wǒ.
if he hates who then tell me ‘If he hates someone, then tell me.’
A-not-A Questions
(22) a. Nǐ zuótiān yǒu-méi-yǒu qù shénme dìfāng?
you yesterday have-not-have go what place ‘Did you go to any place yesterday?’
b. Yǒu-méi-yǒu shéi zài kàn?
have-not-have who ASP watch ‘Is anyone/someone watching?’
Non-factive Verbs
(23) a. * Nǐ xǐhuān shénme (dōngxī). (Li 1992) You like what (thing)
‘You like something/anything.’
(Li 1992) b. Wǒ yǐwéi /rènwéi/cāi/xīwàng nǐ xǐhuān shénme dōngxī.
I think/think/guess/hope you like what thing ‘I think/guess/hope that you like something.’
18
Modals that Express Uncertainty
(24) a. *Tā xǐhuān shénme. (Li 1992) he like what
‘He likes something.’
b. Tā dàgài/kěnéng xǐhuān shénme. (Li 1992) he probably like what
‘He probably likes something.’
To summarize all the contexts indefinite wh-words occur, Li (1992) lists the followings:
(25) a. contexts where the truth value is negated: negation
b. contexts where the truth value is not fixed: questions, conditionals, non-factive verb complements
c. contexts where the truth value is not asserted directly: seem, probably contexts, circumstantial le
Syntactically, the c-command is the relevant structural notion. This could be corroborated by the unavailability of the indefinite wh-word in (26) where the negation occurs within a sentential subject, in (27) where the negation is part of the adverb:
(26) *Tā bù lái duì shénme rén zuì hǎo. (Li 1992) he not come to what man most good
‘That he does not come is the best for someone.’
(27) *Tā bù-gāoxìng dì zuò shénme. (Li 1992) he not-happy DE do what
‘He did something unhappily.’
19
Thus, Li concludes that an indefinite wh-word must be c-commanded by its licenser at S-Structure.
2.2.2 Cheng’s (1994) Analysis
Cheng (1994), following Nishigauchi’s (1986, 1990) work on wh-words in Japanese and Heim’s (1982) work on indefinites, proposes that wh-words in Mandarin Chinese are indefinite NPs, which do not have inherent quantificational force. She further claims that judging from the environments in which the indefinite readings of wh-words are in, wh-words are polarity items. That is, wh-words are like indefinite NPs but need a licenser.
In Mandarin Chinese, wh-words can be interpreted as interrogative-words, existential quantifiers and universal quantifiers. Cheng assumes that a polarity item is an existential quantifier. Table 1 below illustrates both the interrogative and
existential reading of wh-words in Mandarin Chinese:
20
Table 1 The interrogative and existential readings of wh-words in Mandarin Chinese
examples as question words as quantifiers
shéi ‘who’ ‘anybody’
shénme ‘what’ ‘anything’
nǎ ‘which’ ‘any’
héshí ‘when’ ‘any time’
nǎlǐ ‘where’ ‘any place’
zěnme ‘how’ ‘any way’
wéishénme ‘why’ ‘any reason’
A-not-A ‘whether A or not’ ‘no matter whether A or not’
The existential reading of wh-words is either optional or obligatorily depending on the existence of yes-no questions. In other words, the interpretation of a wh-word varies depending on another element in the sentence. Cheng (1994) summarizes the environments of wh-words as follows:
(28)
a. Q
wh----wh (interrogative reading)
b. Qyes/no----wh (polarity/existential reading)
c. Neg----wh (interrogative or polarity/existential reading)
d. wh----dou (universal reading)
Cheng agrees with Li (1992) in treating wh-words in Chinese as variables. But she argues against Li’s (1992) analysis and treats the yes-no question particles,
A-no-21
A question markers, negation markers, and other negative-like operator as binders determining the quantificational force of indefinite wh-words. In her reasoning, if these operators are binders, then the indefinite wh-words must have a sentential scope just like these operators do. Thus, a wh-word requires both a licenser and a binder to determine its meaning. That is, a wh-word in Mandarin Chinese requires a licenser. At the same time, it also needs a binder to determine its own quantificational force. The wh-question-particle ne, for example, is not only licenser but also serves as the binder of the word(s) in its scope. Thus, when ne is present, the existential reading of wh-words is not possible. Crucially, when the existence of ne is not precluded, the
existential reading is possible. This is why the existential reading of wh-words is either optional or obligatory in the presence of negation since negation does not preclude the existence of ne. Overall speaking, the existential reading of wh-words is attainable only when there is no other binder.
The lack of quantificational force of words thus explains why subject wh-words are not allowed in Mandarin Chinese. Indefinite NPs could not occur in subject position in Mandarin Chinese and so are wh-words, given the fact that they both lack quantificational force. Wh-words or indefinite NPs, as Cheng claims, are never quantificational and thus cannot undergo Quantificational Raising. To receive quantificational force, wh-words have to be bound by existential closure, which only applies to the VP domain. Lowering of the subject NP to VP, however, is not possible in Mandarin Chinese. An A-not-A question is such an example. An A-not-A question can license indefinite wh-words in object position but not in subject position. Such subject-object asymmetries exist in A-not-A questions in Chinese Mandarin as well as TSM as illustrated below:
22
(29) a. Zhāngsān xǐ-bù-xǐ-huān shénme?
Zhangsan like-NEG-like what
‘Does Zhangsan like something/anything?’
b. * Shéi mǎi-bù-mǎi zhè běn shū?
who buy-not-buy this CL book
‘Did someone buy this book or not?’ (Intended reading)
c. A-káu sī-m̄ -sī kah-ì siáⁿmi̍h? (TSM) A-káu BE-NEG-BE like what
‘Does A-káu like something/anything?’
d. *Siáng sī-m̄ -sī kah-ì chit pún chheh? (TSM) who BE-NEG-BE this CL book
‘Does someone like this book? (Intended reading)
The ungrammaticality of (29b) and (29d) stems from the lack of quantificational force as well as the failure of meeting the structural requirement. Wh-words in these
examples are not c-commanded by their licensers. Recall that Li (1992) proposes that the occurrence of a non-interrogative wh-element is subject to a syntactic constraint: It needs a c-commanding licenser in overt syntax.
Another argument turns on a yes-no question with a wh-word as its subject as shown below:
(30)
*
Shéixiǎng chī shuǐguǒ ma?
who want eat fruit Q 'Does anyone want to eat fruits?'
The yes-no question marker ma in C(omplementizer) c-commands wh-word shei
‘who’ in subject position, but the sentence is ungrammatical. An indefinite wh-word needs both a binder and a licenser and ma can only serves as a licenser. This is why subject wh-words cannot have an existential reading. Subject wh-words fall out of
23
existential closure which applies to VP only. As a result, subject wh-words cannot receive quantificational force and the sentence remains ungrammatical even if the c-command requirement is met.
Cheng (1994) claims that the licensers for wh-words in Mandarin Chinese are the wh-question-particle, dou, the yes-no question particle, a A-not-A question marker, and negation for different readings of wh-words as demonstrated in Table 2 below:
Table 2 Various readings of wh-words and their licensers (triggers) and binders proposed in Cheng (1994)
Readings Licensers Binders
Interrogative Wh-question-particle Wh-question-particle
Universal dōu dōu
Existential Yes-no particle, A-not-A
question marker, Negation
VP existential closure
Although Cheng’s (1994) analysis mainly deals with the existential wh-words in Mandarin Chinese, it sheds light on the understanding of licensing contexts of wh-words in TSM. Whether Cheng’s (1994) analysis can be adopted completely to cover the existential wh-words in TSM will be probed into in the next chapter.
2.2.3 Lin’s (1996) Analysis
‘No Entailment of Existence Condition’ (NEEC) proposed by Lin (1996:119(46)) is given below: (EPW stands for Existential Polarity Wh-words.)
24
(31) Non-Entailment-of-Existence Condition on EPWs
The use of an EPW is felicitous if the proposition in which the EPW appears does not entail existence of a referent satisfying the description of the EPW.
NEEC constraint roughly amounts to Giannakidou’s Nonveridicality analysis.
They are different only because NEEC deals exclusively with EPWs.
Lin further points out that if this hypothesis is correct, it will be plausible to regard Chinese EPWs as the weakest type of NPIs in natural language in that they only need to satisfy the non-entailment-of-existence condition.
Non-factive verb complements are examples of NEEC as illustrated by the following examples in Mandarin Chinese:
Non-factive verb complements:
(32) a. Lǐsì yǐwéi nǐ dǎpò shénme (dōngxī)?/.
Lisi think you break what (thing) ‘What did Lisi think you broke?’
‘Lisi mistakenly thought you broke something.’
b. Lǐsì huáiyí nǐ dǎpò shénme (dōngxī)?/.
Lisi suspect you break what (thing)
‘What does Lisi suspect you have broken?’
‘Lisi suspects you have broken something.’
2.3 The Licensing Wh-words in Taiwanese Southern Min
Although the studies of wh-words in TSM are relatively rare, Hsiao (2003) provides insights into the mechanisms of wh-words in TSM.
Hsiao (2003) gives a list of wh-elements in English, Mandarin Chinese and
25
Taiwanese3:
Table 3 Wh-elements in English, Mandarin Chinese and TSM (Hsiao 2003)
English Mandarin Chinese Holo Taiwanese
‘who’ shéi, shénme rén siáng, siáⁿ(-mi̍h) lâng
‘what’ shénme/shá (colloquial) siáⁿ(-mi̍h)
‘when’ shénme shíhòu siáⁿ(-mi̍h) sî-chūn, tang-sî, tī
(tang)- sî
‘where’ shénme dìfāng, nǎér, nǎlǐ siáⁿ(-mi̍h),só͘-chāi, tó(-ūi)
‘how’ zěn(me)yàng, zěn(me) chóaⁿ-iūⁿ, chóaⁿ-( iūⁿ),
àn-chóaⁿ, àn-ná
‘how(many)’ duō(me) gōa
‘why’ wéishénme, zěnme ūi-siáⁿ-mi̍h, àn-chóaⁿ, ná,
thah4
Taiwanese words, like their counterparts in Mandarin Chinese, also have wh-interrogative readings, universal readings, and existential readings in Hsiao’s (2003) analysis:
Interrogative Reading
(33) a. Chi-beng tī tó(-ūi)/ siáⁿ-mi̍h só͘-chāi? (TSM) Chibeng at which place/what place
‘Where is Chibeng?’
3 Hsiao (2003) uses ‘Holo Taiwanese’, while ‘Taiwanese Southern Min’ is employed in this research.
26 ‘(S)he doesn’t have any financial affairs with anyone.’
(33a) and (33b) display a wh-interrogative reading and a universal reading
respectively. As for the existential readings, the licensing contexts include yes-no question in (33c), conditionals in (33d), and negation in (33e).
Seemingly, (34a) seems to be a yes-no question, but the answer to it will not be either yes or no as shown below:
(34) A: Lí ài siáⁿ bô? (TSM)
27
Among the licensing contexts, Hsiao (2003) points out that it is not always easy for a wh-word occurring with m̄ to get an existential reading:
(Hsiao 2003) (35) I m̄ (khéng) chia̍h siáⁿ? (TSM) (s)he NEG be-willing-to eat what
‘What is (s)he not willing to eat?’ (preferred reading) ‘(S)he won’t eat anything.’
On the other hand, she also indicates that not all cases of m̄ are defective in licensing an indefinite wh-word.
‘Volition’ is mentioned to be the key factor in her analysis. The negation m̄ with volition cannot license an indefinite wh-word easily.
Sound as the analysis is, we find that m̄ without volition does not necessarily license wh-words.
28
If ‘volition’ is what determines whether a wh-word is interpreted as existential or not, then the contrast between (36) and (37) cannot be explained. This issue still needs further investigation in the next chapter.
2.4 Summary
In this chapter, I have discussed properties and previous analyses of the licensing contexts such as downward-entailing and nonveridicality for NPIs in English. The latter proves to be a better theory in covering the licensing contexts of NPIs. The syntactic approach is also introduced to deal with long-distance licensing. As for the licensing of wh-words in Mandarin Chinese, Li’s (1992), Cheng’s (1994) and Lin’s (1996) analyses are reviewed. Cheng (1994) argues against Li’s (1992) analysis and treats the yes-no question particles, A-no-A question markers, negation markers, and other negative-like operator as binders determining the quantificational force of indefinite wh-words. She differentiates the licenser and the binder to explain the limited scope of wh-words. However, the requirement of c-command at S-structure proposed by Li (1992) is also insightful. Lin’s (1996) ‘No Entailment of Existence Condition’ (NEEC) corroborates nonveridicality. Hsiao’s (2003) analysis on wh-words in TSM is also valuable. The licensing by negation is not as straightforward as
Mandarin Chinese. In the next chapter, I will examine the Taiwanese data under previous analyses and propose a more comprehensive account.
29
Chapter 3 Toward a Solution
In this chapter I will revise the analyses of the licensing contexts of wh-words in TSM. In section 3.1, the licensing contexts of interrogative, universal and existential wh-words will be discussed. In section 3.2, the structural requirements of wh-words will be taken into consideration. I also deal with the licensing of the NPI jīm-hô ‘any’
in section 3.3. Finally, a brief summary will be given in section 3.4.
3.1 The Licensing Contexts of Wh-words in TSM
In this section, I will first deal with the licensing contexts of interrogative wh-words in TSM.
3.1.1 The Licensing Contexts of Interrogative and Universal Wh-words in TSM
So far Cheng’s (1994) classifications of various readings of wh-words and their
So far Cheng’s (1994) classifications of various readings of wh-words and their