Show Summary Details

Page of

Printed from Oxford Research Encyclopedias, Linguistics. Under the terms of the licence agreement, an individual user may print out a single article for personal use (for details see Privacy Policy and Legal Notice).

date: 29 February 2024

Negative Polarity Items in Chinesefree

Negative Polarity Items in Chinesefree

  • Bo XueBo XueChinese University of Hong Kong
  •  and Haihua PanHaihua PanChinese University of Hong Kong

Summary

Negative polarity items (NPIs) are well known for their limited distribution, that is, their negation-implicating contexts, the phenomenon of which has attracted much attention in generative linguistics since Klima’s seminal work. There is a large amount of research on NPI licensing that aims to (a) identify the range of potential licensors of NPIs, also known as Ladusaw’s licensor question; (b) ascertain the semantic/logical properties shared by these licensors; (c) elucidate the licensing dependency, for example, whether the dependency between an NPI and its licensor involves a structural requirement like c-command, and (d) shed light on the nature of polarity-sensitive items in natural languages and, more generally, the architectural organization of the syntax–semantics and semantics–pragmatics interfaces.

Theories of NPI licensing on the market abound, ranging from Klima’s affectivity to the influential Fauconnier–Ladusaw downward-entailingness (DE) as well as some weakened versions of Ladusaw’s licensing condition like (non-)veridicality and Strawson downward-entailingness. These theories are primarily concerned with pinpointing the logical properties of NPI licensors and elucidating the dependency between a licensor and its licensee. Broadly speaking, NPIs are assumed to be in the scope of some negative element. On the licensor side, various logical properties have been identified, resulting in a more fine-grained distinction between different negative strengths including downward-entailing, anti-additive, and anti-morphic. Moreover, a diverse class of NPIs has been uncovered and differentiated, including English weak NPIs like any/ever, for which simple DE would suffice, and stronger NPIs like in years/the minimizer sleep a wink, which are more selective and correlate with a stronger negative strength, namely, anti-additivity. Further theoretical developments of NPI licensing shift to the nature of NPIs and their communicative roles in a discourse, unearthing important properties like domain-widening in need of semantic strengthening (with its recent implementation in the alternative-and-exhaustification framework), which advances the understanding of their polarity-sensitive profiles.

Chinese NPIs include renhe-phrases (similar to English any) and wh-items, and minimizers, all of which are also confined to certain negative semantic contexts and not acceptable if they occur in simple positive episodic sentences without Chinese dou ‘all’. Descriptively, among canonical affective contexts(those including sentential negation, yes–no/wh questions, intensional verbs, if-clauses, imperatives, modals, adversative emotive predicates, adverb dou ‘all’, and the exclusive particle zhiyou ‘only’), renhe-phrases, and wh-items can be licensed by sentential negation, yes–no questions, intensional verbs, if-clauses, imperatives, modals, and the left restrictor of dou ‘all’, whereas minimizers like yi-fen qian ‘one penny’ display a more constrained distribution and can only be licensed by sentential negation, yes–no rhetorical questions, concessive if-clauses, and the left restrictor of dou.

There are at least two research questions worth exploring in the future. First, the affective contexts licensing Chinese renhe-phrases, wh-items, and minimizers are not totally the same, with minimizers being more constrained in their distribution. What could explain the unique behavior of Chinese minimizers? Why are these minimizers deviant in modal contexts and in need of the likelihood reasoning? Second, the affective contexts licensing Chinese NPIs do not totally overlap with those licensing English any. What could explain the divergent distributions of NPIs cross-linguistically?

Subjects

  • Applied Linguistics
  • Semantics

1. Introduction

Negative polarity items (NPIs) are well-known for their limited distribution, the phenomenon of which has attracted much attention in generative linguistics since Klima (1964). Specifically, the English NPI any results in ungrammaticality if it appears in an episodic positive sentence as shown in (1a). Note that “*” is used to indicate ungrammaticality throughout the article. However, sentence (1a) can be rescued by sentential negation as in (1b), indicating that an NPI like any requires a negative environment for grammatical licensing.

(1)

Such a licensing environment for NPIs has been described as affective by Klima (1964), who shows that in addition to sentential negation, a wide variety of negative operators including negative determiner phrases (DPs) and adverbial elements qualify as licensors for English any, as shown in (2) and (3) with NPI licensers underlined. Examples in (4) showcase more negation-implying semantic licensors for NPI any like the exclusive particle only and if-clause. The examples in (2)(5) are all taken from Homer (2021).

(2)

(3)

(4)

Homer (2021) reminds us that in some cases, it is not straightforward to identify a licensor of any because it can occur in questions and the than-clause of comparatives, as exemplified in (5).

(5)

This article concerns itself with NPIs in Chinese, including renhe-phrases, wh-items, and minimizers that are also confined to certain semantic contexts like English any as in the examples given in this section. It is organized as follows. Section 2 starts with a concise survey of the theories of NPI licensing on the market, ranging from the influential downward-entailingness (DE) story as detailed in Fauconnier (1975), Ladusaw (1980), and Zwarts (1998) to various ramifications of the DE-based theory and finally to the recent alternative-and-exhaustification framework as proposed by Krifka (1995) and Chierchia (2013). Against the theoretical background set up in Section 2, Section 3 provides an empirical description of the polarity-sensitive items in Chinese. Section 4 discusses some developments on NPI licensing in Chinese and shows how the theories proposed in the literature can be applied to account for the limited distribution of Chinese NPIs. Section 5 concludes the chapter.

2. A Concise Survey of the Theories of NPI Licensing

This section provides a critical survey of the theories of NPI licensing, which sets the stage for the semantics of Chinese NPIs to be discussed in Section 3. The research on NPI licensing primarily aims to (a) identify the range of potential licensors of NPIs, also known as the licensor question, as discussed by Ladusaw (1996); (b) ascertain the semantic/logical properties shared by these licensors (see Homer, 2021 for a recent overview); (c) investigate the licensing dependency, that is, whether the dependency involves a structural requirement like c-command in some level of a grammatical representation; and (d) shed light on the nature of polarity-sensitive items in natural languages as well as the semantics–pragmatics and syntax–pragmatics interfaces.

As discussed in Section 1, “Introduction,” it was pointed out by Klima (1964) that besides sentential negation, NPIs can also be licensed by a range of licensors, and a semantic property referred to as affectivity and an NPI must appear in an affective environment created by such a licensor. To make explicit the semantic properties of affectivity, which presumably correlate with polarization of a sentence (negation or affirmation), Ladusaw (1980) argues that what unites the class of licensors is that they denote downward-entailing functions that serve to reverse the direction of logical entailment. For instance, (6a) entails (6b) as the set of people who came in the afternoon must be a subset of those who came, that is {x | x came in the afternoon} {x | x came}. However, such a direction of inference from a subset to a superset gets reversed if negation appears in a sentence as demonstrated in (7), wherein the superset-to-subset inference is valid; namely that, from the fact that John didn’t come, it can be concluded that he didn’t come at a specific time, that is, in the afternoon.

(6)

(7)

Ladusaw (1980) proposes that negation as well as other canonical licensors of NPIs as shown in (2)(3) admit downward-entailing inferences, which is why they qualify as NPI licensors, and an NPI is grammatical only if it appears in the scope of such a downward-entailing licensor, which at the same time explains why upward-entailing operators fail to license any as shown in (2a’), (2b’) and (2c’). Below are definitions for downward entailing functions and downward entailing environments following much of the work of Ladusaw (1980), Zwarts (1998), von Fintel (1999), and Gajewski (2005).

(8)

Downward-entailingness (DE) as proposed by Ladusaw (1980) can generalize beyond sentential negation and therefore has the potential to cover those affective contexts mentioned in Klima (1964), which at the same time shows convincingly that NPIs in natural languages are governed by logical properties and some parts of grammar are determined by the logicality of human language. DE has proven quite successful in capturing the limited distribution of weak English NPIs like any/ever, and strong NPIs consisting of English minimizers like sleep a wink, budge an inch, and so on ; and strict NPIs like in years have been observed to be more constrained in their distribution (see Israel (2011) for a comprehensive catalog of such items in English). Specifically, the determiner at most licenses a weak NPI any in its nuclear scope but not the minimizer sleep a wink or the strict NPI in years, as illustrated in (9).

(9)

Since merely relying on DE is not sufficient to predict the distribution of all kinds of NPIs, particularly, the group of strong NPIs including minimizers like sleep a wink and strict NPIs like in years, Zwarts (1998) argues that distinct kinds of NPIs, that is, the weak NPI any/ever and strong NPIs sleep a wink/in years correlate with varying degrees of negative strength ranging from downward entailing and anti-additive (AA) to antimorphic. Specifically, for the examples in (9), according to Zwarts (1998), strong NPIs like sleep a wink/in years are only licensed by an AA operator that is stronger than vanilla DE, explaining why strong NPIs are more constrained and only licensed by a subset of downward-entailing semantic operators. The definition of AA is shown in (10), based on (Zwarts, 1998). Examples in (11) demonstrate that the determiner at most does not denote an AA function though it is downward entailing. Therefore, making a more fine-grained distinction among various licensers in terms of their negative strength helps explain why a strong NPI like sleep a wink or in years is not licensed by at most whereas a weak NPI any is.

(10)

(11)

Moreover, following the lead of the DE tradition, von Fintel (1999) offers an account based on the notion of Strawson downward entailingness1 to explain that the exclusive particle only (as well as adversative emotive factives like surprised/sorry) can license weak NPIs like any in its scope. This shows that a DE-based account and the extension of it can be very promising to provide an answer to the licenser question, that is, what semantic properties qualify certain operators as legitimate licensers for NPIs? For an alternative view based on the notion of (non-)veridicality2 to derive polarity sensitivity, see Giannakidou (1998), whose work considers a wider range of affective contexts in which NPI any gets licensed, including questions, intensional verbs, generics, imperatives, comparatives, superlatives, subjunctives/modals, and so on. For a detailed discussion on limited distributions of cross-linguistic polarity-sensitive items with respect to an exhaustive list of such semantic contexts, see Giannakidou (1998).

Interestingly, the fact that only DE contexts, including negation and antecedents of conditionals among others, are crucial for NPI licensing also raises an important question concerning the nature of NPIs. What kinds of semantic and pragmatic functions, or rather communicative roles, do such polarity-sensitive items perform in a sentence that make them so susceptible to downward-entailing contexts? Kadmon and Landman (1993) provide an answer based on the following conversation. Suppose B is a cook. Notably, after the first turn of information exchange, B uses stressed any in the second utterance to convey that no potatoes are available, even including those that would be considered as marginal ones. Kadmon and Landman argue that the use of NPI any in this case extends the range of its nominal restrictor potatoes along some contextually determined dimension, giving rise to the domain-widening effect.

(12)

Kadmon and Landman further submit that the domain-widening effect of an NPI must lead to a logically stronger interpretation, explaining why an NPI must be licensed by a DE operator, with an overall effect of semantic strengthening. Kadmon and Landman advance the understanding toward the semantic and pragmatic functions of NPI any, deriving its polarity sensitivity in a principled manner (Homer, 2021). Hence, the need for DE results from the domain-widening effect and the semantic-strengthening requirement of an NPI.

There have been further developments of the semantic-strengthening analysis argued by Kadmon and Landman (1993) including Krifka (1995) and Chierchia (2013). Specifically, Chierchia (2013) draws our attention to one major observation that is referred to as the “scalar implicature (SI) and polarity-sensitive item (PSI) parallelism.” One signature property of SIs is that they tend to disappear under DE contexts that happen to provide licensing conditions for NPIs.

For instance, the SI some but not all for the object quantifier can be derived in (13a), which is stronger than the plain existential quantification. However, such an SI disappears in (13b) under sentential negation. The reason why SIs abhor DE is that embedding SIs under DE leads to semantic weakening, which clashes with the original motivation for deriving SIs, that is, to obtain a stronger interpretation. Hence, Chierchia argues that what underlies the distribution of both SIs and NPIs relates to semantic strengthening, and both are particularly sensitive to polarity-affecting contexts.

(13)

Based on this, Chierchia (2013) proposes an alternative-and-exhaustification account to explain the well-noted restricted distribution of NPIs. Concretely, Chierchia argues that NPIs like English any obligatorily activate a set of subdomain alternatives that must be checked by an exhaustifier which requires that the use of any lead to a strengthened interpretation. Hence, the major difference between Kadmon and Landman (1993) and Chierchia (2013) is that the exhaustification view requires the strengthening effect to be checked by a grammatical operator EXH akin to only, a version of which is provided in (14). Simplifying a lot, EXH is a binary focus-evaluating operator taking a prejacent p and its alternative set Alt(p)<<s, t>,t> as arguments, affirms its prejacent and negates all stronger alternatives to the prejacent.

(14)

According to Chierchia, NPI any, like plain indefinites, amounts to existential quantification. However, NPI any is different from canonical existential quantification in that any obligatorily activates a set of alternatives (Rooth, 1985) that must be consumed by a focus-evaluating operator, namely, EXH. Hence, the traditional licensing dependency must be checked via a grammatical exhaustification operator.

Viewed from the alternative-and-exhaustification system, the reason why any is not allowed in a positive episodic sentence in (15a) is that the obligatory exhaustification of the activated subdomain alternatives (DA alternatives) of any cookies leads to a contradiction as shown by the derivation in (15b). The contradictory result is further ruled out by the G-triviality that postulates that the logical skeleton of a grammatical sentence should not be trivial, that is, a tautology or contradiction. To unpack the proposal, suppose there are two cookies a and b in the model. Then the assertion there are any cookies left is x {a, b}. cookies-(x) left-(x), which consists of two domain alternatives, namely, a. cookies-(a) left-(a) and b. cookies-(b) left-(b). The assertion is incompatible with the result of exhaustification, that is, negation of its two stronger subdomain alternatives: ¬[a. cookies-(a) left-(a)] ¬[b. cookies-(b) left-(b)], as shown in (15b). Note that σ-alternatives mean scalar competitors in Chierchia’s framework and for the example at hand, the σ-alternative of an existential quantifier which essentially involves disjunction is a conjunction and negating the scalar alternative will not incur contradiction.

(15)

However, with the help of a downward-entailing operator, for example, negation in (16a), the sentence does not result in ungrammaticality because negation reverses the entailment, as shown in (16b). Exhaustifying above the negation is vacuous since the assertion ¬[x {a, b}. cookies-(a) left-(a)] in this case is stronger than its subdomain alternatives. Hence, no contradiction is derived.

(16)

In a nutshell, Chierchia’s alternative-and-exhaustification system crucially centers on the semantic-strengthening process as proposed by Kadmon and Landman (1993) and further formalizes Krifka’s (1995) insights to derive a more complete typology of a wide range of polarity-sensitive items, which can also account for the intervention effects observed by Linebarger (1987). Moreover, such an alternative-based approach brings together many interesting semantic/pragmatic phenomena, for example, scalar implicatures and free-choice inferences, by extending the framework of Alternative Semantics (Hamblin, 1973; Kratzer & Shimoyama, 2002; Rooth, 1985).

This section contains a concise survey of the major theories of NPI licensing, providing necessary background for us to discuss Chinese NPIs and recent developments on NPI licensing in Chinese. The next section offers an empirical description of Chinese NPIs, including renhe-phrases, wh-indefinites, and minimizers.

3. The Empirical Landscape of Chinese NPIs

This section focuses on three types of Chinese NPIs including renhe-phrases, wh-items, and minimizers, offering an empirical description of their limited distribution. Starting with Chinese renhe-phrases, it has been well documented (e.g., Giannakidou & Cheng, 2006;3 Liao, 2010) that a renhe-phrase is unacceptable in a plain episodic positive sentence, as shown in (17a), whereas they become acceptable in a downward-entailing context with sentential negation as shown in (17b).

(17)

Relatedly, it has also been observed that Chinese wh-items have noninterrogative indefinite uses and behave like polarity-sensitive items (for treating Chinese wh-phrases as polarity-sensitive items, see Cheng, 1997; Chierchia & Liao, 2015; Giannakidou & Cheng, 2006; Li, 1992; Liao, 2010; Lin, 1996; Y.-M. Xiang, 2020, among many others). Using Chinese wh-item shenme ‘what’ as an example throughout the article, the contrast in (18) demonstrates that shenme ‘what’ cannot appear in an episodic positive clause and needs a downward-entailing operator for semantic licensing. Note that (18b) has two readings. The first reading means that Zhangsan didn’t read any book. The second reading conveys that Zhangsan didn’t read particular books, that is, interesting books worth mentioning. This article focuses on the first reading and the second reading will be left for future discussion.

(18)

Importantly, Chinese minimizers behave like renhe-phrases and noninterrogative wh-items. (19a) shows that the lowest-quantity-denoting minimizer yi-fen qian ‘one-CL money’ is weird when used out of the blue, since there are so many cases that could make this claim true and one penny is the most likely amount to be spent away. “#” is used to indicate pragmatic infelicity throughout the article. However, (19b) illustrates that, if such a minimizer appears within the c-commanding domain of negation, its acceptability significantly improves. In this respect, yi-fen qian ‘one-CL money’ is just like English minimizers, including lift a finger, sleep a wink, and budge an inch, which requires a downward-entailing operator to deliver a stronger interpretation.

(19)

In addition to sentential negation, it is worth exploring under what kinds of other affective contexts renhe-phrases, wh-items, and minimizers can be licensed. The sentences in (20) showcase the distribution of renhe-phrases in nine (nonexhaustive) affective contexts involving yes–no/wh nonrhetorical questions, intensional verbs, if-clauses, imperatives, modals, adversative emotive predicates, the universal quantifier dou, and the exclusive particle zhiyou ‘only’. Among these nine canonical affective contexts, a renhe-phrase can only be licensed by a desiderative intensional verb in (20c), an if-clause in (20d), an imperative in (20e), a modal in (20f), or Chinese dou as shown in (20i), which demonstrates that a renhe-phrase in Chinese can be quite selective about its licensors, unlike English any. For instance, English any can appear in a nonrhetorical wh-question, as in Who saw anything? or a sentence with only, as in (4b), which, however, are infelicitous contexts for Chinese prenominal renhe. Moreover, (20g) and (20h) show that Chinese renhe is incompatible with adversative emotive factive predicates like jingya ‘surprised’ or yihan ‘sorry’, which create the so-called Strawson downward-entailing contexts and license English any, as argued by von Fintel (1999).

(20)

Among those contexts where a renhe-phrase can be licensed, three remarks are in order. First, a renhe-phrase seems to prefer desiderative attitudes over doxastic ones, as illustrated in (20c) and (20c’). Second, like English any in Pick any card!, Chinese renhe can be licensed by an imperative in (20e), showing that Chinese renhe can derive a free-choice interpretation. The presence of renhe in (20e), just like English any, indicates that the addressee can take any book to fulfill the speaker’s request and the addressee is welcome to choose freely any one from a range of contextually determined books available. In a similar vein, in (20f), renhe is c-commanded by a deontic modal keyi ‘can’ and also yields a free-choice interpretation conveying that all kinds of books are available for the subject to choose; see Giannakidou and Cheng (2006) and Y.-M. Xiang (2020) for more discussion. Finally, the minimal pair of (20i) and (20i’) further shows that renhe is only licensed if it appears to the left of dou and picks up dou’s universal quantificational force. If renhe appears to the right of dou as in (20i’), the sentence becomes ungrammatical. This suggests that NPI renhe must occupy a certain region, that is, the left of dou, in a dou-construction to be properly licensed, a point of which will be elaborated in Section 4.2. It should be stressed that NPI renhe can appear in simple positive episodic sentences with dou, which makes Chinese different from many other languages.

Moving on to Chinese wh-items, the sentences in (21) show how the wh-item shenme ‘what’ behaves with respect to the same set of affective contexts. For more detailed discussion on the varieties of affective contexts for licensing Chinese indefinite wh-items, see Li (1992) and Lin (1996, 1998). The contexts that license Chinese indefinite wh-items include a yes–no nonrhetorical question as in (21a); an intensional context involving attitudinal predicates like desiderative xiang ‘want’ in (21c) and doxastic renwei ‘think’ in (21c’); a downward-entailingness (DE) if-clause in (21d); an imperative in (21e); and an epistemic modal in (21f), as well as the universal quantifier dou in (21i). Like Chinese renhe, the wh-item shenme ‘what’ cannot be licensed by a wh-nonrhetorical question in (21b), adversative predicates in (21g)(21h) or the exclusive particle zhiyou ‘only’ in (21j). Moreover, when a Chinese wh-indefinite occurs in a dou-sentence in (21), it has to appear to the left of dou and gives rise to the universal quantification interpretation (Cheng, 1997; Feng & Pan, 2018; Jiang & Pan, 2013; Lee, 1986; Lin, 1998; Pan, 2006, among others). However, a noninterrogative wh-indefinite cannot be licensed to the right of dou, as exemplified in (21i’).

(21)

It has been established by Li (1992) and Lin (1996, 1998) that overall, the felicity of a Chinese noninterrogative wh-item in a sentence is governed by a complex range of syntactic and semantic factors. More specifically, Chinese noninterrogative wh-items occur in semantically affective contexts that the truth value of a sentence is negated or yet to be determined.

By investigating those intensional contexts and beyond that license Chinese indefinite wh-items, Li (1992) observes that these wh-items generally go hand in hand with those semantic operators that convey epistemic or inferential tentativeness, an insight of which has been further formalized as the Non-Entailment-of-Existence Condition in Lin (1998). The Non-Entailment-of-Existence Condition states that an indefinite wh-item in Chinese only occurs in a proposition that does not entail existence of a referent satisfying the restrictor of the wh-phrase involved, which has the potential to unify a wide range of semantic licensing contexts for Chinese wh-indefinites.

Concretely, the feel of tentativeness in licensing Chinese wh-indefinites can be witnessed in (21e) wherein the wh-indefinite shenme ‘what’ is more comfortable in an imperative with a tentative tone or a weak illocutionary force. Chinese wh-indefinites are also well-known for inducing some epistemic effect as shown in (21f), in which an attitude holder, i.e., the speaker, may not know the actual name of the book involved. Furthermore, Chinese wh-indefinites like shenme ‘what’ can convey extra pragmatic meanings, like insignificancy, cf. Z. Chen (2021) for a recent discussion.

The last type of Chinese NPIs examined in this section concerns minimizers. As already shown in (19), the minimizer yi-fen qian ‘one-CL penny’ denotes the lowest amount of money by occupying the smallest end of a scale, whose contribution in most cases would be the most uninformative as there are so many everyday transactions involving a larger amount of money that asymmetrically entails one penny. Unless the minimizer yi-fen qian ‘one-CL penny’ appears in a downward entailing or scale-reversing context, the lowest-quantity-denoting minimizer results in pragmatic infelicity.

The sentences in (22) show the distribution of the minimizer yi-fen qian ‘one-CL penny’ with respect to the same set of affective contexts. Interestingly, yi-fen qian ‘one-CL penny’ only yields a sensible interpretation when it appears in a yes–no rhetorical question in (22a), a concessive if-clause in (22d’) or to the left of Chinese dou in (22h). Notably, the minimizer yi-fen qian ‘one-CL penny’ results in pragmatic infelicity in the remaining affective contexts.

(22)

It is worth remarking that when the minimizer yi-fen qian ‘one-CL penny’ is licensed in a yes–no question in (22a), the question conveys a negative rhetorical flavor. That is, (22a) is biased toward the negative answer, meaning that he never spent even one penny on the addressee, which is a peculiar property of minimizers in questions, see Guerzoni (2003). Note that the wh-indefinite shenme ‘what’ that appears in a yes–no question as in (21a) does not make the question biased toward its negative answer. In other words, shenme ‘what’ is interpreted neutrally as an indefinite in a yes–no question, whereas the minimizer yi-fen qian ‘one-CL penny’ is not. The same contrast can be replicated in English as shown in (23) and (24), which shows that the presence of English minimizers like lift a finger, the least bit of, and the faintest idea of in (23) renders a yes–no question biased toward a negative answer, whereas English garden-variety NPIs like any/ever denote plain indefinites and contribute neutral yes–no questions in (24). An account that explores the semantics and pragmatics of English minimizers in questions has been offered by Guerzoni (2003).

(23)

(24)

Moreover, the minimizer yi-fen qian ‘one-CL penny’ is not licensed by a plain if-clause as shown in (22d). However, yi-fen qian ‘one-CL penny’ can be licensed by a concessive if-clause in (22d’), giving rise to the scale-reversal interpretation that no matter how small the amount of money the addressee picks up, he/she has to hand it in, which is equivalent to universal quantification, that is, in the case of all money the addressee picks up, he/she has to hand it in. It has been pointed out by Fauconnier (1975), Krifka (1995), Lahiri (1998), Krifka (1995), Guerzoni (2003), Chierchia (2013) and many others that minimizers tend to occur in contexts that support scalar reasoning involving the scalar particle even (Karttunen & Peters, 1979). Finally, just like renhe-phrases and wh-indefinites, the minimizer yi-fen qian ‘one-CL penny’ has to appear to the left of dou, also yielding the universal quantification interpretation as in (22h).

All kinds of affective contexts—including negation, questions, intensionality, if-conditional, the universal quantifier dou, and the exclusive particle zhiyou ‘only’—that could potentially function as semantic licensors for the three types of Chinese NPIs, namely, renhe-phrase, wh-items, and minimizers, are summarized in Table 1, which provides a panoramic view of the limited distribution of Chinese NPIs. Note that the list of potential semantic licensors discussed in this article is by no means exhaustive, which only serves as a springboard for further discussion and investigation of Chinese NPIs. A similar range of affective contexts has been discussed in the literature including, for example, Li (1992), Lin (1998), and Giannakidou (1998).

Table 1. Chinese NPIs and Their Licensers

Semantic licensers for NPIs

renhe-phrases

wh-items

minimizers

Sentential negation

✔︎

✔︎

✔︎

Yes–no question

✔︎nonrhetorical

✔︎ rhetorical

Wh-question

#

Intensional verb

✔︎

✔︎

#

If-clause

✔︎

✔︎

✔︎concessive

Imperative

✔︎

✔︎

#

Modal

✔︎

✔︎epistemic

#

Adversative emotive predicate

#

The left of dou

✔︎

✔︎

✔︎

Exclusive particle zhiyou ‘only’

#

As summarized in Table 1, the affective contexts licensing Chinese renhe-phrases, wh-items, and minimizers are not totally the same, with minimizers being more constrained in their distribution. Specifically, there are at least two issues worth exploring, which concern differences within these three classes of Chinese NPIs. First, although both renhe-phrases and wh-items can be easily licensed by epistemic or deontic modals, minimizers are still deviant in such contexts, suggesting that more research is needed concerning the semantics and pragmatics of Chinese minimizers, which have not received enough attention in the literature. Undeniably, the more limited distribution of Chinese minimizers strongly correlates with the varying strengths of DE functions as discussed by Zwarts (1998), and a finer-grained categorization of the affective contexts licensing Chinese minimizers can characterize their behaviors more precisely. On a related note, Chinese minimizers need to be licensed by a concessive if-clause but not a simple if-clause, the latter of which readily licenses renhe-phrases and wh-items. Second, the semantics of Chinese renhe-phrases and wh-items in modal contexts differ from each other: a renhe-phrase delivers the free-choice interpretation, that is, any option is available to the agent involved, whereas a wh-item in a modal context tends to induce some epistemic effect relative to a certain attitude holder. More investigation is needed to explore such delicate differences between Chinese renhe-phrases and wh-items.

4. Some Recent Developments

4.1 Applying the Alternative-and-Exhaustification Approach

Liao (2010) extends the alternative-and-exhaustification approach to derive the polarity-sensitive profile of Chinese noninterrogative wh-items. Liao (2010) mainly focuses on three cases of Chinese noninterrogative wh-items. First, a wh-item c-commanded by sentential negation, as in (25a), behaves like English any. Second, a wh-item in construction with a modal element like adverbial haoxiang ‘seem’ receives an existential construal, shown in (25b). Third, a wh-item can occur in a positive episodic context only with the presence of dou, deriving a universal interpretation in (25c), as has been discussed in the previous section. The following examples and glosses are taken from Liao (2010).

(25)

The proposal in Liao (2010), following Krifka (1995) and Chierchia (2013), in a nutshell, is that Chinese wh-items always contribute existential quantification, as exemplified by shenme ‘what’ in (26a), and noninterrogative uses of such wh-items lexically require an exhaustifier EXH (whose semantics has been offered in (14)). Moreover, Chinese wh-items in their noninterrogative uses are hypothesized to activate subdomain alternatives and scalar alternatives, as shown in (26b) and (26c) respectively. Subdomain alternatives represent a subset of the wh-indefinite quantificational domain and the scalar alternative to the wh-indefinite is assumed to be universal quantification, all of which have to be exhaustified by EXH in a recursive manner following the system as laid out by Chierchia (2013).

(26)

Building on the Non-Entailment-of-Existence Condition proposed by Lin (1998), Liao (2010) argues that Chinese noninterrogative wh-items must appear under a propositional nonveridical operator that does not entail its prejacent. Such nonveridical operators consist of sentential negation in (25b) and a modal element in (25c). Crucially, without such nonveridical operators, exhaustifying over an alternative set including both subdomain and scalar alternatives of a wh-indefinite results in contradiction, an undesirable effect.

The derivations in (27) illustrate why recursive exhaustification over a positive episodic sentence containing shenme ‘what’ results in contradiction. Supposing that shenme ‘what’ ranges over two alternatives a and b in a model, and the assertation of Zhangsan chi-le shenme ‘Zhangsan ate-perf what’ can be abbreviated as a b with its subdomain and scalar alternatives shown in (27a) and (27b). Liao (2010) assumes that exhaustification over the subdomain alternatives applies twice with the first round of exhaustification being able to feed the second round. The first round of exhaustification is underlined in (27c), which then gets exhaustified again and yields a strengthened conjunctive interpretation, a b, that is, for both a and b, ZS ate them.

(27)

However, by exhaustifying over the subdomain alternatives, the strengthened implicatures clash with the scalar implicature of the sentence, which is obtained by negating the scalar alternative in (27b), i.e., ¬x. [x {a, b} thing ate (ZS, x)]. Liao (2010) assumes that the contradictory implicatures derived by recursive exhaustification is the reason why the sentence of Zhangsan chi-le shenme ‘Zhangsan ate- perf what’ without a nonveridical operator is ungrammatical. When a wh-item is embedded under a nonveridical operator like sentential negation or a modal adverb, no such contradiction emerges. Therefore, Liao (2010) essentially treats Chinese noninterrogative wh-items as on a par with English any, and the explanation as to why a wh-indefinite in Chinese cannot appear in a positive episodic sentence is the same as that for English any, already shown in (15b) and (16b).

Turning to the only case involving Chinese dou by which a wh-indefinite can be licensed in a positive episodic sentence, Liao (2010) following Mok and Rose (1997) and M. Xiang (2008) argues that Chinese dou as a focus marker functions like English even, which carries a scalar presupposition, stating that dou’s prejacent is the most unlikely/noteworthy among its triggered alternatives as shown in (28a). Dou itself is semantically vacuous. Moreover, Liao assumes that dou carries an additive presupposition, requiring that there be at least one alternative to dou’s prejacent q and such alternatives p are all possible, as shown in (28b).

(28)

Hence, the reason why a positive episodic sentence with dou licenses a noninterrogative wh-item and incurs the resecuring effect is that dou’s presupposition in (28b) renders the scalar implicature of dou’s prejacent unavailable and circumvents a contradictory exhaustification. Specifically, when dou is present in the same positive episodic sentence Zhangsan chi-le shenme ‘Zhangsan ate- perf what’ as illustrated in (28c), recursive exhaustification over the subdomain alternatives of the sentence derives the same result as (27c). However, (28c) is different from (27c) in that in (28c), dou’s presupposition preempts the derivation of the scalar implicature, that is ¬x. [x {a, b} thing ate (ZS, x)]. Therefore, there is no contradiction derived for a sentence with dou like (28c), yielding a universal reading via recursive exhaustification of the subdomain alternatives.

(28)

Although Liao (2010) offers an alternative-based account aiming to explain the limited distribution of Chinese noninterrogative wh-items by treating such wh-items as on a par with English any, one has to bear in mind that the licensing contexts for Chinese noninterrogative wh-items and English any are not totally the same, an observation already pointed out by Lin (1998). Most importantly, as discussed in Section 3, English any can be licensed by only or adversative emotive predicates, and such contexts do not admit Chinese wh-indefinites. Why does the Strawson downward-entailingness (SDE) involving only and adversative emotive predicates fail to license Chinese wh-indefinites, renhe-phrases, and minimizers? This question needs to be addressed in future research.

Another problem associated with Liao (2010) is that although it is recognized that dou licenses noninterrogative wh-items, the account developed by Liao (2010) fails to capture a more complicated distributional asymmetry with respect to Chinese NPIs in sentences involving dou. That is, as shown in (20i-i’), (21i-i’), and (22h-h’), a renhe-phrase, wh-item and minimizer must situate to the left of dou to be licensed but not to the right of dou, which raises the question as to how to account for such a distributional asymmetry within dou-sentences. Section 4.2 introduces an LF-downward-entailingness (DE) account proposed by L. Chen and Pan (2020), which reaffirms the traditional DE licensing strategy and shows that by analyzing dou as a universal quantifier, the problem of the distributional asymmetry of Chinese negative polarity items (NPIs) in dou-sentences can be resolved straightforwardly combined with the DE-based theory on NPI licensing.

4.2 An LF–DE Account and a Case Study of Chinese Dou

Section 3 shows that Chinese renhe-phrases, wh-items, and minimizers behave like English NPI any, which displays the same propensity to be licensed by semantically downward-entailing contexts. In a positive episodic sentence containing dou, a renhe-phrase, wh-item, or minimizer must be situated to the left of dou to be licensed but not to its right, suggesting that dou has some rescuing effects on such items. However, the fact that these three types of NPIs must appear to the left of dou raises the question as to why this is the case. The relevant examples discussed in Section 3 are repeated below.

(29)

(30)

(31)

As stressed in Section 4.1, the alternative-based account developed by Liao (2010) owes us an explanation as to why only dou’s left region licenses Chinese NPIs whereas its right region does not. This section introduces another account proposed by L. Chen and Pan (2020), which shows that analyzing Chinese dou as a universal quantifier (Jiang & Pan, 2013; Lee, 1986; Lin, 1998; Pan, 2006; among many others) combined with the traditional DE-based theory has the advantage of capturing the distributional asymmetry of NPIs in dou-sentences. Specifically, L. Chen and Pan (2020) offer an LF–DE account to address the distributional asymmetry of NPIs in dou-sentences. The upshot of their analysis is that Mandarin NPIs must be licensed by semantic downward-entailingness and evaluated at an abstract level of a dou-sentence, namely, its logical form (LF).

Before introducing their LF–DE analysis, (32) provides more examples to sharpen their observation that only dou’s left region is capable of licensing renhe-phrases,5 deriving the universal quantification of renhe-phrases. Since the major question concerns the limited distribution of NPIs in dou-sentences, possibility modals are intentionally left out in the examples below to focus on dou’s licensing capability.

(32)

Crucially, if dou serves as a universal quantifier, it is predicted that Chinese NPIs like renhe-phrases should only be licensed in the restrictor of the tripartite structure of dou, that is, being associated with dou to its left, as shown in (32). Dou’s behavior is thus reminiscent of the universal quantifier every in English which only licenses the NPI any in its restrictor, which is semantically downward entailing (for English every, see Barwise & Cooper, 1981). Furthermore, the following three contrasts in (33) further corroborate that only the area to the left of dou, namely the restrictor region, can create a downward-entailing context, rendering renhe-NPs acceptable in (33a), (33c), and (33e), whereas the right of dou, that is, its nuclear scope, is upward entailing, precluding renhe-NPs from being licensed, as shown in (33b), (33d), and (33f). Note that (33e) is telling in two respects. First, the comparison standard containing the renhe-phrase must appear to the left of dou and the post-dou renhe-phrase results in deviance, as shown in (33f).6

(33)

It appears that at LF, dou demarcates a boundary of two regions, the left of which instantiates DE, therefore, licensing Chinese NPIs, namely, renhe-NPs as in (33a), (33c), and (33e). The right area of dou is upward-entailing modulo English every where renhe-phrases cannot be licensed to the right of dou, as shown in (33b), (33d), and (33f).

More examples involving Chinese minimizers are offered in (34). Note that such minimizers also can only get licensed to the left of dou, as exemplified in (34a), whereas the one to the right of dou in (34b) is ungrammatical, as the minimizer to the left of dou in (34a) yields the scale-reversal interpretation under a downward-entailing context that he spent all of the money and his pocket was completely empty. Additionally, the contrast between (34c) and (34d) indicates that the quasi-minimizer biaodian fuhao ‘punctuation marks’ can only appear to the left of dou, expressing that the thoroughness as well as a high degree of carefulness is involved during the speaker’s reading process since in this case, punctuation marks are the most unlikely things present in reading materials for any reader to pay close attention to.

(34)

Thus, to account for the directional asymmetry of dou’s NPI licensing capacity, L. Chen and Pan (2020) propose that the left restrictor of dou as a universal quantifier at LF creates a downward-entailing context and is responsible for licensing Chinese NPIs. According to their LF–DE licensing account, dou is a universal quantifier and its LF is determined by two mapping rules, as made explicit by Pan (2006) and Jiang and Pan (2013).7 Specifically, if an NPI appears to dou’s left, such an associate gets mapped to dou’s restrictor (according to the topic-mapping rule as stated in Note 7), which is a downward-entailing context at LF, as dou functions as a universal quantifier like English every. The restrictor of a canonical universal quantifier, for example every, exemplifies a well-known case that can create a friendly environment for polarity-sensitive items, but its nuclear scope fails to accommodate polarity items (Fauconnier, 1975; Kadmon & Landman, 1993; Krifka, 1995; Ladusaw, 1980; Zwarts, 1998, among many others).

To be more specific, the examples in (35) illustrate dou’s monotonicity profiles. When dou associates with a plural subject yielding the universal distributive reading (as per the topic-mapping rule in Note 7), the restrictor student enrolled in this class supports the superset-to-subset inferences as shown in (35a,i) and (35a,ii), whereby {x De | x is a female/male student x enrolled in this class} {x De | x is a student x enrolled in this class}. However, the nuclear scope of (35a) is upward entailing since it fails to validate the superset-to-subset inference, that is, {x De | x went to a lecture} {x De | x went to a math dept. lecture} as shown in (35a, iii). Instead, it supports the subset-to-superset inference as in (35iv) whereby {x De | x went to a lecture} {x De | x participated in an activity} is clearly legitimate.

(35)

Entailment patterns of Chinese dou as detailed in (35) have convincingly demonstrated that only dou’s restrictor determined by dou’s mapping rules at LF can create a downward-entailing context that is responsible for licensing renhe-phrases and minimizers like yifenqian ‘one penny’, as dou is essentially a universal quantifier in itself, contra (Liao, 2010). Hence, the LF–DE analysis of L. Chen and Pan (2020) makes the prediction that if an NPI appears to dou’s right and gets mapped into dou’s nuclear scope (as per the focus-mapping rule in Note 7), an NPI will fail to be licensed by dou since the nuclear scope of dou is upward entailing. In a nutshell, the LF–DE analysis proposed in L. Chen and Pan (2020) aims to explain the limited distribution of NPIs in certain contexts like those involving English only and Mandarin dou, in which DE cannot be straightforwardly inferred based on their surface structure but can be detected once their LF is further determined by quantificational mapping rules. Based on Mandarin dou, L. Chen and Pan (2020) argue that an abstract level like LF with a sentence’s tripartite quantification clearly sorted out shows that NPIs are only licensed by the downward-entailing restrictor region of dou-sentences. L. Chen and Pan (2020) also compares their LF–DE account with the influential Strawson DE account as proposed in von Fintel (1999). The interested reader is referred to L. Chen and Pan (2020) for more details.

In what follows, some compositional details for licensing renhe/yi-fen qian ‘one penny’ phrases in dou-sentences that have not been elaborated by L. Chen and Pan (2020) will be provided to concretely show how their insights neatly explain why NPIs in (33b), (33d), (33f), (34b), and (34d) are ungrammatical/infelicitous. Notably, L. Chen and Pan (2020) reaffirms an important assumption from a series of studies following Ladusaw (1980) and Kadmon and Landman (1993) that NPIs including Chinese renhe-phrases and minimizers, for example, yi-fen qian ‘one-CL penny’ should appear within a downward-entailing context that leads to strengthened interpretations.

Specifically, renhe consists of ren ‘any’ as a modifying ignorance marker with a wh-item he ‘what’ that has a certain archaic feeling compared to shenme ‘what’. Here renhe is analyzed as a modifier that can further compose with a nominal phrase xuesheng ‘student’ of type <e,t> by Predicate Modification (Heim & Kratzer, 1998), yielding another more restricted set only true of students as illustrated in (36a). It is also assumed that a sentence containing renhe amounts to a weak existential claim and the modifier renhe makes salient any individual member of the set in question, also known as the domain-widening effect (Kadmon & Landman, 1993) as discussed in Section 2. The underlying association pattern of (36b) is shown in (36c). Since it associates with dou to its left, the renhe-phrase gets mapped into the restrictor of the tripartite structure as in (36c).

(36)

Note that the renhe-phrase in question amounts to a weak existential quantifier asserting that at least one of the individuals in (36a) has some property. Hence, without a scale reverser, namely a downward-entailing context, a sentence containing the renhe-phrase that expresses a trivialized proposition, namely, a true claim compatible with so many situations, as it can be entailed by so many alternative relevant propositions. It is hypothesized that this type of trivialized sentence should be avoided due to a violation of Gricean Maxims of Quantity (another way to implement this is to appeal to the G-triviality following Chierchia, 2013, as discussed in Section 2).

To be more precise, for instance, in a model where the denotation of renhe xuesheng ‘any student’ amounts to at least one student from the set of {a, b}, the derivation of the renhe sentence in (37ii) without a scale reverser like dou conveys a low volume of information as it is entailed by many other relevant propositions like a bought this book, b bought this book, and a & b bought this book (corresponding to domain/scalar alternatives in Chierchia, 2013) as discussed in Section 2. However, with the help of dou, whose restrictor delineates a downward-entailing context, qualified as a scale reverser, the sentence as a whole becomes the most informative one entailing other relevant propositions as shown in (37i), thus explaining why it is grammatical. Note that according to Chierchia (2013) as introduced in Section 2, it is assumed that polarity-sensitive items obligatorily activate subdomain alternatives that must be exhaustified by an exhaustifier akin to only inserted in the structure, and NPIs in plain episodic sentences lead to contradiction. Though the analysis laid out here does not strictly follow Chierchia’s contradiction-based explanation, what matters most is that the NPI renhe-phrase must appear to the left of dou because only in this case will it get mapped to the restrictor of dou at LF according to the topic comment rule, which can create a downward-entailing context for an NPI like renhe.

(37)

(38) illustrates how the minimizer yi-fen qian ‘one-CL penny’, which expresses the lowest/most unnoteworthy amount of money, gets licensed in a dou-sentence. (38a) is the denotation of yi-fen qian ‘one-CL penny’ and (38c) is the underlying association pattern for Example (38b). Just like the renhe-phrase, yi-fen qian ‘one-CL penny’ also amounts to a very weak trivialized claim (true statement compatible with so many situations). Hence, it is not acceptable without a scale reverser as it is very uninformative for information-communicating purposes. Note that the minimizer yi-fen qian ‘one-CL penny’ must be displaced to a pre-dou position because if it stays in the postverbal position, the background focus-mapping rule is triggered and the minimizer will get mapped to the upward-entailing nuclear scope of dou, which will incur ungrammaticality.

(38)

The reason why the lowest amount minimizer fails to yield a sensical interpretation without a scale reverser is shown in the step of (39ii) wherein so many alternative relevant amounts that could be spent will asymmetrically entail one penny, the lowest quantity. However, the minimizer meshes well with the rest of the sentence when it gets mapped to dou’s restrictor as shown in (39i), which is essentially a downward-entailing context. Hence, sentence (38b) as a whole expresses that the mother gave her daughter all the money that she had, which entails other propositions in the alternative set with a larger amount other than one penny. The alternative set is evaluated by the squiggle operator following Rooth (1985). Remarkably, in this case, the even reading surfaces and the minimizer introduce a set of alternatives which serves as dou’s quantificational domain.

(39)

To summarize, the case study discussed here introduced the LF–DE analysis proposed by L. Chen and Pan (2020) and filled in relevant compositional details showing how dou’s left region but not its right one can license Chinese NPIs like renhe-phrases and minimizers, suggesting that analyzing dou as a universal quantifier combined with the traditional DE-based account can explain the directional asymmetry of NPIs in dou-sentences. An account like this can also be easily extended to explain the same directional asymmetry of Chinese noninterrogative wh-items in dou-sentences because noninterrogative wh-items are also NPIs that must be licensed by a downward-entailing context. Given that only the left restrictor region of dou’s quantification is semantically downward entailing, it is not surprising noninterrogative wh-items only appear to the left of dou, which further supports the analysis that dou is a universal quantifier.

5. Conclusion and Outlook

Negative polarity phenomena have attracted much attention in formal semantics and pragmatics, which are connected to a variety of important semantic/pragmatic topics including Alternative Semantics, scalar implicatures, free-choice items, and epistemic indefinites. What the article has touched on is just the tip of the iceberg. Though limited in scope, the article has offered a concise overview of the theories of NPI licensing, summarized the limited distribution of three Chinese NPIs including renhe-phrases, wh-items, and minimizers, and finally introduced two recent developments to NPI licensing in Chinese.

There are two research questions worth flagging here and exploring in the future. First, as discussed in Section 3, the affective contexts licensing Chinese NPI renhe-phrases, wh-items, and minimizers are not totally the same, with minimizers being more constrained in their distribution. What could explain the unique behavior of Chinese minimizers? Why are these minimizers still deviant in modal contexts and in need of the likelihood reasoning for licensing? Second, as shown in Section 3, the affective contexts licensing Chinese NPIs do not overlap with those licensing English any (Lin, 1998 compares Chinese noninterrogative wh-items with English any). What could explain the divergent distributions of NPIs between English and Chinese? How could we account for the cross-linguistic variations on the limited distribution of NPIs in various languages with respect to a given set of affective contexts?

Discussion of the Literature

For the influential downward-entailingness account of NPI licensing, see Fauconnier (1975), Ladusaw (1980), and Zwarts (1998). For the alternative-and-exhaustification framework, see Krifka (1995) and Chierchia (2013). For an alternative view based on the notion of (non-)veridicality to derive polarity sensitivity, see Giannakidou (1998), whose work considers a wider range of affective contexts in which NPI any gets licensed, involving questions, intensional verbs, generics, imperatives, comparatives, superlatives, subjunctives/modals, and so on. For a detailed discussion on limited distributions of cross-linguistic polarity-sensitive items with respect to an exhaustive list of such semantic contexts, see Giannakidou (1998).

Acknowledgments

This work was supported by the National Social Science Fund of China (Grant No. 22&ZD295).

Further Reading

  • Chierchia, G. (2013). Logic in grammar: Polarity, free choice, and intervention. Oxford, UK: Oxford University Press.
  • Fauconnier, G. (1975). Pragmatic scales and logical structure. Linguistic Inquiry, 6(3), 353–375.
  • Giannakidou, A. (1998). Polarity sensitivity as (non)veridical dependency. Amsterdam, The Netherlands: Benjamins.
  • Krifka, M. (1995). The semantics and pragmatics of polarity items. Linguistic Analysis, 25, 209–257.
  • Ladusaw, W. A. (1980). Polarity sensitivity as inherent scope relations. New York, NY: Garland.
  • Zwarts, F. (1998). Three types of polarity. In F. Hamm & E. Hinrichs (Eds.), Plurality and quantification (pp. 177–238). Dordrecht, The Netherlands: Kluwer.

References

  • Barwise, J., & Cooper, R. (1981). Generalized quantifiers and natural language. Linguistics and Philosophy, 4(2), 159–219.
  • Chen, L., & Pan, H.-H. (2020). Jixiang lilunzhong yantui guanxide pinggu cengmian [Entailment evaluation in the theory of polarity sensitivity]. Zhongguo Yuwen [Studies of the Chinese Language], 2, 188–200.
  • Chen, Z. (2021). The non-uniformity of Chinese wh-indefinites through the lens of algebraic structure [Doctoral dissertation, City University of New York].
  • Cheng, L.-S. (1997). On the typology of wh-questions. London, UK: Taylor & Francis.
  • Cheng, L.-S., & Giannakidou, A. (2013). The non-uniformity of wh-indeterminates with free choice and polarity in Chinese. In K.-H. Gil, S. Harlow, & G. Tsoulas (Eds.), Strategies of quantification (pp. 123–154). Oxford, UK: Oxford University Press.
  • Chierchia, G. (2013). Logic in grammar: Polarity, free choice, and intervention. Oxford, UK: Oxford University Press.
  • Chierchia, G., & Liao, H.-C. (2015). Where do Chinese wh-items fit. In L. Alonso-Ovalle & P. Menéndez-Benito (Eds.), Epistemic indefinites: Exploring modality beyond the verbal domain (pp. 31–59). Oxford, UK: Oxford University Press.
  • Fauconnier, G. (1975). Pragmatic scales and logical structure. Linguistic Inquiry, 6(3), 353–375.
  • Feng, Y. L., & Pan, H. H. (2018). Zailun dou de yuyi: Cong qiongjinxing yu paitaxing tanqi [Revisiting the semantics of dou: From the perspectives of exhaustivity and exclusiveness]. Zhongguo Yuwen [Studies of the Chinese Language], 2, 177–194.
  • Gajewski, J. (2005). Neg-raising: Polarity and presupposition [Doctoral dissertation, Massachusetts Institute of Technology].
  • Giannakidou, A. (1998). Polarity sensitivity as (non)veridical dependency. Amsterdam, The Netherlands: Benjamins.
  • Giannakidou, A., & Cheng, L.-S. (2006). (In)definiteness, polarity, and the role of WH-morphology in free choice. Journal of Semantics, 23(2), 135–183.
  • Guerzoni, E. (2003). Why even ask? On the pragmatics of questions and the semantics of answers [PhD dissertation, Massachusetts Institute of Technology].
  • Hamblin, C. L. (1973). Questions in Montague English. Foundations of Language, 10, 41–53.
  • Heim, I., & Kratzer, A. (1998). Semantics in generative grammar. Oxford, UK: Blackwell.
  • Homer, V. (2020). Domains of polarity items. Journal of Semantics, 38, 1–48.
  • Homer, V. (2021). Negative polarity. In D. Gutzmann, L. Matthewson, C. Meier, H. Rullmann, & T. E. Zimmermann (Eds.), The Wiley Blackwell companion to semantics. USA: John Wiley & Sons.
  • Huang, C.-T. J. (1982). Logical relations in Chinese and the theory of grammar [Doctoral dissertation, Massachusetts Institute of Technology].
  • Israel, M. (2011). The grammar of polarity: Pragmatics, sensitivity and the logic of scales. Cambridge, UK: Cambridge University Press.
  • Jiang, J. Z., & Pan, H. H. (2013). Dou de yuyi fenhe ji jieshi guize [How many DOUs do we really need?]. Zhongguo Yuwen [Studies of the Chinese Language], 1, 38–50.
  • Kadmon, N., & Landman, F. (1993). Any. Linguistics and Philosophy, 16(4), 353–422.
  • Karttunen, L., & Peters, S. (1979). Conventional implicature. In C. K. Oh & D. A. Dinneen (Eds.), Syntax and semantics 11: Presupposition (pp. 1–56). New York, NY: Academic Press.
  • Klima, E. S. (1964). Negation in English. In J. A. Fodor & J. Katz (Eds.), The structure of language (pp. 246–323). Englewood Cliffs, NJ: Prentice Hall.
  • Kratzer, A., & Shimoyama, J. (2002). Indeterminate pronouns: The view from Japanese. In Y. Otsu (Ed.), The proceedings of the third Tokyo Conference on Psycholinguistics (TCP) (pp. 1–25). Tokyo: Hituzi Syobo.
  • Krifka, M. (1995). The semantics and pragmatics of polarity items. Linguistic Analysis, 25, 209–257.
  • Ladusaw, W. A. (1980). Polarity sensitivity as inherent scope relations. New York, NY: Garland.
  • Ladusaw, W. A. (1996). Negation and polarity items. In S. Lappin (Ed.), The handbook of contemporary semantic theory (pp. 321–341). Oxford, UK: Blackwell.
  • Lahiri, U. (1998). Focus and negative polarity in Hindi. Natural Language Semantics, 6(1), 57–125.
  • Lee, H. T. (1986). Studies on quantification in Chinese [Doctoral dissertation, University of California, Los Angeles].
  • Li, Y. H. (1992). Indefinite wh in mandarin Chinese. Journal of East Asian Linguistics, 1(2), 125–155.
  • Liao, H. C. (2010). Alternatives and exhaustification: Non-interrogative uses of Chinese WH-words [Doctoral dissertation, Harvard University].
  • Lin, J. W. (1996). Polarity licensing and wh-phrase quantification in Chinese [Doctoral dissertation, University of Massachusetts Amherst].
  • Lin, J. W. (1998). Distributivity in Chinese and its implications. Natural Language Semantics, 6(2), 201–243.
  • Linebarger, M. C. (1987). Negative polarity and grammatical representation. Linguistics and Philosophy, 10, 325–387.
  • Mok, S.-S., & Rose, R. (1997). The semantics and pragmatics of dou: A non-quantificational account. In L. J. Xu (Ed.), The referential properties of Chinese noun phrases (pp. 141–166). Paris: Ecole des Hautes Etudes en Sciences Sociales, Centre de recherches linguistiques sur l'Asie orientale.
  • Pan, H.-H. (2006). Jiaodian, sanfen jiegou yu hanyu “dou” de shiyi [Focus, tripartite structure, and Dou quantification]. Yufa Yanjiu Yu Tansuo [Grammatical Study and Research], 13, 163–184.
  • Rooth, M. (1985). Association with focus [Doctoral dissertation, University of Massachusetts Amherst].
  • von Fintel, K. (1999). NPI-licensing, Strawson-entailment, and context dependency. Journal of Semantics, 16(2), 97–148.
  • Xiang, M. (2008). Plurality, maximality and scalar inferences: A case study of Mandarin dou. Journal of East Asian Linguistics, 17(3), 227–245.
  • Xiang, Y.-M. (2020). Function alternations of the Mandarin particle dou: Distributor, free choice licensor, and even. Journal of Semantics, 37(2), 171–217.
  • Zwarts, F. (1998). Three types of polarity. In F. Hamm & E. Hinrichs (Eds.), Plurality and quantification (pp. 177–238). Dordrecht, The Netherlands: Kluwer.

Notes

  • 1. Strawson DE (shown below) contains an extra definedness condition added to the original DE definition in (8).

  • 2. In the (non)veridicality-based framework, the limited distribution of polarity items is sensitive to the semantic property, (non)veridicality, which can roughly be defined as follows.

    The given definition is oversimplified and does not refer to dynamic semantics. See Giannakidou (1998) for a dynamic implementation of nonveridicality as well as the relation with the varying strengths of DE functions discussed by Zwarts (1998). Due to space limitations, a comparison between DE- and (non)veridicality-based frameworks to polarity licensing must await another occasion.

  • 3. More precisely, renhe is analyzed as a free-choice item, see Cheng and Giannakidou (2013).

  • 5. This section will provide more examples that involve Chinese renhe-phrases and minimizers in dou-sentences as the syntax and semantics of noninterrogative wh-items have already been the focus of so many works since Huang (1982), Li (1992), and Lin (1996).

  • 6. Note that a comparison standard/class can appear to the right of dou if it does not contain an NPI as illustrated in (i), which clearly shows that it is possible for dou to precede the comparison class; in this case, dou yields the already reading.

  • 7. The two mapping rules are as follows: