2.1 Mandarin Dou: The Most Studied Quantificational Morpheme in Chinese

The universal quantifier ∀ is one of the two basic quantifiers in modern logic; thus, universal quantification is always the ‘protagonist’ on the stage of quantification. In Mandarin Chinese, the expressions related to universal quantification include adverbs dou and quan, which are rough counterparts of the English all, as well as ge, a rough counterpart of the English each, and nominal determiners like suoyou and quanbu, both of which can be translated into all in English, and mei ‘every.’ Among all these expressions, dou is the most studied. Also, since it many a time co-occurs with other quantificational expressions, it can be regarded as a thread connecting the studies of other Chinese quantifier expressions.

Lee’s (1986) pioneering work on quantification in Chinese treats dou as a universal quantifier that can induce a distributive interpretation, as in (6).

(6)

Lee’s observation is further supported by a number of studies (see Cheng, 1995; Jiang, 1998; Jiang and Pan, 2013; Liu, 1990; Pan, 2000, 2006; Wu, 1999; among others). Lin (1998a) observes that dou does not necessarily distribute down to atoms and formalizes dou as a generalized distributive operator exerting its universal force on covers in the sense of Schwarzschild (1996). His proposal is based on examples like (7).

(7)

In (7), the universal force of dou does not relate the VP property of being couples with each atomic individual in the noun phrase denotation; rather, the VP denotation is distributed over pairs of people, i.e., a cover generated by the plural set of tamen.

Aside from distributivity, dou can express exhaustivity and exclusiveness as in (8) and (9), respectively, and convey unexpectedness as in (8) and (10).

(8)

(9)

(10)

To countenance the facts above, one may assume that dou has different meanings in different sentences; as an alternative, one may adopt the unified approach which assigns dou a core meaning and derive the different uses from it. Jiang (1998), Pan (2006), and Jiang and Pan (2013) follow the second option. Assuming that dou is primarily a universal quantifier, they extend the universal quantificational approach to the various uses of dou by allowing it to quantify over different sorts of entities and by regulating the ways of quantificational mapping. For instance, in a scalar context like (3), dou operates on a set of ordered alternatives generated by the focused constituent marked by lian in the topic part of the sentence, i.e., the set of problems ordered by their levels of difficulty, and requires that all the alternatives satisfy the condition imposed by the rest of the sentence. The universal force endows the sentence with the interpretation: he works out the most difficult problem, let alone the less difficult ones.

Example (8) can be followed by a sentence as in (11), which seems to exclude the possibility of universal quantification over all the scalar alternatives.

(11)

To bypass the problem with universal quantification, one may simply assume that dou in these cases expresses the maximal level of unlikelihood or anti-expectation (Chen, 2008; Xiang, 2008). Yet, if one chooses to stick to the universal quantificational approach, one can then argue that the ordered alternatives to the surface proposition are conversational implicatures, which can be canceled or negated, and that dou actually exerts its universal force on implicatures as well as the surface proposition. In this sense, the analysis of dou puts a lot of weight on the level of pragmatics (see Jiang, 1998). To take the load off pragmatic factors, one may also argue that in scalar contexts like (8), dou universally quantifies over the worlds where the speaker’s belief for the actual world holds and relates each possible world with an alternative proposition to the surface proposition. In words, (8) then means that in all the possible worlds where the speaker’s belief for the actual world holds, some more likely alternative proposition(s) is (are) true, yet in the actual world, the surface proposition, which is the most unlikely, is true. In this sense, by expanding the range of entities on which it operates, dou still exerts its universal force at the level of semantics (see Feng, 2014).

2.2 Quantificational Mapping of Dou and Other Expressions

Example (9) is concerned with another interesting phenomenon of dou. In examples like (6), the quantificational mapping is regulated by syntactic structures, that is, the material left to dou is mapped to its domain, and the rest of the sentence to its scope. However, to yield the proper interpretation of (9), repeated below as (12), the mapping strategy cannot be the same as that for (6).

(12)

In (12), obviously, the singular subject ‘he’ cannot serve as the domain of dou. Instead, in such an example, the mapping of dou is constrained by the focus-background structure. Specifically, the focused constituent, i.e., mantou, is mapped to the scope, and the background, to the domain. Focus-background mapping can yield the exclusive reading over the focused constituent, meaning ‘For all the x that he eats, x has the property of steamed bums but not that of other things’ (see Pan, 2006, and Jiang & Pan, 2013, for more details).

In still other cases, the domain of dou is not contributed by any overt constituent in the sentence but substantiated with contextual information, as in (13).

(13)

According to Jiang (1998), in (13), the domain of dou is contributed by contextual information, that is, it quantifies over a set of salient situations where ‘he’ goes shopping for coats (see also Yuan (2007)). Beaver and Clark (2003) observe that the quantificational structure of the English always and only are different despite that they both encode universal quantification. Always operates on contextual information, while the mapping of only is regulated by the focus-background structure only. Taking the Chinese data of dou into consideration, we can further propose: the mapping strategies of quantificational expressions are diverse. In English, different quantifier words, i.e., always and only, follow their respective mapping rules, while in Mandarin Chinese, different strategies coincide in one single morpheme, i.e., dou, which is endowed with multiple uses or ‘meanings.’ In other words, the core semantics of dou is always universal quantification; however, since it can follow different mapping strategies and operate on domains of various sorts (e.g., individuals, ordered alternatives, worlds), dou-sentences are interpreted in different ways and take on various effects including distributivity, exhaustivity, exclusiveness, scalarity, etc. This is exactly what has been proposed in Pan (2006) and Jiang and Pan (2013), which assume a unified approach to the various uses of dou. Essentially, they argue that dou is basically a universal quantifier which can either operate on an ordered or unordered domain, and its quantificational structure is regulated by the following two rules:

P1: If the material left to dou can either serve as the quantificational domain or provide the quantificational domain through pragmatic inferences, it is mapped to the domain and the rest of the sentence, to the scope.

P2: If the sentence contains a contrastive focus in the comment, the focused constituent is mapped to the scope and the rest, to the restrictor, which induces an exclusiveness interpretation on the focus.

Only when dou quantifies over an ordered domain can the sentence in question display an effect of unexpectedness, and only when a dou-sentence is regulated by the P2 mapping rule can the sentence in question receive an exclusive interpretation on the focus, as illustrated in (12). The classification of the different uses of dou, as proposed by Jiang and Pan (2013), is shown in Table 1.

For Mandarin dou, its quantificational structure can be determined by grammar, contextual information, or focus-background. Furthermore, the mapping of the Cantonese universal verbal affix –saai is even more interesting. Lee and Pan (2008) and Lee (2012) observe that the mapping of –saai, similar to dou, is concerned with both syntax and focus structure; moreover, the data show that the relevant interpretation involves two independent processes: the selection process and the mapping process. The selection process applies first to select the material with which –saai is associated according to the grammatical function hierarchy (henthforth GFH) in (14).

(14)

Then the mapping process comes into play. If the selected constituent is in focus, it is mapped to the scope of –saai, and the rest of the sentence to the domain, following the background-focus mapping; otherwise, the selected constituent is mapped to the domain, and the rest of the sentence to the scope. For example, in both (15) and (16), the affix –saai is associated with the direct object rather than the plural subject, according to GFH. However, their semantic mappings are determined by different factors: by background-focus in (15), with the focused direct object mapped to the scope, and by the lexical semantics of -saai in (16), with the unfocused direct object mapped to the domain. Hence only (15) has an exclusive reading on the focused direct object.

(15)

(16)

The Cantonese data are particularly interesting in that they blur the borderline between D-quantification and A-quantification. Conventionally, D-quantificational structures pertain to syntax, while A-quantificational structures pertain to information structure. However, the interpretation of –saai involves both syntax and information structure, which are at work at their respective stages: syntax determines the item for association, while information structure or the lexical semantics of the quantifier in question decides the mapping of the selected item. It should be noted that the quantificational mapping of Mandarin dou also echoes with the two-process approach to quantification related to Cantonese –saai, that is, when dou associates with a contrastive focus in the topic, its mapping is determined by the topic-comment structure, and when dou associates with a contrastive focus in the comment, its mapping is regulated by the background-focus structure, leading to the exclusiveness interpretation on the focus in question, as discussed earlier in this subsection. Although dou is associated with a contrastive focus in both cases, the mappings to the quantification structure are different depending on whether the focus in question is in the topic or the comment, and only the one in the comment leads to an exclusive interpretation of the associated item in question, which argues strongly for the separation of the selection process—association—from the mapping process for quantification including dou.

2.3 Dou’s Interaction with Various Noun Phrase Types

The research on dou also sheds light on noun phrase interpretations. In Mandarin, certain noun phrases, e.g., mei+(yi ‘one’)+CL+N ‘every N’ and suoyou/quanbu+ (de)+N ‘all N’, often require the support of dou to participate in universal quantification, while other noun phrases, e.g., daduoshu+(de)+N ‘a large majority of N’, henduo+de/CL+N ‘many N’, do not have to but still can co-occur with dou. Still other noun phrases, e.g., henshao+N ‘few N’, bu+dao+san+CL+N ‘fewer than three N’, however, normally cannot be quantified by dou. The dividing line between dou-quantifiable and non–dou-quantifiable noun phrases is never very clear. Lin (1998a), Wu (1999), as well as Cheng (2009) all generalize that dou can associate with strong noun phrases but not weak ones (see Milsark, 1979, for strong-weak distinction of quantifier expressions), despite the fact that they base their generalization upon different assumptions. Yet, the strong–weak distinction of noun phrases only partly coincides with dou-quantifiablity, for instance henduo ‘many’ and percentage phrases like baifenzhi sanshi ‘30%’ are weak quantifier determiners, yet they can still co-occur with dou.

Liu (1997) examines an almost exhaustive list of Mandarin noun phrases and proposes to tease apart dou-quantifiable and non–dou-quantifiable noun phrases by the notion of G(eneric)-specificity,³ arguing that only G-specific noun phrases are dou-quantifiable. A G-specific noun phrase can be scope-independent as well as induce inverse scope on the object position, while a non–G-specific noun phrase can never be scope-independent. Liu’s observation and generalization concerning the scope behaviors of noun phrases are inspiring, yet it should be noted that G-specificity, a semantic notion defined upon scope data, is used to explain the scope relation between dou and other noun phrases, which may involve some looping logic.

Lee et al. (1989) and Xu (2007) propose that dou can only associate with a noun phrase expressing a large cardinality relativized to the speaker or the context. Jiang (2011) further relates the subjectively large cardinality with the concept of exhaustivity, as discussed in Fiengo (2007). Essentially, if a noun phrase is exhaustive, then its domain is determined relative to all the entities in the denotation of the common noun contained in it; for instance, the domain of mei+CL+N phrases or suoyou+de+N phrases takes into consideration all the (contextually salient) entities in the denotation of N. Similarly, if one judges that the plurality in the denotation of daduoshu+de+N has some property, one needs to evaluate the proportion of the plurality with respect to the totality of N so as to make sure that it accounts for the majority. Phrases like shaoshu+(de)+N ‘a minority of N’ are not exhaustive since one need not evaluate the pluralities these phrases pick out relative to all the entities in the noun denotation; rather, one can just determine the quantity according to their subjective judgment or experience. However, phrases like mou+xie+N ‘some certain Ns’ are not exhaustive, yet they can co-occur with dou, as in (17).

(17)

It can be concluded from the discussion above that a consensus is not reached as to dou’s compatibility with noun phrases. If some semantic property could be singled out to help draw a dividing line between dou-quantifiable and non–dou-quantifiable noun phrases, then our understanding of the morpheme as well as the various noun phrase types would surely be deepened.

The last thing that needs to be mentioned is the concern of double quantification. Dou is conventionally analyzed as a universal quantifier which denotes a binary relation between the domain and the scope. In the meantime, dou can co-occur with various types of noun phrases which are also analyzed as generalized quantifiers denoting binary relations (Barwise & Cooper, 1981). The co-occurrence of multiple quantificational phrases may cause the problem of double quantification, that is, the same nominal domain is operated over by two quantifier words. For instance, in (18), both dou and the nominal determiner mei are associated with the domain of students, and both seem to contribute to the universal reading of the sentence.

(18)

To solve the problem of double quantification, one may assume that the determiner is not truly a generalized quantifier but a one-place operator which sums together all the students and provides the domain for dou. Also, it can be argued that mei contributes the universal force in (18) and dou is trivialized into a distributive operator which matches students with events of liking linguistics. Sections 2.4 to 2.6 present a more detailed discussion on the interaction of dou with other quantificational nominals.

2.4 More on Universal Nominals

As mentioned in Section 2.3, phrases headed by mei/suoyou/quanbu normally need the support of dou to express universal quantification, which are referred to as universal nominals below. Lin (1998a) gives a uniform analysis of such phrases as contributing pluralities to the domain of dou. Following Beghelli and Stowell’s (1997) feature-checking approach, Lin argues that universal nominals carry a strong [dist] feature, which must be checked off at the Spec of a distributive phrase whose head is occupied by dou. Yet language facts show universal nominals do not necessarily require the support of dou, as shown in (19).

(19)

Pan (2005) proposes that mei is ambiguous between a universal quantifier and a sum operator, that is, in cases like (19), mei is a universal quantifier with its own full-fledged quantificational structure, while in (20), mei is ambiguous between a universal quantifier and a sum operator.

(20)

In (20), when mei is treated as a universal quantifier, dou is regarded as a matching function in the sense of Rothstein (1995) which pairs the smiling events with individual people. If dou is treated as a universal quantifier, then mei is reduced to a sum operator which collects all the atomic individuals to provide the quantificational domain for dou.

Luo (2011) also adopts an ambiguity approach; however, he argues that mei is primarily quantificational and shifted to a determiner (i.e., a sum operator) as a last resort to satisfy interpretability. The difference between mei and dou lies in the entities they can distribute, i.e., mei distributes over individuals, and dou, over events. As a result, when an indefinite or a reflexive is within the scope of mei, it can exert distributivity without dou. The type-shifting analysis is a bit problematic for cases like (21), though.

(21)

In (21), there are no individuals for the students to form a dependency with; hence, mei cannot be interpreted as a quantifier. Naturally, as a repairing strategy, mei could be shifted to a definite determiner that simply outputs a plurality (following Luo’s idea of type-shifting) and the sentence thus means that a certain plurality of students came; however, (21) is not acceptable without dou. Supposedly, after the type-shifting of mei, there is no type mismatch or violation of interpretability. It is thus unclear why type-shifting plus the explicit insertion of dou becomes the winning strategy, which is obviously less economical.

2.5 Other Analyses of Dou

Conventionally, dou is treated as a universal quantifier; yet, the semantics of dou does not command universal assent in the linguistic circle. In recent years, it has also been evaluated as a marker of sum formation (Huang, 1996; Yuan, 2005), maximization (Giannakidou & Cheng, 2006; Xiang, 2008), anti-expectation (Chen, 2008), high degrees (Xu, 2007, 2014), large amounts (Li, 2013), and focus-sensitivity plus additive/scalar presupposition (Liao, 2011). Among these proposals, some only tackle certain uses of dou and cannot be deemed as a unified account; some are still a few steps away from compositionality. Under this context, if the universal quantificational approach seeks to win general acceptance, the potential problems of these proposals and ways to integrate their evidence into the quantificational framework need to be studied. Whether the notions mentioned above are semantic primitives or derived effects from quantification would be an interesting issue for further research.

2.6 Other Quantificational Adverbials in Chinese

The other line of research on quantification in Chinese focuses on how to capture the difference and interaction among various quantificational adverbials (i.e., ge ‘each’, quan ‘all’, yi-CL-yi-CL ‘one by one’) (see Feng, 2014; Lee et al., 2009, 2013; Lin, 1998c; Soh, 1998; Tomioka & Tsai, 2005). Just like dou, the other quantificational adverbials can also induce distributivity, though with distinct semantic requirements on their domain and scope. Lee et al. (2009) generalize that the occurrence of the distributive ge needs to be licensed by a variable in its c-commanding domain, which can be instantiated as an indefinite numeral, a bare noun phrase, a reflexive or a wh-phrase. Tomioka and Tsai (2005) compare the difference between dou and quan and point out quan is not a standard universal quantifier; rather, it is a mere domain restrictor, while Lee et al. (2013) argue that quan is ambiguous between a universal quantifier and a domain restrictor. Intuitively, some adverbials, e.g., dou and quan, stress uniformity, while other adverbials, e.g., ge, convey diversity. It is still unclear how such differences can best be formally captured, and research in this direction will surely enrich the theoretical framework of quantification.

The most complicated yet fascinating puzzle concerning quantificational expressions in Chinese is the co-occurrence of multiple quantifiers, as illustrated in (22) and noted in Lee et al. (2009), which may cause a clash of semantic types.

(22)

Examples like (22) contain multiple quantifier expressions: determiner suoyou or mei-ge, and three adverbs quan, dou and ge. To derive the semantic interpretation of (22) successfully, some expressions need to be reduced to one-place operators. A possible solution is to treat the determiner as a maximality or sum operator, quan as a domain restrictor, dou as a universal quantifier, and ge as a matching function. The first two expressions locate the domain for dou, while ge, in the scope of dou, is sensitive to the event variable and matches an individual in the domain of dou with a distinct donating event (see Feng, 2014). Type-shifting of some adverbials into sum operators over the nominal domain can solve the type-logical clash, yet in the meantime it also increases the ambiguity of these morphemes.

2.7 Scope Interaction Among Different Quantifiers

A sentence may contain multiple quantifier expressions, and the scope interaction among them may yield multiple interpretations. Interestingly, Huang (1982) observes that the Chinese sentence (23) is not ambiguous, while its English counterpart is. To explain the difference, Huang proposes an Isomorphic Principle, which requires the scope relation of two quantifiers to correspond to their c-commanding relation at S-structure, while English (but not in Chinese) has a restructuring process that can switch the c-commanding relation.

(23)

Based on Huang (1982), Aoun and Li (1993) further provide a comprehensive structural account of the difference between English and Chinese with respect to scope interaction between quantifiers and wh-phrases in active and passive sentences and double object constructions. They argue that both languages follow a minimal-binding requirement and a scope principle and the interpretational difference is due to the lack of subject-raising in Chinese (but not in English). Liu (1997), however, investigates the scope behavior of almost every Chinese quantifier expressions and proposes G-specificity to distinguish ready inverse scope takers from poor inverse scope takers. Kuno et al. (1999) further propose that the availability of multiple interpretations is determined by weighing syntactic, semantic, and discourse factors. In this light, we may conclude that quantifier scope is an area where multiple linguistic subfields interact, and further study is definitely needed.

3.1 The Semantics of hen and Its Covert Counterpart

In Chinese, simple gradable adjectives cannot appear as the predicate in the sentence without the support of overt degree morphology, e.g., hen ‘very’; hence, (24) is infelicitous out of context without hen.

(24)

Kennedy (2005) thus claims that hen is an overt positive morpheme conveying context sensitivity. As a result, (24) can be understood as follows: the degree of Zhangsan’s height meets the standard comparison determined by the context and returned a positive value. Liu (2010) further argues that aside from the overt hen, there also exists a covert positive operator, which behaves like a polarity item licensed by a number of constructions, i.e., negation, contrastive focus, ma-particle questions, and the embedded clause of epistemic modality. For instance, in (25), the adjective gao ‘tall’ is contrasted with ai ‘short’ and thus need not be licensed by any overt positive morpheme.

(25)

Now the question becomes why Chinese gradable adjectives require overt positive morphology and allow the application of covert positive operator in certain contexts. Grano (2012) explains the distribution of hen and covert positive morpheme through the interaction between degree semantics and syntax in Chinese following Gu (2008). Essentially, he proposes that covert positive semantics is not projected in syntax and consequently cannot provide a syntactic constituent of the proper type to serve as the complement of the Tense node. Instead, overt positive hen is needed to induce the projection of a degree phrase, a functional projection that can be the complement of T. Other functional projections, e.g., FocusP in the case of (25), can also make the sentence observe the relevant syntactic constraint.

3.2 Chinese Comparatives

Unlike English, Chinese does not have markers like –er or more, which can combine with adjectives to express explicit comparison. Instead, explicit comparison is marked the morpheme bi, followed by a term that indicates a standard of comparison, then a gradable adjective, and finally a differential expression which may be optional, as exemplified in (26); alternatively, the gradable adjective can precede the standard of comparison if bi is absent, as in (27), which is referred to as the transitive comparative (see Erlewine, 2007; Grano & Kennedy, 2012; Lin, 2009; Liu, 1996, 2011; Xiang, 2005; among others).

(26)

(27)

Semantically, Erlewine (2007) treats bi as the head of a verbal functional phrase and defines its meaning using event-based semantics, i.e., bi establishes the comparison between two eventualities along a scale determined by the gradable predicate. Lin (2009) proposes to treat bi on a par with the English comparative morpheme –er, arguing that bi syntactically functions as the head of a degree phrase. Liu (2011), however, rejects the analysis of bi as a comparative operator based on syntactic evidence; for instance, unlike typical degree phrases, bi and the gradable predicate it modifies can be intervened by other syntactic constituents. In this light, Liu argues that the meaning of comparison is not contributed by bi but by a covert comparative morpheme or the covert geng ‘even more’, which is in complementary distribution with the differential phrase. Bi per se only introduces the standard of comparison. It can be concluded that the semantics of bi is intertwined with its syntactic status, both of which still attract heated discussion.

Transitive comparatives, as in (27), are subject to stricter well-formedness conditions. Xiang (2005) observes that the differential expression is obligatory and that the gradable adjective must establish a scale with a conventional measurement system. Hence, gradable predicates like meili ‘beautiful’ which are evaluated with respect to subjective measurement cannot appear in transitive comparatives. More importantly, Xiang concludes that the gradable predicates in transitive comparatives have to allow differential measure phrases, e.g., wu-limi ‘five centimeters’, as shown in (27). Concerning the obligatory occurrence of the differential measure phrase, Grano and Kennedy (2012) define a covert comparative measure function μ‎ which takes the gradable predicate as its argument.

Another type of comparatives is referred to as differential verbal comparative (Cheng, 1966; Li, 2015). As exemplified in (28), a differential verbal comparative involves bi, the standard of comparison, a non-gradable verb preceded by duo ‘much’ or shao ‘little’, and a differential phrase.

(28)

Such types of comparatives are interesting because: (i) they only allow duo and shao but not other gradable predicates; (ii) the occurrence of a differential phrase is obligatory; and (iii) the differential phrase can be a definite noun phrase. Concerning the defining properties of differential verbal comparatives, Li (2015) proposes a degreeless mapping analysis, according to which (28) can be paraphrased as: there exists a mapping function f which maps every entity in domain B, i.e., the domain of objects Lisi read, to domain A, i.e., the domain of objects Zhangsan read, and the difference between the two domains is the book entitled Jane Eyre. The mapping-based analysis fares better than the degree semantic analysis in that the difference between the objects read by the two individuals are compared at the taxonomic level rather than by pure amounts or quantities, i.e., the book of Jane Eyre is not equivalent to the book of Wuthering Heights, although the quantity is the same.

6.1 Does Chinese Have Tense?

Unlike English, Chinese does not have overt tense morphology, and hence some believe that Chinese only has an aspect system but not a tense system (see Chen & Chen, 2005); still other studies postulate a possibly empty inflectional node which holds tense features just like in English (see Huang, 1998; Li, 1990; Lin, 2011; Sybesma, 2007).

Lin (2006) explores various kinds of devices for tense interpretation in Chinese including temporal adverbials, modals, aspectual particles, etc., and concludes that it is not necessary to assume an inflectional node and covert tense features; rather, the information offered by default viewpoint aspect, tense-aspect particles, and pragmatic reasoning combine to determine the temporal interpretation of Chinese sentences. Lin also points out that if tense is simply construed as the ordering of temporal intervals, then some particles can be deemed as tense-markers in the semantic sense. For instance, the tense-aspect particle guo requires that the time of the development of an eventuality is included within the topic time, which precedes the evaluation time; thus, the particle can be used to express the past relation while conveying perfectiveness.

Wu (2007) also agrees that tense does not have a syntactic projection in Chinese, and he further proposes investigating the Chinese tense system at the level of discourse by adding discourse time into discourse models such as DRT. In particular, Wu points out that the endpoint of an activity assigned by rhetoric relations such as narration can be accessed by the following-up utterances; hence, temporal location in Chinese is not assigned at the sentential-level by some empty syntactic projection but determined when a coherent discourse is formed.

Although multiple studies propose that Chinese is a non-tensed language from the syntactic perspective. However, it is still argued that Chinese does have an implicit distinction of finteness and non-finiteness, and some particular criteria are proposed to test the existence of such a distinction. For instance, Huang (1998) argues that only finite clauses allow the occurrence of aspectual markers and modals as auxiliaries; Li (1990) proposes that the existence of finite clauses can be verified by the co-occurrence constraint on the temporal adverbial congqian ‘before’ and aspectual markers, since finiteness can block the association of the adverbial and the aspectual marker in the embedded clause; Li (1990) also uses the licensing of negative polarity items as a diagnostic, arguing that a negation marker can only license negative polarity items in the same finite clause with the licensor. Hu et al. (2001), however, gives a detailed review of all the available tests of finite/non-finite distinction and generalizes that none of the tests is utterly reliable. Also, since they cannot find any other evidence supporting such a distinction, they conclude that the assumed existence of finiteness in Chinese clauses is not based on solid language facts.

6.2 The Semantics of Aspect Markers in Chinese

The Chinese words le, guo, zhe, and zai are normally treated as aspect makers which express different aspectual meanings. Conventionally, le is deemed as a perfective marker, guo, an experiential marker, zhe, an imperfective marker, and zai, a progressive marker. The following discussion focuses on guo and shows how the meaning of guo is approached from different angles and how the analysis of aspect undergoes continuous refinements.

Like the Chinese le, guo can express the perfective meaning since it implies the culmination of an event; however, guo also conveys discontinuity of the event and the point of reference (Iljic, 1990). Therefore, (42) means that Lisi’s leg was broken prior to the speech time and is healed when the sentence is uttered.

(42)

Aside from discontinuity, guo is also thought to require repeatability (Li & Thompson, 1981), which explains why guo is incompatible with predicates like si ‘die’ which express a once-only event, as illustrated in (43). The requirement of repeatability is formally presented as a plurality condition on the domain of temporal quantification (de Swart, 1991; Smith, 1997).

(43)

However, Pan and Lee (2004) argue that repeatability and discontinuity are not strictly required by guo, as exemplified in (44) and (45).

(44)

(45)

In (44), the missing of the book continues at the speech time, and in (45), the predicate ‘young’ expresses an unrepeatable state, yet it can still co-occur licitly with guo. In this light, Pan and Lee (2004) pin down the meaning of guo as expressing a change of state instead of discontinuity or repeatability. When guo co-occurs with predicates like ‘young’, the change of state is possible since one can change from being young to not young; while with predicates like ‘dead’, the change of state is impossible, which thus excludes the use of guo. They further stress that the change-of-state property of guo pertains to pragmatics rather than semantics. As illustrated in (44), the book’s changed state from missing to being found is canceled by the following-up utterance. Ljungqvist (2007), however, takes a more radical pragmatic approach and proposes that the conventionally assumed aspect markers in Chinese convey the speaker’s attitude toward the information in the sentence. Hence, Ljungqvist chooses to tackle these particles in the framework of the Relevance Theory and emphasizes the procedural information they encode and their influence on the processing of discourse information.

Unlike the aforementioned studies, some still try to explain the properties of guo with purely semantic analyses. Lin (2007) points out the problems with the analysis of guo as expressing a change of state and still chooses to analyze guo from a purely semantic perspective. His evidence against Pan and Lee (2004) involves guo’s interaction with in(definite) object noun phrases, as in (46).

(46)

Lin claims that Pan and Lee’s analysis would judge (46) as ill-formed regardless of the definiteness of the object NP; however, the actual grammaticality of the sentence varies with the choice of object NPs. Based on Lin (2006), Lin (2007) defines guo as having two semantic components, the assertion part and the presupposition part. The assertion part requires that the development of an event e in the denotation of the predicate that guo combines with is wholly before the speech time. If the event e in the predicate denotation has a target state (Parsons, 1990), then the presupposition part requires the possible existence of another event e’ distinct from e which also satisfies the predicate denotation and holds in an interval containing the speech time. The assertion part excluded cases where guo combines with predicates like ‘dead’, since the state of being dead cannot be wholly before the speech time; rather, it will hold along the time span. The presupposition part can explain the variance of acceptability caused by (in)definiteness of objects. Indefinite NPs but not definite ones allow flexible value assignment and can make events in the VP denotation distinct from each other. Lastly, it should be noted that the existence of distinct events is a possibility and thus not necessarily true in the actual world, which yields a similar effect of implicatures as in Lee and Pan (2004).

Wu (2008), however, approaches the semantics of guo adopting the notion of terminablility. Guo exerts its force on the whole semantics of a situation with which it combines and requires that the situation in question be terminable, i.e., the situation can be either completed (when it is bounded) or terminated (when it is unbounded). From terminablity, Wu thus derives the relevant properties of guo, i.e., discontinuity and repeatability/recurrence.

To summarize, the study of tense and aspect in Mandarin Chinese is related to a wide range of concepts including situation types, event structures, possible worlds, quantification, etc. Tense and aspectual interpretation also involves the interaction between semantics and pragmatics, and the division of labor between the two modules is not yet a settled issue.⁴

7.1 The Properties of Long-Distance Reflexive ziji

The long-distance reflexive ziji is interesting in that its antecedent is normally the subject. For instance, in both (48) and (49), ziji is not anaphoric to the non-subject noun phrase Lisi.

(48)

(49)

Second, the long-distance ziji tends to be interpreted with a de se reading in intentional contexts (see Pan, 1995, 1997; Anand, 2006; Huang & Liu, 2001), as illustrated in (50).

(50)

It was generally held that the sentence in (50) can only fit in Scenario 1 but not in Scenario 2, that is, the reflexive ziji is anaphoric to the subject noun phrase and the sentence expresses a belief de se. However, Wang and Pan (2014, 2015) argue that the de se reading of long-distance ziji is not obligatory, especially in indirect speech contexts, though it is the preferred one.

The most puzzling issue pertaining to long-distance ziji is the blocking effect. Essentially, first- and second-person pronouns can prevent ziji from being anaphoric to a third-person noun phrase, but not vice versa as in (51) (Pan, 1995, 1997); the anaphoricity of long-distance ziji can be blocked by an expression which does not even occupy the subject position as in (52); the demonstrative use of a third-person NP can block the long-distance binding of ziji as in (53); when a plural NP and a singular NP appear in the same sentence, the blocking effect may vary, as shown in (54) (see Cole et al., 2005; Huang & Liu, 2001; Pan, 2001).

(51)

(52)

(53)

(54)

7.2 Theoretical Analyses of Long-Distance Bound ziji

This section reviews four major accounts which relate to the above-mentioned properties of long-distance ziji.

Concerning the fact that ziji tends to be anaphoric to the animate subject noun phrase of an attitude verb and the blocking effect induced by first and second person pronouns, Pan (1997) relates long-distance ziji with de se interpretations as well as the notion of self-ascription. To explain the anaphoricity of ziji, Pan proposes the condition of self-ascription and the condition of discourse prominence, both of which concert to determine the entity noun phrase ziji refers back to.

Huang and Liu (2001) approach the interpretation of ziji with the concept of logophoricity. Essentially, they argue that the long-distance ziji is the instantiation of a logophor in the sense of Sells (1987), who defines the antecedent of a logophor with the following three discourse roles: source (i.e., the individual who initiates the speech act), self (i.e., the individual whose mental state or attitude is described in the relevant proposition), and pivot (i.e., the temporal anchoring of the discourse). Huang and Liu (2001) further apply the theory of logophoricity to long-distance ziji and propose that ziji’s antecedent should assume at least one of the discourse roles mentioned above. However, their analysis is not watertight; for instance, according to the notion of logophoricity, ziji should be able to co-index with the source, i.e., the speaker, which is not the case in Chinese, however (see Pan, 1997). Furthermore, Wang and Pan (2012) observe that there are also cases where ziji is anaphoric to the hearer.

Anand (2006) notices that the de se interpretation of ziji and its strong tendency to refer back to the subject resembles very much the behavior of shifted-indexicals (Schlenker, 1999). Long-distance ziji can be interpreted as a shifted indexical because it is sensitive to the operator OP_auth which can shift the speaker and thus the context. The blocking effect of first/second-person pronouns is thus explained as the exclusion of such pronouns in the scope of OP_auth.Yet, it should be noted that Anand’s theory cannot countenance all the facts about long-distance ziji, for instance, in extensional contexts, ziji does not necessarily receive a de se reading; hence, one may further question whether long-distance ziji is a typical shifted indexical and what exactly its status is in extensional contexts.

Recently, based on the theory of empathy (Kuno, 1987; Kuno & Kaburaki, 1977; Oshima, 2006, 2007), Wang and Pan (2014, 2015) give a new account of Chinese long-distance ziji. Their main claim is that Chinese long-distance ziji is always empathic and in intensional contexts it is also logophoric, though there are exceptions in some particular cases, e.g., the indirect speech contexts. They show that (i) their proposal avoids the problem associated with Sells’s (1987) so-called unified notion of logophoricity, as used in Huang and Liu (2001); (ii) it also explains the special properties of Chinese long-distance ziji such as the different properties of ziji in intensional and extensional contexts, the possible non-de se reading of ziji in indirect speech, etc., and (iii) the non-de se reading of long-distance ziji leads to a trichotomy of attitude reports, i.e., de re, de se, and indirect de se, contra the traditional de se and de re distinction.