Copyright by Qianping Gu 2020

2020

The Dissertation Committee for Qianping Gu Certifies that this is the approved version of the following Dissertation:

Noun Incorporation and Resultative Verb Compounding

in Mandarin Chinese

Committee:

John Beavers, Supervisor

Stephen Wechsler

David Beaver

Johan Kamp

Patience Epps

Zhiguo Xie

Noun Incorporation and Resultative Verb Compounding

in Mandarin Chinese

Qianping Gu

Dissertation

Presented to the Faculty of the Graduate School of The University of Texas at Austin in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

The University of Texas at Austin

May 2020

Dedication

Dedicated to my grandma, Buying Zhou.

Acknowledgements

I thank my advisor John Beavers for giving me this great opportunity to study linguistics at UT Austin, and for giving me great advising throughout my time in Austin as a PhD student. It is wonderful and challenging to work with John. What amazes me most is his ability to organize pieces of linguistic facts in such a way that leads to a certain point which in turn explains those seemingly unrelated linguistic facts. This is an important lesson for me as a beginner in linguistics, or rather a beginner in science. I also appreciate his patience in advising me. The feedback from him is always full of details, which helps me learn and gives me a lot of ideas on how to make improvements and also how to think more clearly. What is also important is that I thank John for offering a wonderful course on syntax, which does not just give us a solid research grounding in syntax but also shows us how beautiful and elegant syntax can be. I thank him for cultivating a great interest in linguistics in me. I thank Steve Wechsler for giving me a lot of academic support, which has broadened my knowledge of different fields of linguistics. When I decided to work on noun incorporation years ago, I actually had little knowledge about this topic. Steve recommended some work on noun incorporation, which helped me to get started. Later when it turned out that the data I investigated is about hearer-new information, he recommended some work which investigates similar phenomena such as egophoricity. These, among others, have helped me a lot in studying the topic of my dissertation. I feel lucky to study at a department that has David Beaver on faculty. There are two important ideas in my dissertation that were inspired by two of David’s lectures. One is a lecture on definiteness where I learned about how familiarity might influence the choice between definite and indefinite noun phrases when the speaker is referring to an entity. I extended this idea to propositions when I was thinking of the difference between V-le sentences and S-le sentences in Chinese, and found that the

v two sentences parallel the contrast between definite and indefinite noun phrases in familiarity. The other one is a discussion on communicative effects, which gave me the idea that we might seek pragmatic evidence for a semantic study, and later this helped as a guiding idea when I found that S-le sentences give rise to non- compositional evaluative meanings when used in non-informative contexts. It is a great honor to have Hans Kamp on my committee. I thank him for helping me revise the literature review of presupposition in my dissertation in my dissertation, plus giving me a lot of critical feedback on how to frame and sharpen my discussion of the discourse function of S-le sentences. And more importantly, I feel extremely lucky that I could have a chance to learn Discourse Representation Semantics directly from him. It showed me as a beginner a different dimension of language – the discourse – and the class broadened my semantics knowledge. I also would like to thank Pattie Epps and Zhiguo Xie for their valuable comments on my dissertation. Pattie gave me some really nice suggestions about how to think about noun incorporation, and her ideas about comparative categories saved me from making too strong a conclusion. Zhiguo asked me some good questions about tense and aspect as well as different meanings of the particle le, which helped me develop the argument. It takes a village to cultivate a doctor. I thank the other professors whose classes I have attended, including David Quinto-Pozos, Scott Myers, Tony Woodbury, Megan Crowhurst, Danny Law, Rajka Smiljanic, and Katrin Erk. It is a wonderful experience to learn linguistics in different research areas. I thank them for offering these courses, which helped me to become a better linguist. I thank our graduate coordinator Ben Rapstine for making my life in Austin easier. Finally I thank my colleagues, especially Jaime Glez, Alex Rosenfeld, Cornelia Loos, Grace Neveu, Gladys Camacho, Ambrocio Gutierrez Lorenzo, Eric

Adell, and Zach Childers, and also my friend Ying Hao. I’m very glad to have met them in Austin, and learned from them about a lot of things. I thank their help and company as a friend in Austin, which I will always cherish at the bottom of my heart. vi

Abstract

Noun Incorporation and Resultative Verb Compounding in

Mandarin Chinese

Qianping Gu, Ph.D

The University of Texas at Austin, 2020

Supervisor: John Beavers

This dissertation investigates two types of compounding (in a broad sense) in Mandarin Chinese, namely, noun incorporation and resultative verb compounding. I argue that the object of the so-called S-le sentence structure is a case of noun incorporation. Syntactically, the object is constrained as it prefers a bare noun or a small noun phrase but rejects (indefinite) articles and quantifiers entirely. Semantically, the object has those properties that an incorporated noun typically has. It has narrow scope with respect to modality and quantifiers, a number neutral reading if it is a bare noun, and is discourse-opaque as it cannot serve as the antecedent of an anaphoric pronoun. The S-le sentence is also argued to be neutral regarding grammatical aspect as it allows a range of aspectual interpretations, depending on the context. The proposed semantic analysis for the S-le sentence is that it expresses informativeness, which is construed as a presupposition that the proposition is new to the hearer. The evidence for this analysis has three sources. One is the distribution that it is naturally used in a context where the proposition is new to the hearer, building on Liu (2002). The second piece of evidence is a Gricean effect when it is used in a context in which the proposition is not new to the hearer, generating an additional non-compositional evaluative meaning that resembles an implicature. The third piece of evidence is the significantly increased acceptability of S-le sentences vii with heavy NP-internal modifiers, which are usually syntactically dispreferred in S-le sentences, in an informative context when the speaker intends to provide new information. For resultative verb compounding, I investigate what semantics the two resultative morphemes, -wán and -diào, contribute to the aspectual meaning of the entire compound. I propose that -wán expresses termination and -diào culmination (or completion). Both yield telicity but through different avenues. Termination yields telicity by constraining the run time of event while culmination (or completion) sets the constraint on the patient.

viii

Table of Contents

List of Tables ...... xii Chapter 1: Introduction ...... 1 1.1 Introduction ...... 1 1.2 The S-le Sentence ...... 2 1.3 Resultative Verb Compounding ...... 5 1.4 Organization of the Dissertation ...... 9 Chapter 2: Testing for Noun Incorporation in Chinese ...... 11 2.1 Introduction ...... 11 2.2 An Overview of Studies of Noun Incorporation ...... 12 2.3 Formal Characteristics of Noun Incorporation ...... 14 2.3.1 Morpho-syntactic close relations between the incorporated noun and the verb ...... 14 2.3.2 Restrictions on modification of incorporated noun ...... 16 2.4 Testing for Noun Incorporation in Chinese Formally ...... 18 2.4.1 Morpho-syntactic close relations between the object and the verb 18 2.4.2 Restrictions on modification of the object ...... 20 2.4.2.1 Bare noun objects ...... 21 2.4.2.2 Objects with NP-internal modifiers ...... 22 2.4.2.3 Objects with numerals ...... 24 2.4.2.4 Objects with quantifiers and articles ...... 26 2.4.3 Summary ...... 28 2.5 Semantic Properties of Noun Incorporation ...... 29 2.5.1 Number neutrality ...... 30 2.5.2 Narrow scope ...... 31 2.5.3 Discourse opacity ...... 34 2.5.4 Well-establishedness effects ...... 38 2.6 Testing for Noun Incorporation Semantically in Chinese ...... 39 2.6.1 Number neutrality ...... 40 2.6.2 Narrow scope ...... 42 2.6.2.1 Negation ...... 42 2.6.2.2 Modality ...... 43 2.6.2.3 Universal quantifier ...... 45

2.6.3 Discourse opacity ...... 46 2.6.3.1 Zero pronouns ...... 47 2.6.3.2 Overt pronouns ...... 50 2.6.3.3 Referential capacity of the object of the S-le sentence ... 52 2.6.4 Well-establishedness effects ...... 55 2.6.4.1 Use of metonymy ...... 56 2.6.4.2 Formation of VPs with non-canonical objects in contexts ...... 58 2.7 Summary ...... 61 2.8 Incorporation and Compounding in Chinese ...... 64 2.8.1 Verbal compounds formed with a dummy object ...... 65 2.8.2 Verbal compounds formed with a modifier ...... 69 2.8.3 Summary ...... 71 2.9 Conclusion ...... 72 Chapter 3: Discourse Uses of S-le Sentences ...... 73 3.1 Introduction ...... 73 3.2 Another Two Cases of Sentence Final le ...... 76 3.2.1 The inchoative le ...... 77 3.2.2 The phase-complement le ...... 78 3.3 S-le Sentences: Hearer-new Information ...... 80 3.3.1 Hearer-new information ...... 85 3.3.2 When the presupposition fails ...... 90 3.3.2.1 Wait a minute… ...... 91 3.3.2.2 Gricean effects ...... 94 3.3.3 An effect of using S-le sentences in the informative context ...... 98 3.3.3.1 The S-le sentences with NP-internal modifiers ...... 99 3.3.3.2 The S-le sentences with NP-internal numerals and the quantifier every ...... 102 3.3.4 Two common implications: Change of state and contrary to expectation ...... 105 3.3.5 Attitudinal uses of S-le sentences ...... 115 3.3.6 Summary ...... 117 3.4 The S-le Sentence: Free of Grammatical Aspect ...... 118 3.4.1 Prior studies ...... 119

3.4.2 The S-le sentence: Free of grammatical aspect ...... 130 3.5 Informative Contexts with Different Focuses and Givenness ...... 134 3.5.1 Contexts for S-le and the V-le sentences ...... 135 3.5.2 Informative contexts where S-le and V-le sentences are excluded ...... 137 3.5.2.1 Object Preposing structure: The hearer has prior knowledge of the event ...... 137 3.5.2.2 Ba-sentence: The object is given ...... 139 3.5.3 The S-le sentence: The only structure that expresses informativeness ...... 142 3.5.4 Summary ...... 143 3.6 Chapter Summary ...... 144 Chapter 4: Telicity in Mandarin Chinese ...... 146 4.1 Introduction ...... 146 4.2 Termination and Culmination ...... 153 4.3 A Diagnostic Test for Telicity ...... 160 4.4 Previous Studies of Telicity ...... 165 4.4.1 Krifka’s model on aspectual composition ...... 166 4.4.2 Zucchi & White’s (2001) maximal participants ...... 170 4.5 Non-culminating accomplishment predicates ...... 176 4.6 The Analysis ...... 179 4.6.1 Analyzing -wán ...... 184 4.6.2 Analyzing -diào ...... 190 4.7 Conclusion ...... 194 Chapter 5: Conclusion ...... 197 References ...... 199

List of Tables

Table 1: Morpho-syntactic properties of noun incorporation in Chinese, West

Greenlandic, Hungarian, Hindi, and Niuean...... 62 Table 2: Semantic properties of noun incorporation in Chinese, West Greenlandic,

Hungarian, Hindi, and Niuean...... 63 Table 3: The four sentence structures and their presuppositions: the S-le, the V-le,

the Object Preposing, and the ba-sentence...... 144

xii

Chapter 1: Introduction

1.1 INTRODUCTION

This dissertation explores two types of compounding (in a broad sense) in Mandarin Chinese (Chinese henceforth). Chinese is typologically classified as an analytic language that has almost no inflection for grammatical relationships (Huang

2015). But it has a variety of ways to derive words by combining morphemes of various categories (Lu 1964). Some types of compounding derive words such that that the meanings of the derived words are related somewhat compositionally to the meanings of the components, and there are also many types of compounding where the meanings of the derived words are rather idiosyncratic. The two types of compounding I explore here, namely, noun incorporation and resultative verb compounding, are two systematic cases where the semantics of the compound is predictable. Noun incorporation is a typologically common phenomenon (e.g., Mithun 1984, van Geenhoven 1998, Massam 2001), but it might not be obvious at first sight that Chinese has noun incorporation.1 This is because the syntactic and semantic status of a bare noun in Chinese is arguably ambiguous and it is hard to tell whether a bare noun in a given sentence is incorporated or a regular argument with an indefinite (or definite) interpretation. Bare nouns in Chinese do not obligatorily require a determiner or inflection in order to be an argument (Chierchia 1998). They can be used in different grammatical positions of a sentence of different sentence structures and allow a definite interpretation even though it is a bare noun. In this study, however, I provide evidence that the object of a sentence structure called the S-le sentence is a special type of indefiniteness, and I argue that it is a case of noun incorporation.

1 It is notable that in the more recent literature, Huang (2015: 5-6) has mentioned that Chinese also exhibits noun incorporation specifically Pseudo Noun Incorporation (Massam 2001). However, the topic of that paper is not incorporation, and he did not provide any evidence for incorporation in Chinese. So the current study is the first one that addresses incorporation in Chinese. 1

1.2 THE S-LE SENTENCE

The S-le sentence is one of the most studied structures in the literature of Chinese linguistics (e.g. Wang 1959, Chao 1968, Spanos 1979, Li & Thompson 1981, Li et al. 1982, Zhu 1982, Sybesma 1999, Liu 2002, Soh 2009). A prominent feature of this structure is that the sentence ends with a particle le, which is why it is called the S-le sentence. An example is presented in (1). (Note that the hyphen in (1) is artificial, as I argue that the object is incorporated in this case):2

(1) tā kàn-shū le. (S-le sentence)

he read-book LE ‘He read a book.’

What is notable about the morpho-syntactic properties of S-le sentences is that the object of such sentences is constrained, and further modification of the object is usually dispreferred, which, however, has not received much attention in prior studies. For example, the sentence would be unacceptable if the object is modified by an indefinite article.

(2) *tā kàn-běn-shū le. (S-le sentence)

he read-IND-book LE ‘He read a book.’

The ungrammaticality of (2) suggests that the object of S-le sentences is syntactically

2 Note that the provided translation is simplified in four respects. First, the bare noun shū ‘book’ in the three sentences is number neutral, that is, it can be interpreted as either singular or plural. So the singular form of the noun in the provided translation is only for the purpose of simplicity. Second, the bare noun shū ‘book’ may refer to a book that is either identifiable or not to the addressee. Mandarin does not have a definite article. So the indefinite article used in the translation is also for simplicity. Third, Chinese is a tenseless language so I assume a past tense for the sentence in the translation. Fourth, the S-le sentence has no inherent aspect and its aspectual interpretation is determined by the context (see Section 3.4). So the provided translation here represents only one of the possible aspectual interpretations. 2 constrained, providing evidence for noun incorporation in (1a). There is a similar sentence structure that is also commonly used and sometimes interchangeable with S-le sentence for talking about the same event but which does not show such modificational restrictions. This sentence structure is called the V-le sentence, which has the particle le realized as a verbal suffix marking for perfective aspect (e.g., Chao 1968, Li & Thompson 1981, Smith 1997, Bisang and

Sonaiya 1997, Huang et al. 2009, Soh 2014). The corresponding V-le sentence for (1) is presented in (3).

(3) tā kàn-le shū. (V-le sentence)

he read-LE book ‘He read a book.’

The two sentences, (1) and (3), have quite similar meanings, as indicated by the translation. Both of them can be used to talk about an event of reading a book that happened in the past, and there is no apparent difference regarding whether one of them is associated with any other kind of meanings that does not exist for the other one. The interpretation of the bare noun shū ‘book’ is also the same in both sentences. In either case, it allows an indefinite or definite interpretation, and singular or plural interpretation, depending on the context. However, the readily V-le sentence allows all sorts of modifiers that are unpreferable or prohibited for S-le sentences. For example, the object of a V-le sentence allows an indefinite article.

(4) tā kàn-le běn shū. (V-le sentence)

he read-LE IND book ‘He read a book.’

The contrast in grammaticality between the two sentences in (2) and (4) shows that the two sentence structures are syntactically distinct and the object of S-le sentences has an unusual syntactic status.

The discussion of semantics of the S-le sentence in the literature has largely focused on the discourse meaning of the sentence (e.g., Chao 1968, Zhu 1982, Li et al. 1982, Chu 1998, Liu 2002, Soh 2014, Wang 2018), which is different from the discussion of semantics of noun incorporation (e.g., Mithun 1984, van Geenhoven 1998, Farkas & de Swart 2003, Chung & Ladusaw 2004, Dayal 2011). The former views the S-le sentence as a whole without considering the meaning of its components, in particular, the meaning of the incorporated object. The S-le sentence structure is associated with a particular meaning that makes it distinct from other sentence structures such as the V-le sentence. Work on incorporation largely focuses on the semantic properties of the incorporated object such as its scope relation and referentiality. As semantic evidence for incorporation, I present data in Chapter 2 to show that the object of the S-le sentence also has semantic properties that are normally associated with incorporated nouns. But the semantic analysis I propose is for the whole S-le sentence rather than the incorporated object, which captures its use in context. As for the semantics of the incorporated bare noun, I simply follow previous studies such as van Geenhoven (1998) that the bare noun denotes a property. A common description of the S-le sentence in a grammar book (e.g., Chao 1968, Zhu 1982, Chu 1998) is that it describes a new situation. Liu (2002) further pointed out that many of the S-le sentences describe situations that are specifically new to the hearer. For example, consider the pair of discourses in (5), which only differ in the first sentence.

(5) a. tā zǎoshàng sǎo-xuě le. zhīhòu qù-le xuéxiào.

he morning sweep-snow LE afterwards go-LE school b. tā zǎoshàng sǎo-le xuě, zhīhòu qù-le xuéxiào.

he morning sweep-LE snow afterwards go-LE school

‘He swept the snow in the morning. Afterwards, he went to school.’

The two discourses convey the same information, as indicated by the translation, but

(5a), of which the first sentence is an S-le sentence, is natural only in a context in which the hearer does not know of the event of sweeping the snow. Conversely, the discourse (5b), of which the first sentence is a V-le sentence, is acceptable regardless of whether the hearer knows of the event of sweeping the snow or not. In prior works on S-le sentences, the usual assumption is that the hearer-new semantic property of S-le sentences only applies in some uses and not necessarily in others (e.g. Sybesma 1999, Liu 2002), and also that in some prior studies the exact constraint has instead been assumed to be narrower, such as Soh (2009) in which the semantic properties taken into consideration are correcting incorrect beliefs on the part of the hearer, and introducing that a change-of-state had occurred. My proposal is that all of these prior proposals are subsumable under a more general assumption of hearer-newness, which in various contexts gives rise to these more specific uses of S-le sentences, and which is also extendable even to cases that previous authors had suggested did not have anything to do with hearer-newness.

1.3 RESULTATIVE VERB COMPOUNDING

The other type of compounding investigated here is what is called resultative verb compounding in the literature (Li & Thompson 1981, Tai 1984, Smith 1990, 1997). It is a kind of compounding where a verb that typically denotes an activity (e.g., chī ‘eat’, xiě ‘write’, dǎ ‘hit’) is combined with an element that spells out the result of the activity. The resultative element can be an adjective (e.g., gānjìng ‘clean’, gān

‘dry’, hóng ‘red’), an inchoative verb (e.g., pò ‘break’, kāi ‘open’, shàng ‘ascend, up’, dào ‘arrive’), and a morpheme that indicates the second phase of an event (e.g., -wán ‘finish, terminate’, -diào ‘off’, -zháo ‘be on target’). Two examples are presented as follows.

(6) a. tā tī-pò-le mén.

he kick-break-LE door ‘He kick-broke the door.’ 5

b. tā cā-gānjìng-le zhuōzi.

he wipe-clean-LE table ‘He wipe-cleaned the table.’

In the two examples, the verbs tī ‘kick’ and cā ‘wipe’ of the compounds describe the activities of kicking and wiping while pò ‘break’ and gānjìng ‘clean’ spell out the results of the activities.

The particular aspect of resultative verb compounding that I investigate here is the semantics that the resultative element contributes to the aspectual interpretation of the entire compound. The specific aspectual meaning I deal with is telicity. I focus on telicization of predicates by different resultative morphemes and investigate the specific semantics of those different resultative morphemes that can account for their interactions with the base predicates that give rise to a telic interpretation while the base predicate may or may not be a telic predicate. Resultative morphemes telicize any predicates that they compound with, as resultative compounds pass the telicity test. For example, an activity predicate chī píngguǒ ‘eat apples’, which is atelic as it is compatible with a for-adverbial as shown in (7a), becomes a telic predicate when it is compounded with the resultative morpheme -wán as the compound is incompatible with the for-adverbial as shown in (7b).

(7) a. Mǎlì chī-le wǔ fēnzhōng píngguǒ.

Mary eat-LE five minute apple ‘Mary ate apples for five minutes.’ b. *Mǎlì chī-wán-le wǔ fēnzhōng píngguǒ.

Mary eat-TERM-LE five minute apple

The resultative morpheme can also compound with an accomplishment such as chī yī-ge píngguǒ ‘eat an apple’, and gives rise to a telic predicate. See Chapter 4 for the

6 telicity test applied to accomplishments. A specific question I investigate is what the resultative morpheme -wán really means. Prior studies have found that Chinese is one of those languages that does not have a culmination entailment for accomplishments like eat an apple which entails culmination in other languages such as English (e.g. Filip 2017). For example, when the accomplishment is used in a V-le sentence, the culmination reading can be denied, which results in no contradiction.

(8) Mǎlì chī-le yī-gè píngguǒ, hái shèng liǎng-kǒu, bèi wǒ chī-le.

Mary eat-LE one-CL apple still left two-bite PASS I eat-LE ‘Mary ate one apple. There was some left over and I ate it.’

Prior studies have claimed that in order to obtain a culmination reading, a resultative morpheme such as -wán is needed, which forms a resultative verbal compound with the verb (e.g., Tai 1984, Smith 1990, 1997, Sybesma 1997, 1999, Soh 2009). In such cases, the discourse would be contradictory if culmination is denied by the following sentence.

(9) *Mǎlì chī-wán-le yī-ge píngguǒ, hái shèng liǎng-kǒu, bèi wǒ chī-le.

Mary eat-TERM-LE one-CL apple still left two-bite PASS I eat-LE ‘*Mary finished eating one apple. There was some left over and I ate it.’

Thus it seems that the morpheme -wán expresses culmination or completion. In fact some studies assumed exactly this analysis for -wán (e.g., Soh 2009). But I argue that this is not what -wán really means. If -wán really expresses completion, one of the predictions would be that it should select a verb that selects a patient so that it denotes that some result obtains for the patient. But the fact is that -wán does not have such a selectional restriction for the verb it compounds with. Consider the following example where it compounds with an activity yóu(-yǒng) ‘swim’.

(10) tā yóu-wán-le yǒng jiù zǒu-le.

he swimV-TERM-LE swimN then leave-LE ‘He left when he finished swimming.’

I propose that -wán instead expresses a meaning of termination. It introduces the run time associated with the denoted event and expresses the point at which the event stops. This semantic meaning explains why -wán is compatible with a variety of event predicates and entails termination. The culmination entailment for the case that involves non-culmination accomplishments as the base predicate as illustrated by (8) and (9) can be explained if we assume there is an interaction between termination and non-culmination accomplishments that give rise to an interpretation of culmination. However, my purpose of the current study is not to analyze the notion termination as a formal object; instead, I suggest that we may analyze terminated predicates as subtype of telic predicates. In particular, while telicity is often assumed to derive from some result state obtaining for the patient in a way that usually relies on mereological properties of the patient and/or some scale along which the patient changes state (see e.g. Beavers 2012), I will claim that in the case of terminated predicates telicity arises because the time course of the event is constrained to have particular boundaries regardless of whether there is a patient that undergoes a change. Other than -wán, I also study another common resultative morpheme used for a consumption event, which is -diào. The morpheme -diào expresses culmination or completion with respect to the patient directly. It also entails culmination. Consider the following example.

(11) *Mǎlì chī-diào-le yī-ge píngguǒ, hái shèng liǎng-kǒu, bèi wǒ chī-le.

Mary eat-CULM-LE one-CL apple still left two-bite PASS I eat-LE ‘*Mary ate up one apple. There was some left over and I ate it.’

The morpheme -diào presupposes a specific amount of patient involved in the event and it specifies the result for the event as that the event achieves culmination or 8 completion with respect to the patient. The compound with -diào is also telic and it represents the normal case of obtaining telicity through the patient. I propose an analysis of -wán and -diào based on the model developed by Krifka (1989, 1992, 1998) and Zucchi & White (2001), which provides a way of analyzing obtaining telicity through time and patient.

1.4 ORGANIZATION OF THE DISSERTATION

The remaining part of the dissertation is organized in the following way. Chapter 2 focuses on presenting data to show that S-le sentences exhibit the formal characteristics and semantic properties normally associated with noun incorporation, providing evidence for noun incorporation in S-le sentences. Syntactically, the object of S-le sentences prefers small noun phrases and prohibits quantifiers entirely, whereas such restrictions do not apply to V-le sentences.

Semantically, the object of the S-le sentence overall also exhibits the four semantic properties associated with noun incorporation that are established in the literature, namely, narrow scope (relative to modality, and universal quantifiers), discourse opacity, number neutrality, and well-establishedness effects (i.e. noun incorporation expressions are more likely to be about culturally common activities, but it varies from language to language). In addition, I also present Chinese data of typical cases for compounding (in a narrow sense) and show that this is a distinct phenomenon from noun incorporation.

Chapter 3 provides evidence for the semantic analysis of informativeness of S-le sentences. In this chapter, I focus on discourse contexts that S-le sentences occur in, and the ones they do not occur in, with an attempt to give a relatively comprehensive accounting of the uses of S-le sentences. There are many distinct proposals in the literature about what S-le sentences mean. However, drawing on and expanding Liu (2002), Sybesma (1999), and Soh (2009), I suggest that they have one core use of introducing hearer-new information, which subsumes and accounts for a

9 wide range of previously proposed functions. My analysis validates these authors’ intuition about S-le sentences and I also show it extends beyond cases they were concerned with. However, S-le sentences may also be used in non- informative contexts where the sentence is not used with an intention to provide new information but to convey certain evaluative meanings. This is a distinct use of S-le sentences and can be reconciled with the informative use if we assume that informativity is the core function and the additional evaluative meaning in the non- informative context fills in the communicative function. In addition, I show that S-le sentences are blocked in two types of informative contexts where other more specific sentence structures are used. Chapter 4 presents a study of telicity in Chinese. By studying the two resultative morphemes -wán and -diào that are normally used to telicize predicates, I show that they represent two different avenues to telicity: the former derives telicity by setting constraints on time and the latter on patient. The analysis is cast in the model developed by Krifka (1989, 1992, 1998) and Zucchi & White (2001), which consists of structures of event, patient, and time that are homomorphically related. This model shows that a constraint on either patient or time can generate a bounded event through homomorphism. By using this model, some facts associated with expressions containing -wán and -diào can be explained. Chapter 5 provides a summary of the major findings and concludes the study.

Chapter 2: Testing for Noun Incorporation in Chinese

2.1 INTRODUCTION

The aim of this chapter is to show that the object of S-le sentences also exhibits many of the syntactic and semantic properties that we have found in various cross-linguistic cases of noun incorporation, and based on these similarities, I suggest that we might consider S-le sentences to exemplify the cross-linguistic comparative category of noun incorporation. Although S-le sentences show some language particular features in how the properties I discuss below manifest, as is the case in all languages, given the overall set of properties discussed below it is reasonable to view the case of S-le sentences as noun incorporation. The typical morpho-syntactic insight regarding a noun incorporation construction is that the object shows a closer relationship to the verb than the object in a corresponding non-incorporation construction in which the object is more independent. Furthermore, incorporated nouns show more restrictions in morphological markings and nominal modifications when compared with unincorporated nouns, though languages vary with respect to the extent to which such restrictions apply. This is also true in the case of S-le sentences. S-le sentences show a morpho-syntactic close relationship between the object and the verb at the surface level and place severe restrictions on modificational possibilities of the object which do not apply to V-le sentences. Semantically, noun incorporation structures usually have some more constrained and less fully referential interpretation than the corresponding object in a non-incorporation structure. The most commonly accepted semantic property of noun incorporation is that the object has a narrow scope relative to other operators, plus other more cross-linguistically variable properties. I show below that the object of

S-le sentences also has many of those canonical semantic properties while the corresponding V-le sentences tend to lack those properties, suggesting that it belongs to this category even as it shows language particular features. 11

Below I first provide an overview of studies of noun incorporation. Then the formal characteristics and semantic properties of incorporation discussed in the literature are reviewed, each of which is followed by a presentation of the relevant Chinese data which show to what extent those properties apply to S-le sentences. Then I provide a summary of those properties and compare the extents to which they apply to different noun incorporation languages. An overall comparison with other noun incorporation languages supports the claim that the case of S-le sentences is plausibly noun incorporation. Finally, I provide some Chinese data to show that noun incorporation is distinct from typical cases of compounding.

2.2 AN OVERVIEW OF STUDIES OF NOUN INCORPORATION

The phenomenon of noun incorporation was introduced into linguistics by linguists who worked on Native American languages about a century ago (e.g.

Kroeber 1909, Sapir 1911). It typically refers to a construction where a verb and one of its nominal arguments, typically the direct object, form a morphologically complex verb. The incorporated noun is usually a noun stem without a determiner or case marking. Meanwhile, in a noun incorporation language there also exists a counterpart where the noun phrase takes a regular syntactic argument position. An example from Mohawk is presented below. Sentence (1a) illustrates a standard transitive verb construction with a regular object and (1b) is the corresponding sentence with noun incorporation.

(1) a. Wa’-k-hninu-’ ne ka-nakt-a’. [Mohawk, Baker 1996]

FACT-1sS-buy-PUNC NE Ns-bed-NSF ‘I bought the/a bed.’ b. Wa’-ke-nakt-a-hninu-’.

FACT-1sS-bed-∅-buy-PUNC ‘I bought the/a bed.’

Since the 1980s, a large number of unrelated languages have been reported to exhibit the phenomenon of noun incorporation, including West Greenlandic (van Geenhoven 1998), Hungarian (Farkas & de Swart 2003), Hindi (Dayal 2011), Turkish (Öztürk 2009), Chamorro (Chung & Ladusaw 2004), Mayan languages (Mithun 1984), Uto-Aztecan languages (Haugen 2008), among others. Earlier theoretical studies of noun incorporation in the 1980s mainly focused on the morpho-syntax of this construction (e.g. Sadock 1980, Baker 1988, Rosen 1989), while the semantics largely remained descriptive (e.g. Mithun 1984). The term noun incorporation is thus sometimes understood in its narrow sense as referring to a morpho-syntactic process of incorporating a noun stem to form a complex verbal compound. In the 1990s, groundbreaking studies on the formal semantic properties of noun incorporation were conducted (e.g. Bittner 1994, van Geenhoven 1998), and since then a set of semantic properties have been proposed in the literature to characterize noun incorporation (see Borik & Gehrke 2015). More recently, due to these semantic characterizations, some linguists relate noun incorporation to other linguistic phenomena that show similar semantic properties but do not involve any compounding, including bare nouns in Romance languages (Espinal & McNally 2011), weak definiteness (Borik & Gehrke 2015), and imperfective aspect in Russian (Mueller-Reichau 2015).

Meanwhile, it is noticeable that more recent studies have started to use the term Pseudo Noun Incorporation or Pseudo Incorporation to refer to a structure which shares semantic properties with noun incorporation but the involved nominal, morpho-syntactically incorporated or not, is a larger constituent than a noun stem. For example, the nominal may take NP-internal modifier or can be coordinated or retain case marking. The word pseudo is from Massam (2001), who used the term pseudo noun incorporation to refer to such an expanded NP in order to distinguish it from incorporation of a noun stem in the phenomenon of noun incorporation. Borik & Gehrke (2015) noted that these phenomena, including noun incorporation, pseudo noun incorporation, and objects with weak referentiality, share many semantic 13 properties in common and sometimes it is hard to draw a clear line between them. In the current study, I use noun incorporation or incorporation to refer to the broader phenomenon.

2.3 FORMAL CHARACTERISTICS OF NOUN INCORPORATION

Morpho-syntactically, incorporated nouns show close relationship with the incorporating verbs, and are restricted in morphological markings and nominal modifications. Data from other languages are reviewed below. Note that languages show diversity in showing those morpho-syntactic properties, especially in the extent to which they allow morphological markings and nominal modifications. Nevertheless, they all show such restrictions for the incorporation constructions more or less while the unincorporated expressions do not have such restrictions.

2.3.1 Morpho-syntactic close relations between the incorporated noun and the verb

Object nouns in incorporation constructions and the corresponding non- incorporation construction have different morpho-syntactic statuses in relation to the verbs. In the non-incorporation construction the object is a full-fledged argument that is an independent constituent from the verb, which I assume here is a DP. In the incorporation construction the incorporated noun holds a closer relation to the verb than its counterpart in the non-incorporation construction. Such a close relation is exhibited in different ways across languages. For example, in some languages, such as Ponapean, the incorporated noun occurs inside inflectional and/or derivational verbal morphology. An example is provided in (2). In this example, the completive suffix -la immediately follows the verb kang ‘eat’ in the syntactic counterpart sentence (2a).

However, in the incorporation construction (2b), the suffix follows the entire VN complex.

(2) a. I kanga-la wini-o. [Ponapean, Mithun 1984: 850, (6)]

I eat-COMP medicine-that ‘I took all that medicine.’ b. I keng-winih-la.

I eat-medicine-COMP ‘I completed my medicine-taking.’

In West Greenlandic, the close relation between the verb and the object is evidenced by an unusual marking on the verb, namely, intransitive marking on what should be a transitive verb. Consider the pair of sentences in (3).

(3) a. Angunguu-p aalisagaq neri-v-a-a. [W. Greenlandic, van Geenhoven 1998: 13, (2)]

A.-ERG fish.ABS eat-IND-[+tr]-3SG.3SG ‘Angunguaq ate the/a particular fish.’

b. Arnajaraq eqalut-tur-p-u-q. [W. Greenlandic, van Geenhoven 1998: 15, (5)]

A.-ABS salmon-eat-IND-[-tr]-3SG ‘Arnajaraq ate salmon.’

The transitive verb neri ‘eat’ is marked by a transitive suffix -a in the syntactic counterpart (3a), whereas, transitive verb tur ‘eat’ is marked by an intransitive suffix -u in the incorporation sentence (3b). Such a contrast in transitivity marking suggests that the object of (3b) is not an independent syntactic constituent. Moreover, in some languages, the incorporated noun cannot be separated from the verb by adverbs or other clause level syntactic constituents, whereas the unincorporated noun can be separated from the verb by modifiers. Consider an example from Kusaien.

(4) a. Sah el twem upac mitmit sac. [Kusaien, Mithun 1984: 851, (7)] Sah he sharpen diligently knife the ‘Sah is sharpening the knife diligently.’

b. Sah el twetwe mitmit upac. Sah he sharpen knife diligently ‘Sah is diligently knife-sharpening.’

In the non-incorporation (4a), the object mitmit sac ‘the knife’ is separated from the verb by the manner adverb upac ‘diligently’; whereas in the incorporation (4b), the manner adverb follows the VN complex.

The data presented above show that the incorporated noun is morpho- syntactically closer to the verb than the object NP in the non-incorporation sentence, in whatever way is applicable in a given language.

2.3.2 Restrictions on modification of incorporated noun

Furthermore, incorporated noun bears often more restrictions on modification than the object in the non-incorporation construction. According to Mithun (2000), nominal morphological markers that are used to modify nouns are normally not incorporated, including number affixes, case markers, determiners, quantifiers, among others. Consider the following example that she used to illustrate this point. (Similar effects can be seen in (2)-(4) above.)

(5) a. gam-nan tə-ntəwat-ə-rkən utkučʔ-ən. [Chukchee, Mithun 2000: 916, (2)]

1.SG-ERG 1.SG-set-0-PRESENT trap-ABS

‘I am setting a trap.’ b. gəm t-otkočʔə-ntəwat-ə-rkən.

1.SG-ABS 1.SG-trap-set-0-PRESENT ‘I am trap-setting.’

In the case of incorporation (5b), the noun otkočʔə ‘trap’ carries no case marking, whereas in the case of non-incorporation (5a) the noun utkučʔ-ən ‘trap’ is in the absolutive case. Mithun argued that the nouns have a different syntactic status in the

16 two sentence structures and the incorporated noun is not a syntactic argument. NP-internal modification of the incorporated noun is largely restricted. For example, modifiers usually cannot appear with the incorporated noun in Mapudungun but they can appear with the unincorporated noun.

(6) a. Pedro ngilla-ﬁ-y küme pulku. [Mapudungun, Baker 2009: 153, (8)]

Pedro buy-3O-IND.3sS good wine

‘Pedro bought good wine.’ b. Pedro ngilla-(*küme)-pulku-pe-y.

Pedro buy-good-wine-PAST-IND.3sS ‘Pedro bought (*good) wine.’

However, there is variation in this across languages. Some studies show that, at least for some languages, the incorporated noun can be modified to some extent, for example, it can be modified by adjectives. Consider an example in Hindi (Dayal 2011).

(7) anu sirf puraanii kitaab becegii. [Hindi, Dayal 2011: 136, (26a)]

Anu only old book sell-FUT ‘Anu will only sell old books.’

Note that such modification is subject to a condition that the activity is a common activity (see Section 2.5.4), at least, this is the case for Hindi (see Massam 2001 for some data of Niuean that do not involve common activities). In this example, puraanii ‘old’ can but bhaarii ‘heavy’ cannot modify the incorporated noun kitaab ‘book’ since sell old book is a common activity but sell heavy book is not. In many cases, modification of the incorporated noun is restricted in such a way that the modifier is not expressed as an NP-internal dependent incorporated together with the head noun but instead expressed as an element external to the V+N unit, which creates a discontinuous dependency between the modifier and the

17 incorporated noun. Consider an example from Cree where the modifier is a relative clause.

(8) a. kiic -ikonam -w maskisina kaa-osihtaayit John -a [Cree, Mellow 1989: 56, (44)]

remove TI AGR shoes that made AGR John AGR ‘She removed the shoes that John made.’ b. kiit -askisin -ii -n -iiw kaa-osihtaayit John -a

remove shoes AI TA AGR that made AGR John AGR ‘She removed the shoes that John made.’

In the non-incorporation structure in (8a), the relative clause forms a constituent with the head noun, whereas in the incorporation structure in (8b), though the interpretation is still that the relative clause modifies the object askisin ‘shoes’, it is syntactically placed outside of the V+N complex. Though Mellow (1989) did not provide a corresponding sentence which would show whether the relative clause can form a constituent with the incorporated noun, which would show a restriction on the modification of the incorporated noun, it seems to me that the discussion suggests that that would be an unacceptable sentence. The examples provided above suggest that modification of the incorporated noun is often restricted. NP-internal modification is not categorically prohibited across languages, but it can be constrained.

2.4 TESTING FOR NOUN INCORPORATION IN CHINESE FORMALLY

Chinese S-le sentences also exhibit these morpho-syntactic properties, which provide some support for the idea that S-le sentences reflect a type of noun incorporation. The data are provided below in this section.

2.4.1 Morpho-syntactic close relations between the object and the verb

First, the object of an S-le sentence also shows a close relation to the verb.

Consider example (9). It is apparent that the object shū ‘book’ in (9b) is closer to the verb kàn ‘look’ than its counterpart to the verb in the V-le sentence (9a), as the verb and the object are separated by the particle le in the latter case.

(9) a. tā kàn-le shū.

he look-LE book b. tā kàn-shū le.

he look-book LE ‘He read a book.’

Modification by a sentence-level adverb, however, cannot be used to show that the object of an S-le sentence has a closer relationship to the verb than the object of a V-le sentence. This is because in Chinese the adverb is always preverbal regardless of whether it is an S-le sentence or a V-le sentence (Ernst 2014). An example is provided below. The two sentences in (10) show that the adverb rènzhēnde ‘diligently’ can only be preverbal in a V-le sentence, and the two sentences in (11) show that it is the same in an S-le sentence.

(10) a. tā rènzhēnde kàn-le shū.

he diligently look-LE book b. *tā kàn-le rènzhēnde shū.

he look-LE diligently book ‘He read a book diligently.’ (11) a. tā rènzhēnde kàn-shū le.

he diligently look-book LE b. *tā kàn-rènzhēnde-shū le.

he look-diligently-book LE

‘He diligently read a book.’

The data show that the object of an S-le sentence at the surface level is closer to the

19 verb than the object of a V-le sentence, but they do not provide evidence that the two structures are syntactically distinct in terms of the object’s grammatical status. In the next subsection, I provide more data to show that they are distinct on these grounds.

2.4.2 Restrictions on modification of the object

In this section, I provide data to show that the S-le sentence also exhibits modification restriction, whereas such a restriction does not apply to the corresponding V-le sentence, which provides evidence for the claim that the VP in the S-le sentence shows some type of incorporation. To be specific, I show that S-le sentences with different kinds of materials modifying the object form a hierarchy of acceptability as follows when presented in an out of the blue context.1

(12) bare nouns > NP-internal dependents > numerals > quantifiers, article

Acceptability decreases from left to right. S-le sentences with bare nouns are always judged as perfectly acceptable, those with NP-internal dependents shows overall decreased acceptability, those with numerals are barely acceptable, and the S-le sentences with quantifiers and the (indefinite) article are entirely rejected. By contrast, the V-le sentences with any of the four types of objects are always all acceptable. This suggests that the two constructions are syntactically distinct. I follow a common assumption that adjectival and PP adjuncts, plus restrictive relative clauses, and possessors are all NP-internal, and I also include demonstratives2 in this category based on consultants’ judgements, though technically we might want another category for demonstratives. I assume that functional categories like quantifiers, numerals, and articles are above the NP in some higher functional project, which for convenience I

1 Judgments of some of those S-le sentences can be influenced by the context presented to the consultants. As shown later in Section 3.3.1, acceptability of those S-le sentences with objects modified by NP-internal materials is increased when the sentence is presented in a specific context in which the proposition is explicitly clarified to be new for the hearer. However, it does not significantly increase acceptability of those with numerals, and those with quantifiers are unacceptable in all contexts. So the hierarchy still holds across contexts. 2 The demonstrative noun phrases used in those S-le sentences as the object can either be a deictic use or an anaphoric use (Wolter 2004). 20 will simply refer to here as DP (though it could be that there are a multiple layers of functional projections) (Abney 1987, Stowell 1989, Li 1998, Cheng & Sybesma 1998). The data below show that incorporated objects are generally NPs (i.e. largely lacking the higher functional material), whereas the object of a V-le sentence can easily have higher functional elements (i.e. be a full DP). Note that judgments of acceptability of those sentences used in the current study were collected from native speakers in informal discussions, with a goal of having a general sense of speaker responses but not producing quantitative results per se. For most of these sentences presented below as well as those in Section 3.3.1, I consulted anywhere from 3 to 21 native speakers either face-to-face or electronically. The notations that are used to indicate the various degrees of acceptability are explained as follows. The symbol ‘?’ indicates that consultants judge the S-le sentence acceptable but at least one consultant judges it less natural or shows a preference for the corresponding V-le sentence over the S-le sentence. The symbol ‘??’ means the consultants generally find it unacceptable but not everyone rejects it entirely. The symbol ‘*’ means no consultant accepts the sentence.

2.4.2.1 Bare noun objects S-le sentences with bare noun objects are judged as perfectly acceptable by all speakers. Two examples are presented below. The event in (13) is an event of mopping the floor and the event in (14) is an event of writing homework.

(13) a. tā cā-le dìbǎn.

he mop-LE floor b. tā cā-dìbǎn le.

he mop-floor LE ‘He mopped the floor.’

(14) a. tā xiě-le zuòyè.

he write-LE homework

b. tā xiě-zuòyè le.

he write-homework LE ‘He wrote homework.’

Five consultants accepted the two S-le sentences in (13) and (14) immediately and judged them as acceptable as the corresponding V-le sentences.

2.4.2.2 Objects with NP-internal modifiers The NP-internal modifiers concerned below include adjective, PP, possessor, demonstrative, relative clause, and a mixture of those. These S-le sentences received slightly different judgments from consultants. Some S-le sentences were accepted by some consultants as just as acceptable as the V-le sentences while some native speakers considered the S-le sentences less natural than the V-le sentences. Overall, we can conclude that these S-le sentences with NP-internal modifiers are acceptable but less natural than the corresponding V-le sentences. Thus acceptability of the S-le sentences in this case is slightly lower than those with bare noun objects. The modifier in the first example is an adjective. I received judgments from eight consultants. Seven of them accepted (15b) and one found it less natural.

(15) a. tā chuān-le hēisè de yīfū.

she wear-LE black DE clothes

b. ?tā chuān-hēisè-de-yīfū le.

she wear-black-DE-clothes LE ‘She wore black clothes.’

The modifier in (16) is a PP, which unambiguously modifies the object only and not the VP – a VP-level adverbial would be preverbal. I received judgments from four consultants and they all consider (16a) more natural than (16b). One of them accepted

(16b) and the other three judged it awkward or less natural.

(16) a. tā sǎo-le wūdǐng-shàng de xuě.

he sweep-LE house.roof-on DE snow b. ?tā sǎo-wūdǐng-shàng-de-xuě le.

he sweep-house.roof-on-DE-snow LE ‘He swept the snow that was on the roof.’

The modifier in (17) is a possessor. I received judgments from eight consultants and seven of them preferred (17a) to (17b) and the other judged the two sentences as equally acceptable. Five accepted (17b) and the other three found it less natural, but still acceptable.

(17) a. tā dǎsǎo-le Mǎlì de chúfáng.

he clean-LE Mary DE kitchen b. ?tā dǎsǎo-Mǎlì-de-chúfáng le.

he clean-Mary-DE-kitchen LE ‘He cleaned Mary’s kitchen.’

The modifier in (18) is a demonstrative which is optionally followed by a classifier. I received judgments from eight consultants. Six of them accepted (18b) and two found it less natural. Two considered the two sentences equally acceptable and six preferred (18a).

(18) a. tā kàn-le nèi(-běn) shū.

he look-LE that-CL book b. ?tā kàn-nèi(-běn)-shū le.

he look-that-CL-book LE ‘He read that book.’

The modifier in (19) is a relative clause. I received judgments from four consultants. Three of them accepted (19b) and the fourth found it not quite acceptable. They all

23 preferred (19a) to (19b).

(19) a. tā kàn-le Mǎlì mǎi de shū.

he look-LE Mary buy DE book b. ?tā kàn-Mǎlì-mǎi-de-shū le.

he look-Mary-buy-DE-book LE ‘He read the book that Mary bought.’

Finally, the modifier in (20) is a mixture of modifying materials including a demonstrative, a PP and a relative clause. I got responses from 12 native speakers. Two of them accepted (20b); nine found it less natural but still acceptable; one of them could not accept it.

(20) a. tā kàn-le wǒ mǎi de nèi-běn guānyú gǔxīlà shénhuà de shū.

he look-LE I buy DE that-CL about ancient.Greek mythology DE book b. ?tā kàn-wǒ-mǎi-de-nèi-běn-guānyú-gǔxīlà-shénhuà-de-shū le.

he look-I-buy-DE-that-CL-about-ancient.Greek-mythology-DE-book LE ‘He read that book about the ancient Greek mythology that I bought.’

The six examples provided above show that, overall, the S-le sentences with NP-internal modifiers are judged less natural than the corresponding V-le sentences, but are still judged as roughly acceptable. This contrasts with judgments of S-le sentences with bare noun objects, suggesting that the VP in the S-le sentence involves incorporation since, as described in the literature, incorporated nouns tend to be formally smaller units.

2.4.2.3 Objects with numerals When the object is modified by a numeral, acceptability of the S-le sentence is more significantly reduced. An example is presented in (21). I received judgments from four consultants. One found (21b) unacceptable and the other three found it hard to accept but did not entirely 24 reject it.

(21) a. tā chī-le sān-wǎn mǐfàn.

he eat-LE three-bowl rice b. ??tā chī-sān-wǎn-mǐfàn le.

he eat-three-bowl-rice LE

‘He ate three bowls of rice.’

There is more evidence in the literature for the reduced acceptability of S-le sentences with numerals. In a survey of uses of le, Spanos (1979: 49-50) also had a similar result. In this study, 39 native speakers3 were asked to insert le (V-le, S-le, or both) into a number of sentences in which le had been removed, and they were also asked to indicate whether the insertion of le at the position is optional or obligatory. One of the test items involves an object modified by a numeral, which is repeated in (22). The participants were asked to insert le into the first sentence. The insertion may generate three possible sentence structures: an S-le sentence, a V-le sentence, and a double le sentence.4 The judgments are diverse. Only 9 participants accepted the S-le sentence and 30 rejected it.5

3 The total number of subjects involved in the study is 62. But only 39 of them considered Mandarin as the dominant language. 4 In addition to S-le and V-le sentences where the particle le shows up at the end of a sentence and as a verbal suffix respectively, there is another sentence structure called double le sentence which contains both the final particle le and the verbal suffix le. The double le sentence is equivalent to the perfect aspect sentence in English. An example is presented as follows.

(1) tā hē-le sān-wǎn tāng .le he drink-LE three-bowl soup LE ‘He has drunk three bowls of soup.’

The double le sentence, though also with the final le, is syntactically distinct from the S-le sentence as its object can be an NP or DP whereas the object of S-le sentence can only be an NP. 5 The details of the results of this task is as follows: One participant chose an S-le sentence as the only acceptable structure in this discourse; another eight judged an S-le sentence as one of the acceptable structures in this case; 30 did not accept an S-le sentence here. By contrast, 16 chose a V-le sentence as the only acceptable structure here; 9 chose a V-le sentence as one of the acceptable structure; 14 did not accept a V-le sentence here. (Of the 14 who did not accept a V-le sentence, 7 chose a double le sentence as the only acceptable structure; 6 chose both a double le sentence and an S-le sentence as acceptable; one chose an S-le sentence as the only acceptable structure.) It is notable that there is a discourse effect here. As indicated by the translation, the best expression for the first sentence of the discourse is a sentence with perfect aspect, that is, the double le sentence. This is likely why 14 participants did not choose the V-le sentence. 25

(22) wǒ yǐjīng kàn sān-běn shū. [Spanos 1979: 49, (T7)]

I already read three-CL book (wǒ hái yào zài kàn liǎng-běn.)

I still want again read two-CL ‘I’ve already read three books. I want to read two more.’

Overall, the result of the participants’ judgments for the three possible sentence structures of the first sentence of (22) is consistent with the judgments provided by the consultants of the current study. The significantly reduced acceptability of the S-le sentence with a numeral suggests that the VP of the S-le sentence involves incorporation, and the object is an NP, not DP.

2.4.2.4 Objects with quantifiers and articles When the object is modified by a quantifier or an indefinite article, the S-le sentence is entirely rejected. First consider the universal quantifier, which is obligatorily followed by a classifier. An example is provided in (23). All three native speakers I consulted judged (23b) unacceptable.

(23) a. tā kàn-le měi-běn shū.

he look-LE every-CL book

b. *tā kàn-měi-běn-shū le.

he look-every-CL-book LE

‘He read every book.’

Note that though they considered (23a) acceptable, they also pointed out that the most natural expression for a universal quantifier is an SOV order sentence. The example is provided as follows (see Section 3.4.2.1 for more discussion of SOV sentences).

(24) tā měi-běn shū dōu kàn-le.

he every-CL book all look-LE

‘He read every book.’

Mandarin has no definite article but it has an indefinite article, which is restricted to the object position.6 The indefinite article can be used in the V-le sentence but not in the S-le sentence. Consider the following example.

(25) a. tā kàn-le běn shū.

he read-LE IND book b. *tā kàn-běn-shū le.

he read-IND-book LE ‘He read a book.’

Soh (2009: 638) had similar judgments for the S-le sentences that contain a downward-entailing quantifier. One of her examples is presented in (26). Note that

(26b) actually consists of two sentence structures, an S-le sentence (without the verbal suffix -le) and a double le sentence (with the verbal suffix -le),7 though she did not make this explicit in her discussion. It seems that she was considering them both to illustrate the same structure that has a final le, but in fact they are two syntactically and semantically distinct constructions. Regardless, her judgment is that (26b) is unacceptable or infelicitous, by which she seems to mean that both sentences are unacceptable or infelicitous. What is important for us here is that she judged the S-le sentence in (26b) unacceptable or infelicitous. This is consistent with our claim that the object of S-le sentences cannot be modified by a quantifier.

(26) a. tā hē-le bùdào sān-wǎn tāng. [Soh 2009: 638, (31)]

he drink-LE less.than three-bowl soup

6 The indefinite article is actually a set of classifiers. A classifier is usually used with a numeral to form a classifier construction. But it may also be used without a numeral, in which case it functions like an indefinite article and it shows agreement with the noun. In Mandarin this use of the classifier is restricted to the object NP. In other Chinese languages such as Cantonese, the classifier without the numeral functions like a definite article and is restricted to the subject NP. See Cheng & Sybesma (1998). 7 See Footnote 4 for an explanation of the double le sentence. 27

b. */#tā hē(-le) bùdào sān-wǎn tāng le.

he drink-LE less.than three-bowl soup LE ‘He drank less than three bowls of soup.’

However, she considered the corresponding sentence with upward-entailing quantifiers acceptable. One of her examples is provided in (27b). Again, (27b) represents two sentence structures, the S-le sentence and the double le sentence, and she did not distinguish them and simply referred to them as S-le sentences and judged them acceptable. My judgment is that the double le sentence in (27b) is acceptable but the S-le sentence is unacceptable. My consultants have the same judgment. None of my five consultants judged the S-le sentence (27b) acceptable. The corresponding V-le sentence (27a), which is provided by me, is the standard expression for the intended meaning.

(27) a. tā hē-le zhìshǎo sān-wǎn tāng.

he drink-LE at.least three-bowl soup b. tā hē(-le) zhìshǎo sān-wǎn tāng le. [Soh 2009: 639, (36b)]

he drink-LE at.least three-bowl soup LE ‘He drank at least three bowls of soup.’

The data above show that the object of the S-le sentences cannot be a quantificational noun phrase.

2.4.3 Summary

In this section, I have shown that S-le sentences show some morpho-syntactic properties that provide some support for the idea that they are a type of noun incorporation. The object of S-le sentences at the surface level shows a close relationship with the verb as it is immediately adjacent to the verb. More significantly, the object of S-le sentences is restricted in its modificational properties as modification of the object increasingly decreases acceptability of the sentence in 28 accordance to the hierarchy repeated in (28).

(28) bare nouns > NP-internal dependents > numerals > quantifiers, article

2.5 SEMANTIC PROPERTIES OF NOUN INCORPORATION

Early semantic studies of noun incorporation were largely descriptive, and much of the discussion of its semantic properties was focused on the referentiality of the incorporated noun. According to Baker (1988: 78-81), the incorporated noun can serve as an antecedent for an anaphoric pronoun in the subsequent discourse, which he took as evidence to support the claim that incorporated nouns are referential. Mithun (1984, 1986, 2000), however, claimed that incorporation does not specify referentiality for incorporated nouns. According to her, the referential system of polysynthetic languages like Mohawk does not work in the same way as English.

Instead, pragmatics plays a more important role in the use of pronouns in such languages. Moreover, she proposed that the incorporated noun is a qualifier and plays a function of modifying the verb, narrowing its scope. Formal semantic analyses of noun incorporation started with Bittner (1994) and van Geenhoven (1998) (see also Chung & Ladusaw 2003, and Farkas & de Swart 2003). Van Geenhoven related incorporation to English bare plurals and German split topics and argued that they share some semantic properties, in particular the nominals of these constructions denote properties and are all narrow scope elements and thus they can be analyzed along similar lines. She unified these phenomena under the cover term of semantic incorporation. More recently, a number of semantic properties have been proposed to characterize noun incorporation or pseudo noun incorporation, including number neutrality, narrow scope (relative to negation, universal quantification, and modal operators), discourse opacity, and well-establishedness effects (see Borik & Gehrke 2015, Dayal 2011). But note that details with respect to how a noun incorporation language exhibits these semantic properties differ from language to language. For

29 example, while incorporated nouns are typically bare nouns and have number neutral interpretations, some languages also allow number marked nouns to be incorporated such as Hindi (Dayal 2011) and Hungarian (Farkas & de Swart 2003). Moreover, noun incorporation is found to be a productive process in some languages (Baker 1988); however, in some other languages such as Hindi (Dayal 2011), it is found that incorporation constructions are subject to conceptual restrictions and are expressions that are formed to describe institutionalized or common activities. So, although noun incorporation often shows these properties across languages, it can vary which properties a given language shows. Below I present data from other languages to illustrate the four semantic properties. Later I will show that many of those semantic properties can also be tested in the S-le sentences, indicating that it is likely to be a case of noun incorporation.

2.5.1 Number neutrality

The incorporated noun is usually a noun stem. However, singular incorporated nouns are usually semantically number neutral in the sense that they may receive either a singular or a plural interpretation, despite the fact that the language may otherwise morphologically mark for the contrast of singularity and plurality. An example from West Greenlandic is presented below. The incorporated noun kaage ‘cake’ is a bare noun stem and it can be interpreted as singular or plural.

(29) Kaage-liur-p-u-t. [W. Greenlandic, Van Geenhoven 1998: 23, (34)]

cake-make-IND-[-tr]-3PL ‘They made cake/a cake/cakes.’

A way to express a numeral in West Greenlandic is to mark a noun with the

INSTRUMENT case while it is still incorporated. An example is presented as follows.

(30) Marlun-nik ammassat-tur-p-u-nga. [W. Greenlandic, Van Geenhoven 1998: 19, (17)]

two-INST.PL sardine-eat-IND-[-tr]-1SG 30

‘I ate two sardines.’

Note that van Geenhoven did not mention if West Greenlandic allows numerals to be expressed without the object being incorporated. So it is not clear if (30) is the only way to express numerals in West Greenlandic. Note that it is found that it is often the case that incorporated nouns are simply noun stems without number affixes or any other grammatical markers, though the language may have a morphological contrast between singular and plural (Mithun 2000). Unfortunately, Mithun (2000) did not provide relevant data to show that plural forms cannot be incorporated in those languages. However, in some other languages, such as Hungarian and Hindi, both morphologically singular and plural nouns can be incorporated. A Hungarian example is presented below, where the incorporated noun in (31a) is morphologically singular and the one in (31b) is plural.

(31) a. Mari verset olvas. [Hungarian, Farkas & de Swart 2003: 12, (21)]

Mari poem.ACC read ‘Mari is reading a poem/poems.’ b. Mari verseket olvas.

Mari poem.PL.ACC read ‘Mari is reading poems.’

The singular incorporated noun in (31a) is semantically neutral like the incorporated noun in sentence (29). However, the plural incorporated noun in (31b) can only receive a plural interpretation, which contrasts with the number neutrality reading of the singular incorporated noun.

2.5.2 Narrow scope

Since Bittner (1994) and, especially van Geenhoven (1998), who related incorporated nouns to indefinite NPs in English and German, the property of

31 incorporated nouns taking narrow scope with respect to negation, universal quantification, and modals has been considered a especially prominent characteristic of incorporation (Borik & Gehrke 2015). First consider an example from Hungarian which illustrates the narrow scope of the incorporated noun relative to a modal operator ‘must’ (Farkas & de Swart 2003: 7).

(32) a. Mari kell olvasson egy verset. [Hungarian, Farkas & de Swart 2003: 7, (7)]

Mari must read.SUBJ. a poem.ACC ‘Mary must read a poem.’ b. Mari verset kell olvasson. [Hungarian, Farkas & de Swart 2003: 7, (8)]

Mari poem.ACC must read.SUBJ. ‘Mari must read a poem/poems.’

In (32a), the object egy verset ‘a poem’ is a normal indefinite noun phrase and it is scopally ambiguous. Under a wide scope reading, the interpretation is that there is a specific poem that Mary has to read. Under the narrow scope reading, Mary fulfills her obligation if she reads any poem. In (32b) the object verset ‘poem’ is incorporated. There is only one interpretation available for this sentence, that is, Mary fulfills her obligation if she reads any poem. The incorporated noun also has a narrow scope interpretation relative to negation. A Hungarian example is provided in (33). Note that, according to Farkas & de Swart (2002), the negation sentence has no minimal pair in which the object is a full-fledge DP (which takes a determiner). This is different from the other cases such as the modal sentence (32) and the sentence with a quantifier (35), in which there is a minimal pair.

(33) Mari nem olvas verset. [Hungarian, Farkas & de Swart 2003: 7, (9)]

Mari not read poem.ACC ‘Mari is not reading a poem/poems.’

However, Hindi shows such a pair of negative sentences (Dayal 2011), and the incorporated noun has a narrow scope interpretation relative to negation.

(34) a. anu bacca nahiiN sambhaalegii. [Hindi, Dayal 2011: 128, (8)]

Anu child not look-after-FUT ‘Anu will not look after children.’ b. anu ek bacce-ko/bacce-ko nahiiN sambhaalegii.

Anu one child-ACC child-ACC not look-after-FUT ‘Anu will not look after a particular/the child.’

Sentence (34a) contains an incorporation structure where the incorporated noun bacca ‘child’ is not case-marked. This sentence means that Anu will not look after any children. Under this reading, the negation operator takes wide scope relative to the existential quantifier for the incorporated noun. But in the case of (34b), which is the non-incorporation counterpart with the object marked in accusative case, the existential takes wide scope over the negation. When the object is the indefinite form ek bacce-ko ‘one child-ACC’, it is readily interpreted with the existential taking a wide scope to negation. (It is not clear to me if it also allows a narrow scope interpretation.)

When the object is a bare singular noun with accusative case bacce-ko ‘child-ACC’, it has a definite reading and the sentence means that there is a particular child, contextually salient, that Anu will not look after. The incorporated noun also receives a narrow scope interpretation relative to a universal quantifier in the subject. Consider the following example. According to Farkas & de Swart (2003), the incorporated noun in (35) has a narrow scope interpretation while the object of the corresponding unincorporated sentence is scopally ambiguous.

(35) Minden gyerek verset olvas. [Hungarian, Farkas & de Swart 2003: 7, (10)]

every child poem.ACC read ‘Every child reads a poem/poems.’ 33

The data above show that incorporated nouns consistently receive narrow scope interpretations relative to other scopal operators in the clause. The property of narrow scope has been attested in many languages, and is considered one of the most stable properties of noun incorporation (Borik & Gehrke 2015: 14).

2.5.3 Discourse opacity

Discourse opacity is also considered a prominent characteristic of noun incorporation, though there is considerable variability in whether and how this manifests across languages. The reason is that studies of incorporation in different languages make different claims about whether the incorporated noun is referential and can establish a discourse referent. A number of claims in the literature are reviewed below, including earlier and more recent studies.

Sadock (1980) observed that in West Greenlandic the incorporated noun can serve as an antecedent of a personal suffix which functions like English personal pronouns. An example is provided as follows.

(36) Suulut timmisartu-liur-p-u-q. [W. Greenlandic, Sadock 1980: 311, (37-38)]

S∅ren airplane-made-IND-[-tr]-3SG Suluusa-qar-p-u-q aquute-qar-llu-ni-lu

wing-have-IND-[-tr]-3SG rudder-have-INF-3SG.PROX-and

‘S∅ren made an airplanei. Iti has wings and iti has a rudder.’

The third and fourth person suffixes in the second sentence refer to the incorporated noun immisartu ‘airplane’ in the preceding sentence. Baker (1988: 78-81) also argued that incorporated nouns are referential. He observed that though an incorporated noun often refers to a generic or unspecific class, like a noun root in a compound, it can also refer to a specific object introduced earlier in the discourse. An example from Mohawk is provided as follows.

(37) No:nv akwe: yo-stathv no-:nvhst-e sok nu:wa [Mohawk, Baker 1988: 79, (9)]

when all 3N-dry PRE-corn-SUF then now v-tsaka-nvhst-aru:ko.

FUT-1PS-corn-takeoff ‘When the corn was completely dry, it was time to shell it (the corn).’

The incorporated noun nvhst ‘corn’ in the second sentence seems to co-refer with the

NP nvhst ‘corn’ in the first sentence, which suggests that incorporated noun is referential. Mithun (1984: 871), however, argued that the basic function of incorporated nouns is to qualify the verb, and that incorporated nouns are not normally used to establish discourse referents. She further argued that the pronominal system of polysynthetic languages is different from English. She used the following discourse to show that the referent of a pronominal suffix is often determined pragmatically.

(38) K-atenún-hah-kwe. Áh tsi yehétkv. [Mohawk, Mithun 1984: 871, (112)]

I-watch-HAB-PAST ah how she.ugly ‘I was babysitting. Boy, is she ugly!’

The first sentence K-atenún-hah-kwe contains no incorporated noun and thus there is no antecedent for the pronominal prefix ye- ‘she’ in the second sentence. The referent of ye- ‘she’ is determined pragmatically. In more recent literature, Massam (2001) observed that incorporated nouns in Niuean do not show uniformity with respect to discourse transparency or opacity. In general, incorporated nouns are non-referential and cannot be the antecedent of an anaphoric pronoun. Discourse transparency is limited to cases where the incorporating verbs have existential semantics, in which cases the incorporated nouns can serve as antecedents of anaphoric pronouns in the subsequent discourse. Farkas & de Swart (2003: 13) noticed that whether incorporated nouns are discourse transparent or not depends on three factors, namely, the specific language under consideration, the 35 morphological number of the incorporated nouns, and the type of anaphor considered. Below I illustrate the last two factors with data from Hungarian provided in their work. Languages like Hindi and Hungarian, which allow both morphological singular and plural incorporated nouns, show discourse opaqueness for singular nouns while showing discourse transparency for plural nouns. Consider a Hungarian example in (39) (Farkas & de Swart 2003: 18-19). Note that pro stands for a covert pronoun. The two discourses differ in the morphological number of the incorporated noun. Farkas & de Swart remarked that the judgment on the discourse transparency is subtle, but their participants agreed that (39a), where the antecedent of a regular pronoun is a singular incorporated noun, is significantly worse than (39b), where the antecedent is a plural incorporated noun.

(39) a. Jánosi betegetj vizsgált a rendelőben.

J. patient.ACCj examine.PAST the office.in ‘Janos patient-examined in the office.’

??proi Túl sulyosnak találta őtj és

proi too severe.DAT find.PAST he.ACCj and ‘He found him too sick and sent him to hospital.’

b. Jánosi betegeketj vizsgált a rendelőben.

J. patient.PL.ACCj examine.PAST the office.in ‘Janos patients-examined in the office.’

proi Túl sulyosnak találta őketj és

proi too severe.DAT find.PAST he.PL.ACCj and

beutaltatta proj a korházba.

Intern.Cause.PAST proj the hospital.in ‘He found them too sick and sent them to hospital.’

Both Farkas & de Swart (2003) and Dayal (2011) noted that the discourse

36 transparency of the incorporated noun may also be affected by the nature of the anaphor. For example, in Hungarian, singular incorporated nouns cannot be antecedents to regular overt pronouns; however, some native speakers find them transparent with respect to covert pronouns. Consider the two discourses in (40) (Farkas & de Swart 2003: 19). The data show that the singular incorporated noun beteget ‘patient’ cannot be the antecedent to an overt pronoun őt ‘he.ACC’ but it can be taken up by a covert pronoun pro in the following discourse.

(40) Jánosi betegetj vizsgált a rendelőben.

J. patient.ACCj examine.PAST the office.in ‘Janos patient-examined in the office.’

a. ??proi Túl sulyosnak találta őtj és beutaltatta proj

proi too severe.DAT find.PAST he.ACCj and Intern.CAUSE.PAST proj a korházba. the hospital.in

b. proi Túl sulyosnak találta proj és beutaltatta proj

proi too severe.DAT find.PAST proj and Intern.CAUSE.PAST proj a korházba. the hospital.in ‘He found him too sick and sent him to hospital.’

As noted by Farkas & de Swart (2003: 18-19), judgments on discourse transparency are subtle, so consultants may have different judgments for the acceptability of a given discourse. Nevertheless, the data presented above show that the referential capacity of morphologically singular incorporated noun is relatively weaker, though the transparency is increased when the noun is morphologically plural. The incorporated noun sometimes cannot serve as an antecedent for a regular anaphoric pronoun in the subsequent discourse but may serve as the antecedent of a covert anaphor.

2.5.4 Well-establishedness effects

Mithun (1984) noted that many noun incorporation structures describe kinds of activities that are culturally recognizable, common, and well-established in the community. This has also been observed in other languages (e.g. Massam 2001, Dayal 2011, Barrie & Li 2015). In the more recent literature, well-establishedness effects are considered one of the properties of incorporation (e.g. Borik & Gehrke 2015,

Mueller-Reichau 2015). But as with many properties claimed to be associated with noun incorporation in the literature, we can see that languages differ with respect to the extent to which well-esstablishedness effects apply. In some languages, noun incorporation is found highly productive (Baker 1988), and many examples do not seem to describe recognizable, common or institutionalized activities, though we cannot be entirely sure what activities are viewed as common or well-established in a given culture. Hindi is one of those languages that show productive incorporation but also show gaps for the construction (Dayal 2011). For example, in Hindi, the verb dekhnaa ‘see’ can be incorporated with a nominal laRkii ‘girl’ but not aurat ‘woman’. Examples are presented in (41). The incorporation of laRkii-dekhnaa ‘girl-seeing’ has an idiomatic meaning. It does not describe an event of seeing a girl; instead, it refers to a culturally recognizable event of looking for a girl as a prospective bride.

(41) a. laRkii-dekhnaa [Hindi, Dayal 2011: 133, (21)] girl/seeing b. *aurat-dekhnaa

woman-seeing

Similarly, modification of the incorporated noun is also subject to certain restrictions.

Both Massam (2001) and Dayal (2011) point out that not all kinds of modification are allowed. In the Hindi example in (42), where the incorporated noun is puraanii kitaab ‘old book’, Dayal claimed that it is unacceptable to use an adjective meaning ‘heavy’ 38 instead of the one meaning ‘old’ in this sentence, and she attributed this to the fact that selling old books is a prototypical activity while selling heavy books is not.

(42) anu sirf puraanii kitaab becegii. [Hindi, Dayal 2011: 136, (26a)]

Anu only old book sell-FUT ‘Anu will only sell old books.’

Conversely, we find incorporation for activities of ‘make an airplane’ from West Greenlandic as in (36), ‘take medicine’ from Ponapean as in (2b), and ‘buy a bed’ from Mohawk as in (1b). It is hard to see those activities could be common activities in any community. Those data suggest that incorporation in those languages is not motivated by a need to conceptualize a common activity, though it is still likely that productivity of noun incorporation is also restricted in those languages.

2.6 TESTING FOR NOUN INCORPORATION SEMANTICALLY IN CHINESE

Those semantic properties that we normally find in noun incorporation also largely apply to S-le sentences. Like some of those noun incorporation languages, S-le sentences show those semantic properties to some extent and vary in some details of those properties. To be specific, they show number neutrality if the object is a bare noun, though nouns marked with plurality can also be incorporated. The object has narrow scope interpretation relative to modal operators and the universal quantifier.

But there is no convincing evidence for narrow scope relative to negation. Discourse opacity applies as the object sometimes cannot be the antecedent of an anaphoric pronoun, though as I will show certain sentence types do allow this. Finally, some expressions seem to show the well-establishedness effect, though I suggest this can be accounted for by specific discourse factors. Overall, S-le sentences show those semantic properties to a relatively great extent, although these diagnostics do not provide as strong evidence as might be found in other languages for language particular reasons. However, given those diagnostics, we can conclude that it is

39 plausible that S-le sentences involve incorporation.

2.6.1 Number neutrality

Common nouns in Chinese can appear as bare nouns without any determiner in argument position (Chierchia 1998, 2015), regardless of whether the denoted object has a definite shape or not (cf. count nouns vs. mass nouns in English). Expression of number requires a specific construction called the classifier construction if the noun is a count noun or the massifier construction if the noun is a mass noun. The category of the classifier is a closed set of monosyllabic words and the use of the classifier varies according to the noun. Four examples are provided in (43).

(43) a. yì-běn shū b. sān-zhī māo

one-CL book three-CL cat ‘one book’ ‘three cats’ c. sì-bēi shuǐ d. sān-sháo yán

cup spoon four-MS water three-MS salt ‘four cups of water’ ‘three spoons of salt’

Examples (43a-b) show a classifier construction since the nouns shū ‘book’ and māo ‘cat’ are count nouns, and the noun shū ‘book’ requires the classifier běn as in (43a) while the noun māo ‘cat’ requires the classifier zhī as in (43b). Examples (43c-d) are massifier constructions, since the nouns shuǐ ‘water’ and yán ‘salt’ are mass nouns, and bēi ‘cup’ and sháo ‘spoon’ are massifiers which are used as units for measuring the quantity. The bare noun itself, however, is number neutral. In either a V-le sentence or an S-le sentence, number interpretation for a bare noun can be either singular or plural.

Consider the following two sentences.

(44) a. tā zuótiān chī-le píngguǒ.

he yesterday eat-LE apple b. tā zuótiān chī-píngguǒ le.

he yesterday eat-apple LE ‘He ate part of an/an/some/ apple(s) yesterday.’

The bare noun píngguǒ ‘apple’ in both the V-le sentence (44a) and the S-le sentence

(44b) has the same number interpretation, which can be partitive, singular, or plural, depending on the context. There is a plural morpheme in Mandarin, namely, -men, which, however, is subject to many restrictions (Li & Thompson 1981: 11-12). For example, it is largely restricted to humans and sometimes animals but rarely used for inanimate entities, and it cannot co-occur with the numeral classifier construction (see (43)). Nouns inflected in -men can also be incorporated. Consider (45), where (45a) contains the bare noun xiǎoháizi ‘child’ and (45b) contains the plural form xiǎoháizi-men ‘children’.

(45) a. tā zuótiān zhàogù-xiǎoháizi le.

he yesterday look.after-child LE ‘He looked after a/some child(ren) yesterday.’ b. tā zuótiān zhàogù-xiǎoháizi-men le.

he yesterday look.after-child-PL LE ‘He looked after children yesterday.’

The bare noun xiǎoháizi ‘child’ in (45a) allows a singular or plural interpretation. The plural form xiǎoháizi-men ‘children’ in (45b) only has a plural interpretation. Ultimately the number neutrality test, which might provide evidence for incorporation in other languages, does not provide direct evidence for incorporation in

Chinese since both constructions allow a number neutral interpretation for a bare noun. Nonetheless, it does not rule out that the S-le sentence involves incorporation since number neutrality is exactly what we would expect under incorporation. 41

2.6.2 Narrow scope

The object of the S-le sentence also has a narrow scope interpretation relative to scopal operators such as modal verbs and universal quantifier. There is no clear evidence for narrow scope relative to negation. However, the data are still consistent with the analysis. The relevant data are discussed below.

2.6.2.1 Negation First, I discuss the negation test, which cannot provide clear evidence for narrow scope of the object of S-le sentences. The problem is that negation of the S-le sentence and that of the V-le sentence have the same surface string,8 so we do not know for sure which sentence structure we are actually dealing with. For example, we might represent negation of an S-le sentence as in (46a) while negation of a V-le sentence as in (46b). But the two sentences in (46) are phonologically the same. (Note that the perfective suffix -le and the negative particle méi are in complementary distribution and -le is neutralized in negative sentences (Li & Thompson 1981: 430). The final particle le of S-le sentences somehow is not realized either in negation.)

(46) a. tā méi chī píngguǒ.

he NEG eat apple ‘He didn’t eat an apple.’

b. tā méi chī-píngguǒ.

he NEG eat-apple ‘He didn’t eat an apple.’

The sentence (46) allows two interpretations. One is that there was a specific apple but he did not eat it. For example, two speakers are talking about the reception last

8 Here we only consider the case when the S-le sentence expresses a past event. But as shown in Section 3.3.3, the S-le sentence is neutral with respect to grammatical aspect (and has no tense), and its temporal interpretation is determined by context. With different aspectual interpretations, the negation of an S-le sentence is different. For simplicity, here we only focus on the case when an S-le sentence is used to talk about a past event. 42 night at which everyone was served an apple in addition to other food. One speaker asks the other one whether he ate the apple. The sentence (46) can be used by the other speaker as an answer if he did not eat that specific apple. The other interpretation of (46) is that he ate no apples at all. For example, one speaker asks the other one whether he ate any apple at all yesterday without referring to any specific apple. The other speaker may use (46) if he did not eat any apple at all yesterday.

The negation diagnostic does not provide evidence for our analysis of incorporation, but it is still consistent with our analysis. Below I show that the other two operators, modality and universal quantifier, can provide direct evidence for narrow scope of the object of S-le sentences.

2.6.2.2 Modality Although negation does not work as a scopal operator to explore the scopal properties of S-le sentences, modals do, and in sentences with modal auxiliaries S-le appears to have narrow scope relative to the modal. Consider the example (47).

(47) a. tā bìxū chī-le miàntiáo.

he must eat-LE noodles ‘He must eat the noodles.’ b. tā bìxū chī-miàntiáo le.

he must eat-noodles LE ‘He must eat noodles now.’

The V-le sentence (47a) means that the person has an obligation to eat some specific noodles which are identifiable to the hearer.9 That person does not satisfy the

9 In addition, my intuition is that (47a) has a telic interpretation, entailing culmination. Thus the sentence means that the person has the obligation to eat all of the noodles. The verbal suffix -le in (47a) can be replaced by the resultative morpheme -diao, resulting in a sentence which seems to me is equivalent to and interchangeable with (47a), as shown below. See Chapter 4 for more discussion on -diao and telicity in Chinese.

(1) tā bìxū chī-diào miàntiáo. he must eat-CULM noodles ‘He must eat up the noodles.’ 43 obligation if he eats any other noodles. This is consistent with a wide scope analysis of the object quantifier over modal operator. For the S-le sentence (47b), the object has a narrow scope reading relative to the modal operator. The sentence means that the person fulfills his obligation if he eats any noodles now. For example, suppose the person that the subject of (47b) refers to has an obligation of eating noodles at noon every day, and now the time is approaching 12. The speaker may utter (47b) at this moment, by which he means that that person must eat noodles now, which he can get from anywhere. Given that the reading for V-le sentences is obligatorily wide scope, an alternative might be that the object here is not a wide-scope indefinite but is instead interpreted as some kind of definite, and thus is non-scopal. The idea that V-le sentences can be treated as showing some kind of definiteness comes from the fact that a sentence without le can also be used with a modal verb, and in this case the bare noun allows both a narrow scope and a wide scope reading. Consider the example (48).

(48) tā bìxū chī miàntiáo. he must eat noodles ‘He must eat some noodles.’

Under the narrow scope reading of the object relative to the modal operator, the person fulfills his obligation if he eats noodles. For example, suppose tomorrow is John’s birthday, and according to the Chinese tradition, people must eat noodles on their birthdays. The speaker may say (48) – although it is better to add the temporal adverbial míngtiān ‘tomorrow’ – to mean that John must eat noodles tomorrow. Under the wide scope reading, there are particular noodles the person has to eat. For example, at a dinner, a variety of food is ordered and noodles are the only main food in the order. The speaker believes that for every dinner you must eat some main food. So he may use (48) to mean that the person the subject refers to must eat (at least some of)

44 the ordered noodles. This thus looks more like a case of true variable scope. However, even on this analysis it is still true that the S-le sentence shows what appears to be narrow scope relative to the modal operator, which is all that matters here. To sum up, the examples show that the S-le sentence and the V-le sentence are semantically distinct and the object of S-le sentences seems to have narrow scope relative to modal verbs.

2.6.2.3 Universal quantifier The object of S-le sentences also has narrow scope relative to a universal quantifier in the subject, whereas the object of V-le sentences is ambiguous in its scope reading with respect to the universal quantifier. First consider the reading when the universal quantifier has wide scope to the existential quantifier. For example, John was talking to his colleagues in the office and they all said they watched a movie at home last night, though they are not necessarily the same movie. Both an S-le sentence and a V-le sentence as in (49) can be used in this case. This suggests that the object diànyǐng ‘movie’ in both sentences allows a narrow scope reading relative to the universal quantifier.

(49) a. měi-ge rén dōu kàn-diànyǐng le.

every-CL person all watch-movie LE

b. měi-ge rén dōu kàn-le diànyǐng.

every-CL person all watch-LE movie ‘Every person watched a movie.’

Now consider the reading when the existential quantifier has a wide scope to the universal quantifier. For example, suppose there was a free movie in the park of the capitol last night and from the conversation with his colleagues, John learned that all of his colleagues watched that movie in the park last night. In this case, the V-le sentence (49b) is more appropriate than the S-le sentence (49a). This suggests that the object of V-le sentences allows a wide scope reading relative to the universal

45 quantifier whereas that of S-le sentences does not. Note that for S-le sentences as in (49a), it is also possible that the reason why we get a narrow scope reading is not because there are two operators and the existential operator scopes lower than the universal operator, but because the object has no scopal properties at all. For example, the object diànyǐng ‘movie’ in (49a) also allows a kind interpretation (Chierchia 1998). On this analysis, everyone saw the sort of thing known as a movie. In this case, the object diànyǐng ‘movie’ does not really introduce a quantifier per se, and there is only one scopal operator, which is the universal quantifier introduced by the subject, that takes the entire VP of watching a movie. However, if the objects of S-le sentences are somehow treated as kinds then this still aligns with the idea that S-le sentences exhibit a kind of incorporation structure, since broadly speaking incorporation shows a distinct and weaker kind of referentiality than the corresponding object in a non-incorporation structure, as noted in the introduction.

2.6.3 Discourse opacity

One factor that influences discourse transparency/opacity of the object of S-le sentences is the specific anaphoric form used for referring back to the entity introduced by the incorporated noun. There are three anaphoric forms that are regularly used in Chinese, namely, zero anaphora/zero pronouns, overt pronouns, and repeated NPs (e.g. Pu 1997, Christensen 2000, Simpson et al. 2016). As shown below, although the object of S-le sentences shows discourse opacity to overt pronouns, it shows discourse transparency to zero pronouns. Another factor that influences discourse transparency/ opacity of the object of S-le sentences is the specific sentence structure that contains the overt anaphoric pronoun. As shown below, the object of S-le sentences shows discourse opacity if the sentence structure used for the overt anaphoric pronoun is a V-le sentence, whereas it shows discourse transparency if the sentence structure is a ba-sentence. Given the two factors, we restrict the conclusion

46 of discourse opacity for the object of S-le sentences to the condition of using an overt anaphoric pronoun in a V-le sentence. We consider zero pronouns and ba-sentences as special cases and do not draw the conclusion based on those data. Below I review some previous studies of zero pronouns and overt pronouns in Chinese, and then provide data to show that the object of the S-le sentence is somewhat less transparent than the corresponding object of a V-le sentence when the anaphora is an overt pronoun used in a V-le sentence.

2.6.3.1 Zero pronouns Li & Thompson (1979) found that the interpretation of the referents of zero pronouns in Chinese cannot be predicted by structural properties. Instead, the interpretation is inferred on the basis of pragmatic knowledge. The data they used are a number of excerpts from the two Chinese novels Water Margin (or All Men Are Brothers) and Romance of Confucian Scholars. They showed that although for some cases of zero pronouns the interpretation of the referent can be explained in terms of structural properties such as being the topic of a topic chain, or in terms of the semantics of the verb, there are cases where the interpretation of the zero pronoun can only be explained by pragmatics. Overall, the data they examined suggest that the fundamental strategy in the interpretation of the referent of a zero pronoun is pragmatic knowledge (Li & Thompson 1979: 320). One of the examples discussed in their work is shown in (50), which is a speech from Water Margin. The symbol “∅” represents a zero pronoun.

(50) ēn xiàng fàng-xīn. [Li & Thompson 1979: 320-321, (7)] kind sir rest-assured xiǎo-jiàng bì-yào qíng cǐ bèi-yì zhi zéi.

humble-servant will capture this treasonous REL.CL.marker brigand

zhè-jiān∅ 1 hé tā dòu shí, ∅2 gùn-fǎ yi-zì luàn-le.

just-now with he fight when fighting-technique already disoriented-ASP

lái-rì ∅3 jiào ēn xiàng kàn wǒ lì-zhǎn cǐ zéi. tomorrow let kind sir see I decapitate this brigand ‘Kind sir, rest assured. I will capture this treasonous brigand. Just now, as (I) was fighting with him, (he) was already getting disoriented in his fighting technique. Tomorrow, (I) will let you see me decapitate this brigand.’

Their discussion focuses on the second zero pronoun, which is indicated by ∅2. This example provides sufficient evidence for the claim that it is pragmatics rather than structural factors involving grammatical relations or grammatical functions that determines the interpretation of the referent of a zero pronoun. This is because the interpretation of the zero pronoun ∅2 in this broad discourse context is different from its interpretation when the sentence is interpreted in isolation. The judgment of the four consultants in Li & Thompson’s study is that the preferred interpretation for the zero pronoun ∅2 when the sentence is interpreted in isolation is that it refers to the subject (I) of the preceding subordinate clause as shown in (51). However, when the sentence is interpreted within the discourse context in (50), the interpretation for the zero pronoun ∅2 is that it refers to the oblique complement (with him) of the preceding clause.

(51) wǒ hé tā dòu shí, [Li & Thompson 1979: 321, (9)]

I with he fight when

∅ gùn-fǎ yi-zì luàn-le.

fighting-technique already disoriented-ASP ‘As I fought with him, I/he was already getting disoriented in [the] fighting technique.’

When the sentence is interpreted within the discourse context, pragmatic knowledge determines that the referent of the zero pronoun ∅2 is the brigand. According to Li & Thompson, the speaker began by saying that he would capture the brigand, which provides the hint that the brigand is not as good a warrior as the speaker is. It thus 48 follows that it must be the brigand not the speaker who became disoriented in the fighting technique in the earlier encounter. Li & Thompson (1981: 675) found that there are two syntactic environments in which zero pronouns cannot occur. One is the case when the NP immediately follows a coverb.10 For example, when the NP is the object of a preposition as in (52a), the NP cannot be a zero pronoun. An example of the grammatical counterpart is presented in (52b), where an overt pronoun is used. When the NP is the object of a ba-sentence, in which case it follows the coverb ba as in (53a) (see Section 3.4.2.2 for more discussion of ba-sentences), the NP cannot be a zero pronoun either. An example of the grammatical counterpart is presented in (53b).

(52) a. *wǒ gēn ∅ xué Yīngwén. [Li & Thompson 1981: 675, (39)] I with learn English

b. wǒ gēn tā xué Yīngwén. I with her learn English ‘I learn English from her.’

(53) a. *tā bǎ ∅ dài-lái-le. [Li & Thompson 1981: 675, (40)]

he BA bring-come-LE b. tā bǎ tā/Mǎlì dài-lái-le.

he BA her bring-come-LE

‘He brought her.’

The other case is the case when the NP is the pivot NP in a serial verb construction, which is the subject of the second verb and the direct object of the first verb. Two examples are presented here. The grammatical counterparts are provided in (b).

10 Li & Thompson (1981: 356) use the term coverb to refer to a set of morphemes that are used to introduce a noun phrase which usually is not the object of the main verb. Many of them play a grammatical function similar to prepositions in English. These morphemes include gēn 'with', cóng 'from', cháo 'facing', yán 'along', lí 'be apart from', zài 'at (locative)', bǎ (as for the ba-sentence), bǐ (the comparative morpheme), bèi (passive marker), etc. 49

(54) a. *tā mìnglìng∅ yòng dāozi. [Li & Thompson 1981: 675, (43)] he order use knife b. tā mìnglìng wǒ yòng dāozi. he order me use knife ‘He ordered me to use a knife.’

(55) a. *wǒ quàn ∅ bié hē jiǔ. [Li & Thompson 1981: 675, (44)] I advise not drink alcohol b. wǒ quàn tā bié hē jiǔ. I advise her not drink alcohol ‘I advised her not to drink alcohol.’

In the two cases, a certain syntactic constraint applies such that an overt noun phrase must be used, either a pronoun or a full NP.

2.6.3.2 Overt pronouns Li & Thompson (1981: 134) and Tai (1978: 297) found that the use of pronouns in Chinese is primarily for humans. It is rare to use third person pronouns to refer to animals and even rarer to use one when referring to inanimate entities. Consider the following example.

(56) A: nǐ xǐhuān nèi-běn shū ma? [Li & Thompson 1981: 134, (137)]

you like that-CL book Q ‘Do you like that book?’ B: wǒ xǐhuān(?? tā). I like it ‘I like it.’

An appropriate answer in this example is a sentence without the overt third person pronoun as the object. It is hardly acceptable to use the pronoun tā ‘it’ here to refer back to the book introduced in the question, which is inanimate. By contrast, if it is a human, then it is acceptable to use the third person pronoun to refer to it. Consider the 50 following example.

(57) A: nǐ xǐhuān Mǎlì ma?

you like Mary Q ‘Do you like Mary?’ B: wǒ xǐhuān( tā). I like her

‘I like her.’

As an answer, it is acceptable to use an overt pronoun tā ‘she, her’ or a zero pronoun. But there are exceptions. For example, some sentences that contain the third person pronoun used to form a genitive phrase with the particle de are acceptable. Consider the following two examples. The glosses for (58b) are mine and the translation is provided by Tai (1978).

(58) a. Hélán dìfāng fēicháng xiǎo, [Li & Thompson 1981: 135, (138)] Holland place unusually small

wǒmen zhīdào tā-de jīngjì wèntí bu jiǎndān.

we know it-DE economy problem not simple ‘Holland is a small country. We know its economic problems are not simple.’ b. fánshi tīng-le yí-ge xiāoxi, [Tai 1978: 299, (57)]

whenever hear-LE one-CL news nǐ děi zhīdào tā-de láiyuán.

you must know it-DE source ‘Whenever you hear news, you have to know its source.’

Another case where it is acceptable to use the third person pronoun to refer to an inanimate entity is the ba-sentence. Consider the pair of sentences in (59) which are used in the same discourse. The sentence (59a) is a V-le sentence and (59b) is a ba-sentence. The glosses are mine and the translation is provided by Tai (1978). (See

51 also Shen 2002 for similar pairs of sentences.)

(59) wǒ zuótiān mǎi-le zhè-běn shū,

I yesterday buy-LE this-CL book

a. *wǒ yǐjīng kàn-wán-le tā. [Tai 1978: 298, (53b)]

I already read-TERM-LE it b. wǒ yǐjīng bǎ tā kàn-wán-le. [Tai 1978: 299, (56)]

I already BA it read-TERM-LE ‘Yesterday I bought this book, and I have already finished reading it.’

The example above shows that while the object of the V-le sentence cannot be the third person pronoun that refers to an inanimate entity, it can be used for the object of the ba-sentence. (The semantic difference between V-le sentences and ba-sentences is discussed in Section 3.4.2.2.)

2.6.3.3 Referential capacity of the object of the S-le sentence As shown below, whether the object of an S-le sentence can serve as an antecedent for an anaphoric form in the subsequent discourse is influenced by two factors, namely the specific anaphoric form and the sentence structure that contains the anaphoric pronoun. A complete set of data are provided below to show how the two factors influence the judgment of sentence. Overall, the data suggest that the object of an S-le sentence has reduced referential capacity when compared with its equivalent in a V-le sentence. Note that we focus on human individuals as the objects of these sentences since the third person pronoun in Chinese is primarily for humans. The first pair of sentences is from Wang (2018), which is repeated in (60). This pair of sentences illustrates that the object of the S-le sentence cannot serve as an antecedent of an overt pronoun of the subject of the immediately following sentence.

The object of both sentences is rén ‘person’, so potentially it can serve as the antecedent of an anaphoric form in the subsequent discourse. Wang’s judgment is that the object of a V-le sentence as in (60a) can serve as the antecedent of the pronoun of

52 the subject of the following sentence, whereas the object of the S-le sentence as in (60b) cannot serve as an antecedent for the pronoun. For both sentences, the interpretation of the pronoun tā ‘he’ in the following sentence can be Zhāngsān, which is the subject of the preceding sentence.

(60) a. Zhāngsāni zuó-wǎn shā-le rénj. tāi/j shì ge lǎoshī. [Wang 2018: 227, (21)]

Zhangsan last-night kill-LE person he COP CL teacher

‘Zhangsani killed the personj last night. Hei/j is a teacher.’

b. Zhāngsāni zuó-wǎn shā rénj le. tāi/*j shì ge lǎoshī.

Zhangsan last-night kill person LE he COP CL teacher

‘Zhangsani killed someonej last night. Hei/*j is a teacher.’

Wang’s judgment supports the claim that the object of the S-le sentence has reduced referential capacity since in (60b) the pronoun tā ‘he’ refers to Zhāngsān but not the person killed by Zhangsan. The object of an S-le sentence, however, may serve as an antecedent for an overt pronoun in the subsequent discourse if the pronoun is used in a ba-sentence, a case Wang (2018) did not consider. Compare (61a) and (61b). Both discourses are acceptable. The object of the ba-sentence in both discourses, which is a pronoun, has a co-referential reading with the object of the preceding sentence.

(61) a. Zhāngsāni zuó-wǎn shā-le rénj, zhīhòu ∅i yòu bǎ tāj mái-le.

Zhangsan last-night kill-LE person afterwards also BA him bury-LE

b. Zhāngsāni zuó-wǎn shā rénj le. zhīhòu ∅i yòu bǎ tāj mái-le.

Zhangsan last-night kill person LE afterwards also BA him bury-LE

‘Zhangsani killed a personj last night. Afterwards, hei also buried himj.’

This pair of discourses shows that the object of the S-le sentence is not always opaque to an anaphoric pronoun contained in ba-sentence; instead, in this case, it is as transparent as the object of the V-le sentence in such a discourse.

Now consider the anaphoric form of zero pronouns as shown in (62). In this case, the interpretation for a zero pronoun is that it refers to the object of the preceding sentence regardless of whether the preceding sentence is a V-le sentence or an S-le sentence. So in this case, the object of the S-le sentence is not opaque for the zero pronoun.

(62) a. Zhāngsāni zuó-wǎn shā-le rénj. ∅*i/j shì ge lǎoshī.

Zhangsan last-night kill-LE person COP CL teacher

‘Zhangsani killed the personj last night. He*i/j is a teacher.’

b. Zhāngsāni zuó-wǎn shā rénj le. ∅*i/j shì gelǎoshī .

Zhangsan last-night kill person LE COP CL teacher

‘Zhangsani killed a personj last night. He*i/j is a teacher.’

When the zero pronoun is the object of the subsequent sentence, the object of the preceding S-le sentence can also serve as its antecedent as shown in (63b). This is the same for the case when the preceding sentence is a V-le sentence as shown in (63a).

(63) a. Zhāngsāni zuó-wǎn shā-le rénj. zhīhòu ∅i yòu mái-le ∅j.

Zhangsan last-night kill-LE person afterwards also bury-LE

b. Zhāngsāni zuó-wǎn shā rénj le. zhīhòu ∅i yòu mái-le ∅j.

Zhangsan last-night kill person LE afterwards also bury-LE

‘Zhangsani killed a personj last night. Afterwards, hei also buried himj.’

The data in (62) and (63) show that the object of the S-le sentence is transparent to a zero pronoun. However, as pointed out by Li & Thompson (1979), interpretation of zero pronouns is largely based on pragmatic knowledge. So we might just consider this a special case as well.

I have provided data above to show that the object of an S-le sentence is less transparent than the corresponding object of a V-le sentence if we restrict to the case where an overt anaphoric pronoun as opposed to a zero pronoun is used and it is

54 contained in a V-le sentence as opposed to a ba-sentence. The evidence is obtained from (60), which shows that the object of an S-le sentence cannot serve as an antecedent for an overt pronoun which is the subject of the following sentence, whereas the object of a V-le sentence can, though the interpretation in the latter case is ambiguous. Finally, we might want an explanation for the overall pattern of different anaphoric forms being used to refer back to incorporated objects: (a) overt pronouns in general cannot refer back to incorporated objects but covert pronouns can, (b) overt pronouns can refer back to incorporated objects when they are used in ba-sentences, and (c) covert pronouns (when used as an object) and ba-sentences are in complementary distribution. A possible explanation for this overall pattern (suggested to me by John Beavers, p.c.) is that in general incorporated objects cannot be the antecedent of an overt pronoun but can be the antecedent of a covert pronoun, possibly related to the fact discussed above that the latter largely have their antecedents determined pragmatically. However, for independent reasons ba-constructions disallow covert pronouns, and thus the paradigmatic distinction between them is neutralized. In this case the overt pronoun serves whatever functions the overt and covert pronouns would have served in other contexts, and allows interpretations more like covert pronouns. In this case it is expected that they might allow incorporated antecedents, though more work is needed to determine if this analysis is correct.

2.6.4 Well-establishedness effects

As discussed in Section 2.5.4, like many other properties, noun incorporation languages vary with respect to the extent to which the well-establishedness effects apply in the language. In Chinese, coinage of VPs of S-le sentences is not in general restricted to common activities as many of those expressions do not describe common activities. However, this does not mean well-establishedness effects do not exist at all

55 in the case of S-le sentences. We do see a class of S-le sentences that show well-establishedness effects. These expressions are what are called non-canonical objects in the literature (Lin 2001, Li 2014, Barrie & Li 2015, and Huang 2015). I discuss this set of data below as an illustration of how well-establishedness effects manifest in Chinese. Contra previous studies of non-canonical objects, I take a closer look at those data and provide a more specific explanation for them. I discuss two types of such VPs which cover most of the data discussed in previous studies. The first one is a case where the objects are actually metonyms, so they are in fact underlyingly canonical objects. The second one is a true case of non-canonical object and show that well- establishedness effects do exist for them, though I show that explicit contexts must be used to establish the relevant activity rather than some more general notion of being well-established.

2.6.4.1 Use of metonymy

One of the common examples used to show that Chinese allows non- canonical objects is a VP which is headed by the verb chī ‘eat’ and has a location noun as the direct object such as shítáng ‘canteen’ and fàndiàn ‘restaurant’ (e.g. Li 2014, Barrie & Li 2015, Huang 2015). The resulting expression is used in those studies as evidence to show that Chinese allows semantically unselected locations as objects. Meanwhile, some linguists, such as Zhang (2005, as cited in Li 2014: 300), noted that the location noun sometimes may not refer to the location of the eating activity. The example is repeated in (64).

(64) tā píngcháng dōu chī shítáng. he normally all eat canteen ‘He normally eats canteen (food).’

Zhang pointed out that shítáng ‘canteen’ in the above sentence means takeout food from the canteen, and the eating location can be somewhere else. Judgments from

56 more native speakers also show that the location noun refers to the food rather than the location. Consider the VPs in the V-le sentence (65a) and the S-le sentence (65b).

(65) a. tā chī-le shítáng.

he eat-LE canteen b. tā chī-shítáng le.

he eat-canteen LE

‘He ate the food provided by the canteen.’

Seven out of eight consultants accepted the two sentences in the context in which the food bought from the canteen was eaten at home, but also judged the two sentences unacceptable in the context in which the food eaten at the canteen was cooked at home. The other consultant could only accept the two sentences in the context in which the food is provided by and eaten at the canteen. Overall, the judgments suggest that the noun shítáng ‘canteen’ should not be interpreted as location of an eating event. Instead, it means the food provider and refers to the food provided, in other words, it is a canonical object. This in turn suggests that the formation of this kind of VP is a use of the figure of speech of metonymy (Lakoff & Johnson 1980: 35-40), and this is not necessarily evidence that Chinese allows non-canonical objects. The metonymy of food provider for food can be applied in a productive way. Another example is provided in (66) in which the object cānchē ‘food truck’ is literally a vehicle. The sentences in (66) can be accepted when given a rich context. For example, assuming the background that buying food from a food truck is a common activity in daily life, seven out of eight consultants accepted the two sentences in a context that the food was bought from a food truck.

(66) a. tā chī-le cānchē.

he eat-LE food.truck

b. tā chī-cānchē le.

he eat-food.truck LE ‘He ate the food provided by the food truck.’

This suggests that a food provider can be the object of the verb chī ‘eat’, and it is metonymically interpreted as the food. The examples discussed above suggest that some of those so-called non- canonical objects are just applications of metonymy, and they should be considered canonical objects selected for by the verbs.

2.6.4.2 Formation of VPs with non-canonical objects in contexts Chinese, however, does allow VPs with non-canonical objects which are usually expressed as PPs in their English counterparts. The evidence that these NPs are objects rather than adjuncts is that they cannot co-occur with canonical objects in the postverbal object position, that is, they are in complementary distribution (Barrie & Li 2015). Those non-canonical objects may bear various thematic relations such as instrument, location, time, reason, etc. (Li 2014, Lin 2001, Huang et al. 2009: 46-75). (For the sake of simplicity, the discussion below focuses on the thematic relation of instrument. The discussion also applies to other thematic relations.) Many of those expressions show well-establishedness effects, as they describe common activities.

But I show that what explicit context needs to be used for those expressions to be acceptable. Barrie & Li (2015) claimed that forming VPs with non-canonical objects is not highly productive in Chinese, and only those VPs that denote institutionalized activities are acceptable, while equivalent VPs that denote similar but less common activities are unacceptable. Consider the following example from them (p. 171).

(67) a. nǐ chī zhè-shuāng kuàizi ba!

you eat this-CL chopsticks SFP ‘You eat with this pair of chopsticks!’

b. *nǐ chī zhè-bǎ chāzi ba!

you eat this-CL fork SFP ‘(Intended)You eat with this fork!’

According to Barrie & Li, the VP chī kuàizi ‘to eat with chopsticks’ is acceptable because it is a culturally common activity, whereas chī chāzi ‘to eat with a fork’ is unacceptable because it is not a culturally common activity.

However, their judgments for (67) are different from the judgments of myself and my consultants. Both sentences received similar acceptability judgments from my consultants when the judgments were elicited in this context: You hand over a pair of chopsticks or a fork to your friend when you are about to eat. With this context, 25 of 28 consultants accepted (67a), and 20 of them accepted (67b). (The slightly higher acceptability of (67a) can be explained by a higher frequency of use.) This example suggests that the narrow context in which the activity of eating food is understood and the instrument is emphasized, licensing the non-canonical object. Indeed, narrow context plays an important role in licensing the non-canonical objects more broadly. What is notable about (67a) is that it is an imperative sentence, which is typically used in narrow contexts like the one mentioned in the last paragraph. So the acceptability of (67a) is actually largely due to the narrow context brought to mind when reading the sentence. There is further evidence that it is the narrow context and not something more institutionalized that makes such expressions acceptable. The evidence is that a narrow context can significantly increase acceptability of non-imperative S-le sentences and V-le sentences containing the VP chī kuàizi ‘eat with chopsticks’, which are largely unacceptable out of context. First consider the pair of sentences in (68) out of context. None of my twelve consultants accepted the V-le sentence (68a), as they default to the literal meaning, that is, he ate the chopsticks. Two of them accepted the S-le sentence (68b), which might be because the S-le sentence is restricted to a particular context that I call an informative context (see Section 3.3 and 3.4), and thus it is likely that the two consultants were thinking of

59 a certain informative, narrow context when judging this sentence.

(68) a. *tā chī-le kuàizi.

he eat-LE chopsticks b. ??tā chī-kuàizi le.

he eat-chopsticks LE ‘(Intended)He ate with chopsticks.’

By contrast, with a narrow context supplied, acceptability of the two sentences is significantly increased. Consider the dialogue below. Seven of the same twelve consultants accepted the V-le sentence and eight accepted the S-le sentence.

(69) A: zuótiān wǎnyè-shàng yǒu bùshǎo wàiguórén,

yesterday dinner-at have many foreigner gěi tāmen zhǔnbèi dāochā le ma?

for them prepare knife.fork LE Q ‘There were many foreigners at the dinner yesterday. Did you prepare knives and forks for them?’ B: méiyǒu. ?tāmen chī-le kuàizi. /?tāmen chī-kuàizi le. gēn wǒmen yíyàng.

not.have they eat-LE chopsticks they eat-chopsticks LE with us same

‘No. They ate with chopsticks. Like us.’

The unacceptability of VP chī kuàizi ‘eat with chopsticks’ out of context and its increased acceptability in a narrow context suggest that context plays a more important role in licensing the non-canonical object than being an institutionalized or common activity in any broader sense. We may consider another example of a VP which denotes an uncommon activity in daily life. The dialogue is presented in (70). The situation is that two conference staff are preparing for a speech given by some speaker. Speaker B uses the

VP jiǎng huàtǒng ‘speak with a microphone’ and the non-canonical object huàtǒng ‘microphone’ is the instrument involved in the activity. Eleven out of 20 consultants accepted the sentence in this context, six of them judged it not quite acceptable, and three of them rejected it.

(70) A: zuò zài hòupái de rén kěnéng tīng bú dào.

sit LOC back.row DE people may listen NEG arrive

‘People sitting in the back row may not hear (him).’ B: yǒu huàtǒng ma? ?ràng tā jiǎng huàtǒng.

have microphone Q let him speak microphone ‘Do we have a microphone? Let him speak with a microphone.’

The various judgments suggest that the VP jiǎng huàtǒng ‘speak with a microphone’ is not a default or natural expression for the intended meaning – the default expression is that huàtǒng ‘microphone’ is expressed as a syntactic adjunct.11 However, the fact that it is largely accepted in this context suggests that a narrow context can license such a VP with a non-canonical object. The two VP examples discussed above suggest that formation of VPs with non-canonical objects is at least not solely licensed by being a common activity but contexts are also needed to establish them as a feature of this effect. It is also noticeable that those VPs with non-canonical objects can also be used in V-le sentences if there is a supporting context. So well-establishedness effects are not exclusive for noun incorporation.

2.7 SUMMARY

The purpose of this chapter is to provide morpho-syntactic and semantic

11 The expression is presented as follows, where huàtǒng ‘microphone’ is a syntactic adjunct.

(1) yòng huàtǒng jiǎng with microphone speak ‘to speak with a microphone’ 61 evidence for noun incorporation of S-le sentences. Those properties are summarized below. By comparing S-le sentences with other noun incorporation languages against those properties, we can see that the evidence for noun incorporation of S-le sentences is relatively robust. I have shown that the object of S-le sentences largely shows the morpho- syntactic properties of noun incorporation. At the surface level it shows a close relationship with the verb and is restricted in nominal modification, which shows a hierarchy of acceptability with different types of modifications, as repeated in (71).

(71) bare nouns > NP-internal dependents > numerals > quantifiers, article

The morpho-syntactic properties are summarized in Table 1 with a comparison to some other noun incorporation languages.

morph-syntactic quantifiers/ number case NP-internal properties articles languages marking marking modification modification

Chinese S-le yes N/A yes no

W. Greenlandic yes no yes no

not Hungarian yes yes no mentioned

Hindi yes no yes no

not Niuean no yes no mentioned

Table 1: Morpho-syntactic properties of noun incorporation in Chinese, West Greenlandic, Hungarian, Hindi, and Niuean.

Note that although a general tendency for incorporated nouns is that they tend to be bare nouns (Mithun 2000) or small NPs, showing number marking and NP-internal modifications clearly do not provide counterevidence for noun incorporation, as many of those languages list in Table 1 show those properties. The overall picture shown in 62 the table supports the claim that the S-le sentence is a case of noun incorporation. I have also shown that the S-le sentence has four semantic properties often used to characterize incorporation in the literature, although the strength of the evidence varies from property to property. Those properties are summarized in Table 2 together with a summary of those properties in some other languages for comparison. Note that some of the properties for some of these languages are not discussed in the literature, which I indicate as not mentioned in the table.

semantic number narrow discourse well-establishedness properties languages neutrality scope opacity effects

for overt Chinese S-le (yes) yes yes (but productive) pronoun

W. Greenlandic yes yes no not mentioned

for singular noun & Hungarian yes yes not mentioned overt pronoun

for singular Hindi yes yes yes (but productive) noun

depends on not Niuean yes verb’s yes (but productive) mentioned meaning

Table 2: Semantic properties of noun incorporation in Chinese, West Greenlandic, Hungarian, Hindi, and Niuean.

As shown in the table, Chinese S-le sentences show all four of these properties. But note that the details of some of those properties are ignored in this table. For example, although bare noun objects of S-le sentences are number neutral, we should bear in mind that bare nouns always allow number neutral interpretations in Chinese

63 regardless of sentence structures. We have evidence for narrow scope reading relative to modal verbs and universal quantifier but we do not have direct evidence for narrow scope relative to negation. The object of an S-le sentence is discourse opaque unless the anaphora is zero anaphora or the anaphoric pronoun is the object of a ba-sentence. Some S-le sentences show well-establishedness effects, but S-le sentences are productive in general.

2.8 INCORPORATION AND COMPOUNDING IN CHINESE

In the literature on incorporation (e.g. Rosen 1989, van Geenhoven 1998, Haugen 2008, Mithun 2010, Barrie & Mathieu 2016), incorporation is considered a distinct morpho-syntactic process from compounding, though some linguists (e.g. Mithun 2010) use the term compounding in a broader sense such that incorporation is classified as a particular kind of compounding. The arguments provided in the literature for a distinction between the two morpho-syntactic processes are based on evidence that nominals in the two forms show contrasting properties. For example, the nominal of incorporation is referential and can sometimes be referred back to by a pronoun in discourse. An example of Ojibwe is provided below (Barrie & Mathieu 2016).

(72) ngii-moonahapnii mii dash ngii-giziibiiginigan

n- gii- moonah -apnii -e mii dash n- gii- giziibiiginig -an

1- PST- dig -potato -VAI and then 1 PST- wash -3PL ‘I dug up potatoes, and then I washed them.’

However, the noun stem in a compound cannot be the antecedent of a pronoun in subsequent discourse. For example, the pronoun it in the following sentence cannot refer back to the noun dog in the compound doghouse (Barrie & Mathieu 2016: 4,

(5)).

(73) #The new doghouse seems to disturb it (the dog).

The contrast in discourse opacity/transparency as illustrated by the two examples above suggests that incorporation and compounding are distinct, though incorporated nouns are sometimes also discourse opaque (see Section 2.5.3). In this section, I provide Chinese data to show that there is a separate process akin to compounding which is distinct from incorporation. The argument for the distinction, however, is not based on evidence regarding the property of discourse transparency/opacity of the nominal. Instead, I aim to show that a compound word is a word and thus stored in the lexicon and needs to be memorized, whereas incorporation is apparently an independent rule (whether a lexical or grammatical rule). Furthermore, some types of compounding show word orders distinct from incorporation.

Since Chinese is rich in compounds (Lu 1964, Chao 1968, Li & Thompson 1981), the data discussed below include only two types of compounding, which is sufficient to show that compounding and incorporation are distinct phenomena.

2.8.1 Verbal compounds formed with a dummy object

One type of compounding in Chinese uses a V + N template. The data of such combinations are further classified into subgroups according to the degree of idiosyncrasy of the compound’s meaning (T. Wang 2001). The N of a more idiosyncratic combination is more likely to be a bound morpheme and less acceptable when used as the subject of a passive sentence. In a less idiosyncratic case, the N is more likely to be a free morpheme and more acceptable as the subject of a passive sentence. A compounding V + N and a phrasal V + Noun Phrase share many syntactic properties, such as aspectual marking, as shown below.

These compound words are also known as líhécí ‘separable words’ in the traditional literature of Chinese linguistics (e.g. Lu 1957, Lü 1979, Yu 1989, T. Wang

2001). The term líhécí was coined based on the empirical fact that those compounds, though denoting unitary concepts like a typical word and thus expected to be used as a whole unit, can be separated by some linguistic elements inserted between the V and N, including aspectual markers, modifiers, demonstratives, possessor, classifiers, etc., resulting in a linguistic form which contains an underlying concept expressed by V + N but formally realized in a discrete way. Given this fact, some other Chinese linguists (e.g. Hong 1957, Shi 1983) argue that these combinations are not words but phrases, or at least duǎnyǔcí ‘phrasal words’. The data considered below are V + N compounds where the noun root is highly integrated in the sense that it does not seem to contribute a compositional meaning and the combination is idiosyncratic. In Sybesma (1999), these noun roots are called ‘dummy objects’. Some examples are provided as follows.

(74) a. yóu-yǒng

swimv-swimn ‘to swim’ b. zǒu-lù walk-road ‘to walk’ c. pǎo-bù

run-step

‘to run’

The example in (74a) is a compound that consists of a verb and a nominal cognate. In the other two examples, the noun roots seem to contribute senses to the meanings of the complex units but do not seem to refer. The noun roots in the three examples are highly integrated.

Two pieces of evidence are provided below to show that these complex units are indeed compound words. The first piece of evidence is that the meanings of these

66 compounds are totally changed if the noun roots are changed, in which case new compound words are formed to denote different concepts. The corresponding examples are provided in (75).

(75) a. yóu-jiē

swimv-street ‘to parade’ b. zǒu-xíng walk-shape ‘to be out of shape’ c. pǎo-tí run-topic

‘to be off the topic’

The three examples above show that the change of noun roots changes the meanings of compound words dramatically, suggesting that meanings of these combinations are somewhat idiomatic and they should be considered compounds. The second piece of evidence is that these dummy objects normally cannot be the subject of a passive, in which case they give rise to coerced readings which are barely acceptable. This contrasts with the case of a typical object, suggesting that the dummy object is not a subcategorized NP. An example is provided in (76b), and (76a), which contains a typical object, is provided for comparison.

(76) a. píngguǒ bèi tā chī-le.

apple PASS him eat-LE ‘The apple was eaten by him.’

b. ??yǒng bèi tā yóu-le.

swimN PASS him swimV-LE ‘The swimming was done by him.’ 67

The contrastive acceptability of the sentence with a dummy object and the sentence with a typical object in the passive form suggests that the noun root is used to form a compound.12 The data provided above support the claim that these complex units in (74) are compound words as they denote unitary concepts and have idiomatic meanings. Such a word-formation process is distinct from incorporation. In the case of incorporation, the meaning is largely interpreted compositionally (though see Chapter 3). If the VP itself has an idiomatic meaning, then it is taken into consideration when composing the meaning of an incorporation sentence containing such an idiomatic VP as well as when composing the meaning for a non-incorporation sentence. But incorporation itself does not create idiomatic meaning. The idiomatic meaning is created in compounding. Consider an example in (77) which shows that the idiosyncratic meaning of a compound word can carry over to both the S-le sentence as well as the

V-le sentence. The VP in this example is the compound yóu-yǒng (swimv-swimn) ‘swim’.

(77) a. tā yóu-le yǒng.

he swimv-LE swimn b. tā yóu-yǒng le.

he swimv-swimn LE ‘He swam.’

This example shows that the idiomatic meaning of swimming is not created in incorporation (77a) since such an idiomatic meaning shows up in both sentence constructions. The idiomatic meaning must therefore be created in an independent word- formation process. Note that this is also true for other V + N compounds. So

12 Sentence (76b) might be accepted by some native speakers in certain contexts. For example, there was a chance to swim in a fancy swimming pool for free, but it only gave the chance to one person. Suppose two people were competing for this chance. Then the person who lost it might say (76b) to mean that the other person got the chance to swim for free (or, as the translation suggests, the swimming was done by him). 68 this type of compound word provides evidence for the claim that compounding and incorporation are two distinct phenomena.

2.8.2 Verbal compounds formed with a modifier

Another type of compound word discussed here is a combination of a noun root naming an instrument plus a verb root, in which the instrument modifies the action denoted by the verb root. The instrument modifier precedes the head verb, and the resulting compounds are transitive. Some examples are shown in (78).

(78) a. qiāng-jī gun-shoot ‘to shoot with a gun’ b. pào-hōng

cannon-bomb ‘to bomb with a cannon’ c. zhēn-qīng needle-clean ‘to clean with a needle’

Three pieces of evidence are provided below to show that these complex units above are compound words with idiosyncratic meanings. First, such combinations are not productive, suggesting that those acceptable combinations are idiomatic and stored in the lexicon. Some unacceptable examples are provided as follows.

(79) a. *kuàizi-chī chopsticks-eat ‘to eat with chopsticks’

b. *dāo-qiē knife-cut ‘to cut with a knife’ c. *bǐ-xiě pen-write ‘to write with a pen’

The intended meanings of these combinations are common activities, so we would expect them to be acceptable expressions since the examples in (78) suggest that instrument + V is a possible word formation pattern. The ungrammaticality of the examples in (79) suggests that the examples in (78) are stored in the lexicon. Another piece of evidence that these examples in (78) are stored in the lexicon is that the verb roots are fixed and cannot be replaced by semantically equivalent verbs. The monosyllabic verb roots in those combinations have bisyllabic equivalents. However, they cannot replace the monosyllabic verb roots in (78). The corresponding examples are shown in (80).

(80) a. *qiāng-shèjī gun-shoot ‘to shoot with a gun’

b. *pào-hōngzhà cannon-bomb ‘to bomb with a cannon’ c. *zhēn-qīnglǐ needle-clean ‘to clean with a needle’

The unacceptability of the three examples above suggests that those complex units in (78) are formed in an idiomatic way and stored in the lexicon. The third piece of evidence is the compatibility of such a verbal compound 70 with a syntactic instrument adjunct. An example is provided as follows.

(81) tā zài yòng nà-bǎ láifúqiāng qiāng-jī zuìfàn.

he PROG with that-CL rifle gun-shoot criminal ‘He is gun-shooting the criminal with that rifle.’

This suggests that the noun root qiāng ‘gun’ in the verbal compound does not serve as a syntactic adjunct but only a morphological component of a compound verb. The data provided above show that the combinations in (78) are words with idiosyncratic meanings that are stored in the lexicon. The derived compound verbs are transitive. Like normal transitive verbs, they can be used in an incorporation sentence as well as a non-incorporation sentence, and their idiosyncratic meanings carry over to the two sentence structures equally. In the following example, (82a) is a non-incorporation sentence while (82b) is an incorporation sentence. In both sentences, the VP has the same meaning, that is, to shoot a criminal with a gun.

(82) a. tā qiāng-jī-le zuìfàn.

he gun-shoot-LE criminal b. tā qiāng-jī-zuìfàn le.

he gun-shoot-criminal LE ‘He shot a criminal with a gun.’

The pair of sentences above suggests that it is compounding and not incorporation that determines the idiosyncratic meaning. Thus the data in (78)-(82) also prove that compounding and incorporation are distinct phenomena.

2.8.3 Summary

In this section, I provided two sets of data to argue that incorporation is a distinct phenomenon from compounding. The two sets of data are two types of compounding words in Chinese. For each type, I first provided evidence to show that

71 these expressions are indeed compounding words and stored in the lexicon. Then I showed that the idiosyncratic meanings of the compounding verbs carry over equally to the incorporation sentence and the non-incorporated sentence, suggesting that the idiosyncratic meaning is not determined by incorporation. Therefore, incorporation and compounding, as claimed in some studies, are indeed distinct phenomena.

2.9 CONCLUSION

I argued in this chapter that the VP contained in the S-le sentence is a case of incorporation. I showed that the S-le sentence also has formal characteristics and semantic properties of incorporation discussed in prior literature. The data provide sufficient evidence for the claim. Morpho-syntactically, the object of the S-le sentence is closer to the verb, and its NP-internal modification is subject to a condition of size. More importantly, I have shown that the object of the S-le sentence and the object of the V-le sentence are syntactically distinct as the former shows restrictions in terms of possible modification of the object, whereas such restrictions do not apply in the latter case. I have also shown that S-le sentences have four semantic properties that we have found in noun incorporation in other languages, namely, number neutrality, narrow scope, discourse opacity, and well-establishedness effects. In addition, I have provided Chinese data to show that incorporation and compounding are two distinct phenomena. Those data together provide evidence for the claim that S-le sentences involve incorporation.

Chapter 3: Discourse Uses of S-le Sentences

3.1 INTRODUCTION

In the last chapter I argued that the VP in the S-le structure should be considered a case of noun incorporation, as I showed that it exhibits some of the formal and semantic properties that are typically associated with noun incorporation.

In this chapter, I present a semantic analysis for the S-le sentence to account for certain semantic phenomena associated with S-le sentences. The analysis I propose here concerns a different aspect of the phenomenon of noun incorporation from previous semantic studies (e.g. van Geenhoven 1998, Chung

& Ladusaw 2004, Dayal 2011). A major concern of those prior studies is the semantics of the incorporated noun and the relation it bears with the incorporating verb. Accordingly, prior analyses proposed semantic analyses which, as pointed out by Borik & Gehrke (2015), are essentially the same in that the incorporated noun denotes a property and modifies the incorporating verb so that the meaning of VP gets narrowed down. More recently, Mueller-Reichau (2015) extended the semantic study of noun incorporation to the sentence level. He identified the imperfective VP with a bare noun object in Russian, in its factual use, as a case of (pseudo) noun incorporation and the corresponding perfective VP as the syntactic non-incorporation counterpart, and he proposed an information structural analysis of the imperfective VP and perfective VP to account for the four semantic properties of noun incorporation that are also found in the imperfective VP. The current study, though not cast in the framework adopted by Mueller-Reichau, also focuses on the semantics of the sentence that contains noun incorporation (i.e. the S-le sentence) as opposed to the semantics of the incorporated noun, which I assume the property-denoting analysis proposed in prior studies.

What I am going to do in this chapter is to show that we might understand S-le sentences as an expression that is specifically used for providing hearer-new information. When the speaker uses an S-le sentence as opposed to any other sentence 73 structure to provide new information, he makes it more apparent that he believes that the information is new to the hearer and so is his intention to inform the hearer of the new proposition. We can see this from the fact that S-le sentences normally require that the proposition is new to the hearer while other sentence structures have no such requirement. Consider an example in (1). The two discourses only differ in that the first sentence of (1a) is an S-le sentence and that of (1b) is a V-le sentence. They convey the same information as indicated by the translation.

(1) a. tā zǎoshàng -sǎo xuě le. zhīhòu qù-le xuéxiào.

he morning sweep-snow LE afterwards go-LE school b. tā zǎoshàng sǎo-le xuě. zhīhòu qù-le xuéxiào.

he morning sweep-LE snow afterwards go-LE school ‘He swept the snow in the morning. Afterwards, he went to school.’

However, (1a) is generally natural only in a context in which the hearer does not know of the event of sweeping the snow, and unacceptable if the hearer already knows it (with one caveat I discuss below), whereas (1b) is acceptable regardless of whether the hearer knows the event of sweeping the snow or not – in the case when the hearer already knows the event, the V-le sentence is used to remind this event in order to introduce a familiar temporal point that the new introduced event of going to school can relate to.

To show how we may understand the informativity associated with S-le sentence, I examine below how S-le sentences are used in discourse context and specifically how they are used in two types of discourse contexts. In one type of context in which the proposition is new to the hearer, an informative context, S-le sentences are perfectly acceptable. In contrast, in a non-informative context where the proposition is already known to the hearer, S-le sentences normally are unacceptable. But they are not totally banned in such a context. They can be acceptable provided the speaker uses them for a purpose of expressing evaluative meanings. My tentative 74 conclusion is that S-le sentences presuppose that the information is hearer-new and require an informative context, and we may explain the emotiveness meaning arising in a non-informative context in terms of deriving non-compositional meanings in an otherwise inappropriate context in a kind of Gricean way. Prior work has also noted a condition of hearer-new information as one condition on S-le sentences (e.g. Sybesma 1999, Liu 2002, among others). However, it has not been fully appreciated in these studies that hearer-new information is at the core of the use of S-le sentences. This is largely because S-le sentences have various particular interpretations which may not seem at first blush to involve hearer-newness. Yet I will show that those sentences are also used in informative context and they can be accounted for in terms of hearer-new information as well. In addition, other prior studies have analyzed S-le sentences in light of the notions of tense or aspect. However, I will show that tense and aspect are actually irrelevant for the semantics of

S-le sentences. Finally, not all informative contexts license S-le sentences. The kind of informative context appropriate for S-le sentences is a general one. When the informative contexts become more complicated, S-le sentences are blocked by other sentence structures. I show below two types of such informative contexts, one which licenses a structure called the Object Preposing structure and the other one the ba-sentence. This chapter ends with a big picture of the four sentence structures: the S-le sentence, the V-le sentence, the Object Preposing sentence, and the ba-sentence. Throughout I show that the V-le sentence is a relatively neutral structure in the sense that it does not carry any of the kinds of presuppositions the other three sentence structures do. The analysis I propose for each sentence type is theoretically neutral, and I will discuss some theoretical analyses for those sentence structures in the literature. The rest of the chapter is organized as follows. In Section 3.2, I delineate away two cases of using le as a sentence final morpheme distinct from the data concerned in 75 the current study. One is what I call inchoative le and the other one is known as phase- complement le in the literature. In Section 3.3, I review some previous studies of S-le sentences with a focus on those that also considered hearer-newness for S-le sentences, and show that hearer-new information can account for all S-le sentences. I also discuss two types of cases where S-le sentences are used in non-informative contexts and give rise to pragmatic effects. I suggest that these can be understood if we assume that S-le sentences express hearer-newness and require informative contexts. Meanwhile, following Soh (2009), I suggest that informativeness is analyzed as a kind of presupposition characterized in terms of common ground (Stalnaker 1998, 1999, 2002). I propose that S-le sentences presuppose that the proposition is new to the hearer, or is not accepted on the common ground. In Section 3.4, I present an argument that S-le sentences are free of grammatical aspect, and suggest that one trend of discussion of S-le sentences in terms of tense and aspect in the literature is probably on the wrong track of studying S-le sentences. In Section 3.5, I discuss discourse uses of four sentence structures and show that S-le sentences are blocked by other sentence structures in some narrower subtypes of informative contexts. A summary of this chapter is provided in Section 3.6.

3.2 ANOTHER TWO CASES OF SENTENCE FINAL LE

There are many sentences in Chinese that end in le, however, many of them are not the kind of S-le sentences we are concerned with in the current study and thus they must be excluded. Two cases are discussed below where the final le does not express informativeness. In one case, le is a suffix that expresses change of state, which I call inchoative le. In the other case, le is used for a compounding purpose, which is called phase-complement le in the literature (e.g. Chao 1968). The two uses of le are not as productive as the informative le and they are subject to some lexical restrictions and thus can easily be distinguished.

3.2.1 The inchoative le

The inchoative le is largely restricted to adjectival predicates. It is used as a derivational morpheme like -en in English that turns the adjectival root into a change-of-state word. Two examples are provided below. The two (a) sentences describe a property of the subject and the two (b) sentences express a change of the property of the subject.

(2) a. mǎlù hěn kuān. road very wide ‘The road is wide.’ b. mǎlù kuān-le.

road wide-LE ‘The road widened.’

(3) a. tā hěn gāo. he very tall ‘He is tall.’ b. tā gāo-le.

he tall-LE ‘He is taller.’ Or ‘He has grown.’

Both (2b) and (3b) entail a change of state with respect to the property. They do not require the hearer to have any knowledge about the object’s property before the change in order to make the utterance felicitous. Note that such sentences may also take an object. Consider an example from Sybesma (1999) as in (4). The NP wǔ-gōngfēn ‘five centimeters’ is expressed as the syntactic object.

(4) Zhāngsān gāo-le wǔ-gōngfēn. [Sybesma 1999: 68, (19b)]

Zhangsan tall-LE five-centimeter ‘Zhangsan is 5 centimeter taller: has grown 5 centimeter.’

I consider sentences like (2b), (3b), and (4) ambiguous (also see Soh 2009). Depending on the context, such sentences might be analyzed at the underlying level as a structure that consists of not only the inchoative le but also the perfective aspect

V-le or the final particle S-le or both. But they all surface as one le due to phonological restrictions.

3.2.2 The phase-complement le

The phase-complement le is also distinguished from the other uses of le in the literature (e.g. Chao 1968, Lü 1980, Li & Thompson 1981, Sybesma 1999, Soh 2009).

The term phase complement is due to Chao (1968), who described a phase complement as a morpheme that “expresses the phase of an action in the ﬁrst verb rather than some result in the action or goal” (p. 446). According to Li & Thompson (1981), phase complement le expresses “something more like the type of action described by the ﬁrst verb or the degree to which it is carried out than its result” (p. 65). Sybesma (1999) treated the phase-complement le as a neutral telic marker which “indicates that an action has been accomplished more or less successfully, but that they … do not specify the resulting state” (p. 71). It seems to me that the phase-complement le is a morpheme that is used for the purpose of compounding or lexicalization, and it does not contribute any meaning to the larger linguistic units. On the one hand, as noted in the literature, the use of the phase-complement le is not productive and it only applies to a restricted set of monosyllabic stems/verbs.1 On the other hand, what has not been noted in the literature is that some of those

1 Lü (1980: 315) listed 28 monosyllabic verbs that can compound with the phase-complement le. The list includes: wàng ‘forget’, diū ‘throw/get ride of’, guān 'close, shut', chī 'eat', rēng‘throw, get rid of’, sǎ ‘spill’, tú‘scribble’, cǎi ‘step on’, shā ‘kill’, pèng ‘bump’, qiē ‘cut’, tūn ‘swallow’, chōng ‘flush’, huǐ ‘destroy’, etc. 78 monosyllabic stems/verbs have bisyllabic counterparts and le becomes optional for the bisyllabic counterparts. Consider an example in (5) which contains the monosyllabic wàng ‘forget’. The sentence would be unacceptable if the particle le is deleted.

(5) tāz ǒngshì wàng*(-le) nǐ-de míngzì. [Shi 1990: 109, (26a)]

he always forget-LE your name

‘He always forgets your name.’

The monosyllabic wàng ‘forget’ has a bisyllabic counterpart wàngjì ‘forget’, in which case the phase-complement le becomes optional.

(6) tāz ǒngshì wàngjì(-le) nǐ-de míngzì.

he always forget-LE your name

‘He always forgets your name.’

Soh (2009) argued that the phase-complement le should be treated independently from other uses of le. In particular, she showed that it is distinct from the perfective aspect marker V-le. Her evidence is that the perfective -le usually cannot be used in a sentence that has future verbal elements such as huì ‘will’ and dǎsuàn ‘plan’ as in (7b), however, the phase-complement le is obligatory when the sentence has such future verbal elements as in (7a).

(7) a. tā huì/dǎsuàn wàng*(-le) nǐ-de míngzì. [Soh 2009: 627, (10)]

he will/plan forget-LE your name ‘He will/plans to forget your name.’ b. tā huì/dǎsuàn xiě(*-le) yī-běn shū. [Soh 2009: 627, (9a)]

he will/plan write-LE one-CL book

‘He will/plans to write a book.’

Note that in this case the bisyllabic wàngjì ‘forget’ also allows le, suggesting that it is 79 a case of the phase-complement le rather than the perfective -le, and it is optional, as shown in (8).

(8) tā huì/dǎsuàn wàngjì(-le) nǐ-de míngzì.

he will/plan forget-LE your name ‘He will/plans to forget your name.’

The phonological restriction against two les in sequence also applies in this case (Chao 1968, Li & Thompson 1981, Soh 2009). For example, in some cases, a surface le might represent both the phase-complement le and the perfective aspect marker V-le but the two les cannot surface together. The phase-complement le does not express informativeness and thus those data should be distinguished from S-le sentences.

3.3 S-LE SENTENCES: HEARER-NEW INFORMATION

The S-le sentence has received considerable attention in the literature of Chinese linguistics. The discussion largely follows two directions. One direction focuses on temporal properties of S-le sentences, claiming that such sentences express a certain aspectual meaning or past tense (e.g. Chao 1968, Rohsenow 1978, Li & Thompson 1981, Ross 1995, Liu 2002, Bao 2005). The other direction focuses on the contextual conditions and pragmatic effects associated with the use of S-le sentences

(e.g. Chao 1968, Li & Thompson 1981, Sybesma 1999, Liu 2002, Soh 2009). In this vein, prior studies have noted that many S-le sentences are used to express hearer-new information and some S-le sentences may allow other interpretations. In the current study, I suggest that the second direction is the right one to study the semantics of S-le sentences. Later in Section 3.4, I show that studies of the tense or aspectual properties of S-le sentences do not account for their distributions, and thus I suggest that we should not study S-le sentences in terms of aspectual properties including tense and aspect.

Following Soh’s (2009) analysis of S-le sentences using the framework of common ground developed by Stalnaker (1978, 1998, 1999, 2002), I also adopt this framework for analyzing the notion of hearer-new information or informativeness. The theory of common ground is a formal theory of speaker presupposition. According to Stalnaker, a discourse develops against an ever-changing background made up of a set of propositions that have been accepted by the participants, that is, the common ground. When a proposition is asserted, it has such an effect that it will update the common ground by adding the content of the proposition if there are no objections from the other participant. When a speaker presupposes something, he takes it for granted, or at least to act as if he takes it for granted, as background information in the common ground. I suggest that we may analyze the notion informativeness in terms of this kind of presupposition. By using an S-le sentence, the speaker presupposes that the hearer does not know the proposition expressed by the sentence. In terms of common ground, we may more specifically say that it is not part of the common ground that the proposition is true, which could correspond to a case in which ¬p is accepted as true or a case in which there is no presupposition about P.2 It is notable that this use of the term presupposition has a different sense than the one that is normally used in the literature (Beaver & Geurts 2014), which is understood as a type of inference triggered by certain expressions with certain

2 Note that we might want to say that S-le sentences illustrate a case of antipresupposition as described by Percus (2006). According to Percus, to presuppose a proposition means the speaker takes it for granted as mutually accepted information (e.g. Mary knows that Jane is pregnant presupposes that Jane is pregnant.). Antipresupposition is a kind of presupposition for the opposite case where the use of some sentences requires that a certain proposition should not be taken for granted (e.g. Mary thinks that Jane is pregnant antipresupposes that Jane is pregnant). In the case of S-le sentences, I will ultimately claim that they presuppose that the information is hearer-new and require that the proposition should not be taken for granted. So in terms of the distinction proposed by Percus, we might want to say that S-le sentences illustrate a case of antipresupposition rather than (canonical) presupposition. However, in the current study, I use the more general term presupposition instead of antipresuppotion for analyzing S-le sentences. My concern is that expressions involving antipresupposition, as described by Percus (2006), come as a complementary element of a pair that implies there is a corresponding expression that involvs canonical presupposition (e.g. know and think). However, there are no sentence structures that would form such a pair with S-le sentences which involve (canonical) presuppositons of the same propositions – V-le sentences do not inherently presuppose the propositions. Since there is no such complemtary expression for S-le sentences, they might not really be a case of antipresuppostion in the exact sense described by Percus. So I use the more general term presupposition instead of antipresupposition to avoid any misuse of the term. 81 properties (e.g. projection). In the framework of Discourse Representation Theory (Kamp, Van Genabith & Reyle 2011), where the information obtained at a certain stage of discourse processing is formally treated as a semantic representation (the ‘Discourse Context’, a Discourse Representation Structure or DRS), presuppositions are also handled as representational items, which need to be “resolved” in the context (of which the Discourse Context is part).3 However, I use the term presupposition as a more general concept, which roughly means “assumed to be taken for granted” or “is characteristic of the common ground as of this point”. With this general definition, we may analyze S-le sentences as having a presupposition that the speaker assumes the proposition not accepted by the hearer, which is formally represented as that the proposition is not in the common ground. It is also notable that the notion of hearer-new information or new information has been used in many prior works (e.g. Chafe 1976, Prince 1981, Gundel et. al 1993,

Delancey 1997). Those studies investigate a very different issue from the current study and thus what those authors mean by hearer-new information is quite different from what I mean by this notion. For example, there are a number of studies that investigate the relationship between linguistic forms of referring expressions and cognitive statuses of the referents in the addressee’s consciousness which is judged by the speaker on the basis of an assessment of the communication situation or discourse context (e.g. Chafe 1976, Prince 1981, Gundel et. al 1993). In Prince’s (1981) study, hearer-newness is considered against a continuum which consists of a variety of values that reflect the degrees to which a referent is assumed to be familiar to the hearer by the speaker. At one end of the continuum, the referent is a kind of discourse entity that is brand new to the hearer, and the corresponding referring form normally is an indefinite noun phrase such as the noun phrase a guy I work with in (9a). At the other end of the continuum, the referent is already in the context and thus is familiar

3 According to Kamp & Bende-Farkas (2019), the canonical view of presuppositions has it that presuppositions largely concern how the hearer or interpreter is to justify them (the interpreter-oriented view). The obligation of the speaker is to choose the expressions she uses in such a way that the hearer will be in a position to justify the presuppositions triggered by those expressions 82 to the hearer such as the pronoun he in (9a), which refers to an individual introduced earlier in the discourse. Between the two ends of the continuum is an intermediate case where the referent is a kind of discourse entity that the speaker assumes the hearer can infer such as the noun phrase the driver in (9a), which is inferrable from a bus with the common knowledge that buses have drivers.

(9) a. A guy I work with says he knows your sister. [Prince 1981: 233, (22d)]

b. I got on a bus yesterday and the driver was drunk. [Prince 1981: 233, (22c)]

In this study, hearer-newness is understood as the cognitive status of a new item that the speaker judges as being out of the addressee’s consciousness. However, this is not what I mean by hearer-new information in the current study, though the use of S-le sentences definitely requires an evaluation by the speaker of a cognitive status of new information for the hearer. What I mean by hearer-new information is not a cognitive notion that is used in those studies; instead, I focus on the contextual condition for S-le sentences, which requires that the proposition expressed by an S-le sentence is new to the hearer. The notion of new information is also used by Delancey (1997) as an analysis for a widely attested phenomenon called mirativity. Mirativity is identified in Delancey (1997) as a grammatical category whose fundamental function is to mark sentences which express propositions that are new or surprising to the speaker.

Though sometimes it does seem that the speaker is expressing his surprise at the proposition by using the S-le sentence, or the proposition is new to the speaker, I will claim that the fundamental function of S-le sentences is to mark new information for the hearer, and for those cases where the proposition is new to the speaker, as argued below, they should be analyzed as that the speaker is treating himself as the hearer, so they are still cases of providing new information for the hearer. So the kind of new information that I am concerned with here is different from the kind of new information concerned by Delancey for the category of mirativity. The former is about

83 new information to the hearer whereas the latter is about new information to the speaker. Note that our analysis of S-le sentences as being inherently associated with a hearer-new presupposition does not prevent us from explaining why other sentence structures (e.g. V-le sentences) can also be used to provide new information for the hearer. Presumably, all sentences of any type of structure can be used to provide certain new information for the hearer, depending on the knowledge state of the hearer in contexts. The S-le sentence differs from other sentence structures in that it marks for hearer-new information and generally requires an informative context (i.e. a context in which the proposition is new for the hearer), whereas other sentence structures are neutral in this respect and thus can be used in more diverse contexts. The structure of this section is as follows. In Section 3.3.1, I review Liu (2002) and Sybesma (1999), who also noted a condition of hearer-new information for S-le sentences. However, the two authors did not analyze all S-le sentences in terms of hearer-new information. Instead, for some S-le sentences where hearer-new might not be an obvious meaning, they proposed a different analysis. However, I show that hearer-new can account for those S-le sentences as well. In Section 3.3.2, I discuss two types of cases when the hearer-new presupposition fails, giving rise to certain pragmatic effects. In Section 3.3.3, I provide more evidence for the generalization that

S-le sentences express informativeness. In Section 3.3.4, I review a theoretical analysis proposed by Soh (2009) who also adopted the Stalnakerian framework (1998,

1999, 2002). However, Soh did not consider hearer-new information for S-le sentences; instead, she limited her analysis to account for interpretations of “change of state” and “contrary to one’s expectation”, which arise in certain contexts when S-le sentences are used. I thus propose to extend her analysis to account for a wider range of hearer-new utterances. In Section 3.3.5, I consider data of S-le sentences that are described in terms of attitude in earlier literature (e.g., Wang 1959) and show that those data can also be accounted for in terms of hearer-new information. So I suggest that we should take hearer-new as the basic meaning for S-le sentences. A summary of 84 this section is provided in Section 3.3.6.

3.3.1 Hearer-new information

Prior to Liu (2002), a number of linguists (e.g. Chao 1968: 798-800, Li & Thompson 1981: 238-300, Zhu 1982: 235-237) claimed that S-le sentences are used to express a new situation, however they have different definitions of “new situation”.

According to Chao (1968), the new situation is a situation that is either new or only new to the speaker. Zhu (1982) was not specific about what the definition of ‘new situation’ is, but seems to imply that it is a temporally new situation. Consider the following example.

(10) xià-yǔ le. [Chao 1968: 798, Zhu 1982: 235]

fall-rain LE

‘It’s raining.’

According to Chao, (10) can be used in a situation when it has been raining for some time but the speaker only just discovered that it is raining. So it is a situation new for the speaker. He also implied that it could also be a temporally new situation, that is, a change of state of the weather occurs recently. According to Zhu, (10) describes a new situation because it implies that it was not raining. So he seems to imply that it is a temporally new situation.

Contrary to those earlier studies, Liu (2002) pointed out that the ‘new situation’ in cases like (10) is not a temporally recent new situation, though the situation can be recent, but rather new information for the hearer. He drew this conclusion on the basis of two observations about the use of S-le sentences. First, the proposition does not need to be a temporally recent event. One of his examples is repeated in (11). The glosses and translation are mine.

(11) wǒ qùnián huí lǎojiā le. [Liu 2002: 72]

I last.year return hometown LE ‘I returned to my hometown last year.’

He pointed out that the reason why we get the impression that the S-le sentence is used to talk about a temporally new situation is because we normally talk about temporally new situations in our daily life.

Second, Liu (2002: 72) pointed out that S-le sentences are usually appropriate in two kinds of contexts: One is when the proposition is totally unknown to the hearer, and the other one is when the hearer’s knowledge is opposite to or different from the proposition. For either case, the proposition is new to the hearer. Liu argued for this by noticing that an appropriate context for (11), for example, is that the hearer does not know that the speaker returned to his hometown last year. As for the case that the expressed proposition is opposite to or different from the hearer’s knowledge, he illustrated it with negative S-le sentences. Consider one of his examples provided as follows.

(12) bù xià-yǔ le. [Liu 2002: 71]

NEG fall-rain LE ‘It’s not raining now.’

According to Liu, an appropriate context for this sentence is that the interlocutors knew it was raining, which is the background, and now they just found that it is not raining any more. The speaker may use this sentence and pretend that this is a new situation for the hearer and assume that the hearer is still holding the idea that it is raining, though he knows the hearer has also seen that it is not raining. So in this case, the hearer’s knowledge is assumed to be the opposite of the proposition (at least as far as the speaker is concerned). However, it is unacceptable to use this sentence if they have walked together for a while when there is no rain. In this case, the speaker cannot treat it as new information for the hearer any more. This means that the 86 sentence normally cannot be used if the proposition is clearly not new to the hearer. However, Liu did not propose that an S-le sentence expresses hearer-new information. Rather, he described some S-le sentences in terms of new state of affairs rather than new information for the hearer. For example, he considered S-le sentences with adjectival predicates to simply describe a new state of affairs. But what he meant by a new state of affairs for such sentences is neither new information for the hearer nor a temporally new state of affairs or new property of the subject; instead, what he meant is a difference from the standard with respect to a property that holds for the same sort of things.4 An example is presented in (13). The first sentence of (13a) describes a state, and the corresponding S-le sentence is in (13b).

(13) a. jiàqián tài guì, mǎi bù qǐ. [Liu 2002: 73]

price too expensive buy NEG afford

b. jiàqián tài guì le, mǎi bù qǐ.

price too expensive LE buy NEG afford ‘It is too expensive. (I) can’t afford it.’

According to Liu’s analysis of these two sentence structures, sentence (13a) expresses that the product has the property of being too expensive, whereas (13b) expresses that the price of the product exceeds the pricing standard for such products. However, to my understanding, the two interpretations are equivalent – to exceed the pricing standard for certain goods is what the adjective guì ‘expensive’ means; they do not show us how (13b) differs from (13a). In fact the generalization of hearer-new

4 Note that Liu’s discussion of the data that involve adjectives and le is based on a misclassification of the data. He did not distinguish the case when le is the inchoative le combining with an adjective, resulting in a change-of-state predicate (see Section 3.2.1) from the case like (13b) where le is a true sentence final le expressing informativeness. In both cases, an adjectival predicate is involved but they are actually two distinct cases. He simply assumed that when the sentence contains an adjective and a final le, the sentence expresses a change of state. He also noted that data like (13b) do not express change of state in a normal sense, and thus attempted to provide a more general characterization of the notion of change-of-state. In the normal sense, it refers to a change-of-state that occurs temporally, which is construed as a comparison between the state of an entity before the change and its state after the change. There is no such temporal change-of-state for (13b). So he extended the notion to a comparison between entities of the same sort which allowed him to account for (13b) in terms of a new state of affairs. 87 information. Rather, he described some S-le sentences in terms of new state of affairs information also applies to S-le sentences with adjectival predicates (see an example in (35) from Soh 2009). So the difference between the two sentences is that (13b) expresses that it is new information for the hearer that the product is too expensive, whereas (13a) is simply a predication of the subject, that is, it expresses that the product is too expensive.

Sybesma (1999: 59-61) made a two-fold generalization of S-le sentences and he also noted a use of hearer-new information for S-le sentences. For him, the final particle S-le marks change, which can be either objective or subjective. When it marks an objective change, the S-le sentence can be paraphrased like this: “‘the state of affairs [expressed by the part of the sentence preceding le] has begun’ or ‘it is now the case that [the state of affairs denoted by the sentence preceding le]’ with the implication that the state of affairs did not hold just a moment ago” (p. 60). One of his examples is repeated in (14). It can be paraphrased as follows: “the state of affairs that I understand it has begun”. When it expresses a subjective change, the sentence indicates that the state of affairs described by the sentence is new for the hearer. An example is provided in (15).5

5 The use of sentence (15) can be illustrated by two kinds of contexts. Consider the first context. The hearer gets the impression that the person referred to by the subject is young and may still go to school, but in fact this person graduated many years ago and already has two children. Consider the following conversation:

A: tā míngtīan bú shàngxué ma? he tomorrow NEG go.to.school Q ‘Doesn’t he go to school tomorrow?’ B: shàngxué? tā yǒu liǎng-ge háizi le. zǎo bìyè le. go.to.school he have two-CL child LE early graduate LE ‘Go to school? He has two children now! He graduated many years ago.’

So by using the S-le sentence Speaker B provides new information and the sentence here also implies that the hearer holds an incorrect belief that that person is young. Now consider another context. The interlocutors are talking about one of their old friends. The hearer does not know this friend has two children now.

A: tā xiànzài zěnmeyàng? he now how ‘How is he now?’ B: tǐng hǎo de. tā yǒu liǎng-ge háizi le. quite good DE he have two-CL child LE ‘Quite good. He has two children now.’

In this case, Speaker B is only being informative by using the S-le sentence. 88

(14) wǒ míngbái nèi-jiàn shì le. [Sybesma 1999: 60, (4a)]

I understand that-CL thing LE ‘Now I understand it.’ (15) tā yǒu liǎng-ge háizi le. [Sybesma 1999: 60, (4b)]

he have two-CL child LE ‘He has two children now.’

But note that there is no need to make a distinction between a subjective change and an objective change. The two examples above can be further generalized such that both sentences express that the proposition is new for the hearer. This is because (14), which is used to illustrate objective change, is actually also restricted to the kind of context in which the proposition is new to the hearer, that is, the hearer does not know that the speaker understands it. It is just that this is a special case where the hearer additionally holds certain background knowledge about the situation, which is in contrast to the new situation expressed by the S-le sentence, namely that the speaker did not previously understand it. Thus the hearer can infer that there was a moment at which a change of state occurred, that is, an objective change, and the resulting state holds up to the utterance time. So the kind of context for (14) is that the hearer believes that the speaker did not understand it but now is being informed that the speaker has understood it now, and the hearer can then infer that there was a moment when the speaker came to understand it. In the case where there is no hearer, the sentence can be analyzed as that the speaker is talking to himself, that is, he treats himself as a hearer who holds the belief that the speaker does not understand it. So the simpler generalization for the S-le sentence is that it expresses that the proposition is new for the hearer. Note that most of the data we have discussed so far are cases where something new has come about and thus it becomes new information. So those sentences always imply that there was a change of state that occurred at a particular time and the hearer can infer it. If the analysis is right, that is, if the S-le sentence is all about new

89 information for the hearer rather than change of state resulting in a new state or situation, then we should predict that it can also be used to talk about universal truths that involve no change in history, but the sentence forces an interpretation that there was a change of state, though no real change of state has ever occurred. This prediction is borne out. An example of this kind is mentioned in Soh (2009), which is repeated in (16). Sentence (16a) is a factual statement and (16b) is the corresponding

S-le sentence.

(16) a. dìqiú rào tàiyáng xuánzhuàn. [Soh 2009: 633, (19)] earth circle sun turn ‘The Earth circles around the Sun.’ b. dìqiú rào tàiyáng xuánzhuàn le.

earth circle sun turn LE

‘The Earth circles around the Sun (, which it did not before/contrary to what one may expect).’

According to Soh, the particle S-le in (16b) forces an inference that the earth did not rotate around the sun before, or that the proposition is contrary to what the hearer expects. Soh did not provide any context to illustrate the reading of “contrary to the hearer’s expectation” for (16b). It actually is quite hard to find such a context and native speakers’ judgment might differ. My intuition is that it is acceptable to use (16b) with the intended implication of contrary to the hearer’s expectation in the context where the speaker announces it as a new, remarkable, and unbelievable discovery to the public who have been holding the belief that the sun rotates around the earth. So this example supports the generalization that the S-le sentence expresses informativeness.

3.3.2 When the presupposition fails

S-le sentences presuppose that the proposition is new to the hearer. They

90 require a context in which the proposition is hearer-new, and are normally unacceptable if the proposition is already known to the hearer. An example is presented above in (1), which is repeated in (17).

(17) a. tā zǎoshàng -sǎo xuě le. zhīhòu qù-le xuéxiào.

he morning sweep-snow LE afterwards go-LE school

b. tā zǎoshàng sǎo-le xuě. zhīhòu qù-le xuéxiào.

he morning sweep-LE snow afterwards go-LE school ‘He swept the snow in the morning. Afterwards, he went to school.’

The discourse (17a) is acceptable if the hearer does not know the event of sweeping the snow but generally unacceptable if he already knows it, whereas (17b) is acceptable regardless of whether the hearer knows the event or not. In cases where a speaker uses an S-le sentence in a non-informative context one of two things happens. In one case, an accommodation failure effect arises, which supports the proposal that S-le sentences presuppose that the hearer has not already accepted the proposition. In the second case, the S-le sentence may be acceptable without an accommodation failure effect, but gives rise to an additional evaluative meaning. However, I suggest that this is still compatible with the idea that S-le sentences presuppose that the hearer does not accept the proposition. The two types of cases are discussed in Section

3.3.2.1 and Section 3.3.2.2 respectively.

3.3.2.1 Wait a minute…

Presuppositions generally need to be satisfied in the discourse context. When a presupposition cannot be accommodated by the hearer, the hearer may hold the conversation flow and push back on the presupposition. But it is infelicitous if the hearer pushes back against the asserted content in the same way. The “Hey, wait a minute” test is a common test used when the hearer fails to accommodate the presupposition. Consider an example from von Fintel (2004: 271, (3)), as in (18).

(18) A: The mathematician who proved Goldbach’s Conjecture is a woman. B: Hey, wait a minute. I had no idea that someone proved Goldbach’s Conjecture.

B’: #Hey, wait a minute. I had no idea that that was a woman.

Speaker A takes it for granted that the hearer knows someone has proved Goldbach’s

Conjecture. However, the hearer B does not know this. It is legitimate if B pushes back on the presupposition, as in (18B). However, it sounds odd if B pushes back against the asserted proposition, as in (18B’). It is hard to provide evidence for a full parallel of the two-folded pragmatic effects in the case of S-le sentences. On one hand, we can see a similar pragmatic effect arises in using S-le sentences when the hearer-new presupposition fails to be accommodated by the hearer. This is a case when the speaker misinforms the hearer, that is, the hearer actually already knows the proposition but the speaker does not know that the hearer already knows of it. For this kind of context, if the hearer takes informativeness of the S-le sentence serious, he may respond with an acknowledgement that he already knows the proposition. Consider an example in (19).

(19) Context: The hearer already knows that the speaker returned to his hometown last year but the speaker does not know that the hearer already knows of it.

A: wǒ qùnián huí-lǎojiā le.

I last.year return-hometown LE ‘I returned to my hometown last year.’ B: děngdeng, wǒ yǐjīng tīngshuō-le nǐ qùnián huí-lǎojiā le.

wait.a.minute I already hear.say-LE you last.year return-hometown LE

‘Wait a minute. I’ve already heard that you returned to your hometown last year.’

In this example, the speaker A presupposes that the hearer B does not know that he returned to his hometown last year by using an S-le sentence. However, this presupposition fails to be accommodated by the hearer because the hearer already knows the speaker returned to his hometown. The hearer’s reply as in (19B) is a response to this incorrect presupposition rather than a response to the event itself. It is similar to the reply “Hey, wait a minute, I had no idea that someone proved

Goldbach’s Conjecture.” as a response to the incorrect presupposition that someone proved Goldbach’s Conjecture as in (18). On the other hand, consider the case when the hearer pushes back on the asserted proposition expressed by an S-le sentence. In this case, the “wait a minute” test does not provide direct evidence that hearer-newness is a presupposition of S-le sentences. Consider the example in (20). Note that in this case, an informative/non-informative context is irrelevant.

(20) A: wǒ qùnián huí-lǎojiā le.

I last.year return-hometown LE ‘I returned to my hometown last year.’ B: děngdeng, wǒ bù zhīdào nǐ qùnián huí-lǎojiā le.

wait.a.minute I NEG know you last.year return-hometown LE

‘Wait a minute. I don’t know that you returned to your hometown last year.’

The intuition is that (20B) seems fine as a continuation of the conversation, though it is rare to have such kind of continuation for an S-le sentence. However, this does not provide counterevidence for our analysis, because in the (20B) the hearer is in fact essentially accepting the putative presupposition that the fact that the speaker returned home last hear is news to the hearer, which should in principle be acceptable. More generally, since what is presupposed for S-le sentences on the proposed analysis is that the hearer was unfamiliar with the entire proposition, for the hearer to push back on the asserted content of an S-le sentence is tantamount to accepting the 93 presupposition, and we would not expect this to be unacceptable the way pushing back on the asserted content in (20B) is. In sum, the data above provide some evidence for our analysis that S-le sentences presuppose hearer-newness, though this presupposition seems to be different from a typical presupposition. Like a typical case of presupposition, when the hearer-new presupposition fails to be accommodated by the hearer, the hearer may also hold the conversation flow and respond to the incorrect presupposition. However, unlike a typical case of a sentence involving a presupposition, which does not allow a push back on the asserted proposition, we cannot obtain evidence for an equivalent in S-le sentences, since the nature of the presupposition makes it hard to push back on the asserted content in a way distinct from accommodating the presupposition.

3.3.2.2 Gricean effects

Normally S-le sentences are unacceptable in non-informative contexts, or at least generate a kind of accommodation failure effect, which suggest that they presuppose that the hearer does not accept the expressed proposition already. However, there is one acceptable use of S-le sentences in a non-informative context, albeit it is one that has a very specific non-compositional, evaluative meaning. While this might seem to suggest that S-le sentences are acceptable in non-informative contexts as well as informative contexts, I will ultimately suggest that this particular use is compatible with the analysis of S-le sentences as carrying a presupposition that the hearer does not accept the proposition, and in fact may even rely on this analysis. Three examples are presented below. I use the term implicature for the additional evaluative meaning since it gives us a feeling of Gricean implicature (Grice 1975).6 Note that the examples provided below show that the additional evaluative meaning involves a reference to the cultural values, or moral, ethical principles

6 It might not be accurate to call the evaluative meaning an implicature because an implicature can be cancelled but in this case it actually is what the speaker intends to convey and thus it would be odd to cancel it. I adopt the term "implicature" here because it is not a compositional meaning inherently associated with the sentences but a kind of meaning which seems to come out in order to compensate that the contextual condition for using the sentences is not met, which is similar to Gricean implicature, though they do not seem to be the exact same phenomenon. So "implicature" is just the best term I can think of for it. 94 associated with that specific S-le sentence, which are shared between the interlocutors. This reference to the shared values and principles creates certain communicative effects in the discourse, which we do not find if the corresponding non-S-le sentence is used instead. Consider the S-le sentence in (21b) in a non-informative context where the hearer knows that the person denoted by the subject was reading a book yesterday.

(21) Context: The hearer knows that the person was reading a book. a. tā zuótiān kàn-le shū, nǐ ne?

he yesterday look-LE book you Q b. tā zuótiān kàn-shū le, nǐ ne?

he yesterday look-book LE you Q ‘He read a book yesterday. How about you?’

The specific implicature in this case depends on the common value on the activity of reading. If reading is considered a glorious activity and the hearer knows that the speaker has a negative opinion of him, then the implicature is that the speaker is approving of the person who read yesterday while also conveying doubt that the hearer did anything as good yesterday. So the hearer can perceive that the speaker shows dissatisfaction with him. On the contrary, if the activity of reading is considered contemptible, then the implicature is that the speaker shows his contempt for the person who read yesterday, meanwhile indicating to the hearer that he hopes the hearer did not do anything as contemptible. By contrast, the V-le sentence in (21a) only makes a factual statement, without any such implicature unless the sentence is uttered with an unusual intonation. Consider another example. The event described in (22) is happening at utterance time, so the non-S-le sentence used here is marked with the progressive aspect. This discourse might be used in a situation where the parent is asking one of his children to help his sister clean the kitchen, and that child has also seen that his

95 sister is cleaning the kitchen. So in this case the hearer knows of the event of his sister cleaning the kitchen.

(22) Context: The hearer knows that his sister is cleaning the kitchen. a. nǐ mèimei zài dǎsǎo chúfáng. nǐ yàobúyào bāng tā?

you sister PROG clean kitchen you would.like help her b. nǐ mèimei dǎsǎo-chúfáng le. nǐ yàobúyào bāng tā?

you sister clean-kitchen LE you would.like help her ‘Your sister is cleaning the kitchen. Would you like to help her?’

The difference between the two sentences in this context is that by using the S-le sentence as in (22b), the parent seems to blame the child for not helping his sister, an inference rooted in a sense of moral obligation to help around the house. However, the progressive form in (22a) conveys no such judgment and is instead interpreted as a factual description. This contrast derives different communicative effects. It is harder for the child to refuse to help if the parent uses the S-le sentence because the refusal would violate the obligation more blatantly than refusal as a response to the progressive form. A third example is provided in (23). The context is that the wife is blaming her husband for their child’s failure to get admitted to the university. The event the hearer already knows of is that their neighbor’s child got admitted to the university.

(23) Context: The hearer knows that the child next door has got admitted to the university. a. gébì-jiā-de háizi kǎoshàng-le dàxué, zán-jiā ne?

next.door-family-DE child get.admitted.to-LE university our-family SFP b. gébì-jiā-de háizi kǎoshàng-dàxué le, zán-jiā ne?

next.door-family-DE child get.admitted.to-university LE our-family SFP nǐ měI tiān zhǐ gù dǎ májiàng, cóngbù guānxīn háizi. you every day only care.about play mahjong never concern child 96

‘The child next door was admitted to the university. How about our child? Every day, you only care about playing mahjong, and are never concerned with the child!’

The use of the S-le sentence (23b) in this case implies that the speaker is very disappointed that her child did not get admitted to the university, rooted in the general importance of going to college. In contrast, the use of the V-le sentence in (23a) does not express this disappointment with the same rhetorical force as the S-le sentence, though it is apparent from the discourse that the speaker is disappointed. The V-le sentence here is more a simple description of a fact. However, the use of the S-le sentence makes the speaker’s disappointment about their child’s failure more prominent. The data presented above illustrate a seemingly contradictory use of S-le sentences from their canonical use since in the former case they are used in non-informative contexts while in the latter case they are used in informative contexts. However, we may reconcile the two uses by considering the informative use the default use of S-le sentences and the non-informative use a derivative of it. Suppose S-le sentences presuppose hearer-new information, but when a speaker intentionally uses it in a context where the truth of the proposition is already part of the common ground, it creates an additional meaning to account for the fact that this condition failed – since it does not mean what it normally means it takes on a distinct meaning.

This phenomenon of using S-le sentences for a non-canonical meaning in the non-default contexts reminds us of the phenomenon of Gricean implicature noted above. According to Grice’s (1975) theory of conversational implicatures, participants in a communicative exchange observe the Cooperative Principle and follow several maxims of conversation. A speaker obeying the maxims permits the hearer to draw certain inferences not directly expressed. However, when a maxim is blatantly flouted, implicatures also arise. For example, if A says Tehran’s in Turkey, isn’t it?, a response by B that says And London’s in Armenia, I suppose would flout the maxim of quality

97 since the speaker knows that the hearer knows that London is not in Armenia and is being clear about flouting. This gives rise to two kinds of implicatures, one is that Tehran is not in Turkey, and the other one is that B thinks that A is an idiot for suggesting it. It seems to me that this might be a way for us to analyze the non- compositional, evaluative meanings of S-le sentences when they are used in non-informative contexts with an assumption that informative contexts are but non- informative contexts are not the default contexts for S-le sentences. The speaker uses an S-le sentence which literally expresses the belief that the proposition is new to the hearer; however, it is mutually known that it is not new for the hearer, so he blatantly flouts a maxim such as quality, explaining the implicature. In the current study, I do not have a detailed analysis for this issue, and I leave it to future studies.

3.3.3 An effect of using S-le sentences in the informative context

I have shown above that the generalization that the S-le sentence expresses that the proposition is new for the hearer can account for all the data discussed in the literature. In this section I provide more evidence to support this generalization. Recall what I have shown in Chapter 2 about judgments of S-le sentences with and without NP-internal modifiers. S-le sentences with bare nouns are easily accepted, those with NP-internal modifiers overall are judged less natural, and those with DPs are unacceptable. Those judgments suggest a syntactic size effect going on as acceptability of heavier NP-internal modifiers decreases. In Section 3.3.2.1 below, I show that context also plays a role in judgments of S-le sentences. The data show that acceptability of S-le sentences with NP-internal modifiers is significantly improved if they are judged in an informative context, a kind of context where the speaker informs the hearer of new information. Nonetheless, S-le sentences with DPs are still unacceptable in informative contexts, suggesting a syntactic constraint. The data are presented in Section 3.3.2.2. These facts can be explained by the analysis that the S-le sentence expresses

98 informativeness and the VP of this sentence structure involves noun incorporation. Noun incorporation requires a bare noun or at least a fairly small NP, whereas the V-le sentence allows noun phrases of any size. So S-le sentences with bare nouns or small NPs are acceptable but those with NP-internal modifiers are judged less natural or sometimes harder to be accepted while in this case the corresponding V-le sentences are strongly preferred. But since the S-le sentence has a “hearer new” requirement and is the dedicated way of expressing that, then provided that the context is more explicit, the speaker can tolerate larger NPs when necessary to convey that meaning.

3.3.3.1 The S-le sentences with NP-internal modifiers The S-le sentences presented below are those S-le sentences with NP-internal modifiers that receive reduced acceptability when judged out of context (see Section 2.4.2). They are presented here with an explicit informative context. All consultants accepted them when judged in such a context. The significant increase in acceptability in the informative contexts suggests that the informative context is the right context for the S-le sentence. The S-le sentence in (24A) has an NP-internal adjective modifier. Speaker A tells the hearer B that Mary wore black clothes yesterday, which he assumes that the hearer does not know. All 21 speakers I consulted accepted the S-le sentence in this context.

(24) Context: Speaker B does not know that Mary wore black clothes yesterday. A: Mǎlì zuótiān chuān-hēisè-de-yīfù le.

Mary yesterday wear-black-DE-clothes LE ‘Mary wore black clothes yesterday.’ wǒ zhè shì dìyīcì kànjiàn tā chuān hēisè.

I this COP first.time see she wear black

‘This was the first time I saw her wear black clothes.’

B: tā hǎoxiàng quèshí hěnshǎo chuān hēisè. she seem indeed rarely wear black 99

‘It does seem that she rarely wears in black.’

Similarly, the S-le sentence in (25B) with a PP modifier corrects an incorrect belief or misconception the hearer holds. Speaker A’s utterance suggests that he holds the belief that Mary did not do anything today, which is not true in this context. The speaker, therefore, informs the hearer A of the event of Mary sweeping the snow on the roof. All 21 consultants accepted the S-le sentence in this context.

(25) Context: Speaker A does not know that Mary swept the snow on the roof.

A: Mǎlì jīntiān shénmeshì dōu méi gàn.

Mary today anything at.all NEG do ‘Mary didn’t do anything at all today.’ B: bú shì a. tā sǎo-wūdǐng-shàng-de-xuě le.

NEG COP SFP she sweep-house.roof-on-DE-snow LE

‘It’s not like that. She swept the snow on the roof.’

The S-le sentence in (26B), with a possessor modifier, provides information that the hearer requires. Speaker A’s question suggests that he did not know what John did yesterday and wants the information. B accordingly tells A the information he wants, that is, John cleaned Mary’s kitchen. All 21 consultants accepted the S-le sentence in this context.

(26) Context: Speaker A does not know that John cleaned Mary’s kitchen. A: Yuēhàn zuótiān gàn-shénme le?

John yesterday do-what LE ‘What did John do yesterday?’ B: tā dǎsǎo-Mǎlì-de-chúfáng le.

he clean-Mary-DE-kitchen LE ‘He cleaned Mary’s kitchen.’

100

The S-le sentence in (27B) with a demonstrative provides information that is helpful to the hearer. The conversation suggests that A wants someone to tell him about the book but does not know who has read the book. B informs A of the event that John read the book and thus the hearer may ask him for help. All 11 speakers I consulted accepted the S-le sentence in this context.

(27) Context: Speaker A does not know that John read that book.

A: nǐ gěi wǒ jiǎngjiang nèi-běn shū ba!

you to me talk.about that-CL book SFP ‘Please tell me about that book!’

B: nǐ wèn Yuēhàn ba, tā kàn-nèi-běn-shū le. wǒ méi kàn.

you ask John SFP he read-that-CL-book LE I NEG read ‘You may ask John. He read that book. I haven’t.’

Finally, the S-le sentence in (28A2) has a relative clause modifier and provides information which is used as an explanation for an observed situation. The conversation suggests that the hearer B may not know the possible situation that John and Mary have reconciled, and he does not know that John read the book bought by Mary. A informs B of the event that John read the book bought by Mary, which is used as an explanation for his conclusion that they probably have reconciled. All 21 consultants accepted the S-le sentence in this context.

(28) Context: Speaker B does not know that John read the book Mary bought.

A1: Yuēhàn dàgài hé Mǎlì héhǎo le.

John probably with Mary reconcile LE ‘John probably has reconciled with Mary.’

B: nǐ zěnme kàn-chūlái ?de

you how look-out DE ‘How can you tell?’

101

A2: tā kàn-Mǎlì-mǎi-de-shū le. zhè shuōmíng tāmen xiànzaì guānxì búcuò.

he look-Mary-buy-DE-book LE this suggest they now relation not.bad ‘He read the book Mary bought. This suggests their relation is not bad now.’

The data provided above show that the lowered acceptability of those S-le sentences with NP-internal modifiers when judged out of context is significantly increased when they are judged in an explicitly informative context. This suggests that the informative context is the right context for the S-le sentence, and thus it supports the claim that the S-le sentence expresses a proposition new for the hearer.

3.3.3.2 The S-le sentences with NP-internal numerals and the quantifier every The informative context, however, does not significantly increase the acceptability of those S-le sentences with NP-internal numerals and the quantifier every. As shown below, S-le sentences involving numerals are increased to some extent but still not accepted by all consultants and the V-le sentences are strongly preferred for the intended meaning. S-le sentences involving the quantifier every in an informative context are still judged as unacceptable by all consultants. This further confirms the analysis that the S-le sentence does not syntactically project a full DP in the object position; instead, it is an NP. In this case, the syntactic constraint overrides the intended semantics of informativeness. The DP cannot be tolerated even though it is understood that the sentence is intended to be informative. The relevant data are presented below. The dialogue below contains an S-le sentence with an NP-internal numeral. In this example, B informs A, with sarcasm, that Mary ate three bowls of rice, correcting the hearer’s incorrect belief that she did not do anything. Two of 21 consultants judged the S-le sentence in this context unacceptable and another three did not reject the sentence but pointed out that they still prefer the V-le sentence. The other 16 accepted the sentence.

102

(29) Context: Speaker A does not know that Mary ate three bowls of rice.

A: Mǎlì zuótiān shénmeshì dōu méi gàn! tài lǎn le!

Mary yesterday anything at.all NEG do too lazy LE ‘Mary didn’t do anything yesterday! She’s too lazy!’ B: bù shì a. ?tā chī-sān-wǎn-mǐfàn le. wǒ kànjiàn le.

NEG COP SFP she eat-three-bowl-rice LE I see LE

‘It’s not like that. She ate three bowls of rice. I saw it.’

The corresponding V-le sentence is provided below.

(30) tā chī-le sān-wǎn mǐfàn.

she eat-LE three-bowl rice ‘She ate three bowls of rice.’

When the NP-internal modifier is the universal quantifier every, the S-le sentence is still judged as unacceptable in an explicit informative context. Consider (31). The context here is an informative context in which the speaker corrects the hearer by providing information contrary to the hearer’s knowledge. In this example, A thought that Mary only read one book, which is not true in this context. B informs him that Mary read every book. All 11 consultants judged the S-le sentence unacceptable.

(31) Context: Speaker A does not know that Mary read every book.

A: Mǎlì zài shūfáng zhǐ kan-le yī-běn shū.

Mary LOC study only look-LE one-CL book ‘Mary only read one book in the study.’ B: bù shì a. *tā kàn-měi-běn-shū le.

NEG COP SFP she look-every-CL-book LE

‘It’s not like that. She read every book.’

Consider the same S-le sentence in another explicit informative context below. B 103 provides A new information which is used to answer A’s question, yet the S-le sentence is also judged as unacceptable.

(32) Context: Speaker A does not know that the person read every book.

A: tā zài shūfáng kàn-le jǐ-běn shū?

he LOC study look-LE how.many-CL book ‘How many books did he read in the study?’

B: jùwǒsuǒzhī, *tā kàn-měi-běn-shū le.

as.far.as.I.know he look-every-CL-book LE ‘As far as I know, he read every book.’

The corresponding V-le sentence, which is provided below, is acceptable in both informative contexts.

(33) tā kàn-le měi-běn shū.

he look-LE every-CL book ‘He read every book.’

That said, the most standard way to express the quantifier every at the object position uses a different sentence structure, which is shown below.

(34) tā měi-běn shū dōu kàn-le.

he every-CL book all look-LE

‘He read every book.’

The data above show that an explicit informative context increases acceptability of the S-le sentences with NP-internal numerals to some extent but such sentences still sound less natural and might be rejected by some native speakers; meanwhile it does not increase acceptability of S-le sentences involving the quantifier every at all. These data provide evidence for that the object position of S-le sentences is a NP not a DP. In the case of bare nouns or small NPs, there is no syntactic problem 104 regarding incorporation, so I assume that acceptability of those sentences without making the context explicit arises from hearers just accommodating that it is meant to be new information. This contrasts with the lower grammaticality of the bigger NPs where the violation of the syntactic constraint is jarring unless the context is clearer. Meanwhile, the corresponding V-le sentences with any type of noun phrase object are always acceptable even without an explicit context, suggesting that the S-le sentence and the V-le sentence are syntactically distinct.

3.3.4 Two common implications: Change of state and contrary to expectation

Some S-le sentences are often found associated with implications that are normally labeled as “change of state”7 and “contrary to one’s expectation” (or “correcting one’s wrong assumption”) in the literature (e.g. Li et al. 1982, Van den Berg 1989, Liu 2002, Soh 2009). Consider an example from Soh (2009).

(35) a. zhè-piàn xīguā hěn tián. bú bì jiā táng. [Soh 2009: 632, (17)] this-slice watermelon very sweet no need add sugar ‘This watermelon is sweet. It is not necessary to add sugar.’ b. zhè-piàn xīguā hěn tián le. bú bì jiā táng le.

this-slice watermelon very sweet LE no need add sugar LE ‘This watermelon is sweet (contrary to what one may assume/expect). It is not

necessary to add sugar (contrary to what one may assume/expect).’

The two sentences in (35a) assert that the watermelon is sweet and there is no need to add sugar. However, the two S-le sentences in (35b) imply that there is someone in the context who holds a different assumption, that is, it seems to the speaker that the

7 Note that what is meant by “change of state” here is different from the notion of change of state discussed in the literature of lexical semantics, which says that change of state means that something undergoes a change of property. The kind of “change of state” associated with the S-le sentences is more about a change of situation, or a change of state of affairs. The change is not an assertion but an implication drawn by inference. In such a case, the hearer assumes one state of affairs, and then later learns that the state of affairs is different, and thus derives from that that there is a change of state. But the S-le sentence itself does not entail that per se. So the reading of “change of state” only arises when the hearer knows the old situation or state of affairs before the change. 105 hearer assumes that the watermelon is not sweet and needs sugar. Soh (2009) proposed an analysis of S-le sentences in a Stalnakerian framework to account for the two interpretations. She also suggested that S-le sentences are associated with presuppositions. But the specific presupposition she proposed for S-le sentences is different from the one I propose here. The discrepancy between the two analyses is largely due to our different views of the fundamental meaning of S-le sentences. While the current view of S-le sentences is that they indicate hearer-new information, Soh’s view is that they indicate a certain kind of change. But note that even in those cases where we obtain meanings of change of state or being contrary to expectation, which Soh analyzed in terms of two types of change, the hearer-new condition still applies. It is just that in those two cases, there are other propositions in the common ground that interact with the asserted proposition, which give rise to these interpretations by inference. Below I first outline

Soh’s analysis and then, instead of rejecting her analysis, I show a way of modifying her analysis so that we can extend it to account for hearer-new information as well as the two interpretations of change of state and being contrary to expectation concerned in her study. Soh showed that the two interpretations of “change of state” and “contrary to expectation” are always available for the S-le sentence regardless of whether the verb predicate is stative, activity, achievement or accomplishment. Data of the four types of verbs are presented below.8 First, consider one of her examples of S-le with a stative verb.

8 Note that Soh also presented her data in pairs. She compares S-le sentences with corresponding sentences that have no le in order to show the semantic effect of adding the final particle le. Her assumption is that the base sentences (i.e. sentences without le) do not have a meaning of change and thus the change meaning associated with S-le sentences is contributed by the final le. However, it is notable that those base sentences with activities, accomplishments, and achievements, which describe events, are arguable nonfinite. Though they can be used as independent sentences with certain interpretations, the interpretations actually are partially determined by contexts which provide the tense and aspectual meanings. So the translations Soh provided for those base sentences are actually inaccurate. The canonical corresponding finite sentences of those base sentences are sentences marked with an aspectual morpheme such as V-le sentences. All of the three eventive predicates, activities, accomplishments, and achievements, can be expressed by V-le sentences, which have a simple past interpretation by default. 106

(36) a. tā xiàng bàba. [Soh 2009: 631, (16)] he resemble father ‘He resembles his father.’ b. tā xiàng bàba le.

he resemble father LE ‘He resembles his father (, which he did not before/contrary to what one may

expect.)

The sentence (36b) may imply that the subject did not resemble his father before but now he resembles him, and it may also imply that his resembling his father now is contrary to one’s expectation. However, sentence (36a) does not have such an implication. An example with an activity verb is given in (37).

(37) a. tā mà tā de háizi. [Soh 2009: 632, (18)]

he scold he POSS child ‘He is scolding his child.’ b. tā mà tā de háizi le.

he scold he POSS child LE ‘He is scolding his child (, which he was not doing before/contrary to what

one may expect).’

The sentence (37b) may imply that the subject was not scolding his child before, or it is contrary to the hearer’s expectation that the person is scolding his child, depending on the context. However, (37a) only asserts that he is scolding his child and has no such implication, though by common knowledge the hearer can infer from (37a) that he was not scolding his child before.

An example with an achievement verb is provided in (38).

107

(38) a. tāmen gānggāng dàodá shān-dǐng. [Soh 2009: 633, (21)] they just reach mountain-top ‘They just reached the top of the mountain.’ b. tāmen gānggāng dàodá shān-dǐng le.

they just reach mountain-top LE ‘They just reached the top of the mountain (, which they hadn’t done before/

contrary to what one may expect).’

Soh admitted that it is difficult to see how S-le in (38b) contributes a meaning of “change of state” since the achievement verb is associated with such meaning. Both sentences seem to describe the same kind of change, that is, the people referred to by the subject were not at the top of the mountain but now they are. However, she assumed that there is a difference between the two sentences. One type of change of state is a property of the predicate, with respect to which the achievement is the same as an accomplishment. The other type of change of state is associated with the S-le sentence and she has a specific analysis for it, which is presented below. Regarding the accomplishment predicate, Soh discussed what she called atelic accomplishments and telic accomplishments in Chinese. As discussed in Chapter 4, accomplishments that are telic in English may not be telic in Chinese. Usually, Chinese requires resultative morphemes to telicize accomplishment predicates.

Despite the complexity with accomplishments, she argues that the two interpretations of “change of state” and “contrary to expectation” are also available in this case. Consider (39).

(39) a. tā xiě nà-fēng xìn. [Soh 2009: 635, (25)]

he write that-CL letter ‘He is writing that letter.’

108

b. tā xiě nà-fēng xìn le.

he write that-CL letter LE ‘He is writing that letter (, which he was not doing before/contrary to what one may expect).’

The sentence (39b) may imply that he was not writing that letter before, or it may imply that it is contrary to one’s expectation that he is writing that letter. Such implications are not available for (39a). Soh also discussed a restriction on the distribution of S-le sentences. She argued that a telic predicate with a downward-entailing quantifier, which specifies information relating to the endpoint, cannot be used in the S-le sentence, whereas telic predicates with non-downward-entailing quantifiers may occur in the S-le sentence. The data is presented in Section 2.4.2.4, and I have argued that the restriction here is syntactic not semantic, as I provided more evidence that generalized quantifiers, either downward-entailing or non-downward-entailing, are largely prohibited at the object position of the S-le sentence if the verbal predicate is dynamic. Soh proposed a unified analysis for the two interpretations of “change of state” and “contrary to expectation” in the framework of pragmatic presupposition developed by Stalnaker (1998, 1999, 2002). She assumed that S-le essentially marks change and the two interpretations are construed as two different kinds of change which are analyzed as an effect of the result of an inference of two contrasting propositions which are obtained in different ways for the two cases. Her analysis of S-le sentences is presented in (40) (Soh 2009: 642).

(40) The speaker using a sentence with S-le

(i) asserts a proposition p at speech time (ts),

(ii) presupposes [¬p before speech time (ts)], and (iii) either accepts or rejects the inclusion of the presupposition in the subsequent common ground.

109

Note that she assumed two temporally adjacent common grounds. The first one temporally occupies an interval ti and it contains the speech time ts, and a subsequent one occupies an interval tj, and at this common ground the participants accept or reject the asserted proposition. The temporal relation is diagrammed in the time line below (Soh 2009: 642).

(41) Time Line

ts ------[------[-----]-]-----[------]------

ti tj

The analysis in (40) says that when the speaker utters an S-le sentence, he does not just assert a proposition p which is expressed by the sentence, but also presupposes that [¬p before ts], that is, the speaker believes that ¬p is part of the common ground at ti. Whether the S-le sentence gives rise to the interpretation of “change of state” or “contrary to expectation” depends on whether the speaker accepts or denies the presupposition [¬p before ts] for the subsequent common ground at tj. The reading of “change of state” arises if the speaker accepts the presupposition [¬p before ts] for the subsequent common ground at tj. A schematic representation of this reading is provided below (Soh 2009: 643). P1, P2, and P3 refer to the propositions involved. P3 is the asserted proposition as in (40i). P1 is the presupposition [¬p before ts] as in (40ii). P2 represents that the speaker accepts the presupposition at tj as in (40iii).

(42) Change of State

Common groundi at ti Common groundj at tj

P1: ¬p before ts P2: ¬p before ts [presupposition] [acceptance of presupposition]

P3: p at ts [assertion]

Soh proposed that by accepting the previous presupposition [¬p before ts] in the subsequent common ground (P2) and proposing to add the asserted proposition to the

110 subsequent common ground (P3), the speaker expresses a belief about the existence of a change in the truth of p (from ¬p to p) across the temporal domain from before ts to ts. This change becomes part of the subsequent common ground when the interlocutors accept P2 and P3. So the change in this case is derived from P2 and P3 within the subsequent common ground. Now consider how this analysis can explain the change of state reading for (36b), which is repeated in (43).

(43) tā xiàng bàba le.

he resemble father LE ‘He resembles his father (, which he did not before.)

According to this analysis, (43) asserts the proposition that the person the subject refers to resembles his father at speech time (ts), which is represented as P3 in (44) (Soh 2009: 645), and meanwhile the speaker presupposes that the person did not resemble his father before ts, represented as P1 in (44). When the speaker accepts the presupposed proposition after assertion, represented as P2, the contrast between the two accepted propositions, P2 and P3, gives rise to the reading of “change of state” regarding this person’s resemblance to his father.

(44) Change of State

Common groundi at ti Common groundj at tj P1: He did not resemble his P2: He did not resemble his father

father before ts. before ts. [presupposition] [acceptance of presupposition]

P3: He resembles his father at ts. [assertion]

The reading of “contrary to expectation” is obtained if the speaker rejects the presupposition [¬p before ts] for the subsequent common ground. The schematic representation of this reading is provided below (Soh 2009: 643).

111

(45) Contrary to expectation

Common groundi at ti Common groundj at tj

P1: ¬p before ts P2: p before ts [presupposition] [rejection of presupposition]

P3: p at ts [assertion]

The rejection of the presupposition [¬p before ts] for the subsequent common ground results in [p before ts] for P2. Soh proposed that by rejecting the presupposition, the speaker also expresses a belief about the existence of a change, that is, across the temporal domain from ti to tj, there is a change in the hearer’s belief about the truth of

[p before ts] (from [¬p before ts] to [p before ts]) if the asserted proposition is accepted by the hearer. In this case, the change is derived from the propositions P1 and P2 across the two common grounds. Now consider how this analysis explains the reading of contrary to expectation of (35b), which is repeated in (46).

(46) zhè-piàn xīguā hěn tián le. bú bì jiā táng le.

this-slice watermelon very sweet LE no need add sugar LE ‘This watermelon is sweet (contrary to what one may assume). It is not necessary to add sugar (contrary to what one may assume).’

According to the analysis, the first sentence of (46) asserts that the watermelon is sweet at ts, represented as P3 in (47), and the speaker presupposes that the watermelon was not sweet before ts, represented as P1 in (47). The “contrary to expectation” interpretation arises because when the speaker rejects the presupposed proposition after assertion, the resulting proposition that is accepted, represented as P2, contrasts with the presupposed proposition, giving rise to an inference that the speaker is correcting the hearer’s false assumption.

112

(47) Contrary to expectation

Common groundi at ti Common groundj at tj P1: The watermelon was not P2: The watermelon was sweet

sweet before ts. before ts. [presupposition] [rejection of presupposition]

P3: The watermelon is sweet at ts. [assertion]

Soh’s study was aimed to account for the two interpretations of “change of state” and “contrary to one’s expectation” that are often implied from the S-le sentence given a certain context. Her analysis provides an adequate account for these two interpretations. However, her analysis cannot account for other uses of S-le sentences. Consider one of Liu’s (2002) examples as in (48).

(48) wǒ qùnián huí lǎojiā le. [Liu 2002: 72]

I last.year return hometown LE ‘I returned to my hometown last year.’

Soh’s analysis says that the speaker presupposes [¬p before ts] when asserting p using an S-le sentence. However, for (48), which has no implication of “change of state” or “contrary to one’s expectation”, the speaker does not seem to presuppose that he did not return to his hometown last year. Instead, it seems to be that the speaker expresses a belief that the hearer does not know that he returned to his hometown last year, that is, hearer-new information, which I suggest can be characterized in terms of a presupposition that the proposition that the speaker returned to his hometown last year is new to the hearer, or there is no presupposition about this proposition. Note that it is not enough as an analysis for S-le sentences if we simply say that an S-le sentence assert a proposition, which is part of the analysis proposed by Soh for S-le sentences as in (40i), because, for example, the corresponding V-le sentence of (48) can perform the same function of asserting the proposition that the speaker returned to his hometown last year. So the proposed analysis that S-le sentences are associated with a

113 presupposition can distinguish them from other sentence structures such as V-le sentences. In order to account for both the basic meaning of hearer-new as well as the two interpretations of “change of state” and “contrary to one’s expectation”, I follow Soh (2009) but suggest a more general analysis for S-le sentences as in (49), which also assumes the Stalnakerian framework adopted by Soh.

(49) The speaker using a sentence with S-le (i) asserts a proposition p, (ii) presupposes p is not accepted in the current common ground.

The condition (49ii) subsumes the two broad categories we are concerning here. On one hand, to presuppose that p is not accepted on the current common ground can include a case where there is no presupposition about p, which requires that the hearer has no knowledge of p at all, so the use of S-le sentences is simply to provide hearer-new information. On the other hand, it can be a case where ¬p is accepted as a presupposition, which is concerned in Soh’s study. In this case, according to Soh’s analysis, interpretations of “change of state” or “contrary to one’s expectation” arise when the presupposition is accepted or rejected after assertion.

To sum up, I reviewed Soh’s (2009) analysis for S-le sentences, which was aimed to account for interpretations of “change of state” and “contrary to one’s expectation” that arise in certain contexts. However, Soh did not consider the case when S-le sentences are simply used to provide new information for the hearer and neither of the two interpretations is attested. I suggested an analysis, which still assumes the Stalnakerian framework Soh adopted, to extend her analysis to cover all cases of using S-le sentences. However, I leave a full-fledged explication of the proposed analysis for future work.

114

3.3.5 Attitudinal uses of S-le sentences

Before concluding this section, it is also worth mentioning some uses of S-le sentences that are described as expressing an attitude in earlier literature (Wang 1959: 160-163, Zhu 1982: 235-237, Lü 1982: 262-263). However, as I show below, those attitudinal uses can also be accounted for in terms of hearer-new. So I suggest that we should take hearer-new as the essentail meaning for S-le sentences.

Those earlier descriptive studies analyzed the final particle le as a functional word that expresses yǔqì, i.e. mood or speaker attitude. They group S-le with a number of sentence final particles that express different kinds of yǔqì, including ne, a, me, la, bale, etc.. These particles are called yǔqìcí ‘attitudinal particles’ and most of them are toneless. According to Wang (1959) and Lü (1982), S-le expresses a definitive attitude (juédìng yǔqì) of the speaker towards the proposition, that is, the speaker shows his certainty about the proposition. According to Wang, the definitive attitude is used when the speaker claims his perception, determination or judgment. He illustrated each use with several sentences, albeit without giving further elaboration for each sentence. Three of those sentences are repeated below, and they illustrate a different use. The glosses and translations are provided by me.

(50) wǒ bùnéng sòng-nǐ le. [Wang 1959: 161, (1E)]

I can’t see.off-you LE ‘I can’t see you off (now, but I planned to see you off).’

For (50), an appropriate situation is that the hearer knows that the speaker planned to see him off, but now the speaker cannot. According to Wang, S-le is used here because the speaker perceives a certain situation. But Wang was not specific about what is special about the situation so that the perception of it would require an S-le sentence. It seems to me that what he meant is that for this example the speaker perceives a new situation that he cannot see off the hearer, which is opposite to the

115 original plan that he would see off the hearer. So in this example, the proposition that the speaker cannot see off the hearer is new information for the hearer. Now consider one of the examples that Wang used an illustration of the speaker expressing his determination with the S-le sentence.

(51) cóngcǐ zhǐ jiào nǐ shīfu, [Wang 1959: 161, (2B)]

from.now.on only call you master zàibú jiào-jiějie le.

no.longer call-elder.sister LE ‘From now on, I will only call you master, and no longer call you sister.’

For (51), Wang is not specific about this example either. It seems to me that what he meant is that the speaker does not just convey the literal meaning that he will call the hearer master rather than sister from now on but also expresses that s/he has made some decision about it. The hearer-new analysis can also account for this example. The hearer knows that the speaker has been calling her ‘sister’ but does not know the speaker has changed her mind and will call her ‘master’, which is new information for the hearer. Now consider one of the examples that Wang used as an illustration of the speaker expressing a judgment with the S-le sentence.

(52) zài húshuō, wǒ jiù dǎ le. [Wang 1959: 161, (3A)]

again bullshit I then hit LE

‘If you bullshit again, I’ll hit you.’

Example (52) is a conditional sentence and the consequent clause is an S-le sentence.

Again, Wang is not specific about this example. But what he seemed to mean is that the speaker expresses that he has a judgment regarding the consequent of the hypothetical situation which is described by the antecedent – that is, if the hearer talks

116 nonsense again – and the consequent is that he will hit the hearer. (Note that both the subject of húshuō ‘bullshit, talk nonsense’ and the object of dǎ ‘hit’ are null anaphora, which refer to the hearer.) For this example, the speaker assumes that the hearer holds the assumption that he will not hit him if he keeps talking bullshit, so the speaker now tells him that he will hit the hearer if he continues to talk bullshit, which is new information for the hearer.

I have shown above that a single generalization about hearer-newness can account for those data that are described in terms of their uses in the earlier literature. So I suggest that we should take hearer-new as the basic meaning for S-le sentences.

3.3.6 Summary

In this section, I showed that S-le sentences are normally used in informative context in which the information is hearer-new and the speaker has the intention to provide the new information for the hearer. This has also been noted in some prior studies (Sybesma 1999, Liu 2002). However, distinct from those studies, which only consider it a condition for some S-le sentences, I suggest that hearer-new is a generalization for all S-le sentences. I also showed that how the analysis of hearer-new information helps us understand the pragmatic effects generated when S-le sentences are used in contexts in which the hearer-new presupposition fails.

Though I did not provide a full-fledged formal analysis for S-le sentences, I suggest an analysis on the basis of the analysis proposed by Soh (2009) and show how we may extend her analysis in order to account for hearer-newness. Following her, I assume that informativeness is a kind of presupposition which is characterized in terms of common ground in a Stalnakerian framework. I propose that S-le sentences presuppose that the asserted proposition p is not accepted on the current common ground, meaning either ¬p is accepted, which is studied by Soh (2009) to account for interpretations of “change of state” and “contrary to one’s expectation”, or there is no presupposition about p at all, which corresponds to the basic informative case. This

117 analysis can provide an account for both cases; however, I leave the details for future work.

3.4 THE S-LE SENTENCE: FREE OF GRAMMATICAL ASPECT

Many analyses of the S-le sentence have focused on their apparent contribution to tense or aspect rather than discourse (e.g. Rohsenow 1978, Li &

Thompson 1981, Li et al. 1982, Huang & Davis 1989; Shi 1990, Bisang & Sonaiya 1997, Liu 2002, Yang 2003). But now it has been widely accepted that Chinese is a language that has no grammatical tense and the interpretation of time is actually determined by the context or temporal adverbials. In this section, I focus on discussing the issue of the aspectual properties associated with S-le sentences. Different from prior studies, which argue for a certain aspectual property for S-le sentences (e.g. Li et al. 1982, Shi 1990, Bisang & Sonaiya 1997), I argue that S-le sentences are free of any specific grammatical aspect. The aspectual peculiarities of S-le sentences that I discuss below have partially been noted in prior studies (e.g. Shi 1990, Bao 2005); however, I show that there actually is a bigger picture that those prior studies did not notice by which S-le sentences actually have no inherent aspectual meaning at all. In Section 3.4.1 I first review arguments from some prior studies that suggest a specific tense or aspectual analysis for S-le sentences. In Section

3.4.2 I present the argument that S-le sentences are free of grammatical aspect. It is notable that the temporal approach to S-le sentence adopted in prior studies has an historical origin. According to Spencer (1970: 4-5), in early studies on the verbal system of Chinese, Indo-European linguists attempted to equate the Chinese verbal system with the Indo-European tense system, which, however, turned out to be unsatisfactory. Meanwhile Slavic grammarians felt that the Chinese verbal system is closer to the aspect system of Slavic languages, and introduced the concept of aspect into the study of Chinese languages. Those early studies on the Chinese verbal system by Western linguists have greatly influenced later studies of Chinese.

118

Here I present several analyses of S-le sentences in terms of tense or aspect and discuss their limitations and inadequacy in accounting for the data.

3.4.1 Prior studies

Below I outline three prior studies that argued for a tense or aspect analysis for S-le sentences. I will discuss the limitations of those studies and show that the data that were used to motivate the analysis actually can be accounted for by the generalization of hearer-new information. As reviewed in Section 3.3.1, Liu (2002) argued that S-le sentences have the function of providing hearer-new information and express deviation from the standard when used with adjective predicates. However, he also argued that S-le marks past tense, and he considered past tense the ultimate analysis for S-le. But a problem in his argument is that he did not say anything about how the past tense analysis of S-le can account for the other two meanings. In fact we know that Chinese is a tenseless language 9 and the interpretation of time is determined by contexts or specific temporal expressions. Below I outline his argument for a past tense analysis, discuss how the analysis fails to account for the data, and show that those examples can also be explained by the analysis of hearer-new information. Liu (2002) observed that S-le is usually used to indicate that the denoted activity or reported state of affairs has been done or is over by utterance time, so it should be considered a marking for past tense. Since the denoted event may not be a past event, he proposed that S-le expresses past tense with a flexible Reference Time (RT), that is, the denoted event precedes RT. RT is utterance time by default but it may change to some other time depending on other temporal clues. Consider one of his examples that illustrate utterance time is the RT.

9 If we assume the distinction between relative past tense and absolute past tense, then Chinese has relative past tense, which is marked by perfective V-le (Comrie 1976: 2). 119

(53) A1: wǒ 38 suì le. [Liu 2002: 75]

I 38 year LE ‘I’m 38 years old.’ B: shénme, nǐ bú shì 64 nián chūshēng de ma?

what you NEG COP 64 year born DE Q ‘What? Weren’t you born in 1964?’

A2: duìbùqǐ, wǒ shuō de shì míngnián 38 suì le.

sorry I say DE COP next.year 38 year LE ‘Sorry, what I meant is that I will be 38 years old next year.’

According to Liu, this dialogue suggests that the interpretation of (53A1) by the hearer B is that the speaker has turned 38 years old by utterance time. Note that the analysis of hearer-new information also applies to (53A1), that is, it is new information for the hearer that the speaker is 38 years old. Without le, the sentence simply describes a property of the subject. Consider another two of his examples that illustrate a different RT.

(54) 25 wàn nián qián, zhōukǒudiàn-yuánrén [Liu 2002: 75] 25 ten.thousand year ago homo.erectus.pekinensis jiù zhīdào yòng-huǒ le.

already know use-fire LE ‘250,000 years ago, the homo erectus pekinensis already knew how to use fire.’ (55) míngnián, wǒ gāi dàxué bìyè le. [Liu 2002: 75]

next.year I shall university graduate LE ‘I shall graduate from university next year.’

The example (54) was used to illustrate a past event and the RT is in the past, which is indicated by the temporal adverbial ‘250,000 years ago’. The example (55) was used to illustrate a future event and the RT is also in the future, which is indicated by the temporal adverbial ‘next year’. Note that the hearer-new analysis also applies to the 120 two examples. For example, an appropriate context for (54) is like this: the hearer told the speaker that he learned that some homo species knew how to use fire 200,000 years ago and he thought it was marvelous, which suggests that he did not know that some homo species started to use fire even earlier than this date. The speaker uses (54) to provide new information as a response. Since the new information is more marvelous, the speaker is implying that what the hearer learned is nothing marvelous.

An appropriate context for (55) is like this: The speaker does tutoring jobs while attending university and sometimes the hearer helps him find students for tutoring. This time the hearer is talking about looking for students to tutor in the next year. The speaker uses (55) to inform the hearer that he is graduating next year, implying that he does not want to do tutoring work next year. Note that the speaker may still use (55) as a response even if the hearer knows that he is graduating next year; however, this is usually a case where the hearer is not aware of this fact in the immediate communication situation. But there are many data that actually cannot be accounted for by a past tense analysis with flexible RT. As Liu noted, some S-le sentences do not denote relative past events but events that are about to happen. However, given this set of data, he did not reject the past tense analysis; instead, he provided a specific account for those data regardless of the past tense analysis. For example, the sentence in (56) is usually used in a context when the meal is about to begin, so it does not denote a past event or describe a past situation.

(56) kāi-fàn le. [Liu 2002: 77]

open-meal LE ‘The meal is ready.’

Liu’s explanation for the near future reading of (56) is that the sentence expresses that the new state of affairs (i.e. to have the meal) is temporally moving close, giving rise to an interpretation, due to a certain psychological effect, that the new activity is about

121 to come about. But he did not explain why the meaning of past tense as he proposed for S-le is prohibited in this case. In fact, (56) can also be explained by the hearer-new analysis. The context for using (56) with a near future reading is a context in which the hearer has been anticipating the meal but does not know that the meal is ready now. So by using (56) the speaker is informing the hearer that the meal is ready now. The near future interpretation is determined by the hearer’s knowledge state that the meal will be ready soon. However, a tense analysis cannot be the right analysis for the S-le sentence. Liu missed a simple generalization that the S-le sentence does not specify any inherent temporal meaning for the denoted event. The time of the event denoted by an S-le sentence may be interpreted as relatively past to the RT, future to the RT, or overlapping with the RT, depending on the context. For example, it was noted by Bao (2005) that there are two aspectual interpretations associated with the S-le sentence.

One is what he called inceptive aspect, which means that an event is about to happen, as in (57i), and the other one is what he called inchoative aspect, which means a generic event will start to happen but this kind of event never happened before, as in (57ii) (The translation for (57ii) is provided by me). The two aspectual interpretations indicate that the event described by an S-le sentence may not be an event for which the event time is interpreted as past relative to the RT; instead, it could be an event for which the event time is in the future relative to the RT.

(57) wǒmen chī-liúlián le. [Bao 2005: 242, (11b)]

we eat-durian LE (i) ‘We are about to eat durian.’ (ii) ‘We eat durian now (and we never did so before).’

Now consider another example from Soh (2009) shown in (58). According to Soh, the

S-le sentence may not entail termination for the denoted event, that is, the event may still be progressing at utterance time. The continuation of the discourse (58) suggests

122 that the event of writing the letter described by the S-le sentence is still going on at utterance time. This example shows that the event time might overlap with RT if we assume that the utterance time is RT in this case.

(58) tā jīntiān zǎoshàng xiě nà-fēng xìn le. [Soh 2009: 636, (26b)]

he today morning write that-CL letter LE xiě dào xiànzài hái bù kěn tíng.

write until now still NEG willing stop ‘He was writing the letter this morning (, which he was not doing before/ contrary to what one may expect). He has been writing till now and is still unwilling to stop.’

So tense actually is irrelevant for the S-le sentence as the temporal interpretation for the event described by an S-le sentence is actually determined by the context and background knowledge or temporal adverbials. Li et al. (1982) is the most influential study of S-le sentences under the notion of aspect. They proposed that S-le expresses a particular kind of Perfect aspect which they called Currently Relevant State (CRS). CRS expresses the notion that “a state of affairs has special current relevance to some particular Reference Time” (p. 22). The RT is determined by the context, but is often understood as the utterance time.

Consider the sentence presented below, which is one of the examples that they used to illustrate the notion CRS. For this sentence, they assumed the following scenario: Someone calls Mr. Liao but he is out, so the person who answers the phone utters (59).

(59) tā qù mǎi dōngxi le. [Li et al. 1982: 23, (24)]

he go buy thing CRS

‘He’s gone for shopping.’

They used this example to illustrate the notions “current” and “relevant” contained in

123 the definition of CRS. First, according to them, S-le in this example expresses that his having gone shopping is “current” with respect to the RT, which is the utterance time given this scenario. In other words, it expresses a temporal relation for the denoted event or state of affairs and the RT. Second, according to them, S-le is used to indicate that the state of affairs expressed by the sentence is “relevant” for the speaker and the hearer, and the speaker assumes that the hearer can figure out in which way it is relevant to him. In this example, S-le here indicates that the state of affairs his having gone shopping is “relevant” for the hearer. Since the hearer wants to talk to Mr. Liao, once having learned that Mr. Liao is gone for shopping, the hearer should be able to figure out that he cannot talk to him. From this example, we can see that Li et al. (1982) actually included two types of relations of relevance in their analysis without considering which one is more basic. One is a temporal relation, that is, they assume that there is an RT and S-le sentences express that the state of affairs is relevant to it. The other one is a non-temporal relation, that is, the state of affairs is relevant to the speaker and the hearer in the communicative situation. However, we should take the non-temporal relation of relevance as the basic relation for an analysis of the S-le sentence because what shows up in all the data is the non-temporal relation, not the temporal relation. Below I discuss another two more of their examples and show that the non-temporal relation alone is sufficient to account for them. First, consider their explanation of why the S-le sentence may sometimes have an interpretation of change of state: the state of affairs denoted by the proposition of the S-le sentence represents a state that holds now but did not hold before. Therefore, the relevance of the state of affairs to the present situation, which is expressed by CRS, gives rise to the reading of change. In this account, Li et al. largely relied on the temporal relation between the state of affairs and the RT, which is the utterance time.

However, we have reason to believe that the temporal relation is only an implication in the cases when an interpretation of change of state obtains for the S-le sentence. As mentioned above, the S-le sentence gives rise to an interpretation of change of state 124 only when it is used in the context in which the hearer knows the earlier state of affairs before the change but does not know that later there was a change which resulted in a new state of affairs. The speaker uses the S-le sentence not because the described state of affairs is relevant to the present situation, but because it is new information for the hearer, or in Li et al.’s term, it is relevant for the hearer. That there is a moment of time at which the change of state occurs is a precondition for the interpretation of change of state. The hearer can infer a change occurs since the new state of affairs he just learned about is in contrast with his background knowledge. So it is really just about the hearer’s knowledge of the proposition. Consider another example which Li et al (1982) used to illustrate the interpretation of correcting the hearer’s incorrect assumption (or being contrary to the hearer’s expectation). In this example, the hearer, which is a duck flying north, holds the incorrect assumption that he is flying south.

(60) (The dog Scamp calls out to his friend the duck who is flying past him supposedly migrating south) [Li et al. 1982: 34, (66)] hāi, Guā-Gua! nǐ wǎng běifāng qù le!

Hey Quack-Quack you toward north go CRS ‘Hey, Quacky! You’re going north (and not south, as you obviously are assuming!)’

In this case, Li et al. claimed that it is not about that the described state of affairs is relevant to the current situation; instead, S-le here is used to express that the described state of affairs is relevant for the hearer. So they employed the non-temporal relevance relation to account for the data in this case. Overall, we can see that of the two types of relations of relevance included in Li et al.’s analysis, the non-temporal relation is the most fundamental as it can account for all of the data. The temporal relation of relevance relative to the RT might be implied in some cases, such as the case when the interpretation of change of state is

125 obtained. However, it is not general enough to account for all the data. Shi (1990) alternatively proposed an analysis rooted in lexical aspect. Following the general approach of lexical aspect such as Vendler (1967) and Dowty (1979), Shi (1990) proposed two types of situations. One type of situation, which he called an unbounded situation, is a situation that has an initial boundary in the intuitive sense that a situation has a beginning but no specific terminal boundary, including states and activities. The other type of situation, which he called a bounded situation, is a situation that has both initial boundary and terminal boundary, including accomplishments and achievements. He did not consider just S-le sentences but also V-le sentences. He focused on similarities rather than differences between the two sentence structures and proposed a unified analysis for V-le and S-le. The similarity is that both V-le and S-le sentences entail completion when the predicate denotes a bounded situation, and both of them may have an inceptive reading (i.e. inception of a new situation or state of affairs) if the predicate denotes an unbounded situation. The data are provided below. Sentences (61a) and (62a) respectively illustrate a V-le sentences with a VP that denotes a bounded situation and a V-le sentences with a VP that denotes an unbounded situation, which has an inceptive reading (note that (62a) might be better translated as ‘Zhangsan acquired a big sum of money’ as the sentence does not entail Zhangsan possesses a big sum of money at the speech time.).

Sentences (63a) and (64a) illustrate the same for the S-le sentence respectively.

(61) a. tā shā-le yī-zhī jī. [Shi 1990: 107, (23)]

he kill-LE one-CL chicken ‘He killed a chicken.’ b. tā shā yī-zhī jī.

he kill one-CL chicken

‘He is to kill a chicken.’

126

(62) a. Zhāngsān yǒu-le yí dà bǐ qián. [Shi 1990: 108, (24)]

Zhangsan have-LE one big sum money ‘Zhangsan now has a big sum of money.’ [‘Zhangsan acquired a big sum of money.’] b. Zhāngsān yǒu yí dà bǐ qián. Zhangsan have one big sum money

‘Zhangsan has a big sum of money.’ (63) a. tā qù Niǔyuē le. [Shi 1990: 110, (29)]

he go New.York LE ‘He went to New York.’ b. tā qù Niǔyuē. he go New.York ‘(It is expected of him that) He goes to New York.’

(64) a. tā chī ròu le. [Shi 1990: 110, (28)]

he eat meat LE ‘He now eats meat.’ b. tā chī ròu. he eat meat ‘He eats meat.’

Shi proposed that V-le and S-le denote the same function in that they are a kind of temporal operator, which he called relative anteriority marker, that refers to the posterior of the terminal boundary if the situation is a bounded situation, which entails completion, and refers to the posterior of the initial boundary if the situation is an unbounded situation, which, by stipulation, gives rise to an inceptive reading. This is illustrated by the diagram below (Shi 1990: 106).

127

(65) a. bounded situations

initial boundary terminal boundary

Time axis Relative anteriority marker

b. unbounded situations

initial boundary

Time axis Relative anteriority marker

This analysis explains straightforwardly why a V-le or S-le sentence entails completion if the VP denotes a bounded situation (e.g., kill a chicken and go to New York), and why the sentence has an inceptive reading if the VP denotes an unbounded situation (e.g. have a big sum of money and eat meat). It is the boundedness property of the VP predicate rather than V-le or S-le that determines the interpretation of the sentence; the particles V-le and S-le are only used to refer to a place on the time axis relative to the terminal boundary or the initial boundary that are defined by the boundedness of the VP. It is notable that Shi’s (1990) study was not aimed at addressing the question of how the V-le sentence and the S-le sentence differ from each other, in which case he would need to consider the condition that the proposition of the S-le sentence is new to the hearer. Instead, he was largely concerned with the aspectual interpretation of the sentence when the sentence structure, the V-le or S-le sentence, interacts with VPs of different lexical aspectual properties. Below I review discussion from the literature on one set of data discussed in Shi’s study. Shi discussed how a stative verb like yǒu ‘have’ as in (62) has an inceptive reading when followed by V-le. However, it is notable that V-le cannot be productively suffixed with stative verbs, and those data used in Shi’s study should be

128 treated as special cases. Consider some examples from Soh (2009) and Ross (1995) below. Sentence (66) shows that V-le may occur with the stative zhīdào ‘know’; however, sentences (67), (68) and (69) show that V-le cannot occur with stative verbs in general.

(66) tā zhīdào-le nèi-gè xiāoxī.

he know-LE that-CL news

‘He got to know/learned that piece of news.’ (67) *wǒ xiǎng-le jiā. [Soh 2009: 626, (5a)]

I miss-LE home (Intended)‘I came to miss home.’ Or ‘I missed home.’ (68) *tā xiàng-le bàba. [Ross 1995: 110, (31)]

he resemble-LE father

(Intended)‘He came to resemble his father.’ Or ‘He resembled his father.’ (69) *tā ài-le Mǎlì.

he love-LE Mary (Intended)‘He fell in love with Mary.’ Or ‘He loved Mary.’

For those stative verbs that can occur with V-le such as yǒu ‘have’ and zhīdào ‘know’, Shi was right in saying that they have inceptive readings. But such verbs are very few.

Soh (2009) and Ross (1995) assumed that those sentences describe achievements rather than states. I agree with them. An alternative way of analyzing this set of data is to analyze the verbal suffix le as the inchoative le (see Section 3.2.4), which turns the stative verbs yǒu ‘have’ and zhīdào ‘know’ into lexical items that have the meaning of change of state, that is, ‘to become to have’ and ‘to become to know’. This is consistent with the analysis assumed in Soh (2009) and Ross (1995) since both change-of-state verbs and achievements imply in their denotations that there is a moment at which the subject undergoes a certain change of state. To sum up, I presented three theoretical analyses under the notion of

129 temporality proposed in the literature. For Liu’s (2002) tense analysis for S-le, I argued that the tense interpretation of an S-le sentence is actually understood from the context or the discourse. For Li et al.’s (1982) grammatical aspect analysis, which says S-le expresses a Currently Relevant State (CRS), that is, the state of affairs expressed by the proposition is relevant to the RT, I argued that they actually conflated two relations in the notion CRS, one a temporal relation between the state of affairs and the RT and the other one a non-temporal relation between the state of affairs and the hearer. However, as argued above, what is fundamental for the S-le sentence is the non-temporal relation, while the temporal relation is not always relevant. Finally, I reviewed Shi’s (1990) analysis of V-le and S-le sentences, which is based on the framework of lexical aspect. He proposed a unified analysis for V-le and S-le and provided an account for their similarity with respect to their lexical aspectual interpretations. However, this study cannot explain how the S-le sentence differs from the corresponding V-le sentence. Overall, we can see that temporality cannot provide a sufficient account for what we have observed for the S-le sentence.

3.4.2 The S-le sentence: Free of grammatical aspect

The S-le sentence is a sentence structure that is free of grammatical aspect. The data presented below provide evidence for this claim. The data show that S-le sentences do not seem to have an intrinsic grammatical aspect which would restrict it to the kind of informative context that requires that specific grammatical aspect for the proposition. Instead, the data show that S-le sentences can be used in different informative contexts which require different aspectual meanings for the propositions. This can only be explained if we assume that the S-le sentence is free of grammatical aspect, and it is the contextual information and background knowledge that provide the basis for the inference of a given aspectual interpretation for an S-le sentence. So an aspectual analysis will not work since the S-le sentence is actually unspecified for aspect.

130

In order to show that there really is an aspectual interpretation involved in each case, I also provide the corresponding sentence overtly marked with the grammatical aspect (the Asp-sentence) that the context requires, includes perfective, progressive, ingressive, prospective, and future generic. These Asp-sentences explicitly express the aspectual meanings of the propositions implied if the S-le sentences are used instead. In terms of language use in the daily life, the S-le sentences and the Asp-sentences are interchangeable in informative contexts. Of course, the Asp-sentences and the S-le sentences are semantically (and syntactically) different – for example, S-le sentences express a belief that the proposition is new to the hearer while Asp-sentences do not.

Perfective aspect10 Consider the following conversation where B tells A an event that A does not know about that happened in the past. Both the S-le sentence (70B) and the V-le sentence (70B’), which is marked with perfective aspect (the verbal suffix V-le), are equally acceptable here, and they convey the same information to the hearer.

(70) A: Yuēhàn zuótiān gàn-shénme le?

John yesterday do-what LE ‘What did John do yesterday?’

B: tā dǎsǎo-Mǎlì-de-chúfáng le.

he clean-Mary-DE-kitchen LE

10 It has largely been agreed on that the V-le is a perfective aspect in Mandarin (e.g., Chao 1968, Li & Thompson 1981, Smith 1990, Soh 2014). However, perfectivity is characterized differently for different linguists. For example, Smith (1997) considered it a perfective aspect marker that does not entail culmination but analyzed it as expressing termination. Li & Thompson (1981) analyzed it as expressing boundedness. Note that in Chinese, perfectivity does not entail telicity (see Chapter 4). In the current study, I assume that perfectivity in Chinese, which is expressed by V-le, is a temporal precedence relation of event time of the denoted event and a reference time; it presupposes a time point/interval as the event time for the denoted event. This analysis should be apparent from the data given in this section as this set of Asp-sentences only differ in the temporal relation between the event time and the reference time. 131

B’: tā dǎsǎo-le Mǎlì de chúfáng.

he clean-LE Mary DE kitchen ‘He cleaned Mary’s kitchen.’

Progressive aspect Consider the following conversation where A informs B of a situation that B does not know about that happens at utterance time. Both the S-le sentence in (71A) and the progressive sentence in (71A’) are equally acceptable here.

(71) A: tā tōu-wǒmen-jiā-de-yǘ le. wǒmen kuài qù zǔzhǐ tā.

he steal-our-family-DE-fish LE we quickly go stop him A’: tā zài tōu wǒmen jiā de yǘ. wǒmen kuài qù zǔzhǐ tā.

he PROG steal our family DE fish we quickly go stop him ‘He is stealing our fish. We should go to stop him quickly.’

B: shì me? nà wǒmen kuài qù.

COP Q then we quickly go ‘Really? Then let’s go quickly.’

Ingressive aspect Consider the following conversation where B tells A something A does not know. The denoted event just started and thus the aspectual meaning is ingressive.

Note that Chinese uses the aspectual auxiliary verb kāishǐ ‘start’ to express ingressivity. Both the S-le sentence (72B) and the ingressive form in (72B’) are acceptable and receive the same interpretation.

(72) A: Mǎlì zuìjìn hǎoxiàng hěn máng a.

Mary recently seem very busy SFP

‘Mary seems busy recently.’ B: shì a, tā xiě-bìyèlùnwén le.

yes SFP she write-dissertation LE

132

B’: shì a, tā kāishǐ xiě-bìyèlùnwén le.

yes SFP she start write-dissertation LE ‘Yes. She has started to write the dissertation.’

Prospective aspect Consider the following conversation where A tells B something that B does not know. The denoted event is about to happen and thus the appropriate aspect here is prospective. One way to express prospective aspect in Chinese is to use the modal verb yào ‘want to, be about/going to’. Both the S-le sentence in (73A) and the prospective form in (73A’) are acceptable and have the same interpretation here.

(73) A: wǒ sǎo-xuě le. nǐ yào yìqǐ sǎo ma?

I sweep-snow LE you want together sweep Q A’: wǒ yào sǎo-xuě le. nǐ yào yìqǐ sǎo ma?

I be.about.to sweep-snow LE you want together sweep Q ‘I’m going to sweep the snow. Do you want to do it together?’ B: hǎo-de. Okay ‘Okay.’

Future generic11

Consider the following conversation where B tells A something that A does not know as a response. The conversation assumes certain background knowledge shared between the two interlocutors, that is, John never read books written by American authors before. When A tells B that he saw John reading The Old Man and The Sea, which is surprising given the background knowledge, B informs him that John actually has started to read books written by American authors, which is thus hearer new information. The aspectual meaning of (74B) is a kind of change of state which

11 Bao (2005) called this aspect inchoative (see (56) above), which captures a reading of change of state. I call it future generic here since the interpretation is that the proposition will hold true generically from now on. 133 involves a future generic reading. Chinese uses the aspectual auxiliary verb kāishǐ ‘begin’ to express such an aspectual meaning. Both the S-le sentence in (74B) and the Asp-sentence in (74B’) are acceptable and receive the same aspectual interpretation here.

(74) A: wǒ zuótiān qù Yuēhàn jiā. I yesterday go John home

‘I went to John’s home yesterday.’ tā jūrán zài kàn Lǎorén yǔ Hǎi.

he surprisingly PROG look old.man and sea ‘Surprisingly, he was reading The Old Man and The Sea.’ B: shì a, tā kàn-měiguó-zuòjiā-xiě-de-xiǎoshuō le.

yes SFP he look-American-writer-write-DE-novel LE

B’: shì a, tā kāishǐ kàn měiguó zuòjiā xiě de xiǎoshuō le.

yes SFP he start look American writer write DE novel LE ‘Yes. He has started to read novels written by American authors.’

To sum up, the data provided above show that a corresponding Asp-sentence may occur equally well in an informative context where an S-le sentence occurs. The data suggest that the S-le sentence, therefore, does not express any aspect by itself since it can be used in contexts that define different aspectual meanings and can have the same aspectual interpretation as a corresponding sentence with any overt aspectual marker. The aspectual information, which is needed for interpretation, is understood from context and background knowledge.

3.5 INFORMATIVE CONTEXTS WITH DIFFERENT FOCUSES AND GIVENNESS

The notion of an informative context has been understood so far as the kind of contexts in which the proposition the speaker tells the hearer is new for the hearer. In this section, I show that there are four sentence types, each with its own discourse use,

134 and two of them also occur in informative contexts of a narrower type and in those contexts they block S-le sentences. The four sentence structures we deal with below include V-le sentences, S-le sentences, Object Preposing sentences, and ba-sentences. An example of each is presented as follows.

(75) a. wǒ mǎi-le zìxíngchē. (V-le sentence)

I buy-LE bike ‘I bought a bike.’ b. wǒ mǎi-zìxíngchē le. (S-le sentence)

I buy-bike LE ‘I bought a bike.’ c. wǒ zìxíngchē mǎi-le. (Object Preposing sentence)

I bike buy-LE ‘I bought the bike.’ d. wǒ bǎ zìxíngchē mǎi-le. (ba-sentence)

I BA bike buy-LE ‘I bought the bike.’

The four sentence structures are used in different ways, corresponding to different types of informative context.

3.5.1 Contexts for S-le and the V-le sentences

The S-le sentence and the V-le sentence are equally acceptable in the kind of informative context where the focus of the proposition is either the VP (verb + object) or just the object. Following Jackendoff (1972), what I mean by focus is the new information for the hearer in an utterance. The two kinds of focuses are illustrated below by (76) and (77) respectively. For the data of the V-le sentence provided below, see Hole (2012: 46) for the same judgment regarding focus.

135

In the following example, the focus of (76B) is the VP, i.e. the entire proposition expressed by the VP is unknown to the hearer. The conversation suggests that the hearer A has almost no background knowledge relative to the proposition expressed by the speaker B except the agent.

(76) A: nǐ xiàwǔ zài jiā gàn-shénme le?

you afternoon LOC home do-what LE

‘What did you do at home this afternoon?’

B1: wǒ xiū-le zìxíngchē.

I repair-LE bike

B2: wǒ xiū-zìxíngchē le.

I repair-bike LE ‘I repaired a/the bike.’

The V-le sentence (76B1) and the S-le sentence (76B2) are equally acceptable in this context, suggesting that they are compatible with an informative context with a focus on the VP. In example (77), the focus of (77B) is the object. The conversation suggests that the hearer A knows of the subject and the action described by the verb but does not know what the patient of the verb was, i.e. the object introduces a new referent new to the hearer.

(77) A: Mǎlì shuō nǐ xiàwǔ zài jiā xiū dōngxi.

Mary say you afternoon LOC home repair thing ‘Mary said you were repairing things at home this afternoon.’

B1: shì a. wǒ xiū-le zìxíngchē.

COP SFP I repair-LE bike

B2: shì a. wǒ xiū-zìxíngchē le.

COP SFP I repair-bike LE ‘Yes, I repaired a/the bike.’ 136

Both the V-le sentence (77B1) and the S-le sentence (77B2) are equally acceptable in this context, suggesting that they are compatible with an informative context with a focus on the object. The two examples above show that both the S-le sentence and the V-le sentence can equally be used in the kind of informative contexts in which the focus of the informing proposition is the VP (or event) or the object of the VP.

3.5.2 Informative contexts where S-le and V-le sentences are excluded

The data presented below illustrate cases when neither the V-le sentence nor the S-le sentence is generally felicitous, if not totally unacceptable. There are two types of informative contexts in which the two sentence structures are excluded. The first one is the kind of informative context in which the hearer knew the denoted event was likely to happen but did not know whether it actually did happen. In this case, the

Object Preposing sentence structure is preferred. The other context is a kind of informative context in which the entity denoted by the object of the VP is established in the immediate context and the hearer presupposes that something happens to the entity but does not know what. In this case, the ba-sentence is used.

3.5.2.1 Object Preposing structure: The hearer has prior knowledge of the event Example (78) illustrates the kind of informative context where the hearer knew that the denoted event was likely to happen, for example, because s/he and the speaker talked about it earlier. In this example, the background is that the interlocutors talked about buying a bike last week, so the hearer knew an event of buying a bike was likely to happen. The most natural sentence structure used in this context is a structure called Object Preposing in the literature (e.g. Ernst & Wang 1995, Shyu

1995, 2001, Paul 2002, 2005, 2015) as shown in (78A3). It is called Object Preposing because the object in this structure is preposed from the canonical postverbal position to a non-canonical preverbal position. The V-le sentence (78A1) and the S-le sentence

(78A2) are not quite acceptable in this context.

137

(78) Background: The interlocutors talked about buying a bike last week.

A1: #wǒ mǎi-le zìxíngchē.

I buy-LE bike

A2: #wǒ mǎi-zìxíngchē le.

I buy-bike LE

A3: wǒ zìxíngchē mǎi-le.

I bike buy-LE ‘I bought the bike.’

The negation of an Object Preposing sentence also presupposes that the hearer knew that the denoted event was likely to happen. An example of the negation of (78A3) is provided as follows.

(79) wǒ zìxíngchē méi mǎi. ràng tā zuò gōngjiāochē qùxiào xué ba.

I bike NEG buy let him take bus go school SFP ‘I didn’t buy the bike. Let him take the bus to school.’

This discourse is only compatible with a situation where the hearer knew there might be an event of buying a bike, for example, the interlocutors talked about buying a bike earlier, but now learns that there was not.

The meaning of Object Preposing sentences can be paraphrased as “As for the

VPing that we discussed before, it has been done”. They have a discourse function that they can be used to shift the flow of communication to an earlier topic. Therefore, they are the appropriate expression for the type of informative context illustrated by (78) where the speaker is making reference to the hearer’s background knowledge about the VPing. The S-le (as well as the V-le) sentences do not have such a discourse function and thus are excluded from this context, despite the fact that the occurrence or non-occurrence of the event is new to the hearer.

138

3.5.2.2 Ba-sentence: The object is given There is another type of informative context in which the S-le sentence as well as the V-le sentence and the Object Preposing sentence are excluded, but the ba-sentence is the most natural structure to use. The focus of the informing sentence is also the event (VP) and the object of the sentence is given or in the immediate context. Consider the example (80), in which case, the bike has been given in the immediate context. The V-le sentence (80B1), the S-le sentence (80B2), and the Object

Preposing sentence (80B3), if compared with the ba-sentence (80B4), do not fit well in this conversation.

(80) A: zěnme méi kànjiàn tā-de zìxíngchē?

why NEG see his bike ‘Why I don’t see his bike?’

B1: #tā fùqin mài-le zìxíngchē.

his father sell-LE bike

B2: #tā fùqin mài-zìxíngchē le.

his father sell-bike LE

B3: #tā fùqin zìxíngchē mài-le.

his father bike sell-LE

B4: tā fùqin bǎ zìxíngchē mài-le.

his father BA bike sell-LE

‘His father sold the bike.’

The judgments for the examples above are consistent with what Li & Thompson (1981: 483-484) considered as the appropriate sentence structure in the same kind of context. Their judgment is that the ba-sentence is more likely to be used when the object is prominent in the speech context. One of their examples is presented below with minor modification.

139

(81) a. qǐng nǐ qiē xiē cōng. [Li & Thompson 1981: 483-4, (75)] please you cut some green.onion ‘Please cut some green onions.’ b. qǐng nǐ bǎ (zhè-xiē) cōng qiē-yíxia.

please you BA (this-PL) green.onion cut-ASP ‘Please cut these green onions.’

According to them, if the situation is that the speaker wants the hearer to cut some green onions for a dish the speaker is making, and the green onions are not within sight, for example, they are in the refrigerator, then a non-ba-sentence like (81a) might be used as a request. However, if the green onions stand out in the speech context, for example, they are being pointed at or being held in the speaker’s hand, then the ba-sentence (81b) is more likely to be used.

The ba-construction is quite similar to those cases when a definite article is used for languages like English, and thus it is the most appropriate construction for a context in which the entity denoted by the object noun phrase is prominent for the interlocutors. Therefore, in the informative context (80), since the entity denoted by the object NP zìxíngchē ‘bike’ is prominent and identifiable for the interlocutors, the ba-sentence is preferred over the S-le sentence, the V-le sentence and the Object Preposing sentence, none of which carry this definite-like property.

But note that the ba-sentence is subject to a lexical semantic constraint which would exclude it from such informative contexts. The constraint is that it is usually restricted to verbs that have the meaning of affectedness or disposal (Wang 1945/2002, Li 1974, Sybesma 1999, Shen 2002).12 Consider (82). In this informative context the

12 Two of the examples from Wang (1945/2002) are presented below. The two verbs ài ‘love’ and kànjiàn ‘see’ do not have the meaning of “disposal” or “affectedness” and they cannot be used in ba-sentences.

(1) a. wǒ ài tā. [Wang 2002: 165] I love him b. *wǒ bǎ tā ài. I BA him love ‘I love him.’

140 object lóutī ‘stair’ is given and prominent in the discourse but the ba-sentence (82B1) is excluded; instead, a V-le sentence (82B2) or S-le sentence (82B3) is more natural in this context.

(82) A: zhè lóutī zěnme huài-le?

this stair how break-LE ‘How come this stair is broken?’

B1: #Zhāngsān xiàwǔ bǎ lóutī shàng-le, kěnéng shì tā nòng-huài de.

Zhangsan afternoon BA stair ascend-LE maybe COP he make-break DE

B2: Zhāngsān xiàwǔ shàng-le lóutī, kěnéng shì tā nòng-huài de.

Zhangsan afternoon ascend-LE stair maybe COP he make-break DE

B3: Zhāngsān xiàwǔ shàng-lóutī le, kěnéng shì tā nòng-huài de.

Zhangsan afternoon ascend-stair LE maybe COP he make-break DE

‘Zhangsan ascended the stairs this afternoon. Maybe he broke it.’

Note that in this example, it is not that the ba-sentence (82B1) is unacceptable in this context. It is rather that the sentence itself is unacceptable or unnatural because the ba-construction needs to be used with verbs that have the meaning of affectedness or disposal, but the verb shàng ‘ascend, go up’ does not have such meaning. Although ba-NP overtly expresses definiteness and thus we would expect ba-sentence to be the most appropriate structure in this context, the example suggests that the requirement of the meaning of affectedness or disposal is more important. Also note that given the context (82), the object lóutī ‘stair’ of the V-le sentence (82B2) and the S-le sentence

(82B3) should be interpreted as the stair(s), though there is no overt marking for definiteness – Mandarin has no definite article and bare nouns can be interpreted as a definite noun phrase, depending on other factors.

(2) a. wǒ kànjiàn tā. [Wang 2002: 165] I see him b. *wǒ bǎ tā kànjiàn. I BA him see ‘I see/saw him.’ 141

The ba-sentence has received much discussion in the literature (e.g., L. Wang 1945/2002, Lü 1955, Chao 1968, Thompson 1973, Mei 1978, Li & Thompson 1981, Sybesma 1999, Shen 2002, Li 2006, Lipenkova 2014, Paul 2015). On the semantics of the ba-sentence, a large body of the discussion has focused on a reading of disposal or affectedness. This reading is a major concern for studying the meaning of ba-sentence since most ba-sentences have such interpretation. Another line of discussion is centered on the notion of transitivity (e.g., Thompson 1973, Hopper & Thompson 1980, Tsao 1987, Sun 1995), and the proposal is that the ba-sentence expresses a transitive relation between the ba-NP and the rest of the sentence. For the purpose of the current study, we will not go through the details of the discussions in those studies.

3.5.3 The S-le sentence: The only structure that expresses informativeness

We have discussed four sentence structures above, which are summarized as follows.

(83) a. Subj V-Obj le. (S-le sentence) b. Subj V-le Obj. (V-le sentence) c. Subj Obj V-le. (Object Preposing) d. Subj ba-Obj V-le. (ba-sentence)

In Section 3.3, I showed that the S-le sentence shows what may be a kind of Gricean effect when used in a non-informative context, i.e., a context in which the proposition is already known to the hearer. But note that S-le sentences are normally unacceptable in non-informative contexts if no such implicature is intended, as in (1). This contrasts with the other three constructions (83b-d) as they can be used in non-informative contexts without conveying additional implicatures. This can be explained if we assume that non-informative contexts are marked for informativeness but allows for constructions that are neutral in informativeness. The set of data are presented in (84).

142

(84) Context: The hearer already knows the event of buying a bike. a. *wǒ mǎi-zìxíngchē le, zhīhòu jiù huí-jiā le. (S-le sentence)

I buy-bike LE afterwards then return-home LE b. wǒ mǎi-le zìxíngchē, zhīhòu jiù huí-jiā le. (V-le sentence)

I buy-LE bike afterwards then return-home LE c. wǒ zìxíngchē mǎi-le, zhīhòu jiù huí-jiā le. (Object Preposing)

I bike buy-LE afterwards then return-home LE d. wǒ bǎ-zìxíngchē mǎi-le, zhīhòu jiù huí-jiā le. (ba-sentence)

I BA-bike buy-LE afterwards then return-home LE ‘I bought a/the bike, afterwards, I went home.’

The data (84) shows that of the four structures in (83), the S-le sentence is the only one that cannot occur in a non-informative context, assuming that there is no intention to express an additional implicature. This confirms the analysis for the S-le sentence, that is, it expresses informativeness, presupposing that the proposition is new to the hearer. The other three constructions, namely, the V-le sentence, the Object Preposing sentence, and the ba-sentence, are neutral with respect to informativeness and thus can be used in a non-informative context.

3.5.4 Summary

In this section, I discussed four sentence structures, and showed that each of them has a distinct discourse use. S-le sentences are used in a general kind of informative context, but blocked by other sentence structures in some narrow types of informative contexts. The four constructions and their analyses are summarized as follows.

143

Structures Presuppositions

S-le The proposition is new to the hearer.

Object Preposing The hearer has prior knowledge of the proposition.

Ba-sentence The referent denoted by ba-NP is established in discourse.

V-le (neutral)

Table 1: The four sentence structures and their presuppositions: the S-le, the V-le, the Object Preposing, and the ba-sentence.

The generalization of the distribution of the four structures in an informative context is as follows: If there is a more specific sentence structure that better fits the discourse (Object Preposing, ba-sentence), that structure blocks the S-le sentence, else an S-le sentence is used. These discourse conditions do not restrict the use of ba-sentences (i.e. ba-sentences can be used in all of the four types of informative contexts) but ba-sentences are subject to certain lexical semantic (i.e. affectedness) and definiteness constraints. The V-le sentence is neutral and has almost the same distribution as the S-le sentence in an informative context. Finally, note that the object of the S-le sentence and the Object Preposing sentence is syntactically constrained. The noun phrase tends to be small: bare nouns are most preferred and modification by quantifiers is unacceptable.

3.6 CHAPTER SUMMARY

In this chapter, I argued that the S-le sentence, which I argued in Chapter 2 involves noun incorporation, expresses informativeness, which I characterize as a kind of presupposition in a Stalnakerian framework. It presupposes that the proposition is new to the hearer or the proposition is not accepted in the common ground. This analysis explains why S-le sentences are normally used in informative contexts, and give rise to pragmatic effects when the information is not new to the hearer. It also explains readings of “change of state” and “contrary to one’s expectation” in terms of an inference that arises from the contrast of background 144 knowledge and the new proposition. S-le sentences are restricted to more general and simple informative contexts but excluded from more specific kinds of informative contexts where some other factors are involved. The Object Preposing sentence is the most natural structure to use if the discussion is returning to an earlier topic and the speaker is providing new information about it. The ba-sentence is the most natural structure to use if the object is given and prominent in the discourse since the ba-NP is a definite NP. But the use of ba-sentence is subject to an additional constraint, that is, the verb needs to have the meaning of disposal or affectedness. I also showed that the V-le sentence is largely neutral in the sense that it does not have the kinds of presuppositions the other three structures have. Given the condition, it might be blocked by the Object Preposing sentence and the ba-sentence.

145

Chapter 4: Telicity in Mandarin Chinese

4.1 INTRODUCTION

Since Vendler (1957) it has been assumed that verbal predicates possess certain inherent temporal-aspectual properties which associate the events described by the predicates with an inherent, internal temporal structure. Vendler identified four lexical aspectual classes, namely statives, activities, accomplishments, and achievements. Later the class of semelfactive was added to the inventory (Smith 1997). The aspectual properties of the four basic verbal classes are usually characterized on the basis of three binary features, namely, [±dynamic] (or [±static]), [±durative], and [±telic] (Smith 1997: 20). The dynamic feature is used to distinguish statives from activities, accomplishments, and achievements while the three dynamic predicate classes are distinguished from each other by the features [±durative] and [±telic]. Durative predicates denote events that have duration as opposed to instantaneous or punctual events, which include activities and accomplishments. Telic predicates denote events that have a culmination endpoint, which include accomplishments and achievements. The three dynamic predicates are illustrated by English data as follows:

(1) a. activities – [+durative] and [-telic] (e.g. swim, walk around)

b. accomplishments – [+durative] and [+telic] (e.g. build a house) c. achievements – [-durative] and [+telic] (e.g. arrive at the city)

The aspectual classification of a verbal predicate indeed is influenced by the meaning of the head verb. As pointed out by Beavers (2012: 24), different verbs describe different types of events and certain event types already require or preclude duration and/or culmination for the denoted event. So we can easily infer for some verbs what real-world changes and processes make up an event of that type.

146

However, it has also been noticed that the reference type of certain event participants may also influence the aspectual composition and determine the aspectual class of the dynamic event predicate (e.g., Verkuyl 1993; Dowty 1977, 1979; Krifka 1989, 1992, 1998; Rothstein 2004, Beavers 2012, 2014). For example, for consumption/creation predicates (e.g., eat, drink, build), the aspectual class is determined by the expression of the patient. If the patient is expressed as a noun phrase which specifies the quantity of the patient involved in the event, the predicate is telic; if the patient is expressed by a mass noun or the plural form of a count noun, which has no such specification for the quantity of the patient, the predicate is atelic. I probe for telicity using in/for-adverbial modifications (Dowty 1979: 56-58). Normally only telic predicates are compatible with in-adverbials and atelic predicates are more natural with for-adverbials.1

(2) a. John drank a glass of wine in/?for an hour. (telic) b. John drank wine for/??in an hour. (atelic)

A telic predicate like drink a glass of wine describes a wine-drinking event in such a way that the patient a glass of wine “measures out” the drinking event such that the event ends when the glass of wine is entirely consumed and vice versa. So it entails culmination. By contrast, the object wine of an atelic predicate like drink wine does not set any boundary for the drinking event the predicate describes. The predicate only entails that a certain amount of wine is consumed but not necessarily all of some specific amount of wine.

(3) a. John drank a glass of wine (??but he didn’t finish it). (telic) b. John drank wine (but he didn’t finish the glass of wine). (atelic)

1 Note that the judgments for (2) are best considered defaults. Some speakers might accept for-adverbials for accomplishments as in (2a) if they are viewed as partitive somehow, and in-adverbials with activities as in (2b) might also be acceptable if the context clarifies that there is a set terminal point. 147

Such kind of internal argument, which enters the aspectual composition of the entire VP, is called an incremental theme (Dowty 1991: 567-571). In a series of influential works, Krifka (1989, 1992, 1998) developed a mereological model to account for the incremental theme effects. He modeled the event and the corresponding incremental theme as two algebraic part structures which are homomorphically related via a

θ-relation wherein the mereological properties of one structure are reflected in the other. With this model, Krifka analyzed incremental theme effects as a transfer of mereological properties from the structure of the patient to the structure of the event. Depending on how specific the nominal predicate is about the quantity of the incremental theme, the homomorphic θ-relation determines how specific it is about the boundedness of the event. However, cross-linguistically, it has been found that there are many languages in which such kind of accomplishments with an incremental theme do not entail culmination unless some other element is added. For example, in some languages, a culmination reading depends on morphological marking, which might be the perfective aspect marker as in Russian (Filip 2001, 2017, Borik 2002, Rothstein 2004) or the accusative case as in Finnish (Kratzer 2004). In those languages, if the verb is marked with the imperfective aspect or the object is marked by the partitive case respectively, culmination is not entailed. In Hindi (Singh 1991), a complex perfective aspect is needed in order to obtain a culmination reading for an accomplishment, whereas a normal or simple perfective aspect is not enough to yield a telic predicate that entails culmination. Filip (2017: 175-176) lists a number of languages in which even with the perfective aspect marker, the accomplishments do not entail culmination, including Chinese, Tamil, Korean, Thai, Japanese, Tagalog, among others.

In a number of studies on aspects in Mandarin Chinese (Chinese hereafter, Tai 1984, Smith 1990, 1997, Sybesma 1997, 1999, Soh 2009), it has been pointed out that the equivalent expression of English accomplishment does not entail culmination. 148

An example is presented as follows:

(4) Mǎlì chī-le yī-gè píngguǒ, hái shèng liǎng-kǒu, bèi wǒ chī-le.

Mary eat-LE one-CL apple still left two-bite PASS I eat-LE ‘Mary ate an apple. There was some left over and I ate it.’

In this discourse, the second part implies that the event of Mary’s eating an apple does not culminate, which suggests that the accomplishment chī yī-gè píngguǒ ‘eat one apple’ does not entail culmination. The authors of those studies have pointed out that in order to obtain a culmination reading, the accomplishment needs to be expressed in a complex construction called Resultative Verb Compound (RVC). The RVC construction consists of two elements (Li & Thompson 1981: 54-70). The first is a verb that describes the activity or indicates the event type (e.g., chī ‘eat’, xiě ‘write’, dǎ ‘hit’), and the second describes the result of the action, which can be an adjective

(e.g., gānjìng ‘clean’, gān ‘dry’, hóng ‘red’), an inchoative verb (e.g., pò ‘break’, kāi ‘open’, shàng ‘ascend, up’, dào ‘arrive’), or a morpheme that indicates the second phase of an event (e.g., -wán ‘finish, terminate’, -diào ‘off’, -zháo ‘be on target’). The two elements form a verbal compound that denotes a bounded event with the result specified by the resultative morpheme. Thus accomplishments like the one in (4) may not entail culmination, but the culmination reading is guaranteed if they are expressed in the RVC construction. For example, prior authors have noted that an accomplishment like that in (4) can be compounded with the morpheme -wán ‘finish, terminate’, resulting in a complex predicate that entails culmination. The corresponding sentence is presented in (5). The resultative morpheme -wán is glossed as TERM(INATION).

(5) Mǎlì chī-wán-le yī-ge píngguǒ.

Mary eat-TERM-LE one-CL apple ‘Mary finished eating an apple.’

149

With this complex accomplishment predicate, a continuation that implies non- culmination for the denoted event would lead to contradiction:

(6) *Mǎlì chī-wán-le yī-ge píngguǒ, hái shèng liǎng-kǒu, bèi wǒ chī-le.

Mary eat-TERM-LE one-CL apple still left two-bite PASS I eat-LE ‘*Mary finished eating an apple. There was some left over and I ate it.’

The culmination reading of the accomplishment brought out by -wán seems to indicate that the semantic contribution of -wán to the ultimate aspectual meaning is culmination or completion, an analysis assumed in Soh (2009). I argue here that this is not what -wán means. I show that -wán is not a morpheme that expresses culmination or completion per se but a morpheme that expresses termination. To be specific, it introduces information about the progress of the denoted event without setting any direct requirement on the involved patient. It ultimately derives culmination when it is combined with an accomplishment but it does not express culmination by itself. Note that other than -wán, there are several other RVC morphemes commonly used to telicize accomplishment predicates. They express a different meaning from -wán but also give rise to a culmination interpretation. For example, the morpheme -diào, which literally means ‘drop’ or ‘off’ but is grammaticalized as a suffix that designates the second phase of an event when used in the RVC construction, is a common RVC morpheme used for consumption predicates. It requires that the patient that undergoes the change of state must be affected entirely and entails that the event achieves culmination with respect to the entire patient. An example is presented in (7). The morpheme -diào expresses culmination (or completion) and is glossed as

CULM(INATION).

(7) Mǎlì chī-diào-le yī-ge píngguǒ.

Mary eat-CULM-LE one-CL apple ‘Mary ate up an apple.’

150

This sentence is semantically similar to (5) as it also entails that the eating event culminates with respect to the apple. Thus a continuation that implies the eating event was not culminated leads to contradiction:

(8) *Mǎlì chī-diào-le yī-ge píngguǒ, hái shèng liǎng-kǒu, bèi wǒ chī-le.

Mary eat-CULM-LE one-CL apple still left two-bite PASS I eat-LE ‘*Mary ate up an apple. There was some left over and I ate it.’

The morpheme -diào predicates on the patient directly. Different from -wán, which contributes to a culmination reading in an indirect way, -diào expresses culmination directly. Although -wán and -diào express different meanings, they both telicize the predicates compounded with them, yielding a complex predicate that denote bounded events. What is interesting about them is that they show us that there are two ways to describe a bounded event or, to put it another way, there are two different avenues to telicity. The two morphemes suggest that telicity may not just be obtained through patient, which has been the focus of many prior studies (e.g., Krifka 1989, 1992, Zucchi & White 2001, Filip & Rothstein 2006), but it can also be obtained through time. The morpheme -diào represents the canonical case of obtaining telicity through the patient, in which case an incremental theme “measures out” the event. However,

-wán shows that telicity can also be obtained by considering a specific amount of times during which the event progresses incrementally towards the end point. In this case, it is what I call an incremental time that “measures out” the event. With the temporal boundary specified by the predicate, the predicate is telic as it cannot apply to a proper part of the event. For example, an event of eating apples from 3:01 to 3:05 does not have a subevent that can also be called an event of eating apples from 3:01 to 3:05. What is interesting about -wán is that it is a bound morpheme that generically bounds the time, yielding an eventive predicate that does not apply to a proper part of the events in its denotation.

151

Note that what I aim to do in this chapter is not to study termination and culmination as formal notions. Instead, my purpose here is to show a case of deriving telicity from time, which has not received much attention in prior studies, and propose an analysis for analyzing telicity derived from time. The analysis proposed here is couched in the mereological framework developed by Krifka (1989, 1992, 1998), and following Krifka as well as Zucchi & White (2001), Filip & Rothstein (2006), among others, I adopt a quantization analysis for telicity derived not just from patient but also from time. In Section 4.2, I informally characterize termination and culmination, and provide data to illustrate them. In Section 4.3, I use the in-/for-adverbials as a diagnostic test for telicity to show atelicity/telicity for a set of relevant Chinese data. It shows that RVC predicates are telic. Non-RVC predicates are similar to their English counterparts since a bare noun yields an activity and a numeral classifier construction yields a non-culminating accomplishment. In Section 4.4, I provide an overview of some previous studies of telicity including Krifka (1989, 1992, 1998) and Zucchi & White (2001). Zucchi & White proposed an analysis of maximalization of participants to fix a problem in Krifka’s analysis. I adopt their idea that individual arguments can be maximalized. Note that Filip & Rothstein (2006), Filip (2008, 2017), and Kardos (2016) utilized maximalization of events in their accounts for telicity. In the current study, I extend the notion to times. In Section 4.5, I provide a modal analysis for non-RVC accomplishments, which show variable telicity, by assuming that it is optionally associated with a non-culmination operator which is overridden if it is used in the RVC construction. In Section 4.6, I propose an analysis for -wán and -diào rooted in Zucchi & White. I propose that -wán presupposes an event time and denotes a function that picks out the maximal run time of the event (i.e. the event time) at the reference time, and -diào presupposes a certain amount of patient and denotes a function that picks out the maximal patient of the event at the reference time. The account for the data makes reference to the structures of event, patient, and time, which are homomorphically related. In Section 4.7, I provide a summary and 152 conclusion of the current study.

4.2 TERMINATION AND CULMINATION

In this section, I provide a characterization of termination and culmination and discuss how they are attested. A termination predicate ensures termination of an event, entailing that the event stops progressing at some particular time. It presupposes that the event the verbal predicate denotes has duration, otherwise there would be no event progress to terminate. A culmination predicate ensures completion of an event with respect to the patient, entailing that the patient undergoes a change of state completely. It presupposes a certain quantity of patient, whose complete affectedness defines the delimitation of culmination of an event. The two types of predicates are illustrated below with the data involving -wán and -diào. First, if a predicate P entails termination of an event, then when put in a temporal sequence with another event, the first must fully precede the latter. Therefore, if -wán expresses termination, then when combining it with an eventive predicate the denoted event must fully precede another event that occurs in temporal sequence. Consider the discourse in (9). It consists of two sequential events, one that Mary ate a certain amount of apples and the other that she wanted to eat a banana.

(9) Mǎlì chī-wán-le píngguǒ, hái bú gòu, hái xiǎng chī xiāngjiāo.

Mary eat-TERM-LE apple still not enough still want eat banana

‘Mary finished eating apples. It was not enough. She still wanted to eat a banana.’

This discourse gives rise to the inference that the event of Mary eating apples has terminated before she wants to eat a banana. The first event fully precedes the second. By contrast, without -wán, the discourse does not require full precedence. The discourse is presented in (10). A possible context for (10) is as follows: at a party, Mary was talking to John and she was eating an apple, and then she saw bananas on the dining table. She then told John that she wanted to eat a banana while she was still

153 eating the apple. In this context, the event of eating an apple does not fully precede the event of wanting to eat a banana.2 Note that (10) entails that the event of eating an apple at least begins earlier than the event of wanting to eat a banana, though it does not entail full precedence.

(10) Mǎlì chī-le píngguǒ, hái bú gòu, hái xiǎng chī xiāngjiāo.

Mary eat-LE apple still not enough still want eat banana

‘Mary ate apples. It was not enough. She still wanted to eat a banana.’

The morpheme -diào shows similar pattern with -wán in this respect, also giving rise to an inference of full precedence. Consider the discourse (11).

(11) Mǎlì chī-diào-le píngguǒ, hái bú gòu, hái xiǎng chī xiāngjiāo.

Mary eat-CULM-LE apple still not enough still want eat banana

‘Mary ate up the apple. It was not enough. She still wanted to eat a banana.’

This discourse conveys that the event of eating up the apple fully precedes the event of wanting to eat a banana. Second, if a predicate P entails culmination, then there is a non-cancelable inference that the relevant result state has obtained. Therefore, if -diào expresses culmination, then when it is combined with an eventive predicate, the denoted event must entail that the result state with respect to the patient is obtained. This can be verified by the following discourse.

(12) *Mǎlì chī-diào-le (yī-gè) píngguǒ, hái shèng liǎng-kǒu, bèi wǒ chī-le.

Mary eat-CULM-LE one-CL apple still left two-bite PASS I eat-LE ‘*Mary ate up an/the apple. There was some left over and I ate it.’

In this example, the bare noun object píngguǒ ‘apple’ has a definite reading, which is

2 I was not sure whether the judgment is the same for other native speakers, so I asked six native speakers and they all agreed with my judgment. 154 an effect of the meaning of -diào, and it entails that the apple must be consumed entirely. Culmination is also entailed if the object is yī-gè píngguǒ ‘one apple’. The discourse is unacceptable because it is contradictory. By contrast, without -diào, the predicate does not entail culmination. The discourse is presented as follows.

(13) Mǎlì chī-le (yī-gè) píngguǒ, hái shèng liǎng-kǒu, bèi wǒ chī-le.

Mary eat-LE one-CL apple still left two-bite PASS I eat-LE

‘Mary ate (an) apple. There was some left over and I ate it.’

The morpheme -wán may imply culmination as well. Consider (14).

(14) *Mǎlì chī-wán-le (yī-gè) píngguǒ, hái shèng liǎng-kǒu, bèi wǒ chī-le.

Mary eat-TERM-LE one-CL apple still left two-bite PASS I eat-LE ‘*Mary finished eating an/the apple. There was some left over and I ate it.’

This discourse is unacceptable regardless of whether the object is a classifier construction or a bare noun. When the object is yī-gè píngguǒ ‘one apple’, the interpretation is that the apple was consumed entirely, so the discourse is contradictory. When the object is a bare noun píngguǒ ‘apple’, which refers to an unspecified amount of apples, the interpretation is that that amount of apples was consumed entirely. The noun phrase liǎng-kǒu (two-bite) ‘part of some food’ in the second sentence is anaphoric, referring to part of the amount of apples denoted by the bare noun píngguǒ ‘apple’ in the first sentence. So the discourse is contradictory as the first part implies culmination with respect to an unspecified amount of apples while the second part contradicts the culmination inference. The data presented above suggest that -wán and -diào are quite similar since they both have termination and culmination readings, and in fact, in some cases, their differences are so subtle they can be used interchangeably. But there are also some cases in which it is acceptable to use -wán but not -diào. In those cases, the event has terminated but it does not culminate with respect to the amount of patient that is

155 considered prominent in the discourse. For example, consider this context: Mary and John were eating some apples and there was no plan for how many apples each of them should eat. Mary then went to bed after eating part of the apples. In this case, the discourse (15), which uses -wán, is acceptable.

(15) Mǎlì chī-wán-le píngguǒ jiù qù shuìjiào le.

Mary eat-TERM-LE apple then go sleep LE

‘Mary finished eating apples and then went to bed.’

But it is unacceptable to use -diào.3

(16) #Mǎlì chī-diào-le píngguǒ jiù qù shuìjiào le.

Mary eat-CULM-LE apple then go sleep LE ‘Mary ate up the apples and then went to bed.’

The interpretation for the bare noun píngguǒ ‘apple’ in (16) is that it refers to all apples involved in the context. It is unacceptable since Mary did not eat all the apples involved in the context. Third, since most events terminate at a particular time point regardless of whether the event is denoted by a predicate that has an inherent endpoint or not, we should expect that -wán can be combined with eventive predicates that have no inherent endpoints as well. The prediction is borne out. First consider the case of activity verbs. Two examples are provided in (17). The morpheme -wán is combined with the activity verb xiào ‘laugh’ in (17a) and yóu-yǒng ‘swim’ in (17b). For both examples, the RVC expression designates a terminated activity.

(17) a. tā xiào-wán-le jiù zǒu-le.

he laugh-TERM-LE then leave-LE

‘He laughed and then left.’

3 I use “#” to indicate that the sentence/discourse in itself is grammatical but unacceptable for the given context. 156

b. tā yóu-wán-le yǒng jiù zǒu-le.

he swimV-TERM-LE swimN then leave-LE ‘He left when he finished swimming.’

Fourth, since termination presupposes duration for the event, we may expect that -wán cannot be compounded with achievements since instantaneous events denoted by achievements have no duration. This is also borne out. Consider the following example which involves the achievement verb dào ‘arrive’. The sentence is unacceptable if the verb dào ‘arrive’ is compounded with -wán.

(18) tā dào(*-wán)-le jiā jiù shuì-le.

he arrive-TERM-LE home then sleep-LE ‘When he arrived home, he went to bed.’

Note that -wán is also compatible with semelfactives (Smith 1997: 29-30). The event represented by a semelfactive verb has the temporal property of being dynamic and atelic, and it can be used to refer to a single-event activity and a multiple-event activity. When it is compounded with -wán, it allows an interpretation of a multiple-event activity as well as an interpretation of a single-event activity. Consider two examples in (19). The semelfactive verb in (19a) is qiāo ‘knock’ and the one in (19b) is ké ‘cough’. Both allow interpretations describing a multiple-event or single-event activity.

(19) a. tā qiāo-wán-le mén jiù pǎo-kāi-le.

he knock-TERM-LE door then run-away-LE ‘After knocking (once) on the door, he ran away.’ b. tā ké-wán-le zhīhòu hē-le diǎn shuǐ.

he cough-TERM-LE after drink-LE some water ‘After coughing (once), he drank some water.’

157

When the aspectual expression yíxià ‘once’ is overtly used, the semelfactives can still be compounded with -wán, as in the following.4

(20) a. tā qiāo-wán-le yíxià yòu qiāo-le yíxià.

he knock-TERM-LE once again knock-LE once ‘After he knocked once, he knocked once again ’ b. tā ké-wán-le yíxià yòu ké-le yíxià.

he cough-TERM-LE once again cough-LE once ‘After he coughed once, he coughed again.’

Although semelfactives can be used to describe temporally instantaneous events, they are activities in nature and thus are different from the kind of instantaneous events described by achievements. Fifth, it is also worth mentioning compounding of -wán with stative predicates.

This kind of compounding is quite restricted but some such compounds are acceptable under certain conditions. Such RVC expressions do convey literal meanings, that is, a certain state stops holding at a particular time. But usually the purpose of using such expressions is to convey non-literal meanings rather than the literal meanings. Two examples are provided in (21). The stative verb in (21a) is xǐhuān ‘like’ and the one in (21b) is zhōngchéng ‘be loyal to’.5

4 Similar examples are attested on Google:

(1) ān lǎoshi qiāo-wán-le yíxià bù jiěqì, yòu yízhí jìxù liánzhe qiāo. An teacher knock-TERM-LE once not enough again throughout continue continuously knock ‘Teacher An knocked once, which was not enough. He again continued to knock continuously throughout.’ (2) wǒ xiàyìshi xiǎng bǎ shuǐ ké-chūlái, -ké wán yíxià, jiù gǎndào fèi-bù zuānxīn I subconsciously want BA water cough-out cough-TERM once then feel lung-part severely téng-le qǐlái. pain-LE arise ‘Subconsciously I wanted to cough out the water. After coughing once, I felt a severe pain arising in the lung.’ 5 A similar example is attested on Google: dàjiā yǒu-méi-yǒu jīngcháng xǐhuān-wán zhè-ge nánshēng, yòu xǐhuān -nèi ge nánshēng de everyone have-NEG-have often like-TERM this-CL boy.student then like that-CL boy.student DE jīnglì ma? experience Q “Does everyone often have the experience that you start to like another boy when you stop liking some boy?” 158

(21) a. Yuēhàn xǐhuān-wán Mǎlì yòu xǐhuān Sūshān le.

John like-TERM Mary then like Susan LE ‘John started to like Susan when he stopped liking Mary.’ b. Yuēhàn zhōngchéng-wán zhè-ge lǎobǎn yòu zhōngchéng nèi-ge lǎobǎn.

John be.loyal.to-TERM this-CL boss then be.loyal.to that-CL boss ‘John became loyal to that boss when he stopped being loyal to this boss.’

In both examples, the speaker is being sarcastic. In (21a), it literally says the state that John likes Mary stopped holding at a particular time when he started to like Susan, but what the speaker is trying to convey is that John is not a person who is dedicated in a relationship. Similarly in (21b), it literally says the state that John is loyal to this boss stopped holding at a particular time when he became loyal to that boss, and what the speaker is trying to convey is that John is a person who has no loyalty to his boss. To account for the unusual case of statives, I assume that -wán does not just require durative but also dynamic property for the event predicate it is compounded with, so it is normally incompatible with statives but the compounding can be licensed by pragmatic conditions. Finally, note that the morpheme -diào is semantically more restricted than -wán in forming RVC constructions. While -wán can be compounded with almost any eventive verb as long as the event in the denotation has duration, -diào is usually restricted to the type of eventive predicates which denote the kind of events for which the natural outcome involves the disappearance of the patient. For example, other than consumption verbs such as chī ‘eat’ and hē ‘drink’, the verbs that -diào is commonly combined with include transitive verbs such as shāo ‘burn’, cā ‘wipe’, rēng ‘throw’, kǎn ‘cut’, chōng ‘flush (with water)’, mài ‘sell’, shā ‘kill’, and pòhuài ‘destroy, damage’, as well as intransitive verbs such as pò ‘break’, zǒu ‘walk, leave’, and pǎo

‘run’. The range of eventive predicates it may combine with also varies across Chinese dialects (Chao 1968: 466). For example, in Jianghuai Mandarin dialects and

159

Wu dialects, it is acceptable to compound it with kàn ‘read’, resulting in a RVC which means to finish reading (something); however, this RVC is less common in use in standard Mandarin and might be unacceptable for some speakers. Note that diào can also be used as a contentful intransitive verb which literally means ‘drop’. So when it is used as an RVC morpheme, it seems that the requirement that the patient disappears as the result of the event is a metaphoric use of the literal meaning ‘drop’, for example, the food is conceptualized as having been ‘dropped’ or gone at the end of a consumption event. Note that as shown by these examples, -diào is not restricted to the kind of events where a physical change of state occurs, but it can also be used in an abstract sense of change of state. In the current study, I limit my concern to the cases where incremental themes are involved, in which the patient is affected entirely such as a consumption event. To sum up, in this section, I characterized termination and culmination predicates informally and show with Chinese data how they might be attested. However, the purpose of the current study is not to analyze termination and culmination as formal notions. Instead the purpose is to show that there are two different avenues to telicity, which are revealed to us by the meanings of -wán and -diào, and propose an analysis accordingly. So next we return to telicity.

4.3 A DIAGNOSTIC TEST FOR TELICITY

In this section, I use the diagnostic test of in/for-adverbial modification to show that RVC predicates, either with -wán or -diào, are telic. As mentioned above, telic predicates are usually compatible with in-adverbials and but sound less natural when modified by for-adverbials. By contrast, atelic predicates are compatible with for-adverbials and sound less natural when modified by in-adverbials. I show that such modification patterns also apply to RVC predicates, confirming that they are telic predicates. Moreover, I show that although non-RVC accomplishments do not entail culmination, they show similar modification patterns with English accomplishments.

160

So non-RVC accomplishments are ambiguous with respect to telicity and we obtain different results depending on the criteria for telicity. We first consider non-RVC predicates modified by a for-adverbial. The equivalent of for-adverbial in Chinese is a noun phrase, which is syntactically inserted between the verb and the object. The data are presented in (22).6 The durative phrase

6 Note that for-adverbial modification in Chinese is quite a complicated issue. First, the durative phrase is not a PP but a noun phrase. It is usually considered an object rather than an adjunct (see Wechsler & Lee 1996). Second, there is a syntactic constraint: the patient object following the durative phrase must be a bare noun. If the patient object is not a bare noun such as a noun phrase with demonstrative or numeral, the relative order of the object and the durative phrase is reversed – so the relative order of the object and the durative phrase is different in (22a) and (22b). Consider the following example. The object nèi-ge guójiā ‘that country’ is modified by a demonstrative. The durative phrase is placed at the end of the sentence rather than between the verb and the object.

(1) a. tā tǒngzhì-le nèi-ge guójiā bàn-ge shìjì. he rule-LE that-CL country half-CL century b. ??/*tā tǒngzhì-le bàn-ge shìjì nèi-ge guójiā. he rule-LE half-CL century that-CL country ‘He ruled that country for half a century.’

When this condition is met, an accomplishment predicate can also be used but its interpretation is an activity. Consider the following example. The sentence can be paraphrased as that he was involved in an activity of closing a/the door or closing doors for half an hour.

(2) tā guān-le bàn-ge xiǎoshí mén. he close-LE half-CL hour door ‘He closed doors for half an hour.’

Also note that the object in the order Verb-Obj-Dur may have a reading of affectedness, depending on the verb. The verb tǒngzhì ‘rule’ in (1) does not give rise to a reading of affectedness, but the verbs chī ‘eat’ and tuī ‘push’ do. Consider the following examples. Note that accomplishments can be used in this structure as in (3).

(3) a. tā chī-le nèi-ge píngguǒ wǔ fēnzhōng. he eat-LE that-CL apple five minute ‘He ate that apple for five minutes.’ b. tā tuī-le nèi-liàng chē wǔ fēnzhōng. he push-LE that-CL cart five minute ‘He pushed that cart for five minutes.’

For the two examples, the interpretation is that the subject intentionally carried out the activity of eating that apple or pushing that cart for five minutes, and the patient has an interpretation of being affected. Note that the two sentences might have significantly reduced acceptability for some speakers due to the affectedness reading. When the determiner is a numeral classifier construction as in (22b) as opposed to a demonstrative as in (3a) of the footnote, the sentence is simply unacceptable. Third, modification by a durative phrase can also be expressed by the verb-copying construction (see Li & Thompson 1981: 442) and the Object Preposing construction (see Paul 2002, 2005). An example is presented as follows.

(4) tā chī-píngguǒ chī-le wǔ fēnzhōng. (Verb-copying Construction) he eat-apple eat-LE five minute ‘He ate apples for five minutes.’ (5) tā nèi-ge píngguǒ chī-le wǔ fēnzhōng. (Object Preposing Construction) he that-CL apple eat-LE five minute ‘Mary ate that apple for five minutes.’

Since the purpose here is to use for-adverbial modification as a diagnostic test for telicity as opposed to a purpose of studying for-adverbial expressions in Chinese, I assume that the particular structure I use 161 wǔ fēnzhōng ‘five minutes’ is compatible with activity predicate chī píngguǒ ‘eat apples’ as in (22a) and incompatible with accomplishment predicate chī yī-ge píngguǒ ‘eat an apple’ as in (22b). See Footnote 6 for an explanation of the different relative orders of the object and the durative phrase in (22a) and (22b) as well as for (24) and (25) below.

(22) a. Mǎlì chī-le wǔ fēnzhōng píngguǒ.

Mary eat-LE five minute apple ‘Mary ate apples for five minutes.’ b. *Mǎlì chī-le yī-ge píngguǒ wǔ fēnzhōng.

Mary eat-LE one-CL apple five minute ‘?Mary ate an apple for five minutes.’

The activity predicate and the accomplishment predicate also show a contrast when modified by an in-adverbial. Consider the example in (23). When modified by wǔ fēnzhōng nèi ‘within five minutes’, activity predicate chī píngguǒ ‘eat apples’ does not yield a totally unacceptable sentence, but gives rise to two coerced readings. One is that the bare noun píngguǒ ‘apple’ refers to a specific amount of apples and thus it is largely like an accomplishment predicate. The other reading is that the sentence presupposes a specific time point and it says an activity of eating apples occurred within the five minutes from the presupposed time point. It does not entail that the duration of the activity of eating apples lasts for five minutes. Instead, it implies that Mary might also have done something else within the five minutes. So the sentence can roughly be paraphrased as “Within five minutes from a specific time point, what Mary did included an activity of eating apples.” This reading might also be available for some native speakers when the durative phrase modifies the accomplishment predicate chī yī-ge píngguǒ ‘eat an apple’ as in (23b) in certain contexts. But the normal interpretation for the sentence is similar to the English accomplishment

for the diagnostic test is the right structure for testing telicity in Chinese. 162 sentence, as indicated by the translation.

(23) a. Mǎlì wǔ fēnzhōng nèi chī-le píngguǒ.

Mary five minute within eat-LE apple (i) ‘Mary ate (a certain amount of) apples in five minutes.’ (ii) ‘In five minutes, Mary ate apples.’ b. Mǎlì wǔ fēnzhōng nèi chī-le yī-ge píngguǒ.

Mary five minute within eat-LE one-CL apple ‘Mary ate an apple in five minutes.’

So the in-/for-adverbials test shows that the non-RVC activity is atelic and the non-RVC accomplishment is telic, which is similar to the result of the English data. But as mentioned above, non-RVC accomplishments do not entail culmination. So according to some criteria they pattern like telic predicates and according to others do not. Now consider RVC predicates modified by for-adverbial. For both -wán and -diào, the sentences are unacceptable regardless of whether the object is a bare noun as in (24a) and (25a) or a classifier construction as in (24b) and (25b).

(24) a. *Mǎlì chī-wán-le wǔ fēnzhōng píngguǒ.

Mary eat-TERM-LE five minute apple

b. *Mǎlì chī-wán-le yī-ge píngguǒ wǔ fēnzhōng.

Mary eat-TERM-LE one-CL apple five minute (25) a. *Mǎlì chī-diào-le wǔ fēnzhōng píngguǒ.

Mary eat-CULM-LE five minute apple b. *Mǎlì chī-diào-le yī-ge píngguǒ wǔ fēnzhōng.

Mary eat-CULM-LE one-CL apple five minute

This suggests that RVC predicates, either compounded with an activity or an accomplishment, are telic.

163

RVC predicates are always compatible with an in-adverbial. In the case of -wán, a bare noun as in (26a) is interpreted as a definite noun phrase or as having a specific reading, so the sentence means that Mary finished eating the/some apple in five minutes. In other words, in this case, though the object is a bare noun, the predicate is an accomplishment (or telic predicate) rather than an activity (or atelic predicate). It entails that there is a specific amount of apples that is entirely consumed, though the quantity is not specified. An example of an accomplishment compounded with -wán is presented in (26b). The RVC predicate is unsurprisingly compatible with the in-adverbial.

(26) a. Mǎlì wǔ fēnzhōng nèi chī-wán-le píngguǒ.

Mary five minute within eat-TERM-LE apple ‘Mary finished eating the/some apple in five minutes.’

b. Mǎlì wǔ fēnzhōng nèi chī-wán-le yī-ge píngguǒ.

Mary five minute within eat-TERM-LE one-CL apple ‘Mary finished eating an apple in five minutes.’

In the case of -diào, the bare noun píngguǒ ‘apple’ as in (27a) has a definite reading and it refers to a specific amount of apples. This can be explained by the culmination meaning of -diào, which necessarily presupposes that there is a specific amount of food involved in the consumption event. When the object is a classifier construction as in (27b), the interpretation remains indefinite.

(27) a. Mǎlì wǔ fēnzhōng nèi chī-diào-le píngguǒ.

Mary five minute within eat-CULM-LE apple ‘Mary ate up the apple in five minutes.’ b. Mǎlì wǔ fēnzhōng nèi chī-diào-le yī-ge píngguǒ.

Mary five minute within eat-CULM-LE one-CL apple ‘Mary ate up an apple in five minutes.’

164

So the data involving in-adverbial modification also suggest that RVC predicates are telic. To sum up, the bulk of the data presented above suggest that RVC predicates, either with a bare noun or a numeral classifier construction, are telic predicates. They reject for-adverbials entirely and are always compatible with in-adverbials. Non-RVC predicates, either with a bare noun or a numeral classifier construction, are quite similar to the corresponding English predicates in terms of their modification patterns. According to the result of the in/for-adverbial modification diagnostic test, a bare noun yields an atelic predicate and a numeral yields a telic predicate.

4.4 PREVIOUS STUDIES OF TELICITY

The analysis proposed in the current study for -wán and -diào is couched in the event semantic framework proposed by Krifka to account for aspectual composition (Krifka 1989, 1992, 1998). Krifka’s characterization of telic predicates like eat an apple hinges on the property of quantization: a quantized predicate is a predicate that cannot apply to a proper part of an entity in its denotation. This property provides a criteria for re-categorizing mass and count nouns in the nominal domain and atelic and telic predicates in the verbal domain into a category of quantized predicates (count nouns and telic predicates) and a category of non-quantized (or cumulative) predicates (mass nouns and atelic predicates). The correlation between a quantized nominal predicate (e.g. an apple) and a quantized verbal predicate (e.g. eat an apple) is formally explained in terms of a transfer of mereological property from the structure of patient to the homomorphically related structure of event. However, as pointed out by Zucchi & White (2001), Krifka’s account for telicity in terms of quantization runs into problems when the object is an indefinite noun phrase like a sequence, a twig, a quantity of N, and some Ns, which are not quantized predicates, according to the definition. In this case, an event of writing a sequence, for example, may have a proper part which is still an event of writing a

165 sequence, and thus according to Krifka’s analysis, the predicate write a sequence should be a non-quantized, atelic predicate, yet it is telic. Zucchi and White provided a solution to this problem while still maintaining a quantization analysis in this case. They proposed an analysis of maximalization of participants, which restricts the denotation to all the relevant participants involved in the event at reference time, which also predicts quantization. Filip and Rothstein (2006), Filip (2008, 2017), and

Kardos (2016) utilized maximalization of events in their accounts for telicity. In the current study, I extend their analysis to times for an analysis of -wán. In this section, I review the framework proposed by Krifka (1989, 1992, 1998) and the extension provided by Zucchi & White (2001), which provide the basis for the current proposed analysis. For other analyses of telicity, see Kennedy & McNally (2005), Kennedy & Levin (2008), Filip (2008), Filip & Rothstein (2006), Beavers (2012, 2014), Kratzer (2004), and Smith (1991).

4.4.1 Krifka’s model on aspectual composition

Krifka (1989, 1992, 1998) proposed a model to account for the influence of reference types of nominal arguments on temporal properties of verbal predicates. He defined two types of predicates, cumulative and quantized, which connects two types of nominal predicates to the contrast between telic and atelic predicates in the verbal domain. A cumulative predicate can be a mass noun (e.g. beer) or a bare plural noun (e.g. books), that is, if the predicate can apply to both entities, it can apply to their sum.

A quantized predicate can be a count noun (e.g. a book) or a noun phrase that specifies the quantity of the entity (e.g. a glass of beer), that is, if the predicate applies to x, it cannot apply to a subpart of x. The cumulative and quantized predicates are defined with respect to a part structure. The definitions are provided in (28) and (29) respectively (Krifka 1998: 200).

166

(28) cumulative predicates: ∀X[CUM(X) ↔ ∃x,y[X(x) ⋀ X(y) ⋀  x = y] ⋀ ∀x,y[X(x) ⋀ X(y) → X(x⊔y)]]

“A predicate X is cumulative iff X applies to at least two entities, for any x and y to which X applies to, X also applies to the sum of x and y.” (29) quantized predicates:

∀X[QUA(X) ↔∀x,y[X(x) ⋀ X(y) →  y⊂x]] “A predicate X is quantized iff for any x and y to which X applies, y is not a proper part of x.”

The referential property of the object is correlated with the aspectual property of the eventive predicate. When a verb like eat is combined with a cumulative noun phrase like apples, it yields an atelic predicate which has no inherent endpoint, such as (30a). When the verb is combined with a quantized noun phrase like an apple, it yields a telic predicate which has an inherent endpoint, such as (30b).

(30) a. John ate apples (for/??in ten minutes). (atelic) b. John ate an apple (?for/in ten minutes). (telic)

The internal arguments of verbs like drink and eat are characterized as bearing a special thematic role called incremental theme. The extent of the incremental theme is related to the progress of the event via a homomorphic θ-relation which ties each part of the incremental theme with a corresponding part of the event. The homomorphism requires the thematic role of incremental theme to have a number of properties including uniqueness of objects, uniqueness of events, mapping to subevents and mapping to subobjects. These properties are collapsed into two properties mapping to unique subobjects and mapping to unique subevents in Beavers (2012: 28) and the definitions are repeated in (31). Note that these properties define a strict homomorphism θ that Krifka called Strictly Incremental Relation (SINC), which is an isomorphism of subparts that defines being an incremental theme.

167

(31) Strictly Incremental Relation (SINC): Event e is θ-related to patient x such that every unique part of e corresponds to a unique part of x and vice versa, i.e. θ has the MUSO and MUSE properties:

a. Mapping-to-unique-subobjects (MUSO):

∀x,e,e’[θ(x,e) ⋀ e’ ⊂e →∃!y[y⊂x ⋀ θ(y,e’)]] “For all x θ-related to e, for all e’ ⊂ e there is a unique θ-related x’ ⊂ x.” b. Mapping-to-unique-subevents (MUSE):

∀x,x’,e[θ(e,x) ⋀ x’⊂x →∃!e’[e’⊂e ⋀ θ(e’, x’)]] “For all e θ-related to x, for all x’ ⊂ x there is a unique θ-related e’ ⊂ e.”

The MUSO and MUSE properties ensure a strict one-to-one mapping between the proper parts of e and the proper parts of x. It predicts that the event predicate like (32) is quantized, or telic.7

(32) John drank a glass of wine in/?for an hour.

For an event e of drinking a glass of wine x, any proper subpart event e’ ⊂ e is an event of drinking a proper subpart of the glass of wine x’ ⊂ x, by the SINC that holds between e and x. However, a glass of wine has quantized reference, therefore no x’ satisfies this description. This means that no e’ ⊂ e satisfies the description of drink a glass of wine since e’ is an event of drinking less than a glass of wine. So according to the definition of quantized predicate (29), drink a glass of wine is quantized (and also telic). SINC also predicts that the event predicate like drink wine is cumulative, or atelic.

(33) John drank wine for/??in an hour.

For an event e of drinking some amount of wine x, any proper subpart event e’ ⊂ e is

7 I assume Krifka’s (1998) analysis of for-adverbials wherein for-adverbials presuppose the predicate it applies to is atelic. Analogously, I assume that in-adverbials presuppose the predicate it applies to is telic. 168 an event of drinking a proper subpart of that amount of wine x’ ⊂ x by the SINC that holds between e and x. Since wine has cumulative reference, x’ satisfies this description. So e’ ⊂ e satisfies the description of drink wine. Given the definition of cumulative predicate (28), drink wine is cumulative, or atelic. The analysis predicts that event predicates like push a cart are not telic, though the nominal predicate is quantized. The explanation provided by this model is as follows: for a subevent of an event of push a cart, there is no proper part of the object a cart that is the patient of a subevent of the event push a cart. Instead, the patient of the subevent is the same as the patient of the event. In other words, the non-quantized property of the event predicate push a cart is due to that the verb push does not assign the thematic role of incremental theme to its internal argument. Therefore, the difference between eat x and push x boils down to the difference in the thematic role each verb assigns to x.

So far, the analysis can provide an adequate account for events where a subpart of the patient is isomorphically related to a unique subpart of the event, for example, eat an apple. However, it is inadequate to account for events such as read a book where some subparts of the patient can be affected iteratively. To deal with this issue, Krifka made a distinction between two types of homomorphism, namely the strict incremental relation (SINC) and the incremental relation (INC). The SINC is an isomorphism of subparts which consists of the two properties defined in (31) above. In this case, progress of an event is isomorphically related to the extent of the incremental theme, so each subpart of event corresponds to a unique subpart of the theme, and event ends when the theme is entirely affected. By contrast, in the case of INC, an event is incrementally related to the theme in such a way that some subparts of the theme can be affected more than once, so some subparts of an event may correspond to the same subparts of the theme of some other subparts of the event. In the former case, each subpart of the object is a theme of a unique subpart of the event; in the latter case, a subpart of the object can be mapped to more than one subpart of the event. 169

In Krifka (1998), he further extended the analysis to account for motion events such as walk from the university to the capitol. He treated paths of motion as incremental themes as well and defined a θ-relation called Strict Movement Relation (SMR) which is isomorphic in subparts, relating the event e to the path p in their subparts. SMR also preserves the spatial/temporal adjacency: two subparts of e in temporal adjacency corresponds to two subparts of p in spatial adjacency, and vice versa.8 Krifka’s model has received a number of criticisms as it is insufficient to account for a number of empirical linguistic phenomena, and subsequent analyses have been proposed to tackle these issues. Below I will review one of these studies, which will lead to an analysis for telicity in Chinese.

4.4.2 Zucchi & White’s (2001) maximal participants

Zucchi & White (2001) noted that Krifka’s account for telicity is problematic if the object is an indefinite noun phrase like a sequence, a twig and some Ns. According to Krifka’s definition of quantization, such nominal predicates are not quantized since, for example, a subpart of a sequence is still a sequence and a subpart of some Ns is still some Ns. However, these nominal predicates yield quantized verbal predicates. For example, write a sequence is compatible with in-adverbs but not for-adverbs.

(34) John wrote a sequence in/??for ten minutes.

Instead of giving up on Krifka’s idea that quantization of nominal predicates leads to telicity, they modified the definition of quantization, and proposed two

8 Krifka (1998: 207) also provided a weaker version of the definition of telic predicate:

∀X⊆UE[TELE(X) ↔ ∀e,e’∊UE[X(e) ⋀ X(e’) ⋀ e ⊆E e’ → INIE(e’,e) ⋀ FINE(e’,e)]]

It says a predicate is telic iff for any event it applies to it does not apply to any initial or final subevent of that event. According to this definition, quantized predicates are telic but telic predicates may not be quantized. This weaker notion of telicity can solve some problems of telic motion predicates. For example, a motion from point A to point B may involve irrelevant detours in the middle, but all the events in the denotation have the same initial and final points. 170 solutions to account for the data. The intuition that motivates their analysis is that an event of writing a particular object x, be it a sequence or a letter, does not have events of writing x as proper parts. So they included an assignment function in their proposed definition of a quantized predicate which keeps the assignment of individuals fixed. The definition is presented below, where the individuals include both events and ordinary individuals (Zucchi & White 2001: 236).

(35) QUA(P) iff for every model M, assignment g, and individual a, b,

if〚P〛M,g(a)=1 and〚P〛M,g(b)=1, then a is not a proper part of b.

According to this definition, nominal predicates like a sequence and a letter are quantized in the sense that the specific individual, a sequence or a letter, for a single assignment, has no proper part that the predicate can apply to. Note that Zucchi & White assume that mass nouns and plural nouns, which are cumulative predicates in Krifka’s model, denote kinds (Carlson 1977). Under the kind reference analysis, bare nouns introduce a variable that ranges over kinds, which is associated with individual instances via a Realization relation. For example, both an event of writing three letters and its subpart of an event of writing two letters can be denoted by predicate write letters; for both events, the nominal predicate letters have the same denotation, that is, the kind letters. So the predicate write letters is non-quantized because it can apply to a proper part of the event it denotes. The first solution they proposed to deal with predicates like write a sequence is based on the analysis of indefinites proposed by Kamp (1981) and Heim (1982). According to the Kamp-Heim analysis, noun phrases of the form an N and some Ns do not inherently have existential force, instead, they introduce free variables which are existentially bound by closure rules at the discourse level (e.g. after for or in operator applies). Until then the variable introduced by the direct object is treated as a constant by the variable assignment function. The key of Kamp-Heim analysis of indefinite noun phrases in the accounting for the quantized property of predicates like

171 write a sequence is the assumption that indefinite noun phrases do not introduce existential quantifier immediately, which otherwise would predict that write a sequence is a non-quantized predicate. For example, consider the translation of the sentence John wrote a sequence provided in (36), where the noun phrase a sequence introduces a variable z which is not bound existentially. According to the Kamp-Heim analysis, the assignment function cannot apply to z immediately so there is only one single assignment of z to the sequence at a later stage when z is existentially bound. Thus the event predicate write a sequence with the meaning in (36) is quantized, because given the definition of quantization in (35) an event of writing a particular sequence does not have as a proper part an event of writing the very same sequence. Without the assumption of the Kamp-Heim analysis, that is, if we assume that the indefinite noun phrase has existential quantification force and introduces an existential quantifier immediately, then it would allow multiple assignments of z to the sequence and its subparts, since a sequence has proper parts that are still sequences, and thus it would predict that write a sequence is a non-quantized predicate and thus should be compatible with for-adverbials. So this account largely hinges on the fundamental assumption of the Kamp-Heim analysis of indefinite noun phrases.

(36) 〚 John write a sequence 〛 = λe[write’(e) ⋀ AG(e, John’) ⋀ PAT(e, z) ⋀

sequence’(z)]

The other solution is driven by a problem arising from applying the Kamp-Heim analysis to quantifiers like most and less than half of the Ns. Note that a proper part of most fleas can still be most fleas, that is, it is non-quantized. However, derived event predicates like find most of the fleas is quantized as it can be modified by in-adverbs but not for-adverbs.

(37) John found most of the fleas in/??for an hour.

172

The problem that arises when applying the Kamp-Heim analysis to noun phrases like most Ns is that, according to the Kamp-Heim analysis, these quantifiers, unlike indefinites, are inherently quantificational and thus do not introduce free variables. Therefore, this analysis fails to predict that find most fleas is quantized. For example, suppose there are ten fleas and there is an event e of John finding nine of them during an interval i, and during the interval i’⊂i, there is an event e’ of John finding seven fleas. Presumably, e’ ⊂ e. Since the number of fleas John found is greater than one half of the number of fleas in both events, the predicate found most fleas can be applied to both events. Thus it is non-quantized. But this cannot be a right analysis, since found most fleas is telic. Zucchi & White’s solution is that instead of picking out just events, we can pick out individuals involved in an event, which they call maximal participants. The intuition behind this analysis of maximalization is that when you say John wrote a sequence the reference of the object is for all intents and purposes a kind of specific reference, i.e. John wrote a specific thing at reference time that is a list of ordered numbers. Since no proper subpart of a specific sequence involved in a particular writing event at reference time is the same sequence, then the predicate write a sequence is quantized because given the property of mapping to unique subobjects and mapping to unique subevents defined in (31), no proper subpart of an event of writing a given sequence at reference time is the same event of writing that sequence at reference time. In this analysis, the noun phrase is treated like a definite noun phrase which denotes the maximal entity involved in the event. The relation Max is defined as follows (Zucchi & White 2001: 254):

(38) ∀x[Max(P, x) ↔ [P(x) ⋀ ﹁∃y[P(y) ⋀ x⊂y]]]

“An individual is a maximal P iff it is P and it is not a proper part of another P.”

This analysis has certain advantages over the first solution, as it can apply to more cases. First consider the case of indefinite noun phrases of the form an N and

173 some Ns which, in the Kamp-Heim approach, is accounted for by assuming that the indefinite noun phrase has no inherent quantificational force. Consider an example of write a letter. The translation under maximalization provided by Zucchi & White (2001: 259) is given in (39).9 Note that the max operator is introduced by the indefinite quantifier.

(39) 〚write a letter〛= λyλe∃x[write(e) ⋀ AG(y, e) ⋀ PAT(x, e) ⋀Max(λz∃e’[write(e’)

⋀ AG(y, e’) ⋀ PAT(z, e’) ⋀ letter’(z) ⋀ τ(e’)⊆tr], x)]

“An event of writing a letter is a writing event whose patient is maximal

among the individuals in the denotation of letter written at the time tr.”

According to Zucchi & White, there might be more than one maximal element in the denotation of Max(λz∃e’[write(e’) ⋀ AG(y, e’) ⋀ PAT(z, e’) ⋀ letter’(z) ⋀ τ(e’)⊆tr], x) since any singular letter that was written during the interval tr is counted as a maximal element. But the denotation of the predicate write a letter is restricted to a writing event whose patient is a singular letter at tr. Given that the object role of write has the property of mapping to objects, a proper subevent e’ of the event of writing a given letter is not an event of writing that letter since a proper subpart of a given letter is not itself. This means that e’ is not in the denotation of write a letter. Therefore, according to the definition of quantization, write a letter is quantized.

Now consider the case when the noun phrase is a quantifier. For the semantics of DPs like most Ns, they assume that they denote a generalized quantifier that takes the verb as an argument and the verb’s patient argument is saturated by a variable that picks the maximal participants with respect to the reference time tr, which is conditioned by the cardinality being greater than one half of the total number. The representation of most letters they provided is repeated as follows (Zucchi & White

2001: 254):

9 The agent of the event is ignored in the original representation in Zucchi & White (2001). 174

(40) 〚 most letters 〛 = λPλe∃x[P(x)(e) ⋀ Max(λz∃e’[P(z)(e’) ⋀ letters’(z) ⋀

τ(e’)⊆tr],x) ⋀ |x| > 1/2∑(λz[letter’(z)])]

Thus the nominal predicate most letters introduces a set of letters which are the maximal participants in some event described by its first argument P. Translation of the event predicate write most letters they provided is repeated as follows (Zucchi & White 2001: 254):

(41) 〚 write most letters 〛 = λe∃x[write(e) ⋀ PAT(e, x) ⋀ Max(λz∃e’[write(e’) ⋀

PAT(e’, z) ⋀ letters’(z) ⋀ τ(e’)⊆tr],x) ⋀ |x| > 1/2∑(λz[letter’(z)])]

“An event of writing most letters is a writing event whose patient is the

maximal set of letters written by the agent at the reference time tr and the cardinality of this set is greater than one half the number of the letters.”

This analysis predicts that write most letters is quantized. Assuming that the object of write is an incremental theme which has the property of mapping to objects, the maximal sum of letters written by an agent during a given reference time has no proper part which is also the maximal sum of letters written during the reference time, and thus the event of writing such a maximal sum of letters during the reference time does not have a proper part that is an event of writing the maximal sum of letters during the reference time. Therefore, the predicate write most letters is telic.

Zucchi & White also accounted for activity predicates such as rule most countries in terms of maximal participants. The atelicity of such an activity predicate is accounted for as follows: the patient of an event of ruling most countries is the sum of all the countries ruled at reference time tr. But since rule does not have the property of mapping to objects, a subevent may have as its patient the same sum of countries that are ruled at reference time tr, therefore the predicate rule most countries is not quantized. However, Rothstein (2004:152-154) pointed out that this analysis may fail to predict atelic readings for activity predicates which have maximalization as part of the 175 meaning of the object. One of her illustrating sentences is provided as follows (Rothstein 2004: 153, (9a)).

(42) John has owned more than half the houses on this street for the last five years, and he is adding to his property every few months.

Zucchi & White’s analysis makes the incorrect prediction that this sentence is telic. In a possible context where John has been buying up houses on this street incrementally, the event of owning 30 out of 40 houses during the reference time has a subevent of owning 29 houses, both of which can be described as owning more than half the houses but have different maximal sum of participants. Therefore, according to Zucchi & White’s analysis, it is a telic predicate. Yet it is atelic. Thus Rothstein (2004) concluded that quantization of the theme argument does not determine telicity of accomplishment predicates. However, Beavers (2012) noted that this conclusion is too strong. He pointed out that what seems to be relevant for quantized theme arguments calculating telicity is whether the verb introduces incrementality, and own does not.

4.5 NON-CULMINATING ACCOMPLISHMENT PREDICATES

The non-RVC accomplishments like chī yī-ge píngguǒ ‘eat an apple’, as shown above, sometimes pattern like telic predicates and sometimes do not. It patterns like telic predicates insofar as it is incompatible with for-adverbials and compatible with in-adverbials, like English accomplishments. The data are repeated in (43).

(43) a. *Mǎlì chī-le yī-ge píngguǒ wǔ fēnzhōng.

Mary eat-LE one-CL apple five minute ‘?Mary ate an apple for five minutes.’ b. Mǎlì wǔ fēnzhōng nèi chī-le yī-ge píngguǒ.

Mary five minute within eat-LE one-CL apple ‘Mary ate an apple in five minutes.’

176

It patterns unlike telic predicates as it allows a non-culmination interpretation. The data is repeated in (44).

(44) Mǎlì chī-le yī-gè píngguǒ, hái shèng liǎng-kǒu, bèi wǒ chī-le.

Mary eat-LE one-CL apple still left two-bite PASS I eat-LE ‘Mary ate one apple. There was some left over and I ate it.’

In order to account for the variable telicity, I assume that non-culmination is an option for non-RVC accomplishments. I assume the analysis of telicity proposed by Krifka (1989, 1992, 1998) and Zucchi and White (2001) as the basis for analyzing non-RVC accomplishments, and propose that non-RVC accomplishments are optionally associated with a non-culmination operator which is removed when a RVC morpheme is used. There are a number of studies of non-culmination in the literature and the analyses proposed in those studies are based on different observations regarding the data. For example, Bar-el et al. (2005) studied a transitiving morpheme in two Salish languages which removes the culmination entailment of the verb root that it suffixes with. They treated the morpheme as introducing inertia worlds 10 where the culmination point is realized. Tatevosov (2008) observed that in Karachay-Balkar there are two types of non-culmination interpretations. The denoted event might be just an attempt to make the event happen but it fails completely, or the denoted event occurs partially. It was also found that different verbs license different types of non-culmination readings. This analysis is based on a decomposition of the verb’s event structure into three subevental components: initiative, process, and result state, each of which can be realized in some inertia world or in the actual world. The two types of non-culmination differ with respect to what subevental components are

10 The notion of inertia worlds was originally proposed by Dowty (1977, 1979) for analyzing the progressive. According to him, inertia worlds are related to the actual world in such a way that events progressing in the inertia worlds resemble their counterparts in the actual world up to the time in question and the further courses of those events after this time point may differ from each other but they all develop in a way compatible with their past courses. 177 realized in an inertia world or the actual world. Koenig and Chief (2008) analyzed non-culmination, or what they called the Incompleteness Effect, in terms of scalarity. Their observation is that non-culmination is largely associated with those verb stems that involve the kind of change of property that is a matter of degree. The analysis associates those incomplete stems with a scale that measures the degree of property. Beavers and Lee (in press) showed that in Korean caused change-of-state predicates also allow both types of non-culmination interpretations noted by Tatevosov (2008). However, they argued that the zero result reading requires an intention by the subject for the result to come about while such an entailment is missing for the partial success reading. Their analysis combined the modal approach and the scalarity approach, as they proposed a sublexical modality over worlds that defines the subject’s intention to account for the zero result interpretation and a scalar semantics for the partial success interpretation. In the current study, I adopt a modal approach to non-culmination of accomplishments like chī yī-ge píngguǒ ‘eat an apple’. But note that there is no particular reason why I choose the modal approach over the scalarity approach. A scalar analysis can also account for the data as shown in Koenig & Chief (2008). Following those prior studies that adopt a modal analysis for non-culmination accomplishments (e.g. Koenig and Muansuwan 2000, Bar-el et al. 2005, Tatevosov 2008), I assume that non-RVC accomplishments like chī yī-ge píngguǒ ‘eat an apple’ are associated with a non-culmination operator, which is overridden when the accomplishment is used in an RVC construction. Note that such non-RVC accomplishments, which involve an incremental theme, apply to events that at least partially occurr in the actual world (Koenig and Chief 2008). They do not apply to the case where the subject has an intention to make the event happen but no activity is really carried out in the actual world. The semantics of the operator is provided in (45), which is adapted from the imperfective operator proposed by Koenig & Muansuwan

(2000: 163).11

11 Koenig & Muansuwan (2000) studies non-culmination accomplishments in Thai, and suggested that the verb stems in Thai are imperfective. They proposed an imperfective operator built in the lexical entries of accomplishments, which is repeated as follows. 178

(45) Semantics for the non-culmination operator: a. α = Non-culm(e, φ) b. An eventuality e and an event description φ satisfy condition (a) iff

∃e’[e⊆e’ ⋀ ∀w’[w’ is an inertia world w.r.t. w at the beginning of e ⋀ φ(e’)]]

The non-culminating accomplishment chī yī-ge píngguǒ ‘eat an apple’ is translated as follows:

(46) eat one apple

w,g 〚 chi yi-ge pingguo 〛 = λe.Non-culm(e, λe’∃x,t[eat’(e’) ⋀ PAT(e’,x) ⋀

TIME(e’,t) ⋀ apple’(x) ⋀ NU(apple)(x)=1])

The analysis ensures that the denotation of the non-culminating accomplishment chī yī-ge píngguǒ ‘eat an apple’ consists of a set of events in the actual world in which some activity of eating at an apple occurs but culmination with respect to the apple might be realized in the actual world or in an inertia worlds.

4.6 THE ANALYSIS

The proposed analysis is cast in the model developed by Krifka (1989, 1992,

1998), and I also assume that telicity arises from quantization. Essentially, the analysis of -wán and -diào is maximalization over times and patients respectively. Following Krifka (1989, 1992, 1998), I assume that all entities fall into domains of individuals, events, paths, and times, and they form mereological part/whole structures of individuals, events, paths, and times which are related to each other via a series of homomorphic relations. But for purposes of the current study, we

a. α = Impfv(ev, φ) b. An eventuality ev and an event description φ satisfy condition a iff there is an e' which (non-necessarily properly) includes ev and satisfies φ in all 'inertia' worlds—i.e. in all worlds compatible with what it would mean to complete ev without being interrupted. 179 only need a ternary framework, which consists of the structures of events, times, and individuals that are homomorphically related. The homomorphic relations preserve the part/whole structures in such a way that the mereological properties of one structure may transfer to another one. For example, for an event that involves an incremental theme (e.g. drink a glass of wine), the specified quantity of the theme (a glass of wine) determines the boundedness of the event, which is due to an effect of a transfer of the mereological property of the structure of theme to the structure of event via a homomorphism θ that is a SINC. What is proposed here is that telicity may not just arise from a transfer of mereological property from the individual structure to the event structure but also from a transfer of mereological property from the time structure to the event structure. The definition of the homomorphic relation (SINC) between the patient structure and the event structure was presented in (31). So we only need to consider here the homomorphic relation between the time structure and the event structure. First consider a part structure defined as follows (Krifka 1998: 199).

(47) P = is a part structure iff

a. Up is a set of time points;

b. ⊕p, the sum operator, is a function from Up ×Up to Up that is idempotent, commutative, and associative, that is:

∀x,y,z ∊Up [x⊕p x = x ⋀ x ⊕p y = y ⊕p x ⋀ x ⊕p (y ⊕p z) = (x⊕p y) ⊕p z]

c. ≤p, the part relation, defined as ∀x,y ∊Up [x≤p y ↔ x⊕p y=y]

e. ⊗p, the overlap relation, defined as ∀x,y ∊Up[x⊗p y ↔ ∃z∊Up[z≤ px ⋀ z≤ py]

f. Remainder principle: ∀x,y∊Up[x

The part structure assumes an ontological domain structured by the mereological part relation ≤p, defined from the sum operation. The part relation has the properties of being reflexive, transitive, and antisymmetric, that is, a partial order relation.

180

The time structure is defined as a directed path structure. The directed path structure is a part structure which has the additional properties of adjacency and precedence such that for two externally connected or adjacent entities, one must precede the other. The definition of a directed path structure is presented as follows (Krifka 1998: 203-205).

(48) D = < UD, ⊕D, ≤D, is a directed path structure iff

a. < UD, ⊕D, ≤D, is a part structure,

b. ∞D, adjacency, is a two-place relation in UD such that

(i) ∀x,y∊UD [x ∞D y →﹁x ⊗D y]

(ii) ∀x,y,z∊UD [x ∞D y ⋀ y ≤D z → x ∞D z ⋁ x ⊗D z]

c. CD ⊆ UD, the set of convex elements, is the maximal set such that

∀x,y,z∊CD [y,z ≤D x ⋀ ﹁y ⊗D z ⋀ ﹁y ∞D z →

∃u∊CD[u≤D x ⋀ u ∞D y ⋀ u ∞D z]]

d. PD ⊆ CD, the set of path, is the maximal set such that

∀x,y,z∊PD [y,z ≤D x ⋀ ﹁y ⊗D z ⋀ ﹁y ∞D z →

∃u∊CD[u≤D x ⋀ y ∞D u ∞D z]]

e. ∀x,y∊UD [﹁x ⊗D y ⋀ ﹁x ∞D y → ∃u∊PD[x ∞D z ∞D y]

f. DD ⊆ PD, the set of directed paths, is the maximal set, and «D, precedence, is

a two-place relation in DD with the properties:

(i) ∀x,y∊DD [[﹁x «D x] ⋀ [x «D y →﹁y «D x] ⋀ [x «D y ⋀ y «D z → x «D z]]

(ii) ∀x,y∊DD [x «D y → ﹁x ⊗D y]

(iii) ∀x,y,z∊DD [x,y ≤D z ⋀ ﹁x ⊗D y → x «D y ⋁ y «D x]

(iv) ∀x,y∊DD [x «D y → ∃z∊DD[x,y ≤D z]]

The condition for adjacency in (48b) says that adjacent elements do not overlap, and that if an element x is adjacent to y and y is a part of z, either x is adjacent to z or x and z overlap. The condition for convex element in (48c) says that all convex elements that do not overlap or are adjacent are connected by a convex element. The

181 condition for path (48d) says that two disjoint, non-adjacent parts of a path are always connected by exactly one subpath, and (48e) says each two disjoint, non-adjacent elements are connected by a path. The conditions (48a-e) define a basic path structure. The condition (48f) adds the precedence property for a directed path structure. It says that (i) precedence is irreflexive, asymmetric and transitive, (ii) it only holds for non-overlapping elements, (iii) if two subpaths of a directed path do not overlap, one must precede the other, and (iv) only parts of a directed path can stand in the precedence relation to each other. The formal definition of the time structure is presented in (49) (Krifka 1998: 205). The condition for being one-dimensional as in (49b) says that if two times do not overlap, one must precede the other.

(49) T = < UT, ⊕T, ≤T, is a time structure iff

a. < UT, ⊕T, ≤T, is a directed path structure,

b. T is one-dimensional: ∀x,y∊DT [﹁x ⊗T y → x «T y ˅ y «T x]

The time structure is related to the event structure by a homomorphism that isomorphically preserves properties of parthood, adjacency, and precedence. The event structure is defined by Krifka (1998: 206) as in (50).

(50) E= < UE, ⊕E, ≤E, is an event structure iff

a. < UE, ⊕E, ≤E, is a part structure,

b. TE is a time structure < UT, ⊕T, ≤T,

c. τE, the temporal trace function, is a function from UE to UT,

∞E, temporal adjacency, is a two-place relation in UE,

«E, temporal precedence, is a two-place relation in UE,

CE, the set of temporally contiguous events, is a subset of UE, with the following properties:

(i) ∀e,e’∊UE[τE(e⊕E e’) = τE(e) ⊕E τE(e’)]

(ii) ∀e,e’∊UE[e ∞E e’ ↔ τE(e) ∞E τE(e’)] 182

(iii) ∀e,e’∊UE[e «E e’ ↔ τE(e) «E τE(e’)]

(iv) ∀e ∊ CE[τE(e) ∊ PT]

(v) UE is the smallest set such that CE ⊆ UE,

and for every e, e’ ∊ UE, e⊕E e’∊ UE.

The temporal trace function τE maps events to their run times. The properties of temporally contiguous events in (50ci-v) say that (i) the sum of the run time of two events equals the run time of the sum of the two events, (ii) adjacent events also have their run times temporally adjacent, (iii) precedence relations for events also hold for their run times, (iv) temporally contiguous events are events with a contiguous run time, and (v) the set of all events is the closure of the contiguous events under sum formation. I base my analysis on the ternary framework defined above, and assume that telicity arises from quantized reference. In particular, I utilize time in my analysis and assume that predicates over time can have quantized reference as well. I adopt Zucchi & White’s (2001) definition of quantization, which is repeated in (51).

(51) QUA(P) iff for every model M, assignment g, and individual a, b,

if〚P〛M,g(a)=1 and〚P〛M,g(b)=1, then a is not a proper part of b.

Note that in their analysis, the assignment function only applies to events and ordinary individuals, but here I propose that it also applies to times (or temporal individuals). I assume that temporal expressions like from 3:01 to 3:05 are time-denoting expressions. It is a quantized predicate since no subpart of the time from 3:01 to 3:05 is also a time from 3:01 to 3:05. So for the event predicate eat apples from 3:01 to 3:05, we have assignments of events, individuals, and times. The assignment function assigns this predicate a set of events of eating apples, run times from 3:01 to 3:05, and particular but unspecified amounts of apples. The analyses for -wán and -diào are provided below, and I present an account for the data presented above. 183

4.6.1 Analyzing -wán

The proposed analysis for -wán is based on the assumption that event time is a semantic argument of an eventive predicate. It is reasonable to treat event time as one of the semantic arguments of a verbal predicate, though it is apparently different from a regular semantic argument. As pointed out by Krifka (1998), a predicate establishes a relation of a specific type between some number of semantic arguments, and the number and types of the semantic arguments involved in a predication may vary in different cases. For example, the sentence John slept establishes a SLEEP-type relation between the individual John and some event; Mary ate the apple establish an EAT-type relation between Mary, a particular apple, and some event; and John walked from the campus to the capitol establishes a WALK-type relation among John, some event, and some path in space between the campus and the capitol. A semantic argument might be expressed by a syntactic argument such as the subject and the object, and it might also be expressed by a syntactic adjunct such as from the campus to the capitol. Krifka did not consider event time as a semantic argument; however, for the purpose of the current study, I propose that we may take event time as a semantic argument. In the semantic representation, we may treat event time as a special semantic argument like the semantic argument of event e we normally assume for a verbal predicate in the (neo-)Davidsonian approach to verb meaning, that is, it is saturated by existential closure at the end of composition. For example, the meaning of the verb chī ‘eat’ is represented as follows:

(52) 〚chi〛= λxλyλtλe[eat’(e) ⋀ AG(e, y) ⋀ PAT(e, x) ⋀ TIME(e, t)]

Where TIME(e, t)= 1 iff t= τ(e)

In this representation, the eventive predicate chī ‘eat’ is analyzed as a four-place relation that relates a patient x, an agent y, and an event time t to an event e. Event time is defined as the run time of the denoted event, which is the output of the temporal trace function τ mapped onto the domain of event. 184

The proposed analysis for -wán is an extension of the maximal participant account proposed by Zucchi & White (2001) from the domain of entity of patient to the domain of entity of time. I propose that -wán presupposes the predicate it applies to is durative and dynamic, and it denotes a function that maps the event time onto a maximal set of times, which I call the maximal time participant. The representation of -wán is presented in (53).

(53) 〚wan〛= λPλtλe[P(t,e) ⋀ Max(λs∃e’[P(s,e’) ⋀ τ(e’)⊆tr], t)]

This analysis is illustrated by the following sentence in which -wán is compounded with the verb chī ‘eat’.

(54) Mǎlì chī-wán-le yī-ge píngguǒ.

Mary eat-TERM-LE one-CL apple

‘Mary finished eating an apple.’

I assume the following tree as the underlying representation for (54), which undergoes some kind of head movement in order to get the surface representation. The syntactic representation says that the semantics of -wán comes into the compositional system after the composition of the meaning of the clause Mary eats one apple.12 I disregard the aspectual -le in the tree. But the assumption, which follows Zucchi and White (2001), is that the aspectual node introduces a variable which identifies the reference time of the sentence. It is identical to the variable tr. In a simple case where the sentence refers to a particular event, the variable introduced by -le and the variable tr are bound by the same existential quantifier.

12 Note that it is just a tentative assumption that -wán enters the composition after composition of the meaning of the base sentence. We may alternatively assume that -wán is combined with the verb before composing for the meaning of the VP. Either of the two treatments of -wán is fine. 185

(55)

A derivation of this sentence is provided below. Following Krifka (1989), I assume classifiers denote a NU (Natural Unit) function which measures the number of objects. So the analysis for the variable introduced by the indefinite classifier construction assumed here is that it is existentially bound immediately as opposed to the analysis of a variable being bound at the discourse level proposed in the Kamp-Heim analysis for English indefinite NPs.

(56) a. apple

〚pingguo〛= λx[apple’(x)]

b. one apple

〚yi-ge pingguo〛= λPλyλtλe∃x[P(x,y,t,e) ⋀ apple’(x) ⋀ NU(apple)(x)=1]

c. eat

〚chi〛= λxλyλtλe[eat’(e) ⋀ AG(e, y) ⋀ PAT(e, x) ⋀ TIME(e, t)]

d. eat one apple

〚chi yi-ge pingguo〛= λPλyλtλe∃x[P(x,y,t,e) ⋀ apple’(x) ⋀ NU(apple)(x)=1]

(λxλyλtλe[eat’(e) ⋀ AG(e,y) ⋀ PAT(e,x) ⋀ TIME(e,t)])

= λyλtλe∃x[eat’(e) ⋀ AG(e,y) ⋀ PAT(e,x) ⋀ TIME(e,t) ⋀ apple’(x) ⋀

NU(apple)(x)=1] e. Mary

〚Mali〛= λPλe[P(M,e)]

f. Mary eats one apple

〚 Mali chi yi-ge pingguo 〛 = λPλe[P(M,e)](λyλtλe∃x[eat’(e) ⋀ AG(e,y) ⋀

PAT(e,x) ⋀ TIME(e,t) ⋀ apple’(x) ⋀ NU(apple)(x)=1])

186

= λtλe∃x[eat’(e) ⋀ AG(e,M) ⋀ PAT(e,x) ⋀ TIME(e,t) ⋀ apple’(x) ⋀

NU(apple)(x)=1] g. -wan

= λPλtλe[P(t, e) ⋀ Max(λs∃e’[P(s, e’) ⋀ τ(e’)⊆tr], t)] h. -wan (Mary eats one apple)

〚 Mali chi yi-ge pingguo wan 〛 = λPλtλe[P(t,e) ⋀ Max(λs∃e’[P(s,e’) ⋀

τ(e’)⊆tr], t)](λtλe∃x[eat’(e) ⋀ AG(e,M) ⋀ PAT(e,x) ⋀ TIME(e,t) ⋀ apple’(x)

⋀ NU(apple)(x)=1])

= λtλe∃x[eat’(e) ⋀ AG(e,M) ⋀ PAT(e,x) ⋀ TIME(e,t) ⋀ apple’(x) ⋀

NU(apple)(x)=1 ⋀ Max(λs∃e’∃x[eat’(e’) ⋀ AG(e’,M) ⋀ PAT(e’,x) ⋀

TIME(e’,s) ⋀ apple’(x) ⋀ τ(e’)⊆tr], t)] i. Existential Closure:

∃e∃t∃x[eat’(e) ⋀ AG(e,M) ⋀ PAT(e,x) ⋀ TIME(e,t) ⋀ apple’(x) ⋀

NU(apple)(x)=1 ⋀ Max(λs∃e’∃x[eat’(e’) ⋀ AG(e’,M) ⋀ PAT(e’,x) ⋀

TIME(e’,s) ⋀ apple’(x) ⋀ τ(e’)⊆tr], t)]

“An event of Mary eating an apple is an eating event whose event time is the maximal time interval during which Mary was eating an apple at reference time.”

According to this translation, combining an event predicate chī yī-ge píngguǒ ‘eat one apple’ with -wán gives rise to a quantized predicate. Specifically, given the homomorphic relation between the event structure and the time structure, a proper part e’ of an event e in the denotation of chī-wán yī-ge píngguǒ ‘finish eating one apple’ can only correspond to a proper part of the time interval during which the event progresses at the reference time tr. But this means that e’ is not in the denotation of chī-wán yī-ge píngguǒ ‘finish eating one apple’ since by the definition of quantization provided in (35), events in this denotation must have the entire time interval at tr as the time participant. So the resulting predicate is quantized. When -wán combines with atelic predicates like chī píngguǒ ‘eat apples’, 187 which do not specify for the quantity of patient, the resulting predicate is still telic. The final representation of a sentence that contains such a complex predicate is provided as follows:

(57) -wan (Mary eats apples)

〚Mali chi pingguo wan〛= ∃e∃t∃x[eat’(e) ⋀ AG(e,M) ⋀ PAT(e,x) ⋀ TIME(e,t) ⋀

apple’(x) ⋀ Max(λs∃e’∃x[eat’(e’) ⋀ AG(e’,M) ⋀ PAT(e’,x) ⋀ TIME(e’,s) ⋀

apple’(x) ⋀ τ(e’)⊆tr], t)]

“An event of Mary eating apples is an eating event whose event time is the maximal time interval during which Mary was eating apples at reference time.”

According to this translation, combining the atelic predicate chī píngguǒ ‘eat apples’ with -wán also gives rise to a quantized predicate with respect to the time participant. The reason is the same as the one above. A proper part e’ of an event e in the denotation of chī-wán píngguǒ ‘finish eating apples’ only corresponds to a proper part of the time interval during which the event progresses at the reference time tr. This means that e’ is not in the denotation of chī-wán píngguǒ ‘finish eating apples’ since by definition events in this denotation must have the entire time interval at tr as the time participant. As noted above, expressions with -wán always entail termination for the denoted event, which is shown by the entailment of full precedence when two events are put in a temporal sequence. Consider the example (9), which is repeated in (58). As mentioned above, the discourse entails that the event of eating apples fully precedes the event of wanting to eat a banana.

(58) Mǎlì chī-wán-le píngguǒ, hái bú gòu, hái xiǎng chī xiāngjiāo.

Mary eat-TERM-LE apple still not enough still want eat banana ‘Mary finished eating apples. It was not enough. She still wanted to eat a banana.’ 188

In the current analysis, the termination entailment is explained as follows: The function of maximalization over time that -wán denotes necessarily takes the entire run time of the event as the input, which includes the initial and the final points of the event time. This gives rise to an effect such that we can infer that the event reaches some specific final point, which is the point at which the event terminates. So -wán entails termination of the event. To account for the full precedence entailment in (58), we may postulate that terminated events are treated like units and by default they prohibit an overlapping interpretation with other events expressed in discourse. The morpheme -wán does not just entail termination for the denoted event but also culmination with respect to the patient. The patient might be yī-ge píngguǒ ‘one apple’, which specifies the quantity, or a bare noun píngguǒ ‘apple’, which does not specify the quantity. The example is presented in (59), which is repeated from (14).

(59) *Mǎlì chī-wán-le (yī-ge) píngguǒ, hái shèng liǎng-kǒu, bèi wǒ chī-le.

Mary eat-TERM-LE one-CL apple still left two-bite PASS I eat-LE ‘*Mary finished eating an/the apple. There was some left over and I ate it.’

I assume that the two noun phrases are just normal noun phrases in the sense that they have the properties that we normally assume noun phrases have such as referentiality, definiteness, and specificity. My account for the culmination reading is based on an assumption that transferring mereological properties from one structure to another via homomorphism would have an effect that maximalization over one argument also obtains for the other. In this case, maximalization over time, a function of -wán, puts a constraint on the time structure, whose mereological properties are then transferred to the event structure by the homomorphic relationship between them. But since the event is in turn homomorphically (in fact, isomorphically) related to the patient, this will in turns ensures that we get constraints on the patient such that it will have a kind of maximal reading as well. If the noun phrase is yī-ge píngguǒ ‘one apple’, it is one apple that is picked out as the maximal patient. If it is a bare noun píngguǒ ‘apple’, it

189 is a contextually specific amount of apples that is picked out as the maximal patient. In either case, the patient is maximal, which means all parts of it are affected in the event, and the event culminates with respect to it. The proposed analysis of -wán also allows it to be compounded with atelic predicates such as activities e.g. xiào ‘laugh’ and yóu-yǒng ‘swim’, and semelfactives e.g. qiāo ‘knock’ and késòu ‘cough’. In this case, the patient structure is irrelevant as there is no patient involved in those types of events or the patient is not an incremental theme (e.g. qiāo ‘knock’). Such a predicate compounded with -wán is quantized because, for example, a proper part e’ of an event of swimming e can only correspond to a proper part of the time interval during which the swimming event e progresses at reference time, which means e’ is not in the denotation of the predicate yóu-wán-yǒng ‘finish swimming’ as it does not have the entire time interval at reference time as the time participant.

4.6.2 Analyzing -diào

The proposed analysis for -diào is that it denotes a function that maps the patient involved in the event at reference time tr onto a maximal entity, which I call the maximal patient participant. This analysis is decomposed from Zucchi & White’s (2001) analysis of maximal patient (see Section 4.4.2). The formalization is presented in (60).

(60) 〚diao〛= λPλyλxλtλe[P(y,x,t,e) ⋀ Max(λz∃e’∃s[P(y,z,s,e’) ⋀ τ(e’)⊆tr], x)]

This analysis is first illustrated by the following sentence in which -diào is compounded with the verb chī ‘eat’.

(61) Mǎlì chī-diào-le yī-ge píngguǒ.

Mary eat-CULM-LE one-CL apple

‘Mary ate up one apple.’

190

I assume the following tree as the syntactic representation for (61), which is different from the syntactic representation for the corresponding sentence with -wán. In this representation, I assume that -diào applies to the verb as opposed to the clause to form an RVC construction.

(62)

The major steps of the derivation of the sentence (61) are presented as follows.

(63) a. one apple

〚yi-ge pingguo〛= λPλyλtλe∃x[P(x,y,t,e) ⋀ apple’(x) ⋀ NU(apple)(x)=1]

b. -diao

= λPλyλxλtλe[P(y,x,t,e) ⋀ Max(λz∃e’∃s[P(y,z,s,e’) ⋀ τ(e’)⊆tr], x)] c. eat-diao

〚chi-diao〛= λPλyλxλtλe[P(y,x,t,e) ⋀ Max(λz∃e’∃s[P(y,z,s,e’) ⋀ τ(e’)⊆tr], x)]

(λxλyλtλe [eat’(e) ⋀ AG(e,y) ⋀ PAT(e,x) ⋀ TIME(e,t)])

= λxλyλtλe[eat’(e) ⋀ AG(e,y) ⋀ PAT(e,x) ⋀ TIME(e,t) ⋀ Max(λz∃e’∃s[eat’(e’)

⋀ AG(e’,y) ⋀ PAT(e’,z) ⋀ TIME(e’,t) ⋀ τ(e’)⊆tr], x)] d. eat-diao one apple

〚 chi-diao yi-ge pingguo 〛 = λPλyλtλe∃x[P(x,y,t,e) ⋀ apple’(x) ⋀

NU(apple)(x)=1](λxλyλtλe[eat’(e) ⋀ AG(e,y) ⋀ PAT(e,x) ⋀ TIME(e,t) ⋀

Max(λz∃e’∃s[eat’(e’) ⋀ AG(e’,y) ⋀ PAT(e’,z) ⋀ TIME(e’,s) ⋀ τ(e’)⊆tr], x)])

= λyλtλe∃x[eat’(e) ⋀ AG(e,y) ⋀ PAT(e,x) ⋀ TIME(e,t) ⋀ apple’(x) ⋀

NU(apple)(x)=1 ⋀ Max(λz∃e’∃s[eat’(e’) ⋀ AG(e’,y) ⋀ PAT(e’,z) ⋀

TIME(e’,s) ⋀ τ(e’)⊆tr], x)]

191

e. Mary eat-diao one apple

〚Mali chi-diao yi-ge pingguo〛= λPλe [P(M,e)](λyλtλe∃x[eat’(e) ⋀ AG(e,y) ⋀

PAT(e,x) ⋀ TIME(e,t) ⋀ apple’(x) ⋀ NU(apple)(x)=1 ⋀ Max(λz∃e’∃s[eat’(e’)

⋀ AG(e’,y) ⋀ PAT(e’,z) ⋀ TIME(e’,s) ⋀ apple’(z) ⋀ τ(e’)⊆tr], x)])

= λtλe∃x[eat’(e) ⋀ AG(e,M) ⋀ PAT(e,x) ⋀ TIME(e,t) ⋀ apple’(x) ⋀

NU(apple)(x)=1 ⋀ Max(λz∃e’∃s[eat’(e’) ⋀ AG(e’,M) ⋀ PAT(e’,z) ⋀

TIME(e’,t) ⋀ τ(e’)⊆tr], x)] f. Existential Closure:

∃e∃t∃x[eat’(e) ⋀ AG(e,M) ⋀ PAT(e,x) ⋀ TIME(e,t) ⋀ apple’(x) ⋀

NU(apple)(x)=1 ⋀ Max(λz∃e’∃s[eat’(e’) ⋀ AG(e’,M) ⋀ PAT(e’,z) ⋀

TIME(e’,s) ⋀ τ(e’)⊆tr], x)]

“An event of Mary eating one apple is an eating event whose patient is the sum of all parts of the apple eaten by Mary at the reference time and the apple Mary

ate was one apple.”

According to this translation, the derived event predicate chī-diào yī-ge píngguǒ ‘eat up one apple’ is quantized. Specifically, given that the thematic role of the object assigned by the verb chī ‘eat’ has the property of mapping to object, a proper part e’ of an event e in the denotation of chī-diào yī-ge píngguǒ ‘eat up one apple’ corresponds to a proper part of the object, which, however, is not the maximal participant of the event e with respect to the reference time. This means that e’ is not in the denotation of chī-diào yī-ge píngguǒ ‘eat up one apple’, so by the definition of quantization, the predicate is quantized. When the object is a bare noun, it receives a specific or definite interpretation. This can be explained by the meaning of -diào, which requires a specific quantity of the patient. The final representation of the sentence is presented as follows.

192

(64) Mary eat-diao apples

〚Mali chi-diao pingguo〛= ∃e∃t∃x[eat’(e) ⋀ AG(e,M) ⋀ PAT(e,x) ⋀ TIME(e,t) ⋀

apple’(x) ⋀ Max(λz∃e’∃s[eat’(e’) ⋀ AG(e’,M) ⋀ PAT(e’,z) ⋀ TIME(e’,s) ⋀

τ(e’)⊆tr], x)]

“An event of Mary eating some/the apple(s) is an eating event whose patient is the sum of all parts of the apple(s) eaten by Mary at the reference time.”

According to the translation, the predicate chī-diào píngguǒ ‘eat up some/the apple(s)’ is quantized. This is because a proper part e’ of an event e in the denotation of chī-diào píngguǒ ‘eat up some/the apple(s)’ corresponds to a proper part of the object, which is not the maximal patient participant of the event e at the reference time. So e’ is not in the denotation of chī-diào píngguǒ ‘eat up some/the apple(s)’ and thus the predicate is quantized.

Culmination is always entailed when the expression is compounded with -diào, regardless of whether the object is a classifier construction yī-ge píngguǒ ‘one apple’ or a bare noun píngguǒ ‘apple’. In the former case, it entails that the event culminates with respect to a certain apple. In the latter case, the bare noun receives a specific or definite interpretation owing to maximalization of the patient and the sentence entails culmination with respect to the patient. The culmination entailment in the two cases is illustrated by the contradiction of the discourse (65).

(65) *Mǎlì chī-diào-le (yī-ge) píngguǒ, hái shèng liǎng-kǒu, bèi wǒ chī-le.

Mary eat-CULM-LE one-CL apple still left two-bite PASS I eat-LE ‘*Mary ate up an/the apple. There was some left over and I ate it.’

The culmination entailment in this case, as opposed to the culmination entailment in the case when -wán is compounded with an accomplishment (see above), is explained as follows: The function of maximalization over patient that -diào denotes necessarily takes all parts of the patient as the input, entailing that all parts of whatever the

193 contextually defined patient at tr is are affected. This gives rise to an effect such that we can infer that the event culminates with respect to the entire patient. In a consumption event, it entails that all parts of the food are consumed. So we can infer that the event culminates with respect to (all parts of) the patient. We also noted above that expressions compounded with -diào entail termination as well, which is shown by the entailment of full precedence when two events are put in a temporal sequence. The example from (11) is repeated in (66). As mentioned above, the discourse entails that the event of eating an apple fully precedes the event of wanting to eat a banana.

(66) Mǎlì chī-diào-le píngguǒ, hái bú gòu, hái xiǎng chī xiāngjiāo.

Mary eat-CULM-LE apple still not enough still want eat banana ‘Mary ate up the apple. It was not enough. She still wanted to eat a banana.’

Our model of the ternary framework can provide an account for the entailment of full precedence in this discourse. Given the homomorphic relation between the three structures of event, patient, and time, maximalization over the patient would cause a transfer of mereological properties from the patient structure to the event structure, picking out the maximal event, and then a transfer of mereological properties from the event structure to the time structure, picking out the maximal time of the event, which is a time stretch that includes the initial and final points of the event, therefore event termination is entailed. So by using -diào, which directly expresses culmination, the speaker indirectly expresses termination. This explains why (66) entails the event of eating an apple terminates before Mary wanted to eat a banana.

4.7 CONCLUSION

The current study investigates the semantics of telicity in Mandarin Chinese. It starts out with findings in previous studies which said that accomplishments in Chinese do not entail culmination but the culmination reading is obtained if the

194 accomplishment is compounded with a resultative morpheme such as -wán. The for/in-adverbial diagnostics do show that compounding with -wán results in a telic predicate. However, I argue that -wán does not express completion or culmination, which would make it a case of telicization that has no difference from the kind of telicity that has been the focus of much prior work, that is, telicity obtained through a constraint on the patient. I argue that -wán actually expresses terminative – the event comes to an end, which does not even require the verb to select for a patient. Such a predicate is telic because the event with respect to a particular run time (i.e. event time) does not have a proper part that is the same event with respect to the same run time. I specifically analyzed this as time maximalization. This morpheme suggests that there is another avenue to telicity which has not received much attention in previous studies. In addition to -wán, I also studied the RVC morpheme -diào, which is also a common morpheme used to telicize accomplishments. The morpheme -diào shows a more canonical case of telicization – It requires the verb it compounds with selects for a maximal patient and expresses culmination with respect to the patient. So the two morphemes -wán and -diào show us two different avenues to telicity. The analysis is cast in the mereological framework developed by Krifka (1989, 1992, 1998). Following Zucchi & White (2001), Filip & Rothstein (2006), Filip (2008, 2017), and Kardos (2016), I also adopt a maximalization analysis and extend it to time.

The proposed analysis is based on a ternary framework which consists of three homomorphically related structures, namely, event, patient, and time. Telicity through time is analyzed as a quantized reference to time. Such a predicate is quantized because the maximal run time does not have a proper part that is still the maximal run time at reference time, given homomorphic relations, no proper part e’ ⊂ e bears the same predicate applied to e with respect to the run time at reference time. In parallel, telicity through patient is analyzed in terms of quantized reference for a patient DP.

Such a predicate is quantized because the maximal patient does not have a proper part that is still the maximal patient at reference time, given the homomorphic relations, no proper part e’ ⊂ e bears the same predicate applied to e with respect to the same 195 amount of patient.

196

Chapter 5: Conclusion

This dissertation explores two types of compounding in Chinese, namely, noun incorporation and resultative verb compounding. For noun incorporation, it is not obvious at first sight that Chinese has this phenomenon. However, I argued that the object of the S-le sentence is a case of incorporation. For resultative verb compounding, I investigate what semantics the resultative morpheme may contribute to the aspectual meaning of the entire compound, in particular, I studied the meanings of the two resultative morphemes, -wán and -diào. The current study relates the S-le sentence to the typologically common phenomenon of noun incorporation. I have shown that the S-le sentence is subject to certain syntactic constraints that do not apply to a similar structure, the V-le sentence. In the former case, the object is usually a small NP that prohibits the indefinite article and quantifiers; however, in the latter case, no such modificational restrictions apply. This suggests that the object of S-le sentences has an unusual syntactic status and thus it provides syntactic evidence for noun incorporation. I also provided data to show that those semantic properties that are usually associated with noun incorporation also apply to the object of the S-le sentence, including narrow scope relative to negation, modality, and universal quantifier, discourse opacity, and number neutrality if it is a bare noun. This provides semantic evidence for noun incorporation in the S-le sentence. I investigated the semantics of the S-le sentence. Building on Liu (2002), Sybesma (1999), and Soh (2009), I proposed that the S-le sentence expresses informativeness, which I analyse in terms of presupposition in a Stalnakerian framework that the proposition is new to the hearer. This explains why the S-le sentence is typically used in a context in which the speaker provides new information for the hearer. As for the case of using S-le sentences in non-informative context, I suggest that it can be reconciled with the informative use if we assume that S-le

197 sentences express informativeness, but speakers may ignore it intentionally and the evaluative meaning fills in the communicative function. I also showed that the kind of informative contexts for S-le sentences are a general kind of informative contexts. When the contexts become more complicated in the sense that some components of the informing proposition are already known to the hearer, S-le sentences are blocked and other more specialized sentence structures are used.

I also argued that S-le sentences are free of grammatical aspect. I showed that S-le sentences are compatible with a range of (informative) contexts that require different aspectual meanings for the proposition. This can best be explained by the hypothesis that the S-le sentence has no inherent grammatical aspect. As for the two resultative morphemes -wán and -diào, I argued that they reveal two different avenues to telicity. One is through time as suggested by -wán, and the other is through patient as suggested by -diào. I proposed an analysis on the basis of the model developed by Krifka (1989, 1992, 1998) and Zucchi & White (2001), which includes homomorphically related structures of event, patient, and time. The assumption is that telicity is due to quantization, specifically quantization over patient. Krifka did not consider that time could matter for telicity, but the data of -wán show that it can. So the analysis for -wán and -diào says that the former derives quantized reference to time and the latter quantized reference to patient. This extends the original analysis of telicity from the nominal domain to the temporal domain.

198

References

Abney, Steven. 1987. The English noun phrase in its sentential aspect. Doctoral dissertation, MIT, Cambridge, Mass. Baker, Mark C. 1988. Incorporation: A theory of grammatical function changing. Chicago: University of Chicago Press.

Baker, Mark C. 1996. The polysynthesis parameter. Oxford: Oxford University Press. Baker, Mark C. 2009. Is head movement still needed for noun incorporation? Lingua 119:148-165. Barrie, Michael, and Eric Mathieu. 2016. Noun incorporation and phrasal movement. Natural Language Linguistic Theory 34: 1-51. Barrie, Michael, and Yen-Hui Audrey Li. 2015. The semantics of (pseudo) incorporation and case. In The Syntax and Semantics of Pseudo-Incorporation,

eds. Olga Borik, and Berit Gehrke, 159-188. The Netherlands: Brill. Beaver, David I. and Geurts, Bart, “Presupposition”, The Stanford Encyclopedia of Philosophy (Winter 2014 Edition), Edward N. Zalta (ed.), URL = . Beavers, John. 2012. Lexical aspect and multiple incremental themes. In Violeta Demonte and Louise McNally (eds.), Telicity, Change, and State: A

Cross-Categorial View of Event Structure, 23-59. Oxford: Oxford University Press. Beavers, John. 2014. Incrementality in Lexical Aspect. Chronos 11, June 16-18, Scuola Normale Superiore, Pisa, Italy. Beavers, John and Juwon Lee. In press. Intentionality, Scalar Change, and Non-culminationin Korean Caused Change-of-State. Journal of Linguistics. Bisang, W. and R. Sonaiya. 1997. Perfect and beyond, from pragmatic relevance to

perfect: the Chinese sentence final particle le and Yoruba ti’, Sprachtypol. Univ. Forsch. (STUF) 50(2): 143-158.

199

Bittner, Maria. 1994. Case, scope, and binding. Dordrecht: Kluwer. Borik, Olga. 2002. Aspect and Reference Time. Oxford: Oxford University Press. Borik, Olga, and Berit Gehrke. 2015. An introduction to the syntax and semantics of pseudo-incorporation. In The Syntax and Semantics of Pseudo-Incorporation, eds. Olga Borik, and Berit Gehrke, 1-43. The Netherlands: Brill. Carlson, Gregory N. 1977. Reference to kinds in English. Ph.D. Dissertation,

University of Massachusetts, Amherst. Chafe, Wallace. 1976. Givennes, contrastiveness, definiteness, subjects and topics. Subject and Topic, ed. by Charles N. Li, 25-55. New York: Academic Press. Chao, Y. R. 1968. A grammar of spoken Chinese. Berkeley and Los Angeles. University of California. Cheng, L. L.-S., and R. Sybesma. 1998. Yi-wan tang, yi-ge tang: Classifers and massifer. The Tsinghua Journal of Chinese Studies, New Series 28: 385-412.

Chierchia, Gennaro. 1998. Reference to kinds across languages. Natural Language Semantics 6: 339-405. Chierchia, Gennaro. 2015. How universal is the mass/count distinction? Three grammars of counting. In Chinese syntax in a cross-linguistic perspective, eds. Audrey Li, Andrew Simpson, and Dylan Tsai, 1-48. New York: Oxford University Press. Christensen, Matthew. 2000. Anaphoric reference in spoken and written Chinese

narrative discourse. Journal of Chinese Linguistics 28(2): 303–336. Chu, Chauncey. 1998. A discourse grammar of Mandarin Chinese. New York: Peter Lang Inc. Chung, Sandra, and William A. Ladusaw. 2004. Restriction and saturation. MA: MIT Press. Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press.

Dayal, Veneeta. 2011. Hindi pseudo-incorporation. Natural Language and Linguistic Theory 29: 123-167. DeLancey, Scott, 1997. Mirativity: The grammatical marking of unexpected 200

information. Linguistic Typology 1: 33-52. Ernst, Thomas. 2014. Adverbial Adjuncts in Mandarin Chinese. The Handbook of Chinese Linguistics, eds. C.-T. James Huang, Y.-H. Audrey Li, and Andrew Simpson. Malden, MA: Wiley-Blackwell Publishers. Ernst, Tomas and Chengchi Wang. 1995. Object Preposing in Mandarin Chinese. Journal of East Asian Linguistics 4(3), 235-260.

Espinal, M. Teresa, and Louise McNally. 2011. Bare nominal and incorporating verbs in Spanish and Catalan. Journal of Linguistics 47: 87-128. Farkas, Donka and Henriëtte de Swart. 2003. The semantics of incorporation: From argument structure to discourse transparency. Stanford, CA: CSLI Publications. Filip, Hana. 2017. The semantics of perfectivity. Italian Journal of Linguistics 29(1): 167-200. von Fintel, Kai. 2004. Would you believe it? The king of France is back! Presuppositions and truth-value intuitions. In Marga Reimer & Anne Bezuidenhout (eds.) Descriptions and Beyond, 315-341. Oxford University Press. van Geenhoven, Veerle. 1998. Semantic incorporation and indeﬁnite descriptions: Semantic and syntactic aspects of noun incorporation in West Greenlandic. Dissertations in Linguistics (DiLi). Stanford: Center for the Study of Language

and Information. Grice, H. Paul. 1975. Logic and conversation, in P. Cole & J. Morgan (ed.), Syntax and Semantics, 3: Speech Acts, pp. 41–58, New York: Academic Press. Reprinted in H. P. Grice (ed.), Studies in the Way of Words, pp. 22–40, Cambridge, MA: Harvard University Press (1989). Gundel, Jeanette K., Nancy Hedberg and Ron Zacharski. 1993. Cognitive status and

the form of referring expresions in discourse. Language 69(2): 274-307. Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in cross-linguistic studies. Language 86: 663-87. 201

Haugen, D. Jason. 2008. Morphology at the interfaces: Reduplication and noun incorporation in Uto-Aztecan. The Netherlands: John Benjamins Publishing. Heim, Irene. 1982. The semantics of definite and indefinite noun phrase. University of Massachusetts dissertation. Hole, Daniel. 2012. The information structure of Chinese. In The expression of information structure, eds, Manfred Krifka, and Renate Musan, 45-70. Berlin/

Boston: Walter de Gruyter. Hong, Duren. 1957. Ci shi Shenme [what is word]. Shanghai: Shanghai New Knowledge Press. Hopper, P. J. and Thompson, S. A. 1980. Transitivity in Grammar and Discourse. Language 56, 251–299. Huang, James. 2015. On syntactic analyticity and parametric theory. In Chinese syntax in a cross-linguistic perspective, eds. Audrey Li, Andrew Simpson, and

Dylan Tsai, 1-48. New York: Oxford University Press. Huang, Lillian M. and Philip W. Davis. 1989. An aspectual system in Mandarin Chinese Huang, Yan. 1994. The syntax and pragmatics of anaphora: a study with special reference to Chinese. New York: Cambridge University Press. Jackendoﬀ, Ray. 1972. Semantic interpretation in generative grammar. Cambridge MA: MIT Press

Kamp, Hans. 1981. A theory of truth and discourse representation. In Formal Methods in the Study of Language, eds. Jeroen Groenendijk, Theo M.V. Janssen and Martin Stokhof, 277-322. Amsterdam: Mathematical Centre. Kamp, H., J. van Genabith, and U. Reyle, 2011, “Discourse Representation Theory: An Updated Survey”, in D. Gabbay (ed.), Handbook of Philosophical Logic, vol. 15. Berlin: Springer, 125–394.

Kamp, Hans and Ágnes Bende-Farkas. 2019. Epistemic specificity from a communication-theoretic perspective. Journal of Semantics 36: 1-51. Kardos, Éva. 2016. Telicity marking in Hungarian. Glossa: A journal of general 202

linguistics 1:41, 1-37. Koenig, Jean-Pierre & Liancheng Chief. 2008. Scalarity and state-changes in Mandarin (and other languages). In Olivier Bonami & Patricia Cabredo Hofherr (eds.), Empirical issues in syntax and semantics 7, 241–262. Koenig, Jean-Pierre, and Nuttanart Muansuwan. 2000. How to end without ever ﬁnishing. Journal of Semantics, 17: 147–184.

Krifka, Manfred. 1989. Nominal reference, temporal constitution and quantification in event semantics. In Renate Bartsch, Johan van Benthem and Peter van Emde Boas (eds.), Semantics and Contextual Expressions, 75-115. Dordrecht: Foris. Krifka, Manfred. 1992. Thematic relations as links between nominal reference and temporal constitution. In Ivan Sag and Anna Szabolcsi (eds.), Lexical Matters, 29-53. Stanford: CSLI. Krifka, Manfred. 1995. Common nouns: A contrastive analysis of English and

Chinese. In The Generic Book, eds. Gregory Carlson and Francis Pelletier: 398-411. Chicago: University of Chicago Press. Krifka, Manfred. 1998. The origin of telicity. In Susan Rothstein (ed.), Events and Grammar, 197-235. Dordrecht: Kluwer Academic Press. Kroeber, Alfred L. 1909. Noun incorporation in American languages. In XVI. Internationalen Amerikanisten-Kongress, 2d, ed. Franz Heger, 569-576. Vienna and Leipzig: A. Hartleben.

Lakoff, George, and Mark Johnson 1980. Metaphors we live by. Chicago: University of Chicago Press. Lipenkova, Janna. 2014. The syntax-semantics interface in the Chinese ba-construction. Doctoral dissertation. Freien Universität Berlin. Li, Charles, and Sandra Thompson. 1979. Third person pronouns and zero anaphora in Chinese discourse. In Syntax and semantics 12: Discourse and syntax, ed.

Talmy Givon, 311–335. New York: Academic. Li, Charles N. and Sandra A. Thompson. 1981. Mandarin Chinese: A functional reference grammar. Berkeley: University of California Press. 203

Li, Charles N., Sandra A. Thompson, and R. McMillan Thompson. 1982. The

discourse motivation for the perfect aspect: The Mandarin particle le. In Tense-Aspect: Between semantics & pragmatics, eds. Paul J. Hopper and Hopper. 19-44. Amsterdam: John Benjamins Publishing Company. Li, Yingche. 1974. What does ‘disposal’ mean? Features of the verb and noun in Chinese. Journal of Chinese Linguistics. 2 (2): 200-218.

Li, Y.-H. Audrey. 1998. Argument determiner phrases and number phrases. Linguistic Inquiry 29(4): 693–702.

Li, Y.-H. Audrey. 2006. Chinese ba. In The Blackwell companion to syntax, eds. Martin Everaert and Henk van Riemsdijk. Vol. 1, 374-468. Malden, Mass.: Blackwell. Li, Y.-H. Audrey. 2014. Thematic hierarchy and derivational economy. Language and Linguistics. 15 (3): 295-339.

Lin, T.-H. J. 2001. Light verb syntax and the theory of phrase structure. Doctoral Dissertation. University of California, Irvine. Liu, X. 2002. Xiandai Hanyu Juwei “le” de Yufa Yiyi Jiqi Jieshuo. Shijie Hanyu Jiaoxue. 61(3): 70-79. Lu, Zhiwei. 1964. Hanyu de goucifa [Chinese word formation]. Beijing: Science Press. Lü, Shuxiang. 1955. Ba zi yongfade yanjiu [On the usage of ba]. In Hanyu Lunwen Ji

[Collected essays on Chinese grammar]. Beijing: Science Press. Lü, Shuxiang. 1979. Hanyu yufa fenxi wenti [Issues on analyzing Chinese grammar]. Beijing: The Commercial Press. Lü, Shuxiang. 1980. Xiandai hanyu babai ci [Eight Hundred Modern Chinese Words]. Beijing: The Commercial Press. Lü, Shuxiang. 1982. Zhongguo wenfa yaolue [A concise Chinese grammar]. Beijing:

The Commercial Press. Lü, Shuxiang. 2002. Lü Shuxiang xuanji [A collection of Lü Shuxiang]. Changchun: Northeast Normal University Press. 204

Lyons, John. 1982. Deixis and subjectivity: Loquor, ergo sum? In Robert J. Jarvella and Wolfgang Klein (eds.), Speech, Place, and Action: Studies in Deixis and Related Topics, pages 101–124, New York: Wiley. Massam, Diane. 2001. Pseudo noun incorporation in Niuean. Natural Language and Linguistic Theory 19: 153-197. Mei, Kuang. 1978. The ba-sentence in modern Chinese. In Bulletin of the College of

Fine Arts, National Taiwan University No. 27: 145-180. Mellow, J. Dean. 1989. A syntactic analysis of noun incorporation in Cree. M.A. dissertation, Partee McGill University. Mithun, Marianne. 1984. The evolution of noun incorporation. Language 60: 847-894. Mithun, Marianne. 1986. On the nature of noun incorporation. Language 62: 32-37. Mithun, Marianne. 2000. Incorporation. In Morphology, eds. Geert Booij, Christian

Lehmann, and Joachim Mugdan, 916-928. Berlin: deGruyter. Mueller-Reichau, Olav. 2015. Pseudo-incorporation in Russian? Aspectual competition and bare singular interpretation. In The Syntax and Semantics of Pseudo-Incorporation, eds. Olga Borik, and Berit Gehrke, 262-295. The Netherlands: Brill. Paul, Waltraud. 2002. Sentence-internal Topics in Mandarin Chinese: The Case of Object Preposing. Language and Linguistics 3-4, 695-714.

Paul, Waltraud. 2005. Low IP area and left periphery in Mandarin Chinese. Recherches linguistiques de Vincennes 33: 111–134. Paul, Waltraud. 2015. New Perspectives on Chinese Syntax. Berlin/Munich/Boston: De Gruyter Mouton. Percus, Orin. 2006. Antipresuppositions. In Theoretical and empirical studies of reference and anaphora: Toward the establishment of generative grammar as

an empirical science. ed. A. Ueyama, 52–73. Prince, Ellen. 1981. Toward a taxonomy of given-new information. In Radical pragmatics, ed. Peter Cole, 223-256. New York: Academic Press. 205

Pu, Mingming. 1997. Zero anaphora and grammatical relations in Mandarin. In Grammatical Relations: A Functionalist Perspective, ed. T. Givón, 281-321. Amsterdam: J. Benjamins Rothstein, Susan. 2004. Structuring Events: A Study in the Semantics of Lexical Aspect. Oxford: Blackwell. Sadock, Jerrold. 1980. Noun incorporation in Greenlandic: A case of syntactic word

formation. Language 56:300-319. Sapir, Edward. 1911. The problem of noun incorporation in American languages. American Anthropologist 13: 250-282. Shen, Jiaxuan. 2002. Ruhe chuzhi chuzhishi – Lun baziju de zhuguanxing [Can the disposal construction be disposed of? – On the subjectivity of the ba- construction]. Zhongguo yuwen 5: 387-399. Shi, Ziqiang. 1988. Decomposition of perfectivity and inchoativity and the meaning

of the particle LE in Mandarin Chinese. Journal of Chinese Linguistics 18: 95-123. Shyu, Shu-ing. 1995. The Syntax of Focus and Topic in Mandarin Chinese. Ph.D. diss., University of Southern California, Los Angeles. Shyu, Shu-ing. 2001. Remarks on object movement in Mandarin SOV order. Language and Linguistics [Academia Sinica, Taiwan] 2 (1): 93–124. Simpson, Andrew, Zoe Wu and Yan Li. 2016. Grammatical roles, coherence relations,

and the interpretation of pronouns in Chinese. Lingua Sinica 2(2). Smith, Carlota S. 1990. Event types in Mandarin. Linguistics 28, 309-336. Smith, C. S. 1997. The Parameter of Aspect (second edition). Dordrecht: Kluwer. Soh, H. L. 2009. Speaker presupposition and Mandarin Chinese sentence final -le: A unifed analysis of the “change of state” and the “contrary to expectation” reading. Natural Language and Linguistic Theory 27: 623–657.

Spanos, George A. 1979. Contemporary Chinese usage of ‘le’: A survey and a pragmatic proposal. Journal of the Chinese Language Teacher Association 14(1): 36-70, 14(2): 47-102. 206

Spencer, Margret. 1970. The Verbal Aspect System of Standard Colloquial Chinese. PhD Dissertation: The University of Michigan. Stalnaker, Robert. 1978. Assertion. In Syntax and Semantics: 9, ed. P. Cole, 315-322. New York: Academic Press. Stalnaker, Robert. 1998. On the representation of context. Journal of Logic, Language and Information 7: 3–19.

Stalnaker, Robert. 1999. Context and content: essays on intentionality in speech and thought. Oxford: Oxford University Press. Stalnaker, Robert. 2002. Common ground. Linguistics and Philosophy 25: 701–721. Stowell, Tim. 1989. Subjects, specifiers and X-bar theory. In Alternative conceptions of phrase structure, ed. Mark R. Baltinand Anthony S. Kroch, 232–262. Chicago: University of Chicago Press. Sun, Chaofen. 1995. Transitivity, the Ba construction and its history. Journal of

Chinese Linguistics 23(1): 159-195. Sybesma, Rint. 1999. The Mandarin VP. Dordrecht: Kluwer. Tai, James H.-Y. 1978. Anaphoric restraints in Mandarin Chinese narrative discourse. In Anaphora in discourse, ed. John Hinds, 279–338. Edmonton: Linguistic Research. Tai, James H-Y. 1984. Verbs and times in Chinese: Vendler’s four categories. In David Testen, Veena Mishra & Joseph Drogo (eds.), Papers from the Parasession on

Lexical Semantics of the Chicago Linguistic Society, 289-296. Tsao, Feng-fu. 1987. A topic-comment approach to the construction. Journal of Chinese Linguistics 15(1): 1-54. Van den Berg, Marinus. 1989. Modern standard Chinese: Een Functionele Grammatica, Muiderberg: Coutinho. Vendler, Zeno. 1957. Verbs and times. In Linguistics in Philosophy, 97-121. Ithaca:

Cornell University Press. Verkuyl, H.J. 1993. A theory of aspectuality: The interaction between temporal and atemporal structure. Cambridge: Cambridge University Press. 207

Wang, Chen. 2018. The Syntax of le in Mandarin Chinese. Doctoral Dissertation. Queen Mary University of London. Wang, Huan. 1985. Baziju zhong ba de binyu [The object of the ba-sentence]. Studies of the Chinese Language 1: 48-51. Wang, Li. 1945. Zhongguo Yufa Lilun [A theory of Chinese grammar]. Shanghai: Commercial Press.

Wang, Li. 1959. Zhongguo xiandai yufa [Chinese modern grammar]. Beijing: Zhonghua Book Company. Wang, Li. 2002. Wang Li xuanji [A collection of Wang Li]. Changchun: Northeast Normal University Press. Wang, Mingquan. 1987. Transitivity and the ba-construction in Mandarin. Doctoral dissertation, Boston University. Wechsler, Stephen and Yae-Sheik Lee. 1996. The domain of direct case assignment.

Natural Language and Linguistic Theory 14, 629-664. Wolter, Lynsey Kay. 2004. Demonstratives, definiteness and determined reference. In Proceedings of NELS 34, eds. Keir Moulton and Matthew Wolf. Amherst, MA: GLSA. Woodbury, Hanni. 2003. Onondaga-English/English-Onondaga dictionary. Toronto: University of Toronto Press. Yang, Jun. 2003. Back to the basic: The basic function of particle LE in modern

Chinese. Journal of Chinese Language Teachers Association 38(1): 77-96. Zhu, Dexi. 1982. Yufa jiangyi [On grammar]. Beijing: Shangwu Press. Zucchi, Alessandro and Michael White. 2001. Twigs, Sequences and Temporal Sequences of Predicates. Linguistics and Philosophy 24 (2):223-270.

208