現代漢語時間意義的理論研究(II)

行政院國家科學委員會專題研究計畫 成果報告

現代漢語時間意義的理論研究 2/2

中

華

民

國 92 年 8 月 6 日

現代漢語時間意義的理論研究 2/2 Temporal Reference in Mandarin Chinese

計畫編號：NSC 91-2411-H-009-019 執行期限：91 年 8 月 1 日至 92 年 7 月 31 日 主持人：林若望 國立交通大學外文系 一、中文摘要 本計畫主要研究中文— 一個所謂無構 詞時制的語言— 如何決定句子的時間指 涉。我們就前人研究不足的地方，從新檢 視語料，全面性的擴大討論和時間解釋有 關的句子，而不侷限在傳統上只包含 「著」、「了」、「過」等時態助詞的句 子上，因此除了一般性的簡單句外，也對 複雜句的時間解釋做了非常深入的觀察， 有許多的語料都是文獻上第一次被提出來 討論的。除了發掘新的語言事實外，我也 試圖為漢語的時間解釋提出一套完整的理 論系統來詮釋分析漢語句子的時間解釋， 這套理論主要是以模型理論真值條件語意 學為基礎。我提出不帶任何時間副詞或時 態助詞的簡單句，其時間解釋可透過情狀 類型的選擇限制來預測，也就是，「現在 式」必須選擇（非完整態）同質性情狀當 補語，而「過去式」則是選擇（完整態） 異質性情狀來當補語，時間副詞則可凌駕 選擇限制，由時間副詞所指稱的時間來決 定事件所發生的時間。至於帶有「了」和 「過」的句子，我提出了「了」是實現體 標記，「過」是相對過去時體標記的看法， 並且討論這樣的分析和不同情狀類型之間 的交互影響。除此之外，我也討論了句尾 「了」及詞尾「了」的相同點及相異處， 提出兩者具有共同的核心意義這樣的觀 點，其不同處只在於句尾「了」多了一個 「結果狀態必須與講話時間重疊」的限 制。對於從屬子句的時間指稱，我提出證 據證明賓語從屬子句的時間解釋基本上是 由主要子句動詞的語意選擇限制來決定。 至於關係子句和表時間的副詞子句的時間 解釋，我則證明完全無法由主要子句的事 件時間（也就是，句法上的控制理論）來 預測，而是和主要子句動詞的個別語意， 賓語名詞組的語意解釋，限定詞的有定無 定、語用推論及百科知識等種種因素一起 來決定，我們甚至發現一個（非泛指）關 係子句的時間解釋可以同時包含過去、現 在及未來的時間。同樣地，表示時間的副 詞子句的時間解釋也無法藉由句法上的控 制理論來推測，而必須藉由表時間的從屬 連接詞對兩個不同子句的時間前後限制， 加上語用上的「非瑣碎時間限制」及語用 推論來推論句子的時間解釋。 關鍵詞: 時制、時態、時間指涉 Abstr act

This project discusses how Chinese, a so-called tenseless language, determines its temporal reference. For simplex sentences without time adverb or aspectual marker, I show that temporal reference is correlated with aktionsart or grammatical viewpoint. For sentences with an aspectual marker, I discuss the temporal semantics of

le and guo in detail, showing how their tense/aspectual meanings contribute to

temporal reference. I propose to analyze le as an event realization operator and guo as an anteriority operator. For subordinate clauses, I show that temporal reference of

complement clauses of verbs is basically determined by verbal semantics of individual verbs, which may impose some temporal restriction on the temporal location of the embedded event. As for relative clauses and temporal adverbial clauses, many different factors such as lexical verbal semantics, referential properties of determiners, lifetime effect of noun phrases, semantic or

pragmatics constraints on temporal connectives ,inference rules and world

knowledge, etc., all interact to help determine temporal reference. Many data discussed in this paper indicate that there is no evidence of (covert) tenses in Chinese. Therefore, challenging work remains for those who have claimed that Tense Phrase is projected in Chinese phrase structures.

二、緣由與目的 傳統上研究漢語時制現象的學者，其研 究對象多半侷限在帶有時態助詞如「著」、 「了」、「過」的句型上，雖然這些研究對 於提升時態助詞的瞭解有很多的貢獻，但 是因為語料侷限於簡單句，且多假設漢語 的時間解釋和時態助詞、時間副詞或上下 文有關，所以長久以來，很少有人對於整 個漢語時間解釋的影響因素做全面性的討 論，也沒有人提出一套較完整的形式理論 來對漢語的時間解釋作系統性的詮釋，有 鑑於此，我在這個計畫裡，就前人研究不 足的地方，從新檢視語料，並試圖為漢語 的時間解釋提出一套完整的理論系統來詮 釋分析漢語句子的時間解釋。 三、結果與討論 在此計畫裡，我為漢語的時間解釋提出 一套完整的理論系統來詮釋分析漢語句子 的時間解釋，這套理論主要是以模型理論 真值條件語意學為基礎，採用這套理論的 主要原因是西方語言學者研究時制時，很 多都是以這種理論為基礎，因此這套理論 比較有利於日後作對比分析及普遍語法的 研究。我提出不帶任何時間副詞或時態助 詞的簡單句，其時間解釋可透過情狀類型 的選擇限制來預測，也就是，「現在式」必 須選擇（非完整態）同質性情狀當補語， 而「過去式」則是選擇（完整態）異質性 情狀來當補語，時間副詞則可凌駕選擇限 制，由時間副詞所指稱的時間來決定事件 所發生的時間。我們的理論不僅成功地說 明了簡單句的時制指涉，也解釋了帶「得」 字補語的時制意義。至於帶有「了」和「過」 的句子，我提出了「了」是實現體標記，「過」 是相對過去時體標記的看法。我對於「了」 和「過」的分析雖然不是全新的概念，但 是經過利用形式語意學的方法來重新定義 實現體的意義之後，「實現」的意義不僅有 了具體內涵，而且也解決了一個從來不曾 被成功地解釋過的現象，也就是「了」的 「完整態」如「我買了一本書」與「非完 整態」如「我養了一條金魚」的衝突。有 共同的核心意義這樣的觀點，其不同處只 在於句尾「了」多了一個「結果狀態必須 與講話時間重疊」的限制。至於從屬子句 的時間指稱，我提出證據證明賓語從屬子 句的時間解釋基本上是由主要子句動詞的 語意選擇限制來決定，某些動詞如「決定」 要求補語子句的事件時間在主要子句的事 件時間之後，有些動詞如「後悔」則要求 補語子句的事件時間在主要子句的事件時 間之前，還有些動詞如「喜歡」則要求補

語子句的事件時間與主要子句的事件時間 重疊，另外像「認為」這類動詞則對補語 子句的事件時間不做強制性之規範。至於 關係子句和表時間的副詞子句的時間解 釋，我則證明和主要子句動詞的個別語 意，賓語名詞組的語意解釋，限定詞的有 定無定、語用推論及百科知識等種種因素 一起來決定。我們甚至發現一個（非泛指） 關係子句的時間解釋可以同時包含過去、 現在及未來的時間，這些結果暗示漢語可 能沒有句法上時制節點的存在，因而對於 主張漢語應該有隱形時制這樣的論點構成 極大的挑戰。同樣地，表示時間的副詞子 句的時間解釋也無法藉由句法上的控制理 論來推測，而必須藉由表時間的從屬連接 詞對兩個不同子句的時間前後限制，加上 語用上的「非瑣碎時間限制」及語用推論 來推論句子的時間解釋。 四、 計畫成果自評 總而言之，我們的研究結果不僅在語言 事實上有新的發現，在理論分析上也跳出 前人假設的框框，深入討論許多前人不曾 討論過，卻對漢語時間解釋有非常大影響 的因素，因而提升了我們對於漢語時間解 釋系統的全盤性瞭解，這對於日後研究漢 語時制和時態的學者不僅有相當大的啟發 作用，對於有興趣作不同語言的對比分析 研究或是普遍語法研究的學者，也提供了 非 常 有 用 的 比 較 基 礎 ， 我 們 的 研 究 成 果 ”Temporal Reference in Mandarin Chinese”也已經發表在 2003 年國際期刊 Journal of East Asian Linguistics 12: 259-311.。

五、參考文獻

1. Binnick, Robert I. (1991) Time and the Verb: A Guide to

Tense and Aspect, Oxford University Press, New York. 2. Bohnemeyer, Jürgen and Mary Swift (2001) “Default Aspect: The Semantic Interaction of Aspectual Viewpoint and Telicity,” Proceedings of Perspectives on Aspect, Utrecht Institute of Linguistics, 2001.

3. Amherst. Chan, Marjorie (1980) “Temporal Reference in Mandarin Chinese: An Analytic-Semantic Approach to the Study of the Morphemes le, zai, zhe and ne,” Journal of the Chinese Language Teachers’ Association 15, pp. 33–79. 4. N. Carlson and Francis Jeffry Pelletier (eds.), The Generic Book, University of Chicago Press, Chicago, pp. 176–223. 5. Comrie, Bernard (1976) Aspect: An Introduction to the Study of Verbal Aspect and Related Problems, Cambridge University Press, Cambridge.

6. Comrie, Bernard (1985) Tense. Cambridge University Press, Cambridge.

7. Dai, Yaojing (1994) “Le Zai Biaoshi Weilai Yiyi Juzi Zhong De Yongfa [Usage of Le in Sentences Expressing Future Meaning],” in Zhihong Yu (ed.), Xiandai Yuyanxue: Lilun Jianshe De Xin Shikao, Yuwen Chubanshe, Beijing, pp. 115–122.

8. Dowty, David (1979) Word Meaning and Montague Grammar: The Semantics of Verbs and Times in Generative Semantics and in Montague’s PTQ, D. Reidle, Dordrecht. 9. Erbaugh M. and C. S. Smith (2001) “Temporal Information in Sentences of Mandarin,” Manuscript, University of Texas and City University of Hong Kong.

10. Grice, H. P. (1975) “Logic and Conversation,” in P. Cole and J. L. Morgan (eds.), Syntax and Semantics, 3: Speech Acts, Academic Press, New York, pp. 41–58.

11. Heim, Irene (1982) The Semantics of Definite and Indefinite Noun Phrases, PhD dissertation, University of Massachusetts, Amherst.

12. Huang, L. Meei-jin (1987) Aspect: A General System and its Manifestation in Mandarin Chinese, PhD dissertation, Rice University.

13. Jin, Lixin (2002) “Ciwei ‘Le’ De Shiti Yiyi Ji Qi Jufa 14. Kamp, H. and Uwe Reyle (1993) From Discourse to Logic: Introduction to Model Theoretic Semantics of Natural

Language, Formal Logic and Discourse Representation Theory, Kluwer Academic Publishers, Dordrecht.

15. Kang, Jian (1999) The Composition of the Perfective Aspect in Mandarin Chinese, PhD dissertation, Boston University, Boston.

15. Klein, Wolfgang (1994) Time in Language, Routledge, London.

16. Klein, Wolfgang, Li Ping and Henriette Hendriks (2000) “Aspect and Assertion in Mandarin Chinese,” Natural Language and Linguistic Theory 18, pp. 723–770.

17. Kong, Linda (1986) “Guanyu Dongtai Zhuci ‘Guo1’ Han ‘Guo2’ [On Dynamic ParticlesGuo1 and Guo2],” Zhongguo

Yuwen 4, pp. 272–280.

18. Kratzer, Angelika (1994) The Event Argument and the Semantics of Voice, Manuscript, University of Massachusetts, Amherst.

19. Kratzer, Angelika (1998) “More Structural analogies Between Pronouns and Tenses,” SALT 8, Proceedings From the 8th Conference of Semantics and Linguistic Theory, MIT, 1998, pp. 92–110.

20. Krifka, Manfred (1989) “Nominal Reference, Temporal Constitution and Quantification in Event Semantics,” in R. Bartsch, J. van Benthem, and P. van Emde Boas (eds.),

Semantics and Contextual Expression, Foris, Dordrecht, pp. 75–115.

21. Kusumoto, Kiyomi (1999) Tense in Embedded Contexts, PhD dissertation, University of Massachusetts, Amherst. Li, Charles N. and Sandra A. Thompson (1981) Mandarin Chinese: A Functional Reference Grammar, University of California Press, Berkeley.

22. Li, Charles N., Sandra A. Thompson and R. M. Thompson (1982) “The Discourse Motivation for the Perfect Aspect: the Mandarin Particle LE”, in Paul Hopper (ed.), Tense and Aspect: Between Semantics and Pragmatics, John Benjamins Publishing Company, Amsterdam.

23. Li, Tiegen (1999) The Study of Tense in Modern Chinese, Liaoling University Press, Shenyiang.

24. Lin, Jo-wang (2000) “On the Temporal Meaning of the Verbal -le in Chinese,” Language and Linguistics 1(2), pp.

109–133.

25. Lin, Jo-wang (2002) “Selectional Restrictions and Temporal Reference of Chinese Bare Sentences,” Lingua 113, pp. 271–302.

26. Liu, Yuehua (1988) “Dongtai Zhuci ‘guo1’, ‘guo2’, ‘le1’ Yongfa Bijiao [A Compariative Study of the Dynamic Particles Guo1, Guo2 and Le1],” Yuwen Yanjiu 1, pp. 6–16.

27. Liu, Xunning (1988) “Xiandai Hanyu Ciwei ‘le’ DE Yufa Yiyi,” Zongguo Yuwen 5, pp. 321–330.

28. Liu, Xiaomei (1997) Guo Ming Ke Yu De Dongtai Wenfa Tixi Ji Dongtaici De Shangjia Dongmao Yuyi [The Grammatical System of Mood in Mandarin Chinese, Southern Min

Chinese and Haka Chinese and the Aspectual Meanings of Aspectual Markers], Crane Publishing, Taipei.

29. Mangione, L. and Dingxuan Li (1993) “A Compositional Analysis of -Guo and -Le”, Journal of Chinese Linguistics, pp. 65–122.

30. Mochizuki, Keiko (2000) “Hanyu Li De Wancheng Ti [The Perfect Aspect in Chinese],”

Hanyu Xuexi 1, pp. 12–16.

32. Ogihara, Toshiyuki (1989) Temporal Reference in English and Japanese, PhD dissertation, University of Texas, Austin. 33. Ogihara, Toshiyuki (1996) Tense, Attitudes and Scope, Kluwer Academic Publishers, Dordrecht.

34. Partee, Barbara H. (1984) “Nominal and Temporal Anaphora,” Linguistics and Philosophy7, pp. 243–286. 35. Ross, Claudia (1995) “Temporal and Aspectual Reference in Mandarin Chinese,” Journal of Chinese Linguistics 23, pp. 87–135.

36. Shi, Zhiqiang (1990) “Decomposition of Perfectivity and Inchoativity and the Meaning of the Particle Le in Mandarin Chinese,” Journal of Chinese Linguistics 18, pp. 95–124. 37. Stowell, Tim (1993) “Syntax of Tense,” Manuscript, University of California, Los Angeles.

38. de Swart, Hënriette (1998) “Aspect Shift and Coercion,”

Natural Language and Linguistic Theory 16, pp. 347–385. 39. Teng, Shou-hsin (1975) A Semantic Study of Transitivity Relations in Chinese, University of California Press, Berkeley. 40. von Stechow, Arnim (2001) “Temporally Opaque

Arguments in Verbs of Creation,” in Carlo Cecchetto, Gennaro Chierchia, and Maria Tersa Guasti (eds.), Semantic Interfaces, CSLI Publications, Stanford, pp. 278–319.

41. Wang, William S. Y. (1965) “Two Aspect Markers in Mandarin,” Language 41, pp. 457–470.

42. Yeh, Meng (1996) “An Analysis of the Experiential GUOEXP in Mandarin Chinese: A Temporal Quantifier,” Journal of East Asian Linguistics 5, pp. 151–182.

Temporal Reference in Mandarin Chinese*

This paper discusses how Chinese, a so-called tenseless language, determines its

temporal reference. For simplex sentences without any time adverb or aspectual

marker, I show that temporal reference is correlated with aktionsart or grammatical

viewpoint. For sentences with an aspectual marker, I discuss the temporal semantics

of le and guo in details, showing how their tense/aspectual meanings contribute to

temporal reference. I propose to analyze le as an event realization operator and guo as

an anteriority operator. For subordinate clauses, I show that temporal reference of

complement clauses of verbs is basically determined by verbal semantics of

individual verbs, which may impose some temporal restriction on the temporal

location of the embedded event. As for relative clauses and temporal adverbial clauses,

many different factors such as lexical verbal semantics, referential properties of

determiners, lifetime effect of noun phrases, semantic or pragmatics constraints on

temporal connectives, inference rules and world knowledge, etc., all interact to help

determine temporal reference. Many data discussed in this paper indicate that there is

no evidence of (covert) tenses in Chinese. Therefore, it remains a challenging work

for those who have claimed that Tense Phrase is projected in Chinese phrase

structures.

1. Introduction

The study of temporal reference in natural language has been one of the most

important issues in the history of linguistic research. This is especially the case for

Indo-European languages such as English, because distinctions of times in these

languages have been studied for almost twenty-five hundred years since at least the

time of the ancient Greeks and the results are very fruitful as Binnick’s (1991) book

Time and the Verb has shown us. In contrast to Indo-European languages, works on

temporal reference in Chinese are relatively meager and the breadth and depth of

research are far behind those of Indo-European languages. One reason for this,

undoubtedly, is that the Chinese language, unlike Indo-European languages, does not

have the same kind of verbal inflections to indicate distinctions of times. Of course,

not having finite verb forms does not mean that Chinese is not able to express the

notion of time. When hearing a Chinese sentence, any native speaker can immediately

tell whether the situation described by that sentence holds at a past time, a future time

or the speech time. Interesting questions then arise as to how temporal reference of

Chinese sentences is determined and to what extent the mechanisms that the Chinese

language uses are different from those used in Indo-European languages. In this paper,

I will not be able to probe into the second question but I will attempt to give an

answer to the first question in some details based on a wide range of data.

This paper is organized as follows. Section 2 outlines the basic assumptions and

the theoretical framework that I adopt. Section 3 is devoted to an analysis of how bare

sentences in Chinese obtain their temporal reference. Section 4 investigates how

different aspectual markers such as the perfective marker le and the experiential

marker guo affect temporal reference. Section 5 to section 8 discusses temporal

reference of subordinate clauses such as complement clauses of verbs, relative clauses

and adverbial clauses. Section 9 concludes this article.

2. Basic Assumptions and Theoretical Fr amewor k

and aspect that I will be adopting for a better understanding of the discussion that will

follow. Traditionally, tenses are understood as the location of an event or state in time

axis relative to a reference time, which is usually the speech time (Comrie (1985)).

When an event or state takes place or holds before the speech time, the tense is past

tense; when the situation is reversed, the tense is future tense; when a process or state

overlaps with the speech time, the tense is present tense. As for the notion of aspect, it

is often characterized as different ways of presenting a situation as a completed whole,

viewed as if from outside, or as an ongoing, incomplete action or state, viewed as if

from inside (Comrie (1976)). The former is called perfective aspect and the latter

imperfective aspect. The perfective vs. imperfective distinction is often realized

through grammaticalized affixes or auxiliaries. Klein (1994) finds the traditional

definition of aspect imprecise. Therefore, he proposes to replace the definitions of

tense and aspect with temporal relations. He has distinguished three times: the time of

utterance (TU), the time span at which a situation obtains (T-SIT or time of situation)

and the time span about which an assertion is made (TT or topic time). On his

analysis, tense does not express a temporal relation between TU and T-SIT as in the

classical analysis, but one between TT and TU. Aspect, on the other hand, expresses a

temporal relation between TT and T-SIT. In particular, perfective aspect requires that

the situation time is included within the topic time, whereas imperfective aspect is the

other way around or involves an overlap relation.i This paper accepts Klein’s (1994)

distinction of tense and aspect and will recast his notions of tense and aspect within a

framework of model-theoretic semantics. Although my theoretical framework will be

model-theoretic semantics, I will keep the formal mechanisms as few as possible and

plain English will be provided to explain what the intuitive idea is behind the logical

language. Therefore, in most cases, the reader can actually understand the discussion

Now some theoretical assumptions are in order. In this paper, I assume that verbs,

stative or non-stative, have an event argument and that sentence meanings are

properties of eventualities, i.e., λeφ, where φ contains a free occurrence of the eventuality variable e. With the introduction of event arguments to the argument

structures of verbs, I need to assume that in addition to the normal semantic types e

(entity) and t (truth-value), there is a semantic type s, standing for situations, events or

states. In addition, I also assume another semantic type i standing for intervals.

For the syntax, I assume that above VP is an aspectual phrase AspP. The

perfective vs. imperfective distinction is stated at the head of AspP. Above AspP is

AgrsP. Tense Phrase (TP) is located above AgrsP. (The relative order between AgrsP

and TP is not important.) Klein’s topic time occupies the specifier position of TP. In

addition, I adopt the VP-internal subject hypothesis as proposed in Kitagawa (1986)

and Koopman and Sportiche (1991), though this is not crucial.

It is worth noting that the TP projection is assumed for Chinese phrase structures

only as a working hypothesis for comparison. I do not commit myself to the claim that

tenses and TP exist in Chinese, because the information provided by AspP and topic

time seems generally sufficient to explain temporal locations of eventualities denoted

by Chinese sentences. I will bring up this issue from time to time later when I discuss

the Chinese data.

With the above assumptions, I now formalize Klein’s analysis of tense and aspect

as in (1)-(2) (cf. Kratzer (1998); Bohnemeyer and Swift (2001)), where t2 stands for

the topic time and the symbol τ denotes Krifka’s (1989) temporal trace function, a partial function which when applied to an eventuality yields its “run time”. Thus, τ(e) is equivalent to the situation time of the eventuality e. Finally, s* stands for the speech

(1) a. perfective aspect =: λP<s,t>λt2λe[τ(e)⊆ t2∧ P(e)]

b. Imperfective Aspect =:λP<s,t>λt2λe[t2 ⊆τ(e)∧ P(e)]

(2) a. [+ present] =: λP<i,<s,t>>λt2λe[P(t2)(e) ∧ s* ⊆ t2 ]

b.[+ past] =: λP<i,<s,t>>λt2λe[P(t2)(e) ∧ t2 < s*]

To take the English sentence (3) as an illustration, let us see how the above

definitions work. Within Klein’s framework, the meaning of (3) is paraphrased as

follows: The situation time of the proposition J ohn wor ked is included within the

topic time zuotian ‘yesterday’ because of the perfective aspect and this topic time

must precede the utterance time because of the past tense. Therefore, the situation

time of John’s working must be in the past. The temporal meaning of (3) is formally

computed as in (4).

(3) [CP [TP yesterday [T’ T[+past] [AgrsP Johnx [AspP Asp[+perfective][VP x worked]]]]]]

(4) [[VP]] = λework’(x)(e)

[[AspP]] = λP<s,t>λt2λe[τ(e)⊆ t2∧ P(e)] (λework’(x)(e))

= λt2λe [τ(e)⊆ t2∧ work’(x)(e)]

[[AgrsP]] =λxλt2λe[τ(e)⊆ t2∧ work’(x)(e)] (John’)ii

= λt2λe[τ(e)⊆ t2∧ work’(John’)(e)]

[[T’]] = λP<i,<s,t>>λt2λe[P(t2)(e) ∧ t2 < s*](λeλt2[τ(e)⊆ t2∧ work’(John’)(e)])

= λt2λe[τ(e)⊆ t2∧ work’(John’)(e) ∧ t2 < s*]

[[TP]] = λe[τ(e)⊆yesterday ∧ work’(John’)(e) ∧ yesterday < s*]

[[CP]] = ∃e[τ(e)⊆yesterday ∧ work’(John’)(e) ∧ yesterday < s*] (Default Existential Closure)

An assumption commonly held by Chinese linguists is that Chinese conveys temporal

locations of eventualities via temporal adverbs such as zuotian ‘yesterday’, ming nian

‘next year’, aspectual markers such as le and guo or some previous sentences which

set up a time frame for the discourse. However, if we carefully look at the data, we

will find that a large number of Chinese sentences do not contain a time adverb or

aspectual marker. Nor is it necessary to resort to previous utterances to determine their

temporal reference. For example, independently of any context, the sentences in (5)

and (6), which do not contain any time adverb or aspectual marker, can be easily

construed as referring to past and present situations, respectively.

(5) a. Ta dapuo yi ge hua ping

he break one Cl flower vase

‘He broke a flower vase.’

b. Ta ba wo gang-chu jiaoshi

he Ba me drive-out classroom

‘He drove me out of the classroom.’

c. Ta zai Shanghai chu-sheng

he in Shanghai give-birth

‘He was born in Shanghai.’

d. Ta qiangpuo wo xiu ta-de ke

he force me take his class

‘He forced me to take his class.’

e. Didi bang wo xiang-dao yi-ge hen hao de fangfa

brother help me think-of one-Cl very good De method

(6) a. Ta hen congming

he very clever

‘He is very clever.’

b. Wo xiangxin ni

I believe you

‘I believe you.’

c. Diqiu rao taiyang xuanzhuan

earth turn sun around

‘The earth turns around the sun.’

d. Ni da lanqiu ma?

you play basketball Q

‘Do you play basketball?’

e. Ta zai fangjian du shu

he in room study book

‘He is studying in his room.’

If we assume that Chinese has tenses, then the tenses in those examples in (5)-(6)

must be covert tenses. The problem is then to determine the value of those covert

tenses. The suggestion that I would like to make is this. Following de Swart (1998)

and Schmitt (2001), I assume that tenses are subject to selectional restrictions. Thus, a

certain tense can only select a complement with a specific aspectual viewpoint or

aktionsart. On these assumptions, the values of covert tenses in Chinese, if they exist,

can be determined by the following selectional restrictions.

(7) a. Covert present tense must select imperfective AspP as its complement

In order for the above selectional restriction to work, I will rely on Bohnemeyer and

Swift’s (2001) cross-linguistic study about “default aspect” in natural language.iv In

their paper, they have argued that there is a certain correlation between the telicity of

an eventuality description and its aspectual viewpoint. Briefly, a predicate is telic if it

denotes only events that have no part that falls under the same predicate. A predicate

is atelic if the events it denotes have at least one non-final part that falls under the

same predicate. On this definition, eat a fish is a telic predicate, whereas walk on the beach is atelic. According to them, cross-linguistically the default aspectual viewpoint

of telic descriptions is perfective viewpoint, whereas the default aspectual viewpoint

of atelic descriptions is imperfective viewpoint. Moreover, such a correlation may

manifest itself through morphological markedness relations. For instance, there are

languages such as Russian and Yukatek Maya, which have marked imperfective

aspect and unmarked perfective aspect for telic predicates. Chinese can be added to

this category. For telic predicates such as chi yi-tiao yu ‘eat a fish’, they are

interpreted perfectively. But if they are combined with zai such as zai chi yi-tiao yu

‘be eating a fish’, they are interpreted imperfectively. Bohnemeyer and Swift (2001)

have proposed an account for the correlation between (a)telicity and aspectual

viewpoint in terms of the notion of “default aspect”, which is perfective for a telic

predicate and imperfective for an atelic predicate. I will assume this notion of default

aspect without further discussing their formal definitions.

Given Bohnemeyer and Swift’s notion of default aspect, I am now able to explain

why the sentences in (5) have a past interpretation and those in (6) have a present

interpretation: the former all describe perfective telic situations, whereas the latter all

denote imperfective atelic situations. Thus, according to the selectional restrictions

covert tenses in examples like those in (6) must be present tense. The semantic

computation of such sentences is straightforward, so I omit the details.

It should be noted that bare sentences contrast with sentences containing a time

adverb. For the latter kind of sentence, the time adverb determines temporal reference.

For example, in (8), though all the sentences contain the same homogeneous predicate

hen mang ‘very busy’, temporal reference of the sentence varies with the time adverb.

(8) a. Ta zuotian hen mang

he yesterday very busy

‘He was very busy yesterday.’

b. Ta xianzai hen mang

he now very busy

‘He is very busy now.’

c. Wo mingtian hen mang

I tomorrow very busy

‘I will be very busy tomorrow.’

In order to capture the fact that time adverbs override tense selections stated in (7), I

propose that the tense node, if it exists, must agree with the overt time adverb in the

specifier position o f TP (cf. Lin (2002), Erbaugh and Smith (2001)).

Before moving to next section, two remarks are in order. One is that except in few

constructions such as conditionals or imperatives, future tense in Chinese cannot be

an empty tense. That is why we do not have a selectional restriction for covert future

tense in (7). Future time in Chinese must be expressed by an overt expression

indicating future time such as the future time adverb mingtian ‘tomorrow’ or the

‘If he comes, I’ll go’ may express a future time without an overt expression indicating

a future time, it has been argued that such constructions contain an implicit modal in

the matrix clause (Heim (1982); von Fintel (1994), among many others.) If this is

correct, temporal reference of conditionals is determined by an implicit modal

equivalent to hui ‘will’ in force. As for imperatives, such constructions universally

refer to future actions. This, I believe, should be ascribed to the special semantics or

pragmatics of imperatives, which I will not discuss here.

The other remark has to do with the question of whether or not Chinese has

(covert) tenses. In the above discussion, I have assumed that (covert) tenses exist in

Chinese and resort to selectional restrictions to interpret their values in bare sentences.

Can we explain the same facts without assuming existence of (covert) tenses? The

answer seems to be positive. We only need to fill in the value of the topic time

introduced by Asp, which specifies a relation between event time and topic time. If a

sentence does not contain an overt time adverb, the topic time is generally some time

interval determined by the context such as the speech time. However, for a non-future

perfective durative sentence, the topic time must be a past interval rather than the

speech time, because a durative event cannot be included within the speech time.

Similarly, for a non-future perfective instantaneous achievement, the topic time must

also be in the past because the event denoted by an achievement must have already

been completed before one is able to talk about that situation. These are independent

constraints independent of theories of tenses. As for imperfective sentences, the topic

time is the default speech time unless some time adverb appears in the sentence. I

conclude that bare sentences are no evidence for the projection of TP because with or

without covert tenses, one can equally predict the temporal locations of eventualities

4. Aspectual Mar ker s and Chinese Tempor al Reference

Having discussed how temporal reference of bare sentences is determined, I now

turn to sentences with an aspectual marker such as le and guo, discussing how they

contribute to temporal reference of sentences.

The literature on Chinese le and guo is so huge that it is impossible to give even a

brief overview here, due to restrictions of space (Kong (1986); Huang (1987);

Xunning Liu (1988); Yuehua Liu (1988); Shi (1990); Magione and Li (1993); Dai

(1994); Ross (1995); Yeh (1996); Liu (1997); Li (1999); Kang (1999); Lin (2000b);

Klein, Li and Hendrik (2000); to mention just a few). So in this paper, I will focus

more on my own view of these markers, leaving the comparison to the reader.

The verbal suffix le has often been characterized as a perfective marker indicating

completion or termination of an action or inchoativity of a state. To illustrate, consider

(9), which clearly describes a past event.

(9) Ta chi-le yi tiao yu

he eat-Asp one-Cl fish

‘He ate a fish.’

According to Magione and Li (1993), sentences like (9) do not describe just any past

events but past events that occur within a certain reference time. Although sentences

with le usually have a past interpretation indicating completion or termination of an

action, le is actually compatible with a present continuative interpretation (cf.

Xunning Liu (1988); Lin (2000b); Jin (2002)). Consider the following examples.

he raise-Asp one-Cl goldfish

‘He is raising a goldfish.’

b. Wo (zai Boston) zu-le yi-jian gongyu

I in Boston rent-Asp one Cl apartment

‘I am renting an apartment in Boston.’

c. Zhangsan (shou-li) ti-le yi-ge da pixiang

Zhangsan hand-in carry-Asp one-Cl big suitcase

‘Zhangsan is carrying a big suitcase (in his hand).’

d. Ni kan, Lisi qi-le yi-pi ma wang cheng wai fangxiang zou,

you look Lisi ride-Asp one-Cl horse toward town outside direction walk

bu zhidao ta yao qu nar

not know he want go where

‘Look! Lisi is riding a horse toward the direction of the outside of town. I

wonder where he wants to go.’

The syntactic constructions of (10a-c) are identical to the construction of (9), but their

temporal meanings are quite different. Unlike (9), which refers to a past completed

event, (10a-c) do not describe completed or terminated events but present on-going

situations. Although these sentences are not progressive sentences, they are translated

as such to indicate that the event has begun before the speech time and is still

on-going.v An important property distinguishing the sentences in (10) from (9) seems

to be this. When a sentence of the type in (10) is true of an interval, every subinterval

of that interval or a non-final subinterval of that interval can make the same sentence

true. For example, if John rents an apartment from April to August in 2002, then it is

also true that he rents an apartment in May or in June. But if John eats a fish from

This property is known as the subinterval property of atelic predicates (Dowty

(1979)).vi Interestingly and mysteriously, however, not every atelic predicate is

compatible with the verbal le. In fact, many activity predicates are incompatible with le. This empirical fact can be clearly illustrated with the contrast between (11a) and

(11b).

(11) a. *Zhangsan kan-le yi-zhi niao

Zhangsan watch-Asp one Cl bird

‘Zhangsan is watching a bird./Zhangsan watched a bird.’

b. Zhangsan kan-le yi-bu dianying

Zhangsan watch-Asp one-Cl movie

‘Zhangsan watched a movie.’

Although both (11a) and (11b) use the same verb kan ‘watch’, their aktionsart

depends upon the object NP. When the object NP is yi-zhi niao ‘a bird’, the VP is an

atelic activity predicate; when the object NP is yi-bu dianying ‘a movie’, the VP is a

telic accomplishment predicate. However, (11a) is ill-formed regardless of what

interpretation is assigned to it. But (11b) is perfect with a past interpretation. It is not

clear to me what property distinguishes those atelic sentences which are compatible

with le and those which are not and I will not try to provide a solution to this problem.

My main concern will be on the question of how those sentences in (10) obtain a

present continuative interpretation.

Like those non-stative verbs in (10), stative verbs may also sometimes take the

verbal le, giving rise to a present continuative reading.

you have-Asp wife then not want parents

‘You have had a wife. So you don’t want your parents.’

b. Ta yijing zhidao-le na-jian shi

he already know-Asp that-Cl matter

‘He has already known that matter.’

c. Jingguo ta-de jiangjie, wo duoshao liaojie-le yi dian GB lilun

After his explanation I more-or-less understand a little GB theory

‘After his explanation, I more or less understand a little bit of GB theory.’

The examples in (10) and (12) clearly indicate that the verbal suffix le is not an

absolute past tense marker. This is further supported by the fact that le may also

appear in a clause with a future interpretation as in (13).

(13) Deng ni bi-le-ye yihou, wo hui mai yi-bu che gei ni

Wait you graduate-Asp after I will buy one-Cl car for you

‘After you have graduated, I will buy a car for you.’

In (13), the meaning of le seems to indicate anteriority of the embedded clause to the

matrix clause (Dai (1994); Lin (2000b)).

In contrast, if the markerle in (9) and (10) is replaced with the experiential marker guo, the sentences are all unambiguously interpreted as terminated past events. This is

illustrated by (14a) and (14b).

(14) a. Ta chi-guo yi-tiao jinyu

he eat-Asp one-Cl goldfish

b. Ta yang-guo yi-tiao jinyu

he raise-Asp one Cl goldfish

‘He raised a goldfish (before).’

c. Wo (zai Boston) zu-guo yi-jian gongyu

I in Boston rent-Asp one-Cl apartment

‘I rented an apartment in Boston (before).’

Thus, guo is more like a past tense marker than le is. Nevertheless, like le, guo cannot

be an absolute past tense marker, either, because it is also compatible with a future

interpretation when it appears in an embedded clause. This is illustrated by (15).

(15) Deng ni ting-guo ta tan gangqin yihou, ni jiu hui zhidao ta de jiqiao you

wait you hear-Asp he play piano after you then will know he De skill have

duo hao

how good

‘After you have heard him play the piano, you will know how good his skill is.’

Although guo in (15) is not construed as an absolute past tense marker, it still

expresses relative anteriority; namely, the event denoted by the subordinate clause

containing guo must precede the event denoted by the matrix clause.

4.1A Temporal Semantics for Le

As we saw above, temporal interpretation of sentences containing the verbal le is

sensitive to the aktionsart of VP. When the VP is of a type such as chi-yi-tiao-yu ‘eat a

event. When the VP is of a type such as zu yi-jian gongyu ‘rent an apartment’ or yang yi-tiao yu ‘raise a fish’, i.e., atelic predicates, the sentence obtains a present

continuative interpretation. This situation raises a very interesting question. When a

sentence with le refers to a past event, the aspectual viewpoint is certainly perfective.

However, when it has a present continuative interpretation, the aspectual viewpoint

seems to be imperfective., because the situation is incomplete. As a consequence, we

have a dilemma with respect to the interpretation of le, which is sometimes interpreted

perfectively and sometimes imperfectively. In this section, I will propose an analysis

that explains the perfective-imperfective dilemma brought about by le.

Xunning Liu (1988) is the first to suggest that the verbal le is better treated as a

“realization aspect” instead of a completive marker. I agree with his informal idea that

the verbal le indicates that an eventuality is realized and believe that once formalized,

the realization analysis of le can provide a very neat account for the seemingly

contradictory meanings of le, i.e., the perfective-imperfective paradox. I will also

extend the realization analysis to account for sequence of tense in embedded contexts.

The analysis that I will be proposing is based upon the concept of ‘event

realization’ defined by Bohnemeyer and Swift (2001). As mentioned earlier,

Bohnemeyer and Swift have tried to define a “notional aspect operator” that may

derive the fact that the default aspectual viewpoint of a telic predicate is perfective,

whereas that of an atelic predicate is imperfective. In order to achieve this goal, they

define a concept of event realization as given in (16) and utilize it in their definition of

‘notional aspect operator’, which need not concern us here.

(16) ∀P,e,t⊆E[REALE(P,e,t) ↔ P(e) ∧ ∃e’[P(e’) ∧ e’≤E e ∧ tSIT(e’)⊆t]]

that for an event e denoted by P to be realized at a (topic) time interval t, t must

contain the run time tSIT of a part e’ of e such that e’ is also a P. This definition of

event realization has a very interesting consequence for telic and atelic eventualities.

That is, for a telic event to be realized, we need a perfective viewpoint whereas an

atelic eventuality only requires an imperfective viewpoint to entail event realization.

In other words, a telic eventuality is realized only when the eventuality culminates but

an atelic eventuality can be realized as long as a subpart of it holds. Although

Bohnemeyer and Swift’s concern of their paper is not about Chinese le, I will employ

their definition of event realization to account for the meaning of le.

With Bohnemeyer and Swift’s concept of event realization in mind, I would like

to suggest that the verbal le in Chinese is just an event realization operator (cf.

Xunning Liu (1988)), whose definition is given in (17). In plain English, (17) says

that when the meaning of le is applied to a property of events, i.e., sentence meaning,

there must exist an event e denoted by P and a subpart e’ of e that also falls under P is

contained within the topic time t2.

(17) A preliminary version of the meaning of le

[[le]] = λP<s,t>λt2λe ∃e’[ P(e) ∧ P(e’) ∧ e’≤E e ∧ τ(e’)⊆t2]

For example, applying this meaning of le to (10b), we get the following truth

conditions: There is an event e of Ta zu yi-jian gongyu ‘he rents an apartment’ and

this event has a subpart that is included within the topic time. Since (10b) does not

have an overt time adverb, the default topic time is now. This amounts to saying that a

subpart of the eventuality, which is also an eventuality of Ta zu yi-jian gongyu ‘he

rents an apartment’, is included within the time denoted by now. This then entails a

adverbial such as qu nian ‘last year’ is added to the sentence as in (18) below, it is

possible to get a past reading.

(18) Qu nian ta zu-le yi-jian gongyu

last year he rent-Asp one Cl apartment

‘He rented an apartment last year.’

Notice that (18) does not assert that the whole event must be included within the topic

time denoted by qu nian ‘last year’. It only claims that a subpart of it must be

contained within it. Therefore (18) is compatible with a situation where the renting

event is still on-going this year. This is confirmed by the fact that (18) can be

followed by (19).

(19) Bu zhidao jin nian ta shi-bu-shi hai zhu zai na jian gongyu li

not know this year he be-not-be still live in that CL apartment in

‘I wonder if he still lives in that apartment this year.’

However, if the subpart of event that is contained within the topic time happens to be

the whole event itself, we do get a reading according to which the whole renting event

is terminated.

The same remarks and analysis applies to stative sentences with le such as those

examples in (12). So I will not repeat the details.

In my above exposition of the meaning of le, I intentionally left out discussion of

tenses. Now let me make some remarks on this matter. Suppose that like other

perfective or imperfective markers, a realization operator is an aspectual operator that

location of situations described by sentences with le can be determined without

reference to a tense node. In other words, the TP projection is not necessary for

Chinese at least for sentences with the verbal le is concerned.

But what if Chinese does have a tense node? If this is the case, one can assume

that le is a viewpoint-neutral realization operator. The actual aspectual viewpoint of

AspP headed by le is determined by the aktionsart of VP. As Bohnemeyer and Swift

(2001) have argued, “atelic predicates require merely imperfective aspect for the

entailment of realization, whereas telic predicates require minimally perfective aspect

to entail realization”. Thus, when le occurs with an atelic predicate, the aspectual

viewpoint is imperfective; when it occurs with a telic predicate, the aspectual

viewpoint is perfective. It follows from this that when le occurs with an atelic

predicate and no overt time adverb appears, the covert tense must be present tense,

which in turn forces the topic time to be now. However, if an overt time adverb

appears as in (18), the tense must agree with the time adverb as discussed earlier. This

analysis also entails that combination of le with a telic predicate as in (9) yields a past

reading. As mentioned, no subpart of a telic eventuality is the same telic eventuality

unless the subpart is the whole eventuality itself. The meaning of le in (17) thus

entails that e’ is e when e is a telic eventuality and e must be included within the topic

time. This is equivalent to claiming that AspP has a perfective viewpoint. However,

the topic time cannot be now, because a perfective viewpoint is associated with a null

past tense by default. So, the topic time for a telic eventuality can only be some time

interval in the past. Consequently, (9) can only have a past reading with the event

described falling within a past time interval.

So far, the proposed analysis of the verbal le has produced a very good result for

simplex sentences. What about those occurrences of le in subordinate clauses? Can

slight revision seems necessary. As a first step, consider the following sentences:

(20) Zhangsan shuo ta chi-le yi tiao yu

Zhangsan say he eat-Asp one Cl fish

‘Zhangsan said that he ate a fish.’

(21) Xiaozhang hui banfa jiangzhuang gei naxie xiangchu-le daan de ren

principal will give testimonial to those figure-out-Asp answer people

‘The principle will give a testimonial to those who have figured out the answer.’

(20) has a reading on which both the embedded and matrix events took place before

the speech time and the embedded event precedes the matrix event. (21) is compatible

with two situations. In one situation, the matrix event will take place in the future but

the embedded event happened before the speech time. The other situation only

requires that the embedded event precedes the future matrix event and hence the

embedded event can also be located in the future. Such examples indicate that the

meaning of le may involve some kind of relative anteriority. However, the original

definition of the meaning of le as given in (17) allows no parameter to express the

notion of relative anteriority. As a first step to accommodate examples like (20) and

(21), let us first try to add a further condition on the topic time t2 introduced by le

such that t2 must precede the run time of an event epro, which is intended to be a

pronoun-like free variable. When this event variable is free, it is defined to refer to the

utterance event. In this case, τ(epro), i.e., the run time of the utterance event, is

equivalent to the speech time. However, epro can also be co-indexed with another

event argument, giving rise to an anaphoric reading.vii With the introduction of this

pronoun-like event variable, a first attempt to accommodate examples like (20) and

(22) Revised meaning of the verbal le

(i) [[le]] =: λPλt2λe∃e’[ P(e) ∧ P(e’) ∧ e’≤E e ∧ τ(e’)⊆t2∧ t2<τ(epro)]

(ii) [[le]] =: λPλt2λe∃e’[ P(e) ∧ P(e’) ∧ e’≤E e ∧ τ(e’)⊆t2∧ t2≤τ(epro)]

Now consider (21) again. Because the addition of the condition “t2<τ(epro)”, the topic

time of the embedded clause must precede the run time of the speech event or the run time of the matrix event, depending upon whether epro is free or anaphoric. This

accounts for why (21) is compatible with two different situations. Whenepro is free,

the topic time of the embedded clause is required to precede the speech time. When

epro is co-indexed with the matrix event argument, the topic time of the embedded

clause needs to precede the matrix event.

As for (20), when epro is co-indexed with the matrix event argument, the reading

that the embedded event precedes the matrix event is derived. However, when epro in

(20) refers to the speech event, the embedded event is constrained to fall within a past interval but it says nothing about the relation between the embedded event and the matrix event. In principle, there are three possible relations between the embedded event and the matrix event, i.e., the embedded event precedes, follows or overlaps the matrix event. However, (20) has only the reading on which the embedded event precedes the matrix event. How is absence of the other two possible readings to be accounted for when epro is the speech event? Here is one possibility. Let us assume

that indirect speech is transformed from direct speech by leaving out the quotations. Then the content of the indirect speech should match the content of direct speech. In Chinese direct speech, the progressive marker zai is required for an accomplishment

to express an overlapping relation and the future marker hui ‘will’ is needed to

express futurity (e.g., Ta zai gai yi dong fangzi ‘he is building a house; ta hui gai yi dong fangzi ‘He will build a house’). It follows from this that the embedded clause in

(20) cannot express an overlapping or following relation, because the marker zai or hui is not there. Therefore, the only possible reading of (20) is the one where the

embedded event precedes the matrix event when epro refers to the speech event. In

other words, the reading on which epro refers to the speech event happens to coincide

be detected in (20).

Though the attempt given in (22i) works very well when the embedded clause is a

telic sentence, it runs into problems when the embedded clause is atelic such as those

sentences in (10). As mentioned, those sentences require that the run time of a subpart

e’ of the event e be included within the topic time, which is the speech time by default.

This is why they have the present continuative reading. Now if the topic time is

further constrained to precede the speech time, then a contradiction will arise because

the speech time will be required to precede itself. Consequently, if the definition of

(22i) were adopted, the original account for the present continuative reading of those

examples in (10) would be lost. In order to maintain the original result, one might

suggest that an equation symbol, intended to mean an overlapping relation, is added to

make it possible for the topic time to precede or overlap the run time of epro. Thus, a

second attempt to modify the semantics of the verbal le is something like (22ii),

where the topic time of an event e is claimed to either precede or overlap the run time

of epro.viii On this analysis, when epro refers to the speech event, the run time of epro

overlaps the default topic time now. Thus, the present continuative interpretation is

maintained.

The revised semantics of le in (22ii) has a very nice consequence when examples

like (10) are embedded to a verb. It predicts that when a sentence like those in (10) is

embedded to a verb, a simultaneous reading can arise. For example, in (23), when epro

is co-indexed with the matrix event argument, the topic time of the embedded event of

raising a goldfish may overlap Zhangsan’s saying time. It follows from this that the

embedded event many be simultaneous with the matrix event. Indeed, this seems to be

correct.

Zhangsan say he raise-Asp one-Cl goldfish

‘Zhangsan said that he was/is raising a fish.’

If the topic time of the embedded clause with le can overlap the event time of the

matrix clause, this predicts that when a telic sentence is embedded to a verb, it can

have a temporal reading according to which the embedded event is included within an

interval overlapping the matrix event. Is this prediction correct? (24) is a suggestive

piece of evidence for a positive answer. Suppose that Zhangsan has uttered a sentence

like (24a) and later this utterance is reported as an indirect speech as in (24b).

(24) a. Dao muqian weizhi wo yijing chi-le wu-tiao yu

to now until I already eat-Asp five-Cl fish

‘I have so far eaten five fishes.’

b. Zhangsan shuo dao ganggang weizhi ta yijing chi-le wu-tiao yu

Zhangsan say to just-now until he already eat-Asp five-Cl fish

‘Zhangsan said that util just now he had eaten five fishes.’

In (24b), the topic time of the embedded clause is some past interval whose final

subinterval is just now. Since this final subinterval is arguably the same as the initial

subinterval of the time of saying, an overlap relation can be claimed to exist between

the time of saying and the topic time of the embedded clause. If this is correct, then

the addition of the equation symbol to (22i) also makes a correct prediction for telic

situations.ix

Summarizing this section, following Xunning Liu’s (1988) idea, I have analyzed

the verbal le as a realization operator and formally defined its meaning in terms of

that an event e denoted by P is realized if and only if a subevent e’ of e that also falls

under P is included within the topic time. When P is telic, e’ is equivalent to e, hence

entailing that e is perfective. However, when P is atelic, e’ can be a proper subpart of e or is equivalent to e. As a consequence, e is not necessarily included within the topic

time and hence is not necessarily perfective. However, if the focus is only on the

subpart e’ of e, it still can be claimed that e’ is perfective, because e’ is included

within the topic time. In this sense, if le is to be analyzed as a perfective marker as the

traditional assumption holds, perfectivity can only apply to that subpart of event that

is included within the topic time. If the focus is on the whole event, the aspectual

viewpoint depends upon the aktionsart of VP. This analysis thus successfully accounts

for the paradox of the perfective vs. imperfective viewpoint associated with le without

running into a contradiction.

4.2A Temporal Semantics of Guo

As discussed, unlike the verbal le, the temporal meaning of the experiential marker guo always expresses relative anteriority regardless of the aktionsart of the sentence

containing it. In simplex sentences, guo requires that the event time precede the

utterance time, whereas in complex sentences it requires that the event time of the

subordinate clause containing guo precede the event time of the matrix clause or the

speech time. Here are some examples illustrating occurrences of guo in a subordinate

clause.

(25) Ta mai-le yi-jian Daianna chuan-guo de yifu

she buy-Asp one-Cl Diana wear-Asp Rel dress

(26) Wo renshi yi-ge chi-guo she de ren

I know one-Cl eat-Asp snake Rel man

‘I know a man who has eaten a snake.’

(27) Ta (jianglai) hui jia-gei yi-ge zai Harvard du-guo shu de ren

she in-the-future will marry-to one-Cl at Harvard study-Asp book Rel person

‘She will marry a man who (has) studied at Harvard.’

In (25), the event of wearing must precede the event of buying (cf. Li (1999)). (26) is

compatible with two situations. It might describe a situation where the man that I

know ate a snake at a time before I know him; that is, the embedded event precedes

the matrix event. (26) can also be used to describe a situation where the event of

snake eating took place at a time after I came to know the man. In this reading, the

embedded event precedes the speech time but not the event time of the matrix clause.

Finally, (27) is also compatible with two situations. In one situation, the embedded

event occurred before the speech time. So (27) means that she will marry a man who

has studied at Harvard. In the other situation, the embedded event of studying at

Harvard takes place in the future but before the matrix event time of marriage.

To capture the fact that guo always expresses relative anteriority regardless of the

aktionsart of the sentence, I propose that the temporal meaning ofguo be defined as in

(28), which says that when guo is combined with a property of events P, there exists

an event e denoted by P and the running time of e is included within the topic time t2,

which in turn precedes τ(epro).

(28) The temporal semantics of guo

When guo appears in a simplex sentence, epro in (28) is free and hence τ(epro) is

equivalent to the speech time. This guarantees that any simplex sentence with guo has

a past interpretation. On the other hand, if guo appears in a subordinate clause, it can

be free or be co-indexed with the event argument of a higher clause. This explains

why (26) and (27) are compatible with different situations. The reason why (25) does

not have a reading where the event of wearing precedes the speech time but follows

the matrix event might be due to pragmatics, which I will not further explore.

It is interesting, at this point, to compare the temporal semantics of guo with that

of le. If we look at the meaning of guo in (28) and that of le in (22) carefully, it turns

out that the proposed temporal semantics of guo only minimally differs from that of le.

In essence, there are two differences between them. One difference is that while guo

requires that the run time of the whole event is included within the topic time, le only

requires that a subpart of an event is included within the topic time. The other

difference is that while the topic time of a sentence containing guo strictly precedes

the run time of the contextually determined epro, le additionally allows the topic time

to overlap the run time of epro, depending upon the aktionsart. This result, of course, is

not surprising, given that both le and guo have a past-tense like reading in many

similar contexts.

The proposed analyses of guo and le also have something similar; namely, both

seem to incorporate simultaneously the meaning of aspect, i.e., the relation between

an event and its topic time, and the meaning of tense, i.e., the relation between topic

time and a reference time. This result is very desirable, because it explains why some

studies of le and guo have suggested that they are much like a relative past tense

marker, though the traditional assumption has suggested that they are aspectual

markers (cf. Lin (2000b)).

features of my analysis of the verbal le and guo differ from the previous analyses. The

first important feature distinguishing my analysis from most of the other proposals in

the literature is that instead of using descriptive terms that might sometimes be very

vague, the proposal in this paper defines the temporal meanings of the verbal le and guo in a very formal and precise way. This not only enables us to see clearly how the

verbal le and guo differ from each other in their temporal meanings but also explains

straightforwardly why temporal interpretation of sentences with the verbal le is

sensitive to aktionsart, whereas temporal interpretation of sentences with guo is not.x

Most of the references in the literature that have touched this issue are descriptive

without a true explanation, but the analysis proposed in this paper explains it. A

second distinguishing feature of my proposal is that both le and guo seem to have an

aspectual component as well as a tense component. This explains why the verbal le

and guo are like aspectual markers as well as relative tense markers. Finally, the

proposed analysis employs a pronoun-like event (free) variable to formally capture

the flexibility of the reference time of the verbal le and guo so that the reference time

can be either the speech time or an event time in a higher clause. This formal

mechanism renders it unnecessary to say that le and guo are ambiguous as in Li’s

(1999) descriptive work.

4.3. A Temporal Semantics of Sentence-final Le

In addition to appearing as a verbal suffix, le may also occur in the sentence-final

position as illustrated in (29b). The distinction between the verbal le and the

sentence-final le has been traditionally characterized as follows: The former describes

perfectivity of a situation (Wang (1965); Chao (1968); Li and Thompson (1981);

(Teng (1975); Chan (1980); Zhu (1982)), current relevance or perfect (Li, Thompson

and Thompson (1982); Mochizuki (2000)). Although several pieces of evidence have

been adduced to support the twole distinction, it is not agreed by everyone that le as a

verbal suffix and le as a sentence-final particle are two different les. For example, Shi

(1990) has analyzed the two les as having the same meaning. In this paper, I will not

go into the debate, so I will not review the relevant arguments. Instead, I will only

explicate my own view of the sentence-final le. If the analysis to be proposed is

correct, it implies that though the meanings of the two les are not completely alike,

their core meaning is actually the same.

As a first step toward understanding the meaning of the sentence-final le, let us

compare (29a), which has the verbal le with (29b), which has the sentence-final le.

(29) a. Zhangsan mai-le yi-bu xin che

Zhangsan buy-Asp one-Cl new car

‘Zhangsan bought a new car.’

b. Zhangsan mai yi-bu xin che le

Zhangsan buy one-Cl new car Le

‘Zhangsan has bought a new car.’

Looking at the above two sentences alone, it is very difficult to tell what exactly

differentiates them in a very precise way. Both examples require that before the

speech time the event of buying a car be completed. So at first sight the truth

conditions for the two sentences in question seem to be the same. However, if the two

sentences are put into a discourse, their different truth conditions will begin to emerge.

(30) a. Zhangsan zuotian mai-le yi liang xin che, keshi jintian jiu ba chezi

Zhangsan yesterday buy-Asp one Cl new car but today then BA car

mai-gei-le bieren

sell-to-Asp other-people

‘Zhangsan bought a new car yesterday, but he sold it to some other person

today.’

b. ?? Zhangsan zuotian mai yi-liang xin che le, keshi jintian jiu ba

Zhangsan yesterday buy one-Cl new car Le but today then Ba

chezi mai-gei-le bieren

car sell-to-Asp other-people

‘Zhangsan bought a new car yesterday, but he sold it to some other person

today.’

The above contrast indicates that the sentence-final le implies that the car that

Zhangsan bought is still in his possession at the speech time, which makes the

discourse in (30b) incoherent, but there is no such implication for the verbal le. In

other words, the sentence-final le seems to require that the result state brought about

by the buying event must still hold at the speech time.

Another example that points to the same direction is the contrast between (31a)

and (31b).

(31) a. Wo zai meiguo zhu-le ershi nian, cong mei tingshuo-guo zhe-zhong shi

I in America live-Asp twenty year ever not hear-Asp this-kind thing

‘I (have) lived in America for 20 years and (have) never heard this kind of

thing.’