行政院國家科學委員會專題研究計畫 成果報告
漢語時制與時態的進一步研究(2/2)
計畫類別： 個別型計畫 計畫編號： NSC93-2411-H-009-003- 執行期間： 93 年 08 月 01 日至 94 年 07 月 31 日 執行單位： 國立交通大學外國語文學系暨文化研究所 計畫主持人： 林若望 報告類型： 完整報告 報告附件： 出席國際會議研究心得報告及發表論文 處理方式： 本計畫可公開查詢中 華 民 國 94 年 7 月 29 日
漢語時制與時態的進一步研究 2/2 A Further Study of Tense and Aspect in Chinese 2/2
計畫編號：NSC 93-2411-H-009-003 執行期限：93 年 8 月 1 日至 94 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 本研究計畫的主要目是延續我們在過 去兩年有關漢語時制與時態的研究，在過 去兩年，我們已經深入地探討了時態助詞 「著」、「了」、「過」的時間意義，也 對漢語時間指涉的方式作了非常完整的探 討，我們的研究成果已經出在Language and Linguistics 2000, Lingua 2002,
Linguistics 2003, Tsinghua Journal of Chinese Studies 2003以及Journal of East Asian Linguistics 2003。可是我們 過去的研究尚留有幾個議題需要進一步討 論及研究，這一次的研究計畫主要就是要 針對那些尚未觸及的議題作深入之討論， 以便讓語言學家對於漢語的時制與時態有 更完整的瞭解。我們在這次的計畫裡主要 研究的是補語子句及關係子句的時間解 釋，特別是我們研究了文獻上所謂的 Sequence of Tense（時制呼應）、the backward shifted reading（時間后置解 釋） , the forward shifted reading（時 間前置解釋），the simultaneous reading （時間重疊解釋）及the double access reading （時間雙指解釋）在漢語中的情 形。我們提出漢語補語子句的時間解釋主 要受下列因素決定： (i) 補語子句述語的動靜態區別。 (ii) 補語子句的完整態與非完整態區 別。 (iii) 時間副詞是否出現及出現於主要子 句或補語子句的區別。 (iv) 說話者對補語子句中靜態動詞的延 展長度的語用知識。 (v) 包含關係子句的名詞組的範域 (vi) 包含關係子句名詞組的有定與無定 解釋 我們提供豐富的證據論證上述這些因素， 再加上完整態與非完整態的時間意義就可 以很完善地說明漢語補語子句的時間解 釋，我們完全不需要擬設虛語意屬性如 [+present]或是[+past]來解釋漢語的時 間意義，如果我們的分析正確，這暗示著 漢語的句法結構樹其實不包含時制詞組， 時態在漢語中所扮演的角色就如同時制在 英語中所扮演的角色。 關鍵詞: 時制、時態、時間指涉，補語子 句，時制呼應 Abstract
The main purpose of this project is to continue my study of tense and aspect in the past two years. I have studied the temporal semantics of zhao, le and guo and discussed the devices that the Chinese language uses to express temporal information in quite a detail. The results of my previous study are very fruitful. Many of them have already been published in journals such as Language and
Linguisitcs 2000, Lingua 2002, Linguistics
2003, Tsinghua Journal of Chinese Studies 2003 and Journal of East Asian Linguistics 2003. Despite the depth of my previous studies of tense and aspect, several major issues were not touched and I hope to explore those untouched issues in the two forthcoming years, in particular the temporal interpretation of complement clauses and relative clauses in Chinese. I propose that several factors conspire to determine the temporal interpretation of complement clauses in Chinese. They are:
(i) aktionsart properties of the embedded clause,
(ii) the distinction between the perfective and imperfective viewpoint aspect of the embedded clause,
(iii) the presence/absence of an implicit anaphoric or overt temporal adverbial in the matrix or embedded clause,
(iv) world or pragmatic knowledge of the typical duration of the embedded predicate
(v) The scope of the NP containing a relative clause
(vi)The (in)definiteness of the NP containing a relative clause
We argue that the above factors, plus the semantics of perfective and imperfective aspect, are sufficient to determine the temporal interpretation of complement and relative clauses. We also argue not only that Chinese has no morphological tenses but there is no need to resort to covert semantic features under a tense node in order to interpret time in Chinese. It is shown that
aspectual markers in Chinese play the role that tense plays in a tense language. This result implies that the Chinese phrase structure has AspP above VP but no TP is above AspP. 二、緣由與目的 本計畫的緣由主要來自於我們前兩年 有關漢語時間解釋的研究計畫，在前兩年 的計畫裡，我們已經有系統地對影響漢語 時間解釋的因素作了相當程度的研究，特 別是有關核心句子(Bare Sentence)、時體 助詞、副詞子句等的時間解釋作了非常深 入的討論，但是對於補語子句及關係子句 的時間解釋，雖然也有著墨，但是多為描 述性成分居多，缺乏形式分析的具體內 涵，而且在語言事實的掌握上似乎也有所 不足，因此在這次的計畫裡，我們希望能 補足這方面的缺失，對於語言事實的真 相，做進一步的釐清與確認，然後提出具 體的理論分析，以便於日後作跨語言對比 分析時，特別是與英語的比較，能有具體 之理論根據及比較內涵。 三、結果與討論 此次的研究計畫我們已經完全執行完 畢，且撰寫了一篇 54 頁之論文，投稿到國 際期刊 Journal of Semantics，此篇論文已經 被期刊主編接受出版，預計在 2006 年夏季 刊出。現在就此篇論文的重點結果略做說 明。我們在文章中的第一部份首先對漢語 簡單句的時間解釋做出明確的分析，提出 時態是漢語決定時間解是的主要因素，並 且對於個別的時態助詞如「了」、「過」、 「在」、「著」等提出具體分析，修正我在
2003 年所發表的文章的部分結論，我們主 張「了」和「過」是時制與時態的合體， 而「在」和「著」則是單純的時態助詞， 不摻雜時制意義。知道了漢語簡單具是如 何決定時間意義後，我們接著討論英文中 補語子句的各種不同時制對時間解釋的影 響，並介紹 Sequence of Tense（時制呼 應）、the backward shifted reading（時 間 后 置 解 釋 ） , the forward shifted reading （ 時 間 前 置 解 釋 ）， the simultaneous reading （時間重疊解釋） 及 the double access reading （時間雙 指解釋）等概念，並且探討漢語的補語子 句的時間解釋，看看是否也具有和英語類 似的現象，我們對於漢語補語子句的時間 解釋的語言事實看法如下： (1) 如同英語一樣，補語子句的時間重疊 解釋和事件類型有密切關係，只有無 自然終結點的靜態情狀及進行式情狀 才有重疊義，有自然終結點的情狀只 能得到時間後置解釋。 (2) 時間前置解釋需在補語子句裡放上一 個時間副詞或是語境中提供這樣一個 副詞才行。 (3) 時間副詞若出現於主要子句則只能獲 得時間重疊解釋。 (4) 補語字句若無表示未來的助動詞『會』 或其他時間副詞，則得不到時間後置 解釋。 (5) 若無任何時間副詞或時體助詞出現， 而補語子 句是無 自然 終結點的 情狀 時，則可能產生時間雙指解釋。 接著，我們很仔細地評論介紹目前文獻上 有關補補語子句時間解釋的主要理論，我 們共介紹評論了下面幾位語言學家的理 論 ： Toshiyuki Ogihara (1989, 1995,
1996),Paul Portner (2003) 及 Silvia Gennari (2003)，我們舉例說明上述這些 理論都無法真正解釋漢語的補語子句的時 間意義，因此需要另循途徑來解釋。 接著我們就介紹決定漢語補語子句時間解 釋的方式及其實際操作模式。我們舉例說 明漢語帶上非完整態的補語子句雖然可以 得到類似英語的時間後置、時間重疊及時 間雙指解釋，但是時間雙指解釋其實是語 用因素所賦予的解釋，而時間後置解釋則 事實間副詞使用的結果，因此單純的補語 子句真正所具有的時間意義只有時間重疊 解釋，這個結果其實並不使人感到意外， 因為補語子句若為非完整態，它的語意上 必定要求主要子句的事件時間被包含在從 屬子句的事件時間內，因此補語子句的事 件時間必定與主要子句的事件時間重疊。 接著下一個章節則是論述有定指示詞及無 定名詞組如何影響關係子句的時間解釋， 我們觀察到，如果包含關係子句的名詞組 帶有指示詞「這」或「那」，那麼關係子句 就可以得到過去或現在的語意解釋，而這 兩種不同的語意解釋正好和指示詞的直指 或是回指意義有關，我們提出了一個整合 性的指示詞的意義來決定關係子句的參照 時間是說話時間還是一個過去時間，因而 解釋了指示詞對關係子句時間意義的影 響。至於無定名詞組對關係子句的影響， 我們認為主要是和無定名詞組的邏輯範域 有關，當無定名詞組加接到 VP 上時，主要 子句的事件時間就成為關係子句的參照時 間。但無定名詞組若是加接到 IP 上時，因 為範域在動詞組之外，所以不能以主要子 句的事件時間為參照時間，而必須以說話 時間為參照時間，這樣的分析方式不僅適 用於完整態關係子句，也適用於非完整態
關係子句。 文章中最後我們討論了一個主語與賓語的 不對稱現象，也就是主語名詞組裡的關係 子句一定要以說話時間為參照時間，我們 認為這是因為主語名詞組在句法上的位置 一定高於動詞組，因此動詞所指稱的事件 時間不能是參照時間，而必須以說話時間 為參照時間。。 。 四、 計畫成果自評 我們此次計畫的研究結果不僅釐清了 許多前人不曾討論過的有關態度動詞補語 子句的語言事實，在理論分析上也深入討 論許多前人不曾討論過，卻對漢語補語子 句及關係子句時間解釋有非常大影響的因 素，因而提升了我們對於漢語態度動詞補 語子句的時間解釋的全盤性瞭解，這對於 日後研究漢語補語子句的時間解釋的學者 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。此次的研究計畫我們撰寫了一篇 54 頁 之論文，投稿到語意學研究頗負盛名的國 際期刊 Journal of Semantics，此篇論文已經 被期刊主編接受出版，預計在 2006 年夏季 刊出。 五、參考文獻
Abusch, Dorit (1988) “Sequence of tense, Intensionality, and Scope” Proceedings
of the West Coast Conference on Formal Linguistics 7.1-14.
Abusch, Dorit (1994) "Sequence of Tense Revisited: Two Semantic Accounts of Tense in Intensional Contexts." Hans Kamp (ed.), Ellipsis, Tense and
Questions. Shorter version appeared in
P. Dekker and M. Stokhof, eds., Proceedings of the 9th Amsterdam Colloquium, 1993, Department of Philosophy, University of Amsterdam. Presented at Events and Grammar Conference, Bar-Ilan University, Israel, 28 Oct 1993.
Abusch, Dorit (1997) “Sequence of Tense and Temporal De Re.” Linguistics and
Philosophy 20.1-50.
Enç, Mürvet (1987) ”Anchoring Conditions for Tense”, Linguistic Inquiry 18: 633-57. Gennari, Silvia P. (2003) “Tense Meanings
and Temporal Interpretation”, Journal of
Semantics 20: 35-71.
Heim, Irene (1994) “Comments on Abusch’s Theory of Tense”, in Hans Kamp (ed.) Ellipsis, Tense and Questions, DYANA deliverable R2.2.B, Uni. Of Amsterdam, pp. 137-177.
Hu, Jianhua, Haihua Pan and Liejiong Xu (2001) “Is there a finite vs. nonfinite Distinction in Chinese?”, Linguistics 39-6: 1117-1148.
Kamp, Hans 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. Katz, Graham (2003) “On the Stativity of the
English Perfect”, in Aremis Alexiadou, Monika Rathert and Arnim Von
Stechow (eds.) Perfect Explorations, Mouton De Gruyter, Berlin, pp. 205-234.
Kratzer, Angelika (1977) “What ‘must’ and ‘can’ Must And Can Mean”, Linguistics
and Philosophy 1(3): 337-355. Kratzer, Angelika (1998) “More Structural
Analogies Between Pronouns and Tenses”, in Devon Strolovitch and Aaron Lawson (eds.) The Proceedings
of Semantics and Linguistic Theory VIII,
CLC Publications, Cornell University, Thaca, N.Y., pp. 92-110.
Lin, Jo-wang (2000) “On the Temporal Meaning of the Verbal –le in Mandarin Chinese”, Language and Linguistics 1(2):109-133
Lin, Jo-wang (2002) “Selectional
Restrictions of Tenses and Temporal Reference of Chinese Bare Sentences”,
Lingua 113:271-302.
Lin, Jo-wang (2003) “Temporal Reference in Mandarin Chinese”, Journal of East
Asian Linguistics 12: 259-311.
Ogihara, Toshiyuki (1989) Temporal
Reference in English and Japanese,
Ph.D dissertation, University of Texas at Austin. Distributed by Indian University Linguistics Club, Bloomington.
Ogihara, Toshiyuki. (1995) Double-access sentences and reference to states. Natural Language. Semantics, 3:177-210.
Ogihara, Toshiyuki (1996) Tense, Attitudes,
and Scope, Kluwer Academic Publishers,
Dordrecht.
Ogihara, Toshiyuki (1999) “Double –Access Sentences Generalzied”, in Tanya Matthews and Devon Strolovitch (eds.) Proceddings From Semantics and Linguistic Theory IX, February 19-21, 1999, CLC Publications, Cornell
University, Ithaca.
Sandstrøm, Gorel (1993) When-clauses and the temporal interpretation of narrative discourse. PhD thesis, University of Umeå.
Smith, Carlota (1997) The Parameter of
Aspect, Kluwer Academic Publishers,
Dordrecht.
Song, Mean-young (2000) “A Semantics of Sequence of Tense without a Sequence of tense Rule”, Language and
Information 4.2: 93-105: 93-105. von Stechow (1995a) “On the Proper
Treatment of Tense”, In the Proceedings
of Semantics and Linguistic Theory V,
CLC Publication, Cornell University, Ithaca, N.Y., pp. 362-386.
von Stechow (1995b) Tense in Intensional Context: Two Semantic Accounts of Abusch’s Theory of Tense”, In F. Hamm, J. Kolb and A. von Stechow (eds.) The Blaubeuren papers: Proceedings of the Workshop on Recen Development in the Theory of Natural Language Semantics, Seminar für Sprachwissenschaft,
Universtät Tübingen, pp. 379-433. Stowell, Tim (1993) Syntax of Tense . Ms.
University of California, Los Angeles. Stowell, Tim (1996) “The Phrase Structure of
Tense”, in Johan Roryck and Laurie Zaring (eds.), Phrase Structure and the Lexicon, Kluwer Academic Publishers, pp. 277-291.
Jo-wang Lin
National Chiao Tung University
Abstract
This paper outlines a framework of the temporal interpretation in Chinese with a special focus on complement and relative clauses. It argues that not only Chinese has no morphological tenses but there is no need to resort to covert semantic features under a tense node in order to interpret time in Chinese. Instead, it utilizes various factors such as the information provided by default aspect, the tense-aspect particles, and pragmatic reasoning to determine the temporal interpretation of sentences. It is shown that aspectual markers in Chinese play the role that tense plays in a tense language. This result implies that the Chinese phrase structure has AspP above VP but no TP is above AspP.
1. Introduction
It is well-known that Chinese is a language without tense morphology. However, it is sometimes suggested that it has a possibly empty Inflection (INFL) node (Huang (1998); Li (1990); among others).1 If this assumption is correct, it implies that Chinese might have semantic features such as [+present] or [+past] under a phonologically empty tense node which determine the temporal interpretation of a sentence. In this paper, I will argue that Chinese not only has no morphological tenses but lacks semantic features in the above sense. Instead, it utilizes various other factors such as the information provided by default aspect, the tense-aspect particles, and pragmatic reasoning to determine the temporal interpretation of sentences.2 In particular, I will show that aspect in Chinese plays the role that tense plays in a tense language with respect to the temporal interpretation of a sentence. In other words, the Chinese phrase structure has AspP above VP but there is no TP above AspP. This is true not only for simple sentences but for embedded clauses.
Among the many devices that the Chinese language uses to determine the temporal interpretation of sentences, I will in particular explore the following factors:3
(A) Temporal adverbs: zuotian ‘yesterday’, 1996 nian ‘the year of 1996’, etc. (B) Modal verbs: hui ‘will’, yinggai ‘should’, etc.
1 Both Huang (1998) and Li (1990) do not directly claim the existence of TP in Chinese, though they do claim
that the finite-nonfinite distinction in terms of INFL exists. For arguments against Huang’s (1998), Li’s (1990) and other people’s claims about the finite-nonfinite distinction in Chinese, see Hu, Pan and Xu (2001).
2 Also see Li (1999), Lin (2003a, 2003c) and Erbaugh and Smith (in press) for relevant discussions. 3 There are other devices such as imperative operators, covert modality in conditional clauses, de in cleft
(C) Aspectual particles: le ‘perfective/imperfective’, guo ‘perfective’, zai ‘imperfective’,
zhe ‘imperfective’, etc.
(D) Aktionsart of the VP and viewpoint aspect determined by it.
(E) Type (in)compatibility between hui ‘will/would’ and perfective and imperfective aspect
(F) Scope of the DP containing a relative clause.
(G) Definiteness or informational status of the DP containing a relative clause.
In addition to the above factors, there are a number of pragmatic principles that make use of the above pieces of information and determine the temporal interpretation of a clause. To discuss how the above pieces of information are used in interpreting time in Chinese, in what follows, I will discuss the temporal interpretation of Chinese sentences in general and of complement and relative clauses in particular.
2. The Basics of the Temporal Interpretation in Chinese
As noted at the outset of this article, Chinese is usually classified as a tenseless language, as its verbs are not inflected for overt morphological tense markers. Thus, unlike the temporal interpretation in English, which can be determined by morphological tenses, temporal interpretation in Chinese is not determined by tense markers. If we disregard contextual information from the previous discourse, there are at least four main factors which influence the temporal interpretation of simple sentences in Chinese: (i) temporal adverbials, (ii) default viewpoint aspect, (iii) aspectual markers, and (iv) modal verbs. I discuss these factors in turn.4 Before doing this, I want to first spell out my assumption about the Chinese phrase structure. I assume that Chinese has the following phrase structure: [CP…[IP… [ModalP… [AspP… [VP…]]]]].
There is no TP above Asp, because as I will argue later on, Chinese not only has no morphological tenses but does not resort to covert tense features to interpret time.
2.1 Temporal adverbs
It is self-evident that temporal adverbials play an important role in the temporal interpretation of a sentence, as is illustrated by (1).
(1) Zhangsan zuotian qu ni jia Zhangsan yesterday go you house ‘Zhangsan went to your house yesterday.’
Though temporal adverbials themselves do not specify the relation between the temporal interval they indicate and that of the event they modify, the viewpoint aspect or the aspectual class to be discussed below determines which interval is included within the other interval, hence making the relation between the temporal adverbial and the event interval clear.
2.2 Default viewpoint aspect
In Chinese the use of a temporal adverbial is not obligatory. It is not uncommon to find sentences without any temporal adverbials or aspectual markers. Illustrated below are some such examples.
(2) a. Zhangsan hen mang Zhangsan very busy ‘Zhangsan is very busy.’ b. Ni da lanqiu ma? you play basketball Q ‘Do you play basketball?’
(3) a. Zhangsan dapuo yi-ge huaping Zhangsan break one-Cl vase ‘Zhangsan broke a vase.’
b. Ta dai wo qu taibei ta take me go Taipei ‘He took me to Taipei.’
Read in isolation, (2a) is interpreted as equivalent to a present tense sentence and (2b) to a present generic sentence. In contrast, the two sentences in (3) have a past interpretation.5
In Lin (2003c), I have suggested that the temporal interpretation of sentences without any temporal adverbs or aspectual markers is determined via their viewpoint aspect. Namely, a sentence with imperfective viewpoint aspect has a present interpretation, whereas a sentence with perfective viewpoint aspect has a past interpretation.6 Moreover, I proposed to use Bohnemeyer and Swift′s (2001) theory of default aspect to complete the job. According to them, in telicity-dependent languages there is a correlation between the telicity of an
eventuality description and its aspectual viewpoint when the sentence is not overtly marked for viewpoint aspect. Roughly, 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 event it denotes has at least one
5 Note that when a predicate is non-punctual, an aspectual marker is usually needed, unless the construction is a
serial verb construction as in (3b). I have no idea why this is the case.
non-final part that falls under the same predicate.7 According to them, cross-linguistically the default viewpoint aspect of telic predicates is perfective viewpoint, whereas the default viewpoint aspect of atelic predicates is imperfective viewpoint and this can be derived from a notion of event realization. They define default aspect as in (4a), where tTOP is equivalent to
Klein′s (1994) topic time, a time at which a sentence is asserted to be true. Bohnemeyer and Swift′s (2004, p. 286) notion of event realization is defined in (4b).
(4) a. DASP = λPλtTOP∃e[REALE(P, tTOP, e)]
b. ∀P⊆ E, tTOP, e [REALE(P, tTOP, e) ↔∃e′[P(e′) ∧ e′≤E e ∧τ (e′) ≤T tTOP]]
(5) a. Perfective aspect =: λP<i,t>λtTop∃t[t⊆ tTop∧ P(t)]
b. Imperfective Aspect =: λP<i,t>λtTop∃t[tTop ⊆ t ∧ P(t)]
The idea of event realization is that a predicate P is realized by event e at topic time tTOP if
and only if P is true of a part e′ of event e and the run time of e′ is included within the topic time tTOP. From the above definition, we can infer that for a telic event to be realized (to occur
or to happen), the whole event must be completed. In other words, one can assume that when a predicate is telic, the DASP determines that the aspectual node of the sentence is realized as the perfective aspect whose semantics is defined in (5a)—i.e., the perfective viewpoint requires that the event time of a situation be entirely included within the topic time. In contrast, for a (stative) atelic predicate to be realized, it is sufficient for a part of the atelic eventuality to hold at the topic time. Suppose that determination of default viewpoint aspect requires only minimal realization of an event. Then, when a predicate is atelic, the DASP determines that the aspectual node of the sentence is realized as the imperfective aspect whose semantics is defined in (5b), i.e., the topic time is included within the event time.8 Given the above discussion, we can conclude that the default viewpoint aspect of (2a) and (2b) is imperfective, whereas the default viewpoint aspect of (3a) and (3b) is perfective.
Before we examine whether the above theory of default aspect and the definitions of perfective and imperfective aspect may really help derive the temporal interpretation in Chinese as I suggested in Lin (2003c), I will further assume two rules as part of the temporal system in Chinese. One rule, as is given in (6a), is a default rule which will assign the speech time as the value of the evaluation time or topic time variable at the root level. Another rule is (6b), which applies at the IP level to an output translation of type <i,<i,t>>, closing an unfilled topic time variable.9 I will explain the application of (6b) in more detail later on when
7 Atelicity is usually defined in terms of subinterval property. That is, if t is an interval at which an atelic
predicate P holds true, then every proper subinterval of t is also an interval at which P is true.
8 This is a simplified story of Bohnemeyer and Swift (2004). The reader is referred to their article for more
details.
9 Presumably, the rule in question applies to IP, because IP is the level where a topic time should be found as in
Zuotian Zhangsan zou-le ‘Yesterday Zhangsan left’, where zuotian ‘yesterday’ is adjoined to IP. An alternative is
to let this rule apply at CP. This alternative, however, requires that we let quantifier raising get adjoined to CP in (70) and (71) to be discussed later. I will not try to resolve this issue here.
relevant examples are under discussion.
(6) a. An expression φ of type <i,t> that serves as a translation of a matrix sentence is true iff [[φ]] (s*) = 1, where s* is the speech time.
b. Ifφ is an expression of type <i,<i,t>>, apply the formula ‘λR<i,<i,t>>λt1∃t2R(t2)(t1)’ to φ.
With the above formal machinery in mind, the present interpretation of (2a) and (2b) can be derived as follows. Take (2a) as an illustration. This sentence has the imperfective
viewpoint aspect as the default aspect and hence translates as λtTop∃t[tTop ⊆ t ∧ busy′ (he′) (t)].
(6a) then applies, yielding ∃t[s*⊆ t ∧ busy′ (he′) (t)]. Since the speech time is included within the situation time, (2a) has a present interpretation.
In contrast, the default viewpoint aspect of (3a) and (3b) is perfective. So the output translation of (3a) and (3b) are (7a) and (7b), respectively.
(7) a. λtTop∃t∃x[t⊆ tTop∧ vase′(x) ∧ break′(Zhangsan′)(x)(t)]
b. λtTop∃t[t⊆ tTop∧ take′ (me′)(to-Taipei′)(he′)(t)]
Rule (6a) then applies, deriving the result that the situation time is included within the speech time. This gives rise to problems. If an event is durative as in (3b), it is impossible for that event to be included within the speech time, because the latter is only a moment of time. Thus, (3b) is predicted to have no interpretation, let alone a past interpretation. On the other hand, if an event is instantaneous such as (3a), it is theoretically possible for an instantaneous moment to be included within the speech time. However, sentences such as (3a) are not used to
describe instantaneous events at the speech time. They are used to describe past events. The formal mechanism as outlined above does not predict them to have a past interpretation.10 This indicates that perhaps the definition of perfective aspect as defined in (5a) is inadequate for the temporal interpretation of Chinese.
In order to explain the fact that perfective aspect in Chinese always gives rise to a past interpretation, I propose that the definition of perfective aspect in (5a) as given in
Bohnemeyer and Swift (2004) is revised as (8) in Chinese, with a precedence relation between the topic time variable and the evaluation time variable added.
(8) Perfective aspect = λP<i,t>λtTopλt0∃t[t⊆ tTop∧ P(t) ∧ tTop< t0]
It is significant to highlight the precedence relation between the topic time and the local
10 Perhaps one might say that the present interpretation of (3a) is excluded because of a pragmatic reason;
namely, normally by the time the speaker knows that an achievement event has happened, it is past, so even if the event is really instantaneous, the perception of it precedes the statement of it. This pragmatic explanation might be right, but the question still arises as to how the past interpretation of sentences like (3a) is formally
evaluation time designated as t0. This relation actually incorporates the notion of semantic
tense into the semantics of aspect and thus has a strong implication with respect to what tense is in a morphologically tenseless language. That is, the notion of tense in Chinese is expressed in terms of tense-aspectual particles.
As a concrete illustration of (8), let′s apply it to (3a). The result is (9a). Since this output translation is of type <i,<i,t>> with no overt topic time to fill in the value of the topic time variable, rule (6b) applies to existentially close the topic time variable, yielding (9b). Then rule (6a) applies with the speech time filling in the value of the evaluation time variable t0,
yielding (9c).
(9) a. λtTopλt0∃t∃x[t⊆ tTop∧ tTop< t0 ∧ break′ (x)(Zhangsan′)(t) ∧ vase′ (x)]
b. λt0∃tTop∃t∃x[t⊆ tTop∧ tTop< t0 ∧ break′ (x)(Zhangsan′)(t) ∧ vase′ (x)]
c. ∃tTop∃t∃x[t⊆ tTop∧ tTop< s*∧ break′ (x)(Zhangsan′)(t) ∧ vase′ (x)]
In (9c), since the topic time precedes the speech time, it follows that the breaking event precedes the speech time. Therefore, the precedence relation imposed by the perfective aspect in Chinese is functionally like the past tense in English.
It is important to emphasize the proposed distinction between default perfective and imperfective aspect in Chinese. While imperfective aspect expresses a pure aspectual meaning relating the topic time and the event time, perfective aspect incorporates into its temporal meaning an extra relation between the topic time and the evaluation time. In other words, in addition to its aspectual component, perfective aspect has semantic tense as part of its meaning. This makes perfective aspect and imperfective aspect different in their semantic type. As I will show later, this type distinction also persists in overt aspectual particles such as the perfective markers le and guo on the one hand and the imperfective markers zai and zhe on the other hand. That is le and guo have the evaluation time variable t0, but zai and zhe don′t.
Note that the default imperfective aspect and the overt imperfective markers zai and zhe do not have the evaluation time variable not because it is impossible to incorporate such a relation. One can easily add such a relation, namely, tTop = t0, to the denotation of the
imperfective aspect while maintaining the same temporal meaning as before. So the real reason why there is a type distinction between the perfective and imperfective aspect is that only perfective aspect incorporates the meaning of tense and that only semantic tense implies having an evaluation time variable t0. This type distinction will become clear in section 2.4
when I discuss the semantics of the future modal hui ‘will/would’ and its interaction with different aspectual markers.
In addition to the default aspect, there is another aspect, called neutral aspect first
discussed in Smith (1997), that I will employ in interpreting time in Chinese. This aspect will
be used when I discuss the future modal hui ‘will/would’ in section 2.4.
According to Smith (1997), the neutral viewpoint aspect involves the initial point of an eventuality and part of its internal stage but not the end point. It is neither perfective nor imperfective but is aspectually vague, allowing both closed and open interpretations. I will adopt Pancheva′s (2003) formal definition of Smith′s (1997) notion of neutral viewpoint aspect as given below.
(10) Neutral aspect = λP<i,t>λi∃t[i ⊃ t ∧ P(t)], where i ∈ I, the set of temporal intervals, i ⊃ i′
iff i ∩ i′ ≠ ∅ & ∃t[t ∈ i & t ∉ i′ & ∀t′[t′∈ i′→ t < t′]]
What (10) says is that there must be an overlap relation between the topic time and the event time and that there is at least one interval that is included in the topic time and precedes the whole event interval. Since the relation between the endpoint of the event and the topic time is not specified, the event may be completely included within the topic time or may just partially overlap it.
I assume that the neutral aspect gets introduced to the syntax of a clause through an aspectual selectional restriction between the modal hui ‘will’ and its AspP complement; namely, hui ‘will’ must aspectually select a neutral AspP. Moreover, the theory of grammar does not allow projecting an aspectual node without any semantic content. Thus, if no modal like hui ‘will/would’ is present in syntax and no overt aspectual particles are present, the default aspect is then operative, introducing either the perfective or imperfective aspect under the Asp node.11 The motivation of neutral aspect will be discussed later on.
2.3 Aspectual markers
Before discussing examples with aspectual markers, I would like to first make a few remarks on event structure. I assume, roughly following Caudal (1999), that (the time of) an
eventuality canonically breaks down into (the time of) Inner Stage and (the time of) Result State.12 For a dynamic event, the Inner Stage of an eventuality e is the event’s development. For a state, the Inner Stage is the state itself. The notion of Result State refers to the result state of an eventuality. This notion is not problematic for accomplishments and achievements. But are activities and states associated with a result state? Traditionally, one would assume that the answer is no. However, I would like to make a novel assumption here. I propose that activities and states can have results, too. To implement this idea, I assume that a function
11 My view here is different from Smith (1997) in that she claims that the neutral aspect is associated with every
sentence with no overt aspectual markers. I claim that the neutral aspect appears only under the scope of hui ‘will’. The rule of Default Aspect will take care of the other cases.
12 To simplify the logical formulae, I do not introduce event arguments for verbs. So verbs have only time
arguments. Of course, one can assume that verbs have an event argument and time argument simultaneously or just an event argument and derive the time of an eventuality via the trace function.
called Rstate is defined in such a way that it may apply to (the time of) any eventuality and returns (the time of) the result state associated with that eventuality. For example, when applied to an accomplishment such as John goes to America, it returns (the duration of) the result state of John’s being in America. When Rstate applies to activities and states, it yields (the duration of) a result state for them. However, the result states for activities and states are a special kind of eventuality, which I will refer to as empty result states and which I assume exist in all times after the eventuality has occurred. The result states that I am assuming are different from the resultant state associated with the perfect as discussed in Parsons (1990). The resultant state of the perfect is something that cannot cease holding at some later time once an eventuality has occurred. In contrast, the result states that I am assuming are more like Parson′s (1990) notion of target states. Some target states may cease holding at some later time but some may hold forever. For example, the target state of a theorem being proven must last forever, whereas the target state of a door being bolted may just last for a little while (Kratzer 1994). I assume that empty result states are among those that cannot cease holding at some later time once the event has occurred. The motivation of empty result states will
become clearer later. Since we want the inner time and result state time of an event P at time t, the functions Istage and Rstate must depend on P and can be defined as follows.
(11) a. Istage(t,P) is defined if P(t) = 1, in which case
(i) if P is telic, Istage(t,P) = t minus the last point of t; (ii) if P is atelic, Istage(t,P) = t.
b. Rstate(t,P) is defined if P(t) = 1, in which case
(i) if P is telic, Rstate(t,P) = the interval at which the result state of P exists.13 (ii) if P is atelic, Rstate(t,P) = the interval consisting of every moment after t.
Another point to make here is that the literature on Chinese aspectual markers such as le,
guo and zhe is so huge that it is impossible to give even a brief overview here, due to
restrictions of space ((Kong (1986); Huang (1988); Xunning Liu (1988); Yuehua Liu (1988); Shi (1990); Magione and Li (1993); Dai (1994); Ross (1995); Yeh (1996); Smith (1997); Liu (1997); Li (1999); Kang (1999); Lin (2000, 2003a); Klein, Li and Hendriks (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.
2.3.1 The experiential marker guo
Now let us discuss the experiential marker guo, which seems to always imply the pastness of the whole eventuality regardless of the situation type involved. This is illustrated by the
13 It is notoriously difficult to define result states, so I assume that the value of Rstate(t,P) is not definable in
following examples.
(12) a. Lisi he-guo jiu Lisi drink-Asp wine ‘Lisi drank wine before.’
b. Lisi zuo-guo yi-ge qishi dangao Lisi make-Asp one-Cl cheese cake ‘Lisi made a cheese cake before.’
c. Wo xiangxin-guo ni I believe-Asp you ‘I believed you before.’
d. Lisi die-duan-guo zuo tui Lisi fall-broken-Asp left leg ‘Lisi broke his left leg before.’
(12d) is worth special mentioning here, because it involves a result state. What is interesting about the result state is that it may not hold at the utterance (evaluation) time; namely, the broken leg must be cured before the utterance time. This property of guo is known as the “discontinuity effect” in the literature. A very simple way to capture the past interpretation and the discontinuity effect of guo is to say that the whole eventuality, including the Inner Stage and the Result State, must precede the evaluation time.
Although the above analysis successfully accounts for the temporal interpretation of the examples in (12b) and (12d), there are reasons to believe that it might be wrong. A few people have observed that the so-called “discontinuity effect” actually displays a definite/indefinite asymmetry.14 Compare the following pair of sentences:
(13) a. Lisi nong-huai-guo zhe-bu shouti-diannao Lisi make-broken-Asp this-Cl laptop
‘Lisi broke this laptop before.’
b. Lisi nong-huai-guo yi-bu shouti-diannao Lisi make-broken-Asp one-Cl laptop
‘Lisi broke a laptop before.’
While (13a) implies that the laptop is already fixed at the speech time, (13b) does not have such an implication. (13b) is compatible with a situation where the broken laptop is fixed or one where it is not fixed yet. In fact, it is even possible that the laptop in (13b) is not fixable at
constraints on the value of Rstate(t,P), for example, the result must follow t. I thank Paul Portner for this remark.
all. It is not clear how this distinction can be captured under the assumption that the inner stage and result state must precede the evaluation time.15
Another problem with the above analysis is a theory-internal problem. As noted, I
assume that activities and states also have result states and they last forever since their coming into existence. This makes it impossible to claim that both the inner stage and result state precede the speech time.
In view of the above problems, I would like to propose to detach the discontinuity effect from the temporal meaning of guo and derive it from another well-known property of guo, which requires that the eventuality modified by it be repeatable at the evaluation time. The account that I am going to offer is based on Wu′s (2005) idea. In (13a), the object DP is a definite noun phrase. The referent of this definite DP is not only the theme of the first laptop-breaking event but also the theme of any other potentially repeated laptop-breaking event. However, in order for the same laptop to be broken in a repeated event, it should be first fixed before it is broken again. Similar remarks apply to the case of the broken leg in (12d). Before one’s leg can be broken again, it should be healed first. Therefore, the discontinuity effect of guo is actually derivable from the repeatability condition.
What about (13b)? Here, the object DP yi-bu shouti diannao ‘a laptop’ is an indefinite DP, not a definite DP. Therefore, in order for a similar event to reoccur, any indefinite laptop can serve the purpose. Since the laptop involved in the repeated event need not be the same laptop as the one in the original event, there is no requirement that the original laptop be fixed before another one is broken. It might be fixed but this is not a requirement. There is simply no logical connection between the two laptops.
If what I said above is correct, then we can say that the temporal meaning of guo only requires that the Inner Stage of the eventuality modified by guo precede the evaluation time. On the other hand, the repeatability condition will force the discontinuity effect in most cases, unless the theme of the event is an indefinite DP. Given this, I propose that the temporal meaning of guo is the following:
(14) The temporal semantics of guo
||guo|| = λP<i,t>λtTopλt0∃t[P(t) ∧ IStage(t,P)⊆ tTop∧ tTop < t0]16
15
Paul Portner said to me that this theory might come out fine if the result state of the property Lisi breaking a laptop is different from the result state of Lisi breaking the laptop. For example, if the property of Lisi breaking a laptop has the last moment of t as its result state, regardless of what happens to any broken laptops. However, it is not clear why the result state of Lisi breaking a laptop should be restricted to only the last moment of t.
16 If we have event rather than time arguments for verbs, the semantics of guo can be alternatively defined as in
Essentially what (14) says is that the use of guo requires that the time of the inner stage of an eventuality is included within the topic time tTop, which in turn precedes the evaluation time t0.
The repeatability condition can be treated as a presupposition of guo, though I will not spell out the details. To illustrate, according to (14), (12c) has the syntactic representation in (15a) and is interpreted as (15b), where the speech time is the default evaluation time. (I assume that the subject originates in the specifier position of VP and is later moved to the specifier position of IP.)
(15) a. [IP woi [AspP guo [VP ti xiangxin ni]]]
b. ∃tTop∃t[believe′ (you′)(I′)(t) ∧ IStage(t, λt.believe′ (you′)(I′)(t))⊆ tTop∧
tTop < s*]
Since the inner stage of the believing state is included within a topic time that is located before the speech time, (15) is correctly predicted to have a past reading.
The final point about guo that I want to make is that its semantics is not deictic, because when a clause with guo is embedded, the evaluation time can be shifted to the matrix event time. This is illustrated by sentences like Zhangsan shuo ta xiangxin-guo ni ‘Zhangsan said that he had believed you’. Therefore, the role of guo is somewhat like a relative past tense operator. I will come back to this point in more detail later at the end of section 5.
2.3.2 The perfective/imperfective marker le
The next aspectual marker to be discussed is the verbal le, which like the experiential marker
guo, is often assumed to be a perfective marker. However, the temporal meaning of le differs
from that of guo in many crucial ways. The following examples illustrate the semantics of le.
(16) a. ?Lisi he-le jiu Lisi drink-Asp wine ‘Lisi drank wine.’
b. Lisi zuo-le yi-ge qishi dangao Lisi make-Asp one-Cl cheese cake ‘Lisi has made a cheese cake.’
c. Quan xiao de ren dou zhidao-le zhe-jian shi all school Gen person all know-Asp this-Cl matter ‘All the people in the school have known this matter.’
d. Lisi die-duan-le zuo tui Lisi fall-broken-Asp left leg ‘Lisi has broken his left leg.’
When le occurs in an activity as in (16a), the sentence often sounds incomplete and needs another continuing sentence to make it fully grammatical as in Lisi he-le jiu, ye
chang-le ge ‘Lisi drank wine and sang songs as well’. It is not clear to me why this is the case.
Yet to the extent that such patterns are interpretable, they have a past interpretation. (16a) means that the wine-drinking event occurred before the speech time.
When le occurs in an accomplishment, it assigns the sentence a past interpretation too. So (16b) means that the cake-making event took place before the speech time.
Not every stative verb may take le as a verbal suffix. Verbs like zhidao ‘know’, xiangxin ‘believe’, you ‘have’, etc., may do so, but verbs like shuyu ‘belong to’, renwei ‘think’, peifu ‘admire’, etc., may not. When a stative verb is combined with le, the sentence gets an
inchoative interpretation. Thus, (16c) implies a change of state from the state of not knowing to the state of knowing. In fact, we do not need to analyze those stative verbs combined with
le as true stative verbs but can treat them as being type-coerced into achievement verbs. This
explains why examples like (16c) are tinted with dynamicity.
Finally, when le occurs with an achievement verb or a bisyllabic resultative verb, it implies that the ensuing result state must hold at the evaluation time, though the event itself occurred before that time. Thus, (16d) means that Lisi′s leg-breaking event occurred before the speech time and his leg is still broken at the speech time. This interpretation is similar to the inchoative reading that we saw in (16c), which implies that the new state holds at the speech time. In fact, we can say that (16b) has the same type of interpretation in that though the cake-making event occurred before the speech time, the existence of the cake must hold at the speech time.
It is significant to note that in (16b,c,d), there is no overt temporal adverbial in the sentence. Now consider a case where a temporal adverbial is present as in (17).
(17) Lisi shang-ge yue die-duan-le tuei Lisi last-Cl month fall-broken-Asp leg ‘Lisi broke his leg last month.’
In (17), we have a temporal frame adverbial referring to a past interval. Therefore, the leg-breaking event must occur within that past interval. But what about the ensuing result state? Does it still have to hold true at the speech time? It can, but this doesn’t seem to be what is asserted in (17). What is asserted is that the leg-breaking event occurred last month and the state of the leg being broken was true then. The sentence doesn′t say whether Lisi′s leg is still broken or has been healed by the speech time. The situation here is much similar to
the contrast between the following two sentences, where the predicate denotes a state. (18) a. Lisi hen jushang
Lisi very depressed ‘Lisi is depressed.’
b. Lisi shang-ge yue hen jushang Lisi last-Cl month very depressed ‘Lisi was depressed last month.’
The sentence in (18a) does not have a temporal adverbial, so the topic time is the speech time by default and thus the state is asserted to be true at the speech time. However, when a
temporal adverbial is present as in (18b), the state is only asserted to be true of an interval overlapping the interval denoted by the temporal adverbial. The state of being depressed might still be true at the speech time (as can be proved by adding a continuing sentence such as Xianzai ye hai hen jushang ‘now he is still depressed) but this is not part of the assertion made by the speaker.
What we have learned above is this: the temporal interpretation of the result state of an event sentence with le seems parallel to that of simple state sentences and can be independent of the temporal interpretation of the inner stage of that event. This is particularly clear in cases where there is no temporal adverbial as in (16d). In (16d), the inner stage is true of an interval before the speech time but the result state must be true at the speech time. This suggests that each of the inner stage and the result state has an independent time at which they are asserted to be true. However, when an overt temporal adverbial is present as in (17), the topic times for the inner stage and the result state are always the same, i.e., the time denoted by the overt temporal adverbial. How can this be explained? One way to account for this is to say that the topic time for the result state is an anaphor-like time variable and hence must be bound by the overt topic time. When there is no overt temporal adverb, the topic time of the inner stage is existentially closed. Suppose that such existentially closed implicit topic times are incapable of serving as a binder. Then the anaphor-like time variable must look for another appropriate antecedent and chooses the speech time, made available by uttering the speech, as its value.
On the basis of the above discussion, I now propose the semantics of the verbal le as follows.
(19) The Temporal Semantics of Le
||le|| = λP<i,t>λtTopλt0∃t[P(t) ∧ Istage(t,P) ⊆ tTop∧ tTop< t0∧ tana⊆ Rstate(t,P)]
Briefly, what (19) says is this: the use of le requires that the time of the event’s development, i.e., the inner stage, is included within the topic time tTop, which in turn precedes the
variable that needs to be bound or given a value from context.
Notice that like the semantics of guo, the semantics of le in (19) also includes a relation between the topic time and the evaluation time. This is what makes le not a pure aspectual marker but a mixture of tense and aspect, analogous to the analysis of guo.
Another interesting point about the analysis in (19), which crucially differs from every treatment of le in the literature, is that the semantics of le is not a pure perfective marker. On the current treatment, le has a perfective meaning only with respect to the inner stage as the condition “Istage(t,P) ⊆ tTop” requires. However, the meaning of le also has an imperfective
component but this time with respect to the result state. This is reflected by the condition “tana
⊆ Rstate(t,P)” in (19). Therefore, the meaning of le involves both a perfective and an imperfective component.
Notice that the part of imperfective meaning involving the result state is an obligatory component of the meaning of le that I am proposing. A potential problem with this claim is how it can deal with activities, which on the traditional assumption have no result states at all. I would like to argue that this is not a serious problem. As noted, I have assumed that the function Rstate can apply to any eventuality, including activities and states, yielding the interval consisting of all time afterwards. On this assumption, the truth conditions of (16a) are the following, where the evaluation time and the anaphor-like time variable both pick out the speech time as their value.
(20) ∃tTop∃t[drink′ (wine′)(Lisi′)(t) ∧ Istage(t, λt[drink′ (wine′)(Lisi′)(t)]) ⊆ tTop∧
tTop< s* ∧ s* ⊆ Rstate(t, λt[drink′(wine′)(Lisi′)(t)])]
The first half of the truth conditions in (20) has no problem. It asserts that the time of the event′s development occurred before the speech time. However, the second half needs some comments. Here we have conditions saying that the wine-drinking activity has an empty result state and the time of that empty result state includes the speech time. How is this claim to be evaluated? As noted, an empty result state does not cease holding once it comes into existence. Consequently, it overlaps the speech time. Therefore, the existence of empty result state does not affect the truth conditions at all. (20) correctly predicts that the activity occurred before the speech time.
A very nice feature of the above treatment of activities is that with the help of (the time of) empty eventualities we can unify the meaning of le in all contexts, because we don′t have to say that there is a past-tense-like le occurring in activities and an inchoative le which is combined with a type-coerced stative verb, an achievement verb or an accomplishment verb. The superficially different le′s in all contexts in fact utilize exactly the same meaning as given in (19).
Another point about the temporal semantics of le is that just like guo it is not deictic. So in a sentence like Lisi shuo Zhangsan die-duan-le zuo tui ‘Lisi said that Zhangsan fell and
broke his left leg’, the evaluation time of the embedded clause is not the speech time but the matrix event time.
2.3.3 The progressive marker zai
The temporal semantics of the Chinese progressive marker zai seems close to that of the English progressive. However, unlike the English progressive marker, the Chinese zai may not occur with achievement verbs. The use of zai is illustrated below.
(21) a. Lisi zai xi-zao Lisi Prog take-bath ‘Lisi is taking a bath.’
b. Lisi zai xie yi-ben xin shu Lisi Prog write one-Cl new book ‘Lisi is writing a new book.’
c. *Lisi zai ying Lisi Prog win ‘Lisi is winning.’
I propose that the syntactic distribution of zai be explained by imposing a selectional restriction on its lexical semantics, namely, zai can only modify a dynamic durative event. Ignoring the complexity of intensionality, i.e., the modality meaning of the progressive as discussed in Dowty (1979) and many others, I propose that the temporal meaning of zai is as follows.
(22) ||zai|| = λP<i,t>λtTop∃t[P(t) ∧ tTop ⊆ Istage(t,P) ∧ Dynamic(P) ∧ Durative(P)]
What (22) says is that zai requires that the event modified by it be dynamic and durative and that the inner stage, i.e., the event’s development, includes the topic time.
Applying (22) to (21a), for example, we will obtain the following temporal meaning: The inner stage of the bath-taking event includes the topic time, which is the speech time by default. Since the event’s development includes the speech time, it implies that the event is on-going.
Just like le and guo, the tense semantics of zai is not deictic but relative, as is proved by the sentence Lisi shi-fenzhong qian shuo ta zai xi-zao ‘Lisi said ten minutes ago that he was taking a bath’, where the time of taking a bath includes the time of saying.
Chinese has another imperfective marker, the durative marker zhe. This marker only occurs with (possibly stage-level) atelic eventualities (Lin 2003b). The use of zhe is illustrated in (23).
(23) a. Ta zui li jiao-zhe koxiangtang he mouth inside chew-Asp chewing-gum ‘He is chewing a chewing gum in his mouth.’ b. Ta liu-zhe yi-tou chang fa
he wear-Asp one-head long hair ‘He wears his hair long.’
c. *Ta da-puo-zhe beizi he hit-broken-Asp cup ‘He is/was breaking cups.’
d. *Ta xie-zhe ling-pian wenzhang he write-Asp two-Cl articles ‘He is/was writing two articles.’
I define the temporal semantics of zhe as follows:
(24) ||zhe|| = λP<i,t>λtTop∃t[P(t) ∧ tTop ⊆ Istage(t,P) ∧ Atelic(P)]17
According to this analysis of zhe, (23a) has a present interpretation because the chewing activity must overlap the speech time. (23b) is similar.
Like the other aspectual markers in Chinese, the temporal meaning of zhe is relative, not deictic. So in the sentence Lisi zuotian shuo zhuo shang fang-zhe yi-bei cha ‘Lisi said
yesterday that a cup of tea was placed on the table’, the time of the embedded state overlaps the matrix event time.
2.4 The future modal verb hui ‘will’
17 Some examples such as (i) suggest that zhe might also involve the result state of an action as in (i), because
what is durative seems to be the result state of the suit being put on, as opposed to (ii), which describes the on-going action of putting on the suit.
(i) Lisi chuan-zhe xizhuang
Lisi wear-Asp suit ‘Lisi is wearing a suit.’
(ii) Lisi zai chuan xizhang
Lisi Prog put-on suit ‘Lisi is putting on a suit.’
This problem can be avoided if one assumes that the verb chuan in Chinese is lexically ambiguous between a stative and a dynamic meaning. Thus, the verb chuan in (i) lexically denotes a state rather than a change of state.
The future modal verb hui ‘will/would’ has a relative future meaning rather than a deictic one. When hui ‘will/would’ occurs in a simple sentence, the future is relative to the speech time; when it is embedded in a subordinate clause, the future is relative to the matrix event time. This is illustrated by (25a) and (25b), respectively.
(25) a. Wo hui hen mang I will very busy ‘I will be busy.’
b. Zhangsan shuo ta hui hen mang Zhangsan say he would very busy ‘Zhangsan said that he would be busy.’
I assume that the future modal verb hui ‘will/would’ heads the phrase ModalP, which is located above AspP. The basic function of hui ‘will/would’ is to locate the topic time introduced by Asp after the evaluation time. The temporal semantics of hui ‘will/would’ is given as follows:
(26) [[hui]] = λP<i,t>λtλt0[P(t)∧ t0< t]
Applying (26) to (25a) yields the result in (27b), given the LF in (27a), where the feature [+imperfective] is assumed to introduce the default imperfective meaning.
(27) a. [IP Woi [ModalP hui [AspP AsP [+imperfective] [VP ti hen mang]]]]
b. ∃tTop∃t[be-busy′(I′)(t) ∧ tTop⊆ t ∧ s* < tTop]
According to (27b), the time of my being at home includes the topic time, which is located after the speech time. So (25a) has a future interpretation relative to the speech time.
At this point, it is interesting to discuss the interaction between modal verbs and the aspectual markers le and guo. These two markers are incompatible with the modal auxiliary
hui ‘will’ as is shown by (28) (Lin 2000).18
18 The aspectual marker le may appear in future contexts such as a conditional clause or deng-clause as in (i)
below.
(i) a. Yaoshi Lisi dang-le zongtong, wo yiding quanli bangzhu ta if Lisi select-Asp president I definitely all-effort assist him ‘If Lisi is selected as president, I will definitely assist him with all my effort.’
b. Deng ni nadao-le boshi xuewei, wo jiu mai xin che gei ni wait you get-Asp doctor degree I then buy new car for you ‘After you have got your doctor degree, I will buy a new car for you.’
(28) *Lisi hui likai-guo/le bangongshi Lisi will leave-Asp office ‘Lisi will have left the office.’
Why is hui ‘will’ incompatible with le and guo? I claim that (28) is ill-formed due to a type mismatch. As stated in (26), the function denoted by hui ‘will/would’ requires an expression of type <i,t> to serve as its argument. However, as discussed earlier, the output translation of the combination of le/guo with a VP is an expression of type <i,<i,t>> due to the addition of an evaluation time variable. Therefore, semantic computation crashes when hui ‘will/would’ is combined with an AspP headed by le or guo. In fact, the ungrammaticality of (28) can be seen as evidence in support of the proposed treatment of le and guo as involving a precedence relation between the topic time and the evaluation time.
It should be noted that the existential closure rule (6b) cannot apply to the AspP projection in (28), closing the topic time variable first before hui ‘will/would’ is combined with the AspP projection. If this were possible, there would be no type mismatch problem. The operation of topic time closure rule cannot apply here because rule (6b) applies to IP instead of AspP. I speculate that this restriction of rule (6b) might be due to the fact that AspP is not a level where one can be certain that no overt topic time can be found. In many cases, an overt adverbial marking the topic time is attached to ModalP or even higher to IP. In contrast to le and guo, hui ‘will/would’ can be combined with an AspP headed by the progressive marker zai or zhe, as is shown in (29). This is because the output translation of the combination of zai/zhe with a VP is of type <i,t>, which is exactly the type that hui
‘will/would’ requires.
(29) a. (Wangshang) ni hui-bu-hui hai zai jia-ban?
night you will-not-will still Prog work-overtime ‘Will you be still working overtime at night?’
b. Qiang shang hui gua-zhe yi-fu hua Wall on will hang-Asp one-Cl picture ‘There will be a picture hanging on the wall.’
A consequence of the above analysis is that the future modal hui ‘will/would’ is predicted to be always incompatible with a perfective aspect under its scope, including the default perfective aspect, because the output translation of the combination of a VP and the default perfective aspect as defined in (8) is of type <i,<i,t>>, too. This then predicts that the
for example). Such examples are indeed evidence that le is not an absolute past tense marker but they are still compatible with the claim that le has a component of relative past as part of its meaning as I have proposed in this paper. Another thing to note is that even in future contexts as in (ia) and (ib), le and hui may not occur with each other. If hui ‘will ’ is added, the sentence becomes ill-formed.
viewpoint aspect of sentences like those in (30) cannot be perfective. (30) Wo (xiawu) hui xiuli yi-bu che
I afternoon will fix one-Cl car ‘I will fix a car (this afternoon).’
If the viewpoint aspect of (30) is not perfective, what is it then? As a first step to answer this question, let us consider the following dialogue:
(31) A: Ni xiawu hui zuo shenme? you afternoon will do what ‘What will you do in the afternoon?’
B: Wo xiawu hui xiuli yi-bu chezi I afternoon will fix one-Cl car ‘I will fix a car in the afternoon.’
A: Si dian deshihou ni hui hai zai xiu che ma? four o′clock when you will still Prog fix car Q ‘Will you be still fixing the car at four o′clock?
(32) a. B’: Shi-de, na-bu che yuji yao xiu san tian yes that-Cl car estimated need fix three day ‘Yes, it is estimated that the fixing of the car will last three days.’
b. B”: Bu, wo yugu xiu dao san dian keyi wancheng, No I estimate fix until three o′clock may complete
si dian yinggai yijing xiu-hao-le four o′clock should already fix-complete
‘I estimate that I will finish fixing it at three o′clock. It should have been fixed by four o′clock.’
The questions and answers in (31) and (32) show that a future statement such as (31B) may present an open situation. The speaker may intend the future event to be an incomplete (on-going) event as (32B’) indicates or a complete event as (32B”) indicates. In fact, it seems quite appropriate to translate (31B) as ‘I will be engaged in fixing a car in the afternoon’, leaving achievement of the goal of the event open. If the viewpoint aspect of (30) is not absolutely perfective, what is it then? I propose that (30) has a neutral viewpoint aspect as defined in (10) in section 2.2. If the above analysis is right, it is even possible to replace the imperfective aspect in (27) by the neutral aspect. This will make it possible to claim that the aspect under the scope of hui ‘will’ is always the neutral aspect, if no overt aspectual marker is present.
2. 5 Summary of section 2
In this section, I have shown that no matter which of the four factors—temporal adverbials, viewpoint aspect, aspect markers or the future modal hui ‘will’, is involved in interpreting the temporal location of an event, there is no need to postulate covert semantic features such as [+present] or [+past] under an empty Tense node. Therefore, apart from the lack of
morphological tense, it is reasonable to say that Chinese lacks semantic tense under a syntactic tense node, parallel to English ones. This weakens the possible claim that Chinese has empty tense node in syntax.
However, if semantic tenses are simply understood as ordering relations between time spans, Chinese can be said to have them. Normally, aspects are understood as expressing the inclusion relation ⊆ (perfective) and ⊇ (imperfective) between the event time and the topic time and tenses as the ordering relation between the topic time and the evaluation time. In Chinese, the future relation > is expressed by hui ‘will/would’. This element can be seen as a semantic tense. In contrast, the past relation is contained in le and guo. However, le and guo contain both the inclusion and ordering relation. Two parts of the information are thus packed into one morpheme. So they can be said to be semantic tense and semantic aspect at the same time. On the other hand, the progressive marker zai and the durative marker zhe are pure aspectual markers without having any tense meaning.
3. The Temporal Interpretation of Embedded Clauses 3.1 The case of English
The temporal interpretation of embedded tenses/clauses displays interesting properties. Consider the English sentence (33), where a past tense is subordinated to another past tense.
(33) John said that Mary was pregnant. a. John said, “Mary is pregnant”. b. John said, “Mary was pregnant”. c. John said, “Mary will be pregnant”.
(33) has two distinct temporal readings (Costa 1972; Enç 1987; Abusch 1988; Ogihara 1989; among others). The time of the event described by the embedded clause Mary was pregnant may be simultaneous with or prior to the time of the matrix event. The first reading is similar to what (33a) expresses and is sometimes called the simultaneous reading. This reading seems to be derived by converting the present tense of the direct quotation into the past tense in the indirect quotation and is known as the sequence of tense phenomenon in the literature. The
simultaneous reading is often claimed to be possible only with embedded stative predicates (Enç (1987); Ogihara (1989); Stowell (1993); Gennari (2003).19 The second reading is similar to what (33b) expresses and is often referred to as the backward-shifted reading. However, (33) does not have a reading on which the time of the embedded event follows the time of the matrix event. That is, (33) cannot be equivalent to what (33c) says. This
impossible reading is sometimes referred to as the forward-shifted reading.
In contrast to a past tense, if a present tense is c-commanded by a past tense, the event time of the embedded clause has to coincide not only with the matrix event time but with the speech time. This is illustrated in (34).
(34) John said that Mary is pregnant.
Such a reading is referred to as the double-access reading in the literature.
In the above examples, we have a complement clause embedded in an attitude report verb. It has been pointed out that tenses in relative clauses behave differently from tenses in complement clauses. According to Ogihara (1989, 1996), just like the past tense in a
complement clause, the past tense of a relative clause can be understood as simultaneous with a higher dominating tense, displaying a sequence of tense phenomenon. This is illustrated by (35), where the time of the fish being alive can be understood as simultaneous with a future buying time.
(35) John said that he would buy a fish that was alive.
However, Enç (1987), Abusch (1988, 1994, 1997), and Ogihara (1989, 1996) also have observed that tenses in relative clauses differ from tenses in complement clauses in some ways. First, unlike tenses in complement clauses, tenses in relative clauses can have a
forward-shifted reading. For example, in (36), the woman could win the Noble Prize after she married with John.
(36) John married a woman who became a Noble Prize winner.
Second, when a present tense in a relative clause is embedded under a past tense as in (37), there is no effect of obligatory double-access. The embedded event in (37) only needs to be co-temporal with the speech time.
(37) John talked to a woman who is crying.
