ABSTRACT The present paper continues the investigation into the parallel between verbal and nominal expressions in aspectual behavior. Definite and indefinite NPs contribute differently to the aspectual properties of sentences. While indefinite object NPs may contribute to boundedness due to their possible function as virtual verbal measures, definite NPs do not. While indefinite NPs automatically imply total affectedness, definite NPs do not. These differences between the two kinds of NPs are due to the fact that indefinite object NPs can be event-dependent while definite NPs are event-independent. The aspectual difference between indefinite and definite NPs is also related to their different discourse roles, respectively, as first-mentioned and second-mentioned discourse entities. The congruence of aspectual and discourse properties of NPs can be seen in the distribution of these NPs in constructions such as BA, SHI...DE and VERB-COPYING that have specific aspectual and discourse functions. Our hypothesis runs counter to Li&Thompson (1981)'s assumption that definite NPs too help mark event boundaries. It also is inconsistent with a well-known transitivity hypothesis advanced by Hopper&Thompson (1980), which assumes definite NPs to be higher on the transitivity scale and more compatible with perfectivity than indefinite NPs.
I. INTRODUCTION
Since the present paper studies the aspectual properties of definite and indefinite NPs, I should first justify the possibly surprising assumption that noun phrases have aspectual properties at all. After all, aspect is a temporal notion and it is to verbs that we normally attribute aspectual properties. But since Verkuyl (1972), it has been accepted that NP arguments to verbs also contribute to the aspectual properties of sentences. An example would be the change of a verb phrase in situation type from an accomplishment to an activity type when the object changes from a count to a mass NP. The joint contribution to aspectual properties by nominal as well as verbal phrases is ultimately related to a striking parallel between verbal and nominal expressions, pointed out by a number of scholars (Mourelatos 1981, Carlson, L 1981, ter Meulen 1984, Hinrichs 1985, Krifka 1987). Both verbal and nominal expressions can be homogeneous and divisible or heterogeneous and indivisible. Homogeneous verbal expressions such as the activity verb of running have the sub-interval property, i.e., sub-events of an event can be described by the same predicate. The nominal parallel is a mass noun such as water, which can be sub-divided and the parts can still be described as water.
Heterogeneous verbal expressions such as the accomplishment predicate of building a house do not have the sub-interval property; the sub-events of building a house cannot be described by the same predicate. The nominal counterpart is a count NP such as a house, which cannot be divided into parts that can still be called a house. I will also call attention to the fact that whether verbal predicates are homogeneous or heterogeneous has to be jointly determined by the lexical meanings of its parts, including nouns and verbs. This is clearly demonstrated with the examples given in (1):
(1) playing a sonata, playing the piano, building a piano
Although playing a sonata is heterogeneous, playing the piano , with a different noun and a definite article, does not seem to be; but it is not true that the noun piano is incompatible with heterogeneity, since parts of a piano cannot be called a piano and building a piano, with a different verb, again is heterogeneous. To recapitulate, although both the verb play and the noun piano can figure in heterogeneous predicates separately, together they cannot be heterogeneous.
The present paper continues the investigation into the intimate relationship between the aspectual properties of verbs and their NP arguments. Although the contribution of NPs to aspectual properties is well-known, as far as I am aware, no attention has been paid to the possible difference between definite and indefinite NPs in aspectual behavior.
In this paper, I will present some distributional contrasts between definite and indefinite NPs in the post-verbal position. I will argue that the distributional differences between the two kinds of NPs can be explained by whether they can provide virtual verbal measures and hence contribute to boundedness, a necessary condition for the use of the perfectivity aspect in Chinese. This analysis helps to explain the behavior of the two kinds of NPs in constructions such as BA, SHI...DE and VERB COPYING. It also has implications for research into the issue of transitivity and the semantic and discourse roles of noun phrases in general.
II. AN INITIAL OBSERVATION
Definite NPs can take many different forms. They can be bare or marked. To simplify matters, most of the definite noun phrases used in this paper are those that are explicitly marked as definite, i.e., they include a demonstrative. Some examples are zhei ben shu 'this book', nei ge ren 'that person', zhei dianr shui this much water', zhei liang feng xin 'these two letters'. Similarly, indefinite NPs can be referential or non-referential, bare or with explicit marking. Again for simplicity, we use only indefinite NPs that contain some numerical expressions and classifiers. Some examples of indefinite noun phrases are yi feng xin 'one letter', liang ben shu 'two books', san ge ren 'three people', yidianr shui 'some water' and so on.
My initial observation of the contrast between the two kinds of noun phrases is given in the minimal pair in (2). Both sentences have the perfective marker le:
(2) a. Wo zuotian kan le yi ben shu.
I yesterday read LE one CL. book (CL=classifier)
'I read a book yesterday.'
b. ?Wo zuotian kan le zhei ben shu.
I yesterday read LE this CL.book
'I read this book yesterday.'
While the indefinite NP yi ben shu is perfect as the object, the definite NP zhei ben shu is not as good. An interesting fact that will not be dealt with in this paper is that generic NPs pattern with definite NPs1. While the English translation for (2a), with an indefinite NP as object, is unquestionably good, it is not clear whether the English translation of (2b) is as questionable as the Chinese counterpart. An English perfective sentence does not even have to have the object NP. In answer to the question 'what did you do yesterday', it is entirely possible to simply answer 'I read'. This is consistent with Smith (1991)'s analysis that English perfectivity is compatible with all kinds of situation types, including simple activity without any endpoint
The above observation concerning the contrast between definite and indefinite NPs is significant, as the possible difference in aspectual behavior between these two kinds of NPs has not been entertained before in the linguistic literature. This observation is also at odds with Li&Thompson (1981)'s assumption that definiteness, as well as quantity, contributes to boundedness, which is necessary for the use of the perfective aspect. The first two of their four conditions on boundedness are given in (3):
(3) a. quantified event (with NP of extent, duration, frequency)
b. definite/specific (name, pron, NP with modifiers, BA)
Condition a states that an event is bounded if it is quantified by post-verbal NPs of extent, duration, number of occurrences and so on. Condition b states that an event is bounded if it is definite or specific, rendered so by definite NPs of various kinds including proper names, pronouns, those with pre-nominal modifiers such as demonstratives, relative clauses, and finally objects of BA. While our observation confirms the validity of condition a, it does not support condition b. The minimal pair in (2) suggests that only indefinite NPs are acceptable in a perfective context.
Going beyond aspect, we find that our observation is also at odds with a well-known view of transitivity proposed by Hopper &Thompson (1980). According to Hopper & Thompson, definite objects are higher than indefinite ones in transitivity, being in the foreground and associated with perfectivity.
Now if, according to our observation, indefinite objects are more acceptable in the perfective context, then the relative transitivity ranking between definite and indefinite NPs needs to be reexamined.
III. CO-OCCURRENCE WITH VERBAL MEASURES
The contrast between definite and indefinite NPs can be seen more clearly in their different patterns of co-occurrence with verbal measures. Verbal measures are phrases that denote the duration and number of occurrence of events, such as yihuir, yici, yixia and so on. They also include verbal reduplication, as it patterns with other verbal measures.
The pair of sentences given in (4) shows that while indefinite NPs do not co-occur with any of the verbal measures, definite NPs do:
(4) a. *Ta kan le kan/yihuir/yixiar yi ben shu.
He read LE read/a while/a little one CL. book
'*He took a look at/read a while/a bit one book.'
b. Ta kan le kan/yihuir/yixiar zhei ben shu.
He read LE read/a while/a little this CL. book
'He took a look at/read a while/a bit this book.'
The verbal measures in (4) all precede the post-verbal NPs. Interestingly, the same co-occurrence pattern also holds when verbal measures occur after the NPs, as shown in (5):
(5) a. *Wo da le yi ge ren yi dun.
I beat LE one CL. person one VM (VM=verbal measure)
'?I gave a person a beating.'
b. Wo da le nei ge ren yi dun.
I beat LE that CL. person one VM
'I gave that person a beating.'
IV. INDEFINITE OBJECT NPS AS VIRTUAL VERBAL MEASURES
I interpret the mutual exclusiveness between certain indefinite object NPs and verbal measures of any kind as indication that these indefinite NPs themselves function as virtual verbal measures. This is possible when the conditions given in (6) are met:
(6) a. If the referent of the NP is event-dependent
b. If the event has a temporal span
The clearest examples of NPs as virtual verbal measures are cognate objects, exemplified in (7):
(7) Ta xia le yi pan qi.
He play LE one CL. chess
'He played a game of chess.'
The game of chess certainly does not exist apart from the event of playing it. Playing chess also has an inherent temporal span and an inherent boundary. Thus, the indefinite NP yi pan qi serves as a virtual verbal measure. Of course, yi pan qi does not explicitly denote temporal duration or completion, which is only implied.. What it provides is the event boundary necessary for the perfective aspect.
The same is true of 'effected' objects, given in (8):
(8) Ta xie le yi feng xin.
he write LE one CL. letter
'He wrote a letter.'
The difference between 'effected' objects and cognate objects is that effected objects can be referential and are able to exist independently of the events once they are created. But they are still event-dependent, as their referents are created and introduced into the discourse by the very events described in the same sentences. Something similar can be said about the opposite of effected object, namely, 'destroyed' or 'consumed' objects.
'Affected' objects can also be virtual verbal measures, as exemplified in (9):
(9) Ta xi le yi jian chenshan.
he wash LE one CL. shirt
'He washed a shirt.'
The shirt is neither brought about nor destroyed by washing, but it is event-dependent nonetheless because the shirt, in its changed state, is introduced into the domain of discourse by the event of washing.
So far the boundary necessary for perfectivity has been inherent in the naturally individuated entities denoted by indefinite NPs. But such a boundary does not have to be inherent, as has been pointed out by Smith (1991). It can also be imposed by the quantity expression, as in the case of indefinite NPs with mass or generic nouns, shown in (10):
(10)a. Ta he le yixie jiu.
he drink LE some wine
'He drank some wine.'
b. Wo nian le liang ge zhongtou de shu.
I read LE two CL hour DE book
'I read for two hours.'
In (10a), yixie imposes a boundary on the otherwise boundless mass of jiu. In (10b), liang ge zhongtou imposes a boundary on the generic verbal predicate nianshu.
An indefinite NP can fail to be a virtual verbal measure, of course, if one of the conditions given in (6) is not met. Indefinite NPs fail to be verbal measures when condition a is not met, i.e., they are not event-dependent. A minimal pair is given in (11):
(11) a. *Ta tan le yi jia gangqin.
he play LE one CL. piano
'*He played a piano.'
b. Ta tan le yi ge zoumingqu.
he play LE one CL. sonata
'He played a sonata.',
(11a) fails to be perfective because the indefinite NP yi jia gangqin fails to provide a verbal measure. Despite the fact that a piano is a naturally individuated entity with natural boundary, playing the piano does not involve a temporal span. In contrast, (11b) is fine, since playing a sonata does imply a temporal span.
This happens, for example, when they are displaced from the object position and become the object of BA, as seen in (11):
(11) Wo ba yi ge juzi chi le yi ban.
I BA one CL. orange eat LE one half
'I ate half of an (specific) orange.'
The verbal measure in this sentence is not yi ge juzi, the object of BA, but yiban, the complement to the verb chi. yi ge juzi as the object of BA has also lost its quantity meaning of 'any one orange' and come to have the referential or the de re reading of 'a specific orange'.
Indefinite NPs also fail to be verbal measures when condition b, concerning the necessity of temporal span, is not met. Indefinite NPs as virtual verbal measures have one very important property, namely, they are totally affected / effected / consumed or destroyed. This is reasonable, as their very introduction into the discourse depends on the total accomplishment of the events that affect / effect / consume or destroy them. This may also explain why indefinite NPs cannot co-occur with other verbal measures, which may specify partial affectedness.
V. DEFINITE NPS AS EVENT-INDEPENDENT ENTITIES
In contrast to indefinite objects, definite objects are not as felicitous as verbal measures due to the fact that they are event-independent. Like definite descriptions, their referents, whether real or fictitious, have the presupposition of existence. The contrast between definite and indefinite NPs in presuppositional status can be seen in a standard negation test, given in (13):
(13)a. Ta xie le yi feng xin----------->Ta meiyou xie xin.
he write LE one CL. letter he not have write letter
'He wrote a letter.' 'He did not write letters.'
b. Ta xie wan le zhei feng xin------->Ta meiyou xie wan zhei feng xin.
he write finish LE this CL. letter he not have write this CL letter
'He wrote this letter.' 'He did not write this letter.'
The sentence in (13a), with an indefinite NP, does not retain the indefinite NP in the most neutral negation, but the sentence in (13b) with the definite NP does2. This contrast can be interpreted in the same way as negation tests for presupposition are generally interpreted, namely, presupposed elements survive negation. Since the definite NPs survive negation, they are presupposed.
The presupposed existence of definite NPs has important implications for what can be done to them. Unlike indefinite NPs, whose very introduction into the discourse implies their being totally affected, the already existing referents for definite NPs can be only partly affected3. After all, we have the choice of leaving them alone and doing nothing to them. This has the consequence that when they are partly affected, some other verbal measure has to be used to indicate partial affectedness, as can be seen in (14):
(14) Ta kan le yihuir zhei ben shu.
he read LE a while this CL book
'He read this book for a while.'
The verbal measure marking the event boundary is the durational expression yihuir and not the definite NP zhei ben shu. Needless to say, in (14), the book is not totally affected, since the measure yihuir implies that the book is not finished. Definite NPs can of course be totally affected if a perfectivizing complement such as wan is used. The important point however is that wan still has to be used and total affectedness cannot be inferred from the definite NP alone.
Can definite NPs themselves be pressed into service as verbal measures, necessarily indicating total affectedness, when no other verbal measures occur in the sentence? Can the contrast between the two kinds of NPs then be blurred in some cases? The anonymous reviewer for the present paper suggests, among other ideas4, something like Li&Thompson (1981)'s condition c on boundedness:
(15)c. if the verbal semantics suggests an inherent boundary
While conceding that verbal semantics may contribute to boundedness, we believe that this may not have direct bearings on the aspectual contribution of NPs. If a verb already has an inherent boundary, then it does not matter what kind of object NPs are present. Definite NPs can be compatible with boundedness but still do not contribute to it. The minimal pair in (16) suggests that definite NPs cannot be used to mark event boundaries:
(16)a. ?Ta kan le zhei ben shu, keshi mei kan wan.
he read LE this CL. book but not read finish
'He read this bok, but did not finish it.'
b. *Ta kan le yi ben shu, keshi mei kan wan.
he read LE one CL. book but not read finish
'He read a book, but did not finish it.'
Since (16a) seems to produce less of a sense of contradiction than (16b), the definite NP in (16a) then must be less than totally affected. Despite the notoriously slippery judgment concerning (16a) and (16b) individually, as evidenced by disagreements concerning similar data first brought up by Chu (1976) and Tai (1984), the contrast between the two sentences is quite clear. (16a) definitely sounds better than (16b).
In any case, the contrast between the two kinds of NPs remains regardless of whether definite NPs can contribute to boundedness. When indefinite NPs delimit events, they are always totally affected, whereas even if definite NPs can delimit events in cases of total affectedness, it is not possible for them to do so in cases of partial affectedness.
There is an assymmetry between definite and indefinite NPs that has not been explained. In the debate over whether Chinese has SVO or SOV word order, Sun&Given (1985) admits that while definite and indefinite NPs both occur overwhelmingly post-verbally, indefinite NPs are indeed less likely to occur pre-verbally than definite NPs, as was argued by Li&Thompson (1975). They have no direct explanation for this difference. Their explanation for why object NPs occur pre-vrbally at all is that the rather limited cases of pre-verbal NP objects are all contrastive and emphatic in function. Why is it then definite NPs are more likely to be contrastive/emphatic? We suggest that the assymmetry should be explained with our hypothesis that indefinite NPs are event-dependent while definite NPs are event independent.
VI. DISCOURSE CONDITIONS
The verb-indefinite NP complex is best used when the object is not focused on.
Ni chi le shenme? ---> Wo chi le yi wan fan.
Ni chi de shi shenme? ---> Wo chi de shi (yi wan) fan.
Ni zuo le shenme? ---> Wo chi le yi wan fan.
When discussing the differences between definite and indefinite NPs, it is inevitable that their different discourse properties be noticed. The most salient discourse difference between definite and indefinite NPs is that the former can be second-mentioned while the latter has to be first-mentioned. This is illustrated in the mini-discourse in (17):
(17)Wo xie le yi feng xin. Zhe feng xin hen chang.
I write LE one CL. letter this CL. letter very long
'I wrote a letter. This letter is very long.'
The definite NP in the second sentence refers to the same referent introduced by the indefinite NP in the first sentence. This 'second mention' nature of definite NPs is related to their presupposed existence mentioned in the last section.
Does this discourse difference relate to the aspectual difference between the two kinds of NPs? We suggest that the discourse property of 'second mention' is indeed linked to the aspectual properties of both NPs and verbal predicates. First of all, just as NPs can be second-mentioned, verbal predicates can be second-mentioned as well. Second-mentioned verbal predicates are understandably not aspectually salient. The parallel between nominal and verbal expressions can again be seen here. Even though second-mentioned NPs can be used with first-mentioned as well as second-mentioned verbal predicates, they themselves in either case do not contribute to the aspectual makeup of the predicates. Therefore, second-mentioned NPs are most congruent with second-mentioned verbal predicates . First-mentioned verbal predicates have to rely on other elements than second mentioned definite NPs for their aspectual contribution.
One of the structures expressly using second-mentioned verbal predicates is the SHI...DE construction, which is used for focusing on the circumstantial aspects of events, such as location, time, manner, purpose and so on. This presupposes that the events have been mentioned before, in whatever aspectual viewpoint:
(18)a. Wo zuotian kan le yi ge dianying. perfective
I yesterday watch LE one CL. movie
'I watched a movie yesterday.'
b. Wo na shi zheng zai kan yi ge dianying. imperfect
I that time right at watch one CL. movie
'I was watching a movie at that time.'
c. Wo shi zai jia kan zhe ge dianying de. SHI...DE
I be at home watch this CL. movie DE
'It is at home that I watched this movie.'
(18a) first-mentions the event of watching a movie in a perfective viewpoint. (18b) first-mentions the event in an imperfect viewpoint. But (18c) with the SHI...DE construction can be used as the continuation for either the perfective or the imperfect sentence. Hence the SHI...DE construction is inherently aspectually unmarked. Due to the discourse property of 'second mention' and the aspectual property of neutrality, we correctly predict that in the post-verbal position of SHI...DE sentences, definite rather than indefinite NPs should be used, as shown in (19):
(19)a. Wo shi zuotian kan zhei ben shu de.
I be yesterday read this CL. book DE
'It is yesterday that I read this book.'
b. *Wo shi zuotian kan yi ben shu de.
I be yesterday read one CL. book DE
'It is yesterday that I read a book.'
Other verbal measures are not allowed in the SHI...DE construction either, as (20) shows:
(20) *Wo shi zuotian kan kan/yihuir zhei ben shu de.
I SHI yesterday read read/a while this CL.book DE
'It is yesterday that I read a while this book.'
Despite the fact that second-mentioned verbal predicates in constructions such as SHI...DE take only second-mentioned NPs, it is not true that second-mentioned NPs can only occur in second-mentioned verbal predicates. After all, we can do something new and different to an entity that has been mentioned before:
(21)a. Wo qunian mai le yi suo fangzi.
I last year buy LE one CL. house
'I bought a house last year.'
b. Wo jinnian ba zhe suo fangzi xiu le yixia.
I this year BA this CL. house fix LE VM
'I did some repair work on the house this year.'
In (21b), the second-mentioned NP zhe suo fangzi 'this house' occurs after the first-mentioned verb xiu 'repair'. But since the NP does not contribute to the aspectual character of the predicate, verbal measures such as yixia 'a bit' have to be used.
VII. OTHER CONSTRUCTIONS
We have argued that while indefinite NPs contribute to boundedness, definite NPs do not. The latter and not the former however can occur in aspectually unmarked structures, such as the SHI..DE construction. In this section, I will show the contrast between these two kinds of NPs in two other constructions, namely the BA construction and the VERB COPYING construction.
The BA construction is inherently telic in modern Mandarin5. But it is also expressly used to describe what happens to entities that are presupposed to exist, such as definite NPs. Our analysis correctly predicts that, since definite NPs do not themselves contribute to telicity, some other means have to be used to meet the telicity requirement. When no such means exist, even the inherent boundary built into the lexical semantics of some verbs6, the use of verbal measure is necessary. Consider the two sets of sentences given in (22):
(22) a. ?Wo xiu le zhei dong fangzi.
I fix LE this CL. house
'I fixed this house.'
b. Wo xiu le xiu/yixia zhei dong fangzi.
I fix LE fix/a bit this CL. house
'I fixed up this house a bit.'
a'. ?Wo ba zhei dong fangzi xiu le.
I BA this CL. house fix LE
'I fixed this house.'
b'. Wo ba zhei dong fangzi xiu le xiu/yixiar..
I BA this CL. house fix LE fix/a bit
'I fixed up this house a bit.'
Just as sentence (22a) is bad due to the lack of a quantity expression, so is (22a'), the BA counterpart. On the other hand, just as sentence (22b) is good due to the presence of verbal measures, so is sentence (22b'), the BA counterpart. We would like to think that the parallel between the BA and non-BA sentences is not a coincidence. So what at first seems a peculiar characteristic of the BA construction is really rooted in the aspectual properties, or lack of which, of definite NPs.
Given the fact that BA construction is inherently telic but the SHI...DE construction is aspectually unmarked, we predict that the co-occurrence of BA and SHI...DE is not always possible7. This is indeed true, as we see in (23):
(23)a. Wo zuotian ba zhei ben shu kan le yibian.
I yesterday BA this CL. book read LE once
'Yesterday I read this book once.'
b. *Wo shi zuotian ba zhei ben shu kan yibian de.
I be yesterday BA this CL. book read once DE
'It is yesterday that I read this book once.'
c. *Wo shi zuotian ba zhei ben shu kan de.
I be yesterday BA this CL. book read DE
'It is yesterday that I read this book .'
(23b) and (23c) show that there is no way to embed the BA sentence of (23a) into a SHI...DE construction. Keeping the verbal measure of yibian, as in (23b), would violate the aspectual requirement of the SHI...DE construction; getting rid of it, as in (23c), would violate the aspectual requirement of the BA construction.
The VERB-COPYING construction can be exemplified by (24):
(24) Ta pao bu pao de hen kuai.
he run step run DE very fast
'He runs very fast'.
The syntactic analysis of such sentences has been controversial (see C.-R. Huang and L. Mangione 1985 versus J. Huang 1988), the disagreement centering on whether the pao or the kuai should be considered the head or the primary predication. The present paper does not have anything to add to this debate. It is quite clear however that the verb pao 'run' has to be aspectually unmarked. We predict that any post-verbal NPs have to be definite or generic rather than indefinite. This is exactly borne out by the minimal pair in (25):
(25)a. *Ta nian yi ben shu nian de hen laijin.
he read one CL. book read DE very enthusiastic
'He is enjoying reading a book.'
b. Ta nian (zhei ben) shu nian de hen laijin.
he read this CL. book read DE very enthusiastic
'He is enjoying reading (this book).'
From a discourse point of view, the VERB-COPYING construction is mainly used for description rather than narration, as second mention rather than discourse-initial. The preference for definite NPs therefore is not really surprising.
VIII. INDEPENDENT SUPPORT
Support for various ideas in the present proposal can be found in the literature.
The idea that NP arguments can be integral parts of verbal processes has been proposed before. Tenny (1987) argued that some NPs have the function of measuring out events. Dowty (1990) proposed the notion of 'incremental theme', which can be understood informally as NPs whose undergoing of change in a part-to-whole fashion corresponds to the part-to-whole process of accomplishment events. Verkuyl (1993) proposed that the terminative aspect of a sentence arises compositionally from the +ADD TO feature of verbs and the +SQA ( specifiable quantity) feature of nouns. What is common to these proposals is the relationship between the atemporal structure of NPs and the temporal structure of verbal predicates.
The idea that indefinite objects are event-dependent finds support from the joint effects of two proposals in the literature. One is Keenan (1984)'s proposal that patients tend not to exist independently of events; the other is Comrie (1977)'s proposal that subjects are more likely to be definite and objects are more likely to be indefinite. Putting the two proposals together, we arrive at the conclusion that indefinite objects tend to be event-dependent.
Our argument that the mutual exclusiveness between indefinite NPs and verbal measures indicates similarity in function is also supported by the patterns of co-occurrence of object NPs with perfectivizing particles in English. Hopper&Thompson (1980) cited Ellen Prince as providing the contrast given in (26):
(26)a. ?I ate up a sandwich.
b. I ate up the sandwich.
Hopper&Thompson's explanation is that the perfectivizing particle up only co-occurs with definite NPs, which are more referential, higher in transitivity and more closely associated with perfectivity than indefinite NPs. There seems to be some problem with this explanation. Their assumption is that elements similar in function can co-occur while those dissimilar in function do not. But since I ate a sandwich must be interpreted perfectively, we must assume that the indefinite NP can contribute to boundedness. Why is it then that the perfectivizing particle up and the boundedness-contributing indefinite NP do not co-occur? We think the opposite explanation seems to be more reasonable. We assume that similar elements do not co-occur: up is mutually exclusive with indefinite NPs, since both contribute to boundedness; up co-occurs with definite NPs, since definite NPs do not contribute to boundedness.
IX. CONCLUSION
In this paper, we have further explored the connection between atemporal nominal expressions and the temporal notion of aspect. Based on the mutual exclusiveness between verb measures and indefinite NPs, I suggest that indefinite NPs can be virtual verbal measures as well. On the other hand, the possible co-occurrence of verbal measures and definite NPs has been taken as indication that definite NPs do not serve as verbal measures. Event-dependent indefinite NPs are totally affected, while event-independent definite NPs can be partially affected. The aspectual differences between the two kinds of NPs are shown in a number of constructions with built-in aspectual requirements, such as BA, SHI...DE and VERB COPYING.
The present analysis has implications not only for aspectual research; it also has consequences for the issue of transitivity and the semantic and discourse roles of noun phrases in general. Aspectually, the case of Chinese confirms the previous observations concerning the contribution of NPs to the aspectual character of a sentence. But the difference between definite and indefinite NPs in aspectual behavior should open up new avenues for investigation. Going beyond aspect, the present analysis challenges the well-known view on transitivity proposed by Hopper&Thompson (1980), which takes definite NPs to be higher than indefinite NPs in transitivity, being in the foreground, totally affected and more likely to occur in perfective contexts. The analysis in this paper suggests exactly the opposite. Finally, the present research has taken an initial step towards relating the discourse property of first and second mention and the aspectual contribution, or lack of which, of noun phrases.
NOTES
* I am grateful to the organizing committee of The Symposium on the Referential Properties of Noun Phrases for allowing my paper to be presented in my absence due to exceptional circumstances. I thank Dr. Li Ping to present the paper on my behalf. I also thank Dr. Gu Yang and an anonymous reviewer for their constructive feedback. I alone am responsible for any remaining flaws.
1The similarity between generic and definite NPs is shown in the following example:
?Wo zuotian kan le (zhei ben) shu.
I yesterday read LE this CL. book
'I read (this book ) yesterday.'
2The addition of the perfectivizing wan 'finish' in the positive sentence in (13b) is due to the fact that definite NPs cannot by themselves render events perfective. But the contrast between (13a) and (13b) is still clear despite the fact that the two sentences do not exactly constitute a minimal pair.
3I use 'affected' here to stand for the more accurate but also more cumbersome 'affected / effected / consumed or destroyed'.
4The reviewer also points out that topicalization and modification of definite NPs can improve some otherwise awkward sentences. The original examples from the reviewer are given below:
Na chu xi, women zuotian kan le.
Na jian yifu, wo zuotian xi le.
Wo zuotian kan le ni tuijian de na ben shu.
It seems to me that even with topicalization and modification, the sentences are not as good as ones that include some extra verbal measures. The topicalized sentences are not very different in acceptability from their BA counterparts, which are, as argued in the paper, not very good without some extra verbal measures. It is not obvious either that modification significantly improves matters. Nor is it clear to the present author why it should be so, if the reviewer's intuition turns out to be confirmed by other people. In any case, sentences with definite NPs, with or without topicalization and modification, remain to be less acceptable than ones with indefinite NPs.
5In older forms of Chinese, sometimes heard in operas, the telicity requirement apparently does not hold. One line from the Cultural Revolution Era Peking opera Taking the Mount Tiger by Strategy (Zhi Qu Weihushan) shows this clearly:
Zhi pan zhe neng zai ren qian ba hua jiang.
only hope ZHE can at people front BA words speak
'I only hope that I can speak in front of people.'
There is no telicity-contributing element in the whole verbal predicate ba hua jiang 'speak'.
6This is pointed out by the anonymous reviewer, who uses sentences with the verb chi 'eat' and chai 'dismantle' to show that the semantics of the verbs are such that no other telicity-contributing element is necessary to complement the definite NPs:
Ta ba na ge pingguo chi le.
Ta ba na dong fangzi chai le.
7SHI...DE and BA do combine in sentences using perfectivizing RVCs, such as the one below:
Wo shi zuotian ba zhe ben shu kan wan de.
I SHI yesterday BA this CL. book read finish DE
'It is yesterday that I finished reading the book.'
It is clear that RVCs behave rather differently from verbal measures, virtual or literal. We will not go into RVCs in this paper.
REFERENCES:
Carlson, L. 1981. Aspect and quantification. in Syntax and Semantics. V.14. Tense and Aspect. (eds.) by Tedeschi&Zaenen. New York: Academic Press.
Chu, Chuancey. 1976. Some semantic aspects of action verbs. Lingua 40.43-54
Comrie, Bernard. 1977. Subject and direct object in Uralic languages: A functional explanation of case-marking systems. Etudes Finno-Ougriennes 12.5-17.
Dowty, David. 1990. Thematic roles and argument selection. Language 67.547-619.
Hinrichs, Erhart. 1985. A Compositional semantics for aktionsarten and NP reference in English. Columbus: Ohio State University dissertation.
Hopper, Paul and S.A. Thompson 1980. Transitivity in grammar and discourse. Language 56. 251-299.
Hopper, Paul. 1982a. Aspect between discourse and grammar. Introductory essay for Hopper (ed.) 1982b.
Hopper, Paul. 1982b. Tense-Aspect: Between Semantics and Pragmatics. Amsterdam: John Benjamins.
Huang, C.-R. and L. Mangione. 1985. A Reanalysis of de:Adjuncts and subordinate clauses. Proceedings of the 4th West Coast Conference on Formal Linguistics, 80-91. Stanford: Stanford Linguistics Association.
Huang, C.-T.James. 1988 Wo pao de kuai and Chinese phrase structure. Language 64.274-311.
Keenan, Edward. 1984. Semantic correlates of the ergative/absolutive distinction. Linguistics 22.197-223.
Krifka, Manfred. 1987. Nominal Reference and Temporal Constitution: Towards a Semantics of Quantity. FNS-Bericht 17. Universitat Tubingen.
Li, Charles & S.A. Thompson. 1981. Mandarin Chinese: A Functional Reference Grammar. Berkeley:University of California Press.
Mourelatos 1981. Events, processes and states. in Syntax and Semantics. V.14. Tense and Aspect. (eds. by Tedeschi&Zaenen. New York: Academic Press.
Smith, Carlota. 1991. The Parameter of Aspect. Dordrecht: Kluwer Academic Publishers.
Sun, C-F & T.Givon 1985. On the So-called SOV hypothesis. Language.
Tai, H.-Y. James. 1984. Verbs and times in Chinese: Vendler's four categories. Papers from the Parassession on lexical semantics. Chicago Linguistics Society.
Tedeschi, Phillips.&A. Zaenen (eds). 1981. Syntax and Semantics: Tense and Aspect. New York: Academic Press.
Tenny, Carol. 1987. Grammaticalizing Aspect and Affectedness. MIT dissertation.
Ter Meulen, Alice. 1984 Events, quantity, individuals. Varieties of Formal Semantics. Dordrecht: Foris Publications.
Verkuyl, H.J. 1972. On the Compositional Nature of the Aspects. Dordrecht: Reidel Publishing Company: .
Verkuyl, H.J. 1993. A Theory of Aspectuality. Cambridge: Cambridge University Press.
Back to