As is the case with verb-complements, temporal reference of Chinese relative clauses has received little attention in the literature. So the following discussion might be the first attempt to explain temporal reference of Chinese relative clauses.
I will start with a comparison between relative clauses and verb-complements. As noted, temporal reference of complement clauses can generally be determined by the temporal relation they bear to the matrix verb. This temporal relation can be seen as a semantic restriction that the matrix verb imposes upon the complement clause. Unlike complement clauses, relative clauses are not arguments of matrix verbs. So it is impossible for a matrix verb to directly impose a temporal restriction upon a relative clause. To illustrate, consider the following two examples.
(111) a. Ta mai-le Zhangsan xie de shu
he buy-ASP Zhangsan Zhangsan write REL book ‘He bought a book/books that Zhangsan wrote.’
b. Mama na-zou-le wo nan-pengyou ji gei wo de xin mother take-away-ASP my boy-friend send to me REL letter
‘Mother took away letters/the letters that my boy sent to me.’
Because the arguments of the verb mai ‘buy’ and na-zou ‘take away’ are shu ‘book’
and xin ‘letter’, respectively, rather than Zhangsan xie de ‘which Zhangsan write’ and wo nanpengyou ji gei wo de ‘which my boyfriend send to me’, there cannot be a direct temporal relation between the matrix verbs and the relative clauses. Notice also that the relative clauses in (111) do not contain any time adverbial or aspect marker. So temporal reference for these relative clauses cannot be attributed to time adverbials or aspect markers. If temporal reference of relative clauses such as those in (111) cannot be inferred from a lexical temporal relation as in the case of verb-complements or from time adverbs and aspect markers, how is temporal reference for these relative clauses resolved?
One possible hypothesis is that the temporal reference of a relative clause is determined by a higher clause that dominates it. I will call this hypothesis Temporal Control Hypothesis (TCH). For instance, due to the use of –le in (111a) and (111b), the event time of the matrix verb refers to a past interval and therefore the relative clause also has a past tense interpretation. (112a) and (112b) below, where the matrix clause contains a modal auxiliary indicating a future time, may support the same hypothesis. According to TCH, the relative clauses in both (112a) and (112b) should have a future tense interpretation just like the matrix clauses. Indeed, one can
felicitously utter (112a) and (112b) if the events denoted by the relative clauses take place in the future.
(112) a. Ta hui mai Zhangsan xie de shu (ma) he will buy Zhangsan write REL book Q
‘(Will) he (will) buy books that Zhangsan wrote/will write?’
b. Mama hui na-zou wo nan-pengyou ji gei wo de xin mother take-away my boy-friend send to me REL letter
‘Mother will take away the letter that my boy sent/will to me.’
Notice, however, that (112a) and (112b) are also felicitous in a situation where the relative clauses have a past tense interpretation. Such interpretations of the relative clauses in (112a) and (112b) do not conform to the prediction made by the TCH.22 One way out of this problem is to say that the object NPs may optionally undergo quantifier raising (QR). Once an NP containing a relative clause has undergone QR, the relative clause will be outside the scope of the original dominating clause and hence its tense interpretation is no longer dependent upon the higher clause. Suppose we further assume that a relative clause not within the scope of the matrix clause receives its tense interpretation as if they were unembedded, then we can get a past tense interpretation for the relative clauses in (112). So the past tense interpretations of (112a) and (112b) are not problems to TCH.
Another example that might be used to support TCH is (113). In this example, the matrix verb is an individual-level predicate that is inherently generic according to Chierchia (1995). Very interestingly, the relative clause Zhangsan xie de ‘which Zhangsan writes’ in this example has a generic reading, in addition to a past tense reading. That is, (113) can be construed as: I like any poem that Zhangsan writes at any time. Again, we see that the temporal reference of a relative clause seems to depend upon that of the matrix clause.
(113) Wo xihuan Zhangsan xie de shi I like Zhangsan write REL poem ‘I like poems that Zhangsan writes.’
Although TCH seems quite successful in accounting for the above data, it is not without problems. The first problem is that TCH is a device designed merely for temporal reference of relative clauses rather than a general hypothesis for temporal
22 In fact, (112a) can also be uttered in a situation where Zhangsan is still writing the book. I will come back to this reading later.
reference of all subordinating clauses. For example, it does not predict that though the matrix clauses in (114) and (115) have a future and present tense interpretation, respectively, the embedded clauses must have a past tense interpretation.
(114) Ta yiding hui fouren shi ta nazou wo de shu he definitely will deny be he take-away I DE book
‘He will definitely deny that it is he that ha took away my book.’
(115) Zhangsan renwei Lisi shuo huang Zhangsan think Lisi tell lie ‘Zhangsan thinks that Lisi told a lie.’
Notice that the past tense interpretation of the embedded clauses in (114) and (115) cannot be rescued in the same way as we did for the relative clauses in (112), because verb complements are not NPs and hence will not undergo QR. Likewise, as will be discussed later, the tense interpretations of many adverbial clauses are not dependent upon the tense interpretations of the matrix clauses. Therefore, TCH is not a general hypothesis.
In addition to the problem of generality, TCH has empirical problems. In our above discussion, relative clauses are contained in NPs without a determiner. The addition, of a determiner, however, may change the interpretation of a relative clause. Compare (116a) and (116b) with (112b) and (113).
(116) a. Mama hui na-zou wo nan-pengyou ji gei wo de na feng xin mother take-away my boy-friend send to me REL that CL letter
‘Mother will take away the letter that my boy sent to me.’
b. Lisi xihuan Zhangsan xie de na shou shi Lisi like Zhangsan write REL that CL poem ‘Lisi likes the poem that Zhangsan wrote.’
Unlike (112b) and (113), (116a) is not ambiguous between a future tense and past tense interpretations. With the addition of the determiner na feng ‘that CL’, the future tense interpretation— the one predicted by TCH--disappears. Similarly, after the insertion of the determiner na shou ‘that CL’, (116b) does not have a generic reading.
Instead, the relative clause now only has a past tense interpretation. The examples in (116a) and (116b) clearly show that temporal reference of Chinese relative clauses is not a pure matter of temporal control. The syntactic forms of the NPs also matter.
Given the above problems, I would like to pursue another approach to temporal reference of relative clauses in terms of pure semantics and pragmatics inference. To
begin with, I will make some comments on the semantics of bare nouns in Chinese.
Chinese bare nouns may receive various interpretations depending upon the contexts in which they appear. Here are some examples.
(117) Wo mai-le shu (le) Existential or Definite Interpretation I buy-ASP book ASP
‘I bought books/the book(s).’
(118) Shu bu jian le Definite Interpretation book not see ASP
‘The book(s) is/are missing.’
(119) Wo xihuan shu Generic interpretation I like book
‘I like books.’
The interpretational possibilities of Chinese bare nouns are much like English bare plurals as discussed in Carlson (1977) except that the former may get an additional definite reading (cf. Lin 1999). When a relative clause is used to modify a bare noun, the interpretational possibilities are the same. NPs of the form ‘relative clause + noun’
can therefore be analyzed as complex bare nouns.
Some further assumptions are needed before I can explain how relative clauses in Chinese obtain their temporal interpretations. Earlier I said that there is no direct selectional restriction between a matrix verb and a relative clause. However, a verb can directly impose some restriction upon its object NP argument that contains a relative clause. Take the verb mai ‘buy’ for instance. If you want to buy something, that something must have already existed before the time of the buying event or the action of buying is simply impossible.23 Moreover, the life time of the thing that is bought seems to always longer than the time at which the buying event holds. Thus, instead of the usually simplified translation of buy as in (120), I translate it as (121).
(120) [[buy]] = λxλyλtbuy’(x)(y)(t)
(121) [[buy]] = λxλyλt1∃t2[buy’(x)(y)(t1) ∧ EXIST(x)(t2) ∧ t1 ⊆ t2]
23 The existence presupposition sometimes does not seem to hold as in the following example.
(i) Wo xiang mai yi-ben youguan wuaixing ren de shu I want buy one-CL about alien person DE book
‘I want to buy a book about alien people.’
The loss of the existence presupposition is due to the fact that the existential operator is embedded within the operator xiang ‘want’.
In (121) the predicate EXIST is introduced to predicate of the internal argument of buy and a further condition says that the buying event must be included within the interval at which the internal argument exists. The inclusion condition guarantees that the thing that is bought must exist before it is bought and that it can still exist after the buying event is completed. I will not try to discuss whether the existence predicate and the inclusion condition is a presupposition of the verb buy ‘buy’ or part of the truth conditions. It suffices for the purpose of this paper to assume that it is part of the truth conditions.
On the other hand, there are some other predicates which require that their internal arguments exist only after the event denoted by the verb is completed. Verbs of creation are of this type. Roughly following Kratzer (1994), I translate this type of verb as follows, where e represents the event argument.24
.
(122) [[wr ite]] = λxλyλeλt1∃t2[write(x)(y)(e)(t1) ∧ Exist(x)(ftarget(e))(t2) ∧ t2 >< t1]
In plain English, (122) is intended to mean that an object x exists in the target state of a writing event right after the writing event culminated.
As for the semantics of relative clauses, I will assume with Heim and Kratzer (1998) and many others that they translate as predicates and the combination of a relative clause with a head noun translates as a conjunction of both.
Now let us reconsider (111a). The object NP in this example has an indefinite reading. On this reading, (111a) is (roughly) translated as (123) within my system, ignoring irrelevant details.
(111a) Ta mai-le Zhangsan xie de shu he buy-ASP Zhangsan write ASP book
‘He bought books/a book that Zhangsan wrote.’
(123) ∃t1∃t3∃t4∃x∃e[book’(x) ∧ write’(x)(Zhangsan’)(e)(t3) ∧ Exist(x)(ftarget(e))(t4) ∧ t4 ><t3∧ buy’(x)(he’)(t1) ∧ t1 ⊆ t4∧ Now > t1]
From my previous discussion, it should be clear by now that the event time t1 of the buying event must precede the speech time. Now if the buying event must be included within the interval t4 at which the book exists, the book must have already existed
24 von Stechow (2000) has a very detailed review of all the literatures concerning the representation of creation verbs. He has pointed out that Kratzer’s semantics for creation verbs is inadequate in some respects. However, in this paper I still assume Kratzer’s style of representing the semantics of creation verbs, because I do not want to go into too many technical details that are irrelevant to the point that I want to make. In any case, whatever version of semantics for creatin verbs I adopt will not affect my point. So readers who are interested in the semantics of creation verbs are referred to von Stechow’s work.
before the speech time too. It follows that the writing event must precede the speech time as well, because the writing event must precede the existence of the book.
Therefore, the relative clause in (111a) has a past tense interpretation. The truth conditions in (123) are equivalent to the following diagram.
Exist(t4)
(124) ----|/////////// ---|/////////|---|--->
write(t3) buy(t1) Now
Next, let us reconsider (112a), which differs from (111a) in having a future modal auxiliary in the matrix clause instead of the aspect marker –le. The truth conditions of (112a) are minimally different from (123). We only need to reverse the temporal relation between Now and t1. The truth conditions of (112a) is (125).
(125 ) ∃t1∃t3∃t4∃x∃e[book’(x) ∧ write’(x)(Zhangsan)(e)(t3) ∧ Exist(x)(ftarget(e))(t4) ∧ t4 ><t3∧ buy’(x)(he)(t1) ∧ t1 ⊆ t4∧ Now < t1]
According to (125), the buying event, i.e., the time t1, must be located after the speech time, i.e., in the future and is included within the interval t4 at which the book that Zhangsan writes exists. To satisfy these two conditions, there are three possibilities.
The first possibility is that the writing event begins and ends before the speech time but the buying event is located after the speech time. Another possibility is that the writing event and the buying event are both located in the future. Finally, the writing event might begin before the speech time, continues to the future and end in the future but before the buying event. In these three possibilities, the writing event must end before the buying event or the condition “t1⊆ t4” will not be satisfied. The three possibilities are represented as follows.
Exist(t4)
(126) a. ---|///////////// ---|---|////////|--->
write(t3) Now buy(t1) Exist(t4)
b. --|----|/////////// ---|////////|--->
Now write(t3) buy(t1)
write(t3) Exist(t4)
c. ---|//////|////// ---|/////////|--->
Now buy(t1)
These three different possibilities to satisfy the requirement that the buying event must be included within the book’s life span and the requirement that the buying event takes place in the future predicts that the relative clause in (112a) is compatible with a situation where the event denoted by the realtive clause takes place in the past, in the future or is on-going. Indeed, this seems to be true.
The reading represented by (126c) is worth particular mentioning here. Earlier I mentioned that the relative clause in (112a) has both a future and past tense interpretation but in fact the sentence (112c) is also fully compatible with a situation in which he will buy the book that Zhangsan is still writing at the speech time. This reading is a great problem with TCH, because neither TCH nor quantifier raising predicts this reading.
Next, let us now reconsider (113), reproduced below as (127). Recall that this example seems to constitute a piece of evidence in support of TCH. In what follows, I will show that it is not necessary to resort to TCH to explain the fact that a relative clause contained in a complex bare NP has a generic reading when the matrix predicate is an individual-level predicate. The generic reading of the relative clause in (127) can be derived from a device independently needed by universal grammar.
(127) Wo xihuan Zhangsan xie de shi I like Zhangsan write REL poem ‘I like poems that Zhangsan writes.’
To begin with, I assume Chierchia’s (1995) treatment of individual-level predicates as generic polarity items which must be licensed by a Gen operator. The Gen operator is like an adverb of quantification in that it will partition the clause containing it into restriction and scope and can freely bind any free variable. Moreover, when there is no overt restriction on individual-level predicates, a general locative relation in is the restriction. Take (128a) as an example. It translates as (128b) on Chierchia’s analysis.
(128) a. John knows Latin.
b. Gen s [in’(j,s)][know’(j,L,s)]
In plain English, what (128b) says is that whenever John is or might be located, he knows Latin.
It has been argued that bare plurals, like indefinites with the form ‘a + N’, may introduce free variables bound by an adverb of quantification (See Wilkinson (1991) for instance.) Therefore, a sentence such as (129a) may get a translation like (129b)
under Chierchia’s analysis.
(129) a. John likes poems.
b. Gen x,s [poem’(x) ∧ in’(j,s) ∧ in’(x,s)][Like’(j,x,s)]
Since the Gen operator induces universal readings for the variable x, so the bare plural poems in (129) has a universal force.
Returning to the Chinese example (127), we can assume that Chinese bare nouns may introduce free variables just as English bare plurals. Since an NP of the form
‘relative clause + bare noun’ can be seen as a complex bare noun, it may introduce a free variable just like a normal bare noun except that the variable introduced has an additional predicate contributed by the relative clause to restrict it. On the above assumptions, (127) can be analyzed as follows. Suppose that in addition to the in restriction, the Gen operator for the sentence (127) also selects the object NP as its restriction. Then, the relative clause contained in the object NP will become part of the restriction, because it is syntactically part of the object NP. Thus, the translation of (127) should be something like (130).
(130) Genx,s,s’ [poem’(x) ∧write’(Z,x,s’) ∧ in’(Z,s) ∧ in’(x,s)][like’(I’,x,s)]
In (130), the Gen operator binds not only the variables x and s but the variable s’ as well.25 Given that the Gen operator is roughly equivalent to a universal quantifier, the meaning of (130) is something like the following: For any poem x that Zhangsan writes in any situation s’ and for any situation s in which the poem x and I are located, I like x in s. Because the situation variables of the predicate xie ‘write’ and xihuan
‘like’ are bound by the Gen operator, both the matrix and relative clause obtain a generic reading. On this analysis, the fact that the relative clause in (127) has a generic reading is nothing but a side effect of the object NP being quantified over by the Gen operator.
I should note at this point that the situation variable s and s’ in (130) can be replaced with time variables t and t2 without affecting the original conclusion. One can also keep both the situation variable as well as the time variable. The result is still the same. Although this minor revision does not change anything, I mention it in order to make the picture more congruent with the theory of Chinese temporal reference that
25 If Gen also binds s’, then Gen binds a variable in the restriction but does not bind the same variable in the scope. This problem can be avoided by the property of Conservativity associated with
quantifiers/determiners. (See Barwise and Cooper (1981).) According to Conservativity, Q(A,B) is equivalent to Q(A, A∩B). Thus, when the restriction is reproduced in the nuclear scope of the quantifier, the situation variable s’ will appear in both the restriction and the scope. To simplify the
I am proposing.
Now recall that when an extra determiner such as na ‘that’ is added to modify the head noun of the object NP in (127), the relative clause must be understood as having a past tense reading instead of a generic reading. I reproduced the relevant example as (131).
(131) Wo xihuan Zhangsan xie de na-shou shi I like Zhangsan write REL that-CL poem ‘I like the poem that Zhangsan wrote.’
Why must the relative clause have a past tense interpretation? The answer seems to have to do with the semantics of the definite determiner na ‘that’. When na ‘that’ is combined with a common noun, it implies existence of an individual satisfying the description of the common noun. Such a property is known as existence presupposition of definite descriptions in the literature. (See Heim (1982), for instance.) Applying this property to the definite description Zhangsan xie de na-shou shi ‘that poem that Zhangsan wrote’ in (131), this means that there must exist an individual that satisfies both the property of being a poem and the property of being written by Zhangsan. If a poem has already existed and is the product of Zhangsan’s writing, then the writing event must take place before the speech time, because verbs of creation such as write has an end product only when the event described by the
Why must the relative clause have a past tense interpretation? The answer seems to have to do with the semantics of the definite determiner na ‘that’. When na ‘that’ is combined with a common noun, it implies existence of an individual satisfying the description of the common noun. Such a property is known as existence presupposition of definite descriptions in the literature. (See Heim (1982), for instance.) Applying this property to the definite description Zhangsan xie de na-shou shi ‘that poem that Zhangsan wrote’ in (131), this means that there must exist an individual that satisfies both the property of being a poem and the property of being written by Zhangsan. If a poem has already existed and is the product of Zhangsan’s writing, then the writing event must take place before the speech time, because verbs of creation such as write has an end product only when the event described by the