CHAPTER 2 GRAMMATICAL FUNCTIONS IN AN LFG
3.1 E VIDENCE FOR S UBCATEGORIZED E LEMENTS
The notion of subcategorization is a crucial one in most modern linguistic theories. The term standardly refers to differences among verbs as to the syntactic nature of the various constituents they co-occur with to form a complete sentence. However, in Chinese it is not always straightforward to determine whether a constituent is subcategorized for by a verb. In other words, not all arguments are clearly distinguishable from adjuncts. Although this may be a relatively easy matter in many European languages like English and French where such subcategorized constituents often appear obligatorily, in languages like Chinese and Japanese where such items may appear optionally, decisions about the subcategorization status are consequently harder (Ernst 1989). While in the linguistic literature there is no general agreement as to how the difference between subcategorized constituents and
non-subcategorized ones should be characterized in theoretical terms, there are a number of well-accepted, rule-of-thumb syntactic and semantic tests available which serve to make the distinction (e.g. Pollard and Sag 1987, Ernst 1989, Bresnan 1987, Starosta 1988, and Huang 1989a). Many of the following criteria serving to make this distinction are based on the notion that a verb's lexical entry is the proper repository for idiosyncratic information about the verb and elements it co-occurs with. It should also be noted that most of these criteria should not be taken to be sufficient conditions or definitive tests individually; rather, a stronger case can be made when they are used in conjunction with one another. We will now go through these conditions or tests and illustrate with examples in Chinese.
3.1.1 Obligatory Co-occurrence
A constituent is generally considered as subcategorized for by a verb if its co-occurrence with the verb is obligatory. Therefore, if such a constituent is missing, the sentence is ungrammatical or its acceptability is clearly and heavily dependent upon the existence of an appropriate discourse context.
However, this test, which probably provides the strongest, if not sufficient, evidence for subcategorization, is not significantly relevant for our purpose in studying verb subcategorization in Chinese, where, as mentioned before, verbs allow great freedom for their subcategorized elements to be optional.
For instance, in the following simple dialogue, the entirely acceptable response can be a "naked" verb stripped of all arguments discoursally recoverable.
1. Q: Ni3 ba3 qian2 cun2 zai4 yin2hang2-li3 le ma1?
you BA money deposit at bank inside LE MA
‘Did you deposit the money in the bank?’
A: (Wo3) (ba3 qian2) cun2 (zai4 yin2hang2-li3) le.
I BA money deposit at bank inside LE
‘I deposited the money in the bank.’
However, contrary to the common belief that in Mandarin Chinese there are no such cases of obligatory co-occurrence (e.g., Ernst 1989), certain Chinese
VERBSUBCATEGORIZATIONINMANDARINCHINESE 71
verbs, not necessarily small in number, do require their arguments, at least the ones in the predicate, to be overtly present, regardless of what discourse context they actually appear in. We observe that they include verbs like bei4 or zao1shou4 'suffer or incur (adversely)', ting1qi3lai3 'sound', and several other verbs that require a pivotal construction, [V + NP + VP], or [V + OBJ + XCOMP] in LFG terms, such as shi3de2 'make, force'. All the following sentences 1a-d below are thus ungrammatical without the arguments in parentheses, given any discourse context.
2. a. Wo3 bei4 *(ni3 da3).
I BEI you hit
‘I was hit by you.’
b. Ta1 kan4qi3lai3 *(hen3 jian4kang1).
he look very healthy
‘He looks very healthy.’
c. Wo3 po4shi3 *(ta1 *(qu4)).
I force he go
‘I force him to go.’
d. *(Ni3 de gu4shi4) shi3de2 *(ta1 *(hen3 gan3dong4)).
you DE story make he very moved
‘Your story makes him very moved.’
Note that the notation ...*( α )... means that the string is unacceptable when α is optional; in other words, α is required for the string to be well-formed.
Evidence of obligatory co-occurrence in a grammar of Chinese therefore seems to be a sufficient condition for an element to be considered subcategorized-for, but it is definitely not a necessary one, for most of verbs do not require obligatory co-occurrence of their arguments.
3.1.2 Ontological Necessity
Elements that are subcategorized for by a verb are usually associated with semantic roles that are ontologically necessary for a complete interpretation of
the sentence. Thus, if a verb presupposes a certain semantic role normally expressed by a subcategorized constituent, even when the constituent is missing, such a constituent is still considered subcategorized for by the verb.
Such sentences when in isolation thus leave one feeling odd due to their semantic incompleteness. This test is also relatively difficult to enforce in Chinese, again due to the freedom of optional, discoursally recoverable arguments. Furthermore, the ontological nature of this criterion also makes it notoriously difficult to apply. For example, since all events must occur at certain place and time, could we thus assume that all verbs presuppose a location and therefore subcategorize for a locative constituent syntactically?
To allow assumptions like this would make the notion of subcategorization useless, so this criterion has to be applied in conjunction with other syntactic principles, which will be described below. Thus, while ontological necessity is a necessary condition for a constituent to be considered as subcategorized-for, like most other criteria below, it is not a sufficient one.
3.1.3 Constancy of Semantic Content
Non-subcategorized, adjunctive constituents can generally occur with a wide, relatively unrestricted range of head verbs while making a rather uniform contribution of semantic content to that of the sentence;
subcategorized constituents, however, often may only co-occur with a semantically and/or syntactically or even arbitrarily restricted class of head verbs while their contribution to the semantic content is often idiosyncratically dependent upon the particular verb. Let's examine the following examples.
3. a. Wo3 gen1 ta1 pao3bu4/wan2shua3/you2yong3/he1jiu3.
I with he jog play swim drink
‘I jog/play/swim/drink with him.’
b. Wo3 gen1 ta1 wei2xiao4.
I at he smile
‘I smile at him.’
VERBSUBCATEGORIZATIONINMANDARINCHINESE 73
c. Wo3 gen1 ta1 hen3 ke4qi4.
I to he very cordial
‘I am very cordial to him.’
d. Wo3 gen1 ta1 zu1 fang2zi.
I from he rent house
‘I rent a house from him.’
In 3a, the constituent of gen1 ta1 is a comitative adjunctive PP and its comitative semantic content is constant regardless of what head verb it co-occurs with. In contrast, in 3b-d, where the same italicized constituent is considered as subcategorized for by the individual verbs, precisely what semantic contribution to the sentence this PP makes is dependent upon the particular verb of that sentence. As a matter of fact, 3b-d all have a comitative reading as well, though such a reading is weak. This of course also comports with this current criterion and indicates that [gen1 NP] is a comitative adjunctive PP, with a constant semantic content, which may appear with a wide range of head verbs; yet, it may also be subcategorized for by a certain subclass of verbs and in this case its semantic content is dependent on the individual verb and thus the more idiosyncratic readings of 3b-d.
3.1.4 Distributional Restrictions
It is widely assumed that subcategorized constituents are usually restricted in their distribution in the constituent structure; non-subcategorized adjuncts, on the other hand, generally have a freer distribution. We shall illustrate with the following examples.
4. a. Zai4 Tai2bei3, ta1 chang2 chi1-guan3zi.
at Taipei he often eat restaurant
‘In Taipei, he often eats out.’
b. Ta1 zai4 Tai2bei3 chang2 chi1-guan3zi.
he at Taipei often eat restaurant
‘In Taipei, he often eats out.’
c. Yin1wei4 ta1 bu4 lai2, wo3 sheng1qi4.
because he not come I angry
‘Because he is not coming, I am angry.’
d. Wo3 sheng1qi4 yin1wei4 ta1 bu4 lai2.
I angry because he not come
‘I am angry because he is not coming.’
a' Wo3 ba3 ta1 da3 le.
I BA he hit LE
‘I hit him.’
b'*Ba3 ta1 wo3 da3 le.
BA he I hit LE
‘I hit him.’
c' Wo3 da3du3 ta1 bu4 lai2.
I bet he not come
‘I bet that he is not coming.’
d'*Ta1 bu4 lai2 wo3 da3du3.
he not come I bet
‘I bet that he is not coming.’
The italicized constituents in 4a-b, a PP, and in 4c-d, an S, are both adjuncts to the head verbs and thus show a greater freedom in their syntactic distribution. In comparison, the subcategorized elements in 4a'-d' are rather restricted in their syntactic distribution and violations of such restrictions lead to ungrammaticality. However, this condition is neither necessary nor sufficient for a constituent to be considered subcategorized-for. It merely describes a tendency.
3.1.5 One per Sentence
More than one instance of the same adjunct type may co-occur with the same head verb in a single clause; however, at most one instance of a
VERBSUBCATEGORIZATIONINMANDARINCHINESE 75
particular type of subcategorized constituent may appear with a head verb.
Starosta (1988), in his Lexicase theory, distinguishes the adjunctive use of PP's as "outer case" and PP's that are subcategorized for as "inner case." To borrow his terminology, while theoretically there is no limit on the number of instances of an outer case in a clause, there should be at most one instance of each inner case; this is known as the "One-per-Sent" Principle in Lexicase. In LFG, the same effect is achieved by the fact that all subcategorizable grammatical functions have to obey the Functional Uniqueness Condition, while the class of ADJUNCTS does not. Accordingly, 5a-b below are acceptable because the italicized constituents are adjuncts while 5c-d are unacceptable because both of the italicized elements have to be taken as subcategorized-for constituents of the same type, or the same "inner case,"
and the "One-per-Sentence" condition is thus violated. Yet, 5e is good, for although there are three instances of locative PP's, only the postverbal one is the inner locative case subcategorized for by the verb shui4 'sleep', while the preverbal two are adjuncts.
5. a. Zai4 Tai2wan1, ta1 zai4 Tai2bei3 you3 fang2zi.
at Taiwan he at Taipei have house
‘In Taiwan, he has houses in Taipei.’
b. Ta1 wei4 guo2jia1 wei4 ren2min2 xi1sheng1 sheng1ming4.
he for nation for people sacrifice life
‘He sacrificed his life for the nation for the people.’
c.*Wo3 ba3 ta1 ba3 ni3 da3 le.
I BA he BA you hit LE
‘*I hit you, him.’
d.*Wo3 shui4 zai4 chuang2-shang4 zai4 wu1zi-li3.
I sleep at bed top at house inside
‘*I sleep on the bed, in the house.’
e. Zai4 na4 shi2 zai4 wu1zi-li3 wo3 shui4 zai4 chuang2-shang4.
at that time at house inside I sleep at bed top
‘At that time, in the house I slept on the bed.’
3.1.6 Proximity to the Head
The general assumption here is that a subcategorized element tends to appear closer to the verb in proximity than adjuncts. Thus, two useful tendencies can be deduced: a) an element is likely to be subcategorized if it appears closer to the verb than another element which is independently known to be subcategorized; b) an element tends to be adjunctive if it occurs further away from the verb than another constituent known to be an adjunct. These tests, as the way in which they are stated, describe a tendency, not an absolute distinction.
3.1.7 Semantic Selectional Restrictions
While the head verb usually imposes semantic selectional restrictions on its arguments, it rarely does so on adjunctive constituents. This test is closely related to the criterion of constancy of semantic content above.
3.1.8 Possible Internal Gaps
Adjuncts in general tend to disallow unbound gaps; sentential complements, by contrast, allow such long-distance gaps. In the following two noun phrases, 8a's ungrammaticality is thus caused by the relativized subject which is in an adjunctive clause of jia3ru2 'if'. The relativized subject in 8b, however, is in a sentence complement of the verb shuo1 'say' and thus causes no problem.
8. a.*[jia3ru2__bu4 lai2] wo3 hui4 sheng1qi4 de ren2 if not come I will angry de person
‘*The person who [if__is not coming] I will be angry.’
b. Ni3 shuo1 [__bu4 lai2] de ren2.
you say not come DE person
‘The person who you say [__ is not coming].’
VERBSUBCATEGORIZATIONINMANDARINCHINESE 77
3.1.9 Ability of Being an Indirect Question
An adjunctive clause with a wh-element in it can never be interpreted as an indirect question; rather, the entire sentence has to be interpreted as a question.
Yet, a subcategorized predicative complement with a wh-element may be interpreted as an indirect question, though it is also possible for the whole sentence to be interpreted as a question. Logically then, like obligatory co-occurrence, evidence from indirect question constitutes another sufficient but not necessary condition for a constituent to be considered subcategorized-for. Thus, since 9a is ambiguous in that its sentential complement may be a direct or indirect question and that 9b only allows the interpretation of an indirect question, both of the complements in 9a and 9b have to be subcategorized for by the verbs; while neither of the italicized clauses of 9c-d can be an indirect question and thus their status regarding subcategorization has to be determined by other criteria.
9. a. Ta1 xuan1bu4 ming2tian1 shei2 hui4 lai2./?
he announce tomorrow who will come
‘He announces who is coming tomorrow.’
OR ‘Who is coming tomorrow, that he announced?’
b. Ta1men yan2jiu4 ru2he2 shi3 jing1ji4 qi3fei1.
they study how make economy take off
‘They study how to make the economy take off.’
c. Ji4ran2 shei2 bu4 lai2, ta1men dou1 bu4 lai2?
since who not come they all not come
‘All of them won't come because who is not coming?’
d. Ni3 bu4 ren4wei2 shei2 hui4 lai3?
you not think who will come?
‘Who don't you think will come?’
Cheng (1984) conducted a very thorough survey of Mandarin verbs that take statements as arguments, including verbs that may take or require indirect questions. According to this criterion, the embedded clauses of what he calls
verbs of locution, e.g., bao4gao4 'report' and gao4su4 'tell', verbs of cognition, e.g., zhi1dao4 'know' and wang4ji4 'forget' and ask-type verbs should be considered subcategorized for. As for other types of verbs cited by cheng (1984:141-143), their embedded clauses can still be judged as subcategorized complements according to other criteria, such as obligatory co-occurrence, ontological necessity, distributional restrictions, one per sentence, selectional restrictions, and possible internal gaps.
3.1.10 Identification of a Subcategory
Verbs, or more generally predicative elements, in a grammar may be further classified into different subcategories, according to the different kinds of constituents they subcategorize for. It follows logically that a subcategorized constituent should be able to be used to identify a distinctive subcategory of verbs, while adjuncts have no such functions. In the discussion of semantic presupposition of ontological necessity, we mentioned that although all events presuppose a location we do not want to say that therefore all verbs subcategorize for a locative element. Part of the reason is precisely that such a position serves no function in distinguishing a subcategory of verbs from others. Similarly, as discussed earlier in 3.1.3, adjuncts are relatively free to co-occur with a wide range of verbs and thus lack the ability to identify a subcategory of verbs; subcategorized arguments on the other hand only co-occur with a restricted subclass of verbs.
In the following discussion of Mandarin verb subcategorization, we will use evidence from the above tests in justifying our account for Mandarin verb subcategorization. Therefore, it may be worthwhile reiterating that while a couple of the criteria may serve as definitive tests or sufficient conditions for a subcategorized constituent, other conditions may be necessary but by no means sufficient, and still there are others that merely provide a general tendency. Hence, a stronger case can be made when a combination of these conditions is satisfied.