CHAPTER 2 GRAMMATICAL FUNCTIONS IN AN LFG
3.3 S UBCATEGORIZATION OF G RAMMATICAL F UNCTIONS
3.3.5.3 Semantic Restrictions of [ba3 NP]
The universal characteristic of OBJ as a semantically unrestricted grammatical function and OBL-Θ as a semantically restricted one in the LFG theory provides another test for us to decide which grammatical function [ba3 NP] should be assigned to. Recall that semantically unrestricted functions, such as SUBJ and OBJ, may be linked to any thematic role in the thematic structure of a verb. Semantically restricted functions are more intrinsically related to their semantic content and thus may only be linked to thematic roles that are semantically compatible. In the following examples in 3, all post-verbal noun phrases are no doubt direct objects. Note that these OBJs are associated with several different thematic roles: theme, goal, location, and instrument.
3. a. Wo3 da3 le ta1.
I hit LE he
‘I hit him.’ `theme
b. Wo3 yong4 le yan2.
I use LE salt
‘I consumed the salt.’ `theme, OR ‘I utilized the salt.’ `instrument
VERBSUBCATEGORIZATIONINMANDARINCHINESE 131
c. Wo3 jing4 le ta1. `goal I toast LE he
‘I toasted him.’
d. Wo3 qu4 le Tai2bei3. `location I go LE Taipei
‘I went to Taipei.’
However, only 3a-b, whose OBJs are linked with the thematic role theme, have grammatically acceptable ba3 counterparts, as shown in 4 below. Notice also that while 3b is ambiguous in that yong4 could be interpreted as "utilize"
which requires an instrument role for its object or as "consume" which requires a theme role, its ba3 counterpart 4b can only have the reading where [ba3 yan2] is linked with the theme role. Sentences in 4 thus show that the grammatical function of [ba3 NP] prepositional phrases can only be linked with the theme role.
4. a. Wo3 ba3 ta1 da3 le.
I BA he hit LE
‘I hit him.’ `theme
b. Wo3 ba3 yan2 yong4 le.
I BA salt use LE
‘I consumed the salt.’ `theme
‘*I utilized the salt.’ `*instrument
c.*Wo3 ba3 ta1 jing4 le.
I BA he toast LE
‘I toasted him.’ `*goal
d.*Wo3 ba3 Tai2bei3 qu4 le.
I BA Taipei go LE
‘I went to Taipei.’ `*location
Actually, given the semantic content of "dispose" or "process" of ba3 as a preposition, it is rather natural for the [ba3 NP] prepositional phrase to be
associated with only the thematic role of theme or patient. Another type of evidence for the semantic restrictiveness of [ba3 NP] comes from certain transitive verbs, such as jing4 'toast' in 4c above, which do not normally have acceptable ba3 counterparts, but when they are modified by an "extended"
expression, they may indeed co-occur with [ba3 NP]. See the following example.
5. a. Wo3 ba3 ta1 jing4 de ta1 bu4 neng2 zai4 he1 le.
I BA he toast DE he not can again drink LE
‘I toasted him so much that he could not drink any more.’
In this sentence, the added expression indicating the result of the action greatly exaggerates the degree of my toasting him. As Li and Thompson (1981:469) put it, "It is as if one cannot help thinking" that he was affected in some way when he was toasted so much by me that he could not drink any more. Therefore, it seems that ba3 forces its NP to be interpreted as having the theme role in such constructions, although normally, as shown in 3c, jing4 requires a goal role in its predicate.
To sum up, all the evidence provided thus far clearly indicates that ba3 should be classified as a preposition and that [ba3 NP] should be assigned to a subtype of the semantically restricted oblique function, not to the semantically unrestricted direct object. Since the semantic content of ba3 and the semantic restriction on [ba3 NP] seem to require an interpretation of theme role of the NP, we assign OBLTHME to be its grammatical function. Incidentally, it is a well-known fact that [ba3 NP], being always preverbal, has a strong tendency for its noun to have a definite reading; this fact is accounted for by our functionally annotated phrase structure rule that creates VP, where OBLTHME
gets a default value of [ DEFINITE + ] unless it is otherwise marked.
All verbs in this subcategory that subcategorize <SUBJ , OBLTHME> also belong to the previous subcategory of 3.3.4.1 that subcategorize <SUBJ , OBJ>. However, as we have mentioned before, the postverbal object, when unmarked, has a strong tendency to be indefinite. For example,
6. a. Wo3 ba3 ji1 sha1 le.
I BA chicken kill LE
‘I killed the chicken.’
VERBSUBCATEGORIZATIONINMANDARINCHINESE 133
a' Wo3 sha1 le ji1.
I kill LE chicken
‘I killed a chicken.’
b. Wo3 ba3 men2 ti1huai4 le.
I BA door kick-bad LE
‘I kicked and broke the door.’
b' Wo3 ti1huai4 le men2.
I kick-bad LE door
‘I kicked and broke a door.’
FI-V-5:
[ CAT V
FS [ PRED <SUBJ , OBLTHME> ] (↑ OBLTHME PCASE) =c THME ]
sha1:
{ [ FS [ FORM 'sha1' ]
FI-V-4
]
[ FS [ FORM 'sha1' ] FI-V-5 ]
}
ba3:
[ CAT P
FS [ PFORM 'ba3' PCASE THME ]
]
Most resultative compound verbs, such as 6b above, belong to this subcategory and thus all have non-ba3 derivational counterparts. However, an interesting observation is that when a resultative compound takes the form with a potential infix, affirmative de or negative bu2, it cannot subcategorize a [ba3 NP] constituent. The following sentences illustrate this observation.
Again, this fact is accounted for in the individual lexical entries.
7. a. Ming2tian1 wo3 hui4 ba3 men2 da3po4.
tomorrow I will BA door break
‘Tomorrow I will break that door.’
a' Ming2tian1 wo3 hui4 da3depo4 na4 shan4 men2.
tomorrow I will break that CLS door
‘Tomorrow I will break that door.’
a"*Ming2tian1 wo3 hui4 ba3 men2 da3depo4.
tomorrow I will BA door break
‘Tomorrow I will break that door.’
b. Wo3 bu4 neng2 ba3 men2 da3po4.
I not can BA door break
‘I cannot break that door.’
b' Wo3 da34bu2po4 na4 shan4 men2.
I break-not-open that CLS door
‘I cannot break that door.’
b"*Wo3 ba3 men2 da3bu2po4.
I BA door break-not-open
‘I cannot break that door.’
da3po4:
{ [ FS [ FORM 'da3po4' ]
FI-V-4
]
VERBSUBCATEGORIZATIONINMANDARINCHINESE 135
[ FS [ FORM 'da3po4' ] FI-V-5 ]
}
da3depo4:
[ FS [ FORM 'da3depo4' ] FI-V-4
]
da3bu2po4:
[ FS [ FORM 'da3bu2po4' ] FI-V-4
]
3.3.6 <SUBJ , OBLTHME , OBJ>
Examples: bo1 'peel', shang4 'lock', shua1shang4 'paint', zhuang1man3 'fill', zhuang1hao3 'fix', chi1 'eat', da3puo4 'break', fa1she4 'shoot';; gei3 'give', gong1ji3 'provide', song4 'give', jie4 'lend', jiao1 'teach', zu1 'rent', jie4 'lend', ti2gong1 'provide', gao4su4 'tell', dang1 'consider', kan4cheng2 'mistake...as', jiao4zuo4 'call'
1. a. Wo3 ba3 ta1men pian4 le liang3 ge.
I BA they cheat LE two CLS
‘I cheated two of them.’
b. Ta1 ba3 men2 shang4 le liang3 dao4 suo3.
he BA door lock LE two CLS lock
‘He locked the door with two locks.’
c. Ta1 ba3 hu2li2 bo1 le pi2.
he BA fox peel LE skin
‘He peeled the skin off the fox.’
d. Ta1 ba3 hu2li2 bo1 le san1 zhi1.
he BA fox peel LE three CLS
‘He peeled three of the foxes.’
FI-V-6:
[ CAT V
FS [ PRED <SUBJ , OBLTHME , OBJ> ] (↑ OBLTHME PCASE) =c THME ]
As we have argued in the previous section, we reject the double object analysis for sentences in 1 above. We treat the [ba3 NP] PP as having the OBLTHME oblique function. Another possible preposition that may assign this OBLTHME function is jiang1, which is used mostly in a more formal context such as the written text. The OBJ here can either be a full NP with a lexical head, thus with the head feature FORM, such as 1b-d above, or a headless NP construction without the head feature FORM, such as 1a. Note also that semantically when the OBJ is linked with a patient role, there is an interesting relation between the entities in OBLTHME and patient OBJ: the entity in OBJ is a closely related, often inalienable, part of the entity in OBLTHME, either before or after the action denoted by the verb. Thus, it is always the patient OBJ that directly receives the action while the theme OBLTHME is indirectly affected by it. However, notice that the relationship has to be that of whole-part or possession but cannot be that of class-member. This semantic restriction applies without exception to all ba3 sentences with a postverbal patient OBJ, which directly receives the action or undergoes the process denoted by the verb. However, this semantic restriction does not apply to verbs that are derivationally related to ditransitive verbs, such as gei3, whose OBJ is not linked to a patient but more likely to a thematic role of beneficiary, maleficiary, or goal. Our distinction of theme and patient in ba3 sentences is also supported by the studies of Chinese resultative verb compounds and verb-copying by Chang (1989 and 1990a).
This semantic restriction may provide partial explanation for the interesting observation given by Her (1985-6) that a postverbal patient OBJ and a preverbal [ba3 NP] OBLTHME are in complementary distribution.
VERBSUBCATEGORIZATIONINMANDARINCHINESE 137
1. e. Mao1 chi1 le yu2.
Cat eat LE fish ‘The cat ate the fish.’
f. Mao1 ba3 yu2 chi1 le.
Cat BA fish eat LE ‘The cat ate the fish.’
g.*Mao1 ba3 yu2i chi1 le yu2i. Cat ba3 fish eat LE fish ‘The cat ate the fish.’
No doubt the sentence 1g has a well-formed c-structure and f-structure since chi1 does belong to this subcategory and thus subcategorizes <SUBJ , OBLTHME , OBJ>. As a matter of fact, it seems that all formal features are satisfied in this sentence. Therefore, we know for sure that this sentence is ill-formed not because there is any violation in c- or f-structure. As we have seen in sentences 1a-d, it is entirely acceptable for the verb with preverbal ba3 to have a postverbal patient object, provided that this object and the object of ba3 are of a part-whole or possession relationship. In other words, the presence of an overt direct object with the identical form with the noun in OBLTHME is not the cause of the ill-formedness. The culprit is rather the difficulty or impossibility for a part-whole relationship to exist between the direct object and the NP in the ba3 phrase. We could therefore assume that sentence 1g is syntactically well-formed, but it is semantically or pragmatically ill-formed due to the violation of a semantic constraint.
The principles of anaphoric binding may provide another part of the solution to the unacceptability of sentences like 1g. Since full-NPs (or R-expressions in GB terminology) must be free, in 1g, it violates this principle for the NP in OBJ to be bound by the NP in OBLTHME. Thus, the only possible grammatical reading would force the two NPs, yu2 'fish', not to be co-referential. And when such reading is possible within the semantic constraint of whole-part or possession relationship either before or after the action, sentences like 1g will be acceptable, as predicted by our account. Thus, all the following sentences are acceptable, but only if OBJ and OBLTHME are not co-referential and a sensible reading is still possible. Yet, the reading
where the OBL and OBLTHME are co-referential is always anomalous or nonsensical and also violates the binding principle of full NPs.
1. h. Ta1 ba3 tu3i sha3shang4 le tu3j/*i. He BA dirt spread-on LE dirt ‘He spread dirt on the dirt.’
i. Ta1 ba3 yi1fu2i gai4shang4 le yi1fu2j/*i. He BA clothes cover-on LE clothes ‘He covered the clothes with clothes.’
j. Ta1 ba3 pi2i buo1 le pi2j/*i. He BA skin peel LE skin ‘He peeled the skin off the skin.’
k. Ta1 ba3 suo3i jia1 le suo3j/*i. He BA lock add LE lock ‘He added a lock on the lock.’
However, Her (1989a) also assumes that for sentences like 1a where the object is a headless NP the head noun of the OBLTHME must be unifiable with the entire OBJ. Thus, the unacceptability of the following sentence is said to be caused by the fact that the tou2 as a classifier is in conflict or not in agreement with the grammatical classifier of xiao3ji1 'chick', which is zhi1.
1. l.*Hu2li2 ba3 xiao3ji1 chi1-le liang3 tou2.
fox BA chick eat two CLS ‘The fox ate two of the chicks.’
Such an analysis may be incorrect; instead, we would suggest that the whole-part relationship between the OBLTHME and the OBJ is a semantic and discoursal one only and that there is definitely not a control relation where unification takes place between the two functions. This is evident from the following example of a simple discourse.
VERBSUBCATEGORIZATIONINMANDARINCHINESE 139
1. m. Q: Ni3 ba3 Lao3wang3 chi1 le ji3 zhi1 ji1?
you BA Old Wang eat LE how-many CLS chicken
‘How many of Old Wang's chicken did you eat?’
n. A: Wo3 ba3 Lao3wang3 chi1 le liang3 zhi1.
I BA Old Wang eat LE two CLS
‘I ate two of Old Wang's (chickens).’
In the case of 1n where zhi1 cannot be the grammatical classifier of Lao3wang3 and thus the unification between OBLTHME, Lao3wang3, and the OBJ, liang3 zhi1, would definitely fail. The fact that 1n is an entirely acceptable sentence proves that there is no control relation between OBLTHME and OBJ under any circumstance. Thus, our account predicts that 1l is in fact syntactically well-formed and attributes its unacceptability to its semantic anomaly. Therefore, if we can conjure up some discourse context in which such a sentence makes some sense, it will be judged as well-formed. Indeed we can. Imagine someone telling some children a fairy-tale where some smart little fairy chicks have a farm with some cows and there is also this bad giant fox trying to steal and devour some of the cows. The fox finally succeeds.
One of our young listeners might ask the following question to which 1l is a syntactically grammatical as well as semantically acceptable response.
1. o. Q: Hu2li2 ba3 xiao3ji1 chi1-le ji3 tou2 niu2?
fox BA chick eat LE how-many CLS cow
‘How many of the chicks' cows did the fox eat?’
l. A: Hu2li2 ba3 xiao3ji1 chi1-le liang3 tou2.
fox BA chick eat LE two CLS
‘The fox ate two of the chicks' (cows).’
This observation confirms our previous statements that 1) the relationship between OBLTHME and a patient OBJ in a single clause is that of whole-part or possession, but cannot be that of class-member, 2) this semantic restriction applies without exception to all ba3 sentences with a postverbal patient OBJ, and 3) there is no control relation between them. Our analysis thus accounts
for the anomaly of the following sentence nicely, while an agreement account does not.
1. p.?Ta1 ba3 yi1 zhi1 ji1 chi1-le wu3 zhi1.
he BA one CLS chicken eat LE five CLS
‘?He ate five of the chicken.’
All the ditransitive verbs, such as gei3 'give', song4 'give', gong1ji3 'supply', that subcategorize <SUBJ , OBJ , OBJ2>, also have counterparts in this subcategory. However, for these verbs, their OBJ is linked to a beneficiary, maleficiary or goal role and therefore there does not exist a part-whole or possession relation between the OBLTHME and the OBJ.
1. q. Wo3 ba3 shu1 gei3 le ta1.
I BA book give LE he
‘I gave him the books.’
r. Ta1 ba3 qing2bao4 gong1ji3 jun1fang1.
he BA information supply military
‘He supplies information to the military.’
Incidentally, though we will not get into the analysis of it, we note that verbs in the following sentences do not belong to this subcategory; rather, they subcategorize <SUBJ , OBLTHME> only and the postverbal NPs yi1 quan2 and liang2 xia4 are not OBJ. They function entirely like the frequency phrases such as yi1 ci4 'once' and belong to the non-subcategorizable ADJ, quantifying the extent or the frequency of the action.
1. s. Ta1 ba3 ni3 da3-le yi1 quan2.
he BA you hit ASP one fist
‘He hit you with his fist once.
t. Ta1 ba3 ni3 da3-le yi1 xia4.
he BA you hit ASP one time
‘He hit you once.’
VERBSUBCATEGORIZATIONINMANDARINCHINESE 141
3.3.7 <SUBJ , OBLGOAL , OBJ>
Examples: jie3shi4 'explain', bao4gao4 'report', shuo1ming2 'illustrate', du3, da3du3 'bet', biao3shi4, biao3da2 'express', shuo1 'say', cheng2ching1, biao3ming2 'clarify', mai3 'buy', mai4 'sell', tui1xiao1 'promote', kua1zhang1 'exaggerate', zu1 'rent', jie4 'borrow', xun2wen4 'ask'
1. a. Ta1 gen1/tong2/dui4/xiang4/he2 ni3 jie3shi4 Yi4jing1.
He with you explain Yi Ching ‘He explains Yi Ching to you.’
b. Ta1 gen1/tong2/*dui4/xiang4/*he2 ni3 zu1 fang2zi He with you rent house ‘He rents a house from you.’
c. Ta1 gen1/tong2/dui4/xiang4/*he2 ni3 tui1xiao1 fang2zi He with you promote house ‘He promotes the house to you.’
FI-V-7:
[ CAT V
FS [ PRED <SUBJ , OBLGOAL , OBJ> ] ( ↑ OBLGOAL PCASE ) =c GOAL ]
jie3shi4:
[ FS [ FORM 'jie3shi4' ]
FI-V-7
]
zu1:
[ FS [ FORM 'zu1' ]
( ↑ OBLGOAL PFORM ) =c { 'gen1' 'xiang4' 'tong2' }
FI-V-7
]
tui1xiao1:
[ FS [ FORM 'tui1xiao1' ]
~( ↑ OBLGOAL PFORM ) =c 'he2' `PFORM cannot be 'he' FI-V-7 `~ negates the schemata ]
Again, in the default case all prepositions that have [PCASE GOAL] may assign the OBLGOAL function; however, individual verbs within this subcategory that subcategorizes <SUBJ, OBLGOAL , OBJ> may require different prepositions rather arbitrarily as shown in the above sentences 1a-c, and such idiosyncratic information must be specified in the individual verb entry, as exemplified in the above entries.
3.3.8 <SUBJ , OBJ , OBLLOCT>
Examples: xie3 'write', gua4 'hang', diu1 'throw', ban1 'move', ji4 'mail', pai4 'sent', tui1 'push', tiao1 'carry', dai4 'carry', mai4 'sell', kai1 'drive', fang4 'put', ti1 'kick'
1. a. Ta1 xie3 le yi1 ge zi4 zai4/*dao4 zhi3-shang4.
he write LE one CLS character at to paper top ‘He wrote a character on the paper.’
b. Ta1 diu1 le yi1 jian4 yi1fu2 dao4/zai4 chuang2-xia4.
he throw LE one CLS dress to at bed under ‘He threw a dress under the bed.’
c. Ta1 ban1 le yi1 kuai4 shi2tou2 dao4/*zai4 fang2-li he move LE one piece rock to at house inside ‘He moved a rock into the house.’
As we can see from the above examples, although most verbs in this subcategory may appear with either zai4 or dao4, both of which have PCASE LOCT, some verbs may occur only with one of them and the choice seems to be related the meaning of the verb.
VERBSUBCATEGORIZATIONINMANDARINCHINESE 143
FI-V-8:
[ CAT V
FS [ PRED <SUBJ , OBJ , OBLLOCT> ] (↑ OBLLOCT PCASE) =c LOCT ]
diu1:
[ FS [ FORM 'diu1' ] FI-V-8 ]
xie3:
[ FS [ FORM 'xie3' ]
(↑ OBLLOCT PFORM) =c 'zai4' FI-V-8
]
ban1:
[ FS [ FORM 'ban1' ]
(↑ OBLLOCT PFORM) =c 'dao4' FI-V-8
]
3.3.9 <SUBJ , OBLTHME , OBLLOCT>
Examples: xie3 'write', gua4 'hang', diu1 'throw', ban1 'move', ji4 'mail', pai4 'sent', tui1 'push', tiao1 'carry', dai4 'carry', mai4 'sell', kai1 'drive', fang4 'put', ti1 'kick'
1. a. Ta1 ba3 zi4 xie3 zai4 zhi3- shang4.
he BA character write at paper top ‘He wrote the character on the paper.’
b. Ta1 ba3 yi1fu2 diu1 dao4 chuang2-xia4.
he BA dress throw to bed under ‘He threw a dress under the bed.’
FI-V-9:
[ CAT V
FS [ PRED <SUBJ , OBLTHME , OBLLOCT> ] (↑ OBLTHME PCASE) =c THME
(↑ OBLLOCT PCASE) =c LOCT ]
Verbs in this subcategory seem to coincide with the previous subcategory of 3.3.8; thus, we assume that there is a derivational relation between verbs of the two subcategories.
3.3.10 <SUBJ , OBLTHME , OBJ , OBLLOCT>
3.3.10 <SUBJ , OBLTHME , OBJ , OBLLOCT>