• 沒有找到結果。

An Optimality Theoretic Analysis of the Distribution of Hakka Prepositions DI, DO, BUN, LAU, TUNG, ZIONG

N/A
N/A
Protected

Academic year: 2021

Share "An Optimality Theoretic Analysis of the Distribution of Hakka Prepositions DI, DO, BUN, LAU, TUNG, ZIONG"

Copied!
39
0
0

加載中.... (立即查看全文)

全文

(1)

Concentric: Studies inLinguistics 38.2 (November 2012).171-209

An Optimality Theoretic Analysis of the Distribution of

Hakka Prepositions

DI

,

DO

,

BUN

,

LAU

,

TUNG

,

ZIONG*

wd

ob--nNM M訕地 T-w 時 U MW 唱 C 切 hm T扣

This paper has three purposes. The first purpose is to examine the uses of the prepositions:di, 品, b帥, tau,tung,ziong,and generalize an overall structural pattern for the Hakka YP involving prepositional phrases. The second purpose is to distinguish between the PPs functioning as complements and adjuncts of the verb,and argues that the two types of PPs may occur in four different syntactic positions depending on their syntactic and semantic functions. The third purpose is to formulate an Optimality Theoretic analysis to account for the different positioning of PPs. The analysis is achieved through two rounds of aT evaluation. The first round is based on the semantic aT,by which each PP receives its complement or adjunct interpretation when appearing in the YP. The second round is syntactically oriented. By proposing various manifestations of Generalized Alignment Constraints for complement PPs associated with different semantic functions and adjunct PPs exhibiting a relatively more peripheral relation with the verb, this paper will show that the interaction of these alignment constraints provides an alternative account for the distribution of the four different kinds of PPs.

Key words: Hakka,Optimality Theory,prepositional phrase

1.Introduction

Hakka1prepositional phrases follow the general word-order pattern of Chinese languages,in which the head preposition precedes itsNP complement and occurs at the left edge of aPP. Afew examples are given below:2

n ob 址, s nUH --山 AU 刊 u uad -E 。u u3 刀 LHhuh-、 -Y l /S' 、

* I am grateful to the anonymous reviewers for their insightful comments and suggestions on this work. This paper is supported by the National Science Council under the grant #NSC-100-241O -H-032-063. The author is solely responsible for any errors and misrepresentations.

I The two major dialects of Hakka in Taiwan“Si-ien" and“Hoi-liuk" are contrasted mainly in their tonal systems. Fundamentally the Si-ien Hakka has six tones and the Hoi-liuk Hakka has seven tones (Xu 2001); high tones in Si-ien are pronounced as low tones and low tones pronounced as high tones in Hoi-liuk,and vice versa. The two dialects also differ in some oftheir lexicon and collocational expressions; however,due to cultural and geographic proximity,two of them share more and more mixed features in all respects. In this paper,all the data presented is based on Hoi -liuk Hakka spoken in Miaoli,Taiwan.Itshould be noted that the Hakka dialect spoken in Miaoli is now often referred to as the "Hoi-si" dialect. Itis a mixture of Hoi-Iiuk and Si-ien dialects. The structure of Hoi-si is primarily based on the Hoi-Iiuk dialect,and at the same time it borrows some vocabulary from Si-ien,which made this dialect different from other Hoi-liuk dialects spoken in Taiwan.

2 This paper uses the following glosses: ASP “'aspectual marker," CL “classifier," CaMP

“complementizer," MOD "modificational," PART “particle," PASS “passive," PAT "patient,"

(2)

IC I!!I!且E且11II 38.2(November 2012)

(2) di mangin ge vukha

at who poss home

'at whose home' (3)tung ge zhak penyiu

with that CL friend

‘with that friend'

The internal structure of Hakka preposition phrases is illustrated by the tree diagram presented in (4),in which the preposition heads a functional projection and selects an NP as its complement branching on the right.

~\\\\\

P NP

To further investigate the syntactic and semantic properties of PPs in Hakka,this paper is organized as follows. Section 2 examines in detail the function and distribution of PPs projected by various head prepositions. In this section the PPs associated with different semantic roles will be divided into two groups: complements and a句 uncts, and the discussion will emphasize the position they may occur in sentences. Section 3 provides an Optimality Theoretic approach to account for the data,and then followed by a conclusion in section 4.

2. Hakka prepositions

The Hakka prepositions can be classified as a grammatical category with the capability to bring out at least the temporal, locative, directional, associative,

benefactive information and to mark the passive and emphatic functions to the event indicated by the VP. In this section,we first examine the meaning and function of PPs affiliated with various prepositions that very commonly occur in Hakka,and then their distribution and syntactic status (complement or adjunct interpretation) will be discussed.

(3)

Tsenl!:OT on HakkaPrer可 osition且 IPhrases

2.1 Preposition:di

The morpheme di is probably the most uncontroversially identifiable preposition in Hakka. It occurs quite frequently in oral and written expressions, functioning to point out the temporal or locative information to an event.The following examples(5) and(6) illustrate these functions.

(5) Gi di libaingit voi hi hokgau tuk yin刊n.

he on Sunday will go school study English 'He goes to school studying English on Sunday.' (6) Gi di hokgau tuk yinvun.

he at school study English 'He studies English at school.'

Now contrast with the following examples, in which the prepositional phrases headed bydioccur in other positions in the same sentence.

(7) a.(*Di) libaingit gi voi hi hokgau tuk yinvun. on Sunday he will go school study English

‘He goes to school studying English on Sunday.'

b. *Gi voi hi hokgau tuk yinvun di libaingit.

he will go school study English on Sunday 'He goes to school studying English on Sunday.' (8)a.*Di hokgau gi tuk yinvun3.

at school he study English ‘He studies English at school. ' b. *Gi 個k yinvun di hokgau.

he study English at school ‘He studies English at school.'

] One may argue what causes the ungrammaticality of(8a) might not be the positioning ofthe locative

PP,since the sentence turns grammatical as long as anotherdiis inserted between the subject and the verb:

(i) Di hokgau, gi di tuk yinvun. at school he DI study Englsih ‘At school,he is studying English now.'

In effect,(8a) and (i) present two different kinds of constructions. In (i),the seconddibehaves as a linking verb functioning to indicate the occurrence of an action on the subject, and the entire expression can be identified having a“topic-comment" reading,in which an independent clause is taken as the comment made about the sentence-initial topic,and a pause normally exists as an in-between to separate the topic and the comment. Besides,the expression in (i) does not receive the habitual reading as (6) possibly does,but it must be used to describe the current status about the subject.

(4)

W I!且且且盟且IIIJ區 38 .2(November 2012)

When a di-phrase functions to denote place infonnation to the verb,as in (8),it is completely unacceptable to place the di-phrase in any position other than the left edge ofVP. Similarly,when a di-phrase functions to denote temporal infonnation,locating it in the post-verbal position renders absolute ungrammaticality, as shown in the second example of (7). The positioning of the temporal di-phrase in (7a) may sound better,but we can argue that the time indication is preferably to be expressed by an NP if it occurs in the sentence-initial position. Compare (7a) with the following example (9),when the di-phrase occurs at the left edge ofVP,the time indication may alternatively be expressed by a PP or an NP; namely,the temporal preposition di may overtly or covertly emerge in the VP.

(9) Gi (di) libaingit voi hi hokgau tuk yinvun.

he on Sunday will go school study English

‘He goes to school studying English on Sunday.'

The above examples show that di, combining with a noun phrase to fonn an adverbial expression of time and place, preferably occurs at the left peripheral position of a VP. In other words,the canonical position for the PP headed by di is to the immediate left ofthe VP it modifies.

2.2 Preposition: do

The prepositional do occurs with a verb,and the projected PP is in some way associated with locative interpretation. As explained by Lai (1988, 2002), the prepositional use of do functions to introduce the source, the goal, or the place infonnation to further extend the meaning of the verb. Examining the following examples:

(IO) a. Gi zang do miguet zonloi.

he just from America come.back

He just came back from the United States.' b. Gi do hokgau hanglu.

he at school walk

He is walking at school. '

c. Gi oi hanglu do hokgau.

he want.to walk to school

(5)

TsenQ":OT on Hakka Prenositional Phrases

All the above do-phrases co-occur with the main verb,and in the first example of (10),the do-phrase indicates the source location from which the event represented by the verb begins; in the second example,the do-phrase points out the generallocation in which the event of the verb takes place; while in the third example the do-phrase states the goal toward which the action of the verb is directed.

It should be noted that when do appears in the pre-verbal position, it is interchangeable with the prepositiondi; on the other hand,do cannot be replaced by di

if the do-phrase appears post-verbally. This is demonstrated in the following(11).

(11) a. Gi zang di miguet zonloi.

he just from America come.back

He just came back from the United States. ' b. Gi di hokgau hanglu.

he on school walk

He is walking at school. '

c. *Gi oi hanglu di hokgau.

he want.to walk to school

‘He wants to walk to school.'

The following generalization can be made concerning the syntactic position where do

may occur in sentences. When the phrasal projection headed bydo functions to bring

out the source or the generallocative information to the verb,it precedes the verb and canonically occurs at the left edge of VP. In that case the meaning and positioning is similar to the uses of the di-phrase. Contrastive旬, if a do heading 臼nctional

projection acquires the semantic meaning to indicate the direction or the goal ofverbs,

it obligatorily follows the verb.

2.3 Preposition:lau/tung/ziong

The three prepositions tau, tung, and ziong exhibit very similar syntactic and

semantic behavior in some respects

,

and at the same time

,

their differences can be found if we further compare and contrast their constructions when appearing in sentences. Among the three morphemes,tau is the one that displays the most versatile

functions. As proposed by Lai (200站, b), tau contains at least the following five

different senses: as a comitative marker,a source marker,a goal marker,a benefactive marker,and a patient marker. It may be argued that the comitative tau is actually

related to two distinctive grammatical functions: a comitative preposition and a coordinative conjunction. However, since the distinction between them is neither

(6)

|行回 iii ,aa::..38.2(Novemb叮 2012)

straightforward nor uncontroversial,this paper will neglect the comitative use oftau

to avoid causing some unnecessary complexity in our analysis.

(12) a. Oi Jau gei gien diam mai yit tiau nge.

he from that CL store buy one CL 自由

‘He bought a fish from that store. '

b. Oia moie diamdiame Jau gi gong ge kien siicin.

his daughter quietly to he say that CL event

His daughter quietly told him (said to him) that event.'

c. Oia me Jau gi mai dong do suo

his mother for he buy really many book

His mother bought many books for him.'

d. Oi Jau gia moie ga-tet Ie.

he PAT his daughter rnaπy-RYC PART

'He has married his daughter to someone.'

In (12a), the tau-phrase indicates the source from which something has been taken away. In contrast, the prepositional phrase in (12b) denotes the opposite direction ofthe source. The head prepositiontau in this sentence denotes a goal,and it is to this destination that some information has been transferred. In (12c) the

tau-phrase marks the notion of benefaction,in which the object oftau refers to the target

who benefits from the event. Finally, tau in (12d) assigns the patient sense to its objective NP; to be exact, it emphatically marks the NP that has been in some way affected by the event that the predicate describes.

Tung is a near synonym of tau. The two morphemes seem to be very similar in

their syntactic and semantic behavior. Like the prepositional uses oftau,tung can also

be used as a source marker,a goal marker,a benefactive marker,and a patient marker; therefore,all the above four tau-phrases in (12) can be substituted by a tung-phrase,

and with the substitution oftung for tau,the meaning of each sentence remains the same.

(13) a. Oi tung gei gien diam mai yit tiau nge.

he from that CL store buy one CL fish

‘He bought a fish from that store. '

b. Oia moie diamdiame tung gi gong ge kien siicin.

his daughter quietly to he say that CL event

(7)

Tsem!: OT on Hakka PrenositionalPhras尼日

c. Gia me tung gi mai dong do suo

his mother for he buy really many book

‘His mother bought many books for him. '

d. Gi tung gia moie ga-tet Ie.

he PAT his daughter rnaπy-RVC PART

He has married his daughter to someone.'

However,some differences actually exist between the semantic senses related to

lau and tung. In this paper Twill show at least the following two salient differences.

First,as demonstrated in (14),when the preverbal prepositional phrase is associated with the patient role, the NP that follows tung may not be inanimate, but such restriction does not apply to the lau construction.

(14) a. Gi lau/?tung ge bun su mai-zeu Ie

he PAT that CL book buy-RVC PART

He has bought that book.'

b. Gi lau/?tung fan siit-ciang-ciang.

he PAT rice eat-clear-clear

He ate up all the rice. '

Second, when a patient argument is marked in the preverbal position by a prepositional phrase, lau is preferable to tung in heading this preverbal functional

projection if the PP expresses the meaning that is somewhat malefactive. Tn other words, comparing with lau,tung is semantically incompatible with the malefactive

sense,as shown by the two sentences in (15).

(I日 a. Gi lau/?tung ge zhak seilai da-si-tet Ie

he PAT that CL boy beat-dead-RvC PART

‘He beat that boy and which made the boy dead. '

b. Gi lau/?tung sinsang hi do mien-fung-fung.

he PAT teacher irritate COMP face-red-red

He irritated the teacher to the degree that the teacher's face has turned red.' In section 3.1, an explanatory account will be provided with the Optimality Theoretic framework to further illustrate the semantic restrictions related totung.

Another related preposition ziong differs from lau and tung in that it is not

allowed to mark the source,the goal, nor the benefactive senses,as seen in (16). On

(8)

|行...且且星星益團 38.2(November 2012)

the other hand,when used as a patient marker,ziongis probably the default marker, which is not subject to any kind of semantic restriction; see the examples in (17).

(16) a. *Oi ziong gei gien diam mai yit tiau nge.

he from that CL store buy one CL fish

‘He bought a fish from that store.'

b. *Oia moie diamdiame ziong gi gong ge kien siicin.

his daughter quietly to he say that CL event

‘His daughter quietly told him (said to him) that event.'

c. *Oia me ziong gi mai dong do suo

his mother for he buy really many book

His mother bought many books for him.'

(17) a. Oi ziong lepsep fit-tet Ie.

he PAT garbage discard-RYC PART

He dumped the garbage. '

b. Oi ziong sinsang hi do mien-fung-fung.

he PAT teacher irritate COMP face-red-red

‘He irritated the teacher to the degree that the teacher's face has turned red.'

c. Oi oi ziong moie ga bun ge zhak seilai.

he will PAT daughter rna叮Y to that CL man

‘He willrna叮yhis daughter to that man. '

The closest synonym of ziong in Mandarin Chinese is them。中hemeba,which is normally associated with an NP denoting the patient role in the preverbal position. One of the major claims about the grammatical status of ba is to take ba together with its following NP as a PP (Mei 1972,Li 1990). However,the PP account of the ba-construction has been questioned by other linguists.A significant argument is that ba-phrases do not behave exactly like other PPs. Ritter & Rosen (2000) observe that while PPs are generally allowed to appear both preverbally and postverbally,

ba-phrases obligatorily occur in preverbal positions. In addition, as argued in Rhys (1996),when a preverbal PP functions as an adverbial modifier,it may appear in the sentence initial position, but a preverbal ba-phrase never occurs sentence initially; instead, it has to be closer to the main verb and is encoded with the function to indicate a patient argument.The Hakkacounte中artpreposition ziong,when serving to bring in a patient argument to the following verb, may also be identified as a PP parallel to the Mandarinba司construction.In response to the queries about the validity of this PP analysis,this paper will make the following claim: ziong-phrases are known to be associated with particular syntactic position and semantic function,and it is this

(9)

TsenQ: OT onH且kkaPrenositional Phrases

feature which distinguishes them as a distinctive type of preverbal PP. They differ from other preverbal PPs in that syntactically they are nearer the verb comparing to those whose position is closer to the leftperipheηIof sentences,and semantically they introduce a patient or a theme argument to the verb, as opposed to the adverbial function commonly associated with other preverbal PPs.

Regardless of the various meanings and functions that the three prepositions lau, tung,and ziong may associate with,from the above examples(12) throughout(17) we can argue that the only position in sentences where the three prepositions may possibly occur is in the preverbal position; more precisely,the leftmost position of a VP.

2.4Preposition: bun

The Hakka morpheme bun can be used to express a variety of grammatical functions. The most common ones include the main predicate of sentences, a goal marker indicating the recipient for the transference ofpatient!theme from the agent,an agent marker denoting the performer of an action,and a complementizer connecting two clauses and showing the relation between them; see Lai (200 I) for the detailed illustration of the different uses of bun. This paper will focus on the prepositional uses of bun, which is believed to be closely related to two syntactic constructions: ditransitive and passive constructions. In the former construction,bun is seen as the

counte中art of the Mandarin Chinese gei; while in the latter construction,bun is by

and large equivalent to the Mandarin Chinese bei.

(18) a. Ge zhak seingine sung yit liap lingo bun gi.

that CL kid give.as.gift one CL apple to he

'That kid gave an apple to him as a present.'

b. Ge zhak seingine bun gi rna do gieuzii Ie

that CL kid PASS he scold COMP cry PART

That kid was scolded tocηbyhim.'

Bun in(l8a) functions to indicate a recipient for the predication of a giving-verb. In (18b), bun is an agent marker indicating a demoted argument in the passive

construction,which serves as an action performer for the verb predication.

As mentioned earlier, the ditransitive bun corresponds to gei in Mandarin Chinese. Many Chinese linguists have proposed important arguments concerning the syntactic analysis of gei in ditransitive constructions. The arguments state how researchers disagree with each other on determining the syntactic category of gei in

(10)

1 6'1!D'I:且且且呈且138.2(November 2012)

different verbal constructions. While there are linguists proposing a more traditional account in which ditransitive gei-phrases should be treated as PPs functioning to indicate an indirect object to the verb (Tang 1979, Li & Thompson 1981), other linguists argue that gei should be verbal in nature. When gei directly follows the verb,

it is a suffix attachi月 to the verb; on the other hand,when a direct object is inserted between the verb and the gei-phrase,gei is a lexical verb and the sentence should be

identified as a serial verb construction (Huang & Ahrens 1999). In Ting& Chang's (2004) proposal,they counter the serial verb account of gei. They argue instead that the postverbal gei is polysemous and must be analyzed as being associated with different grammatical categories. When gei directly follows the verb,it is the second component of a verbal compound. When gei follows both the verb and the direct object,it may be a preposition or a complementizer depending on the type of phrasal constituent it selects as the complement.

In this paper,I will argue that bun in ditransitive constructions should never be identified as a verb; instead,it is a preposition followed by an NP functioning as an

l吋irectobject of the verb. The prepositional account of bun can be supported by the following three linguistic tests. First, unlike the Mandarin Chinese gei, when bun immediately follows the verb, it is not allowed to be attached an aspectual marker. Examine the examples in (19).

(19) a. Gi zia bun *go

he lend to ASP

ngaI ng Clen yen.

I five thousand dollar

‘He used to lend me five thousand dollars.'

b. Gi na bun *den gia laie ge bun suo

he take to ASP his son that CL book

'He is handing that book to his son.'

c. Gi sung bun *Ie ge zhak hoksange liong gi bit.

he give.as.gift to ASP that CL student two CL pen

He has given two pens to that student as the present.'

Second, while most verbs can be modified by adverbs, bun contradicts this

verbal property by the fact that when it follows both the main verb and the direct object,it cannot be modified by an adverb; instead,the adverbial modifier can only modify the main verb. The ungrammaticality is shown in (20) below.

(20) a. *Gi na ng cien yen taifong bun ngai.

he take five thousand dollar generously give I

(11)

TsenQ: OT on HakkaPrenosition且 I Phrases

b. Gi taifong na ng cien yen bun ngai.

he generously take five thousand dol1ar to

‘He generously took five thousand do

l1ars to me.'

Third,like the case of Mandaringei,when the main verb selects a second VP as the complement, the object NP inside the second VP theoretical1y should be extractable,see (21 a) through (21 c)

(21) a. Gi cut ng cien yen (mai ge Hap Hnggo].

he spend five thousand dol1ar buy that CL apple

‘He spent five thousand do

l1ars to buy that apple.'

b. Gi cut ng cien yen mai t ge (ge Hap Iinggo]

he spend five thousand aol1ar buy MOD that CL apple 'The apple which he spent five thousand dol1ars to buy...'

c. (Ge Hap Hnggo], gi cut ng cien yen mai-do ge t

that CL apple he spend five thousand dol1ar buy-RYC MOD ‘That apple,the one he purchased with five thousand dol1ars...'

However,compare the sentences in (21) with those in (22) below,we therefore can argue that since the Hakka bun cannot be stranded,as shown in (22a) through (22c),it is problematic to analyzebunas a verb in Hakka.

(22) a. Gi sung cien (bun ge zhak seimoi].

he give.for.free money to that CL girl ‘He gave money to that girl as the present.'

b. * Gi sung cien bun t ge (ge zhak seimoi]4

he give.for.free money to MOD that CL girl

‘The girl who he gave money as the presen

t...'

4 One may argue that the unacceptable construction may be rescued as long as we propose a resumptive pronoun in the trace position,as in (i) below:

(i) Gi sung cien bun gi ge [ge zhak seimoi)

he give.fo r.仕ee money to she MOD that CL girl ‘The girl whom he gave money to as the present.

However,if we again compare (i) with the following unacceptable (ii),where the same syntactic process has been applied to(21b)

(ii)*Gi cut ng cien yen mai gi ge (ge Iiap Iinggo)

he spend five thousand dollar buy it MOD that CL apple ‘The apple which he spent five thousand dollar to buy it...'

The ungrammaticality of (ii) helps argue that the examples presented respectively in (i) and (ii) should be identified as two different kinds of constructions,and thebunin (i) behaves differently from the verbalmaiin (ii); thus it is dubious toidentif扯 bun in (i) similarly as a verb analogous to the status ofmaiin (ii)

(12)

|βmu"un ,, 38.2(November 2012)

c. *[Ge zhak seimoi)

,

gi sung cien bun t ge

that CL girl he give.for.free money to MOD

‘That girl,the one he gave money as the present. . . '

Now we should move on to the discussion on the passive bun. As stated above, the equivalence of the Hakka passive bun in Mandarin is the morpheme bei. Thus,we will first examine the status of bei in the research of Chinese linguistics. There are at least two alternative analyses in the interpretation of the Mandarin passive bei. The first analysis identifies bei as a preposition denoting a demotedsubject附 argumentof the verb (Wang 1970,Koopman 1984,Travis 1984, Li 1990). The second analysis treats bei as a verb followed by a clause,and the clause may be related to bei through complementation (Hashimoto 1969,Wei 1994) or control (Cheng,Huang,Li & Tang 1993, Ting 1995). One major difference between the two analyses is the syntactic function of the NP following bei,whether as a nominal complement bound with the prepositional bei,or the subject of the complement clause taken by the verbal bei. Again in this paper I will argue that the Hakka passive bun should be analyzed as a preposition, which we can find evidence from the following arguments. First, the Hakka bun-construction differs from the Mandarin bei-construction in that bei can be followed by either an NP or a VP,see (23).

(23) a. Ta de xiezi bei wo tou-zou Ie.

he poss shoes PASS I steal-away PART

‘His shoes were stolen by me.'

b. Ta de xiezi bei tou-zou Ie.

he poss shoes PASS steal-away PART

His shoes were stolen.'

We may argue that the two b衍's behave differently. As shown in below (24a), while a normal sentence is allowed to be adjoined by an adverbial modifier, if we argue that bei is a verb taking a clausal complement, how can we justify the ungrammaticality of (24b) in which the clause is left-adjoined by an adverb? Comparing with (24c),the left-adjunction is grammatical if the clause is subjectless,

or the adverbi祉-adjunctionoccurs at the VP level. According to (24),the passive bei is closely bound with its following NP (if it is present),and no constituent is allowed to be inserted in-between. Their bound relation argues for a prepositional interpretation on the grammatical bei when it is followed by an overt NP.

(13)

TsenQ: OT on Hakka PrenositionalPhra且尼且

(24) a. Qiaoqiaodi wo tou-zou Ie ta de xiezi.

quietly I steal-away ASP he poss shoes

Quietly,I stole his shoes.'

b. *Ta de xiezi bei qiaoqiaodi wo tou-zou Ie.

he poss shoes PASS quietly I steal-away PART

‘His shoes were quietly stolen by me.'

c. Ta de xiezi bei (wo) qiaoqiaodi tou-zou Ie

he poss shoes PASS I quietly steal-away PART

His shoes were quietly stolen.'

The Hakka passive construction exhibits even stronger bond between the passive morpheme and its subsequent NP. This is illustrated by(2日 below,where bun always

requires the presence of an NP directly following it.

(2日 a. Gi ge hai bun ngai teu-zeu Ie.

he poss shoes PASS I steal-away PART

His shoes were stolen by me.'

b. *Gi ge hai bun teu-zeu Ie

he POSS shoes PASS steal-away PART

‘His shoes were stolen. '

Example (25b) is ungrammatical becausebun is directly followed by a YP. The

syntactic bond between bun and the NP can be further proved by providing the

following examples in (26). In (26),when a temporal expression adjoins to the verbal predicate,it may be adjoined to the left of YP or the lower Y-bar,but it never inserts between the functional bun and the NP following it

(26) a. Gi ge hai di cobungit bun ngai teu-zeu Ie.

he poss shoes at yesterday PASS I steal-away PART

‘His shoes were stolen by me yesterday. '

b. Gi ge hai bun ngai di cobungit teu-zeu Ie.

he poss shoes PASS I at yesterday steal-away PART

His shoes were stolen by me yesterday.'

c. *Gi ge hai bun di cobungit ngai teu-zeu Ie.

he poss shoes PASS at yesterday I steal-away PART

'His shoes were stolen by me yesterday. '

(14)

l 行lI.U.l.I.!L且且III:量 38.2(November 2012)

The second evidence can be found by providing the following conversation in (27) as the test. The underlying argument is that when a speaker repeats part of a conversation to seek for further confirmation about a certain piece of information,it happens naturally and unconsciously that the speaker tends to repeat constituents instead of non-constituents because only a constituent can be singled out as an independent linguistic unit.

(27) A: Ge zhak seimoi bun gia tunghok rna.

that CL girl PASS her classmate scold

‘That girl was scolded by her classmate. ' Bl: Ge zhak seimoi?

B2: Bun gia tunghok rna? B3: Bun gia tunghok? B4: *Gia tunghok rna?

In the statement about the girl's being scolded by her classmate,it is possible to repeat the entire subject NP to check for confirmation as in B I,the entire predicate VP as in B2,the bun-NP combination as in B3,but it is inappropriate if we take the part after bun as a constituent to be repeated as in B4. This argues for the PP analysis of the bun-NP combination,as they form a linguistic constituent that may stand alone as an independent expression to perform a certain pragmatic function,in this case to seek for confirmation about the agent who performs the action.

From the above discussion,we see that both the passive and the ditransitive bun should be analyzed as a PP construction in Hakka. The prepositional phrases headed by bun may occur in both preverbal and postverbal positions, determined by the grammatical function assigned to them. When bun functions as an agent marker in passive constructions,it precedes the verb; when bun functions as a recipient marker in ditransitive constructions,it follows the verb.

2.5Complement and adjunct PP

Before probing into the discussion of the syntactic status of PPs concerning their functions as the complement or the adjunct of verbs, we will first recapitulate the position in relation to the verb where each of the prepositions mentioned above might occur in sentences. As presented in the following table, a PP may occur either preverbally or postverbally,and the position depends on the semantic meaning and the grammatical function they are associated with.

(15)

TseTIl!:OT on Hakka PrenositionalPhr且ses

Table1.Preverbal!Postverbal position of PPs

P~rE~P~OS~It~io~n~~~~~P~o~sit~io~n Preverbal Postverbal

di Temporal and Locative

do Source and Locative Directional/Goal

lau Source,Goal,

Benefactive,and Patient

tung Source,Goal,

Benefactive,and Patient

ZIOn2 Patient

bun Agent Recipient

The position of Hakka PPs follows the word order generalization of Chinese languages,according to which adjunct PPs are bound with preverbal positions,while complement PPs preferably occur in postverbal positions (Mulder& Sybesma 1992, Feng 2003). As I will discuss in the following, Hakka prepositional phrases can basically be divided into two major categories: preverbal a句unct PPs and postverbal complement PPs. The preverbal adjunct PPs can 如此herbe classified into two types based on their functions and syntactic positions in tree diagrams. In addition,we may observe the fourth categ。可 inwhich a preverbal preposition functions to denote the following NP as an agent or a patient argument to the verb. The PPs of this category are complements ofverbs appearing in the preverbal position.

The first type of PPs precedes the verb,acting as a preverbal adjunct modifying the VP. Phrases headed by di and do functioning to indicate the temporal,source and locative information belong to this type. The second type of PPs are also preverbal adjuncts modifying the VP they adjoin to. Phrases headed by lau and tung with the grammatical functions to introduce a source, goal and benefactive argument to the verb can be classified into this type. When the two types of PPs co-occur to modify the same verb,the precedence order shows that the PP of the first type must precede that of the second type,as shown in (28) and (29) below:

(28) a. Gi [di Iibaingit] [tung/lau gieue] sesiin.

he on Sunday for dog bathe

He bathed the dog on Sunday.'

b. *Gi [tung/lau gieue] [di Iibaingit) sesiin.

he for dog on Sunday bathe

He bathed the dog on Sunday.'

(29) a. Gi [do toibet] [tung/lau ge gien diam) mai suigo.

he in Taipei from that CL store buy fruit

‘He bought fruit from that store in Taipei.'

(16)

I rJrJmI!II且11!11 38.2(November 2012)

b. *Gi [tung/lau ge gien diam] [do to泌的] mai suigo.

he from that CL store in Taipei buy fruit

‘He bought fruit from that store in Taipei.'

The third type of PPs constantly occurs in the postverbal position. The PPs that belong to this type include the ones headed by do and bun functioning to indicate

respectively the goal and the recipient to which the action ofVP is directed. As shown in the following (30), this type of PPs is the verbal complement which obligatorily appears after the verb. The absence of this type of PPs causes ungrammaticality.

(30) a. Gi oi hanglu 叮do hokgau).

he want.to walk to school

‘He wants to walk (to school).'

b. Ge zhak seingine sung yit liap

that CL kid give.as.gift one CL

‘That kid gave an apple (to him) as a present.'

lingo *(bun gi).

apple to he

We may find exceptional cases showing the non-obligatory existence of a complement PP. As in (31) below, a verb like mai may optionally be used as a

transitive or a ditransitive verb,depending on the argument structure of the verb.

(31) a. Gi mai [yit bun su ] [bun ngai].

he buy one CL book to I

He bought a book to me. '

b. Gi mai [yit bun su ].

he buy one CL book

He bought a book. '

Both sentences in (31) are syntactically well-formed. In (31 a) the argument structure of the verb requires three participants: an agent,a theme and a recipient to be involved in the activity; in (31 b) though,only two participants are required: an agent and a theme. We argue that thebun heading PP in (31 a) is a complement of the main

verb because we are allowed to swap the position of the PP with the NP which semantically denotes the theme. As demonstrated in (32),when two constituents co-occur at the same side of the predicate, the complement must be closer to the predicate than the adjunct does. If the PP can cross the objective NP to the immediate right ofthe verb,it must take the syntactic function as a complement ofthe verb.

(17)

Tsem,.:OTonHakk且 PrenositionalPhrases

(32) Gi mai [bun ngai] [yit bun su ].

he buy to I one CL book

He bought me a book.'

Some complement PPs may be preverbal,in this paper they are classified as the fourth type of PPs. Prepositions such as tung, tau and ziong,when occurring in the preverbal position,may be used as a patient marker to bring out an essential argument to the verb. Another preposition having the similar function is the morpheme bun. When used in the passive construction,it functions as an agent marker indicating the demoted argument ofthe verb. This type ofPPs is realized as the oblique complement of the verb. They differ from the first two types of preverbal PPs in two ways. First,

unlike the adjunctP恥, the occurrence of a preverbal complement PPs is not optional. The examples in (33) and (34) below show the obligatory occurrence ofthe preverbal PP complements headed respectively by the patient and the agent marker,which in comparison with the optional PP adjuncts,is required by the head verb.

(33) a. Gi [tung ge gien diam] [ziong suigo] mai-ciang-ciang.

he for that CL store PAT fruit buy-clear-clear

HeclearecL句oughtall the fruit in that store.' b. Gi [ziong suigo] mai-ciang-ciang.

he PAT fruit buy-clear-clear

Hecleare cL句oughtall the fruit.'

c. *Gi [tung ge gien diam] mai-ciang-ciang.

he for that CL store buy-clear-clear

Hecleare cL句ought(all the things) in that store.'

(34) a. Gi [di libaingit] [bun gia moie ] rna.

he on Sunday PASS his daughter scold

He was scolded by his daughter on Sunday.'

b. Gi [bun gia moie ] rna.

he PASS his daughter scold

He was scolded by his daughter.' c. *Gi [di libaingit] rna.

he on Sunday scold

He was scolded on Sunday.'

Both (33b) and (34b) are syntactically acceptable with the preverbal adjunct PP being deleted. However, the two (c) sentences in (33) and (34) tum out to be unacceptable when the PP undergoing deletion is the complement PP.

(18)

le I!!且且且且1111138.2(November 2012)

Second,when both the adjunct and the complement PPs co-occur in the same VP, the precedence order must be specified in a designated way. Consider the examples in (35) and (36),which show that the complement PP headed by the patient marker ziong or the agent markerbungenerally follows the different kinds of adjunct PPs.

(35) a. Gi [tung gia moie ] [ziong cien ] van-tet Ie.

he for his daughter PAT money return-RYe PART

‘He paid off all the money for his daughter. '

b. *Gi [ziong cien ] [tung gia moie ] van-tet Ie.

he PAT money for his daughter return-Rye PART

‘He paid off all the money for his daughter. '

(but: He used the money to payoff the debt for his daughter.)

。 6)a. Gi [di gungyen] [bun gieue] dui den.

he in park PASS dog chase ASP

‘He is being chased by the dog in the park. '

b. *Gi [bun gieue] [di gungyen] dui den.

he PASS dog in park chase ASP

He is being chased by the dog in the park.'

(but with a contrastive focus reading:Itis in the park (not any other location) where he is being chased by the dog.)

Thus,the well-forrnedness of a Hakka sentence prefers the oblique complement to be closer to the verb while at the same time pushing the adjuncts to the more peripheral position

Generalizing all these four kinds ofPPs into the syntactic scheme ofthe same VP, the structural template can be depicted as in(37):

(37)

/ \ \

叮m/\

dildo-phrase ↓ ε(小 ADJ) ~

( b e n J m i t f g o a l ) m T p a t H / \

au/tung/ziong/bu叫hrase V (小 OBLdirl間)=↓

PP

do/bun-phrase (directiona l/recipient)

(19)

Tseng: OT on Hakka Prenositional Phrases

In this structure,four 句pes of PPs are recognized. The first two types appear in the adjunct position,which left-adjoin to the VP. Another two types of PPs involve complementation, which can be found as the verb selects a preverbal PP as its complement to express the agent or the patient role,or in cases when a postverbal PP is selected to be the complement providing additional information about the direction or the recipient as an extension of meaning required by the verb.

3. The Optimality Theoretic analysis

In this section,an analysis based on the Optimality Theoretic approach (Prince & Smolensky 1993) is proposed. Before we are ready to develop a syntactic OT analysis on the distribution of different kinds of PP within the VP construction,we would first propose a semantic OT analysis with the purpose to assign appropriate syntactic role,

i.e. complement or adjunct, to a PP which exhibits semantic parsing in

correspondence with the subcategorization of verbs. The tableau evaluation of a syntactic OT approach pursues the

meaning to form" derivation. In this version of analysis the given input is the meaning representation of a syntactic construction, which generated a set of semantically equivalent syntactic forms competing for the grammatical output (Grimshaw 1997). The semantic OT approach, in contrast,

pursues the "form to meaning" derivation; i.e., a grammatical form is taken as the input for the tableau analysis, generating alternative meaning representations as candidates to be evaluated, and the optimal meaning could be derived through candidate competition (Hendriks& de Hoop 2001).

In the Hakka case concerning the distribution of PPs,I suggest that we adopt the OT-based analysis from both directions; specifically, a bidirectional approach that provides the capacity for two rounds of optimization process (Blutner 2000). The first round of optimization is based on the OT-semantics,with which the complement and adjunct interpretations compete for the better representation of a given PP form,and the subcategorization frame of the verb is determinant in defining the constraints in effect. The second round of process is based on the OT-syntax. All the PPs at work are now associated with either the complement or adjunct interpretation, the association has been decided earlier through the previous round of optimization. The possible permutations of these PPs then compete with each other to derive the grammatical word orders. The first round of optimization is presented in section 3.1,

which also provides an extra analysis to illustrate the semantic functions and restrictions with reference to the three semantically related morphemes lau,tungand

ziong.The following sections from 3.2 through 3.5 deal with the second round of OT process. In this round the syntactic analysis is further organized into four stages. The

(20)

I

{J

iliiii4iiiili

38.2(November 2012)

first stage generalizes the positioning of postverbal PP complements and preverbal PP adjuncts with three Generative Alignment Constraints. The second stage deals with the problem that a certain type of complement PPs empirically appears in the preverbal position. The solution is to further propose a more specific constraint for the PP complements based on their semantic meanings and functions. The third stage handles the word order between the two types of PP adjunctions marked with different semantic functions. At this stage,again,a more specific constraint for the PP adjuncts needs to be proposed to help filter out the unwanted permutation during the process of candidate competition. The final fourth stage focuses on the temporal prepositional constructions involving the morphemedi. The analysis proposed in this stage will deal

with the situation that the temporal di-phrases differ from the other PPs in that they may occur at the sentence initial position without the overt presence of the head prepositiondi.

3.1 First round of optimization-semantic OT

The PPs do not inherently pair with a complement or adjunct interpretation; on the

other hand, the interpretation is by and large determined by the strict

subcategorization statement (Chomsky 1965) of a given verb in a clause. This is illustrated in (38):

(38) a. sung:V ‘give as agi缸,

[+一 NP NP NP] [+一 NP NP PAD且 ] <+agent><+recip> <+theme> b. dui:V ‘chase' <+agent> <+theme><+recip> [+一 NP NP] [+ NP nrD且 ] <+agent><+patient> <+patient><+agent> c. mai-ciang-ciang: V

clear (buy all the items provided)'

[+一_ NP NP] [+一_ NP PP]

<+agent><+them巴〉 <+agent><+patIent>

In above examples, each lexical entry contains the syntactic information which includes what syntactic category the word belongs to and what statements/arguments are required in the item's context. In addition,the semantic representation associated with each selected statement is also expressed in these entries. Itshould be noted that the verb subcategorizations presented in (38) do not intend to show the position of verbs with respect to other required statements.

(21)

Tsem!: OT on Hakka Prenositional Phrases

We tum now to the semantic features ofprepositional phrases. Previously in Table I,we have shown the association of the five commonly used Hakka prepositions with their grammatical functions. Below in (39) we will incorporate these indicated functional information into part oftheir lexical properties:

(39) a. di: P (• THEMATIC)=TEMP [+_ NP]

b. di: P (• THEMATIC)=LOC [+_ NP]

c. do: P (• THEMATIC)=SOUR [+_ NP]

d. do: P (• THEMATIC)=LOC [+_ NP]

e. do: P (• THEMATIC)=GOAL [+_ NP]

f. lau: P (• THEMATIC)=SOUR[+_ NP]

g. lau: P (• THEMATIC)=GOAL [+_ NP]

h. Jau: P (• THEMATIC)=BEN [+一_ NP]

i. lau: P (• THEMATIC)=PAT [+_ NP]

j. tung: P (• THEMATIC)=SOUR [+_ NP]

k. tung: P (• THEMATIC)=GOAL [+_ NP]

1. tung: P (• THEMATIC)=BEN [+_ NP]

m. tung: P (• THEMATIC)= PAT [+_ NP]

n. ziong: P (• THEMATIC)=PAT [+_ NP]

o. bun: P (• THEMATIC)=AGENT [+_ NP]

p. bun: P (• THEMATIC)=RECIP [+_ NP]

When given the functional structures as (40):

(40) ( PRED sung‘give as a gift' [+一 NP NP PP]

<+agent> <+theme> <+recip> NP1 gi

he' [+agent]

NP2 yit liap lingo‘an apple' [+theme]

PP1

r

PRED 圳州dii川川(什↑ THEMAT叩TI昀C)=TEMP[+一 N附叫P凹] - OBJ amηlbut仿eu ‘ m喀ght'

PP2 [PRED bun (• THEMATIC)=

REC昕一 NP]J

OBJ gi

him'

A semantic OT analysis offers an approach to determine the complement or adjunct interpretation on the two involved PPs. The constraints comi月 into play include *Subcategorize-COMP and *Subcategorize-ADJ. The former requires all but only the complements of a verb to be specified in the subcategorization frame of the verb, while the latter requires all and only the a句uncts to be noted in the subcategorization frame. The constraints ranking and interaction can be seen in (41) and the tableaux right below in (42) and (43).

(22)

Ie

'';'''4h;.tl

38.2(Novemher 2012)

(41) *Subcategorize-COMP: A verb must be subcategorized for all and only its complements.

*Subcategorize-ADJ: A verb must be subcategorized for all and only its adjuncts.

Constraints ranking: *Subcategorize-COMP» *Subcat巴gorize-ADJ

(42)

PP 1: di ambuteu *Subcatel?:orize-COMP *Subcatel?:orize-ADJ

Complement *'

<7"Adjunct *

(43)

PP2: bun gi *Subcategorize-COMP *Subcategorize-ADJ

<7"Complement *

Adjunct *'

As shown in (40),the predicatorsung is subcategorized for a PP with recipient denotation. Therefore in (42),if the temporal PPdi ambuteureceives the complement interpretation, presented by the first candidate, it must be ruled out since the complement PP is not specified in the subcategorization frame of the verb. Similarly, in (43),ifthe recipient PP bungi receives the adjunct interpretation,as illustrated by the second candidate,it must be ruled out since the adjunctive PP is specified in the subcategorization of the verb sung. Under this evaluation,the optimal mapping is di

ambuteu 今 adjunct; bungi 今 complemen t.

When a verb like dui is used in the passive voice,it is subcategorized for a patient NP and an agent PP,as presented in (44):

(44) ( PRED dui

‘chase' [+_

NP PP]

<+pattent> <+agent> NPI gi

‘he' [+patient]

PPJ

r

PRED di (• THEMATlC)=LOC [+_ NP]

I

~ORJ gungyen‘park'

PP2

r

PRED bun (• THEMATIC)=AGENT [+_ NP]

I

- OBJ gieue ‘ dog' 一

The OT analysis pairs the locative PP with the adjunct inte叩 retation, and the PP denoting the agent role with the complement interpretation, as illustrated in the following(4月 and(46):

(23)

TsenQ: OT on Hakka Prenositional Phrases

(45)

PP I: di gungyen *Subcategorize-COMP *Subcategorize-ADJ

Complement *'

"'"'Adjunct *

(46)

PP2: bun gieue *Subcategorize-COMP *Subcategorize-ADJ

"'"'Complement *

Adjunct *'

The last case is provided with the verbal predicator mai-ciang-ciang,which may have the following functional structure when used in sentences:

(47) ( PRED mai-ciang-ciang‘clear(buy all the items)'[+一_ NP PP] <+agent> <+patlent> NPl gi

‘he' [+agent]

PPl

I

PRED di (• THEMATIC)=TEMP [+_ NP]

I

'-- OBJ cobungit

‘yesterday'

----PP2

r

PRED tung (• THEMATIC)=BEN [+_ NP]

I

'-- OBJ gegien diam ‘that store'

PP3

r

PRED zio喀什 THEMATIC)=PAT [+_ NP]

l

- OBJ suigo ‘ fruit' 一

When three different kinds ofPPs co-occur in the same YP,again the optimization process associate both the temporal and the benefactive PPs with adjunctive interpretations, and here only the PP of patient role indication is interpreted as the complement of predicator. See the three tableaux presented in (48) throughout (50)

(48)

PP I: di cobungit *Subcategorize-COMP *Subcategorize-ADJ

Complement *'

"'"' Adjunct *

(49)

PP2:tun貝貝egien diam *Subcategorize-COMP *Subcategorize-ADJ

Complement *'

G7Adjunct *

(50)

PP3: ziong shuigo *Subcategorize-COMP *Subcategorize-ADJ

G7Complement *

Adjunct *'

(24)

1 6'1!且且且且且且團 38.2(November 2012)

The OT approach developed in this section applies the constraint hierarchy that is based on the subcategorization frame of the predicator verb to all the PP constituents enclosed in the functional structure of the verb. As stated earlier at the beginning of section 3, the purpose is to predict the complement or adjunct interpretation on a givenPP form,which as will be shown in the succeeding section,is pretty crucial in deciding the appropriate syntactic position when we are about to deal with their relative order within clauses.

Finally,at the end of this section,we solve a subsidiary problem and that is the distinction between three semantically related morphemes lau,tung and ziong when

used as a patient marker. To recapture the previously mentioned semantic restrictions related to the three morphem白, I repeat (51) and (52), the earlier examples listed

respectively in(14),(15) and(17).

(51) a. Gi lau/?tung ge bun su mai-zeu Ie.

he PAT that CL book buy-RVC PART

He has bought that book. '

b. Gi lau/?tung sinsang hi do mien-fung-fung.

he PAT teacher irritate COMP face-red-red

‘He irritated the teacher to the degree that the teacher's face has turned red.'

(52) a. Gi ziong ge bun su mai-zeu Ie.

he PAT that CL book bUY-RVC PART

'He has bought that book.'

b. Gi ziong sinsang hi do mien-fung-fung.

he PAT teacher irritate COMP face-red-red

He irritated the teacher to the degree that the teacher's face has turned red.' The examples in (51) illustrate the inappropriateness of using the prepositiontung

in the context where the following NP denotes an inanimate object (51a),or in cases when thePP is encoded with the malefactive meaning (51 b),but the prepositionlau is

not su句 ect to the same restrictions. The examples in (52) show that the use of prepositionziong is similar to lau,which is free from the restrictions that have been imposed upontung.

The explanation for the semantic restrictions on tung lies with the agreement of

semantic features between the preposition and its subcategorized NP statement. Consider the following feature specification of the three prepositions under discussion:

(25)

Tsenl!:OT onHakk 且 Prenositional Phrases

Table 2. Feature specification oflau/tung/ziong'(• THEMATIC)=PAT'

semantic feature lau tung ziong

(animate] [αanimate] [+animate] [αanimate]

r

malefactivel [αmalefactive] malefactive] [αmalefactive]

With this feature specification in mind,the ungrammaticality oftungappeari月 III

(51a) and (5Ib) can be attributed to the disagreement in semantic feature between the predicator preposition and the NP,as depicted in (53) and (54):

(53)

~\\

~\\↑!↓

P

ζ三三〉

mai-zeu Ie tung (•THEMATlC)=PAT (↑ ANIMATE)~+ ge bun su (•PRED)= 'bought' [+_ NP PPJ <+agent> <+patlent> (• ANIMATE)=- (↑ANIMATE)~

\投/

agreement (54) VP

/ \ \

中,\ 恥 V/ hi do mien-fung-fung P

4三三〉

(↑ PRED)~ 'irri扭扭'[+_ NP PP ] <+agent> <+pat1ent>

(•MALEFACTIVE)= + tung + E +V F 口 們UA 曙 ATY 叫 Mulj// n 山 A / 心劃盯圳 M / ••/ TL-A 宇 nrrE 、 =+W\ 前羊口 l\\ VA 叫 11 MMH ELLWYL mMHM W 料科 agreement 195

(26)

le I!且RlnIlll面 38.2(November2012)

The tree in (53) corresponds to the ungrammatical sentence (51 a), in which the preposition tung disagrees with the NP in [animate] feaωre; while the tree in (54) explains the ungrammatical expression (51 b) by showing their disagreement in [malefactive] feature. With an OT model,we can propose the following Faithfulness constraints on features agreement, which outrank the constraint IDENT to force segment replacement when the semantic agreement between the preposition and its NP complement breaks.

(55) Faith-SemF(animate): Semantic agreement must be reached between the preposition and its subcategorized NP in the feature [animate].

Faith-SemF(malefactive):Semantic agreement must be reached between th巴

preposition and its subcategorized NP in the feature [malefactive]

IDENT: Identical values of all the features must be preserved between input and output

The tableaux shown in (56) and (57) work through the cases of (53) and (54), showing that the OT can capture the phenomenon mentioned above.

(56)

Faith-SemF Fa剖it仙h吋 emF

!DENT (animate) (malefactive) . .tung ge bun su

*'

G" . •.Iaug巴 bunsu

*

G " . ..zlOn旦旦ebun su

*

(57) Faith-SemF Faith-SemF lDENT (animate) (malefactive) .札mgSInsang

*'

G" • •.Iau sinsang. . .

*

G " . ..zlOng slllsang...

*

The ranking Faith-SemF(animate)

,

Faith-SemF(malefactive) » IDENT

predicts the replacement oflau or ziongover the inappropriate tung in the semantic context where the complement of the preposition is inanimate,or in cases when the PP functions to express a malefactive event. The former situation fatally violates the constraint Faith-SemF(animate) as shown in (56), while the latter fatally violates Faith-SemF(malefactive),as in (57),both irretrievable violations save the violation on IDENT,rendering the change of linguistic elements possible.

(27)

Tseng: OT on Hakka Prenositional Phrases

3.2 Second round of optimization-syntactic OT

Once we have decided the complement or adjunct interpretation for each PP enclosed in the VP construction,we can continue to the second round of syntactic OT analysis. As shown below, the development of a syntactic OT approach on the distribution ofPPs can be divided into four stages.

3.2.1 Stageon令-preverbaladjunction and postverbal complementation

As mentioned in section 2, Hakka adjunct PPs are always found in preverbal positions, while complement PPs preferably occur in postverbal positions. By adopting the Optimality Theoretic analysis, this word order generalization can be accounted for through the interaction of three Generalized Alignment Constraints (McCarthy & Prince 1993a,b),as defined in (58):

(58) ALIGN-L (X

,

XP): Align X with the left edge ofthe XP that dominates it. ALIGN-L (YP

,

XP): Align YP

,

which appears to be an adjunct of XP

,

with the

left edge ofthe XP that dominates it.

ALIGN-L (ZP

,

XP): Align ZP

,

which appears to be a complement of X

,

with the left edge ofthe XP that dominates it.

The first positional constraint ALIGN-L (X

,

XP)

,

which functions to enforce a non-phrasal head node X to the left edge of its phrasal projection XP,must outrank ALIGN-L (ZP

,

XP)

,

which enforces all complements of X being positioned at the initial position of the same XP projection. The constraint ranking ensures the structure of both VP and PP to be left-headed and right-branching. On the other hand,knowing that in Hakka a VP allows only t伽he left adjoint of PP a吋吐4你恥j抓戶un江In叫1

ALIGN-L (YP

,

XP)

,

which enforces the adjuncts of an XP to be phrasal initial

,

must outrank ALIGN-L (X

,

XP)

,

so that a PP adjunct is allowed to left adjoin to a VP and thus it may precede the head verb within the VP domain. The proposed constraint ranking is generalized in (59):

(59) ALIGN-L (YP

,

XP) » ALIGN-L (X

,

XP) » ALIGN-L (ZP

,

XP)

Now given the following two grammatical sentences in (60),both sentences show an example of the ditransitive construction,which involves a ditransitive verb sung

followed by two complements. The direct object is realized by an NP,and the indirect object is realized by a PP headed by the preposition bun. By comparing the two sentences in (60),we can find that the relative word order between the two postverbal

(28)

IG I!且~且D:I'量 38.2(November2012)

complements, the direct objective NP and the indirect objective PP, is syntactically free; namely,the bun-heading PP can freely occur before or after the objective NP

(60) a. Gi [di ambuteu] sung [yit liap lingo] [bun gi]

he at night give.as.gi ft one CL apple to he

‘That kid gave an apple to him as a present at night.'

b. Gi [di ambuteu] sung [bun gi] [yit liap lingo].

he at night give.as.gi ft to he one CL apple

‘That kid gave to him an apple as a present at nigh

t.'

The grammaticality of the two sentences in (60) can be accounted for by the OT-based analysis presented in the following tableau (61):

(61 )

Verb: sung

‘gIVe as

gl 缸,

ADJ: di ambuteu

‘at night'

ALlGN-L ALlGN-L ALIGN-L

COMPI: yit liap linggo 'an apple' (yP,XP) (X,XP) (ZP,XP)

COMP2: bun gi

‘to him'

sung [di ambuteu] [yit liap linggo] [bun gil *' *****

[bun gil sung [di ambuteu] [yitliapIinggo] *'*

***

[di ambuteu] [bun gil sung [yitliapIinggo] **' ****

[di ambuteu] [bun gil [yit liap linggo] sung **'* ***

。 [diambuteu] sung [bun gil [yitliap linggo] * *****

。 [diambuteu] sung [yitIiap linggo] [bun gil * *****

It should be noted that in this paper, the evaluation of Generalized Alignment Constraints takes into account the degree of violation. Taken in this sense,violations of all the alignment constraints proposed in (58) should be calculatedcumulatively,的

the constraints are designed to measure the distance between the two designated edges of the referring categories

From the above tableau, we can see that assigning the highest ranking to the constraint ALIGN-L (YP,XP) rules out the first two candidates in which the adjunct PP appears in any other position except phrase initial. Then, the dominance of ALlGN-L (X,XP) over ALJGN-L (ZP,XP) in the constraint hierarchy ensures the head verb to appear in the second position right after the adjunct PP,which Ie代 the

last two candidates as the grammatical outputs. The tableau analysis successfully accowlts for the two well-formed sentences in (60),where the adjunct PP occupies the left-most position in the VP, followed by the head verb, and then the two phrasal complements. The precedence order of the two postverbal complements is unconstra ined.

(29)

den.

mal-clang-clang. buy-clear-clear

Tseng: OT onH且kk且 Prenositional Phr且ses

3.2.2 Stage two--preverbal complementation

It may sound ideal to identify the syntactic function of PPs as complements or adjuncts according to their position in relation to the head verb; that is, all the

preverbal PPs are verbal adjunc的, while all the postverbal PPs are verbal

complements. However,it is indeed an overgeneralization to build such an absolute association,as demonstrated in (62) below:

(62) a. Gi [tung ge gien diam] [ziong suigo]

he for that CL store PAT fruit

Hecleare 也借oughtallthe 仕uitin that store. '

b. Gi [di gungyen] [bun gieue] dui

he in park PASS dog chase ASP

He is being chased by the dog in the park. '

The two sentences in (62) repeat the previous examples listed earlier in(3 月 and

(36) in this paper. As it has been argued that the preverbal PP projected by the patient marker ziong in (62a) and the PP projected by the passive marker bun in (62b) are

肌肉allycomplements ofthe verb,the two sentences display counter-examples against the overgeneralization addressed in the previous paragraph.

Now if we go back to the theoretic model we have developed so far,we find that the aT account that has been established up to this point fails to generate the two grammatical sentences in (62), where the PP complement appears preverbally preceding the head verb. The wrong prediction can be seen in the following tableau (63), indicated by the symbol

®

in all the following tableaux. In this tableau the ungrammatical pattern "ADJ Verb CaMP" 的 problematicallyselected as the optimal output when the PP complement is expressed by a ziong-phrase functioning to denote the patient role or by a bun-phrase denoting the demoted agent role. We know the output generated by the tableau in (63) is incorrect because according to (62),the only grammatical pattern allowed in this case is“ADJ CaMP Verb,"where both the verbal adjunct and complement precede the head verb.

(30)

Ie

l.ilii§ lihtl38.2 (November 2012) (63)

Verb: mai-ciang-ciang‘clear out'/dui‘chase'

ALlGN-L ALlGN-L ALlGN-L ADJ:tung .(benehcIIVE)/dl(locatlivvce)) (YP,XP) (X,XP) (ZP,XP) CaMP:ziong... (patient)/ bun... (pass

*' **

Verb AD] CaMP

*'* *

Verb CaMP AD]

* **

@GP'ADJ Verb CaMP

CaMP Verb AD] *'* *

**' *

@ AD] CaMP Verb

CaMP AD] Verb *' **

※®

=

wrong predictions

A solution to this is to fu 閃her propose a more specific constraint ALIGN-L (PPAgent/Patient,X P). This constraint requires the PP complement of verbs surfaces in the phrase initial position as long as the PP functions to indicate the patient or the agent role for the semantic domain ofthe head verb. The constraint is defined in (64):

(64) ALlGN-L (PPAgcnt/Paticnt

,

XP): Align PP with the left edge ofthe XP that dominates it iffthe PP is a complement ofX and it is associated with the agent or the patient semantic role

Now the new aT tableau successfully generates the correct grammatical output pattern. See the following tableau in(6月 for the constraints interaction:

(65)

Verb: mai-ciang-ciang / dui

ADJ: tung... / di... ALlGN-L ALlGN-L ALlGN- L ALlGN-L

COMP:zlongnt()pat1En1) (YP,XP) 叭,XP) (PPAglPat,XP) (ZP,XP) bun... (age

Verb AD] CaMP *' ** **

Verb CaMP ADJ *'* : * *

ADJ Verb CaMP * ** **'

CaMP Verb AD] *'* *

.

。 ADJ CaMP Verb ** * *

CaMP AD] Verb *' ** **

In this tableau the equal ranking between ALlGN-L (X, XP) and ALlGN-L (PPAgentlPaticnt,XP) is decisive in saving the fatal violation identified in the previous tableau (63) on the grammatical pattern

“AD] CaMP Verb." The candidate

AD] Verb CaMP" collects double violations on the constraint ALlGN-L (PPAgentIPaticnt,

XP) by positioning the PP complement postverbally in the phrase final position, which balances the violations of the candidate "ADJ CaMP Verb" on the constraint

(31)

TsenQ: OT onH且kkaPrenositional Phrases

ALIGN-L (X

,

XP) where the PP complement precedes the verb

,

and the verb occurs phrase-finally. Then the constraint ALIGN-L (ZP

,

XP) optimally selects "ADJ COMP Verb" over

ADJ Verb COMP" given the fact that the former candidate has the complement phrase closer to the left edge ofthe VP. The analysis provided in (6日

successfully selects the grammatical pattern“ADJ COMP Verb" as the optimal output based on this proposed constraint ranking: ALIGN (YP

,

XP) » ALIGN (X

,

XP)

,

ALIGN-L (PPAgentJPatient

,

XP»> ALIGN-L (ZP

,

XP).

3.2.3 Stage three--two types of preverbal adjunction

In the earlier section 2.5, I have mentioned that the preverbal adjunct PPs can actually be divided into two kinds. The temporal and locative PPsp叫 ectedby di and

do belong to one kind,and the PP constructions involving the head prepositions tau and tung with the function to introduce a source,goal and benefactive argument to the verb are members of another kind. Ithas also been maintained that if the two kinds of PPs co-occur modifying the same VP, the temporal/locative PP must precede the source!goal!benefactive PP to appear in the more peripheral position.

While the OT analysis established so far works pretty well to generate the positioning of different types of PPs,a problem arises as it fails to account for the restricted word order when the structure of a VP contains simultaneously the PP

吋unctionof these two kinds of preverbal PPs,as shown in (66). The sentence(66的

involves three preverbal PPs. The PP which is closest to the main verb,headed by the preposition ziong,is a preverbal complement bringing out the patient argument to the verb. The other two PPs are a甸 uncts. The di-phrase functions to indicate temporal information,and the tung-phrase expresses the benefactive meaning,stating that the following NP object is benefiting from the action of VP. As shown in (66a), the temporal PP must precede the benefactive PP; if we switch the position of the two preverbal adjuncts,the sentence becomes unacceptable,see (66b).

(66) a. Gi [di cobungit] [tung ge gien diam] [ziong suigo] mai-ciang-ciang.

he in yesterday for that CL store PASS fruit buy-clear-clear

He cleared!bought all the fruit in that store yesterday.'

b. *Gi [tung ge gien diam] [di cobungit] [ziong suigo] mai-ciang-ciang.

he for that CL store in yesterday PASS fruit buy-clear-clear

He cleared/bought all the fruit in that store yesterday. '

(but with a contrastive focus reading: Itis yesterday (not any other time) when he cleared/bought all the fruit in that store.)

(32)

le I!且~且II區 38.2(November2012)

Given the above two sentences in (66), the analysis developed up to this point, however,wrongly allows the free ordering between the two adjunct PPs. The tableau provided in (67) below shows this false prediction:

(67)

Verb: mai-ciang-ciang

ADJ I: di... (temporal) ALlGN-L ALlGN-L ALlGN- L ALlGN-L ADJ 2: tung...(benefactive) (yP,XP) (X,XP) (PPAg/Pat,XP) 帥,XP) COMP: ziong... (patient)

ADJ I ADJ2 V COMP

*

**

***

***'

ADJ2 ADJ I V COMP

*

**

***

***'

G"ADJl ADJ2 COMP V

*

***

**

**

<8>G"ADJ2 ADJ I COMP V

*

***

**

**

COMP ADJ I ADJ2 V

**'*

***

ADJ I COMP ADJ2 V

**'

***

*

*

ADJ I V ADJ2COI\徊3

**'

*

*-

***

V ADJ I ADJ2 COMP

**'*

!

***

-*

The evaluation shown in (67) problematically selects the candidate

ADJ2 ADJI COMP V" 的 the equally optimal ordering pattern, indicated by the symbol <8>. According to (66),when a temporal di-phrase and a benefactive fung-phrase co-occur

in the same VP,the former PP must precede the later one; the swap of position will lead to ungrammaticality.

To eliminate the ungrammatical pa位em from the list of grammatical outputs, I propose another constraint ALlGN-L (PPTemp/Loc,XP),which requires any adjunct PP encoding temporal or locative information to appear at the left edge of VP. Thi constraint is defined in(68)below:

(68)ALlGN-L(PPTemplL肘, XP): Align PP with the left edge ofthe XP that dominates it iff the PP is an adjunct of XP and it i used to express the temporal or the locative information to further modi fy the verb

ow,see the following tableau in(69)for the revised constraints ranking and Il1teractlon:

(33)

TsenQ: OT on Hakka Prenositional Phrases

(69)

Verb: mai-ciang-ciang ADJ 1:di... (temporal)

ALlGN-L ALlGN-L

ADJ 2: tung..

(PPTempl ALlGN-L ALlGN-L 伊 PAglP at, ALJGN-L

(benefactive) (YP,XP) 阱, XP) (ZP,XP)

CO MDE 7L0nb。n 、'‘

Loc,XP) XP)

(patie

ADJ I ADJ2 V CaMP * ** *** ***!

ADJ2 ADJ1 V CaMP *!

-

*** **會

Gi"ADJ1ADJ2 CaMP V * *** ** **

ADJ2ADJ1CaMP V *!

*** ** **

CaMP ADJ1ADJ2 V *! *** ***

ADJ1CaMP ADJ2 V **! ***

*

ADJI V ADJ2 CaMP **!

: *** **會

V ADJJADJ2 COMP *! *** *** ***

In (69), the ungrammatical pa前em “ADJ2ADJl CaMP V" is now ruled out by positioning the temporal PP headed bydi to the right of the benefactive tung-phrase,

leading to a fatal violation on the newly proposed high-ranking constraint ALIGN-L (PPTemp/Loc,XP). Thus,the optimization process left only the pattem“ADJI ADJ2 CaMP V" 的 the winning output. The constraint hierarchy is summarized in the following(70)

(70) ALlGN-L (PPTemp/Loc,XP)» ALIGN(yP,XP)» ALlGN阱, XP),

ALlGN-L (PPAgent/Patient,XP»> ALIGN-L (ZP,XP)

3.2.4Stage four-the sentence initial di-construction

As shown previously in example(7a),when the prepositional di-phrase is used as a time indicator,it may occur at the sentence initial position preceding both the verb and the subjectNP.The cons仙ction is nevertheless subject to the condition that the functional di may not be overtly presented in the sentence. Some examples are listed in the following(71):

(71) a. (*Di) Iibaingit gi voi hi hokgau tuk yinvun.

on Sunday he will go school study English

‘He goes to school studying English on Sunday. b. (*Di) cokbungit gi siit-tet sam van fan.

on yesterday he eat-RvC three CL nce

‘He ate three bowls of rice yesterday.'

數據

Table 1. Preverbal!Postverbal position of PPs
Table 2. Feature specification of la u/ tung/ziong '(• THEMATI C)= PAT'

參考文獻

相關文件

When we want to extend an operation from functions to distributions — e.g., when we want to define the Fourier transform of a distribution, or the reverse of distribution, or the

When the relative phases of the state of a quantum system are known, the system can be represented as a coherent superposition (as in (1.2)), called a pure state; when the sys-

In JSDZ, a model process in the modeling phase is treated as an active entity that requires an operation on its data store to add a new instance to the collection of

6 《中論·觀因緣品》,《佛藏要籍選刊》第 9 冊,上海古籍出版社 1994 年版,第 1

The first row shows the eyespot with white inner ring, black middle ring, and yellow outer ring in Bicyclus anynana.. The second row provides the eyespot with black inner ring

If a contributor is actively seeking an appointment in the aided school sector but has not yet obtained an appointment as a regular teacher in a grant/subsidized school, or he

• In 1976 Hawking argued that the formation and evaporation of black holes leads to a fundamental loss of information from the universe, a breakdown of predictability, as

Strange as it may sound, the little girl escaped being hurt in the accident.. (A) the little girl was hurt when escaping in the accident (B) the little girl didn’t get hurt in