CHAPTER 7

DIRECTIONAL SVC

Now, let’s consider examples of Directional SVCs such as (1) and (2) below.

(1) wo tuei ta xia-qu.

I push him down-go

‘I pushed him down.’

(2) wo na yizhi bi chu-lai

I take one pen out-come

‘I take out one pen.’

The word order of this construction is NP Vt O Vi lai/qu. The first verb is

always transitive, and the second verb is a motional intransitive verb, followed by a

directional verb, either lai (come) or qu (go). Even though lai and qu can be a full

predicate, the constituent Vi + lai/qu behaves like a semantically complete verb.

Hereafter, we will refer the whole constituent of Vi + lai/qu to VP2 or CDC33

33
Li (2001) named the constituent Vi + lai/qu as CDC (Le comple ment directionnel complexe du
verbe; complex directional complement).

106
(complex directional complement).

(3) a. ta qu/lai le

He go/come PERF

‘He went/came.’

b. Xiaohua chu-qu le.

Xiaohua out-go PERF

‘Xiaoxua went out.’

On the other hand, example (1) appears to be very similar to an RSVC in

Baker’s and Stewart’s sense. Nonetheless, resultative or causative constructions in

modern Mandarin Chinese normally surface in a different word order, i.e. as verb

compounds, namely, resultative verbal compounds (RVCs). This is illustrated by the

contrast of (4a) and (4b). A result VP separate from the cause (VP1) by an object NP,

such as (1), is comparatively rare, and the form seems to limit to conditions when the

result VP is a CDC.

(4) a. ta sha-si ren-le.

He kill-die person-PERF

107
 ‘He killed someone dead.’

b. *ta sha ren si-le.

He kill person die-PERF

In addition, example (2) shows that VP2 can be simply a direction of the first

action (V1), not a result. Hence, VP2 can be a result like (1), or a pure direction like

(2). Since VP2 is directional in nature, we classify both sentences (1) and (2) into

Directional SVCs.

Finally, these examples cannot be interpreted in other way, such as CCs, since

there is no subject sharing between the two verbs. Nor can VP2 in this construction

be regarded as a purpose clause, since the subject of a purposive is normally

controlled by the matrix subject. Contrary to all the previously mentioned Chinese

SVCs, this type involves only object sharing, where the object of V1 is semantically

the subject of VP2.

7.1. Varieties of the Construction

In the ancient China, there was a period of time when resultative constructions

appeared in the form of RSVCs, where there was an object NP mediating between the

108
two verbs. Mei (1981) demonstrated that, in Tang dynasty, the result phrase can

appear either after the object NP, as (5), or before it, as (6).

Vt + O + result

(5)a. 吹 歡 [情人] 羅裳 開

Chui xuan qin-ren luo-shang kai

Blow joy lover clothes open

‘(Someone) blows open the lover’s clothes joyfully.’

(子夜四時歌Zi ye shi shi ge, 5AD)

b. 當 打 汝 口 破

dang da ru kou po

should hit you mouth break

‘You deserve to be hit on the cheeks until they broke.’

(幽明錄 You ming lu, 5AD)

Vt + result + O

(6)a. 折 斷 綠 楊 枝。

Zhe duan lui yang zhi

Snap break green poplar branch

‘(Someone) snapped a branch of the green poplar.’

109
 (李白《宣城送劉副使入秦》Libai, shuanchen song liu fushi ru qin,1810)

b. 今日 壓 倒 元白 矣。

Jin-ri ya dao yuan-bai yi

Today press fall Yuanbai YI

‘Today (someone) pressed down Yuanbai.’

(《唐遮言》Tang Zhe Yen)

In modern Chinese, the constituent V + CDC can surface in the following word

orders (Li, 2001). Take na chu lai 'take out’ for example:

A. V+O+ Vi+lai/qu

(7) Na [yizhi bi] chu lai

Take one pen out come

‘Take one pen out.’

B. V+ Vi+O+lai/qu

(8) Na chu [yizhi bi] lai

Take out one pen come

‘Take one pen out.’

C. V+ Vi+lai/qu +O

110
(9) Na chu lai [yizhi bi]

Take out come one pen

‘Take one pen out.’

D. BA+O+V+ Vi+lai/qu

(10) Ba [yizhi bi] na chu lai

BA one pen take out come

‘Take one pen out.’

(Li, 2001, p.180)

Examples (7)-(10) show that the relative positions of the object NP and the

V+CDC are quite free. The object NP can be between the verbs (7-8), after the

verbs (9), or preceding the verbs in a BA-construction (10). Specifically, the

behaviors of (9) and (10) indicate that the object NP is actually the object of the whole

complex predicate (V+CDC), instead of just that of the first verb. Now, let’s

compare sentences when the CDC is a result phrase. Take tuei xia qu ‘push down’

for example.

A. V+O+Vi+lai/qu

(11) a. tuei ta xia qu

111
 Push he fall go

‘Push him down.’

b. tuei [yige ren] xia qu

push one person fall go

‘Push one person down.’

B. *V+Vi+O+ lai/qu

(12) a. *tuei xia ta qu

Push fall he go

‘Push him down.’

b. *tuei xia [yige ren] qu

push fall one person go

‘Push one person down.’

C. V+ Vi+lai/qu +O

(13) a. *tuei xia qu ta

Push fall go he

‘Push him down.’

b. tuei xia qu [yige ren]

push fall go one person

‘Push one person down.’

112
D. BA+O+V+ Vi+lai/qu

(14) a. ba ta tuei xia qu

BA he push fall go

‘Push him down.’

b. ba [yige ren] tuei xia qu

BA one person push fall go

‘Push one person down.’

This set of examples demonstrates that a resultative VP should be by itself a

smaller non-separable constituent, and so that the possible word orders are more

limited. This is manifested by the ungrammaticality of word order B (12), where the

result VP cannot split. In addition, that yigeren is ok, whereas ta is unacceptable in

word order C, as in (13a)-(13b), seems to show that it requires certain phonological

prominence of an object to be located at the sentence final position. Finally, (13b)

and (14), again, show that the object NP is the object of the whole constituent V+CDC.

Hence, we conclude that V+CDC, no matter the CDC is resultative or directional, is a

single event, like E do RSVCs, and, presumably, is better treated as a complex

predicate.

113
7.2 The Syntactic Structures

Following our previous discussion, there are three possibilities of the structure

of the complex predicate Vi+CDC: right-headed, left-headed or double-headed.

Let’s examine each of the possibilities.

The first possibility assumes VP2 as the head. Hence, V1 would be a modifier

to VP2. In this case, modality, negation and question-formation will be all on the

second VP, just like the cases of Circumstantial SVCs.

The second possibility assumes that V1 is the head of the complex predicate.

Theoretically, this is on a par with Huang’s (2006) claim that Chinese has gone

through a change of head directionality, from right-headed to left-headed, and result in

‘de-causativization’ or ‘ergativization’ of verbs34.

The third possibility adopts Baker and Stewart’s double-headedness hypothesis,

where both verbs are the heads of a single verbal projection.

Now, let’s examine the sentence behaviors. The following examples show that

negation, modals and manner adverbs can be placed in front of V1 and scopes over

VP2. On the contrary, VP2 cannot be negated, nor can a modal be placed

immediately before it. If there is a manner adverb specifically modifying VP2, the

34
The argument is based primarily on the derivation of [V sha ‘kill’] NP to [V shi ‘die] NP in Chinese.

114
sentence will be ok with a resultative VP2, but unacceptable with a directional VP2.

(15) a.wo yinggai tuei ta xia-qu.

I should push him fall-go

‘I should have pushed him down.’

b. ta huei na bi chu-lai.

He will take pen out-come

‘He will take out his pen.’

(16) a.wo bu tuei ta xia-qu.

I not push him fall-go

‘I won’t push him down.’

b. ta bu na bi chu-lai.

He not take pen out-come

‘He won’t take out his pen.’

(17) a.wo man-man-de tuei ta xia-qu.

I slowly push him fall-go

‘I push him slowly down.’

b.ta gankuai na bi chu-lai.

He quickly take pen out-come

115
 ‘He take out the pen quickly.’

(18) a.*wo tuei ta bu xia-qu.

I push him not fall-go

b.*ta na bi bu chu-lai.

He take pen not out-come

(19) a.*wo tuei ta keneng xia-qu.

I push him possible fall-go

b.*ta na bi yinggai chu-lai.

He take pen should out-come

(20) a.wo tuei ta man-man-de xia-qu.

I push him slowly fall-go

‘I push him slowly down.’

b.*ta bu na bi gankuai chu-lai.

He not take pen quickly out-come

Examples (15)-(17) show that VP2 falls into the scope of modality and negation

of V1, and hence V2 cannot be the structural head. Examples (18)-(20), especially

the impossibility of negation of VP2, further illustrate that VP2 cannot be an

independent clause. Thus, the first possibility, right-headed analysis, is out. Finally,

116
the contrast of (20a)-(20b) seems to indicate that a result phrase can be a VP

independent from V1, while a pure directional phrase cannot be separate from the first

verb.

Furthermore, analyzing the V+CDC as a complex predicate seems to be on the

right track. A secondary predicate, like (21), reveals different behaviors with a CDC

on negation (22) and modal (23) placement. Crucially, the secondary predicate hen

congming can not raise to the right of jiao-guo and form a legitimate constituent with

it.

(21) a.ta jiao-quo yige xuesheng hen congming.

He teach-EXP one student very smart

‘He once taught a student who was very smart.’

b.* ta jiao-quo hen congming yige xuesheng

(22) a.* ta bu-ceng jiao-quo yige xuesheng hen congming.

He never teach-EXP one student very smart

b. ta jiao-quo yige xuesheng bu-shi hen congming.

He teach-EXP one student not-be very smart

‘He once taught a student who was not very smart.’

(23) a. ta keneng jiao-quo yige xuesheng hen congming.

117
 He possible teach-EXP one student very smart

‘It is possible that he once taught a student who was very smart.’

b. ta jiao-quo yige xuesheng yinggai hen congming.

He teach-EXP one student should very smart

‘He once taught a student who should be very smart.’

Now, we will assume on the second possibility that Directional SVCs are

complex predicates, and V1 is the structural head. V1 must raise to the light v

position, as it is causative in nature. The result VP, as a semantically complete

constituent, can optionally pied-pipe with V1. Moreover, the object yigeren is the

object of the whole complex predicate, instead of just that of tuei, as evidenced by

tuei xia-qu yigeren. This has an extra benefit of avoiding Huang’s PSC. Besides,

as we have mentioned, the result VP can not be further split into smaller units and

consequently neither of the two elements can be raised out from the constituent. In

this respect, the result VP is better re-analyzed as a verb.

118
(24)

vP

DP v’

v VP

CAUSE
DP V’

yigeren
V VP

tuei V

xia-qu

On the other hand, when VP2 is a directional VP, the correlation of the two

elements within the VP is rather loose. This VP is a real VP, and can be further

broken into smaller units, such as chu and lai. Hence, after na raises to the light verb,

either chu or chu-lai can optionally raise along with it to the higher position. The

movement of lai alone is ungrammatical, since it reverses the original word order

chu-lai.

119
(25)

vP

DP v’

v VP

CAUSE
DP V’

yizhi bi
V VP

na
chu-lai

Now, we have claimed on the left-headed structure for Directional SVCs, and

dispensed the possibility of the right-headed structure. Yet there is still one

possibility which is not ruled out by the previous examples: double-headedness

structure proposed by Baker and Stewart. Adopting their analysis, the sentence

structure of a Directional SVC would be like (26). Actually, (26) is roughly the

same as (24), since we have reanalyzed the result VP as a V.

120
(26)

vP…

v VP

DP V’

yigeren V V

tuei xia-qu

Is there a best choice between (24) and (26)? I believe there is. And the

answer should be (24). A double-headed structure like (26) has long been criticized

as neglecting projection rules and -criterion. In the convention, a head projects to a

maximal projection, and, conversely, a maximal projection is projected from a single

head. Moreover, -criterion regulates that an argument should be assigned one and

only one theta role. Clearly, in (26), yigeren inevitably receives two theta roles from

the two heads. On the contrary, our postulation in (24) raises no such problems, for

a maximal projection, VP, intervenes between the two verbs tuei and xia-qu. There

is still one benefit of our analysis. On the VP analysis, that chu or chu-lai in (25) can

121
optionally pied-pipe with V1 is directly accounted for, while it would be a problem

for (26) to explain why chu can raise out of the head verb chu-lai.

In this section, we examine three possibilities of head directionality in

Directional SVCs. We conclude that either left-headed structure or double-headed

structure can explain the sentence behaviors. However, since left-headed analysis

incurs no theoretical problems that double-headed analysis has, it should outshine as

the optimal choice.

7.3. Summary

In this chapter, we presented two types of Directional SVCs: one with a

resultative VP2 and the other with a pure directional VP2. Even though the two

types are different semantically, they are highly similar in the combination of verb

types (both are Vi +lai/qu) and on the event structure—both denote a single event

with the previous Vt. Syntactically, they are analyzed either a left-headed complex

predicate or a double-headed vP. Finally, we argue that the left-headed structure

prevails for theoretical reasons.

122