A study of the Japanese causative construction

A study of the Japanese causative construction

著者（英） Yasuyuki Nakai

journal or

publication title

Doshisha literature

number 28

page range 97‑123

year 1976‑11‑22

権利（英） English Literary Society of Doshisha University

URL http://doi.org/10.14988/pa.2017.0000016500

A STUDY OF THE JAPANESE CAUSATIVE CONSTRUCTION

Y ASUYUKI NAKAI

I This paper deals with the problem of whether or not it is valid to support an analysis of the Japanese causative construction which is made along the lines of discussion of generative semantics in " Lexical Insertion in a Transformational Grammar without Deep Structure" 1 by James D. McCawley. According to his theoretical framework, "kill ", for instance, must be derived from the following tree structure such as (1): 2

(1) x KILL y = S

CA~S

BECOME

~

S

NOT

~

S

ALI~~

It seems doubtful if, by applying his theoretical framework to the Japanese language, we could explain the Japanese causative construc- tion so neatly as McCawley did in his English example. In Japa- nese, there arise many intricate problems and logical inconsistencies in the claim in which we would derive the LCV (lexicalized calj.sa- tive verb) korosu "kill" and the UCVP (unlexicalized causative verb

98

h ) . aseru "cause to die" from the same single deep structure, prase szn

as we will observe'below in this paper. Hence it seems to be valid that, to explain these facts,

we·

should discard the position in which we would derive korosu from sinaseru.

IL 1 In Japanese there seems to be a semantic constraint to the effect that the surface objects of UCVP's must .be some animate things that ha,7e their own volition.³. We will now consider whether or not this constraint is valid not only for sentences with an embedded intransitive sentence but also for' ones with an embedded transitive sentence.⁴ Consider, for instance, such an example with an intransi- tive verb embedded within that matrix· sentence.

a ?Taroo wa hon 0 itiretu ni narabaseta.

"Taro made books be lined in a row."

b Taroo wa hon 0 itiretu ni nambeta.

"Taro lined books in a row."

c *Taroo wa hon ill itiretu ill narabaseta.

d *Taroo wa hon ni itiretu ni narabeta.

(2) e Tareo wa seito 0 itiretu ni narabaseta.

"Taro made students be lined in a row."

£ Taroo wa seito 0 itiretu ni narabeta.

"Taro lined students in a row."

g Tareo wa seito ni itiretu ni narabaseta.

"Taro made students be lined in a row."

h *Taroo wa seito ni itiretu ni narabeta.

(2a) may be grammatical, but it might be interpreted as having a somewhat unnaturalflavor. While (2b) 'is grammatical and quite naturalas a Japanese' sentence; And (2c, d) are ungrammatical. The particle 0 cannot bereplacerl' by ni'in these examples, generating ungrammatical structures. (Za'-d) seem to have the same single deep

structure as is shown in (3 a):

(3a)

I NP I

I

Taroo

I

NP

hon

~

()

NP

VP I 1

VP I V sas£"

I

-=====---=-

^!

Adv

L~

_ltIretu

111

I V narab

I

I

Aux

I I i

t<1

Aux !

ru

I

It seems that in (3a) the matrixsentehce causes the ungrammaticality of the surface sentence which is the output of the causative trans- formation. Since" books" cannot be forced to perform something.

CZe-h) also have the same single deep structure as

is

shown in (3b) : (3b)

Taro() I \va

s,

I

I

VP

I

I

V

:-:;asE'

I

Aux

td

I

VP ~

I

----=-' ....

1 - - ' 1

Adv V Aux

~,,,,

\ I

ltiretu narab ru

ni J

</1

There is in the matrix sentence in (3b) nothing that causes the un- grammaticality of the whole sentence. It is quite adequate to say that "students" can be forced to perform some action. While (2c, cl) are not acceptable as grammatical sentences, (2g) is acceptable.

The particle replacement transformation does not effect the gram- maticality of (2e, g) ; however, it causes (2h) to be ungrammaticaL5.

In some sentences with embedded intransitive verbs, the surface particle ⁰ may be optionally replaced with ni. For (2e) and (2g) are the same in meaning.⁶ Let us consider another instance such as (4) :

a ? Taroo wa kane 0 atumaraseta.

"Taro made the money collect."

b Taroo wa kane 0 atumeta.

"Taro collected the money."

c * Taroo wa kane ni atumaraseta.

cl *Taroo wa kane m atumeta.

(4) e Taroo wa gunsyuu 0 koodoo ni atU7naraseta.

"Taro made the throng gather in the hall."

f Taroo wa gunsyuu 0 koodoo ni atumeta.

"Taro gathered the throng in the hall."

g Taroo wa gunsyuu ni koodoo ni atumaraseta.

"Taro made the throng gather in the hall."

h *Taroo wa gunsyuu ni koodoo ni atumeta.

The deep structure of those examples presented in (4a)-(4d) corres- ponds to (5a) and that of those examples presented in (4e)-(4h) to (5b):

(5a) NP I

j

I

Taroo \va

(5b) NP I

Taroo

~

wa SI

I

_I

VP

I I

I

~

^S2

^I

kane 0

Jp

k~l

^{V I}

~

_atulna

I

l'

I

'{

NP

I

I I

I

^Aux

^I

I

sa se ta

v;P

AJx ru

I

9

I

I

V Aux

I _sase

I

_ta

I

Vr

I I

V Aux

I I

atuma ru if;

It is quite clear that there is no semantic violation in each constituent sentence embedded in (5a) and (5b). Semantic anomaly is met in those matrix sentences of (3a) and (5a). In other words, we cannot force money, books, etc. - i. e., inanimate things in general- to do something or to perform some action. This is why semantic vio- lation occurs in those matrix sentences of (3a) and (5a). Having observed that (2b), (2e), (2f) and (2g) are grammatical and (4b),

(4e), C4f) and C4g) are also grammatical; that (2h) and C4h) are not grammatical; and that (2a) and (4a) may be grammatical but some- what unnatural, we may claim as follows from these facts. In the causative construction with underlying embedded intransitive sentences, the surface objects must be animate noun phrases if sentences come to have UCVP's in the surface structure. ",Vhile surface objects may be animate or inanimate for corresponding LCV's. If the theory in- troduced in the introductory passage of the present paper is valid, then C2g) and (4g) must also be ungrammatical along with (2h) and (4h). Since (2g) and (4g) that contain corresponding UCVP's are grammatical, this claim is not the case. If we can derive UCVP's and LCV's from the same single deep structure, such a problem concerning the particle replacement transformation does not occur.

Shibatani should have developed his claim to the extent that he could point out the possibility of the particle replacement from 0 to ni in certain environments.⁷

n.

2 Next, let us consider whether the semantic constraint men- tioned in the last paragraph and the possibility of the particle repla- cement can be applied to the causative construction which contains underlying transitive sentences embedded within that matrix sen- tence. For example, we can take up the following sentences presented in (6):

(6)

a ?Taroo wa ame ni Ziroo 0 utaseta.

"Taro let the rain fall on Jiro."

b *Taroo wa ame ni Ziroo ni utaseta.

c Taroo wa Ziroo ni kugi 0 utaseta.

"Taro made liro strike the nail."

d *Taroo wa Ziroo ni kugi ni utaseta.

The deep structure of (6a, b) is represented in (7a) as we see below:

(7a) I NP

I

'I"

^{an)o \Y<.l}

^I

^NP

^r

r-~

S2

I

VP I

I

I V

ame 0

r>-!.I----,-,

_Sdse

I

NP VP

I

I

! I r-->--l-.-,

^--~

ame ga NP V Aux

-y ^~I I I

<p Ziroo () utu ru

,j"

~j

Aux

I

ta

I

And the deep structure of (6c, d) IS (7b):

(7b) NP I

I Taroo

I wa

S1

I I

VP

II---~I~---.---_I

,

NP S,

Zi~ ~

- - - . I NP

I~l

Ziroo ga

-r

^{rp ktgi NP I}

^I (~

VP

I

I

V

utu

I

V Aux

I

^{I I I}

sase ta

Aux I

rll

I

9

Again in (7a) the semantic anomaly in the matrix sentence seems to cause the unnaturalness of the whole sentence as we see in (6a).

Therefore, the above semantic constraint seems to be appropriate in the case of (6). For example, we must notice the fact that, if (6a)

has undergone passivization, the output structure IS grammatical as we see in (S):

(8) { :

Taroo wa Ziroo 0 ame ni utaresaseta.⁸

"Taro made Jiro be struck by the rain."

Taroo wa Ziroo ni ame ni utaresaseta.

"Taro made Jiro be struck by the rain."

The deep structure of (S) is represented in (9);

( 9 )

I

I NP

I Taroo \iVa

s, I

I I

NP S,

VP I I

~ 1-

I

V

Zirou () I ~ sa se

I

NP VP

I

---'-r-' - - . -

Z · I I S",----' , I

'wo ga ³ V Aux

---r-I~-I

_{<p NP VP rare}

I I

ru

~ i---LI--'I~

ame ga NP V Aux

.~

^{1 1}

ZlWO () ulu ru

---r

_'P

^~

_'P

Aux

la

I

Applying relevant transformational rules in order, we can get (Sa).

(Sb) can be explained, if we suppose an optional application of the particle replacement transformation, which repl~ces 0 with ni.⁹ Again the constraint on the possibility of the application of the particle replacement transformation seems to be limited to those sentences that contain surface animate objects. This is obvious, since the fol- lowing examples presented in (10) are semantically anomalous.

?Taroo wa hon 0 ame ni utaresaseta.

"Taro let the book be struck by the rain.

*Taroo wa hon ni ame ni utaresaseta.

Therefore, we may probably claim that those objects of UCVP's must be some animate things.

As we have already observed, the following assumption seems to be a strong evidence for not deriving UCVP's and LCV's from the same single deep structure. The particle replacement which trans- forms ⁰ of surface animate objects (i. e., subjects of embedded sen- tences) into ni is possible for sentences with UCVP's, however, this cannot be said about the surface animate objects of LCV's.

III According to Shibatani's opinion, the Japanese verb korosu

"kill" has SASE in its underlying representation, while the Japanese verb phrase sinaseru "cause to die" has sase in its underlying re- presentation. Hence he is to claim two different types of causation.

SASE obligatorily undergoes predicate raising and lexicalization, and sase does not. And the semantic constraint introduced above does not effect the former but it does effect the latter.lo Let us take up the Japanese" do so" construction: " - - , X mo soo suru". As we will see below, a semantic ambiguity occurs in a sentence with a UCVP, while it does not occur in a sentence with an LCV. This fact, for instance, is explicitly shown in the two examples in (11);

(11)

a Taroo wa musume 0 kuruma ni noraseta, suruto Hana- ko mo soo sita.

"Taro made his daughter get on a car, and Hanako also did so."

b Taroo wa musume 0 kuruma ni noseta, suruto Hanako mo soo sita.

l

"Taro made his daughter get on a car, and Hanako also did so."

According to his grammatical theory, (l1a) is two ways ambiguous, for instance, between "Hanako also made her daughter get on a car" and "Hanako herself got on a car." However, this semantic ambiguity does not occur in the case of (11 b). The former inter- pretation of (l1a) has the implication in which" soo suru" obviously refers to the whole part of the UCVP "noraseru ", while in the latter interpretation the relevant verb phrase "soo suru" refers only to the underlying intransitive verb "noru" embedded in the matrix sentence.

Here is the reason why (Ha) is two ways ambiguous in meaning.

On the other hand, the semantic unambiguity in the case of (11b) seems to result in the fact that the construction " - - , X mo soo suru "

cannot refer to nothing but the output which has already undergone the predicate raising and lexicalization. Shibatani appears to suppose that a Japanese sentence with an LCV does not contain an underlying embedded sentenceY However, this is not correct. We may assume that the base structure of (Ha) and (l1b) are presented in the same base phrase-marker. The very difference between them is the fact that in (l1a) the syntactic amalgamation of "noru+saseru" must be performed at the level of the n~n-lexical transformation, i. e., must pert<lin to the transformational component of the grammar, and that in (l1b) . the different lexical item of "noseru" must be already determined in the level of the lexical insertion transformation for it to appear in the surface structure, since "noseru" has an obvious semantic similarity to "noru+saseru". That is to say, the deep structure of (Ha) corresponds to (12), and that of (l1b) to (13)

respectively.

(12)

, - - - -

I

So 1- ^conj

r---'---- -, I

;\II' VP

N

^{I I I suruto}

T"",,

~~,,,,,T}, /'

NP,

,

VP ^I

1

I

I- rx

sa se 1:.1

I

^I

I I Ir---'--.,---.

111USUm" ga Aclv V ,\ux

--r-~

_{kun~i nor}

^t

_IU

I

I

So 1-

-,---~I

{ rp

,

^-

con]

NP VP suruto

I

Taroo wa

~

musume

l2l':~a Aclv

<jJ kuruma ni ru

*

^'P

NP j _

n

Hanako wa

1 S,

!

Jno I~

I

M

soo sita

I S I'

,,---'---,

NP VP

I

~

Hlmako W'I soo sita

1

t n ( )

NOSETA

'\7i[hat (12) and (13) show us is the fact that In (12) "noru" and

" saseru" must be derived separately into the the deep structure by the lexical insertion transformations, and that in (13) "noru +saseru ".

which is to undergo obligatory amalgamation, must already be replaced by "noseru" before the post.lexical structure is developed. This fact seems to suggest a sentence with an LCV and one with a UCVP do not have the same single deep structure.

IV Shibatani again notices one interesting fact of the Japanese language, showing that the reflexive pronoun zibun "self" may be ambiguous in its reference when it appears in a sentence with a UCVP, and that it cannot refer to any other NP than the subjective one when it appears in a sentence with an LCV. For instance, we can take up the following example (14):

(14)

Watasi wa kodomo ni zibun no isi de boosi 0 kabu·

Taseta.

"I made my child put a hat on with my own will."

"I made my child put a hat on with his own will."

W atasi wa kodomo ni zibun no isi de boosi 0 kabu·

seta.

"I made my child put a hat on with my own will."

In (14b), the ~eflexive pronoun zibun only refers to the subjective NP watasi, but in (14a) two different kinds of interpretation can be given. Here we may refer to the rule of Japanese reflexivization.

In Japanese, the reflexive pronoun cannot refer to any other NP than the subjective one in a sentence. Judging from this fact, what is grammatically meant by the fact that in (14b) zibun cannot refer to any other NP than watasi? Obviously it suggests that the embedded transitive sentence in (14b) which contains the LCV kabuseru is deleted or discarded¹⁴ at the level near the post.lexical structure, and that reflexivization occurs afterwards. The semantic ambiguity in (14a) can be easily explained from the following assumption. If the reflexive pronoun is assumed to refer to watasi in the matrix sen·

tence, reflexivization occurs in the level of the matrix sentence. If, on the other hand, it is assumed to refer to kodomo, the relevant transformation occurs in the level of the embedded S cycle. There- fore, we cannot help recognizing that the deep structures of (14a) and (14b) are different from each other, as we see in the following tree-structures (15) to (17). (15)-(16) are associated deep structures for the two kinds of possible interpretation of (l4a).

(15)

j

NP

I

S, !

I I

VP I

I

I I

watasi wa NP Adv V Aux

(16)

I .

\Vatasl

k

^{o omo}

^~d

^(I

^~

^{watasl no IS1} _sase

I

_ta

I

I NP

I

--k C + refl) de

ni 13 . - - - ' - - - .

I

NP

I I !

kodoffiO ga NP

~~

_{ooosi 0}

S, I

VP i I

VP I

I

I

V Aux

I I

kabu ru

'*

^'P

wa NP I

V 1

Aux

I

^{I I}

kodoffiO

l

ni

NP I

I I

k()dOffiO

I VP

I

sa

I

se _ta

I

I I,

ga NP Adv V Aux

bLo~L)u

kodoffiO no isi de

i

C +refl)

(17) s,

I ! I

NP VP

I

I

r-~--~I I,---.I---~---.\----·---.---.I

watasi ~P Adv S, Au"

k()do!TIO ()

~

\~. C

+

reft) ici de

III

I~

NP

kodomo ga

--r--

<P

NP

,-

hnosi ()

.L,

sa se

(17) is the associated deep structure of (I4b). Thus, as we have now observed, those deep structures corresponding to (I4a) and (I4b) are different from each other.

V Next we must notice that there is a semantic difference be- tween a sentence containing an LCV and a sentence containing a UCVP. For example, we can consider (18):

(18)

a Keisatu wa syoonen no hikoo no torisimari no· waku o yurumaseta.

"The police made some ones loose the scope of the control of boys' mischievous action."

b Keisatu wa syoonen no hikoo no torisimari no waku o yurumeta.

"The police loosed the scope of the control of boys' mischievous action."

It is quite obvious that the sefIlantic content of (I8a) and that of (I8b) are not the same. In (I8b), it is the police itself that directly controls boys' mischievous action, while in (I8a) it is some body of

persons rather than the police itself that performs an actual control on boys' mischievous action. That is, the difference between (I8a) and (I8b) seems to be the difference between the directness and the indirectness of the action that is implied in sentences.¹⁶ If it is the case that (I8a) and (I8b) can be derived from the same single deep structure, such a semantic difference as is shown above would not occur. One of the most valid ways of solution is for us to derive them respectively from different deep structures. Hence we will assume that the deep structure of (I8a) corresponds to (I9a). Now let us observe how we can derive the surface structure (I8a) from (I9a).

(l9a) _SI

r--·--·---L---"

NP VP

I I

I

keisatu

I I

Wi! NP S2 ^I I

V Anx

~ I

dareka ⁰ I

I ^I

NP

dareka ga

~

I ,

NP S

I

VP

I

sa:se La

I Aux

,

h ^~13

wakn ()

r---'--~I

yurume ^.fU

NP VP

~ Ir---il~---;I

dareka ga NP V Anx

~~.1.

_(') syoonen no lonSlnla

I

. hikoo ^{(J 17}

First, relativization must be applied to the S3 cycle of the structure of (19 a), simultaneously preposing the whole VP-node just in front of the NP-node that has dominated the embedded sentence. Next,

identical NP deletion deletes "dareka ga" in this S cycle. Thus, (19b) can be generated as we see below:

(19b) NP I I I

keisatu wa 1 NP

dareka ()

~

1 I

NP VP

I VP

I

I V 1

si:t~e

r----1I---,

r-_L_,

~

y~

__

^~

y

^,\ux

rp syounen nu hd,,,() yurume )u no torisimari

no waku 0

*'

<p

Aux I

ta

Furthermore, to the structure of (19b) identical NP deletion is again applied to dareka in S2. And subjectivalization simultaneously applies to the NP-node in the same cycle; consequently, the transitive verb yurumeru" is replaced by the corresponding intransitive verb "yu- rumu ". The output structure of such a set of transformations as are now mentioned is (19c), which is the same as (18a):

(19c)

=

(18a) SI

r!---~---·i I

.

NP VP

1_

-.11_-, 1

I ! I I

keisatu wa NP 52 V

~ /~ I

~~/"

j

^V~sase

1

^{NP V Aux}

\)~\ \

syoonen no hikoo yurunl r~

110 torisimari {.

Aux

ta

I

.)0 \\'aku g:1 1)

And the deep structure of (18b) IS represented In (20a) : (20a)19

NP I

I

NP I

I

\Vaku I

I I

keisatu ga

~

sO Si

I

;:-';P I

I

j I S2 ⁰

J I

VP

I

I

~

syoonen no hikoo NP

()

I VP

I

I I

V Aux

I I

yururne ta

I I

V Aux

.I. I

tonslma ru

v

6

To this structure those relevant transformations to which we have referred in generating (19b) from (19a) must be applied here, de- riving (20b):

(20b)

,

NP i

I

keisatu wa

syoonen no li 00 no torisin1ari no \vaku o

I VP

V

I

yurume

I

I Aux

ta

I

Judging from the facts mentioned above, we cannot deny that the claim of deriving LCV's from corresponding UCVP's involves many

an illogical aspect.

VI Next, let us take up the case of causative sentences in which adverbials occur. And also in this case, there arises a semantic dif- ference between a sentence with a UCVP and a sentence with an LCV. We must argue this point. For example:

(21)

(22)

Taroo wa kuruma 0 kyuuni tOfnaraseta.

"Taro suddenly caused the car to stop."

"Taro caused the car to stop suddenly."

Taroo wa kuruma 0 kyuuni tameta.

"Taro suddenly stopped the car."

Taroo wa kuruma ⁰ murini magaraseta.

"Taro rashly caused the car to curve."

"Taro caused the car to curve rashly."

?Taroo wa kuruma 0 murini mageta.

"Taro rashly tun;'ed the handle of the car."

a Akiko wa musumetati ni kawarugawaru hon 0 mi- saseta.

" Akiko, in turn, made her daughters see the book."

(23) " Akiko made her daughters see the book in turn."

b Akiko wa musumetati ni kawarugawaru hon 0 mi- seta.

" Akiko showed in turn the book to her daughters."

In (21a), a semantic ambiguity occurs by the possibility of the scope of the modification by the adverb kyuuni. The adverb in question may modify "saseru" in the matrix sentence or "tomaru" in the em- bedded sentence. In (21b) kyuuni only modifies the LCV tomeru; as the result, no semantic ambiguity occurs in it. The same line of argument can be made on those examples of (22) and (23).

we can consider those examples represented in (24):

Then

(24)

a Hanako wa musuko 0 katteni oba no ie ni nokorasetil.

"Hanako let her son stay with his own will at her aunt's home."

"Hanako made her son stay with her own will at her aunt's home."·

b Hanako wa musuko ni katteni oba no ie ni nokoraseta.

(two ways ambiguous like a)

c Hanako wa musuko 0 katteni oba no ieni nokosita.

"Hanako made her son stay with her own will at her aunt's home."

d *Hanako wa musuko ni katteni oba no ie ni nokosita.

In (24a, b), katteni, in general, refers to the son's will, however, it can also refer· to Hanako's wil1. Yet this is impossible in (24.~), for the adverb katteni cannot refer to any other person's will than Hanako's. Furthermore, we mllst pay attention to the fact that (24b) is a possible Japanese sentence, but that (24d) is not. The particle replacement cannot be applied to the structure of (24c) , since the application of this transformational rule explicitly results in an un- grammatical structure. This fact has often been referred to pre- viously. In (25),

(25) { :

Asada sensei wa sono seito 0 isoide ie ni kaeraseta.

"Professor Asada mada the student go home hastily."

" Professor Asada promptly made the student go home."

Asada sensel wa sono seito 0 isoide ie ni kaesita.

" Professor Asada promptly made the student go home."

the same claim and discussion made for those examples in (21)-(23) can be applied. The following example is much more difficult than previous examples presented above.

Taroo wa Saburoo to Ziroo no seki 0 sorezore no isi de kawaraseta.

"Taro made Saburo and Jiro exchange their seats

with their respective will."

Taroo wa Saburoo to Ziroo no seki 0 sorezore no isi de kaeta.

"Taro exchanged Saburo's seat and Jiro's with their respective will."

(26a) and (26b) seem to be different III meaning. As for (26a), the . most natural interpretation is that Taro himself did not make an action of exchanging the two seats, but that Saburo and Jiro did ex- change their seats with their own will and Taro only directed them to do so. While, in (26b) , this interpretation is not possible, since the most natural one is that Taro himself exchanged their seats ac- cording to their own will. This becomes even more explicit in the following example which shows the paraphrases of (26a, b), one of which is ungramrhatical for the reason mentioned in footnote 13.²⁰

(27)

Taroo wa Saburoo to Ziroo ni sorezore no isi de seki o kawaraseta.

*Taroo wa Saburoo to Ziroo m sorezore no ISI de seki 0 kaeta.

VII Finally, we will argue the relevant Japanese construction, considering some examples in which we can validate our claim that UCVP's and LCV's cannot be introduced from the same single deep structure. One of the most striking examples that can be observed is the pair between korosu "kill" and sinaseru "cause to die". Let us examine the following pair of sentences:

Watasi 0 korosite kudasai.

"Kill me."

Watasi 0 sinasete kudasai.

"Let me die."

Every Japanese speaker will not deny the fact that Ca) and (b) III

(28) are not the same in meaning. For eample, in (28a) the speaker of this sentence will probably utter like this even if he really does not want to be killed by the person directed by him. (28b) is not a direction, but it is rather a sentence implying the speaker's hope or expectation. This is the intrinsic difference between (28a) and (28b).

Therefore, this distinction becomes much more clear in the following pair:

Syu yo, kono syukuhuku no naka de watasi 0 szna- setamae.

"Lord, let me die in this bliss."

?Syu yo, kono syukuhuku no naka· de watasi 0 koro- sitamae.

"Lord, kill me III this bliss."

Furthermore, we may consider the following example (30):

Taroo wa kawaigatte ita aiken 0 korosite-simatta.

"Taro has killed his pet dog he cherished."

Taroo wa kawaigatte ita aiken 0 sinasete-simatta_

"Taro has caused his pet dog to die, which he cherished."

(30a) is quite preposterous, excepting our making allowance for some particular situation in which some unrecoverable disease suddenly captured the dog and Taro decided to kill it for he could not endure seeing the dog being afflicted and suffering in his presence. While the most normal interpretation for (30b) seems to be one that some action accompanying Taro's careless act became an indirect cause for the dog's death. Judging from the examples presented above, we may claim that koroSll and sinaseru cannot be derived into the surface form from the same single deep structure. Deliberating such kinds of fact on the Japanese causative construction as we have referred to above in this paper, we cannot help upholding the theory in which

we should derive UCVP's and LCV's from different deep structures respectively.

VIII SUMMARY

This paper is an attempt for not deriving LCV's of the Japanese lan·

guage from their corresponding UCVP's. For this purpose we have offered many examples and thereafter submitted some problematic aspects against the theory proposed in the introductory passage of this paper. Concluding this paper, we claim as follows: Japanese LCV's cannot be derived from corresponding UCVP's transformationally (of course, in the level of the transformational component of the gram- ma!"). They must have different underlying sources and different derivations from different deep structures. For the validation of the theory, we have pointed out the following three kinds of significance concerning the construction in question.

( I) Sentences with UCVP's and ones with LCV's have some different semantic resrictions concerning, for instance, the feature specification of [±Animate], etc.

( II ) The possible occurrence of some adverbials in sentences with UCVP's is not allowed in sentences with LCV's.

(Ill) Hence, sentences with UCVP's and ones with LCV's are semantically different.

Notes

1. James D. McCawley, "Lexical Insertion in a Transformational Grammar without Deep Structure," Grammar and l\1eaning (Tokyo: Taishukan Pub- lishing Company, cl973).

2. ibid., pp. 157ff.

3. M. Shibatani, "Three Reasons for not Deriving' Kill' from 'Cause to

die' in Japanese," Syntax and Semantics, Vol. I, ed. J. P. Kimball (Tokyo:

Taishukan Publishing Company, 1972), p. 128.

4. It is a generally recognized fact that the causative construction has an underlying embedded structure. See S.-Y. Kuroda, Generative Grammatical Studies in the Japanese Language (Unpublished Ph.D. dissertation, lVI. I.

T., 1965), K. Inoue, A Study of Japanese Syntax (The Hague: Mouton, 1965), T. Ishiguro, "A Study of Japanese Verb Phrase Embedding Construc- tions," Doshisha Literature No. 25 (Kyoto: English Literary Society of Doshisha University, 1969), etc.

5. Obviously, so-called LCV's function as transitive verbs in sentences. Hence they cannot take NP's accompanied with ni as their direct objects, since the particle which marks NP's as direct objects of transitive verbs is o. For this reason the replacement under discussion is not possible in this case.

6. According to Kuroda's claim, o-causativization and ni-causativization are derived from different deep structures. However, this claim is probably untenable. If, as Chomsky and others say, the maintenance that the intrinsic part of semantic representation must be given in the deep structure is the- oretically valid, o-causativization and ni-causativization must have different semantic interpretation respectively in the surface structure. However, this is not the case, as we have observed in (2e, g) and (4e, g). This fact seems to support the analysis of not deriving o-causativization and ni·causa- tivization from different deep structures. Second, there is a strong evidence for upholding Inoue's analysis in which she tries to explain ni is a surface variant of o. If we uphold the theory by Kuroda, we are only to complicate the analysis of the deep structure of the Japanese causativization and also we must contrive some, otherwise unnecessary, gratuitous transformations such as Constituent Subject Extraction and the deletion of the objects of matrix sentences. On the other hand, if the theory with the transformation- al apparatus of the particle replacement is valid, the grammar of the Japa- nese causativization becomes even more explicit in its descriptive and ex- planatory adequacy. For the possibility of the particle replacement seems to be restricted to those sentences with UCVP's whose objects are (

+

Animate).

Therefore we may formalize the particle replacement in question as follows:

NP-<?->NP-ni / NP h~s a feature specification

I

X--UCVP Y

(optIOnal) (

+

Ammate) I

See Kuroda, op. cit., pp. 175-177, and Inoue, op. cit., p.52.

7. Furthermore, the grammaticality of (2g) and (4g) shows a fact of the Japanese causative construction. Since in the case in which the surface ob- jects of UCVP's (i. e., the underlying subjects of embedded intransitive sen- tences) are inanimate, the particle replacement from 0 to ni is not possible, the particle replacement transformation can be resricted to those sentences which have animate noun phrases as subjects of underlying intransitive (and probably, transitive) sentences embedded within the matrix clauses. See footnote 6.

8. This example I owe to T. Ishiguro. In this case the particle of the surface object, ^0, can be replaced with ni, since the surface object has a feature (+ Animate). See T. Ishiguro, op. cit., p. 80.

9. See K. Inoue, op. cit., p. 50. Her explanation is that Ci) can be derived from (ii) through the deletion of the unspecified object.

(i) Taroo wa Ziroo ni ame ni utaresaseta.

(ii) Taroo wa Ziroo ni nanika 0 ame ni utaresaseta.

Yet this way of explanation cannot explain the semantic identity between (8a) and (8b), since (i) (=8b) is semantically ambiguous in two ways.

One of the possible interpretations is that (i) is the output of the deletion of the unspecified object "something" nanika 0 in (ii). The other inter- pretation is that it is the output of the application of the particle replace- ment to the structure of (8a).

10. Indeed, this constraint, in general, seems to be valid in many cases. How- ever, we can present some possible exception in the korosu-sinaseru distinc- tion: "korosu" and "sinaseru" seem to necessitate animate objects. Of course, it is significant to point out the semantic difference between UCVP's and LCV's. However, much more important is to find out whether we can formalize the syntactic difference between UCVP's and LCV's which can validate the theory for not deriving LCV's from UCVP's. As for the details, see Shibatani, op. cit., p. 129.

11. M. Shibatani, op. cit., p. 13l.

12. Since" soo suru" is an expression essentially anaphoric in its sense, the particle "wa" must be replaced by "mo" obligatorily.

13. In (13), S3 is to be deleted when" noru+saseta" is replaced by "noseta".

Adverb raising applies to "kuruma ni "; therefore, it is not deleted at this level. As for the details concerning the deletion of an embedded S-node, see J. R. Ross, "A Proposed Rule of Tree-Pruning," }110dern Studies in

English, edited by D. A. Reibel and S. A. Schane (Englewood Cliffs, N. J. : Prentice-Hall Inc., 1969), p_ 299.

14. Here I said "deleted or discarded ", since there is no definitive reason for us to determine whether the elements of an embedded sentence are deleted in the transformational component of the grammar or have already been discarded in the obliteration fulfilled at the level of lexical insertion. In each case, there is no doubt for the assumption that reflexivization is applied afterwards.

Concerning this problem, a serious discussion raises itself. According to Grinder's paper, the notion of the transformational cycle is doubted. How- ever, our introduction of the notion of the transformational cycle obviously enables us to explain the semantic ambiguity in (14a). The notional neces- sity of the cycle is supported from the relation between causativization and reflexivization. If we do not uphold the theory with the cycle, how can we explain the semantic ambiguity in (14a)? Therefore, the notion of the cycle, as G. Lakoff claims, seems to be motivated from the viewpoints of syntax and semantics. J. Grinder, "On the Cycle in Syntax," Syntax and Semantics, Vol. 1. pp. 81-111, and G. Lakoff, "Discussion," Syntax and Semantics, Vol. I, pp. 113-115.

S. Kuno gives a similar discussion on the relation between the Japanese reflexivization and the Japanese causativization. S. Kuno, Nihon-Bunpoo Kenkyu (A Study of Japanese Grammar) (Tokyo: Taishukan Publishing Company, 1973), pp. 192-194.

15. As for the occurrence of the particle ni, confer Inoue, op. cit., p. 49.

16. Chomsky, "Deep Structure, Surface Structure and Semantic Interpreta- tion," Studies on Semantics in Generative Grammar (The Hague: Mouton, 1972), p.72, fn 7, and G. Bedell, "The Arguments About Deep Structure,"

Language, VoL 50, No. 3 (The Linguistic Society of America: The Waverly Press Inc., September 1974), p. 425, fn 6.

17. In a strict sense, this P-marker is not the deep structure of (18a), since

"syoonen no hikoo" is an output structure of nominalization. The structure that underlies this structure seems to be "syoonen ga hikoo 0 suru."

Furthermore from the correct assumption that an LCV is derived through obligatory predicate-raising and lexicalization from, for instance,

"yurumu+saseru ", we must recognize the following fact. The grammatical function of "causation" through which yurumeru is derived in the lexical

insertion level from its semantically similar structure "yurumu +saseru "

and that through which yurmnaseru is derived in the transformational com- ponent from its underlying structure" yurumu+ saseru " are quite different.

These two kinds of causation are of different nature in their grammatical relation. Obviously enough the existence of yurltmerzt in the deep structure level suggests the fact that the verb yurltmerlt must already be generated in the level of the lexical insertion.

18. In Japanese there seems to be a transformational rule which deletes an unspecified NP. For example, we can consider the following sentences:

( i) Taroo wa isya 0 yobi ni yatta_

"Taro asked someone for a doctor_"

(ii) Watasi wa hon 0 ageta.

"I gave a book (to someone)."

(iii) vVatasi wa Taroo ni ageta.

"I gave something to Taro."

In these cases, everyone can discern the existence of some unspecified NP's, for instance, dareka ni in (i), (ii) and nanika 0 in Ciii). The following example, however, is ungrammatical.

(iv) *sore 0 sita.

There is no doubt for the assumption that unspecified NP's operating as subjects of sentences cannot be deleted. But this problem is out of the domain of this paper; therefore, we will not go into the details of this pro- blem here.

19. For the ease of exposition, I presented this phrase-marker in a form which has already been simplified.

20. The ungrammaticality of (27b) seems to result, directly, in the fact that there is a surface structure constraint to the effect that an LCV cannot take both a dative object and an accusative object.

SELECTED BIBLIOGRAPHY

Chomsky, N. "Deep Structure, Surface Structure and Semantic Interpreta- tion," Studies on Semantics in Generative Grammar. The Hague:

Mouton, 1972.

Bedell, G. "The Arguments About Deep Structure," Language, Vo!. 50, No.

3. Published by The Linguistic Society of America. The Waverly Press Inc., September 1974.

Grinder, J. "On the Cycle in Syntax," Syntax and Semantics, Vo!. 1. Edited by J. P. Kimball. Tokyo: Taishukan Publishing Company, 1972.

Inoue, K. A Study of Japanese Syntax. The Hague: Mouton, 1965.

Ishiguro, T. "A Study of Japanese Verb Phrase Embedding Constructions,"

Doshisha Literature, No. 25. Kyoto: English Literary Society of Doshisha University, 1969.

Kuno, S. Nihon-Bunpoo Kenkyu CA Study of Japanese Grammar). Tokyo:

Taishukan Publishing Company, 1973.

Kuroda, S.-Y. Generati-ve Grammatical Studies in the Japanese Language_

Unpublished Ph.D. dissertation, M. I. T. 1965.

Lakoff, G. "Discussion," Syntax and Semantics, Vo!. I. Edited by J. P.

Kimball. Tokyo: Taishukan Publishing Company, 1972.

McCawley, J. D. "Lexical Insertion in a Transformatinal Grammar without Deep Structure," Grammar and 111eaning: Tokyo: Taishukan Pub- lishing Company, c1973.

Ross, J. R. "A Proposed Rule of Tree-Pruning," 1110derll Studies in Englsh:

Readings in Transformational Grammar. Edited by D. A. Reibel and S. A. Schane. Englewood Cliffs, N. J.: Prentice-Hall, Inc., 1969.

Shibatani, M. "Three Reasons for not Deriving 'Kill' from 'Cause to die' in Japanese," Syntax and Semantics, Vo!. I. Edited by J. P. KimbalL Tokyo: Taishukan Publishing Company, 1972.

(January, 1975)