‘Neg-sensitive’ Elements, Neg-c-command and Scrambling in Japanese
Kiyoko Kataoka
NihonUniversity, College of Economics
1. Goal
Most languages have some expressions which must occur with sentential negation. XP-sika and rokuna-N are instances of such elements in Japanese.
(1)a.Taro-wa manga-sika yoma-nai. / *yomu (koto)
Taro-TOP comics-all:but read-Neg / *read (Comp)
‘Taro does not read any kind of book but comics.’
b.Taro-sika manga-o yoma-nai /* yomu (koto)
Taro-all:but comics-ACC read-Neg / *read (Comp)
‘Nobody but Taro reads comics.’
(2)a.Saikin rokuna-sakka-ga syoo-o {tora-nai / *toru}.
recently good-writer-NOM award-ACC get-Neg / get
‘Recently, no good writers have got an award.’
b.Taro-wa itumo rokuna-koto-o {si-nai / *suru}.
Taro-TOP always good-thing-ACC do-Neg / do
‘Taro always does damn things.’
In this paper, I will refer to those ‘Neg(ation)-sensitive’ elements as NPIs for ease of exposition.[1] In general, the condition in (3) is assumed, and this assumption is widely accepted in the Japanese literature, too (Kato 1994, 2002, Kuno 1995, and others).
(3)An NPI must be c-commanded by Neg at LF. (Klima 1964)
The goal of this paper is to show, contrary to this general assumption, that not all NPIs in Japanese obey the condition in (3). More specifically, I will argue for (4) and (5) below.
(4)Rokuna-N must be c-commanded by Neg at LF.
(5)XP-sika must c-command Neg at LF.
I first argue that rokuna-N is, but XP-sika is NOT, subject to (3). For that purpose, we need to examine whether those items can occur in a position outside the c-command domain of Neg at LF. Though it appears that every NP in Japanese can be in the c-command domain of Neg at LF as seen in (6) below, I will demonstrate that some element in a certain configuration cannot be in the c-command domain of Neg at LF, and I will show how to identify such element. Making use of the construction in question, I will first argue for (4), and then propose (5), on the basis of scopeinteractions between Neg and QP in the sentences containing XP-sika and another QP. I will then discuss some consequences.
2. Scrambling construction, A-scrambled object and Neg
[HH1]It has been observed that the scope relation between Neg and QP is ambiguous in Japanese (Kuno 1980),[2] suggesting, given the general assumption that the scope of α is its c-command domain at LF, thatevery NP α can be in the c-command domain of Neg where α and the Neg are clause-mates.
(6)a.[goninizyoo-no seito]QP-ga so-no hon-o yoma-na-katta (koto)
five:or:more-GEN student-NOM that-GEN book-ACC read-Neg-Past (Comp)
‘Five or more students did not read that book.’ QPNeg, NegQP
b.[gosatuizyoo-no hon]QP-o so-no seito-gayoma-na-katta (koto)
five:or:more-GEN book-ACC that-GEN student-NOM read-Neg-Past (Comp)
‘Five or more books, that student did not read.’ QPNeg, NegQP
A close examination of sentences of the O(bject)-S(ubject) V order, however, reveals that this is not quite correct.
2.1. Scrambling construction and its structural ambiguity
While it has beenassumed since the mid 1980s that the subject asymmetrically c-commands the object in a non-scrambling sentence, the structural relation between the subject and the object is said to be ambiguous in the scrambling construction, and each case reveals its own properties as illustrated below (Saito 1992, Ueyama 1998).[3]
(7)a.A-scrambling: A- properties (e.g., the absence of WCO effects)
PF : NP-ACC/DAT NP-NOM V
LF : [ NP-ACC/DAT [ NP-NOM V ] ]
b.A’-scrambling: A’- properties (e.g., reconstruction effects)
PF : NP-ACC/DAT NP-NOM V
LF : [ NP-NOM [ NP-ACC/DAT V ] ]
2.2. A-scrambling or A’-scrambling?
I will show that the object in an A-scrambling case (henceforth A-scrambled object) is outside the c-command domain of Neg at LF. Since it is necessary to know, for that purpose, whether the relevant object is an A-scrambled object or not, I will first introduce two tests in order to identify an A-scrambled object.[4]
2.2.1. Test 1: bound variable construal (BVA)
The availability of bound variable construal (BVA) has been regarded as a test to determine whether a given syntactic position is an A-position or an A’-position, and the NP in question is said to be in an A-position if the binding is possible from there. The same test has been applied to the scrambled object in Japanese; if the scrambled NP can serve as a ‘binder’, it is considered as an A-scrambled NP (Ueyama 1998, Yoshimura 1992).[5] See (8), where the object QP is an A-scrambled object when the BVA obtains.[6]
(8)[Itutuizyoo-no ginkoo]i-ni sokoi-no torihikisaki-ga Nissan-o
five:or:more-GEN bank-DAT it-GEN client-NOM Nissan-ACC
suisensi-ta.
recommend-Past
‘To [each of five or more banks]i, itsi client recommended Nissan.’
2.2.2. Test 2: resumption
Since it has been observed that ‘resumption’ is allowed in A-scrambling but not in A’-scrambling (Hayashishita 1997, Hoji & Ueyama 1998, and Hoji 2003), its use is another way to force the scrambled NP to be an A-scrambled NP. As in (9a), ‘resumption’ (soko) is allowed when the object NP succeeds in binding, which shows that it is an A-scrambled object. However, it is not allowed in (9b) when the BVA is forced through reconstruction; i.e., it is not possible when the object is an A’-scrambled object.[7]
(9)a.[Itutuizyoo-no ginkoo]i-ni sokoi-no torihikisaki-ga soko-ni
five:or:more-GEN bank-DAT it-GEN client-NOM it-DAT
Nissan-o suisensi-ta.
Nissan-ACC recommend-Past
‘To [each of five or more banks]i, itsi client recommended Nissan to it.’
b.*Sokoi-no torihikisaki-ni [itutuizyoo-no ginkoo]i-ga soko-ni
it-GEN client-DAT five:or:more-GEN bank-NOM it-DAT
Nissan-o suisensi-ta.
Nissan-ACC recommend-Past
‘To itsi client, [each of five or more banks]i recommended Nissan to it.’[KK2]
2.3. A-scrambled objects and Neg
I will now argue, by making use of the two methods above, that the A-scrambled object is outside the c-command domain of Neg at LF.
2.3.1. BVA
The QP in (10), which is an A-scrambled object, cannot be in the scope of Neg, while the QP in (11), which can be an A’-scrambled object, can be in the scope of Neg. (See footnote8.)
(10)[Itutuizyoo-no ginkoo]i-ni sokoi-no torihikisaki-ga Nissan-o
five:or:more-GEN bank-DAT it-GEN client-NOM Nissan-ACC
suisensi-na-katta.
recommend-Neg-Past
‘To [each of five or more banks]i, itsi client did not recommend Nissan.’ with BVA obtained,ok5-or-more>Neg, *Neg>5-or-more
(11)[Itutuizyoo-no ginkoo]-ni Toyota-ga Nissan-o suisensi-na-katta.
five:or:more-GEN bank-DAT Toyota-NOM Nissan-ACC recommend-Neg-Past
‘To five or more banks, Toyota did not recommend Nissan.’
5-or-moreNeg, Neg5-or-more
2.3.2. Resumption
The QP in (12), which is forced to be an A-scrambled object by ‘resumption’, cannot be in the scope of Neg. Compare this again with (11) above, whose object can be in the scope of Neg. (See footnote8.)
(12)[Itutuizyoo-no ginkoo]-ni Toyota-ga soko-ni Nissan-o
five:or:more-GEN bank-DAT Toyota-NOM it-DAT Nissan-ACC
suisensi-na-katta.
recommend-Neg-Past
‘To five or more banks, Toyota did not recommend Nissan to it.’
5-or-moreNeg, *Neg5-or-more
Given the assumption that the scope of α is its c-command domain at LF, the observation above shows that the A-scrambled object cannot be in the c-command domain of Neg at LF. Thus it is reasonable to assume that the A-scrambled object is outside the c-command domain of Neg before QR. The structures before QR of the A/A’-scrambling cases should be as illustrated in (13). Notice here that the general condition on movement in (14) should be assumed.
(13)Structures before QR
a.A-scrambling: [ NP-ACC/DAT [NegP [VPNP-NOMV] [Neg -nai]]]
b.A’-scrambling: [NegP [VP NP-NOM [ NP-ACC/DAT V]] [Neg -nai]]
(14)Movement cannot be downward.
3. XP-sika, rokuna-N as A-scrambled objects
Now I will turn to XP-sika and rokuna-N, and show that XP-sika can, but rokuna-N cannot, occur as an A-scrambled object.
3.1. A-scrambling case with BVA
With the BVA interpretation, the scrambled object is forced to be an A-scrambled object, as seen in 2.2.1. See(15) and(16a). XP-sika can be a ‘binder’ in this position, while rokuna-N cannot. (Cf. (16b)).[8][KK3]
(15)Mosi [Tokyoginkoo-ni-sika]i sokoi-no torihikisaki-ga
if Bank of Tokyo-DAT-all:but it-GEN client-NOM
syootaizyoo-o okura-na-katta-ra, ....
invitation-ACC send-Neg-Past-if
If, to [any company except Bank of Tokyo]i, itsi client did not send an invitation, ....’ ok BVA(Tokyoginkoo-ni-sika, soko)
(16)a.*Mosi [rokuna-ginkoo]i-ni sokoi-no-torihikisaki-ga
if good-company-DAT it-GEN client-NOM
syootaizyoo-o okura-na-katta-ra, ....
invitation-ACC send-Neg-Past-if
*BVA(rokuna-ginkoo, soko)
b.Mosi [rokuna-ginkoo]i-ga sokoi-no-torihikisaki-ni
if good-company-NOM it-GEN client-DAT
syootaizyoo-o okura-na-katta-ra, ....
invitation-ACC send-Neg-Past-if
okBVA(rokuna-ginkoo, soko)
3.2. A-scrambling case with ‘resumption’
‘Resumption’ also forces the scrambled object to be an A-scrambled object. As seen below, XP-sika can, but rokuna-N cannot, occur with resumption.
(17)Mosi Tokyoginkoo-ni-sika Mitubisi-ga soko-ni syootaizyoo-o
if Bank of Tokyo-DAT-all:but Mitsubishi-NOM it-DAT invitation-ACC
okura-na-katta-ra, ....
send-Neg-Past-if
‘If, to any company except Bank of Tokyo, Mitsubishi did not send there an invitation, ....’
(18)*Mosi rokuna-ginkoo-ni Mitubisi-ga soko-ni syootaizyoo-o
if good-company -DAT Mitsubishi-NOM it-DAT invitation-ACC
okura-na-katta-ra, ....
send-Neg-Past-if
The observations above show that XP-sika can, but rokuna-N cannot, occur as an A-scrambled object. Given (13) and (14), I conclude that rokuna-N must be c-commanded by Neg at LF, but that XP-sika can be outside the c-command domain of Neg at LF. What then is the necessary condition on XP-sika? I will propose that XP-sika MUST C-COMMAND Neg at LF.
4. Proposal
I will argue that XP-sika must c-command Neg at LF, as stated in (5) above.
4.1. Scope interaction between XP-sika and QP
The argument for the proposal comes from the scopeinteraction between XP-sika and QP. As noted above with the example in (6), the scoperelation between Neg and QP is generally observed to be ambiguous. This scopeambiguity can be accounted for by assuming, as in (19), that there are two distinct landing sites for QR, an instance of an adjunction to a maximal projection (Hasegawa 1991, section 1).
(19)a.PF : ... QP1 ... V-nai
b.LF1 : [ QP1 [ [VP ... t1 ... ] [Neg -nai] ] ] → QP1 > Neg
c.LF2 : [ [VP QP1 [VP ... t1 ... ] ] [Neg -nai] ] → Neg > QP1
However, when XP-sika and a QP are in the same clause, the ambiguity disappears. As in (20), where the XP-sika is the subject of the sentence, the QP, which is c-commanded by XP-sika before QR, cannot take wide scope over Neg. On the other hand, as shown in (21), a QP which c-commands XP-sika before QR cannot be in the scope of Neg.
(20)Yamada-sensei-sika [sanninizyoo-no gakusei]QP1-o Mitubisi-ni
Yamada-professor-all:but three:or:more-GEN student-ACC Mitsubishi-DAT
syookaisi-nai (koto)
introduce-Neg (Comp)
‘All professors but Prof. Yamada do not introduce three or more students to Mitsubishi.’ (i) * QP1>Neg, (ii) ok NegQP1
(21)[sanninizyoo-no sensei]QP2-ga Taro-o Mitubisi-ni-sika
three:or:more-GEN student-NOM Taro-ACC Mitsubishi-DAT-all:but
syookaisi-nai (koto)
introduce-Neg (Comp)
‘Three or more professors do not introduce Taro to any company but Mitsubishi.’ (i) ok QP2 Neg, (ii) * Neg QP2
4.2. Analysis
The unambiguity suggests that, ifc-commanded by XP-sika before QR, a QP cannot raise to a position outside the c-command domain of Neg at LF, while it cannot be in a position c-commanded by Neg at LF, if it c-commands XP-sika before QR.
Assuming that XP-sika undergoes QR, and that the Scope Rigidity Principle (henceforth the SRP) holds in Japanese, I propose that XP-sika must be in NegP-Spec at LF.
(22)Scope Rigidity Principle (SRP):The c-command relation between maximal projections XP1 and XP2 cannot be altered through LF movement. (Huang 1982, Hoji 1985)
(23)XP-sika must be in NegP-Spec at LF. [9],[10]
Given that the SRP prohibits the c-command relation between QPs to be altered after QR, a QP, if it is c-commanded by XP-sika before QR, cannot raise beyond the XP-sika, and therefore must stay inside the c-command domain of Neg at LF, since XP-sika must be in NegP-Spec at LF after QR. (In what follows, ‘ok’ means that the condition/principle is satisfied.)
(24)XP-sika1 QP2 V-nai,where XP-sika1 c-commands QP2 before QR
a.LF1: [[NegP XP-sika1 [Neg' [VP QP2 [VP t1 [V' t2 V]]] [Neg -nai]]]]
ok(23) forXP-sika, ok(22) SRP
b.*LF2: [IP [IP QP2 [NegP XP-sika1 [Neg' [VP t1 [V' t2 V]] [Neg -nai]]]]]
ok(23) forXP-sika, * (22) SRP
c.*LF3: [IP XP-sika1 [IP QP2 [IP [NegP [Neg' [VP t1 [V't2 V]] [Neg -nai]]]]]]
* (23) forXP-sika, ok(22) SRP
If the QP2 raised to the position outside the NegP, it would violate the SRP as in (24b), or the condition on XP-sika would be unsatisfied as in (24c). Consequently, the QP which is c-commanded by XP-sika cannot be outside the c-command domain of Neg at LF, making its wide scope over Neg impossible; see (20).
On the other hand, if it c-commands XP-sika before QR, a QP must raise beyond the XP-sika, due to the SRP, and therefore must be outside the c-command domain of Neg, since XP-sika must be in NegP-Spec at LF.
(25)QP2 XP-sika1 V-nai, where QP2 c-commands XP-sika1 before QR
a.LF1: [[IP QP2 [IP [NegP XP-sika1 [Neg' [VP t2 [V' t1V]] [Neg -nai]]]]]]
ok(23) forXP-sika, ok(22) SRP
b.LF2:[[NegP XP-sika1 [Neg' [VP QP2 [VP t2 [V' t1V] ] ] [Neg -nai]]]]
ok(23) forXP-sika, * (22) SRP
c.LF3: [IP [NegP [Neg' [VP QP2 [VP XP-sika1 [VP t2 [V' t1 V]]]] [Neg -nai]]]]
* (23) forXP-sika, ok(22) SRP
In order for the QP to be in the c-command domain of Neg at LF, the QP must either be in the c-command domain of the XP-sika violating the SRP as in (25b), or be in the c-command domain of Neg with the condition on XP-sika unsatisfied as in (25c). Therefore the QP which c-commands XP-sika before QR cannot be in the c-command domain of Neg at LF.
Thus the condition in (23), together with (22), accounts for the observation in (20) and (21); A QP which is c-commanded by XP-sika before QR cannot take wide scope over Neg since it cannot be outside the c-command domain of Neg at LF, while a QP which c-commands XP-sika before QR cannot be interpreted in the scope of Neg since it cannot be in the c-command domain of Neg at LF.
5. Some Consequences and Predictions
I argued in section 3 that rokuna-N must be c-commanded by Neg at LF ((4)). I then argued in section 4 that, given (22) and (23), a QP which c-commands XP-sika before QR cannot be in the c-command domain of Neg at LF, while, if c-commanded by XP-sika before QR, it must be in the c-command domain of Neg at LF. Assuming that rokuna-N undergoes QR at LF (see (16b) above), it follows that rokuna-N, if it c-commands XP-sika before QR, must c-command the XP-sika at LF due to the SRP, and, hence be outside the c-command domain of Neg, violating (4). We thus deduce that rokuna-N cannot occur in the position c-commanding XP-sika before QR.
5.1. XP-sika and rokuna-N in non-scrambling construction
Given the assumption that the subject asymmetrically c-commands the object in a non-scrambling sentence, the first prediction we make is (26).
(26)Prediction 1: Rokuna-N cannot occur as the subject of a non-scrambling sentence where XP-sika occurs as the object in the same sentence, being related to the same Neg.
(27)a.LF1: *[IP rokuna-N2 [IP [NegPXP-sika1 [Neg' [VPt2 [V' t1 V]] [Neg -nai]]]]]
* (4) forrokuna-N,ok(23) forXP-sika, ok(22) SRP
b.LF2: *[IP [NegPXP-sika1 [Neg' [VProkuna-N2 [VP t2 [V't1 V]]] [Neg -nai]]]
ok(4) forrokuna-N, ok(23) forXP-sika, * (22) SRP
This prediction is borne out, as illustrated in (28), which should be compared with acceptable (29a), where the subject XP-sika c-commands the object rokuna-N.
(28)*[Rokuna-sensei-ga] [Taro-sika] Mitubisi-ni syookaisi-nai
good-professor-NOM Taro-all:but Mitsubishi-DAT introduce-Neg
(koto)
(Comp)
(29)a.[Mori-sensei-sika] [rokuna-kaisya-o] Taro-ni syookaisi-nai
Mori-professor-all:but good-company -ACC Taro-DAT introduce-Neg
(koto)
(Comp)
(Roughly)‘All professors but Prof. Mori do not introduce any decent company to Taro.’
b.LF: ok [IP [NegPXP-sika1 [Neg' [VProkuna-N2 [VP t1 [V' t2 V]]] [Neg -nai]]]
ok(4) forrokuna-N, ok(23) forXP-sika, ok(22) SRP
The status of (29) is as expected; a QP which is c-commanded by XP-sika before QR can be in the c-command domain of Neg at LF, and hence there can be a legitimate LF representation like (29b), with all the conditions in question satisfied.
5.2. XP-sika and rokuna-N in a scrambling construction
The second prediction is regarding the scrambling construction. Given the structural ambiguity of ascrambled sentence, the two distinct c-command relations between rokuna-N and XP-sika are possible at LF for the phonetic string of the form rokuna-N-ACCXP-sika(subj)V.
(30)a.A’- scrambling: XP-sika(subj)rokuna-N-ACC
b.A-scrambling: rokuna-N-ACCXP-sika(subj)
(where α > β means α c-commands β.)
It is thus expected that (31), which is the scrambled counterpart of (29a), can be acceptable since it can be analyzed as an A’-scrambling case, where rokuna-N is c-commanded by XP-sika before QR.
(31)[Rokuna-kaisya-o] [Mori-sensei-sika] Taro-ni syookaisi-nai
good-company -ACC Mori-professor-all:but Taro-DAT introduce-Neg
(koto)
(Comp)
(Roughly) ‘All professors but Prof. Mori do not introduce any decent company to Taro.’
Now, if we ‘force’(31) to be an instance of A-scrambling, rokuna-N should be in a position c-commanding XP-sika before QR, and hence could not be in the c-command domain of Neg at LF, given (14). Thus we predict (32), since ‘resumption’ forces it to be an instance of A-scrambling, as seen in section 2.2.2.
(32)Prediction 2: ‘Resumption’ makes (31) unacceptable.
The prediction is also borne out, as indicated in (33a), whose LF representation should be (33b).
(33)a.*[Rokuna-kaisya-o] [Mori-sensei-sika] Taro-ni soko-o
good-company -ACC Mori-professor-all:but Taro-DAT there-ACC
syookaisi-nai (koto)
introduce-Neg (Comp)
b.LF: [IP rokuna-N1-ACC [IP t1 [NegP XP-sika2 [Neg' [VP t2 [V'soko-ACC V]] [Neg -nai]]]]
* (4) for on rokuna-N
6. Concluding remarks
I first demonstrated that the ‘A-scrambled’ object is outside the c-command domain of Neg at LF. I then argued for (4) and (23) (which means (5) by definition).
(4)Rokuna-N must be c-commanded by Neg at LF.
(5)XP-sika must c-command Neg at LF.
(23)XP-sika must be in NegP-Spec at LF.
References
Aoyagi, H. and T. Ishii. 1994. On NPI Licensing in Japanese.Japanese/Korean Linguistics 4: 295-311.
Fauconnier, G. 1975. Pragmatic Scales and Logical Structure.Linguistic Inquiryvol.6, No.3: 353-375.
Hasegawa, N. 1991. On Non-Argument Quantifiers: Floating Quantifiers and the Narrow Scope Reading.Metropolitan Linguistics 11: 52-78.
Hayashishita, J.-R. 1997. On the Scope Ambiguity in the Scrambling Construction in Japanese. ms. USC.
Hayashishita, J.-R.2004. Syntactic Scope and Non-Syntactic Scope. Doctoral Dissertation. USC.
Hoji, H. 1985. Logical Form Constraints and Configurational Structures in Japanese. Doctoral Dissertation, University of Washington.
Hoji, H. 2003.Falsifiability and Repeatability in Generative Grammar: A Case Study of Anaphora and Scope Dependency in Japanese.Linguavol.113, No.4-6: 377-446.
Hoji, H. and A. Ueyama. 1998. Resumption in Japanese. ms. USC.
Huang, C.-T. J. 1982. Logical Relations in Chinese and the Theory of Grammar. Doctoral Dissertation. MIT.
Imani, I. 1993. Hiteeryokabun-o Zenken-ni Motu Jyokenbun-ni-tuite (On Conditional Clauses including quantifiers and negative element).Nihongo-no Jyoken-Hyogen, ed. T. Masuoka, 203-222. Tokyo: Kurosio Publishers.