Two Types of Neg-Sensitive Elements
v.6 Kiyoko Kataoka 4/02/05
1. Introduction
2. Preliminary discussion: structural characteristics of Japanese negative sentences
2.1. Scope ambiguity between Neg and QP
2.2. Different landing sites of QR
2.3. Summary and a problem
3. O(bject) S(ubject)-construction
3.1. Structural ambiguity
3.1.1. Bound Variable Anaphora (BVA) Interpretation
3.1.2. Wide Scope Distributive Readings
3.1.3. Resumption
3.2. Deep DL : an element outside the c-command domain of Neg
3.3. Summary
4. Rokuna-N: an N(egation)S(ensitive)E(lement) that must be in the c-command domain of Neg
4.1. Rokuna-N as an QP
4.2. Rokuna-N (as?) a Deep DL
4.2.1. Rokuna-N and BVA
4.2.2. Rokuna-N and resumption
4.3. Summary
5. XP-sika: an NSE that must be outside the c-command domain of Neg
5.1. XP-sika as an QP
5.2. XP-sika as a Deep DL
5.2.1. XP-sika and BVA
5.2.2. XP-sika and resumption
5.3. Proposal
5.3.1. Scope interaction between XP-sika and QP
5.3.2. Analysis
5.3.3. Examples in OS-construction
5.4. Summary
6. Consequences: XP-sika and rokuna-N
6.1. Prediction 1 : XP-sika and rokuna-N in a S(ubject)-O(bject) construction
6.2. Prediction 2: XP-sika and rokuna-N in a O(bject)-S(ubject) construction
6.3. Summary
7. Further consequences and implications
7.1. (N-CM) dare-mo/nani-mo as a Deep DL
7.2. (N-CM) dare-mo/nani-mo and XP-sika
7.2.1. SO-construction
7.2.2. OS-construction
7.3. An alternative analysis
7.4. (N-CM) dare-mo/nani-mo and rokuna-N
7.4.1. SO-construction
7.4.2. OS-construction
7.5. Summary
8. Concluding remarks.
Appendix: Judgmental fluctuation and variation
References
1. Introduction
As is well known, most languages have elements which must occur with sentential negation. Japanese expressionssuch as XP-sika'all but XP'and rokuna-N'good/decent N'are among those as illustrated below.
(1)a.Taro-wa manga-sika yoma-nai. / *yomu.
Taro-TOP comics-all:but read-Neg / read
(Roughly) 'Taro does not read any kind of book but comics.'
b.Taro-sika manga-o yoma-nai. /* yomu.
Taro-all:but comics-ACC read-Neg / read
(Roughly) 'Nobody but Taro reads comics.'
(2)a.Jiro-wa kyoositu-de rokuna-koto-o iwa-nai. / *iu.
Jiro-TOP classroom-in good-thing-ACC say-Neg / say
(Roughly) 'Jiro does not say any good thing in the classroom.'
b.Rokuna-gakusei-ga gakkai-de happyo-o si-nai. / *suru.
good-student-NOM conference-LOC presentation-ACC make-Neg /make
(Roughly) 'No good students make a presentation at the conference.'
As seen above, these expressions require sentential negation. We henceforth refer tothem as'Neg(ation)-sensitive' elements (NSE).[1] Based on the observation that an NSE in English such as any(-)cannot occur outside the c-command domain of the negative element (not), as observed in (3), it has been generally agreed that the condition (4) holds of NSEs.
(3)a.He didn't [ invite anybody.]
b.*Anybody didn't [invite him.]
(4)An NSE must be c-commanded by Neg at LF. (Klima1964)
Following this analysis, Japanese NSEs have also been treated in general as falling under the condition in(4) (Kato 1985, 1994, 2002, Kuno 1995 and many others), and this condition has been widely accepted in the field. (I will hereafter refer to the condition in (4) as the Neg-c-command condition.)
[The aim of this paper is to demonstrate that not all NSEs obey the condition in (4), and that there are at least two types of NSEs in terms of the c-command relation with Neg, and more specifically rokuna-N obeys (4) but XP-sika does not. <==Is this the aim of the paper? If so, you might want to declare that the aim of this paper is descriptive in nature, or something like that. See my remarkd on the concluding section.]
there are some NSEs in Japanese which MUST C-COMMAND Neg, rather than being c-commanded by Neg, such as XP-sika in (1). I thus claim that, though there are some NSEs which are subject to the condition in (4) such as rokuna-N in (2), not all NSEs obey the condition in (4), and that there are at least two types of NSEs in terms of the c-command relation with Neg.
The organization of the paper is as follows. In the next section, after introducing several assumptions we should adopt regarding the structure of negative sentences in Japanese, I will point out that it is necessary to identify an element which is independently demonstrated to be outside the c-command domain of Neg, in order to discuss the validity of the condition (4). In section 3, I will first provide[three methods to disambiguate the structure of 'scrambling' construction, and then demonstrate that the so-called 'A-scrambled' object cannot be in the c-command domain of Neg at LF. Making use of the 'A-scrambled' object, <==I will later comment on this; but I will leave it uncommented on for now.] I will argue, in section 4, that rokuna-N is subject to the condition (4), and, in section 5, show that XP-sika is not subject to the condition (4). I will then propose a structural condition for XP-sika based on the scope interpretation in a sentence that involves XP-sika, Neg, and another QP. In section 6, I will provide supporting evidence for the hypotheses put forth in this paper, by showing that [the predictions made under the proposed analyses <==You may want to make their content known to the reader, if that is possible] are borne out. Section 7 is to discusses consequences and further implications from the discussions here in regard to xxx. I will suggestsuggest will be concluded that dare-mo/nani-mo, another NSE in Japanese must be outside the c-command domain of Neg at LF as well as XP-sika. Section 8 is for provides the conclusion and remaining issues.
2. Preliminary discussion: structural characteristics of Japanese negative sentences
2.1. Scope ambiguity between Neg and QP
In this subsection, we will go over the scope phenomena which have been reported in the literature, and, in consideration of them,introduce the assumptions we should adopt regarding the structural characteristics of Japanese negative sentences.
It has been observed that, in Japanese, both a subjectNP and non-subjectNPs can be within the scope of negation (Kuno 1980, Hasegawa 1991 (1993), Imani 1993). Let's see the examples below.
(5)a.QP-NOM NP-ACC/DAT V-Neg
[QPgoninizyoo-no gakusei ]-ga so-no hon-o yoma-na-katta (koto)
5:or:more-GEN student-NOM that-GEN book-ACC read-Neg-Past (Comp)
'Five or more students did not read that book.' 5-or-moreNeg, Neg5-or-more
b.NP-ACC/DAT QP-NOM V-Neg
so-no hon-o [QPgoninizyoo-no gakusei ]-ga yoma-na-katta (koto)
that-GEN book-ACC 5:or:more-GEN student-NOM read-Neg-Past (Comp)
'That book, five or more students did not read.' 5-or-moreNeg, Neg5-or-more
(6)a.NP-NOM QP-ACC/DAT V-Neg
so-no gakusei-ga [QPgosatuizyoo-no hon ]-o yoma-na-katta (koto)
that-GEN student-NOM 5:or:more-GEN book-ACC read-Neg-Past (Comp)
'That student did not read five or more books.' 5-or-moreNeg, Neg5-or-more
b.QP-ACC/DAT NP-NOM V-Neg
[QPgosatuizyoo-no hon ]-o so-no gakusei-ga yoma-na-katta (koto)
5:or:more-GEN book-ACC that-GEN student-NOM read-Neg-Past (Comp)
'Five or more books, that student did not read.' 5-or-moreNeg, Neg5-or-more
All the sentences above are ambiguous with respect to the scope interpretation between the QP and Neg: one reading in which the QP takes scope over Neg (henceforth, QP >Neg), and the other in which Neg takes scope over the QP (henceforth,Neg> QP). The two interpretationsfor (5a), for instance, can be informally describedas in (7)and (8), respectively.
(7)QP > Neg : 5ormorex (x= student) NOT (xread that book)
The interpretation where the number of the students who did not read that bookis concerned.
(8)Neg > QP : NOT 5or morex (x= student) (xread that book)
The interpretation where the number of the students who read that bookis concerned.
The latter interpretation Neg > QP is often observed to be more or less marginal. It is pointed out, however, that the judgmental variation is attributed to pragmatic factors (Kato 1985: Chapter xxx, Imani 1993), and theinterpretation obtains more easily if the relevant sentence is embedded in some context, such as the conditional clause, as pointed out by Imani (1993).
(9)[QPGoninizyoo-no gakusei ]-ga so-no hon-o yoma-na-kereba, ....
5:or:more-GEN student-NOM that-GEN book-ACC read-Neg-if
'If five or more students did not read that book, ....'
In anycase, it seems to be the case that every NP in Japanese can be in the scope of negation, which must be derived by whatever analysis one might take.
2.2. Different landing sites of QR
We adopt the widely-accepted assumption that every NP, i.e., the subject NPand non-subject NPs, base-generates in a position inside the VP(or vP in the recent framework), and assume, according to Kuroda 1988 and Kitagawa 1986,that the subject NP does not obligatorily move up to the Specifier position of the IP. Furthermore if we assume (10), as is generally assumed, regarding the syntactic position of the negative element,every NP in Japanese can be in the c-command domain of Neg (-nai) at LF.
(10)Negative element Neg (-nai) is in the sister position of the VP at LF.
(Masuoka 1989, Pollock 1989)
Given all and the general assumption that the c-command domain of an element α will be the scope of α (Reinhart 1983), it can be given an account that every NP in Japanese can be in the scope of Neg. Now if we adopt the assumption, following Hasegawa 1991, that Quantifier Raising (QR) at LF[2] (May 1977, 1985) can either be an IP-adjunction (whose landing site is a position c-commanding Neg) or a VP-adjunction (whose landing site is a position c-commanded by Neg), the LF representations as illustrated in(11) are available.
(11)a.[IP QP [IP [NegP [VP ...t ... V ]-Neg ] ] ]
b.[IP [NegP [VP QP [VP ... t ... V ]]-Neg ] ]
Once these two LF-representations are possible,the observationgiven above that the scope relation between an NP and Neg are ambiguous can be derived. Thus it is reasonable to adopt these assumptions, and many studies in the Japanese literature, including Hasegawa 1991 (1993), have adopted the assumptions along this line.
2.3. Summary and a problem
Based on the discussion above, wWe will adopt the followings assumptions.
(12)Assumptions:
(i)The subject NP as well as the non-subject NP can remain inside the VP at LF. (Kuroda 1988 and Kitagawa 1986)
(ii)Negative element Neg (-nai) is in the sister position of the VP at LF.(Pollock 1989, Masuoka 1989)
(iii)QRat LF can either be an IP-adjunction (whose landing site is a position c-commanding Neg) or a VP-adjunction (whose landing site is a position c-commanded by Neg) at LF. (Hasegawa 1991)
Once we accept all the assumptions abovein (12), as is generally done in the literature, the distributional characteristics of Japanese NSEs that they can occur not only in the subject position but also in the non-subject positions can also be given an account by the analysis is consistent withunder the Neg-c-command condition in (4), without giving rise to any problem, since every NP in Japanese can be c-commanded by Neg at LF under the assumptions above.
(4)An NSE must be c-commanded by Neg at LF. (Klima 1964)
Given the possibility of VP adjunction of QR, the subject and the object QP can both be within the scope of Neg at LF. It is perhaps It can be said that it is for this reason, one may suggest, that the validity of the condition (4) in Japanese has never been questioned seriously.
[Notice, however, that, given all the assumptions above, it cannot be verified whether the condition (4) actually works or not on the licensing of NSEs, since, under those assumptions, every NP in Japanese can be inside the c-command domain of Neg at LF so that there cannot be a position which is necessarily outside the c-command domain of Neg. This only means that the distributional characteristics of Japanese NSEs can be accounted for by the condition (4) under the assumptions above, but not that the condition is necessary for Japanese NSEs. <==Try to shorten this part.] In order to show that an NSE is subject to the condition (4), it is necessary to show that the NSE cannot occur in a position where the condition (4) is not satisfied, i.e., a position outside the c-command domain of Neg. It is therefore necessary to identify an element which is independently demonstrated to occupy a position outside the c-command domain of Neg at LF in order to examine the validity of the condition (4).
[In the following sections, I will first demonstrate that, if we examine in detailso-called 'scrambled' sentencesin the Object-Subject-Verb order (NP-ACC/-DAT NP-NOM V) (henceforth, we will refer to themas OS-constructions,according tofollowing Ueyama 1998), there can be an elementwhich cannot be in the c-command domain of Neg at LF, and that it is possible to identify the relevant NP through interactions with other phenomena such as QP-scope interpretation or bound variable anaphora (BVA). Then, making use of the NP in question, I will testhow valid the condition (4) is, and then proposethe syntactic conditions which work on NSEs. <==Redo.]
3. O(bject) S(ubject)-construction
3.1. Structural ambiguity
It is generally agreed in the Japanese literaturethat, whilesentences with the word order of Subject-Object-Verb (NP-NOMNP-ACC/-DAT V) (henceforth, SO-construction, according to Ueyama 1998) have only one kind of structural relation between the subject and the object at LF, where, corresponding to its phonetic form (PF), the subject NP asymmetrically c-commands the objectNP (Hoji 1985, 2003), two kinds of structural relation at LF are available to sentences in the OS-construction(Saito 1992, Ueyama 1998): the one where the subject NP c-commands the object NP, and the other where the object NP c-commands the subject NP. Thus, for the sentence with the PF (13), the two LF-representations are possible as indicated in (14).
(13)PF: NP-ACC/-DAT NP-NOM V
(14)a.LF-representation in which the subject NP c-commands the object NP
NP-NOM
NP-ACC/-DAT V
b.LF-representation in which the object NP c-commands the subject NP
NP-ACC/-DAT
NP-NOM V
This analysis is based on the observations, as is introduced in thenext subsection, regarding binding phenomena and scope interpretation among others. The case which is assumed to have the LF-representation in (14a) has been called 'A-bar-scrambling', since it shows 'A-bar-properties' such as reconstruction effects, and the other which is assumed to have the LF in (14b) has been called 'A-scrambling' with 'A-properties' such as absence of Weak Cross Over effects (WCO).
In this paper, we will, according to Ueyama 1998, refer to the sentence with the PF (13)and the LF (14a), i.e., 'A-bar-scrambling' case, as Surface OS, and its object NP as Surface DL (DL means 'dislocated element'), and we will refer to the sentence with the PF (13) and the LF (14b), i.e., 'A-scrambling' case, as Deep OS, and its object NP as Deep DL.[3]
It is necessary to identify on which LF-representation the interpretation of the sentence in question is based if we discuss structural issues based on the observations of sentences in the OS-construction. In the following subsection, we will introduce the analyses of Bound variable anaphora (henceforth, BVA)interpretation, wide-scope-distributive reading (henceforth, DR), and resumption, to which we can have recourse in order to identify the structure of a sentence with the PF of (13), and, more crucially, to identify a Deep DL, i.e., a dislocated object in a Deep OS case. Then I will argue that elements which are identified to be a Deep DL in the OS-construction cannot be in the c-command domain of Negby demonstrating that they cannot be interpreted in the scope of Neg.
3.1.1. Bound Variable Anaphora (BVA) Interpretation
The first way to identify the structure of a sentence in the OS-construction is to examine the availability ofBVA. The necessary condition for a BVA interpretation between a QP α and a dependent term β (henceforth, BVA(α,β)) to obtain is that the trace of α left by QR at LF should c-command β as argued in Ueyama 1998 and Hoji 2003. In other words, the relevant QP must c-command the dependent term in the position which the QP occupies before QR. Thus we can identify the c-command relation between the subject and the object in the OS-construction by making use of BVA.
Let's start from the examples in theSO-construction. Given the assumption that only one type of LF, where the subject NP c-commands the object NP, is possible for the SO-construction (Hoji 1985, 2003), it is predicted that the BVA interpretation cannot obtain in (15a), where the relevant QP does not c-command the dependent term (soko) in the pre-QR position. On the other hand, it is expected that the BVA interpretation can obtain in (15b), where the QP c-commands the dependent term in the pre-QR position. This prediction is borne out as shown bythe examples in(16), and the unavailability of BVA in (15a) is a so-called WCO effect.
(15)a.PF: [ ...soko ... ]-NOM QP-ACC/-DAT V* BVA(QP, soko)
b.PF: QP-NOM [ ... soko ... ]-ACC/-DAT VokBVA(QP, soko)
(16)a.Sokoi-no torihikisaki-ga [Itutuizyoo-no ginkoo]i-ni Nissan-o suisensi-ta.
it-GEN client-NOM 5:or:more-GEN bank-DAT Nissan-ACC recommend-Past
'Itsiclientrecommended Nissan to[each of five or more banks]i.'
* BVA(itutuizyo-no-ginkoo, soko)
b.[Itutuizyoo-no ginkoo]i-ga sokoi-no torihikisaki-ni Nissan-o suisensi-ta.
5:or:more-GEN bank-NOM it-GEN client-DAT Nissan-ACC recommend-Past
'To[each of five or more banks]i, itsiclientrecommended Nissan.'
ok BVA(itutuizyo-no-ginkoo, soko)
Next, it is observed that the BVA interpretation obtains both in (17a) and (17b), which are in the OS-construction.
(17)a.Sokoi-no torihikisaki-ni [Itutuizyoo-no ginkoo]i-ga Nissan-o suisensi-ta.
it-GEN client-DAT 5:or:more-GEN bank-NOM Nissan-ACC recommend-Past
'Toitsiclient, [each of five or more banks]irecommended Nissan.'
okBVA(itutuizyo-no-ginkoo, soko)
b.[Itutuizyoo-no ginkoo]i-ni sokoi-no torihikisaki-ga Nissan-o suisensi-ta.
5:or:more-GEN bank-DAT it-GEN client-NOM Nissan-ACC recommend-Past
'To[each of five or more banks]i, itsiclientrecommended Nissan.'
okBVA(itutuizyo-no-ginkoo, soko)
Now if we assume that two types of c-command relation between the subject and the object,as indicated above,are available to the sentences in the OS-construction, each of the sentences in (17)has possibly two LF representations; one in which the subject-NP c-commands the object-NP, and the other in which the object NP c-commands the subject NP. (Henceforth, α means α c-commands .)
(18)a.PF: [ ... soko ... ]-ACC/-DAT QP-NOM Vok BVA(QP, soko)
b.LF1 : QP-NOM[ ... soko ... ]-ACC/-DAT
c.LF2: [ ... soko ... ]-ACC/-DAT QP-NOM
(19)a.PF: QP-ACC/-DAT [... soko... ]-NOM Vok BVA(QP, soko)
b.LF1: [... soko... ]-NOMQP-ACC/-DAT
c.LF2:QP-ACC/-DAT[... soko... ]-NOM
Thus the BVA interpretation between the QP and the dependent term soko may obtain in both sentences with the PF (18a) and sentences with the PF(19a), since, for each of them, the LF-representation where the trace of the QP after LF-movement c-commands soko is available.
Since it has been said that the NP in questionshould bein an A-positionif it binds successfully a dependent term from its position, the availability of BVA has been also regarded as a test for determining whether a given syntactic position is an A-position or an A'-position. Given the condition for BVA, if the scrambled object gives rise to a BVA interpretation with the dependent term in the subject NP, the object should be in an A-position c-commanding the subject NP. Thus the BVA test has been applied to sentences in the OS-construction in order to identify their structural characteristics (Yoshimura 1992, Ueyama 1998).