Scrambling, resumption, and scope of Neg
Kiyoko KATAOKA draft v.4 9/014/05
1. Introduction
1.1. Scope ambiguity between Neg and a QP
1.2. Absence of scope ambiguity in a scrambling case
1.2.1. A/A'-properties of the scrambling construction
1.2.2. Absence of Neg>NP reading for the A-scrambled object
2. Possible account 1: A-movement analysis for A-scrambling
2.1. Blocking of LF-lowering by Neg for A-moved NP
2.2. Problem for A-movement analysis: 'resumption'
3. Alternative account: base-generation analysis for A-scrambling (Ueyama 1998)
3.1. Absence of Neg>QP for A-ScramObj
3.2. 'Resumption' under base-generation analysis
4. Problem in derivational analysis for A/A'-properties of ScramObj (Saito 2003)
5. A supporting argument for base-generation analysis: QP-scope interpretation
5.1. A-ScramObj and 'resumption'
5.2. Absence of Neg>QP for A-ScramObj
6. Conclusion
References
As a general comment, I would say that you should (perhaps indirectly?) address or respond to the comments by the NELS reviewer(s). You should also try to attain a higher level of repeatability across speakers (even within the GGES) re. your empirical claims before you publish it in the JK proceeding or submit it to a journal.
1. Introduction
The aim of this paper is to provide a discussion on how the so-called 'scrambling' construction, i.e., sentences with the word order of Object-Subject-Verb, should be derived, based on the observations involving negation and resumption. I will examine three possible analyses proposed in the field, and argue that the sentence-initial object in the scrambling construction should be assumed to be base-generated in its surface position when it shows A-properties, as argued by Ueyama (1998), and Ueyama (2003), and contrary to the widely adopted assumption that every 'scrambled' object should be base-generated in its theta-position inside the VP (or vP in the recent frame work), as argued by Saito 1992, 2003 and Miyagawa 2001.
The organization of the paper is as follows: in this section for Introduction, afterproviding a generally agreed observation regarding the scope interaction between Neg and a QP, I will report a peculiar phenomenon involving negation in the scrambling construction, then introduce A/A'-properties, which are observed in the scrambling construction and have long been discussed in the literature, and claim that the observation is subsumed under the generalization that the 'scrambled' object with A-properties cannot be in the scope of Neg. In section 2, in order to give an account for that generalization,I will first examine the A-movement analysis by Miyagawa (2001), among others, for the 'scrambled' object with A-properties, and point out its problem. In section 3, I will show that the analysis under the base-generation analysis by Ueyama 1998 can give an account for the generalization without any problem. In section 4, I will discuss, as the third possible account for the generalization, the derivational analysis by Saito (2003) for A/A'-properties of the scrambling construction, and point out the problems which we will suffer from in giving an account for the generalization under his analysis. In section 5, I will provide a supporting discussion for our conclusion making use of the QP-scope interpretations. Section 6 is for the conclusion of this paper.
1.1. Scope ambiguity between Neg and a QP
It has been observed that the scope relation between Neg and a QP is ambiguous in Japanese (e.g., Kuno 1980). As in (1), the subject NP as well as non-subject NPs can be in the scope of Neg.
(1)a.[QPitutuizyoo-no ginkoo ]-ga Toyota-ni yuusisi-nak-atta (koto)
5:or:more-GEN bank-NOM Toyota-DAT finace-Neg-Past (Comp)
'Five or more banks did not finance Toyota.'
5-or-moreNeg, Neg5-or-more
b.UFJ-ginkoo-ga [QPitutuizyoo-no kigyoo ]-ni yuusisi-nak-atta (koto)
UFJ Bankt-NOM 5:or:more-GEN company-DAT finance-Neg-Past (Comp)
'UFJ Bank did not finance five or more companies.'
5-or-moreNeg, Neg5-or-more
The interpretations in question of the example (1a), for instance, can be informally described as the following.
(2)a.QP > Neg : 5 or more x (x= bank) NOT (xfinance Toyota)
The interpretation where the number of the banks which did not finance Toyota is concerned.
b.Neg > QP : NOT 5 or more x (x= bank) (xfinance Toyota)
The interpretation where the number of the banks which financed Toyota is concerned.
As observed in (3), the ambiguity continues to obtain in the 'scrambled' counterparts.
(3)a.Toyota-ni [QPitutuizyoo-no ginkoo ]-ga yuusisi-nak-atta (koto)
Toyota-DAT 5:or:more-GEN bank-NOM finace-Neg-Past (Comp)
5-or-moreNeg, Neg5-or-more
b.[QPitutuizyoo-no kigyoo ]-ni UFJ-ginkoo-ga yuusisi-nak-atta (koto)
5:or:more-GEN company-DAT UFJ Bank-NOM finance-Neg-Past (Comp)
5-or-moreNeg, Neg5-or-more
1.2. Absence of scope ambiguity in a scrambling case
However, it is observed that the Neg>QP reading cannot obtainin the following examples, which are in the scrambling construction:if, as in (4a),thebound variable anaphora (BVA) interpretation is forced to obtain betweenthe object QP1 and the dependent termsoko 'it' in the subject NP, the 'scrambled' object (henceforth ScramObj) cannot be in the scope of Neg. As in (4b), if what seems to be 'resumption' is involved, the ambiguity disappears as well in the scrambling construction.
(4)a.[Itutuizyoo-no kigyoo]QP1-ni [soko-no torihiki-ginkoo]-ga yuusisi-nak-atta.
5:or:more-GEN-company-DAT it-GEN-dealing:bank-NOM finance-Neg-Past
'To five or more companies, its own bank did not finance.'
with BVA, *Neg QP1
b.[Itutuizyoo-no kigyoo]QP1-ni UFJ-ginkoo-ga soko-ni yuusisi-nak-atta.
5:or:more-GEN-company-DAT UFJ Bank-NOM it-DAT finance-Neg-Past
*Neg QP1
These examples should be compared with (3b) above.
1.2.1. A/A'-properties of the scrambling construction
It has been known that the scrambling construction displays A'-properties such as reconstruction effects as in (5a), and A-properties such as an absence of weak crossover effects as in (5b)(Saito 1992, Ueyama 1998, among others).
(5)a.Sokoi-no torihikisaki-ni [itutuizyoo-no kigyoo]QP1-ga kitaidoori tousisi-ta.
it-GENclient-DAT 5:or:more-GEN company-NOM expectedly invest-Past
'To itsi client, [each of five or more companies]i invested expectedly.'
okBVA
b.[Itutuizyoo-no kigyoo]QP1-ni sokoi-no torihikisaki-ga kitaidoori tousisi-ta.
5:or:more-GEN company-DAT it-GENclient-NOM expectedly invest-Past
'To [each of five or more companies]i, itsi client invested expectedly.'
okBVA
Given the general assumption that, in order for the BVA interpretation to obtain between a QP α and a dependent term , α must c-command at pre-QR position (Reinhart 1983, Ueyama 1998, Hoji 2003), the subject must c-command the object at pre-QR position in (5a),and the object must c-command the subject at pre-QR position in (5b), in order for the BVA to obtain, respectively. The former case has been said to show reconstruction effects, which are A'-property, and the latter has been said to show an A-property since binding relation obtains for the ScramObj from that surface position.
Thus, based on the observations above, it has been assumed that the two kinds of c-command relation at LF between the subject and the object are available to the sentences in the scrambling construction: the one in which the subject c-commands the object, and the other in which the object c-commands the subject. The former (e.g., (5a)),corresponding to A'-properties, has been called A'-scrambling, and the latter (e.g., (5b)), corresponding to A-properties, has been called A-scrambling.
Furthermore, it is pointed out and discussed in Hayashishita 1997, Hoji & Ueyama 1998 and Hoji 2003 that what appears to be 'resumption' is compatible with A-properties, but it is not compatible with A'-properties. As in (6), resumptive element (soko) is possible with the BVA obtained, which means that the object is aScramObj with A-properties (henceforth, A-ScramObj), whereas, as in (7), the sentence involving resumption is not acceptable if we force the BVA interpretation through 'reconstruction', which means that the object in that case is a ScramObj with A'-properties (henceforth, A'-ScramObj).
(6)[Itutuizyoo-no kigyoo]QP1-ni sokoi-no torihikisaki-ga soko-ni tousisi-ta.
5:or:more-GEN company-DAT it-GENclient-NOM it-DAT invest-Past
'To five or more companiesi, itsi client invested to it expectedly.' (with BVA)
(7)*Sokoi-no torihikisaki-ni [itutuizyoo-no kigyoo]i-ga soko-ni tousisi-ta.
it-GEN client-DAT 5:or:more-GEN bank-NOM it-DAT invest-Past
As seen above,'resumption' seems compatible with A-properties but not with A'-properties.[1]
1.2.2. Absence of Neg>NP reading for theA-scrambled object
Given the observation regarding the A/A'-properties of the scrambling construction in the previous subsection, the generalization can be drawn, from the observation in (4), that the A-ScramObjcannot be in the scope of Neg: the object in (4a), with the BVA obtained, which means that the object then c-commands the subject showing an A-property, cannot be in the scope of Neg, nor can the object in (4b), with resumption, which shows that the object then is an A-ScramObj. I would like to add the example below, which is a negative counterpart of (6) above. Crucially, the object cannot be in the scope of Neg, either.
(8)[Itutuizyoo-no kigyoo]QP1-ni sokoi-no torihikisaki-ga soko-ni tousisi-nak-atta.
5:or:more-GEN company-DAT it-GENclient-NOM it-DAT invest-Neg-Past
with BVA(QP1, soko), *Neg QP1, okQP1 Neg
The object in (8) is also an A-ScramObj, with the BVA obtained and resumption involved. Thus we can draw the generalization below.
(9)Generalization 1: the A-ScramObjcannot be in the scope of Neg.
The goal of this paper is to provide an account for the observation in (9), as exemplified by (4), and explore its implications for the analyses of the scrambling construction. Though there have been disputing views regarding how the scrambling construction is derived, I will argue, based on the observations provided in this paper, against theA-movement analysis of theA-ScramObj(Miyagawa 2001 among others) and the derivational account of A/A'-properties of the scrambling construction (Saito 2003), and argue in support of thebase-generation analysis of theA-ScramObjproposed in Ueyama 1998.
2. Possible account 1: A-movement analysis for A-scrambling
I will first examine, as possible account 1, theA-movement analysis of theA-ScramObj(Miyagawa 2001 among others), and point out a problem which we will face if we adopt this analysis.
(10)Possible account 1: theA-movement analysis of A-ScramObj (Miyagawa 2001 among others):
theA-ScramObjis moved to its surface position from its theta-position inside the VP by A-movement.
2.1. Blocking of LF-lowering by Neg for A-moved NP
It has been said, since it was argued by May (1977),that the A-moved NP undergoes 'quantifier-lowering'; as seen in (11),(11a) allows the reading in(11b).
(11)a.Someone is likely to address the rally.
b.It is likely that someone will address the rally.
It is claimed that the narrow scope for someone obtains by someone adjoining to the lower IP at LF and being inside the scope of likely at LF.
It is pointed out, however, that LF-lowering of the A-moved NP is blocked by a negative predicate (Lasnik & Saito 1991); as seen in (12),(12a) does not allow the reading in(12b).
(12)a.Someone is unlikely to address the rally.
b.It is unlikely that someone will address the rally.
If we accept this analysis and assume that 'quantifier-lowering' for the A-moved NP is blocked by Neg, the absence of the wide-scope reading over Neg for A-ScramObj, which is generalized as in (9), couldbe accounted forunder the assumption that theA-ScramObjis moved to its surface position from its theta-position inside the VP by A-movement as argued by Miyagawa (2001, among others), provided that the scope of α is its c-command domain at LF and that Neg is in the VP (or vP in the recent framework)-adjoined position at LF. Thus it seems that we can give an account for the generalization in (9) by assuming that the A-ScramObj is moved by A-movement from its theta-position.
2.2. Problem for A-movement analysis: 'resumption'
However, we willsuffer from a problem if we adopt theA-movement analysis and assume that the A-properties of the ScramObj come from A-movement.
As already mentioned above with the examples (6) and (7), what appears to be 'resumption' is compatible with A-properties, butit is not compatible with A'-properties,which has been discussed in Hayashishita 1997, Hoji & Ueyama 1998 and Hoji 2003. In addition to them,the ScramObj in(8), which is a negative counterpart of the example of A-ScramObjin(6) and repeated below, crucially, cannot be in the scope of Neg in accordance with the generalization in (9).
(8)[Itutuizyoo-no kigyoo]QP1-ni sokoi-no torihikisaki-ga soko-ni tousisi-nak-atta.
5:or:more-GEN company-DAT it-GENclient-NOM it-DAT invest-Neg-Past
with BVA(QP1, soko), *Neg QP1, okQP1 Neg
On the other hand, it has been generally assumed thatA-movement does not allow resumption (Ross 1967: 6). Therefore the observation above regarding 'resumption' is problematic if we assume that the A-properties of theScramObj come from A-movement, since it is inconsistent with that general assumption.
3. Alternative account:base-generation analysis for A-scrambling (Ueyama 1998)
Next, we will examine an alternative analysis by Ueyama (1998).
(13)Possible account 2: abase-generation analysis of A-ScramObj (Ueyama 1998)
It will be shown that, if we adopt thebase-generation analysis for theA-ScramObjas in (14), as argued by Ueyama (1998), the observations above can be accounted for without any problem.
(14)The A-ScramObj is base-generated in its surface-position c-commanding the subject. (Ueyama 1998)
3.1. Absence ofNeg>QPfor A-ScramObj
The generalization in (9) that the A-ScramObjcannot be in the scope of Neg, as observed in(4a, b), which are repeated below, is also as predicted under the base-generation analysis.
First, in (4a), in order for the BVA reading to obtain, the object QP must c-command the subject NP at pre-QR position. Under the base-generation analysis, the ScramObj QP is assumed to be base-generated in its surface position c-commanding the subject NP, and thus its LF should be as illustrated in (15).
(4)a.[Itutuizyoo-no kigyoo]QP1-ni [soko-no torihiki-ginkoo]-ga yuusisi-nak-atta.
5:or:more-GEN-company-DAT it-GEN-dealing:bank-NOM finance-Neg-Past
'To five or more companies, its own bank did not finance.' with BVA, *Neg QP1
(15)LF: [ [ ... ]QP1-DAT [VP [ ... soko ... ]-NOM V] ]
Now if we assume that the base-generated position of the A-ScramObj is outside the c-command domain of Neg, as extensively argued in Kataoka 2004 and to appear, the LF representation of (4a) will be (16) below.
(16)LF: [ [ ... ]QP1-DAT[NegP [Neg'[VP [ ... soko ... ]-NOM V] [Neg -nai] ] ] ]
The ScramObjin (4a), consequently, can never be in the c-command domain of Neg throughout its derivation, given the assumption that movement is never downward, thus making the A-ScramObj impossible to be in the scope of Neg.
3.2. 'Resumption' underbase-generation analysis
Next, the ScramObj in the example (4b), which is repeated below, is also assumed under Ueyama's analysis to be base-generated in its surface position, since the ScramObj should be regarded to be an A-ScramObjinvolving resumption, which is, as observed above in 1.2.1, compatible only with A-properties but not with A'-properties.
(4)b.[Itutuizyoo-no kigyoo]QP1-ni UFJ-ginkoo-ga soko-ni yuusisi-nak-atta.
5:or:more-GEN-company-DAT UFJ Bank-NOM it-DAT finance-Neg-Past
*Neg QP1
The LF-representation should be as schematized below, and the ScramObj cannot be in the c-command domain of Neg throughout its derivation as well as the one in (4a).
(17)LF: [ [ ... ]QP1-DAT [NegP [Neg' [VP NP-NOM soko-ni V] [Neg -nai] ] ] ]
According to the discussions in Hoji & Ueyama 1998 and Hoji 2003, the possibility of 'resumption'in the case oftheA-ScramObjis precisely what is expected under the base-generation analysis of theA-ScramObj. It is assumed that what appears to be resumption is actually an overt counterpart of an empty argument which occupies the theta-position, and that it functions like a null operator in the English tough construction in order to make a syntactic relation betweenthe theta-position and the ScramObj.
4. Problem in derivational analysis for A/A'-properties of ScramObj (Saito 2003)
Finally, I will discuss, as Let us now turn to the third possible account, a derivational account of A/A'-properties of the scrambling construction put forth in Saito 2003. It will be shown that the absence of the wide-scope reading for Neg over the A-ScramObj, as exemplified by (4), also argues against Saito's (2003) derivational analysis for A/A'-properties of the ScramObj.
Saito's (2003) analysis is can be summarized as in (18).
(18)Possible account 3: a derivational account of A/A'-properties of 'scrambling' (Saito 2003)
a.Japanese scrambling is not feature-driven and is truly an optional movement operation (Saito 2003: section 3.2). It does not have subtypes like A-scrambling and A'-scrambling with different landing sites (Saito 2003: section 1).
b.Scrambling is subject to radical reconstruction at LF. (Saito 2003: section 4)
c.Adopting a copy and deletion analysis of movement by Chomsky (1993), chains, which are formed by movement of features, are interpreted as they are formed, and the deletion for interpretation applies as soon as a chain is formed in the derivation (Saito 2003: section 1).
Under his this analysis,it is assumed that every dislocated NP in the scrambling construction is base-generated in its theta-position, and that, though it moves up in the process of derivation, it must always be 'reconstructed' at LF in any case as in(see (18b)).[2] It is also assumed, as in (18c), that the relevant formal relation underlying the A/A'-properties, such as the BVA reading, is established or checked,derivationally, at some stage of the derivation.
Those assumptions adopted by Saito (2003), however, [do not make correct predictions==>or "make a wrong prediction? Which did you mean?] correctly regarding the scope relation between Neg and an NP. , i.e., the absence of the wide-scope reading for Neg over the A-ScramObj, as exemplified by (4). If a scope relation between α and were assumed to obtain based on their c-command relation at some stage of the derivation, even an A-ScramObjwould be expected to be in the scope of Neg, since even an A-ScramObj would certainly be in the c-command domain of Neg while being in its theta position. IOr, if it were determined on the basis of the LF-representation, the same would be expected since even an A-ScramObjshould be 'reconstructed' at LF being inside the c-command domain of Neg at LF. Thus the absence of the wide-scope reading for Neg over the A-ScramObjcould not be predicted correctly. It is not clear how one could account for the impossibility of that reading under Saito's(2003) analysis.
5. A supporting argument forbase-generation analysis: QP-scope interpretation
In this section, I will provide another argument in support of the conclusion reached above. The argument will be constructed based on QP-scopeinterpretations which involve a wide-scope distributive reading (DR). We will make crucial reference to the availability of scope reconstruction depending on the presence/absence of 'resumption', and the availability of the wide scope reading for Neg over the ScramObj, and see that the observed phenomenaare also in accordance to the generalization in (9) above. We will then show that they can also be given an account by the analysis put forth here under the base-generation analysis by Ueyama (1998).