Negative Concord Items in coordinate structure Ryoichiro Kobayashi

Negative Concord Items in coordinate structure
Ryoichiro Kobayashi and Taihei Asada
Whether syntactic verb raising exists in Japanese has been a topic of debate in the literature of
Japanese syntax. Due to its agglutinative and strict head-finality, such verb raising, if it exists
at all, does not show any overt effects on the surface order of elements. In this paper, we examine the distribution of Negative Concord Items (NCI) in Non-Constituent Coordination
(NCC) in Japanese. We argue against Koizumi’s (2000) verb-raising approach to NCI, since it
makes an incorrect prediction on the distribution of NCIs inside coordinate structure. We conclude that it should be analyzed under the gapping approach proposed by Fukui and Sakai
(2003). First, we introduce some seemingly problematic data for the gapping analysis, which
turn out to be a counterexample against the verb-raising analysis. Following Kato’s (2006)
representational approach to the Coordinate Structure Constraint (CSC), we argue that the
gapping approach correctly captures the distribution of NCIs, while the verb-raising analysis
reaches a dead end. The conclusion of this paper supports Fukui and Sakai’s argument that
there is no string-vacuous overt verb raising in Japanese.
1. Introduction: NCIs in NCC
Since Koizumi (2000) it has been hotly debated whether Japanese has syntactic verb raising,
which must be string-vacuous since Japanese is a strictly head-final language. In NonConstituent Coordination (NCC; 1a), Koizumi assumes that verb-to-Tense-to-Comp (V-T-C)
raising occurs in Narrow Syntax, as in (1b). On the other hand, Fukui and Sakai (2003) argue
that gapping occurs in the first conjunct in deriving NCC with no recourse to syntactic verb
raising, as in (1c). This section aims to point out that the verb-raising approach encounters a
problem concerning the licensing of Negative Concord/Polarity Items in NCC. When NCC
contains XP-sika ‘XP-except,’ it becomes ungrammatical, as in (2). The sentence in (2) can
be schematically represented under Koizumi’s verb-raising approach as (3a), and under the
gapping analysis as (3b).
Proceedings of ConSOLE XXIV, 2016, 254-262
http://www.sole.leidenuniv.nl
© Ryoichiro Kobayashi and Taihei Asada
NCIs in coordinate structure
(1)
255
a. George-wa
kinoo
ringo-o
mit-tu to
George-TOP yesterday apple-ACC 3-CL
CONJ
Mary-wa
kyoo
banana-o
ni-hon tabe-ta
Mary-TOP
today
banana-ΑCC 2-CL
eat-PAST
‘George ate three apples yesterday and Mary ate two bananas today.’
b. [[[George … [VP ... tV] tT] & [Mary … [VP ... tV] tT]] V-T-C]
|___|_________________|___|____↑
c. [[[George … [VP … V] T] & [Mary … [VP … V] T]] C]
(2)
*George-wa
ringo-sika
kinoo
to
George-TOP
apple-except
yesterday CONJ
Mary-wa
banana-sika
kyoo
tabe-nakat-ta
Mary-TOP
banana-except today
eat-NEG-PAST
Intended: ‘George only (ate) an apple yesterday, and Mary only ate a banana today.’
(3)
a. [TP SU [NegP NCI [VP ... tV] tNEG] tT] & [TP SU [NegP NCI [VP ... tV] tNEG] tT] V-NEG-T
|___|___|_______________________|___|___|____↑
b. [TP SU [NegP NCI [VP … V] NEG] T] & [TP SU [NegP NCI [VP … V] NEG] T]
It has been observed that the licensing of XP-sika requires a one-to-one relation; that is, there
cannot be more than one XP-sika for each NEG, as in (4).
(4)
*Taro-sika
ringo-sika
tabe-nakat-ta.
Taro-except apple-except eat-NEG-PAST
‘Only Taro only ate apples.’
(Kato 1985:155)
In (2), although NEG does not have its phonological content in the first conjunct, it should be
interpreted in-situ in LF, licensing the XP-sika in the first conjunct. Therefore, on the surface,
it seems that there is an apparent problem with the gapping approach, since it may predict
grammaticality for what should be ungrammatical. On the other hand, if we take the verbraising analysis of Koizumi, there will be only one NEG, which cannot license two XP-sikas,
as illustrated in (4). Given these observations, data such as (2) appear problematic to Fukui
and Sakai’s gapping approach. We argue, however, this is not the case. Rather, the verbraising approach makes incorrect predictions if we closely investigate the peculiar nature of
XP-sika in Japanese.
The rest of the paper is organized as follows. Section 2 reviews Kato’s (2006) analysis
of a coordinate structure that we largely follow in this paper. In section 3, the nature of two
types of NCIs, wh-mo and XP-sika, are presented, followed by our proposals for the gapping
analysis of NCC in section 4 and 5. We argue that the gapping approach correctly captures the
distribution of NCIs, while the verb-raising analysis fails. Section 6 is the overall summary of
the contributions of this paper.
Ryoichiro Kobayashi and Taihei Asada
256
2. The component structures and NCC
Kato (2006) convincingly shows that the Coordinate Structure Constraint (CSC) is a representational constraint that must be satisfied in LF. To briefly summarize his discussion, a coordinate structure is well-formed only when each of the component structures satisfies LFrequirements of grammaticality. The component structure is defined as a structure composed
of one of the conjuncts and the material that is not included in the coordinate structure (Kato
2006:3). A typical CSC-violating example can be found in (5), whose ungrammaticality is
explained under Kato’s proposal: Sentences such as (5) become ungrammatical since one of
the component structures - (6) - violates the LF-constraints on vacuous quantification.
(5)
*Whati did Mary [send ti on Monday] and [receive the parcel on Wednesday]?
(6)
Component structures of (5):
a. what did Mary send t on Monday
b. *what did Mary receive the parcel on Wednesday
(Kato 2006:3)
Given that the licensing of NCIs is an LF phenomenon, a natural suspect for the source of
ungrammaticality of NCI coordinate items is its component structure. Under Koizumi’s
(2000) verb-raising approach, V-NEG-T amalgams are created due to the Head Movement
Constraint. Since the V-NEG-T, tabe-nakat-ta ‘eat-NEG-PAST’ in C (or in T) is shared by each
conjunct, the relevant component structures will be something like (7).
(7)
Component structures of (2) under the verb-raising approach (tabe-nakat-ta is shared):
a. George-wa
ringo-sika
kinoo
tabe-nakat-ta
George-TOP apple-except yesterday eat-NEG-PAST
‘George only ate apples.’
b. Mary-wa
banana-sika
kyoo
Mary-TOP
banana-except today
‘Mary only ate bananas.’
tabe-nakat-ta
eat-NEG-PAST
(7) predicts (2) to be grammatical, which is not the case. I propose, on the other hand, that the
gapping approach correctly explains the relevant ungrammaticality of (2) above.
3. Two types of NCIs in Japanese
Before we turn to the gapping analysis on NCC, let us briefly review two well-established
facts concerning NCIs in Japanese. There is another type of NCI, which consists of wh plus
the focus particle -mo. They cannot appear without NEG that agrees with its [neg] feature.
NCIs in coordinate structure
(8)
257
a. *Dare-mo
ki-ta.
WH-FOC
come-PAST
Lit. ‘Anyone came’
b. Dare-mo
ko-nakat-ta.
WH-FOC
come-NEG-PAST
‘No one came.’
Kuno (2008) proposes that NCIs are licensed via agreement of [neg] features on it and
clausemate NEG, as illustrated in (9).
(9)
a. … NEGprobe …
NCI
[neg]
[foc][neg][one]
|____________↑Agree
b. … NEG …
NCI
[neg]
[foc][neg][one]
↑the locus of Negation
Interestingly, WH-mo-type NCIs such as dare-mo ‘anyone’ and nani-mo ‘anything’ can be
licensed in an elliptical fragment answer, which is one of the diagnostics of NCIs (Vallduví
1994).1
(10) Q: Nani-o
mi-ta-no?
WH-ACC
see-PAST-Q
‘What did you see?’
A: Nani-mo
WH-FOC[NCI]
‘Nothing.’
mi-nakat-ta
see-NEG-PAST
cf. Q: What did you see?
A: *Anything[NPI].
(Adapted from Watanabe 2004:564)
Why is an elliptical answer possible even though there is no negative antecedent mi-nakat-ta
‘didn’t see’ in the preceding context? Kuno (2008) argues, following Giannakidou (2006),
that the semantics of questions provide a solution: A question denotes the set of its true answers that also contains a negative proposition (Karttunen 1977). Therefore, an elliptical answer can be derived from its non-elided answer counterpart in the domain of true answers.
1
The basic diagnostics for NCIs (Vallduví 1994, Giannakidou 2006 and others):
a. Ability to appear in nonnegative contexts
b.Ability to appear in preverbal position
c. Ability to be modified by expressions like almost
d.Ability to be used as an elliptical answer
e.Clause-boundedness
Ryoichiro Kobayashi and Taihei Asada
258
Some may say that this line of argument may not distinguish NCIs from Negative Polarity
Items, which cannot be licensed in elliptical answers. Kuno (2008:209) notes that Giannakidou’s (2006) analysis is compatible with the fact that elliptical answers are not possible with
Negative Polarity Items. Since a Negative Polarity Item must move out of the elided constituent, it ends up outside the c-commanded domain of NEG, while the NCI can be licensed due
to its intrinsically negative nature.
(11) a. What did you see?
b. *Anything.
c. *Anything [I didn’t see anything].
↑________________|
In the next section, we move on to our proposal, which is compatible with, and lends credence
to Fukui and Sakai’s (2003) gapping analysis.
4. Proposal
Interestingly, the wh-mo type of NCI may readily be licensed in similar constructions to (2),
as in (12).2
(12) [Taro-ga
nani-mo
mik-kakan]
to
Taro-NOM
WH-FOC
3-days
CONJ
[Hanako-ga
nani-mo
yok-kakan]
tabe-nakat-ta
Hanako-NOM WH-FOC
4-days
eat-NEG-PAST
‘Taro (didn’t eat) anything for 3 days, and Hanako didn’t eat anything for 4 days.’
We propose that the relevant acceptability arises from the wh-mo as an NCI: While wh-mo
can be licensed through concordance of [neg] between wh-mo and NEG, XP-sika needs to be
licensed under the scope of NEG. The prediction is borne out that XP-sika cannot be licensed
or become marginal at the very least as an elliptical answer.
(13) Q: Dare-ga
WH-NOM
‘Who came?’
ki-ta-no?
come-PAST-Q
A: *?Taro-sika
ko-nakat-ta.3
Taro-except come-NEG-PAST
‘Only Taro.’
2
The reviewer pointed out that sentences like (12) still need a long pause between conjuncts. Moreover, it
was questioned as to whether the readers parse the sentence as a Right-Node-Raising like sentence without -to,
which is perfectly grammatical in Japanese. We agree that some pause makes the sentence sound better. However, there is no way to test whether (12) is parsed without -to. I assume that (12) is a genuine NCC for argument’s
sake.
3
The same reviewer noted that adding certain lexical items such as zannennagara ‘unfortunately’ improves
the sentence. We agree with this judgment, but such words contain somewhat negative flavor. Therefore, we
believe that it does not affect the discussions in this paper.
NCIs in coordinate structure
259
A: Taro-sika
ko-nakat-ta.
T.-except
come-NEG-PAST
‘Only Taro came.’
XP-sika cannot be licensed under elided NEG. In (14), although the elliptical answer has its
negative antecedent, the fragment answer is ungrammatical.
(14) Q: Dare-ga
ko-nakat-ta-no?
WH-NOM
come-NEG-PAST-Q
‘Who didn’t came?’
A: *Taro-sika
ko-nakat-ta.
Taro-except come-NEG-PAST
‘Only Taro.’
We assume that XP-sika shows parallelism between English Negative Polarity Items like anybody, which can never be licensed in an elliptical answer. Note that in Japanese, XP-sika can
be licensed in the subject position (Kato 1985).
(15) a. Dare-ga
ki-ta-no?
WH-NOM
come-PAST-Q
‘Who came?’
b. *?Taro-sika.
Taro-except
‘Only Taro.’
c. *Taro-sika [Taro-sika ko-nakat-ta.]
↑_________|
Given these observations, we argue that XP-sika itself is not negative in nature, following
Kuno (2011) and others. Although some classify XP-sika as a NCI since it cannot be licensed
without syntactic NEG, we assume that wh-mo and XP-sika minimally differ in that the latter
always requires overtly pronounced NEG. It automatically follows that XP-sika cannot be
rescued in an elliptical fragment answer in (15), while wh-mo type NCIs can be, as in (16).
(16) Nani-o
mi-ta-no?
WH-ACC
see-PAST-Q
‘What did you see?’
a. okNani-mo.
WH-FOC
‘Nothing.’
↓ ̄ ̄ ̄ ̄ ̄|
licensed via agreement in (9)
b. Nani-mo [Nani-mo[neg] mi-nakat[neg]-ta].
ok
Ryoichiro Kobayashi and Taihei Asada
260
↑_________|
In the next section, we further show that the proposal above correctly captures the distribution
of XP-sika and wh-mo type NCIs in a coordinate structure.
5. Focus movement of XP-sika and the component structure
Let us return to the data in (1), repeated here as (17). The schematic representation would be
something like (18) under the verb-raising approach. There is no gapping in the first conjunct.
(17) *George-wa
ringo-sika
kinoo
to
George-TOP apple-except
yesterday
CONJ
Mary-wa
banana-sika
kyoo
tabe-nakat-ta
Mary-TOP
banana-except
today
eat-NEG-PAST
‘George only (ate) an apple yesterday, and Mary only ate a banana today.’
(18) [TP SU [NegP XP-sika [VP... tV] tNEG] tT]&[TP SU [NegP XP-sika [VP... tV] tNEG] tT] V-NEG-T
|___|___|_________________________|___|___|____↑
We have already noted that Koizumi’s (2000) analysis incorrectly predicts that sentences like
(17) become acceptable since the component structures in (19) both satisfy grammatical constraints.
(19) Component structures of (17) under the verb-raising approach:
a. George-wa
ringo-sika
kinoo
tabe-nakat-ta
George-TOP apple-except
yesterday
eat-NEG-PAST
‘George ate only apples yesterday.’
b.Mary-wa
banana-sika
kyoo
Mary-TOP
banana-except yesterday
‘Mary ate only bananas yesterday.’
tabe-nakat-ta
eat-NEG-PAST
On the other hand, the gapping analysis fully explains the deviance of (17) with XP-sika. It is
schematically represented in (20), where gapping occurs in the first conjunct. In order for the
ellipsis of identical V, NEG and T to occur, the elements inside TP, NegP and VP, including
the subject and XP-sika, must be extracted from them, precisely what occurs in elliptical answers with regular NCIs. The ungrammaticality stems from XP-sika being unlicensed outside
the scope of NEG in the first conjunct. The component structures of (20) are exemplified in
(21). 4
4
It does not affect the argument here even if we assume that XP-sika cannot be licensed by covert NEG that
is elided, say without [focus] features, while wh+mo can. The component structures of (2) under this line of
argument would be (i).
(i) Component structures of (2) under the gapping approach:
a. *George-wa ringo-sika
kinoo
tabe-nakat[covert]-ta
George-TOP apple-except yesterday eat-NEG-PAST
Intended: ‘George ate only apples yesterday.’
b. Mary-wa
banana-sika
kyoo
tabe-nakat-ta
Mary-TOP
banana-except today
eat-NEG-PAST
NCIs in coordinate structure
261
(20) [SU XP-sika α [TP SU [NegP XP-sika [VP α V] NEG] T]] &
[SU [NegP XP-sika [VP…V] NEG] T]]
(21) a. *Georgei-wa
ringo-sikaj
kinook
[TP ti [NegP tj [VP tk
George-TOP
apple-except
yesterday
Intended: ‘George ate only apples yesterday.’
b. Mary-wa [NegP banana-sika
kyoo
Mary-TOP
banana-except today
Intended: ‘Mary ate only bananas today.’
tabe-nakat-ta]]]
eat-NEG-PAST
tabe-nakat-ta]
eat-NEG-PAST
Next, let us re-examine the data in (12), repeated below as (22), which can be schematically
represented as (23). Just like the other NCCs, gapping may occur in the first conjunct under
identity. Although wh-mo ends up outside the scope of NEG in the first conjunct, the NCI can
be licensed via agreement with a [neg] feature on NEG.
(22) [Taro-ga
nani-mo
mik-kakan] to
Taro-NOM WH-FOC
3-days
CONJ
[Hanako-ga nani-mo
yok-kakan] tabe-nakat-ta
Hanako-NOM WH-FOC
4-days
eat-NEG-PAST
‘Taro (didn’t eat) anything for 3 days, and Hanako didn’t eat anything for 4 days.’
(23) [SU wh+mo α [TP SU [NegP wh+mo[neg] [VP α V] NEG[neg]] T]] &
[SU [NegP wh+mo[neg] [VP…V] NEG[neg]] T]]
(24) Component structures of (22) under the gapping approach:
a. Taroi-ga
nani-moj mik-kakank [TP ti [NegP tj [VP tk tabe-nakat-ta]]]
Taro-NOM
WH-FOC
3-days
eat-NEG-PAST
‘Taro didn’t eat anything for 3 days.’
b. Hanako-ga
[NegP nani-mo
yok-ka-kan
Hanako-NOM
WH-FOC
4-days
‘Hanako didn’t eat anything for 4 days.’
tabe-nakat-ta]
eat-NEG-PAST
Licensing of wh-mo is done via agreement of [neg] features between the NCI and NEG; hence
the relevant NCI, wh-mo, need not remain in-situ within the scope of NEG. Therefore, the
gapping approach correctly predicts that sentences like (22) become grammatical. Whatever
analysis on Negative Concord we take, this is empirically clear that wh-mo does not necessarily need overt NEG, as long as the sentence satisfies prerequisites for ellipsis, as in (24).
6. Conclusions
‘Mary ate only bananas yesterday.’
Ryoichiro Kobayashi and Taihei Asada
262
The discussions above support Fukui and Sakai’s (2003) gapping approach to NCC: It shows
that Koizumi’s (2000) verb-raising analysis fails to explain the distribution of these NCIs in
Japanese since it does not distinguish the component structures of grammatical and ungrammatical NCC. Although the discussions in this paper have left open many important issues,
we hope to have shown that the existence of syntactic verb raising in Japanese NCC is suspicious. Those who support the verb-raising analysis must at least take into consideration the
component structure analysis of a coordinate structure.
Acknowledgements
We would like to thank Takaomi Kato for the valuable comments on the earlier versions of
this paper. We wish to thank the reviewer, the organizers of and audience at the ConSOLE 24
in York. All remaining errors and inadequacies are of course ours. The first author is supported by Grant-in-Aid for JSPS Fellows 16J00637.
Ryoichiro Kobayashi1,2 and Taihei Asada1
Graduate School of Languages and Linguistics,
Sophia University1
Japan Society for the Promotion of Science2
Corresponding Author (Kobayashi):
[email protected]
https://sites.google.com/site/ryoichirokobayashi/
References
Fukui, N. & H. Sakai. (2003). The visibility guideline for functional categories: Verb raising in Japanese and
related issues. Lingua, 113:4, pp. 321-375.
Giannakidou, A. (2006). N-words and negative concord. Everaert, M & H. van Riemsdijk (eds.), The syntax
companion 3, Blackwell Publishing, Oxford, pp. 327-391.
Karttunen, L. (1977). Syntax and semantics of questions. Linguistics and philosophy 1, pp. 3-44.
Kato, T. (2006). Symmetries in coordination. [Ph.D thesis], Harvard University, Cambridge, Mass.
Kato, Y. (1985). Negative sentences in Japanese. Sophia University, Tokyo.
Koizumi, M. (2000). String vacuous overt verb raising. Journal of East Asian Linguistics, 9:3, pp. 227-285.
Kuno, M. (2008). Negation, focus, and negative concord in Japanese. Toronto Working Papers in Linguistics 28,
pp. 195-211.
Kuno, M. (2011). Separating Feature Interpretability from Feature Values: Evidence from Negative Concord in
Japanese. Baunaz, L, Bentea, A & J. Blochowiak (eds.), Proceedings of Swiss Workshop in Generative
Grammar: Generative Grammar in Geneva 7, Geneva, pp. 23-32.
Vallduví, E. (1994). Polarity items, n-words, and minimizers in Catalan and Spanish. HCRC Publications, University of Edinburgh.
Watanabe, A. (2004). The genesis of negative concord: Syntax and morphology of negative doubling. Linguistic
inquiry 35:4, pp. 559-612.