Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Tomohide KINUHATA
Osaka University
1 Introduction
Kataoka (2006a, 2006b) proposes the following structural relationship held be-
tween XP-sika and negation.
(1) Kataoka’s proposal
XP-sika must be in NegP-Spec at LF. Kataoka (2006a: 229)
Sika is one of the most discussed NPIs in the literatures on Japanese which needs
one negation for one occurence of it in the same clause (Oyakawa 1975).
(2) Each sika must co-occur at least with one negation.
a. Taro-wa manga-sika yoma-nai
Taro-Top comic-sika read-Neg
‘Taro does not read any books other than comics.’
b. ∗ Taro-wa manga-sika yomu.
Taro-Top comic-sika read
∗
c. Taro-sika manga-sika yoma-nai.
Taro-sika comic-sika read-Neg
Kataoka’s proposal nicely explains those properties of sika since Spec is in one
to one correspondence with its head, obviously, in the same clause. 1
In addition, her analysis can account for the scope of negation relative to the
QPs occuring in the sentence with sika. Consider the following:
Assuming that the subject asymmetrically c-commands the object (see Hoji
2003 and the referencs therein), the XP-sika in (4a) c-commands the QP whereas
in (4b) the QP c-commands the XP-sika. Since XP-sika is in the Spec of negation
in Kataoka’s proposal, the negation takes scope over the QP in (4a) whereas the
QP outscopes the negation in (4b). The relavant structures are given in (5).
As the structures in (5) indicate, (4a) can only mean (6a) where the negation
takes wider scope than the QP and (4b) only means (7b) where the QP is wider
than the negation. 2 To see that (4a) does not have the truth condion in (6b),
it suffices to consider the situation in which someone other than Taro ate more
1
Kataoka (2006a, 2006b) uses the term Spec-head only to capture the structural
relationship between XP-sika and negation, and does not claim that those have
features checked by agreement contra Aoyagi and Ishii (1994). Kataoka (2006a: fn
10) notes ‘(t)he important point of my proposal, however, is not whether there is
some kind of agreement or not, but that, at LF, XP-sika must c-command and be
close enough to Neg just like the structural relation between Spec and head.’
2
The subject position can be in the scope of negation in Japanese as well as in English.
i) 3-nin-izyoo-no kyooju-ga Taro-o suisensi-nakat-tara, Taro-wa
3-CL-more.than-Gen professor-Nom Taro-Acc recommend-Neg-if, Taro-Top
syuusyoku deki-nai
get.job can-not.
‘If more than three professors does not recommend Taro, Taro cannot get a job.’
ok
Neg>QP
than three bowls of noodles. (4a) is clearly false in this situation but (6b) is
not. In the same way, consider the situation where there are five professors who
recommended everyone but Taro. This situation makes (7a) false, but (4b) is
true if there are other five professors who recommended only Taro.
This paper argues from a semantic viewpoint why XP-sika and negation
seem to occupy the closest position at LF as in (5). This question might not be
taken up seriously by researchers working in syntax. Because there is a plenty
of phenomena which can be explained by agreement between Spec and head,
e.g. ‘Negative Concord’ (Haegeman 1995)3 , and sika and negation might be
considered as one instance of this phenomena (cf. Aoyagi and Ishii 1994).
However, as suggested by Fukui (1986) and Kuroda (1988), Japanese might
lack functinal categories which constitute the basis of feature agreement. Thus
I believe it to be worth pursuing whether or not a semantic account is viable.
Since syntactic features are highly abstract constructs, agreement based accounts
should be taken into consideration after going through intuitive accounts such
as those based on lexical meanings.
This paper is organized as follows. In section 2, I will look at Kataoka (2007)’s
explanation for the syntactic position of XP-sika and negation. Her account
is rather semantic but cannot be maintained. Section 3 provides new data on
dake [only] and igai [except], which are parallel with the data of sika, and are
closely related to the meaning of it. In section 4, I will give the semantics of sika
capturing the parallels of it with dake and igai. Section 5 concludes the paper.
Though my argument is indirect in that it relies on the correlations found in
different lexical items, they provide us with an empirical ground on which we
should give those items similar semantics, which explains the scope phenomenon
of the negation of sika without syntactic features.
Rejecting the view that XP-sika and negation have agreement features, Kataoka
(2007) considers that they form a ‘Subject-Predicate structure’ in the sense
of Kuroda (1992). I argue, however, in this section that XP-sika and negation
3
But see Giannakidou (2000) for the explanation of negative concord which does not
assume any features.
need not form a Subject-Predicate structure, and even if they constitute that
structure, negation does not necessarily take wider scope than the QP occurring
after the subject.
Referring to the Ueyama (1998)’s observation that the clause expressing
‘predicational propositoins’ cannot be embedded in the complement of percep-
tual verbs, Kataoka (2007) claims that the following data suggest that XP-sika
and negation constitute the Subject-Predicate structure.
(9) Context: Hanako said to Taro that she buys three types of meats for the
dinner: chiken, pork and beef. But in a grocery store,
Taro-ni-wa [Hanako-ga gyuuniku-sika tora-nai-no]-ga mieta.
Taro-Dat-Top Hanako-Nom beef-sika take-Neg-FN-Nom saw
‘Taro happened to see that Hanako does not take any meats but beef.’
I made explicit the event which Hanako perceived in (9): that Hanako passed the
corner of chicken and pork without taking them. This shows that XP-sika and
negation does not necessarily constitute the Subject-Predicate structure, since
the sentence containing sika can denote an event, which the Subject-Predicate
structure cannot denote (Kuroda (1992), Ueyama (1998), Kataoka (2007)).5
4
The same oddity can be found with the sentences in which sika is replaced by igai
which is not an NPI as we will see in section 3.2.
i) ♯ Taro-ni-wa [Hanako-ga piano-igai-o hika-nai-no]-ga mieta (kikoeta).
‘Taro-Dat-Top Hanako-Nom piano-sika play-Neg-FN-Nom saw (heard)’
♯
‘Taro happened to {see/hear} that Hanako did not play any instruments but
piano.’
5
Another evidence that XP-sika and the predicate do not constitute the Subject-
Predicate structure is that the XP-sika does not have to occupy the sentence initial
positions as opposed to the typical subject marker wa. XP-wa which occupies non-
sentence-initial positions is called huku-joshi in Kuroda (1969-70) and is not regarded
as a Subject.
A more serious problem of Kataoka (2007)’s analysis is that even if XP-sika
and negation constitute the Subject-Predicate structure, it does not lead to the
conclusion that the negation appearing in that sentence occupies the closest
position to XP-sika. That is, the following two options are still available for the
structural relation between the XP-sika and the negation.
When XP-wa, a typical subject marker, occupies the sentence initial positions,
the negation does or does not scope over the QP in the object position. Thus
analyzing XP-sika as a subject still leaves the question why the LF associated
with the sentence is not (10b) but (10a).
I will show in the next section that the data presented as illustrations of the
property of sika can be observed in other lexical items, namely dake and igai.
3 Implicit Negation
The sentence with dake [only] can convey the same truth conditional meaning
as the sentence with sika. (11a) and (11b) both mean that Taro reads comics
(henceforth, ‘positive component’) and he does not read other genres of literature
(‘negative component’).
Although it is still under debate what status those components have (see Beaver
and Clark (2008) for the survey of discussions), it is ascertained that they are lex-
ically induced uncancelable meanings, since neither ‘Taro does not read comics’
nor ‘Taro also reads novels’ can follow those sentences.
Given the truth conditional meanings of sika and dake, readers may notice
that dake as well as sika includes negation in its meaning though the negation
is not pronounced. What I want to show here is that the negation involved in
the meaning of dake behaves semantically in the same way as the pronounced
negation in the sentence with sika. In (12), I repeat the examples of sika from
(4), and (13) is the dake-counterparts of (12), both with same truth conditional
meanings.
In the basic SOV order, when XP-sika and XP-dake precede QPs, as in (12a)
and (13a), only the interpretation in which the negation, pronounced or not
pronounced, takes wider scope than the QP, i.e. (14a), is possible: the situation
where someone other than Taro ate more three bowls of noodles, under which
(14b) can still be true, makes (12a) and (13a) false.
(14) Interpretation of (12a) and (13a)
a. Everyone but Taro ate less than three bowls of noodles. Neg>QP
b. ∗ For more than three bowls of noodles, everyone but Taro did not eat
it. QP>Neg
On the other hand, when a QP precedes an XP-sika and XP-dake as in (12b)
and (13b), the QP must take wider scope than the negation as in (15b): when
just two professors recommended only Taro, (12b) and (13b) are obviously false
but (15a) is not.
(15) Interpretation of (12b) and (13b)
a. ∗ Less than three professors recommended everyone but Taro. Neg>QP
b. For more than three professors, he did not recommend anyone but
Taro. QP>Neg
Not only in the basic word order but also in the scrambled sentence do sika
and dake have the same scope interpretations. It is acknowledged that when the
object is scrambled, it can occupy both higher and lower positions relative to
the subject (Ueyama 1998). That is, the PF in (16a) has two LFs such as (16b)
and (16c).
When an XP-sika is in the subject position, since the negation occupies the clos-
est position to it (Kataoka’s proposal), the QP in the object position can take
higher and lower positions than the negation. These structures give the scram-
bled sentence two possible interpretations, i.e. QP>Neg in LF1 and Neg>QP in
LF2. This ambiguity also arises when an XP-dake occupies the subject position
without a pronounced negation. Consider the scrambled counterparts of (12a)
and (13a) below.
We saw in the previous subsection that the negation involved in the semantic
interpretation of dake behaves in the same way as the pronounced negation in
the sentences with sika. This subsection turns to another lexical item which
introduces negation in its interpretation but lacks the phonetic equivalent.
Igai, meaning ‘except’ or ‘other than’, can be used with or without negation.
When used with negation as in (19a), the sentence seems to convey the same
meaning as the sentence with sika such as (2a). However, the scope restriction
of QPs and negation observed in sentences with sika do not hold for igai.
(20) a. Taro-igai-ga 3-bai-izyoo-no raamen-o tabe-nakat-ta.
Taro-igai-Nom 3-CL-more.than-Gen noodle-Acc eat-Neg-Past
‘Everyone but Taro didn’t eat more than three bowls of noodles’
b. 3-nin-izyoo-no kyoozyu-ga Taro-igai-o
3-CL-more.than-Gen professor-Nom Taro-igai-Acc
suisensi-nakat-ta.
recommend-Neg-Past.
‘More than three professors didn’t recommend everyone but Taro.’
For example (20a) can mean ‘everyone but Taro ate less than three kinds of
noodles’ (Neg>QP) and ‘for more than three kinds of noodles, everyone but
Taro did not eat it’ (QP>Neg).6 This means that the negations in (20) are not
necessarily in the closest position to XP-igai contrary to the negation introduced
by sika.
However, an interpretation of the sentence with igai can induce the negation
which works in the way the negation of sika does. In order to illustrate this, let
us use the sentence without a pronounced negation.
(21a) has the meaning given in the translation and, at the same time, can imply
that Taro did not eat more than three bowls of noodles. But the implication
must be the one where the negation scopes over the QP. That is, (22a) is the
possible implication of (21a) but (ex1b) is not.
Those facts coincide with the data of sika and dake: when sika, dake and igai,
attaching to NP, take higher positions than a QP, the negation must scope over
the QP and when those resides in lower positions than a QP, the QP must take
wider scope than the negation.
When the object is scrambled, the implication becomes ambiguous depending
on the LF associated with it. Thus (24a) can imply both (22a) and (22b), and
(24b) can imply both (23a) and (23b).
The rise of ambiguity here parallels with the cases we found in the scrambled
sentences including sika and dake.
It should be noted, however, that the meaning of igai is not the same as
the meaning encoded by dake and sika. We considered the negative and positive
component of sika and dake as lexically induced meanings, showing that they
cannot be canceled in usual contexts. On the other hand, the meanings described
as (22a) and (23b) are neither assertions nor presuppositions: the meanings given
in (22a) and (23b) can be canceled (though the cancelation is less natural when
universal quantifiers are interpreted to quantify over the rest of the people as we
will discuss).
Those facts indicate that the negative component of igai is conversational im-
plicatures. The unpronounced negation is not involved in the pragmatic inter-
pretation when the implicatures are canceled. However, when that meanings are
not canceled, the negation must be interpreted contingent upon the position of
XP-igai.
4 Explaining Correlation
The reasoning behind the inferences indicated by ‘;’ in (26) is stated as: if it is
unnecessary to extract an element from the set, you should not extract it. This
means that the extracted elements are to be the smallest sets, which, ‘Smallest
Extraction Condition’, is formulated as in (27).
In order to see the condition clearly, consider the situation in (28) and compare
the appropriate sentences in (26) and the inappropriate ones in (29). Note that
all sentences in (26) and (29) are true under the situation (28).
In addition, dake and sika divide the relevant group into two and only two
groups: in other words, they extract the elements from the contextually rele-
vant set with the universal determiner taking the remaining members as the
restriction. Thus an extraction condition is imposed which in this case says: if
it is necessary to extract an element, you should extract it. This requires the
exhaustive extraction of elements if necessary, which is expressed as:
The condition in (35) leads to the interpretation in which all remaining mem-
bers have the property ¬P .
(36) ∀S[S ⊆ (P ∩ A) → S ⊆ C]
=(P ∩ A) ⊆ C
=(A − C) ⊆ ¬P
(38) ensures the scopal property of negation observed in (4), (17a) and (18a), in
stead of (1).
In this section, we have discussed that the conditions on extraction introduce
not only the negation of igai but also that of dake and sika. Those negations are
regarded as coming out of the same womb, since the exhaustive condition and the
smallest extraction condition constitute the necessary and sufficient condition for
extraction (with universals) respectively.
5 Conclusion
Appendix: Interpretation
In this paper, I assume that quantifiers are interpreted in-situ and negation works
as boolean complements, not sentential negation. To illustrate the feasiblity of
this idea, I use the tools from Keenan and Faltz (1985) and Keenan (1988).
Analysis Tree (4b)
(4a) 3more P rof.(taro sika(x recom)):P0
taro sika(3more noodle(x eat)):P0
Aoyagi, Hiroshi, Ishii, Toru, 1994. On NPI licensing in Japanese. In: Akatsuka,
N. (Ed.), Japanese/Korean Linguistics 4. CSLI, Stanford, pp. 295–311.
Beaver, David, Clark, Brady, 2008. Sense and Sensitivity: How Focus Determines
meaning. Blackwell, Cambridge.
Fukui, Naoki, 1986. A theory of category projection and its aplications. Ph.D.
thesis.
Giannakidou, Anastasia, 2000. Negative... concord? Natural Language and Lin-
guistic Theory 18, 457?523.
Haegeman, Lilian, 1995. The Syntax of Negation. Cambridge University Press,
Cambridge.
Hoji, Hajime, 2003. Falsifiability and repeatability in generative grammar: A
case study of anaphora and scope dependency in Japanese. Lingua 113 (4-6),
377–446.
Kataoka, Kiyoko, 2006a. ‘Neg-sensitive’ elements, neg-c-command and scram-
bling in Japanese. In: Timothy, J., Jones, K. (Eds.), Japanese/Korean Lin-
guistics 14. CSLI, Stanford, pp. 221–233.
Kataoka, Kiyoko, 2006b. Nihongo Hiteibun no Kōzō: Kakimazebun to
Hiteikoōhyōgen [The structure of negative sentences in Japanese: Scrambling
sentences and Neg-sensitive expressions]. Kuroshio Publisher.
Kataoka, Kiyoko, 2007. Gendainihongo sika no hurumai to tōgotekijōken [syn-
tactic behavior of -sika in Japaneseand its syntactive condition]. In: Nihon-
gengogakkai dai135kai Yokoosyuu. Nihongengogakkai.
Keenan, Edward, 1988. On semantics and the binding theory. In: Hawkins, J.
(Ed.), Explaining Language Universals. Blackwell, Cambridge, pp. 105–144.
Keenan, Edward, Faltz, Leonard, 1985. Boolean semantics for natural language.
D. Reidel Publishing Company, Dordrecht.
Kuroda, S.-Y., 1969-70. Remarks on the notion of subject with reference to words
like also, even, and only. In: Annual Bulletin, Research Institute of Logopedics
and Phoniatrics 3-4. University of Tokyo, Tokyo, pp. 111–129, 127–152.
Kuroda, S.-Y., 1988. Whether we agree or not: A comparative syntax of English
and Japanese. Linguisticae Investigationes 12, 2–44.
Kuroda, S.-Y., 1992. Judgment forms and sentence forms. In: Japanese syntax
and semantics: Collected papers. Kluwer Academic Publisher, Dordrecht, pp.
13–77.
Oyakawa, Takatsugu, 1975. On the Japanese sika-nai construction. Gengo
Kenkyu 67, 1–20.
Ueyama, Ayumi, 1998. Two types of dependency. Ph.D. thesis.
von Fintel, Kai, 1993. Exceptive constructions. Natural Language Semantics 1,
123–148.