11112313

100%

LIBRARY
Michigan State

 

University

—-<-. - -

 

This is to certify that the
thesis entitled

CONSTRUCTING CONCESSIVE CONDITIONALS
IN JAPANESE

presented by

Al MATSUI

has been accepted towards fulfillment
of the requirements for the
MA. degree in Linguistics and
Germanic, Slavic, Asian and

African Languages

 

fit/lajor' Professor’s Signature
4 1/ 3 0/0 8”

Date

MSU is an afi‘innative-action, equal-opportunity employer

—-__---——

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:/Prolecc&Pres/ClRCIDateDue indd

CONSTRUCTING CONCESSIVE CONDITIONALS IN JAPANESE
By
Ai Matsui

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Department of Linguistics and
Germanic, Slavic, Asian and African Languages

2008

ABSTRACT
CONSTRUCTING CONCESSIVE CONDITIONALS IN JAPANESE
By
Ai Matsui

This thesis investigates the syntactic, semantic and pragmatic nature of concessive
conditionals, e.g. even if sentences in English. Since Karttunen & Peters (1979), the
word even has been viewed as an element that adds an existential and scalar presup—
position. The existing analyses of even if sentences (e.g. Bennett (1982), Guerzoni
& Lim (2007)) based on this view are not directly applicable to the Japanese con-
cessive conditional which has no conditional morpheme but consists of a gerundive
clause and the additive particle m0. In my compositional analysis, the concessive
conditional meaning is build up from the gerundive construction, which is a clausal
conjunction ‘and’ that also reflects the speaker’s knowledge about the normal situa-
tions, and mo that measures and compares the degree of expectedness derived from
the gerundive construction. This study may also shed light on theoretical connections
among conditionals, modals, and questions, as well as some issues of semantics and

pragmatics of clausal conjunctions.

ACKNOWLEGDMENT

This thesis originally developed from a piece of work submitted as a term paper for
LIN 837: Advanced Studies in Semantics and Pragmatics offered in Fall Semester 2006 by
Professor Marcin Mozycki. I would like to express my sincere appreciation to Professor
Morzycki for being helpful and patient not only during the time I took the course, but also
from the very beginning of my learning of formal semantics. I thank Professor Mutsuko
Endo—Hudson for guiding me to a better understanding of J apanesedata. I am also thankful
to Professor Cristina Schmitt for the comments that helped to clarify my argument.

My thanks go as well to the graduate students of the linguistics program at MSU for
talking with me about English data and for their encouragement. I especially owe a lot to
Hsiang-Hua Chang, David Hasty, and Walter Sistrunk for suggestions and corrections of
my drafts.

I thank my family for their warm support from overseas.

Remaining oddities in the thesis, if any, are of course mine.

iii

TABLE OF CONTENTS

1 Introduction
1.1 The expression even if ..........................

1.2 A Note on Kinds of Meaning .......................
1.3 An Overview ...............................

2 Previous Studies
2.1 English even if ..............................
2.1.1 Bennett (1982) and the ‘Consequent Entailment Problem’ . .
2.1.2 Guerzoni and Lim (2007) .....................
2.2 Japanese ‘Even (If)’ Expressions .....................
2.2.1 Fujii (1989) ............................
2.2.2 Nakanishi (2006) .........................
2.2.3 Yoshimura (2007) .........................
2.3 Summary .................................

3 Constructing Japanese Concessive Conditionals
3.1 On The Gerundive Construction .....................
3.2 Deriving the Meaning of the Gerundive Construction .........
3.3 On m0 ...................................
3.4 Deriving the Meaning of Gerundive—m0 ................

4 Summary and Some Implications
4.1 The Gerundive Construction Revisited .................
4.2 Mo Revisited ...............................
4.3 Other General Issues ...........................

5 Conclusion

References

iv

(0me

11

11
13
18
18
19
21
22

25
25
30
34
38

43
43
46
50

56

57

1 Introduction

1.1 The expression even if

Let us first consider what the speaker means by uttering an even if sentence.

Suppose one said sentence (2), given the following situation.

(1) A scenario: We have a friend Mary, who is hospitalized now. Among several
visitors she has, she is usually very happy whenever her boyfriend John comes
to see her, whereas it is mostly not so when her ex-boyfriend Bill comes
instead. As long as she can see John though, she is happy. But we have found
that this is not the case on a day she has to get an injection. She hates it so
much that on a day she has one . . .

(2) Even if John came, she would be in a bad mood all day.

An even if sentence like (2) is the one I refer to as concessive conditionals and is the
target expression of this thesis. Given the situation (1), the sentence is grammatical
and felicitous. We understand that it adds a kind of emphasis indicating in what way
this is unexpected compared to the other normal situations in which John comes.
Thus, it is infelecitous to say the following given the scenario (1) because they do

not seem to express unlikeliness or unexpectedness with respect to what we know.

(3) a. # Even if John didn’t come, she would be in a bad mood all day (on her
injection day).
b. # Even if Bill came, she would be in a bad mood all day (on her injection
day).
Knowing that Mary was usually happy when John came, and one might believe that
if John came she would be happy but may not be if he didn’t, (3a) would sound
infelicitous and we might partially reject it by giving a response like “yeah, but, she
would be like that on her injection day anyway if he didn’t come. . . ” Similarly, to (3b),
we might say “yeah, but, she would be like that anyway if Bill came. . .” Hereafter, I
mark the sentence with “7“” to indicate that the sentence is infelicity, as opposed to

“*” for morpho—syntactic ungrammaticality and clear falsity as an intended meaning.

1

Some general questions will be about how to represent the meaning of conces-
sive conditional in a linguistic model, and how to derive the notion of expectedness
or likeliness compositionally. I investigate the questions from the perspective of the
Japanese concessive conditional construction. Since it appears in quite a different
morphological composition from that in English, it gives us a new look at the repre-
sentation of the concessive conditional meaning.

The following sentence is an equivalent to the even if sentence (2).

(4) John-go. kite-mo Mary-tun kigen-ga wami
John-NOM c0me.GER—mo Mary-TOP mood-NOM bad.NONPAST
‘Even if John came, Mary would be in a bad mood.’

Like in English, it expresses how unlikely or unexpected it is for Mary to be in a
bad mood given John’s coming, considering the normal situation. If the subject in
the antecedent clause was ‘Bill’ instead of ‘John’, then it would result in a kind
of rejection ‘yeah but...’ as same as in English (3). Note that, however, unlike
English even if, there is no explicit conditional morpheme. It does not appear to
be concessive ‘conditional’ in that sense. Instead, the particle mo is attached to the
antecedent clause whose verb ‘come’ is in the gerundive form (represented with GER
in the gloss).1

By comparing the morphemes and structure of concessive conditional and other
conditionals in Japanese as presented in (5), it is clear that the Japanese concessive
conditional does not involve an overt conditional construction. The following types

of construction are taken from Kuno (1973:p.168—194).

(5) a. Perfective tam Clauses

John-go kitam Mary-via kz'gen-ga wamz‘
John-NOM come.if Mary-TOP mood-NOM bad.NONPAST

‘If John has come/ came, Mary will be in a bad mood.’

 

1I adopt a traditional term ‘gerundive’ (Kuno 1973) for convenience. Morpho—phonologically
speaking, the gerundive form in Japanese is an afiixation of allomorphic te or de on the verb stem.
The syntactic and semantic properties of the gerundive form in Japanese may not be identical to
that of the gerundives in English.

b. Assertive nam Clauses

John-go kvrv nam Mary-um kz’gen-ga
John-NOM come.NONPAST if Mary-TOP mood-NOM
warm

bad.NONPAST
‘If John is coming, Mary would be in a bad mood.’

c. A clause with the temporal marker to

John-go kur’u—to Mary-too kz'gen-ga warm
John-NOM come.NONPAST-t0 Mary-TOP mood-NOM bad.NONPAST

‘If/When John comes, Mary would be in a bad mood.’

All sentences in (5) show that the verb form of the antecedent clause must take
the past (perfective) form, as in (5a) or the non-past form, as in (5b,c). There is
neither a gerundive form verb for ‘come’ nor a particle like mo. Thus, comparing the
morphological manifestation of concessive conditionals in (4) with that of conditionals
in (5a—c), it is apparent that the former is not simply built up from the' latter. The
situation is a little different from English in which the concessive conditionals even if
clearly has the conditionality realized morphologically with if.

One might hypothesize that the particle mo corresponds to even. However, the
following does not support that view. The sentence (6) is a minimal pair of (4), where
mo has removed.

(6) John-9a kite Mary-we kigen-ga wami
John-NOM come.GER Mary-TOP mood-NOM bad.NONPAST

‘John came, and (so) Mary is in a bad mood.’

Without mo, the sentence is a conjunction ‘and’ of two events. The meaning of
sentence (6) is that the propositions that John came and that Mary is in a bad mood
both must be true, thus no conditionality at all. In addition to the conjunction of
the two events, the construction also expresses some sort of relationship between the
two propositions such as so in English might represent. The expression signals the

speaker’s perspective on the situation such that the Mary’s being in a bad mood is

3

somehow resulted from John’s coming. Thus, there is an causal relationship between
two clauses, although it is less explicitly expressed compared to sentences with an
overt causal conjunction such as Java ‘because’ and —node ‘because’. From this
observation, one might expect that mo is an element that takes this gerundive ‘and’—
conjoined sentence and turns that into a concessive conditional.

Another possible hypothesis is that mo be an element that introduces not only the
notion of unexpectedmess/unlikeliness but also conditionality, i.e. mo as a combination
of even and if. However, this view seems to go too far beyond the core property of mo.
The basic property of m0 is to introduce an additive meaning ‘also’ as the following
shows, and conditionality is not involved.

(7) a. [John]—m0 kita.

John-mo come.PAST
‘John also came (in addition to someone else).’

b. [John-go kum~t0]-mo kangae—mrem
John-NOM come.NONPAST-COMP-m0 think-able.NONPAST

‘It is also conceivable that John will come (besides other conceivable
situations). ’

When it attaches to the noun John, it implies someone other than John that has come
in (7a). When mo attaches to the tense specified embedded clause John-go [cum-to,
it signals some other proposition besides ‘John is coming’. Thus, it is not the case
that mo inherently has conditionality as a part of the meaning.

In sum, the above facts in Japanese raises a question of the emergence of condi-
tionality in concessive conditionals, so to speak. My hope is to derive the conditional
meaning without any overt conditional morpheme and to find a way to represent the
notion of ‘unlikeliness’ or ‘unexpectedness’. Solving this puzzle with a compositional
analysis includes a careful investigation of particles such as mo that is difficult to find
the equivalent in other languages (English, for example) by itself. Furthermore, it re-

quires us to reconsider the meaning of (concessive) conditionals and its representation

4

in a linguistic model.

1.2 A Note on Kinds of Meaning

Let me clarify what I mean by ‘meaning’ and what kind of meaning my analysis is
trying to capture. Briefly put, in considering the meaning of concessive conditionals,
the present thesis considers all the aspects of meaning together on the same level
of representation, either truth-conditional or of conventional implicature, but setting
aside conversational implicature. As it is commonly understood, when we say to
know the meaning of a certain expression, it is to know the truth-conditions (Heim &
Kratzer 1998). That is, it is to know the conditions for the expression to be judged as
true or false. If I say I ate an apple, we know what the condition is for the sentence
is judged as true. Thus, the sentence is judged as true if and only if I, the one who
uttered the sentence, ate an apple.

A sentence could convey more meaning than what is said, for example, the class of
meaning called ‘conversational implicatures’. One of the classic examples in Gamut

(1991:Ch. 6) is a meaning related to the temporal ordering.

(8) a. Annie took off her socks and jumped into bed. (Gamut 1991:197)

b. Annie took off her socks and then she jumped into bed.

By saying (8a), there is first of all a truth-conditional meaning that the two things
have happened, i.e. the event of Annie’s taking off her socks and the event of her
jumping into bed are both true at some point in past. The conversational aspect of
the meaning of (8a) is that it happened in that order as described in (8b) as then
explicate. The conversational meaning that arises in (8a) is said to be ‘cancelable’.

Below, what follows but cancels the ‘then’ meaning.

(9) Annie took off her socks and jumped into bed, but I do not know which she
did first. (ibid.)

If the temporal sequence meaning in (8a) is conventionalized and always there as
a part of truth-condition, then what follows after but. . . should lead it to contradic—
tion.2 There is an approach to derive this kind of meaning, e.g. Grice’s maxims and
the cooperative principles as a part of his theory of language use. As Potts (2005)
remarks, the maxims are ‘independent of language’ (p. 26), and thus not of idiosyn-
cratic properties of the grammar of language. In such a view, the conversational
implicature, the ‘then’ meaning in (8a), is not encoded as a linguistic property of the
conjunction and.3

For the concessive conditionals, on the other hand, it is clear that the meaning of
even if is not of conversational implicatures. The meaning of concessive conditionals,
the notion of unlikeliness/unexpectedness as we saw in the previous subsection, arises
because of the existence of morphemes like even and if, thus the meaning must be
derived conventionally. This should be the case in the Japanese concessive conditional
too, for it is the interaction between the gerundive clause and the particle mo that is
responsible for the ‘even if’ meaning. This is also supported by the cancelablity test

as I show below.

(10) John-ga kite—mo Mary-ma kigen-ga warui
John-NOM come.GER—mo Mary-TOP mood-NOM bad.NONPAST
‘Even if John came, Mary would be in a bad mood.’

a. #...Taitei John-go kam-to Mary-ma kigen-ga
Usually John—NOM come.NONPAST-to Mary-TOP mood-NOM
warui-kedo.

bad.NONPAST-th0ugh

 

2Assuming that a conjunction but does not allow a pair of contradicting truth-conditions to be
conjoined together in the discourse.

3Levinson (2000:199) introduces a following example in English from Wilson (1975): Driving
home and drinking three beers is better than drinking three beers and driving home. This shows
that the temporal sequence meaning is required, or otherwise the comparison between A and B
and B and A in the form A and B is better than B and A would be the same, which should make
the comparatives meaningless because the comparison is made between the two same things. This
is called ‘pragmatic intrusion’ (Levinson 2000) where the truth-condition of the whole sentence,
the comparative, depends on the local and less explicit meaning like ‘then’. This might doubt the
cancelable property of the ‘then’ meaning in ‘and’ conjunction. However, it might depend on how
to analyze the meaning of comparatives, and I will leave such an issue open.

6

‘...Though usually, when John come, Mary is in a bad mood.’4

b. #...Mochiron John-ga kareba, sore-ma
Of course John-NOM come.CONDITIONAL that-TOP

atarimae- da-kedo.
expected-COPULA.NONPAST-though

‘...Though, of course, that is expected if John came.’

Besides the above two aspects, the truth-conditional meaning and conversational
implicatures, there are several other terms and concepts that surround the non-truth-
conditional aspect of meaning. The most common ones are ‘presuppositions’ and
‘conventional implicatures’. Whatever the complicated history of misusage or misun-
derstanding there is behind certain terminology, in my present analysis, I will treat
all the aspects of meaning, other than the one of conversational implicatures, to be
a part of conventionalized meaning along with truth-conditions that can be derived
in terms of linguistic properties rather than language use. The two terms may be
used interchangeably in the present thesis, since the distinction is not crucial for my
purpose.5 I am following the strategy taken by Karttunen & Peters (1979) in the
following sense as described by Gamut: “[. . .] the value of the Karttunen and Peters
article is not that it distinguishes conventional implicatures from presuppositions, but
that it presents a method which demonstrates how non-truth-conditional aspects of
meaning can be dealt with in the same formal recursive manner as truth—conditional
ones” (Gamut 1991:218).

In sum, there are several different terms and notions to label different kinds of
meaning, such as truth-condition, conversational implicature, conventional implica-
ture, and presupposition. What I focus on is all those kinds except conversational
implicature. Thus, it is no important for the purpose of the thesis to make a precise

distinction between truth-condition and conventional implicature (presupposition).

 

4I follow Kuno’s (1973) treatment of to as a ‘temporal marker’.

5However, in Potts (2005, 2007), where he analyzes the ‘supplementary expressions’ as conven-
tional implicatures, the distinction is made more clearly in terms of cancellation under a certain
condition and projection property (projection plugs).

7

Now that it is clearer what kind of meaning we should look for, let us return
to the target expression (2) and (4). It is important to notice that the notion of
‘expectedness’ or ‘likeliness,’ is conventionalized; it arises because of the certain con-
struction (for example having a lexical item even in English), and not just because of
what the situation was like when the sentence was uttered or what the feeling of the
speaker/ addressee was like. For Karttunen & Peters (1979), even is an element that
does not affect the truth-condition of the sentence but that contribute to non—truth-
conditional and conventional meaning" The ‘even’ meaning consists of two parts: (i)
an additive or existential meaning, and (ii) a scalar meaning with respect to likeliness
or expectation. For example, sentence (11) is analyzed as consisting of several aspects

of meaning (12a—c).
(11) Even Bill likes Mary.

(12) a. Bill likes Mary.
b. Other people besides Bill like Mary.

c. Of the people under consideration, Bill is the least likely to like Mary.
(From Karttunen and Peters 1979: (16-18))

In order for (11) to be true and felicitous, it first has to be that Bill likes Mary (12a).
Furthermore, the sentence says that there are people other than Bill who likes Mary
(12b), i.e. the aditive/existential meaning. It also says that Bill is the least likely
one to like Mary of the people under consideration (12c), i.e. the scalar meaning.
In Karttunen and Peters’ sense, the aspect of meaning (12a) is the ‘truth-condition’
of the sentence (11), and the rest (12b—c) is the presupposition (or the conventional
implicature, in their original terminology).

Turning to concessive conditional (2), it is not clear, compared to example (12),

what exactly the truth-condition of the even if sentence is, based on the above distinc-

 

6They call it ‘(conventional) implicatures,’ and others point out that it is actually ‘presupposi-
tions.’ (Potts 2005:Ch.2). I do not think that, in any event, the distinction between truth-conditional
and non-truth-conditioanal is particularly a critical issue in the present analysis, as long as it is not
about the conversational implicatures.

tion. Suppose even only contributes to the non-truth—conditional (but conventional)
aspect of meaning following Karttunen and Peters. Does it follow then, in terms of

the truth—condition, (2) has the meaning (13a)?
(13) A tentative meaning for even if John came, Mary would be in a bad mood:

a. If John came Mary would be in a bad mood.
b. There could be other situations besides John’s coming.

c. Of the situations under consideration, John’s coming is the least likely.

Suppose the description in (13) is a correct view. Then, the sentence (2) even if
John came, Mary would be in a bad mood would mean that John’s coming is un-
likely/ unexpected compared to other situations, from (13b,c). However, let us recall
scenario (1). It does not say that the event of John’s coming itself is unlikely. Rather,
we want to have a meaning that says ‘it is unlikely/unexpected for Mary to be in
a bad mood by John’s coming’. Thus, in extending Karttunen and Peter’s view of
even, one should reconSider when determining the truth-condition of the even if ex-
pressions as well as the other aspects of meaning. As I noted, the main purpose of
this thesis is not to sort the concessive conditional meaning into truth-condition and
other kinds of meaning. Although I do not present a comprehensive analysis on the
Englisheven if, my compositional analysis on the Japanese even if sentences, which
do not exhibit a morphological one-to—one correspondence to the English even and if,
will hopefully shed light on the general question about how to represent the overall

meaning of concessive conditionals.

1.3 An Overview

In section 1, I introduced the target construction, the concessive conditional con-
struction, and explained what the puzzle is. I also noted what I mean by the ‘meaning’
and what kinds are relevant in my analysis. In section 2, first I review some of the

previous works on English even if (Bennett 1982, Guerzoni & Lim 2007 ), and then

9

on the Japanese even if expression (Fujii 1989) as well as on the closely related ‘even’
expression demo (Nakanishi 2006, Yoshimura 2007). Section 3 presents my composi-
tional analysis on the Japanese concessive conditional construction. It consists of two
parts. First I consider the gerundive construction without mo and derive the mean-
ing of ‘and’-conjunction (section 3.1-3.2). Then, I propose a denotation for mo that
would be added on to the analysis of gerundive constructions (section 3.3-3.4). The
meaning is derived step by step via functional application following the framework
laid out in Heim & Kratzer (1998). Section 4 summarizes the main points in my pro—
posal and their implications. Additional data is presented at the end where necessary,
for example, the deontic ‘may’ modal construction, the so-called Universal Concessive
Conditional that shows an interaction of wh-phrases and concessive conditionals, and

morphological variations across different registers/ dialects of Japanese.

10

2 Previous Studies

I will first review Bennett (1982) and Guerzoni & Lim (2007) in section. 2.1,
and point out some achievements and problems in the compositional analysis of the
concessive conditionals. In section 2.2, I will turn to the meaning and usage of the
Japanese concessive conditionals (Fujii 1989). I will also review some of the recent
works on the Japanese even-like item demo (Nakanishi 2006, Yoshimura 2007). Their
works concentrate particularly on the relationship between demo and noun phrases
(e.g. [Mary-demo] kita ‘Even Mary came’). Therefore, their compositional analyses
on the Japanese ‘even’ item requires some modification to my analysis on concessive
conditionals, where it involves conditionality and clausal conjunctions. Nevertheless,
it is worth reviewing their work on demo for my analysis, for demo actually consists

on the same elements as concessive conditionals (i.e. the gerundive form and mo).

2.1 English even if
2.1.1 Bennett (1982) and the ‘Consequent Entailment Problem’

Bennett (1982) makes a distinction between two kinds of even if meaning. He
calls one of them ‘Introduced If Conditionals’ as exemplified in (14a) and the other
‘Standing If Conditionals’ as in (14b). The distinction is based on the generalization
by Pollock (1976), in which he calls the former ‘standard use of even if’ and set aside

the other as exceptional.

(14) a. Introduced if Conditional

(Situation: I stand looking at the ranging water of the river and the ruins
of the bridge. . .)

Even if the bridge were standing, I would not cross.

<——+ I would not cross.

b. Standing if Conditional
(Situation: John’s boss is so puritanical.)

Even if he drank just a little, she would fire him.

11

H She would fire him if he drank some.

According to Pollock (1976) and Bennett (1982), the truth of consequent I would not
cross is implied (¢—)) in the Introduced if conditional (14a), but the consequent she
would fire him in the Standing if conditional (14b) is not. To paraphrase, the natural
reading of (14a) is almost saying as ‘I would not cross (no matter what)’ or ‘I would
not cross if the bridge were standing or not’, whereas in (14b), it is not ‘she would fire
him no matter what’ but rather it is ‘she would fire him if he drank some’. This
contrast is sometimes referred to as the ‘Consequent Entailment Problem’ (Lycan
1991, Baker 1994).

Bennett (1982) then gives a set of conditions (15) for a concessive conditional

sentence S to be felicitously asserted.

(15) o The S" is true,

0 There is a neighbor sentence (8]) which is known and naturally related to
S*,and

o Sj is more expectably true than S* is.

This conditions applies equally to both kinds of even if sentences. However, what
counts as S“ and Sj look different depending on whether it is of Introduced or Standard
if conditionals as shown below.
(16) Introcuded If Conditionals

S: Even if the bridge were standing I would not cross.

Sj: I would not cross.
8*: If the bridge were standing I would not cross.

(17) Standing If Conditionals
S: Even if John drank just a little she would fire him.
Sj: If John drank a lot she would fire him.
8*: If John drank just a little she would fire him.

By applying the conditions in (15), the two kinds of even if meaning can be

described in the following way.

12

(18) o The S" ‘If the bridge were standing I would not cross’ is true,

0 There is a neighbor sentence (Sj ‘I would not cross’) which is known and

naturally related to S", and

o 83- ‘I would not cross’ is more expectably true than S* ‘If the bridge were

standing I would not cross’ is.

(19) o The S" ‘If John drank just a little she would fire him’ is true,

0 There is a neighbor sentence (Sj ‘If John drank a lot she would fire him’)

which is known and naturally related to S", and
o S, ‘If John drank a lot she would fire him’ is more expectably true than
S* ‘If John drank just a little she would fire him’ is.

As Bennett remarks, the contrast is attributed to the difference in the scope that
even may take. According to him, even is a kind of operator that has a certain scope
of operation and that provides what he calls ‘neighbor’ sentences (labeled Sj). Notice
that the Sj of the Introduced if conditional only consists of the consequent clause, but
the one of the Standing if conditional consists of both the antecedent and consequent
clauses. “It’s difference between a sentential vehicle for ‘even’ which happens to be a
conditional, and one whose conditionality is itself a result of the operation of ‘even’,
so to speak.” (Bennett 19822411)

Bennett’s (1982) analysis maybe a good start to grasp the overall meaning of
concessive conditionals. However, it still remains unclear how to obtain a relevant
neighbor sentence for each even if sentence. Furthermore, we do not know the precise

meaning of even itself and how that interact with the conditionals.7

2.1.2 Guerzoni and Lim (2007)

Given such unanswered questions, Guerzoni & Lim (2007) present a compositional

analysis on even if sentences with a denotation for even. For them even is a focus

 

7Baker (1994) points out other potential problem with Bennett’s account on the Introduced if
conditionals.

13

sensitive operator that does not affect the truth-conditional meaning of the sentence
that it applies to, but only adds some kind of presuppositional meaning, following
Karttunen & Peters (1979). The following (20) is the formalization that gives the
two kinds of presuppositional meaning (the additive meaning (20a) and the scalar
meaning (20b)), and the truth-conditional meaning (20c). (Symbols ‘p, q’ stand for
variables for propositions, ‘w’ for an evaluation world, ‘C’ for a set of contextually

salient propositions.)

(20) [I even ]] (C)(p)(w) is defined iff

a- 3 q 60 [q ,5 p & q(W)=1l &
b' V q EC [q 7g p _’ p <likely/ezrpected q l
c. If defined, then [[ even I] (C)(p)(w) = p(w)

Adopted from (4) in Guerzoni and Lim (2007)

The denotation (20) defines a condition for an even if sentence to be felicitous in a
certain discourse context. When even is applied to the sentence p(w) in a certain
discourse context C (i.e. a proposition that takes a variable w, the evaluation world),
it first presupposes that there exists another proposition q that is not the same as p
(the additive meaning (20a)). It also presupposes that comparing p and q, p is less
likely than q (the scalar meaning (20b)). The basic meaning of p(w) does not change
with or without the word even (the truth conditional meaning (20c)).

The following shows the structure assumed for the Standing if conditional.
(21) [even [if John drank one ounce of whiskey he would be fired] ]

This shows that the word even is a sentential operator that applies the whole sen-
tence [if John drank one ounce of whiskey he would be fired]. Following the proposed
denotation (20), the truth-condition of the even ifsentence even if John drank one

ounce of whiskey he would be fired is equivalent to if John drank one ounce of whiskey

14

he would be fired. A difference is that, with even, the sentence is has two kinds of
presuppositions: the additive and scalar meaning.

In order to formalize and explicitly describe what those presuppositional meaning
are, Guerzoni & Lim (2007) adopt Rooth’s focus/ alternative semantics. As Rooth
(1997) explains, there are two distinct levels of semantic interpretations in the se-
mantic model. One is called the ‘ordinary semantics’ level, whose interpretation is
represented with Mo, and the other is the ‘alternative semantics’ level, with fl-flF.
While the ordinary level computes the truth condition, the alternative level derives
the non-truth-conditional meaning such as focus by producing a set of alternatives.
In Guerzoni and Lim’s (2007) View, there is always a focus assigned somewhere in the
even involved sentences. The interpretation of the focus meaning is a key to derive
what Bennett (1982) called the ‘neighbor’ sentences that is crucial for the likeliness
scalar meaning.

In the example (21), the phrase one ounce is focus assigned according to their
analysis. Roughly, at the level of alternative semantics, the following set of alternative
propositions is produced under such condition.8 Symbol ‘d’ stands for a variable of

degree.

(22) a. {p: 3d&p=that if John drank d—much whiskey she would fire him}

b.
that if John drank one ounce of whiskey she would fire him,

that if John drank one and half ounce of whiskey she would fire him,
that if John drank two ounces of whiskey she would fire him,. . .

(Guezoni and Lim 2007 (22))

We may think this as a process of deriving set of alternative propositions (neighbor
sentences) that varies in terms of the amount of whiskey. This set of alternatives is

supposed to fix the range of C in the denotation (20).

 

81 interpreted ‘&’ as ‘such that’.

15

The denotation (20) would predict the following meaning for the Standing if con-
ditional (21).

(23) a. ADDITIVE MEANING
There is a proposition q in the set C(=(22)) that is not same as p(=‘if

John drank one ounce of whiskey he would be fired’)

b. SCALAR MEANING
‘If John drank one ounce of whiskey she would fire him’ is less likely than
all other alternatives in C(=(22))9

c. TRUTH CONDITION10
The truth condition of even if John drank one ounce of whiskey she
would fire him is equivalent to that of if John drank one ounce whiskey
she would fire him.
In this way, their compositional analysis contributes to the formalization of concessive
conditional. It recaptures Bennett’s ( 1982) notion of neighbor sentences by adopting
the focus / alternative semantics.
However, in their attempt to uniformly derive the two kinds of ‘even if’ meanings,
they introduce an additional null morpheme in the analysis to account for Introduced
it conditionals, which may make their proposal less convincing. The null morpheme

AFF appears in the antecedent clause, and it is focused in the Introduced it condi-

tional.
(24) [even [if [AFF]F the bridge were standing I would not cross] ]

This AFF is supposed not to affect to the meaning at the ordinal semantics level as
shown in (25a), but does so at the alternative level when it has a focus assigned as

in (25b).

9As it has been pointed out by Marcin Morzycki, there is no constraint that states the amount
‘on ounce’ to be the minimum amount among the alternative set. So, there could be ‘a drop of
whiskey’ or ‘a sip of whiskey’ for example in the alternative set. However, it should not be that
drinking one ounce and get fired is less likely than drinking a sip of whiskey and get fired.

10Guerzoni and Lim (2007) call it ‘assertion’ instead of ‘truth condition’. The two terms may not
be defined the same. According to Stalnaker (1978) for example, the term ‘assertion’ may include
the additive and scalar meanings as well. I tentatively interpret their ‘assertion’ as ‘truth condition’
for simplicity.

 

16

(25) a. [[AFFBO = A t. t
b. [[AFFHF = {A t. t, /\ t. t=0}

(Adopted from (15) and (17) in Guezoni and Lim (2007))

They explain that the focus assigned AFF produces the following set, which will

determine the range of C in the denotation of even.

(26) C— that if the bridge were standing I would not cross,
— that if the bridge were not standing I would not cross

(Adopted from (18) in Gueraoni and Lim (2007))

This analysis explains how the Introduced if conditional seems to imply the truth of
the consequent ‘I would not cross’ independently. However, from a technical point
of view, it is not clear how a set of truth values {0, 1} can be transformed into a
set of propositions (26). Even though it could, positing a null morpheme should be
done with care or otherwise it may be ad hoc. To support the adequacy of the null
morpheme, we may look for a condition that restricts the distribution of the null
morpheme. For example, there seems to be no reason for the non-existence of the
focus assigned AFF in the Standing If sentence. It is also unclear why it should only
appear in the antecedent clause (if AFF the bridge were standing) and not elsewhere
(e.g. if the bridge were standing I AFF would not cross).

Furthermore, in their account, the set of propositions are in the form of English
conditional ‘if... (then)’ as in (22, 26). However, the English if( then) expression is
not equivalent to a logical connective such as material implication in the propositional
logic (cf. Kratzer (2002, 1986)). Therefore, it may be not adequate to represent a
list of propositions using English if( then), since English if... may be tied to non-
truth—conditional meaning. My concern is that, in Guerzoni & Lim’s (2007) account,

there is no consideration on the conditionality itself and its interaction with concessive

17

meaning. Accordingly, the following contrast may imply the interaction of conditional

and even.

(27) a. If John drank one ounce of whiskey (then) she would fire him.

b. Even if John drank one ounce of whiskey (*then) she would fire him.

The optionality of then is not available when there is even. Although it depends
on how to analyze then, an implication is that the conditional part of concessive

conditional may not be equivalent to the conditional without ‘even’.

2.2 Japanese ‘Even (If)’ Expressions
2.2.1 Fujii (1989)

Fujii (1989) claims that the meaning of the gerundive-mo construction in Japanese
as exemplified in (4) is different from what an even if sentence in English means.
According to her, the Japanese gerundive-mo construction has broader range of use
and it could be translated in even if, even when, even though, or although in English.11
However, her claim is not problematic for my analysis, since her description about
the Japanese gerundive-mo construction, as I summarized in (28), is the property of
concessive conditional in general and it is not limited to the Japanese gerundive-mo

construction.

(28) Fujii’s (1989) generalization: In the Japanese gerundive-mo construction,

a. it is not specified whether or not the antecedent clause is accepted as true
by the speaker, and

b. the consequent clause expressed is contrary to the consequent clause
expected, given the truth of the antecedent clause.

As it will be shown in section 3, my analysis will account for the observation in (28).

 

11It is another issue in what way the related expressions in English (such as even if, even when,
even though, and although) are similar and whether they can be also analyzed as a kind of concessive
conditionals.

18

Although Fujii’s (1989) observation captures the meaning of the Japanese gerundive-
mo sentences, she does not propose a compositional analysis to account for the con-
cessive conditional meaning. To my knowledge, no compositional analysis is available
for the Japanese gerundive—mo construction. I will now turn to some previous studies
on demo, a kind of even expression in Japanese (Nakanishi 2006, Yoshimura 2007).
However, their analyses on the expression demo is not about concessive conditionals.
The reason I still refer to their works here is because they present an application of
Karttunen & Peter’s (1979) view on even to the Japanese data in a compositional
semantic framework. Moreover, since the expression demo can be decomposed into
de-mo (copula do in the gerundive form and mo), I believe that it is construction-
ally related and that my analysis on concessive conditional could be extended to the

understanding of demo in future.

2.2.2 Nakanishi (2006)

For N akanishi (2006), demo and mo are both ‘even’, which I will argue against
later. Her focus of the analysis is where demo/mo appears adjacent to a noun phrase.

(29) a. John-wa [Hon A]F {-mo/-demo/??-dake-demo} yon-da.
John-TOP book A {-even/-even/-only-even} read-PAST
‘John even read Book A.’

b. John-wa [Hon A]F {-mo/-demo/*-dake-demo} yom-ana-katta.
John-TOP book A {—even/—even/—only-even} read-NEG-PAST

‘John didnt even read Book A.’
(adopted from (9) and (10) in Nakanishi (2006))

Similarly to the English even, both sentences (29) involve a scalar meaning. As we
have seen in the ‘even if’ analysis by Gueroni & Lim (2007), the scalar meaning
of ‘even’ can be described in terms of a set of alternatives associated with focus.
Following Rooth’s alternative / focus semantics, the set of alternatives can be achieved
by having a set of propositions that minimally differs where the focus is assigned, i.e.

by substituting the focus marked ‘Book A’ with other candidates.

19

(30) a. John even read [Book A] F

John read Book A,

b. Among the set of propositions C: John read Book B, ,
John read Book O,

[John read Book A] is the least likely thing to happen.

(31) a. John didn’t even read [Book A] F

John read Book C,. ..
[John read Book A] is the most likely thing to happen.

John read Book A,
b. Among the set of propositions C: John read Book B, ,

A puzzle is that, as I bold faced in (30b) and (31b), there are two options for the
scalar meaning (‘the least likely’ or ‘the most likely’) depending on whether there is
a negation in the sentence or not.

A possible solution is to have the scope of even over the negation.

(32) a. even > —1
John didn’t read Book A, }

b. Among the set of propositions C: John didn’t read Book B,
John didn’t read Book C,. . .

[John didn’t read Book A] is the least likely thing to happen

([John read Book A] is the most likely thing to happen).
This would prevent us from having a lexical ambiguity theory that assumes two kinds
of ‘even’: one that derives the ‘least likely’ meaning in (30) and another that derives
the ‘most likely’ meaning in (31).12

N akanishi (2006) argues, based on the interaction with dake ‘only’ as in dake—demo

‘only-even’, that the Japanese even items (demo and mo) support the scope theory.

Essentially, her analysis on demo/mo follows Karttunen and Peter’s ‘even’.

(33) [[—demo][“’ (C)(p) is defined iff VqEC[q7é p —> q>z,-kelyp] (Nakanishi 2006 (26))

 

12However, as Giannakidou (2007) points out, it leaves us with a question why such movement
(the raising of even) is obligatorily required to fix the order even > -., and never allows an option
of the hierarchical order of -1 > even.

20

According to N akanishi (2006), demo, like the English even does not change the truth
condition of the sentence ‘John read Book A’, however, it introduces a scalar meaning,
i.e. ‘John read Book A’ is the least likely thing among all other considerable situations
(e.g. {John read Book B, John read Book O, ...}).13

In short, we may view this as an example of applying an ‘even’ analysis in En-
glish to Japanese data. The problem is that the word demo does not appear in the
concessive conditional construction. If, as Nakanishi (2006) assumes, demo and mo
are both ‘even’, we may adopt the denotation (33) to the gerundive-mo construction
and modify it so that it can take a clausal argument for example. However, I think
demo and mo should be kept distinct from each other from both theoretical and em-
pirical points of view. As Nakanishi (2006) herself mentions in her footnote 6, the
grammaticality judgement of demo-sentences is not stable. As we will see below, this
is explained by that demo and mo are indeed different and their grammaticality or

acceptability are judged on different property of the sentence.

2.2.3 Yoshimura (2007)

Yoshimura (2007) takes the unstable judgment result for demo seriously, and argues
that mo and demo should be analyzed separately because the acceptability of the

two morphemes depend on different licensing conditions. The following data (34)

illustrates that demo and mo appear under different conditions. 14

(34) a. i-tteki-{??demo/*mo} non-da.
one-CLn-me-{DEMO/MOijpitch} drink-PAST
‘#(I) drank even one drop.’

 

13Nakanishi (2006) denies the need of the existential presupposition.

14In this case, the accent pattern of the phrase with mo makes a difference in meaning. Thus, for
example, a phrase i.ehi.pe.e.ji.mo. ‘one—page-mo’ with the LHHHHH pattern (where L: low, H:
high) would be an NPI element (or an Negative Cord Item (Watanabe 2004)) ‘not even one page’,
requiring an overt negation that c-commands it within the same clause, whereas the same phrase
with the LHHHLL pattern would be ‘one page’ accompanied with an implicature/presupposition
‘one page is surprisingly a lot’ without requiring any overt negation in the sentence. Let us set aside
of the latter one.

21

b. i-tteki—{??demo/mo} noma-naka-tta.
one-CLu-me-{DEMO/MONPIm-tch} drink-NEG-PAST

‘(I) didnt drink even one drop.’

c. i-kkai-{ demo/*mo} sensei-ni hanashi-ta no?
one-CLtime-{DEMO/MONplpitch} teacher-to talk-PAST Q
‘Did (you) talk to the teacher even once?’

Yoshimura (2007; Ch.5 (75-77))

As the contrast (34a,c) and (34b) shows, i-tteki-mo should appear with negation. On
the other hand, i-tteki-demo is natural only when the sentence involves a so—called
‘affective operator’, such as Yes-No questions as in (340), conditionals, imperatives,
generics, emotive verbs (e.g. ‘be surprised (at)’), and intentional verbs (e.g. ‘want’).
Furthermore, it is important to note that demo and mo result in different degrees and
kinds of unacceptability. One the one hand, the licensing failure of moms; (34a, 0)
results in ‘sharp ungrammaticality’ as indicated by ‘*’. On the other hand, the failure
results in a ‘pragmatic oddity’ as indicated by ‘#’ in the case of demo as shown in
(34a, b). This suggests that demo and mo are different and the contrast should be

represented in a linguistic model.

2.3 Summary

As for the expression even if in English, there was an issue on the two kinds of ‘even
if’ meaning (Pollock 1973). Bennett (1982) analyzes the two meanings, however, leav-
ing a room for a conventional way to derive the unexpectedness meaning. Guerzoni
& Lim (2007) attempt a compositional analysis for even if to derive the two kinds
of ‘even if’ meaning uniformly by adopting Rooth’s alternative/ focus semantics and
Karttunen & Peter’s (1979) view of even. However, there were some problems with
respect to the null morpheme in the Introduced if conditional and the conditionality

in concessive conditional.

22

For the Japanese concessive construction, first, we examined the basic meaning of
the gerundive-mo construction and its similarity to the meaning of even if in English
(Fujii 1989). We then turned to some application of the English even analysis to
the Japanese demo, another ‘even’-like expression, which is morphologically closely
related to concessive conditional. Like Guerzoni & Lim (2007), Nakanishi (2006)
adopts the focus/ alternative semantics and Karttunen and Peter’s view on even to
derive some ‘even’ meaning in Japanese. While no distinction has been made between
the two kinds of ‘even’ elements (demo and mo) in Nakanishi (2006), Yoshimura
(2007) argues that the two are different.

From these previous studies, I point out some significance of my analysis on the
Japanese concessive conditional. First of all, since there has been no compositional
analysis on the gerundive-mo construction, it would be worth presenting one. There
may be a way of adopting the denotation for demo by N akanishi (2006) or Yoshimura
(2007), and modify it so that we can consider the interaction of the ‘even’ meaning and
conditionals or a similar clausal relationship, extending the case of the ‘even’ element
that is attached to non-clausal constituents. However, it would require a decomposi-
tion of demo, since there is no independent word demo in concessive conditionals, but
the gerundive construction and mo. The decomposition itself is reasonable, knowing
that demo indeed consists of the gerundive form of copula de and mo. The difficulty
is that we do not want to assign conditional meaning to the gerundive form, as we
confirmed in section 1. The gerundive construction is merely a sentential ‘and’ con-
junction. Furthermore, we do not want to take an approach that assumes mo to be
‘even’ by itself, since that would go back to Yoshimura’s (2007) argument. Therefore,
my attempt of deriving the meaning of gerundive construction and proposing a de-
notation for mo that would transform the gerundive into the concessive conditional
will support the view that the distinction between he made between demo and mo.

Moreover, my approach, particularly in deriving the conditionality from the absence

23

of conditional construction, will shed light on the English even if expressions and the
meaning of concessive conditionals in general. I also hope to explore the interaction
between conditionality and the notion of expectedness, which has not been concerned

sufficiently in Guerzoni & Lim (2007).

24

3 Constructing Japanese Concessive Conditionals

This section presents my analysis of the Japanese concessive construction. In order
to see how the meaning of concessive conditional is derived step by step, first, I
will consider the structural relationship of the two clauses involved in concessive
conditionals. In section 3.1, I show that the gerundive construction in Japanese
functions like the sentential ‘and’ conjunction. In section 3.2, I propose a denotation
for the gerundive morpheme to account for the meaning observed in the previous
section. In section 3.3, I will turn to the basic properties of mo, which is the additive
meaning but with the unexpectedness meaning under a certain structural condition.
I then propose a dentation for mo in section 3.4. Lastly, the derivation is shown to see
how the gerundive construction turns into the concessive conditional by the addition

of mo.

3.1 On The Gerundive Construction

As we saw in section 1 and as repeated below, the sentence (35) resulted in an ‘and’
conjunction, as in (36), when there is no mo.
(35) [John-go kite]-mo Mary-wa kigen-ga warui

John-NOM come.GER—mo Mary-TOP mood-NOM bad.NONPAST
‘Even if John came, Mary would be in a bad mood.’

(36) [John-go kite] Mary-wa kigen-ga warui
John-NOM come.GER Mary-TOP mood-NOM bad.NONPAST
‘John came, and (so) Mary is in a bad mood.’

This suggests that the fundamental operation introduced by the gerundive construc-
tion is the clausal conjunction ‘and’. The two conjoined propositions, John’s coming
and Mary’s being in a bad mood, are both true.

The following set of data supports that the gerundive formation is the basic ‘and’

conjunction strategy in Japanese.

25

(37) a. Mary-wa [yasashikute] [kawaii].
Mary-TOP kind.GER cute.NONPAST

‘Mary is kind and cute.’

b. John-wa [hon-o yonde] [terebi-o mita].
John-TOP book-ACC read.GER TV-Acc watch.PAST

‘John read a book and watched TV.’

c. [John-go kite] [Mary-go kaeru].
John-NOM come.GER Mary-NOM went h0me.NONPAST
‘John comes and Mary goes home.’

(As an answer to: What usually happens in the office while I am out?)

(1. [John-wa kite] [Mary-wa kaeru].
John-TOP come.GER Mary-TOP went home.NONPAST

‘John comes and Mary goes home.’

(As an answer to: Do both John and Mary join the reception after the talk?)

There is no free morpheme like and in English that combines adjectival phrases, verb
phrases, and clauses. Instead, the first conjunct (phrase/ clause) takes the gerundive
form.15 Note that the gerundive clause lacks tense, i.e. the interpretation of the
temporal information depends on the other clause that does have the past / non-past
morpheme. For example, (37b) indicates John’s reading and watching TV happened
in past, and (37c) indicates that John’s coming and Mary’s going home both happen
within the same interval of time (in this particular example, ‘while you are out’).
These facts suggest that the gerundive construction in Japanese is a strategy to
combine clauses or phrases that takes tense morpheme.

Let us focus on the clausal conjunctions (37c, d), which consists of two propositions
with different subjects. I set aside constructions (37a, b), which presumably is a
6

coordination of adjective phrases or verbal phrases, and not a clausal conjunction.1

As is also the case in English, a natural interpretation of (37c) may have a temporal

 

15There is a particle to ‘and’ as in [John-to Mary]-ga kita ‘John and Mary came’, but this can
coordinate nominals only, and it cannot connect adjectival/verbal phrases and clauses.

16It is possible to analyze (37a, b) as a clausal conjunction as well, for example, by assuming a
null pronoun (e) as in structure Meryl-we [ e1 yasashikute] [ e1 kawaii]. I will leave this option.

26

sequential meaning ‘then’, i.e. ‘John came and (then) Mary left’. I have explained in
section 1.2 that this aspect of meaning is cancelable, thus it is probably not lexically
encoded as a part of the meaning of and. The following shows that, in Japanese too,
such temporal meaning ‘then’ in (370) can be cancelled by the subsequent utterance

‘I don’t know which happens first though...’.17

(38) (Taitei watashi-ga i—nai-aida-ni) John-ga kite M ary- ga
(Usually I-NOM exist-NEG-while-at) J ohn—NOM come.GER MaryNOM
kaeru.

go h0me.NONPAST
‘(Usually, while I am away,) John comes and Mary goes home.’

...Docchi-ga-saki-ka shiranai-kedo. ..
‘I don’t know which happens first though...’

This fact suggests that the temporal sequential meaning is not conventionalized in the
gerundive morpheme itself, just as it was for the English and (see section 1.2). The
sentence in (37d) too does not necessarily imply which occurred first. As an answer to
a particular question as exemplified in (37d), probably the two things happens at the
same time or within the same interval of time (whenever ‘after the talk’ indicates),
thus no temporal sequetial meaning.

As a note, however, there are certain cases where it is difficult to deny the temporal
meaning. That is when the gerundive clause is subordinated, rather than coordinated,
to the tensed clause. The distinction between subordination and coordination of the
gerundive clause can be made by observing where the topic marker wa appears. For
example, sentence (37c) does not involve the topic marking at all, and the sentence
is coordination. To see sentence (37d), both subjects of the two clauses are equally
topic-marked, and this too is coordination, not subordination. On the other hand,

when there is a topic marking in the second conjunct only and not in the gerundive

 

17I thank Yasutada Sudo for confirming the cancelability of constructions in (38, 39), and for
pointing out a possible interpretation for (40a).

27

clause, it is subordination. As exemplified in (39), in the subordination type gerundive

construction, denying the temporal sequential meaning will result in inconsistency.

(39) (Taitei watashi-ga i-nai aida-ni) John-9a kite Mary—ma
(Usually I-NOM exist—NEG while—at) John-nom come.GER Mary-TOP
kaeru.

go h0me.NONPAST
‘(While I am away,) John comes, and (then/so) Mary goes home.’

#Docchi-ga saki-ka shiranai-kedo...
‘I don’t know which happens first though...’

There is also a sense that the speaker of the sentence (39) thinks that Mary’s going
home is somehow triggered by John’s coming. This sense is based on what the speaker
knows about the relationship between the two events. For example, if the speaker
knows that Mary does not want to see him for some reason, the combination of the
two events ‘John came’ and ‘Mary goes home’ is a reasonable and expected situation.
The sense Of ‘reasonable’ or ‘expected’ based on what is known is signaled in the
asymmetrical type gerundive construction, and it is not cancelable unlike in (38).
Furthermore, the following contrast indicates that the topic marking may change

the structural relationship of the conjuncts.

(40) a. COORDINATION

[Mary-gah John-go kite e1 kaetta.
Mary-NOM John-NOM come.GER go home.PAST

Roughly: ‘John came, and MARY went home.
(It is Mary who went home and not Bill.)’

b. SUBORDINA’I‘ION

[Mary-wah John—go kite e1 kaetta.
Mary-TOP John-NOM come.GER go home.PAST

‘John came, and (then/so) Mary went home.’ (=(39))

In (40a), dislocating the subject of the second conjunct typically results in the focus

meaning, or a kind of contrast between Mary and other people the speaker has in

28

mind. Therefore, dislocation of the subject Mary-go in the coordination structure
changes the original meaning of the sentence in some sense. On the other hand, the
topic marked subject Mary-wa can occur at the beginning of the sentence as in (40b)
without changing the meaning, i.e. without gaining the focus meaning. Compared
to (40a), (40b) sounds neutral and canonical.18 From the fact that no meaning
change occurs in the subordinate gerundive construction, I assume that the gerundive
clause (John—go kite) in (40a) is structurally and semantically subordinated under the
main clause (Mary—wa kaetta), whereas in (40b), the conjoined clauses are a case of
coordination at least semantically.19
Now, let us return to the target construction, repeated below.
(41) a. John-go kite-mo Mary-wa kigen-ga warui

John-NOM come.GER-mo Mary-TOP mood-NOM bad.NONPAST
‘Even if John came, Mary would be in a bad mood.’

b. [Mary-wok John-ga kite-mo e1 kigen-ga warui.
MaryTOP John-NOM come.GER-m0 mood-NOM bad.NONPAST
‘Even if John came, Mary would be in a bad mood.’

The topic marking on the subject noun Mary (but not on the subject of the gerun-
dive clause) suggests that the sentence has some non-cancelable meaning based on
whatever the speaker knows. For the sake of my purpose, I will focus on the meaning
of the subordinate type gerundive construction only, where the causal meaning is
non-cancelable. I will leave the other type aside. I tentatively adopt an idea that the
gerundive morpheme is lexically ambiguous: one for subordination and another for

coordination, for the sake of simplicity.

 

18From a different perspective, Kuno (197 3) also points out that the topicalization (topic marking
with dislocating it across the gerundive close) is only possible when the gerundive clause is being
a subordinate clause of the second conjunct and not when the two clauses are in a coordination
relationship. However, his distinction of subordination-coordination may not be exactly the same
from the one I assume in the thesis.

19Contrary to my argument, Hasegawa (1996a,b) argues that whatever additional meaning that
arises in the gerundive construction is cancelable, and attempts to derive that temporal/causal
meaning from the Gricean pragmatic account. At the same time, however, she also presents some
cases where such meaning is not cancelable.

29

3.2 Deriving the Meaning of the Gerundive Construction

In this subsection, I will compositionally derive the meaning of gerundive construc-
tions (of the asymmetrical type) setting aside mo until later in section 3.3 and 3.4.

First, let me layout the basic rules for my compositional analysis.

(42) Types:

i. s, t are basic types.
ii. If 05B are types then ((1,3) is a type.

iii. Nothing else is a type.

(43) Domains:
i. The domain D of type sis D8 = {w1,w2,w3, ...}.

iii. DOM) is the set of all functions from Do, to Dg

Everything is combined via Functional Application (Heim & Kratzer 1998).

(44) If a is a branching node, {[3, ’7} is the set of a’s daughters, and [[5]] = is a
function whose domain contains [[7]], then [[a]] = [[B]([]'7]]).

As we have confirmed in section 1.2, the main purpose in the present analysis is to
draw a clear line between the truly truth-conditional meaning and presuppositions.
Let us tentatively adopt an idea that all those aspects of non-cancelable meaning con-
tributes to the truth value. That is, making clear a distinction between presupposition
and truth-condition is not important in the present analysis.

For convenience, let us restrict the domain of worlds to just four of the follow-
ing. The purpose of this illustration is to show a way to incorporate the notion of

expectedness or likeliness into the model.

(45) . D8 : {wltw21w3iw4}

o [[John-NOM come] : )‘w.came(John)(w) = {w1, w2}

30

o [[Mary-TOP happyNONPAST ]] = /\w.happy(Mary)(w) = {w1, w3}

This sets up the situation in the following way: in w, both of the two events happened
to be true, just John came is true in wg, John didn’t come and Mary is happy in w3,
and both events didn’t happen in w4. We are thus not considering about propositions
that other visitors such as Bill came or something else happened to Mary (e.g. Her
physical condition started getting worse, she got a wonderful present from Bill. . . etc.),
though, of course, we could if we want to. For now, this is sufficient to represent the
meaning of unexpectedness in the gerundive-mo construction.

In addition to (45), let us further assume that Mary only could be either happy or
otherwise in a bad mood, admitting though that it is too simplistic about the nature

of one’s emotional states.

(46) . D8 :{w11w27w3‘lw4}

o [[John-NOM come] = Aw.came(John)(w) = {1111,1112}
o [[Mary-TOP happy.NONPAST]] = Aw.happy(Mary)(w) = {w1, w3}
o [[Mary-TOP mood-NOM bad.NONPAST]]

=Aw.in-a-bad—mood(Mary)(w) = {w2, w4}

Using the simplified model (46), let us return to the scenario, repeated below, that
allows the concessive conditional even if John came Mary would be in a bad mood to
be felicitous.

Scenario: We have a friend Mary, who is hospitalized now. Among several
visitors she has, she is usually very happy whenever her boyfriend John came
to see her, whereas it is mostly not so when his friend Bill came instead. As
long as she can see John though, she is happy. But we happened found that
this is not the case on a day she has to get an injection. She hates it so much
that on a day she has one... Even if John came Mary is in a bad mood.
With respect to the normal situation, i.e. when it is not the day she gets an injection,

Mary is normally happy (not in a bad mood) if John came. I represent this normal

situation in the following way: Vw[w ENORMH came(John)(w) /\ happy(Mary)(w)].

31

This says that the for all worlds w, such that w is one of the expected or normal
kinds, the two propositions are always either both true or both not.

I will embed this piece of information about the normal situation into the meaning
of the gerundive morpheme in order to derive the non—cancelable meaning that reflects
the speaker’s knowledge about the relationship between the two event under a normal

circumstance.
(47) lGERl = Ap<s.t>/\9<s.t>Aw- p(w) A q(w) A Vw’lw’ ENORMr [p(w’) ... q(w’)ll

The gerundive morpheme takes two clausal arguments and conjoin them together:
p(w) /\ q(w). Furthermore, it tells us what kind of situations are normal (likely,
expected) with respect to what we know in the discourse. This is represented as
Vw'[w’ €NORM—> [p(w’) <—> q(w’)]]. It says that in the normal situations, both
of the propositions p and q have the same truth value (both true or both false).
In other words, there is a particular relationship between the two conjoined events.
Considering (45) we have set up in advance, w (call it an evaluation world) presumably
should be {wl}, where as w’ is {w1, w4}. It says that for all w’ that is in the set NORM
={w1,w4} (what are considered to be the usual situations, the norm, or whatever
assumed to be taken for granted, reflecting the discourse up to that point), the truth
value of John’s coming and Mary’s being happy are the same. That is, either both
true of both false. To see the relationship with w and w’, we notice that w’ = {w1, w4}
happens to include w = {w}, thus it indirectly says that w is in NORM. Let us say
that this gives us the information about how ‘expected’ the proposition p(w)/\q(w)
is. I suggest that the existence of information about the likeliness/expectedness of
p(w) /\ q(w) is why the gerundive construction has the non-cancelable meaning.

Let us derive the meaning for the gerundive construction without mo, using an

example (48).

32

(48) John-ga kite Mary-wa ureshii
John-NOM comeGER Mary-TOP happy.NONPAST

‘John came, and (so) Mary is happy.’

I assume the following structure (49) for simplicity, although it leaves room for dis-
cussion to which node the gerundive clause should be attached. In this structure, the

gerundive clause appears as a sentential modifier.

(49) (s,t)

/\

(st,st) (Sat)

/\/\

John come.GER
Mary-TOP happy.NONPAST

 

For now, I ignore the internal structure of the gerundive clause, especially about tense,
which should be worried about at some point in order to assign a relevant temporal

interpretation for the gerundive clause.20

(50) [[ John-NOM come.GER Mary-TOP happy.NONPAST ][
=came(John)(w) A happy(Mary)(w)/\

Vw’[w’ ENORM~> [came(John)(w’) <—> happy(Mary)(w’)]]

Two things have occured in the evaluation world w, John came and Mary is happy.
There is also another aspect of meaning: about how likely it is with respect to the
normal situation. For all w’, w’ is in the set of normal situations, and it is either both
events, John’s coming and Mary’s being happy, are true or both false. In our very
restricted understanding of the worlds, it should be wl and w4. The evaluation world
w is happened to be one of NORM, thus the present situation, that John came and
Mary is happy, is something that is compatible with what is considered to be normal

and is nothing surprising, hence expected.

 

20For example, we might need a free variable for tense that could be bound by the tense of the
main clause.

33

Now that we have derived the meaning for the gerundive construction, let us move
on to add the particle mo. As it will be shown in the following section 3.3, it is not
only mo that is responsible for the ‘even if’ meaning, but it is the combination of the

gerundive construction and mo.

3.3 On mo

Let us first confirm the basic property of mo. As the following data shows, the

particle mo by itself contributes to the additive meaning ‘also/too’.

(51) a. John-ga san-ji-ni keeki-o tabeta.
John-NOM three-o’clock-at cake-ACC eat.PAST

‘John ate cake at three o’clock.’

b. John-mo san-ji-ni keeki-o tabeta.
John-mo three-o’clock—at cake-ACO eat.PAST

‘Also John ate cake at three o’clock.’
(Presupposes: There is someone else who ate cake at three o’clock.)

c. John-ga san-ji-ni keeki-mo tabeta.
John-NOM three—o’clock-at cake-mo eat.PAST

‘John ate cake too, at three o’clock.’
(Presupposes: There is something else that John ate at three o’clock.)

d. John-go san-ji-ni-mo keeki-o tabeta.
John-NOM three-o’clock-at-mo cake-mo eat.PAST

‘John ate cake at three o’clock too.’
(Presupposes: John ate cake at some other time.)

(67a) shows the one without mo. Mo can attach to a subject noun (67b), an object
noun (67c), or an adjunct (67d).21 When mo appears in a sentence, it raises an addi—
tive meaning, presupposing an alternative situation. Depending on which constituent
it attaches to, the content of presupposition (‘something else’) changes as shown in

(67b—d).

21Mo cannot co—occurr with the nominative case marker (*John-ga-mo/ *John-mo-ga). On the
other hand, it is marginally acceptable with the accusative marker, but the order is fixed (??keeki-
o-mo/ *keeki—mo-o). Other particles such as dative, instrumental, and locative do not get deleted by
the attachment of mo (?*san-ji-mo), and the order of mo and those particles is fixed (*san-ji-mo-ni).

 

34

A morpho—syntactic constraint on mo is that it cannot be directly attached to
adjectives and verbs with a non-past or past tense morpheme.

(52) a. *John-ga keeki—o taberu/tabeta-mo.
John-NOM cake-ACC eat.NONPAST/eat.PAST-mo

Intended: ‘John also EAT/ ATE the cake.’
(Intended to presuppose: John does/did something with the cake; e.g.
Not just watching it, but he also ATE it.)

b. John-go keeki-o tabe-mo shita.
John-NOM cake-AGO eat-mo dO.PAST

‘John also ATE cake.’

(Presupposes: John does/ did something with the cake; e.g. Not just

watching it, but he also ATE it.)
Syntactically, it could appear adjacent to the stern of verb (52b). However, as it
is also the case in English, (52b) is grammatical and felicitous if there is a focus
prosodically marked on tabe-mo ‘eat’. If not, I would say it is a kind of negative
polarity item that requires negation as its licenser. I will leave the cases where mo
associates with focus and / or triggers polarity sensitivity. The point here is only to
show a morpho—syntactic constraint on mo such that mo cannot directly attach to
verbs/ adjectives that has a non-past or past morpheme. As we know, an exceptional
case is the gerundive form as it appears in the concessive conditional construction.

Futhermore, mo could take a clausal one as shown below.

(53) a. Mary-wa [John—go keeki-o tabeta]—to-mo ' itteita.
Mary-TOP [John-NOM cake-ACC eat.PAST]-COMP-mo say.PAST

‘Mary also said John ate cake.’
(Presupposes: Mary said something else.)

b. Mary-wa [John-go keeki-o tabeta]-ka-mo kiita.
Mary-TOP [John-NOM cake-ACC eat.PAST]-whether-mo ask.PAST

‘Mary also asked whether John ate cake.’
(Presupposes: Mary asked something else.)

The additive meaning arises similarly to (67b-d). It seems that the additive meaning

generally appears, as long as the constituent that mo attaches to has assigned a

35

structural nominative or accusative case or has a particle such as ni, to and ka.22:23

Ffom the fact above, we may take the additive meaning as a part of the core
meaning of mo. This sounds plausible, to recall the meaning of ‘even’ in Karttunen
and Peters (1979), in which it says that there are two kinds of presuppositional
meaning: the additive and the scalar meaning. Since our focus of study is ‘even
if’ and not ‘also/ too’, we need the scalar meaning as well. The scalar meaning was
responsible for the unexpectedness or unlikeliness meaning. However, where can we
find the expectedness meaning in mo? As a clarification, it is not the case that what
is said (e.g. ‘John’, ‘cake’, ‘at three o’clock’ etc.) is unexpected in (67) and (53). The
sentences only presuppose that there is something else other than what is said but
do not say which is more likely.

In some certain cases, however, mo expresses that what is said in the sentence
is unexpected, out of the norm, or off the standard. The unexpectedness meaning
appears not in an arbitrary way but in a particular structural environment. Typically,
one is where mo appears with an expression of amount as shown below in (54a—d),

and another is with the gerundive clause, i.e. the concessive conditionals, as we know.

(54) a. John-go keeki-o mi-kire tabeta.
John-NOM cake-ACC three-pieces eat.PAST

‘John ate three pieces of cake.’

b. John-go keeki-o mi-kire-mo tabeta.
John-NOM cake-ACC three-pieces-mo eat.PAST

‘JOhn ate three pieces of cake.’ (Eating 3 pieces of cake is surprisingly a lot!)

0. John-go juuhachi-jikan neta.
John-NOM eighteen-hours Sleep.PAST

‘John slept for 18 hours.’

 

22It is controversial deciding which grammatical category we assume especially for to and ka.
For example, some would say they are the head of CP, i.e. a complementizer like that in English,
but for some they may be a postpositional element. (Fukui & Sakai 2003:p.328-221) concerns the
controversy.

23Sentence (53b) may sound unnatural compared to (53a) to some speakers, although I found
no unnaturalness. My guess is that the unnaturalness in (53b) decrease when there is an overt
complementizer -to in between ka and mo.

36

d. John-go juuhachi-jikan-mo neta.
John—NOM eighteen-hours-mo sleep.PAST

‘John slept for 18 hours.’ (Sleeping 18 hours is surprisingly a lot!)

Unlike the nouns and clauses in (67, 53), numeral expressions that tells the amount
do not take a case or postpositional particle (54a, 0). When such amount expressions
take mo (54b,d), it indicates the speaker’s perspective that the amount mentioned
is off the standard and it is surprisingly a lot.24 The sense of surprise ‘a lot!’ that
the speaker implies with mo could be ‘a lot’ for the speaker, ‘a lot’ in general, or ‘a
lot’ for John based on what the speaker knows. Thus, what is said (‘John ate three
pieces of cake’ or ‘John slept for 18 hours’) is unexpected, off the standard or out of
the norm with respect to the normal amount.

The notion of unexpectedness is the most crucial part of the meaning of concessive
conditional. As Fujii (1989) has generalized as in (28), the gerundive-mo construction
expresses that the consequent is contrary to what is expected given the antecedent.
Thus, the sentence John-go kite-mo Mary-wa kigen-ga warui ‘Even if John came,
Mary would be in a bad mood’ expresses that Mary’s being in a bad mood is contrary
to what is expected to be normal given John’s coming. It is not the case that John’s
coming itself is unexpected or Mary’s being in a bad mood is unexpected. The
expectation is based on the relationship between the two conjoined propositions.

In the following subsection, I will propose a denotation for mo based on the two
fundamental properties of mo: the additive meaning, and the notion of unexpected-
ness. I tentatively distinguish the case in which mo that has these two properties from
the case in which mo only has the additive meaning. This distinction is not arbitrary,

since they appear in different structural conditions. As the contrast between (67, 53)

 

24There is another way to express the notion of off the standard, which is ‘surprisingly few’. In
such a case, there is another morpheme that expresses the less-than—expected meaning, which is
-shika. It is a negative polarity item, and with an overt negation, it means ‘only’ or ‘just’, e.g.
san-jikan-shika ne-nakatta(sleep-NEG.PAST). ‘I only slept for (no more than) 3 hours (which is
remarkably short time)’. Yoshimura (2007) investigates various focus sensitive elements including
-shika.

37

and (54) indicates, we may tentatively generalize that the unexpectedness meaning
arises when it attaches to a kind of adverbial expression that is not case marked or
marked with a postpositional particle. Therefore, my proposal of mo will account
for the cases where mo appears together with such adverbial expressions. Although
this proposal may be limited and far from the comprehensive understanding of mo,

it may still be a starting point of such a big project.

3.4 Deriving the Meaning of Gerundive-mo

Let us assume the following structure.

(55) (s,t)

/\

(st,st) (Sit)

(st,st) mo

/\ Mary-TOP mood-NOM bad.NONPAST

John come.GER

 

As before, the gerundive clause is of type (st,st), a kind of adverbial clause or a
sentential modifier. Suppose mo takes such an adverbial constituent as its argument.
The following shows my denotation for mo based on the structure above.

(56) ImoI =

Afisi,st>x\q<s,ty\w- q(wlA 39’l9’ at q /\ f (9')] /\ WW 75 q —* f (q) <19.er f (9’)]

First, mo takes an argument of type (st,st), a sentential modifier. It will be the
gerundive clause [John come.GER] in the case of concessive conditional construction.
However, mo would not assign the truth value to the content of the gerundive clause.
What it gives us is only about the main clause q in w. This is plausible, since the
speaker does not specify whether John’s coming is true/ actually happened in the
evaluation world w by saying ‘even if John came,...’. The rest of the denotation

gives more information about the likeliness/expectedness, following the Karttunen

38

and Peters’ (1979) approach. There are two kinds of presuppositional meaning: the
additive meaning (3g’[q’ 75 q/\ f (q’ )]) and the scalar meaning (Vq’[q’ 34 q —-> f (q) <h-kezy
f (q’ )]), or the comparative likeliness/expectedness meaning. This is in fact what we
have seen in the analyses by N akanishi (2006) and Yoshimura (2007) of the ‘even’-like
element demo. However, unlike their approach, the denotation (56) is not for demo
but it is for mo.

Now, let us combine the gerundive construction and mo together and see how the
proposed denotations derive the meaning of concessive conditionals compositionally.
As the first step, we have the gerundive construction with some sort of causal meaning,
which is based on what is presupposed to be normal in the discourse.

(57) [[John-NOM come.GER]
=[ApAqu.p(w) /\ q(w) A Vw’[w’ ENORM—e p(w’) = q(w’)]]([[John-NOM come]])

= Aqu.come(John)(w) /\ q(w) /\ Vw’[w’ ENORM-> come(John)(w’) <—-> q(w’) ]

Up to this point of derivation, it has a potential to raise the meaning of expectedness
(i.e. how normal it is) by virtue of having the gerundive construction. What mo does
is to take this gerundive clause, add the other possibilities of what could happen upon

John’s coming, and assign the likeliness/expectedness meaning.

(58) [[John-NOM come.GER-mo]

2: [[m0[[ ( [[John-N OM come.GER]])

A 3 ’ ’ A ’ A
=Af(st,st))‘q(s,t))‘w' q(w) q [q # q f(q )] ([[ John-NOM come.GER ][)
VQ'ICI' 7‘ q —> f ((1) <likely f (4')]

q(w) /\ Elq’[q’ 31$ q /\ [[John-NOM come.GER]](q’)]/\
=Aq(s,t))\w.
Vq’[q’ 7g q -—> [[John-NOM come.GER]](q) <11]:er [[John-NOM come.GER]](q’)]

Finally, the main clause is combined, filling in the slot ‘q’ above. This last step is just

to substitute [[Mary-TOP mood-NOM bad.NONPAST]] with ‘q’.

(59) [[JOhn-NOM come.GER—mo Mary-TOP mood-NOM bad.NONPAST]]

39

q(w) A Elq’[q’ 75 q A [[John-NOM come.GER][(q’)]A
=Aq<sjt>Aw
‘v’q’[q’ 75 q —+ [[John-NOM come.GER]](q)<1,-kezy [[John-NOM come.GER](q’)]

<[[ Mary-TOP mood-NOM bad.NONPAST [D

Since the final step (59) takes a space to fully spell out, let me explain by breaking it

up into three parts as in (A—C) below, which would roughly correspond to the truth

conditional meaning, the additive meaning, and the scalar meaning respectively.

(A)

TRUTH CONDITIONAL MEANING

[Aqu.q(w)](|[Mary-TOP mood-NOM bad.NONPAsr]])

=[[Mary-TOP mood-NOM bad.NONPAST]] (w)

From (46),

bad-mood(Mary)(w) = Mary is in a bad mood in the evaluation world w.

The core part of the meaning of the concessive conditional sentence says that

Mary is in a bad mood in the evaluation world w.

ADDITIVE MEANING

Mary-TOP mood—NOM

AquEiq’[q’ 75 q A [[John-NOM come.GER]](q’)]
bad.NONPAST

=3q’Aw[q’ 7é bad-mood(Mary) A [[John—NOM come.GER]](q’)]
From (57),

A 3 I q’ 75 bad-mood(Mary) A come(John)(w) A q’(w)A
w 9
Vw’[w’ ENORM—> come(John)(w’) <—> q’(w’)]]

The additive meaning says that there is another consequence q’ which is dif-
ferent from q (=‘Mary is in a bad mood’). By virtue of having the gerundive
morpheme as I proposed in (47), the additive meaning also introduces the de-

gree of expectation, or how likely it is for the two events (John’s coming and q’)

40

' to co—occur. Assuming the simplified nature of worlds in (46) that reflects the
current understanding of the discourse, q and q’ can be represented as different
sets of worlds {w2, 104} and {w1, 1113} respectively. The former is in which Mary

is in a bad mood and the latter is in which she is happy.

(C) SCALAR MEANING
[ Aqu.Vq’[q’ 74 q —> [[John-NOM come.GERfl(q) <likely [[John—NOM come.GER][(q’)]
([[ Mary-TOP mood-NOM bad.NONPAST ][)

q’ aé [[Mary-TOP mood-NOM bad.NONPAST]] —>

= Aw.Vq’ [[John-NOM come.GER][([[Mary—TOP mood-NOM bad.NONPAST]])

 

<1“,er [[John-NOM come.GER]](q’)]

 

From (47) and (50),

q’ 75 [[Mary-TOP mood-NOM bad.NONPAST]] —>

come(John) (w”) A bad-mood(Mary)(w”)A
Aw.Vq’ Vw’[w’ ENORM—> [came(John) (w’) H bad-mood(Mary)(w’)]]

come(John)(w”) A q’(w”)A
<likely

 

 

 

 

Vw’[w’ ENORM—I [come(John)(w‘) H 9011,)”

The scalar meaning that is derived by mo is represented as a comparison of
expectedness, as it has been claimed elsewhere. However, it is not the case that
mo itself assigns the degree of expectedness to the antecedent clause or to the
consequent clause. In my analysis, the notion of expectedness is provided by
the gerundive morpheme. What mo compares is the expectedness of the two
different gerundive constructions, i.e. the likeliness of ‘John came and Mary is
in a bad mood’ and ‘John came and Mary is not in a bad mood’ with respect

to normalcy.

The representation of the meaning we have achieved follows the pervious works

we have covered in section 1-2 in that it has the additive meaning and the expected-

41

mess/likeliness scalar meaning. I derived this notion of unexpectedness not from the
particle mo by itself but rather from the interaction between mo and the gerundive
construction that introduces the notion of expectedness with respect to normalcy. As
a result, the scalar meaning in my analysis is not anymore a comparison of how likely
it is between the consequent proposition q and the alternative consequent proposi-
tion q’. The concessive conditional meaning is described in terms of how likely the
consequent clause is given the condition of John’s coming compared to all other con-
sequences given the same condition. This indicates that my version of mo has to take
an argument that already bears a notion of expectedness/likeliness in order to achieve
the unexpectedness meaning. In the remaining part of the thesis, I will consider some

predictions and implications of my approach presented here.

42

4 Summary and Some Implications

4.1 The Gerundive Construction Revisited

I proposed the following denotation (60) for the gerundive morpheme.
(60) [[GER]] = Ap(s‘t))\q(s,t)/\w. p(w) A q(w) A Vw’[w’ ENORM—+ [p(w’) +—> q(w’)]]

The basic property of the gerundive construction was the ‘and’ conjunction in Japanese,
which is represented in the first part of the denotation (p(w) A q(w)). It takes the two
arguments p and q and states the truth condition with respect to the evaluation world
w. The rest of the denotation illustrates what the normal situation should be like. In
the normal situation w’, according to the discourse context and what we know, both
p and q co—occur or neither does. Because it is said that (p(w) A q(w)) is one of such
normal situations, I suggested that this results in the expectedness meaning.

The proposed denotation can account for the basic conjunction property as well

as the special relationship between the two conjoined propositions.

(61) John-ya kite Mary-ma ureshii.
John-NOM come.GER Mary-TOP happy.NONPAST

‘John came, and (so) Mary is happy.’

The gerundive construction says that John came and Mary is happy. It not only says
that those two things occurred, but it also tells us that it is somehow expected for
Mary to be happy upon John’s coming based on the speaker’s knowledge about the
normal situation. I suggested this is reflected as a kind of causal meaning ‘so’. I have
said in section 3.1 that the expectedness meaning is not cancelable when the sentence
involves a topic marking on the subject of the tensed clause as in (61), which I called
the subordination type gerundive construction.

The present analysis only focused on the subordination type gerundive construc-

tion. However, the fact is that mo can appear in either type of the gerundive con-

43

struction, either subordination or coordination. Let us recall the examples of the
coordination type gerundive sentences introduced in (37c, d), as repeated below.
(62) c. [John-go kite] [Mary—ya kaeru].

John-NOM come.GER Mary-NOM went home.NONPAST
‘John comes and Mary goes home.’

(As an answer to: What usually happens in the office while I am out?)

(i. [John-wa kite] [Mary-wa kaeru].
John-TOP come.GER Mary-TOP went home.NONPAST

‘John comes and Mary goes home.’

(As an answer to: Do both John and Mary join the reception after the talk?)

In these cases above, no particular relationship between the two conjoined events is
presupposed. They are just a list of events.

The following is the two sentences with mo attached.
(63) a. [John-go kite]-mo [Mary-ya kaeru].

John-NOM come.GER-mo Mary—NOM went home.NONPAST
Roughly: ‘Although John comes, Mary goes home.’

b. [John-wa kite]-mo [Mary-wa kaeru].
John-TOP come.GER—mo Mary-TOP go home.NONPAST

Roughly: ‘Although JOHN might come, MARY would go home.’

The translations in (63a) and (63b) are mine and they may leave room for discussions.
Let us tentatively assume that ‘although’ is the most accurate translation and that
it is one of concessive expressions. In any event, sentences (63a, b) show that the
gerundive mo construction is not limited to the subordination type but also applicable
to the coordination type. This may suggest that my analysis on the gerundive-
mo should be extended to the coordination type. However, the connection between
the conjuncts, John’s coming and Mary’s going home, may not be the same as the

gerundive-mo of the subordination type. The relationship of the two clauses may not

44

be conditional, but it may rather be contrastive.25 Thus, what is unusual/ unexpected
is not that Mary’s going home given John’s coming, or in other words, it is not that the
speaker thinks Mary would go home if John comes based on the normal situations,
but rather, the speaker is making a contrast between the two events. I will leave
room for discussion on the difference between ‘even if’ and ‘although’. It may be
an interesting discussion if there is some diagnostics to distinguish the two kinds of
concessive meaning.

To give a comprehensive analysis of the gerundive construction in Japanese, one
should consider a question of kinds of clausal relationship that needs to be assigned
for the gerundive conjunction. It is not plausible to implement a piece of informa-
tion about the normal situation Vw’[w’ €NORM—> [p(w’) H q(w’)] for any type of
gerundive form, because that was originally supposed to derive the causal meaning
between the two conjoined propositions that arises in the subordination type and
not necessarily in the coordination type. Thus, in order to account for the gerun-
dive construction in general, one has to look into various kinds of clausal relationship
other than the causal one, for example, the temporal sequential meaning I did not
cover in the present analysis. A hope is to construct a denotation for the gerundive
form that can account not only for the causal meaning, but also for the temporal
sequential meaning for example (cf. Hasegawa (1996a,b) for other possible mean-
ings). As Watanabe (1990) argues, deriving the temporal sequential meaning would
require more careful classification of verbs with respect to the temporal and aspectual
properties, e. g. telicity.

Furthermore, adopting a formal tool for the notion of topichood may help con—
structing a theory of the gerundive conjunction, since the topic marking was the key to

distinguish the subordination type gerundive conjunction (in which a non-cancellable

 

25By ‘contrastive’, I mean a sense in which but might represent. Indeed, the ‘even’ expression
demo as a sentential conjunction (and as an independent word), it functions like ‘but’, e.g. John-go
kita. Demo Mary-wa kaetta. ‘John came. But, Mary went home.’

45

meaning arises) and the coordination type gerundive conjunction (in which the ad-
ditional meaning is cancelable). Further research of the Japanese concessive condi-
tional may lead us to investigate broader issues of the semantics-pragmatics interface

of clausal conjunctions.

4.2 M o Revisited

In some cases, including concessive conditionals, mo expresses unexpectedness.

(64) a. John-go juuhachi-jikan-mo neta.
John-NOM eighteen-hours—mo sleep.PAST

‘John slept for 18 hours.’ (Sleeping 18 hours is unusually a lot!)

b. John-ga konshuu keeki-o san-kai-mo tabeta.
John-NOM this week cake-ACC three-times-mo eat.PAST

‘John ate cake three times this week.’
(Eating cake three times a week is surprisingly frequent!)

In (64a), in- addition to the truth conditional meaning that says John has slept for
18 hours, the sentence also expresses that the speaker thinks sleeping 18 hours is a
lot in general, for the speaker, or for John based on what the speaker knows about
John’s normal lifestyle. Similarly, three times a week is unusually a lot as a frequency
of John’s cake eating in (64b).

To capture the notion of unexpectedness in the concessive conditional, I proposed
the following denotation for mo.
(65) [m0] = Af(3t,st))\q<s,t)/\w. q(w)A

39’l9’ aé q A f (9’)] A Vq’ICI’ 7e 9 -> f (q) <91.er f (9’)]

Mo first takes an argument of type (st,st), a kind of modifier. Then, without af-
fecting the meaning of the second argument q, it gives two additional meanings: the
additive meaning (3q’[q’ 94 q A f (q’ )]) and the comparative meaning with respect to

likeliness/expectedness (‘v’q’[q’ -+‘ q —-> f (q) <1“,er f (q’ )])

46

In the concessive conditional construction, the gerundive clause (John-go kite)
fills the argument variable f(sm). When mo applies, it introduces possible alterna-
tive situations q’, which is different from the situation Mary’s being in a bad mood.
Assuming the simplified world as illustrated in (46), q’ is the situation in which Mary
is happy. Then, it puts the two conjoined propositions (‘John came and Mary is
in a bad mood’ and ‘John came and Mary is happy’) on the likeliness/expectedness
scale, and says that what is said (‘John came and Mary is in a bad mood’) is less
likely/ expected than all (‘John came and Mary is happy’).

My denotation for mo may be applied to the example (64) to account for the
meaning ‘it is surprisingly/ unusually a lot’, assuming that the amount/ frequent ex-
pressions have the same semantic type as a gerundive clause. However, it is not
the case that any kind of a non-case-marked or a postposition-free phrase can take
mo. For example, ‘sometimes’ and ‘(sleep) deeply’, which should not take a case or

postposition particle, cannot have mo.

(66) a. *John-ga gussuri-mo neta.
John-NOM deeply-mo sleep.PAST

Intended: ‘John slept deeply.’ (It was an unusually deep sleep!)

b. *John-ga tokidoki—mo keeki-o tabeta.
John-NOM sometimes-mo cake-ACC eat.PAST

Intended: ‘John sometimes ate cake.’
(‘Sometimes’ is surprisingly a lot because John rarely eats cake!)

We may tentatively say that the phrase that mo attaches to has to have a specific
amount such as ‘for 18 hours’ and ‘three times’ in order to express the unexpectedness.
However, in other cases, mo only expresses the additive meaning and not neces-

sarily the expectedness scalar meaning.

(67) a. John-mo san-ji-ni keeki-o tabeta.
John-mo three-o’clock—at cake-ACC eat.PAST

‘Also John ate cake at three o’clock.’
(Presupposes: There is someone else who ate cake at three o’clock.)

47

b. John-go san—ji~ni-mo keeki-o tabeta.
John-NOM three-o’clock-at-mo cake-mo eat.PAST

‘John ate cake at three o’clock too.’
(Presupposes: John ate cake at some other time.)

c. Mary-wa [John-go keeki-o tabeta]-to-m0 itteita.
Mary-TOP [John-NOM cake-ACC eat.PAST]-COMP-mo say.past

‘Mary also said John ate cake.’
(Presupposes: Mary said something else.)

d. Mary-wa [John-go keeki-o tabeta]-ka-mo kiita.
Mary-TOP [John-NOM cake-ACC eat.PAST]-whether—mo askpast

‘Mary also asked whether John ate cake.’
(Presupposes: Mary asked something else.)

A different denotation for mo may be required to account for these cases. A lexical
ambiguity approach will assume that there are two mos in the lexical entry. However,
such an approach may undermine the observation that the meaning of mo (whether
it is only additive or additive with unexpectedness) correlates with the structural
environment it appears. The mo that appears with a phrase that has a structural case
or postpositional particle introduces the additive meaning and no further than that.
On the other hand, the other mo appears with a subordinate type gerundive clause
or a (specific) amount expression, and it derives the additive and the expectedness
meaning. The proposed denotation for mo should specify the latter mo, and not the
one that appears on a phrase that has a case or postpositional particle. It might be
possible to give a unified account on mo, however it is diflicult to do so unless we
incorporate a theory of case/postposition marking, because that is the factor that
distinguishes the two different mos.26

Another important aspect of mo, which has been well studied but not in this the-

sis, is its interaction with wh-phrases. For example, by substituting dare ‘who’ with

John in the gerundive—mo construction results in ‘whoever’ as shown below.

 

26The relationship of case assignment and mo has been notice elsewhere. For example, Kakegawa
(2003) offers an analysis that treats mo as a case particle.

48

(68) [dare-9a kite]-mo Mary-wa kigen-ga warui.

who-NOM come.GER-m0 Mary—TOP mood-NOM bad.NONPAST

‘Whoever came (No matter who came), Mary would be in a bad mood.’
Sentence (68) shows that the property of wh—phrases in Japanese and in English

are different, since no substitution of a wh—phrase is available in an English even if

sentence.
(69) *Even if who came, Mary would be in a bad mood.

Since no wh-phrase can occur in the conditional if clause in English, a syntactic
theory of wh-phrases may independently account for the ungrammaticality of (69).
As for wh-phrases in Japanese, there have been various syntactic and seman-
tic accounts proposed in order to account for their distribution and interpretations.
Japanese wh-phrases are also called ‘indeterminate pronouns’ following Kuroda’s
(1965) terminology (Shimoyama 2006), since the interpretation of a wh-phrase such
as dare is not determined until it is ‘licensed’ by particular particles such as mo in
(68), the sentence final question marker no as in (70a), or ka/mo that appears adja-
cent to the wh-phrase (70b, c). Without theses licensers, a wh—phrase alone cannot
have an interpretation and the sentence is ungrammatical as shown in (70d).
(70) a. [dare-go kite] Mary-wa kigen-ga warui no?

WhO—NOM come.GER Mary—TOP mood-NOM bad.NONPAST Q
Who x such that x came and Mary is in a bad mood?

b. [dare—ka-ga kite] Mary-wa kigen—ga warui.
who-ka-NOM come.GER Mary—TOP mood-NOM bad.NONPAST
‘Someone came and Mary is in a bad mood.’

c. [dare-mo-ga kite] Mary-wa kigen-ga warui.
who-NOM come.GER Mary—TOP mood—NOM bad.NONPAST
‘Everyone came and Mary is in a bad mood.’

d. * [dare-ga kite] Mary—wa kigen-ga warui.

who-NOM come.GER Mary-TOP mood-NOM bad.NONPAST

49

In order to account for the dependency relationship of wh—phrases and the licensing
particles, various approaches have been proposed. For example, Nishigauchi (1991)
proposes a wh-movement analysis at LF, and Watanabe (1992) argues for an overt
(non-LF) movement analysis of null operators. Shimiyama (2006) and Kratzer &
Shimoyama (2002) present a non-movement analysis based on Hamblin’s (1973) se-
mantics for wh—phrases as sets of alternatives. Although their approaches are different,
one common view is that mo is a universal quantifier that binds variables introduced
by wh-phrases. While such a view explains the meaning (70c), it may not be ade-
quate to account for the gerundive-mo construction that does not necessarily involve
a wh-phrase. However, a unified account for mo should be flexible enough to explain
both cases in which the sentence has a wh—phrase and in which it does not. In this
respect, a further investigation of mo implies a better understanding of wh-phrases
in Japanese.

Furthermore, the research on wh-phrases and concessive conditionals across lan-
guages may also shed light on some empirical and theoretical connections between
conditionals and interrogatives. According to Gawron (1999), the English wh-ever
sentences are indeed examples of a conditional construction. He calls the wh-ever
construction the ‘Universal Concessive Conditional (UCC)’, and argues that the an-

tecedent clause in the UCC is closely related to question formation.

4.3 Other General Issues

I would like to point out three more related issues with respect to the gerundive-mo
construction and concessive conditional in general. First, let us recall the so—called
consequent entailment problem. In short, it was a contrast shown in the two kinds of
even if sentences.

(71) a. Introduced if

Even if the bridge were standing, I would not cross.

50

H I would not cross.
b. Standing if

Even if he drank just one ounce, she would fire him.

H She would fire him if he drank some.
According to Pollock (1976) and Bennett (1982), the consequent in (71a) must be
true independent of the truth of the antecedent clause. The speaker implies his/ her
strong will of not crossing, thus ‘I would not cross if the bridge were standing or not’.
On the other hand in (71b), it is not that the truth of the consequent ‘she would fire
him’ is independent of the truth of ‘if he drank just one ounce’. It is not intuitive to
interpret it as ‘she would fire him if he drank one ounce or not’.

My analysis in section 3 on the Japanese concessive conditional did not concern

about the consequent entailment problem. Interestingly, though, the contrast is ap-

parent in Japanese.

(72) a. Introduced if

Hashi-ga kakatte-ite-mo watashi-wa watara-nai
Bridge-NOM standing-be.GER-mo I—TOP cross-not.NONPAST

‘Even if the bridge were standing, I would not cross.’
b. Standing if

John-ga honno hito-kuchi-de-mo nondara bosu-wa
John-NOM just one—sip-COP.GER—m0 drink.COND(if) boss-TOP

kubi-ni suru
fired-DAT dO.NONPAST

‘Even if John drank just a sip, the boss would fire him.’
The structure of Introduced if (72a) is the one I accounted for in section 3, i.e. the
combination of the gerundive clause and mo. On the other hand, the structure of
Standing if (72b) is different in that, although there is a similar construction (the
gerundive copula and mo), it is embedded in an antecedent clause with an overt

conditional morpheme (...nondara, the perfective tara conditional construction, cf.

51

5a).27 This brings us back to Bennett’s (1982) explanation of the contrast. Ac-
cording to him, the Standing if does involve the conditional construction but the
conditionality in the Intorduced if is “a result of the operation of even” (Bennett
1982: p.411). Indeed, to see the Japanese counterparts, the Standing if is built upon
a conditional construction (72b), but the conditionality in the Introduced if (72a) is
morphologically unclear.

In this respect, my analysis of the Japanese concessive conditional was mainly
about the Introduced if conditional only. This was the type of even if sentences that
Guerzoni & Lim (2007) accounted for by introducing an invisible morpheme AFF,
which is not at all necessary in my analysis. As the data (72) shows, the other type
of concessive conditional, the Standing if conditional does have the same elements
as the Introduced if conditional: gerundive morpheme de, the copula verb, and mo.
However, one has to account for the morphologically overt conditional as well and the
interaction of de-mo and the overt conditional in order to account for the Standing
if conditional.

Accordingly, since the source of conditionality is different in (72a) and (72a), this
might raise a question on the English if construction. For example, although we see if
in both (71a) and (71b), is it the case that both types of constructions share the same
structure? With respect to the structure of the English if sentences, Haegeman (2003,
2006) argues that there are two different structures of if-conditionals in English:
either the antecedent clause being IP adjoined to the main clause, or it being CP
adjoined.

A second thing to be mentioned is a link between conditionals and modality. One
of the modal expressions in Japanese has a parallel construction with the one of con-

cessive conditional.

 

27I thank Yasutada Sudo again for pointing this out.

52

(73)

a. With mo: Concessive Conditional ‘Even if’

[John-go kite-mo] Mary-wa kigen-ga warui.
John-NOM come.GER-mo Mary-TOP mood-NOM bad.NONPAST
‘Even if John came Mary would be in a bad mood.’

b. With mo and iidesu—yo: The Deontic ‘may’

[John-ga kite-mo] iidesu-yo.
John-NOM come.GER-mo good.NONPAST-yo
Lit. ‘It is also ok if John come.’

‘John may(is allowed to) come.’

The deontic ‘may’ in (73b) shows exactly the same antecedent clause (John-go kite)

as the one in (73a). As Kratzer (2002, 1986) argues, there seems to be a link be—

tween modality and the natural language conditional in that the antecedent clause

of conditionals restrict modals. However, we might need to reconsider whether the

notion of normalcy, which is a key to derive the unexpectedness in my analysis, is

also adequate for accounting for the deontic meaning of permission.

A third thing to be mentioned here is about some variations. In this thesis, I have

only introduced one version of concessive conditionals (74a). There is also a colloquial

(casual, less formal) version (74b). (740) is an example from Kansai (Western) dialect.

The last two involve complementizers as glossed with ‘COMP’ below.

(74)

a. ‘Standard’

John-go kite-mo Mary-wa kigen-ga warui.
John-NOM come.GER-m0 Mary-TOP mood-NOM bad.NONPAST
‘Even if John came, Mary would be in a bad mood.’

b. Colloquial

John-go kita-tte Mary-wa kigen-ga warui.
John-NOM come.PAST-COMP Mary-TOP mood-NOM bad.NONPAST

‘Even if John came, Mary would be in a bad mood.’
c. Kansai-dialect

John-ga kita-ka-te M ary-wa kigen-ga warui.
John-NOM come.PAST-ka-COMP Mary-TOP mood-NOM bad.NONPAST
‘Even if John came, Mary would be in a bad mood.’

53

The notable differences are summarized below.

(75) a. The verb of the antecedent clause in (74a) must take the gerundive form,
whereas it is the past tense form in (74b, c).

b. There must be a complementizer tte/te in (74b, c), whereas (74a) does
not allow a complementizer.

c. No mo in (74b, c). In (74c), however, takes the question particle ka.

Like the ‘standard’ version in (68), the other two versions also demonstrate the

property of dependency relationship with wh-phrases.

(76) a. Colloquial

Dare-go kita-tte Mary-wa kigen-ga warui.
Who-NOM come.PAST—COMP Mary-TOP mood—NOM bad.NONPAST
‘Whoever came, Mary would be in a bad mood.’

b. K ansai-dialect

Dare-go kita-ka-te Mary-wa kigen-ga warui.

Who—NOM come.PAST-ka—COMP Mary-TOP mood-NOM bad.NONPAST

‘Whoever came, Mary would be in a bad mood.’
What is interesting is that there is no mo in (76a, b) which is said to be a potential
licenser for the wh—phrase. As we saw in (70a—d), the interpretation and grammat-
icality of wh-phrases depends on particles such as ka and mo. Sentence (76b) does
have the particle ka before the complementizer, however, it is not clear why it is ka,
which is normally considered as a question marker, and not mo as in the ‘standard’
version. A further investigation of concessive conditional should account not only for
the properties of those particles, but also for the licensing condition of wh-pharses
apart from the existence of such particles.

If the meanings of all the variations in (74a-c) should be represented exactly in

a same way, we should also be able to derive the meaning of concessive conditional
without the referring to the gerundive construction and mo. In order to account for

the other versions of concessive conditionals as exemplified in (74b, c), one has to

54

account for the obligatory past tense morpheme on the antecedent clause, the com-
plementizers, and the question particle ka in the (74c). A further research on these
dialectal variations would reveal a range of morphological realization that conven-
tionally derives the concessive conditional meaning. It may eventually lead us to
understand how the meanings of concessiveness and conditionals are constructed and

lexicalized across languages.

55

5 Conclusion

In order to account for the Japanese concessive conditional, I presented a compo-
sitional analysis that concerns an interaction of the gerundive construction and the
additive particle mo. In deriving the meaning of the gerundive construction, I argued
that the construction has a meaning that reflects the speaker’s knowledge about the
relationship between the two conjoined events under a normal circumstance, in addi-
tion to the truth—conditional meaning. This explains how a certain type of gerundive
construction not only represents the combination of more than one events but also
introduces the notion of likeliness or expectedness. Then, I proposed a denotation for
mo that derives the notion of unexpectedness based on the expectedness derived by
the gerundive clause.

I pointed out that there is a limitation on kinds of data that the proposed deno-
tations can account for. Although my denotation for the gerundive morpheme can
account for the meaning of the gerundive construction of a particular type, namely
the subordination type, it may not be adequate for the one of coordination. As for
the denotation for mo, it may not be adequate to explain the meaning of mo that
appears as a particle that attaches to a case-marked phrase or a phrase with a post-
positional particle. To overcome the limitations and to extend the present analysis of
the gerundive form and mo, I suggested that one should also look into theories about
topichood, case-marking, and postposition marking. I also pointed out that further
research may reveal theoretical connections among concessive conditionals and other

kinds of constructions such as modal sentences and questions or interrogatives.

56

REFERENCES

Baker, Stephen. 1994. The Consequent—Entailment Problem for EVEN IF. Linguistics
and Philosophy, 17, 249—260.

Bennett, Jonathan. 1982. Even If. Linguistics and Philosophy, 5, 403—418.

Fujii, Seiko. 1989. Concessive Conditionals in Japanese: A Pragmatic Analysis of the
SI-TEMO S2 Construction. In: Hall, K. (ed), Proceedings fo the Annual Meeting
of the Berkely Linguistics Society, vol. 15.

Fukui, N aoki, & Sakai, Hiromu. 2003. The Visibility Guideline for Functional Cate-
gories: Verb Raising in Japanese and Related Issues. Lingua, 113, 321—375.

Gamut, L. T. F. 1991. Logic, Language, and Meaning - Vol.1 Introduction to Logic.
The University of Chicago Press.

Gawron, Jean Mark. 1999. Universal Concessive Conditional and Alternative NPs in
English.

Giannakidou, Anastasia. 2007. The Landscape of EVEN. Natural LAnguage and
Linguistic Theory, 25(1), 39—81.

Guerzoni, Elena, & Lim, Dongsik. 2007. Even if, factivity and. focus. In: Puig—
Waldmiiller, Estela (ed), Proceedings of Sinn und Bedeutung 11.

Haegeman, Liliane. 2003. Conditional Clauses: External and Internal Syntax. Mind
and Language, 18(4), 317-339.

Haegeman, Liliane. 2006. Conditionals, factives and the left periphery. Lingua, 116,
1651-1669.

Hasegawa, Yoko. 1996a. The (Nonvacuous) Semantics of TE—linkage in Japnaese.
Journal of Pragmatics, 25, 763—790.

Hasegawa, Yoko. 1996b. Toward a Description of T e-linkage in Japanese. Oxford
University Press. Chap. 3, pages 55—75.

Heim, Irene, & Kratzer, Angelika. 1998. Semantics in Generative Grammar. Blackwell

Publishers.

Kakegawa, Tomomi. 2003. The analysis of Japanese case particles as determiners.
Ph.D. thesis, Michigan State University.

Karttunen, Lauri, & Peters, Stanley. 1979. Conventional Implicature. Syntax and
Semantics, 11, 1—56.

Kratzer, Angelika. 1986. On Conditionals. Chicago Linguistic Society, 22.

57

Kratzer, Angelika. 2002. The Notinal Category of Modality. Chap. 12, pages 289-321
of: Paul Portner, Barbara Partee (ed), Formal Semantics: The Essential Reading.
Blackwell Publishers.

Kratzer, Angelika, & Shimoyama, Junko. 2002 (March). Indeterminate Pronouns:
The View from Japanese. In: the 3rd Tokyo Conference on Psycholinguistics.

Kuno, Susumu. 1973. The Structure of the Japanese Language. Cambridge, MA: MIT
Press.

Levinson, Stephan C. 2000. Presumptive MeaningL The Theory of Generalized Con-
versational Implicature. MIT Press.

Lycan, William G. 1991. Even and Even If. Linguistics and Philosophy, 14, 115—150.

Nakanishi, Kimiko. 2006. Even, only, and Negative Polarity in Japanese. In: the
Proceedings of SALT 16.

Pollock, John L. 1976. Subjunctive Reasoning. Dordrecht, The Netherlands: D. Reidel
Publishing Company.

Potts, Christopher. 2005. The Logic of Conventional Implicature. Oxford University
Press.

Potts, Christopher. 2007. Into the conventional—implicature dimension. Philosophy of
Compass, 2(4), 665—679.

Rooth, Matt. 1997. Focus. Chap. 10 of: Lappin, Shalom (ed), The Handbook of
Contemporary Semantic Theory. Blackwell Publishers.

Shimoyama, Junko. 2006. Indeterminate Phrase Quantification in Japanese. Natural
Language Semantics, 14(2), 139—173.

Stalnaker, Robert. 1978. Assertion. Syntax and Semantics, 9, 315—355.

Watanabe, Akira. 1992. Subjacency and S-Structure Movement of Wh-in—situ. The
Journal of East Asian Linguistics, 1, 255—291.

Watanabe, Akira. 2004. The Genesis of Negative Concord: Syntax and Morphology
of Negative Doubling. Linguistic Inquiry, 35(4), 559—612.

Watanabe, Fumio. 1990. A Semantic Analysis of the Japanese Suspensive—Connective
Form: Meaning and Usage of the TE—Form. Gengogakurongyoo, 9, 79—92.

Yoshimura, Keiko. 2007. Focus and Polarity: ‘even’ and ‘only’ in Japanese. Ph.D.
thesis, University of Chicago.

58

 

 

 

ulv.

4

 

 

 

 

 

 

SII'V LIBRARIE

Ulll‘.‘EH

STATE

'ICHIGAH

l.

1293 02956 5185

3

 

3.
I... :1 .
3.1.1 ..

.. ... ...; ...
11y}...
: - N.

..T .
pinum.
.. . ...

9:.
I