An Event-Based Semantics for Japanese Emphatic Particles
Akira Ishikawa
Department of English and Area Studies, Sophia University,
[email protected]
Inspired by Herburger (2000), I will develop an event-based semantics for
Japanese emphatic particles which can address the issue of the mechanism of
association with focus involving the emphatic particles. The proposed semantics
makes use of and modifies Herburger's three key ideas: events as basic entities,
decomposition of predicates into subatomic formulas, and separation of
backgrounded and foregrounded information. I will show the resultant DRS
notation can describe the meaning of the emphatic particles very explicitly in
terms of the discourse referents involved in the described and compared events,
and also account for the interaction of emphatic particles and semantic particles in
a perspicuous way
1. Emphatic particles
Japanese emphatic particles I will take up in this paper correspond to
English adverbs only, even, and also, among others, of which only could
be arguably a determiner when it occurs in prenominal position. Japanese
has three kinds of postnominal particles constituting noun phrases: kaku
zyosi case particles (ga, o, ni, etc.), huku zyosi supplementary particles
(dake 'only', made 'as far as', bakari 'only', etc.), and kakari zyosi lead-in
particles (wa 'TOPIC', mo 'also', sae 'even', sika 'except', etc.). In the recent
Japanese linguistics literature, supplementary and lead-in particles are
grouped together to form a class known as toritate zyosi particles for
taking up (entities). This appellation of the new class is taken to mean that
such particles mark different modes of conceptualization of the entity
taken up for predication (Miyazima and Nitta, 1995: 278). Semantically,
they all have to do with focus, as some of their semantic counterparts in
English suggest. In this paper, I will refer to the class of toritate zyosi as
emphatic particles (EP, henceforth).
The main concern of this paper is to suggest a systematic way to capture
the meanings of Japanese EPs, which is inspired by the treatment of focus
Journal of Cognitive Science 3: 43 - 63, 2002.
c 2001 Institute for Cognitive Science, Seoul National University.
○
44
Akira Ishikawa
and quantification by Herburger (2000) but offers a considerably different
solution to the problem of focus and the meaning of Japanese EPs. I will
show a DRT semantics of EPs, which uses thematic roles as predicates and
is many-sorted in the sense of having variables for events as well as those
for individuals. This semantic treatment is an attempt to reveal the role
played by focus in the overall workings of Japanese post-nominal particle
affixation systems, and not just that of EP affixation.
2. EPs and domain selection
The word focus is used in two different, but unifiable senses in
semantics. In one sense, it is what explains question-answer congruence as
shown in (1)
(1) a. Who did John introduce to Sue?
b. John introduced [Bill] F to Sue.
c. #John introduced Bill to [Sue] F.
In the other sense, focus is the locus of the scope of certain adverbs and
adverbial expressions (focus-sensitive operators) such as only, even, and
also.
(2) a. John only introduced [Bill] F to Sue.
b. John only introduced Bill to [Sue] F.
In alternative semantics theories of focus, as proposed by Rooth (1985,
1992, etc.), the two senses of focus can be uniformly characterized by
means of focus semantic values and association with focus. The focus
semantic value of a sentence like (1b) is a set of alternative propositions
such as (3) which might equally well serve as the answer to the question if
things were differently arranged.
(3) { {John introduced John to Sue.}, {John introduced Bill to Sue.}, {John
introduced Mike to Sue.}, ...} ∈
An Event-Based Semantics for Japanese Emphatic Particles
45
Association with focus captures the felicity of (1b) given (1a), as
opposed to the infelicity of (1c), which has a different set of alternative
propositions {p = ‖John introduced Bill to d ‖ ‖d‖∈ D} which
does not correspond to (1a), or the focus semantic value of (1b). Basically,
the process of association with focus consists of the identification of a
contextually constrained set of alternative propositions given a sentence
containing a focus operator (e.g. (3) in a more contextually constrained
form) and the confirmation of the proposition corresponding to the given
sentence as a (non-unique) member of the set.
As we saw in (1) and (2), association with focus in English is typically
determined by the locus of focus accent in the sentence.1 By contrast, in
the case of Japanese EPs, association with focus is dictated by the
constituent structure in which they occur rather than focus accents. Since
EPs are directly attached to the nominal to be focused, cancellation of
focus effects due to special contextualization is not possible.
(4) a. kome-o saibai-suru hitotati-wa kome-dake-o tabe-ru
rice-ACC grow-PRES people-TOP rice-ONLY-ACC eat-PRES
`People who grow rice only eat rice. '
b. kome-o saibai-suru hitotati-wa kome-wa tabe-ru-dake da
rice-ACC grow-PRES people-TOP rice-TOP eat-PRES-ONLY COPULA
`People who grow rice only do the eating of rice. '
However, Japanese EPs do need an adequate theory of domain selection.
In any alternative semantics theory of focus, the focus semantic value of a
clause containing a focus operator such as (3) needs to be meaningfully
restricted as pointed out by Schwarzschild (cf. Kadmon(2000, p. 307)).2
A very conservative measure proposed in Ishikawa (2001) uses the
concept of directly comparable proposition (DCP, henceforth). The DCP
of a clause with an EP is an F-structure representation of LFG which is
derived from the F-structure of the clause by replacing the PRED value of
the focused grammatical function (i.e., the scope of the EP) by a variable
X constrained to be the comparable elements of the PRED value as shown
in (1c), where the set of three feature-value pairs stands for the semantic
contribution of the EP which induced the DCP.
Akira Ishikawa
46
(5) a. Taro-dake-ga kitaTaro-ONLY-NOM come-PAST
`Only Taro came.'
b.
c.
Although DCPs can avoid Schwarzschild's problem associated with the
alternative semantics theories, the representation suffers from two major
drawbacks which are related with each other. Since F-structure is
essentially a syntactic representation, the derived F-structures used for the
representation of DCPs are mainly concerned with explaining the
cooccurrence restrictions of EPs within a clause. As a result, the semantics
of EPs is indicated only in terms of the combination of the values for four
features: type, self, edge and polarity. This lack of modeltheoretic
semantics for DCPs makes it difficult for this theory to do justice to the
role played by focus in the workings of EPs, which can be seen in the fact
that domain selection has no place in the theory because it is fixed.
Although EPs are directly attached to their scope, so to speak, the actual
focus can be wider than the immediate constituent they form a part of.
Consider (6) in a context where Taro was supposed to finish several chores
besides writing a letter in the morning.
(6) Taro-wa asa tegami-dake kai-ta
Taro-TOP morning letter-ONLY write-PAST
`Taro only wrote the letter in the morning. '
An Event-Based Semantics for Japanese Emphatic Particles
47
(6) can have both VP focus (the intended interpretation here) and N
focus (comparing the letter with the other things to be written). This kind
of ambiguity in focus scope clearly points to a more flexible treatment of
the scope of EPs. We will address this problem by providing a
Davidsonian semantics for Japanese EPs.
3. Structured Davidsonian decomposition
Herburger (2000) treats the problem of focus in connection with its
interaction with negation, adverbial quantifiers, determiners and only and
even. For the semantic representation, she proposes a many-sorted
predicate logic notation, which she call structured Davidsonian
decomposition (SDD, henceforth). The SDD of a clause with a focus is a
logical form containing both event and individual variables (thus,
Davidsonian) in which the predicate of the clause is decomposed into
logical predicates corresponding to the original predicate and its
associated thematic roles (cf. Parsons (1990)). But it is different from a
simple Davidsonian decomposition in that it is structured into the
backgrounded information and foregrounded information of the clause.
(7a & b) have (8 a & b) as their SDDs (Herburger's (1) and (12)).
(7) a. ROSALIA wrote a poem.
b. Rosalía wrote A POEM.
(8) a. [∃e : C(e) & Write(e) & Past(e) & [a x : Poem(x)] Theme(e,x)]
Agent(e,rosalia) & Write(e) & Past(e) & [a x : Poem(x)] Theme(e,x)
b. [∃e : C(e) & Agent(e,rosalia) & Past(e)]
[a x : Poem(x)] Theme(e,x) & Agent(e,rosalia) & Write(e) & Past(e)
In both (8 a & b), the first line represents the backgrounded information
and the second line the foregrounded information of the clause. C in C(e)
stands for the context of utterance. Besides certain possible considerations
of quantification discussed in Schein (1993), the overlap between the
background part and the foreground part is largely due to the procedure of
deriving SDDs from LFs. As far as the semantics of focus is concerned,
48
Akira Ishikawa
there is no strong reason why we should retain such overlapping
information in the semantic representation of focused clauses.
Another problem with Herburger's SDD notation is that the SDD
derivation procedure needs to be augmented with translation rules, which
are called equivalences, such as in (9) in the case of only and even. This
suggests that SDDs might not be amenable to direct modeltheoretic
interpretation as hinted by Herburger (2000, p.19).
(9) [only e : F(e)] G(e) iff F ≠ {}
& ∀f((f ∈ F) → ∃ e(Part(f, of e) & e ∈ G))
For these reasons, I will adapt the three component ideas involved in
her structured Davidsonian decomposition to develop a focus semantics of
Japanese EPs which can overcome the drawbacks of Ishikawa (2001):
events as basic entities, decomposition into subatomic formulas, and
separation between background and foreground information.
4. Emphatic particles An event-based semantics for Japanese EPs:
ESJEP
I will use DRSs which are extended in terms of legitimate basic entities
somewhat along the lines suggested by Asher (1993). Let us first consider
how (6) (repeated here as (10)) should be represented in our ESJEP
notation. (11) and (12) correspond to the N focus reading and the VP focus
reading, respectively. (11) means that Taro wrote the letter in the morning
and nothing else, whereas (12) means that Taro wrote the letter in the
morning and did nothing else.
(10) Taro-wa asa tegami-dake kai-ta
Taro-TOP morning letter-ONLY write-PAST
`Taro only wrote the letter in the morning.'
(11) [e, x, yp1: past(e), write(e), in(e,the-morning), agt(e,x), x = taro,
theme(e,ypl), yp1 = the-letter,
[e': e'⊏ e, past(e'), write(e'), in(e',the-morning), agt(e',x)]⇒[theme(e',y
p1) ]]
An Event-Based Semantics for Japanese Emphatic Particles
49
(12) [ep1, x, y p1: past(e p1), write(e p1), in(e p1,the-morning), agt(e p1,x), x =
taro, theme(e p1,y p1), y p1=the-letter,
[e': e'⊏ e p1, past(e'),in(e',the-morning), agt(e',x)]⇒[write(e'), theme(e',y
p1) ]]
Association of dake `only' with its focus is represented by the
subscripted index p1 and the `semantic' effect of dake by the conditional
statement in each DRS, where e'⊏ e signifies that e' is comparable to e.
The index p1 is intended to indicate the prominence accorded to the
relevant discourse referent by the focus-inducing EP dake. When a
sentence contains more than one focus-inducing construction, there will be
more than one prominence-marked discourse referent in the corresponding
DRS. Such discourse referents might be associated with different degrees
of prominence from each other. In (12), both e and y are marked by the
same degree of prominence because the prominence derives from the same
source, i.e., the EP dake. Let us call these indeces prominence indices.
Like Rooth's focus semantic value, prominence indeces are meant to
capture the focus effects in combination with other DRS conditions. In the
case of dake, its semantic force is directly specified by the implicational
condition involving the subcondition e' ⊏ e. In the case of (11), the
righthand DRS in the conditional statment specifies the uniquness of the
theme argument, whereas, in (12), it is the uniqueness of the writing event
itself that is specified there. The distinction between foregrounded and
backgrounded information, which is the principal mechanism for
calculating a focus in Herburger (2000), is changed into that between the
implicational condition refelecting the semantic force of an EP and the rest
of the conditions of a DRS.
The derivation of ESJEP DRSs starts with the F-structure of a clause
with an EP or EPs. The decompostion of the predicate into subatomic
predicates is mediated by the two levels of lexical information: `semantic'
structure and `argument' structure as proposed by Mohanan (1997).
The following shows a schematic correspondence relationship between
an F-str with the EP dake and its N- and VP-focus ESJEP DRSs. The
implicational conditions specify the universal force of dake in both DRS.
They capture the fact that the focus of dake represents a necessary
50
Akira Ishikawa
condition for any e' to take place.3 We assume that the events form a
lattice whose elements are ordered in terms of an inclusiveness relation
between types. The type of en event is defined in terms of the conjunction
of the conditions describing the event. The conjunction of the set of
conditions in a DRS minus the implicational condition corresponds to the
top element of the lattice. For example, in (11), the top element, or the
most inclusive one, is past(e) & write(e) & in(e,the-morning) & agt(e,x) &
x=taro & theme(e,y)& y=the-letter
(13)
p
(14) [e, x : verb(e), θ 1 ,...,θ i −1 ,θ i ,θ i +1 ,...,θ n ,...
[e': e' ⊂ e, verb(e' ), θ 1 ,...,θ i −1 ,θ i +1 ,...,θ n ] ⇒ [θ i (e' , x p ) ]]
p
p
p
(15) [e , x : verb(e ), θ 1 ,...,θ i −1 ,θ i ,θ i +1 ,...,θ n ,...
[e': e' ⊂ e p , θ 1 ,...,θ i −1 ,θ i +1 ,...,θ n ] ⇒ [verb(e' ), θ i (e' , x p ) ]]
5. The semantics of mo and sae
The semantics of the EP mo `also' requires the existence of a
comparable event in which the thematic role in the scope of mo is replaced
by a different discourse referent as shown in (16a) and (16b).
(16) a. Taro-mo ki-ta
Taro-MO come-PAST
`Taro also came.'
b. Taro-wa kono-tegami-mo kai-ta
Taro-TOPIC this-letter-MO write-PAST
`Taro also wrote this letter.'
An Event-Based Semantics for Japanese Emphatic Particles
51
But, in this case, the nature of DCPs induced by mo is more
presuppositional than potential. Both (16a) and (16b) should be judged
semantically abnormal if no comparable situations are at hand where
different individuals are involved in the same types of events. Compare
this situation with that of (7), where both backgrounded information and
asserted information correspond to parts of the same sentence. By contrast,
in (16a), for example, the semantic force of mo is such that it introduces
the discourse referent for Taro against a backdrop of a similar event which
is not explicitly stated in the sentence but only implicated by virtue of the
use of mo, i.e., ∃x(x≠taro & came(x)) or its existential instantiation using
a contextually adequate individual name.. This implicit information or
presupposition must be anaphorically satisfied in the sense of Geurts
(1999). Moreover, the problem of deciding exactly what proposition we
should take as the presupposition of such a mo-marked sentence in a
particular context has to do with that of domain selection as discussed by
Rooth and others. In (16a), the presuppostion could as well be the
existence of another happy (or fearful) incident at the party, such as an
early departure of Bill, or more generally, another happy (fearful) situation,
such as the venue which everybody liked (dreaded) a lot. All in all, the
schematic correspondence rule refers to the presupposition by means of
the implicational condition.
(17)
p1
p1
(18) [e, x : verb(e),θ1 ,...,θ i −1 ,θ i (e, x ),θ i +1 ,...,θ n ,...
[e', y : e' ⊂ e, verb(e'), θ1 ,...,θ i −1 ,θ i (e, y),θ i +1 ,...,θ n ] ⇒ [ y ≠ x p1 ]]
p1
p1
p1
p1
(19) [e , x : verb(e ), θ 1 ,...,θ i −1 ,θ i (e, x ),θ i +1 ,...,θ n ,...
[e', y : e' ⊂ e, θ 1 ,...,θ i −1 ,θ i (e' , y),θ 'i +1 ,...,θ ' m ] ⇒ [¬ verb(e' ), y ≠ x p1 ]]
(20) and (21) correspond to (16b).
52
Akira Ishikawa
(20) [e, x, yp1: past(e), write(e), agt(e,x), x=taro, theme(e,yp1), yp1=this-letter,
[e', z: e' ⊏ e, past(e'),write(e'), theme(e',z), agt(e',x)] ⇒ [yp1≠ z ]]
(21) [ep1, x, yp1: past(ep1), write(ep1), agt(ep1,x), x=taro, theme(ep1,yp1),
yp1=this-letter,
[e', z: e' ⊏ ep1, past(e'), theme(e',z), agt(e',x)] ⇒[¬write(e'), yp1≠ z ]]
As with dake, the implicational condition is intended to capture the
semantic force of the EP mo. The presuppositional nature of the
comparable situations is captured by the use of underlined discourse
referents e' and z, which correspond to pronominal expressions requiring
antecedents. Such antecedents must be either available in the accessible
domain or accommodated. This means that the conditions containing
underlined discourse referents are usually to be interpreted in the top DRS,
where they are globally accommodated. In other words, the discourse
referents and conditions in the implicational condition could just as well
be registered directly in the top DRS in our correspondence rule. We do
not take this option because we want to have a separate condition
reflecting the semantic force of mo. Since we use the same format to
represent the semantic contribution of any EP, the generality will make for
the ease of comparison in later discussions.
(18) deals with the simple case where the comparable situation is of the
same event-type as the described one except that only the individuals
fulfilling the focused thematic role in the two situations are taken to differ
from each other. (19), on the other hand, is intended to take care of the
difficult case where the comparable situation is not of the same event-type
but only similar to the event-type of the described situation. The θs in the
implicational condition are supposed to reflect those theta roles which are
not found in the described situation. So, in order to do justice to the
intricate nature of this case, we need a theory of resemblance between
different types of events (cf. Kehler (2002)), which is clearly beyond the
scope of this paper.
A piece of evidence which suggests that (19) is on the right track is the
use of mo in which no comparable event is readily available, whether
An Event-Based Semantics for Japanese Emphatic Particles
53
anaphorically or otherwise. The use of mo in such a context only hints that
the described event can be placed within a certain implicitly aknowledged
range of events, thus implying that it is not the only such event in the
context. (22a) is a phrase often used to start a ceremonial speech, and
(22b) to begin a letter. If mo is replace by the subject marker ga, the kind
of day and the cold will be given more prominence than is intended by the
speaker, because they could be the sole such individuals in the context.
Thus, our treatment of the difficult case, i.e., (19), can naturally be applied
to this vague use of mo.
(22) a. honzitu-wa o-higara-mo yoku
today-TOP HON-kind=of=day-MO good
`As today is also a lucky day, ... '
b. hi=itiniti-to samusa-mo yurumi
day=by=day-TO cold-MO abate
`As the cold too has been abating day by day,... '
The EP sae `even' is similar to mo in that it also indicates the presence
of another discourse referent potentially fulfilling the thematic role in its
scope, but, unlike mo, the comparable event e' is not presuppositional and
further restricted by a certain relation of unexpectedness.
(23) a. kaze-sae huki-hazime-ta
wind-SAE blow-begin-PAST
`Even the wind began to blow. '
b. taro-wa tasizan-sae deki-nai
taro-TOP addition-SAE can=do-not=PRES
`Taro cannot do even addition. '
This semantic force of sae is best appreciated if we compare it with that
of made, which is another EP with a sense corresponding to even. Unlike
sae, made also means `as far as' in a locative sense. So its sense of even
can be considered as an extension of the locative one. This may account
for the fact that its semantic force is not so much to convey the
54
Akira Ishikawa
unexpectedness of the instance in terms of a scale as to emphasize its
extremity on the scale. Thus, in (23a) but not in (23b), sae can naturally be
replaced by made.
This situation is represented by using ≺ on top of ⊏ in (25) and (26).
Unlike ⊏ , ≺ presupposes a scale on which different events are located
based on their unexpectedness in the context. The described event e is
considered to be at the highest end in terms of unexpectedness compared
with the comparable events. It should be noted that the relevant scale is
defined in terms of the events rather than the individuals fulfilling the
thematic role involved. This is because such individuals can assume
different degrees of unexpectedness depending on the type of event at
issue: for example, children are more surprising if the event type is that of
making an ass of some adult, but adults are more unexpected if the event
type is that of drinking a large amount of alcohol.
(24)
p1
p1
(25) [e, x : verb(e),θ 1 ,...,θ i −1 ,θ i (e, x ),θ i +1 ,...,θ n ,...
[e' , y : e' ⊂ e, e π e' verb(e' ), θ 1 ,...,θ i −1 ,θ i (e, y),θ i +1 ,...,θ n ] ⇒ [ y ≠ x p1 ]]
p1
p1
p1
p1
(26) [e , x : verb(e ), θ 1 ,..., θ i −1 , θ i (e, x ), θ i +1 ,...,θ n ,...
[e' , y : e' ⊂ e p1 , e' π e p1 θ 1 ,..., θ i −1 , θ i (e' , y), θ ' i +1 ,..., θ ' m ] ⇒ [¬ verb(e' ), y ≠ x p1 ]]
Another crucial consideration is this: since ≺ is a negatively defined
order based on the notion of unexpectedness, y in the implicational
condition of (25) and (26) should represent those individuals which
constitute more usual events. Thus, we can look at the implicational
condition in two opposite angles. On the one hand, it is viewed as
indicating the locus of unexpectedness, which is rather trivial in the case
of (25). On the other hand, it can also be interpreted as describing the
An Event-Based Semantics for Japanese Emphatic Particles
55
more usual cases.
This line of reasoning and our analysis are supported by the behaviour
of sae, when it occurs in the antecedent of a conditional statement like
(27), whose DRS is (28). (28) says that if you are near me and the other
more usual events are such that they do not involve your being near me, I
will be happy. Without going into detail, I will assume that, in interpreting
sae in the antecedent of a conditional, the viewpoint of the more usual
cases is adopted because the antecedent corresponds to a sufficient
condition, and (26) rather than (25) is used because it is more informative
as to the more usual cases.
(27) anata-sae ir-ebae watasi-wa siawase-da
you-SAE xist-COND I-TOP happy-PRES
`If only you are with me, I am happy. '
(28) [[ep1,xp1: be-near(ep1),theme(ep1,xp1), xp1=you
[e', y: e' ⊏ ep1, e' ≺ ep1, theme(e',y)] ⇒ [¬ be-near(e'), y≠ xp1 ]]
⇒[e",z: happy(e"), theme(e",z),z=I]
6. Contrastive wa
Following Ishikawa (2001), topical wa is considered as link in the
sense of Vallduví(1992), marking the presentation of the clause topic. By
contrast, contrastive wa is an EP which interacts with focus and negation.
The difference between contrastive wa and dake can be shown by
introducing a modal operator indicating a lack of information. In (30), the
implicational condition says that it is a distinct possibility to the speaker
that the comparable event involving a different individual fulfilling the
focused thematic role might not have taken place: Taro wrote the letter in
the morning but might not have written anything else.
(29) Taro-wa asa tegami-wa kai-ta
Taro-TOP morning letter-CONTRAST write-PAST
`Taro wrote [the letter]F in the morning. '
Akira Ishikawa
56
(30) [e, x, yp1: past(e), write(e), in(e,the-morning), agt(e,x), x=taro,
theme(e,yp1), yp1=the-letter
[e', z: e' ⊏ e, in(e',the-morning), agt(e',x), theme(e',z), z≠yp1] ⇒[⋄¬
write(e')]]
(31) and (32) show how our DRS captures the interaction of contrastive
wa and negation, where the focused argument and the main predicate
equally attract the negation of the clause. (2) says that Taro did not write
the letter in the morning, but might have written something else.
(31) Taro-wa asa tegami-wa kak-anakat-ta
Taro-TOP morning letter-CONTRAST write-NEG-PAST
`Taro did not write [the letter]F in the morning. '
(32) [e, x, yp1: past(e), ¬write(e), in(e,the-morning), agt(e,x), x = taro,
theme(e,yp1), yp1 = the-letter
[e', z: e' ⊏ e, in(e',the-morning), agt(e',x), theme(e',z), z≠ yp1, ] ⇒
[⋄¬¬write(e')]]
Contrast (31) and (32) with (33) and (34), which do not involve an EP.
(33) Taro-wa asa tegami-o kak-anakat-ta
Taro-TOP morning letter-ACC write-NEG-PAST
`Taro did not write the letter in the morning. '
(34) [e, x, y: past(e), ¬write(e), in(e,the-morning), agt(e,x), x = taro,
theme(e,y), y = the-letter]
Thus, I propose the following correspondence schema for contrastive wa.
(35)
An Event-Based Semantics for Japanese Emphatic Particles
57
7. EPs' interaction with semantic case particles
In ESJEP DRSs, we can capture the difference in meaning between (37
a & b), which has been noted for some time but defied its formal
characterization in semantic terms (cf. Ishikawa (1985)).
(37) a. sono kinko-wa kono kagi-dake-de ak-u
that safe-TOP this key-ONLY-INST open-PRES
`That safe can be opened with this key alone (without using any other)'
b. sono kinko-wa kono kagi-de-dake ak-u
that safe-TOP this key-INST-ONLY open-PRES
`That safe can be opened with nothing other than this key (no other key
works)'
Unlike grammatical case particles (ga, o, and ni), semantic case
particles such as de 'with', kara 'from', etc. show scopal interaction with
EPs.4 In ESJEP DRSs, the two readings can be clearly distinguished as
in (38) and (39). We notice that (39), which corresponds to the case in
which the EP has a case-marked N in its scope, has the structure along the
lines of (14). By contrast, (38), where the EP has a bare N in its scope and
is itself in the scope of the semantic case particle de, has the focus of dake
as part of backgrounded information rather than that of foregrounded.
(38) [e, x, yp1: potential(e), open(e), theme(e,x), x=that-safe, instr(e,yp1),y
p1=this-key
[e': e' ⊏ e, potential(e'), theme(e',x), [z: instr(e',z)] ⇒ [z = yp1] ] ⇒
[open(e')]]
(39) [e, x, yp1: potential(e), open(e), theme(e,x), x=that-safe, instr(e,yp1),
yp1=this-key
[e': e' ⊏ e, potential(e'),open(e'), theme(e',x)] ⇒ [instr(e',yp1)] ]
Since the uniqueness of the individual fulfilling the instrumental role is a
sufficient condition for a comparable event to take place and a necessary
condition in (38) and (39), respectively, let us call the former the
58
Akira Ishikawa
sufficient pattern and the latter the necessary pattern. The relevant
correspondence schemata are given in (40) and (41) for the sufficient
pattern and in (42) and (43) for the necessary pattern.
(40)
p
(41) [e, x : verb(e), θ 1 ,...,θ i 1 ,θ i ,θ i +1 ,...,θ n ,...
[e': e' ⊂ e, verb(e' ), θ 1 ,...,θ i −1 ,θ i +1 ,...,θ n , [ y : θ i (e, y )] ⇒ [ y = x p ]] ⇒ [verb(e' )]
(42)
p
(43) [e, x : verb(e),θ1 ,...,θ i 1 , θ i , θ i +1 ,...,θ n ,...
[e' : e' ⊂ e, verb(e' ), θ1 ,..., θ i −1 , θ i +1 ,..., θ n ] ⇒ [θ i (e, x p )]]
As to the surface case sequence patterns, we notice that N-EP-SC gives
rise to the sufficient pattern, and N-SC-EP the necessary pattern, where SC
stands for a semantic case. To the best of my knowledge, the other
semantic case particles also show the same behaviour with regard to dake
with the exception of ni, which is part semantic and part grammatical.
The sufficient pattern is suggestive of how the focus associated with the
EP dake has its prominence detracted as it is removed from the main
predicate which determines its thematic role in the F-str. We can
conjecture that morphologically embedded EPs cannot make the item in its
scope the primary focus of a clause, which must be sought elsewhere in it.
An Event-Based Semantics for Japanese Emphatic Particles
59
In (44), the ga-marked NP rather than the dake-marked NP gets primary
focus because the former is at the same level with the main predicate. 5
Our DRS notation using explicit indices for prominence degree can
capture this situation.
(44) sono kinko-ga kono kagi-dake-de ak-u
that safe-GA this key-ONLY-INST open-PRES
`It is that safe which can be opened with this key alone (without using
any other)'
(45) [e, xp1, yp2: potential(e), open(e), theme(e,xp1), xp1=that-safe,
instr(e,yp2), yp2 =this-key
[e': e' ⊏ e, potential(e'), open(e') [z: instr(e',z)] ⇒ [z = y p2] ] ⇒
[ open(e') ]]
Stacking two EPs dake and wa within a single phrase and following our
correspondence schemas for them, we predict the following DRS for (1),
which matches our intuitions.
(46) sono kinko-ga kono kagi-dake-de-wa aka-na-i
that safe-GA this key-ONLY-INST-CONTRAST open-NEG-PRES
`It is that safe which cannot be opened with this key alone (without using
any other)'
(47) [e, xp1, yp2: potential(e), ¬ open(e), theme(e,x p1), x p1=that-safe,
instr(e,yp2), yp2=this-key
[e': e' ⊏ e, potential(e'), theme(e',xp1), ¬ [z: instr(e',z)] ⇒ [z = yp2] ]
⇒ [◇¬¬open(e')]]
8. Conclusion
I have shown how our ESJEP DRS notation, which is inspired by
structured Davidsonian decomposition, can make it possible to account for
the basic semantic effects of four EPs, i.e., dake, mo, sae, and wa in such a
way that will lead to their fully compositional treatment. By using DRSs
whose conditions contain subatomic predicates standing for various
60
Akira Ishikawa
thematic predicates, we can represent the intricate semantic contribution of
the EPs. Besides being more explicit as to the role of discourse referents in
the overall meaning of a particular EP, the DRS notation is shown to be
empirically `correct' in the sense that their definition of the basic cases can
predict the semantic effect in the complex cases, such as interaction with
other kinds of particles.
In this paper, we have not addressed the issue of compositionality in the
process of deriving the DRSs from LFG F-structures. We have only
indicated the generality of such derivation by using a common format to
describe the correspondence between the two. It should be the topic of our
future study to identify the components involved in the implicational
conditions of the EPs and to devise a more flexible formulation of the
correspondences.
References
Asher, Nicholas, 1993. Reference to abstract objects in discourse. Kluwer.
Atlas, Jay David, 1991. Topic/Comment, Presupposition, Logical Form and
Focus Stress Implicatures: The Case of Focal Particles. Journal of
Semantics 8: 127-147.
Bresnan, Joan, 2001. Lexical-functional syntax. Blackwell.
Choi, Hye Won, 1996. Optimizing Structure in Context: Scrambling and
Information Structure. Ph.D. thesis, Stanford University.
Foley, William A. and Robert D. van Valin, 1984. Functional syntax and
universal grammar. Cambridge University Press.
Geurts, Bart, 1999. Presuppositions and Pronouns. Elsevier.
Hasimoto, Shinkichi, 1948. Kokugo-hoo kenkyuu. Iwanami Syoten.
Herburger, Elena, 2000. What Counts-Focus and Quantification. MIT Press.
Horiguti, Kazuyoshi, 1995. Wa no hanasi. Hituzi Shobo.
Huebler, Axel, 1983. Understatements and hedges in English. John Benjamins.
Ishikawa, Akira, 1985. Complex predicates and lexical operations in Japanese.
Doctoral dissertation, Stanford University.
Ishikawa, Akira, 2001. Emphatic particles and their scopal interactions in
Japanese. Proceedings of the 15th Pacific Asian Conference: 73-84. City
University of Hong Kong.
Iida, Takashi, 2001. Nihongo keishiki imiron no kokoromi (3) - toritate to hitei.
ms. Keio University.
Kadmon, Nirit, 2000. Formal Pragmatics. Blackwell.
An Event-Based Semantics for Japanese Emphatic Particles
61
Kehler, Andrew, 2002. Coherence, Reference, and the Theory of Grammar.
Stanford, CSLI.
Kim, Mi Kyung, 2000. Dynamics of information packaging in Korean. In
Akira Ikeya and Masahito Kawamori, editors, PACLIC 14: 14th Pacific
Asia Conference on Language, Information and Computation: 177-188.
Logico-Linguistic Society of Japan, February.
Lambrecht, Knud, 1994. Information structure and sentence form. Cambridge
University Press.
Lambrecht, Knud and Laura A. Michaelis., 1998. Sentence accent in
information questions: Default and projection. Linguistics and
Philosophiy: 477-544.
Lee, Chungmin, 2001. Contrastive Topic and Proposition Structure. Presented
at the Asymmetry Conference, UQAM, in May 2001. Published in Anna
Maria di Sciullo (2002) (ed.) Asymmetry in Grammar Vol 1: Syntax and
Semantics. Amsterdam: John Benjamins.
Martin, Samuel E., 1975. A Reference Grammar of Japanese. Yale University
Press, London.
Miyazima, Tatsuo and Yoshio Nitta, editors. 1995. Nihongo ruigihyougen no
bunpoo (zyoo). Kurosio Syuppan.
Mohanan, Tara, 1997. Multidmensionality of Representation: NV Complex
Predicates in Hindi. In Complex Predicates, edited by Alex Alsina, Joan
Bresnan, and Peter Sells, 431-71. Stanford, CSLI.
Numata, Yoshiko, 1992. Mo, dake, sae nado - toritate. Nihongo Bunpo Self
Master Series 5. Taishukan, Tokyo.
Parsons, Terence, 1990. Events in the semantics of English: A study of
subatomic semantics. Cambridge, Mass.: MIT Press.
Rooth, Mats, 1985. Association with focus. Doctoral diss. University of
Massachusetts, Amherst.
Rooth, Mats, 1992. A theory of focus interpretation. Natural Language
Semantics 1, 75-116.
Rooth, Mats, 1995. Indefinites, adverbs of quantification, and focus semantics.
In The generic book, edited by Gregory Carlson and Francis Jeffry
Peletier. Chicago: Univeristy of Chicago Press.
Schein, Barry, 1993. Plurals and events. Cambridge, Mass.: MIT Press.
Teramura, Hideo, 1991. Nihongo no syntax to imi. Kurosio Syuppan, Tokyo.
Vallduví, Enric, 1992. The Information Component. Garland Publishing.
Wlodarczyk, Andre, 1994. Predicative and set-theoretic relations in standard
Japanese - WA and GA. Proceedings of Warsaw Symposium on Japanese
Studies 24-26: 395-416.
Akira Ishikawa
62
Foot notes
1
Such focus effects can be cancelled by suitable contexts as shown in the
next sentence.
(i) People who [grow]F rice generally only [eat]F rice. (Rooth 1992, p. 109)
(ii) They don't eat [meat]F, or [bread]F
In (i), association of only with the focused verb in its scope does not take
place because of the preceding context in the sentence. A natural continuation
is as in (ii).
2
The problem is that the definition of focus semantic value allows unwanted
propositons to be in the set of alternatives.
(i) Sue is [bold]F
(ii) (i) f = { X(s)
g
: g assignment function }
Since λx[beautiful (m)] is also a member of D<e,t> , the set of alternatives will
include such propositon as 'Mary is beautiful' as well as 'Sue is beautiful', etc.
(Kadmon, p. 307)
3
Our formulation of schematic correspondences between F-structure
configuration and DRSs might strike one as stipulative. In fact, the tacit
claim is that an EP can have either its SCOPE or the PRED which
immediately contains it as its target of focusing. Consequently, the sister
grammatical categories governed by the same PRED cannot be such a target.
In section 7, we will see that the focus of dake represents either a necessary or
a sufficient condition for e' to take place depending on the semantic nature of
its focus, which might or might not contain a semantic case marker.
4
For some speakers, (37a) can admit both interpretations. But (37b) is not
ambiguous.
5
Our reviewer wondered whether, in (37a), which has a wa-marked subject
in place of the ga-marked one in (44), the absense of an item bearing primary
focus due to the embedded position of dake might not make the wa-marked
subject carry primary focus, thus forcing a contrast interpretation of wa.
An Event-Based Semantics for Japanese Emphatic Particles
63
Indeed, we can interpret the surface string in (44) as involving contrastive wa.
But, in that case, the primary focus is induced by the presence of the
contrastive wa, as opposed to the topic wa in the original interpretation. Our
claim as to the embedded EP is that the item in its scope cannot carry primary
focus because it becomes part of the backgrounded information of the entire
sentence. This paper does not deal with the mechanism of how exactly the
degree of prominence among multiple focusable items associated with
multiple EPs in a sentence can be determined.