Negative Imperatives in Korean Peter Sells Stanford University
1. Introduction Like many languages, Korean has a special form of negation that is used in imperative clauses (see (1)c), to the exclusion of the usual clausal negation in (1)b: (1) a.
ka-la go-Imp ‘Go!’
b.
*ka-ci anh-ala go-Comp Neg-Imp ‘Don’t go!’
c.
ka-ci mal-ala go-Comp Neg-Imp ‘Don’t go!’
Sadock and Zwicky (1985) noted that negation in imperative(-like) clauses shows special morpho-syntax in many languages, a fact documented in more detail by Zanuttini (1997) or Han (2000). In this paper I will consider the semantic properties of Korean clauses that use the negative form mal-, and suggest a more indirect relationship to the morpho-syntax than has been assumed in previous work.∗ In section 2 I present the basic account of clausal semantics in the HPSG framework of Ginzburg and Sag (2000), and then in section 3 I return to a fuller consideration of data like that in (1).
2. Semantics of Clause Types Ginzburg and Sag (2000) (hereafter, GS00) classify clause meanings into the four types at the bottom of the hierarchy in (2), which shows sub-types of message: (2)
message austinian proposition
prop-constr outcome fact
question
Of these, only proposition (the denotation type of declaratives) and outcome (the denotation type of imperatives) will be pertinent to the points at issue. GS00 also classify certain types of illocutionary relation, which will be the meanings associated with the uses of declaratives or imperatives, among other clause types: ∗I
am grateful to Jonathan Ginzburg, Chung-Min Lee, Jong-Bok Kim, Shin-Sook Kim, Ivan Sag and John Whitman for discussion of the examples and ideas here.
(3)
illoc-rel assert-rel ask-rel order-rel exclaim-rel
Following a general strategy laid out by Portner (1997), Ginzburg and Sag (2000) classify the contents of different clause types, and then specify which illocutionary forces can embed them at the root level. Portner draws analogies between verbs taking different kinds of embedded clause complement (e.g., think takes a proposition and wonder takes a question) and what happens at the root level (one asserts a proposition and asks a question, and not vice versa). (4) a.
assert-rel ⇒ [MSG-ARG proposition]
b.
ask-rel ⇒ [MSG-ARG question]
c.
order-rel ⇒ [MSG-ARG outcome]
d.
exclaim-rel ⇒ [MSG-ARG fact]
A root clause is characterized by the rule in (5), in which a speaker presents a message to an addressee. GS00 argue that (in English) all root clauses are headed by a finite form of a verb, and so this is specified in the rule. The content 2 is the (semantic) content of the clause itself. (5)
root-cl: ROOT
HEAD CONT
+
"
#
v VFORM fin
proposition
illoc-rel UTTERER SOA|NUCL ADDRESSEE
CTXT|C-INDICES
"
MSG-ARG
#
C-SPKR
1
C-ADDR
3
ROOT − + → HIC 1 2 CONT 3 2
For example, the declarative clause Kim sees Sandy has a denotation of proposition, and hence is the kind of object that can be the argument of assert-rel (see (4)a). Unifying various constraints together gives (6) as the analysis of the semantics. The innermost part of (6) is a state-of-affairs soa which involves the see relation holding between two individuals (and no quantifiers are present). The meaning of a declarative clause is a proposition, which consists of a situation s and some soa which the situation supports. 2
(6)
proposition SIT s0
SOA
soa
QUANTS NUCL
h i
assert-rel
UTTERER 4 ADDRESSEE 6 proposition SIT s soa QUANTS h i MSG-ARG SOA see-rel SEER NUCL SEEN
1 2
where 1 = ‘Kim’ and 2 = ‘Sandy’. An imperative like Close the door! receives the analysis in (7): (7)
proposition SIT s0
SOA
QUANTS h i order-rel 4 UTTERER ADDRESSEE 6 outcome SIT s i-soa QUANTS h i NUCL param MSG-ARG INDEX t T-PARAM n o SOA RESTR time(t) close-rel CLOSER 6 NUCL
CLOSED
where 2 = ‘the door’ 3
2
Imperatives involve the type of outcome, which is built up from an soa, like a proposition. GS00 classify soas into two types: realis and irrealis (r-soa and isoa), for the proposition and outcome types respectively. The difference between the two lies in the fact that the latter has its temporal parameter abstracted out. And by (4)c, an outcome is appropriate to be the argument of an order-rel. An outcome has the irrealis quality due to the fact that its temporal parameter is abstracted out, and informally we can characterize the meaning of an outcome like this: (8) a.
The outcome is interpreted relative to some world(s) which are not necessarily extensions of the current world. In other words, the outcome is interpreted as if it were under the scope of a deontic (modal) operator.
b.
The outcome is ‘futurate’ in the sense that it has no backwards history in time (unlike propositions in ‘normal worlds’; see Portner (1997)). The outcome is only forward-looking.
Now in the case of a true imperative, the outcome is supposed to be true of some later time: Close the door! can be analyzed as ‘I want you to make it true at some point in the future that [x close the door] is true’. GS00 (p.98) define a property Fulfilled which can be true of outcomes which have this ‘futurate’ notion of truth associated with their interpretation. While the canonical use of an Imperative form is to issue a directive, the form can have a variety of uses which are quite far from pure commands or orders (see e.g., Schmerling (1982)): (9) a.
Don’t trip on that cord.
b.
Slower traffic keep right.
c.
Get well soon!
d.
Stay (, dammit)! (said to a poster hung on the wall)
Moreover, the way that imperative forms get their illocutionary force involves calculations in discourse. Lascarides and Asher (2002) argue that whether an imperative form expresses an order or not depends on the rhetorical relations between the parts of the conversation. (10) is not a command to eat, but rather a suggestion to avoid doing that, and go left in (11) is not accepted as a directive, because it cannot be complied with, though go right is accepted (as it can be complied with): (10)
Eat that and you’ll die.
(11)
A: What should I do next?
B: Go left.
A: That’s a dead-end.
B: OK, go right then.
4
Using the terminology of GS00, Lascarides and Asher (2002) argue that an imperative expresses an order just in case the futurate content of the outcome is supposed to be Fulfilled. Davies (1986) argues that imperatives are not restricted even to a class that we might call directives, but rather fall in wider class, all of whose members have in common the presentation of some proposition as a future possibility (rather than the assertion of some proposition), as suggested by (12): (12)
present-rel propose-rel request-rel order-rel wish-rel
The ‘proposition’ – an outcome, in the terms used here – is presented as some future possibility (which the speaker at least acknowledges as a possibility). In summary, the semantics of a clause involves assigning a particular denotationtype to the clause itself which is then the argument of some illocutionary relation, which falls into the presentational class in (12) in the case of an outcome. One important part of the analyses either in GS00, or in the approach to sentential force developed in Zanuttini and Portner (2003), is that there is no marker of the illocutionary force in the syntax, in the general case.
3. Negative Imperatives While canonical imperatives in Korean are formed with -e/ala, if host is a stative verb, an exclamative interpretation results (see Sohn (1999, 280)), though this is only true for statives which cannot be coerced into some agentive or intentional interpretation. Hence (13)a/b are imperatives while (13)c/d are exclamatives. (13) a. anc-ala! b. cengcik hay-la! sit down-Imp honest-Imp ‘Sit down!’ ‘Be honest!’ c.
chwuw-ela! d. caymi iss-ela! cold-Imp interesting-Imp ‘(Boy,) It’s cold!’ ‘What fun!’ The imperative form in Korean can be associated with the range of presentationtype relations discussed above. (14) is a warning (cf. (10)), (15)B involves a suggested answer to a question, and (16)B is a somewhat ironic use: (14)
kuke (mek-ki-man) mek-ela kulemyen ne-nun cwuk-ul.kes.i-ta that (eat-Nmlz-only) eat-Imp then you-Top die-Fut-Decl
(15) A:
kyeylan sey kay yeki iss-ta. icey mwuel hay-ya ha-ni? egg three Class. here be-Decl now what.Acc do-Comp do-Q ‘Here are three eggs. What should I do now?’
B:
kukes(-tul)-ul kulus-eyta phwul-ela That(-Plu.)-Acc bowl-in beat-Imp ‘Beat them in a bowl.’ 5
(16) A:
B:
ne nuc-ess-ta-ko pokoha-l.ke-ya you be.late-Past-Decl-Comp report-Future-Emph ‘I’m going to report that you were late.’ kwaynchanh-a. pokohay-la okay (I don’t mind) report-Imp ‘Okay, I don’t mind, report me.’
In other words, -e/ala constrains its clause denotation to be of the outcome type. Negating imperatives involves the special form mal- (see (17)b); normal propositional negation formed with an(h-) does not appear with an imperative (see (19)): (17) a.
ka-la go-Imp ‘Go!’
b.
ka-ci mal-ala go-Comp IrNeg-Imp ‘Don’t go!’
(18) a.
an ka-n-ta Neg go-Proc-Decl ‘(Someone) doesn’t go.’
b.
ka-ci anh-nun-ta go-Comp Neg-Proc-Decl ‘(Someone) doesn’t go.’
(19) a.
*an ka-la b. Neg go-Imp ‘Don’t go!’ (short-form neg.)
*ka-ci anh-ala go-Comp Neg-Imp ‘Don’t go!’ (long-form neg.)
mal- can appear in variety of directive or propositive clause types, all of which correspond to outcome meanings: (20) a.
ka-ci mal-ca go-Comp IrNeg-Propos ‘Let’s not go!’
b.
ka-ci ma-sey-yo go-Comp IrNeg-Hon-Level ‘Please don’t go!’
Han (2000, 163) proposes that the ‘logical form’ of an imperative is directive(irrealis(p)) (see also Han (2001)). The inner irrealis part corresponds to the outcome of GS00 and the outer directive part corresponds to the presentation illocutionary relation described above. For the morpho-syntax, Han and Lee (2002) argue that mal- is a lexical spell-out of a combination of Neg and deontic Mood: (21)
CP ModP NegP
C Mod
VP
Neg
... V
[Neg]
6
[Imp-Op]
[Mood] mal-
If the special [Mood] is not present, negation spells out as an(h)-. Semantically, the proposals by Han (and Lee) for the interpretation of imperatives seem quite appropriate. However, the syntactic representation in (21) is highly underdetermined by the data: there is no overt evidence for [Imp-Op] as a ‘force’ operator, and similarly [Mood] is totally abstract. As I will show below, even mal- does not appear to literally be the ‘spell-out’ of [Neg] + [Mood]. In the context of HPSG, I propose that mal- is a V(-root) which forms a complex predicate with a preceding complement V. This is not at all unusual in the context of Korean syntax; long-form negation as in (19)b has the same structure (see Kim (2000), Sells (1998), Sells (2001)). First, we define types that the different negative forms will inherit from: (22)
aux-lxm
neg-lxm anh-
ir-neg-lxm ma(l)-
Next, we define the properties of the types. An aux-lxm is a verb that directly passes up the arguments from its complement verb to itself, and by default the semantic content is unchanged (though the auxiliary verb may change the semantics – see e.g., Sells (1998)). The GS00 analysis of negation is straightforward: it puts neg-rel in the QUANTS list, as in (23)b/c: A neg-lxm constrains its clause’s content to be of type r-soa while ir-neg-lxm constrains the content to be i-soa. (23) a. aux-lxm: HEAD v v D E+ * SUBJ 1 ARG-ST ⊕ A 1 , COMPS A CONT /3 /3
CONT
" b.
neg-lxm:
" CONT
" c.
ir-neg-lxm:
CONT
##
r-soa QUANTS h neg-rel i
"
## i-soa QUANTS h neg-rel i
The polite form -sey-yo can be associated with an assertion, a question, or a request. However, negating with an, which inherits from neg-lxm, causes the request interpretation to be lost, because an constrains the content to be r-soa (in GS00, questions are built up from propositions (see (2)).
7
(24) a.
b.
ka-sey-yo go-Hon-Level ‘(He) goes.’/‘Does (he) go?’/‘Please go!’ an ka-sey-yo Neg go-Hon-Level ‘(He) does not go.’/‘Does (he) not go?’/*‘Please don’t go!’
Like the long-form negation V-root anh-, mal- takes the full range of verbal inflection. And as it forms a complex predicate with a preceding V0 , it can appear more than once in a single articulated predicate. Consider the negation of (25)a, from Martin (1992), in (25)b: (25) a.
b.
mal hay cwu-sey-yo say.Comp give-Hon-Level ‘Please give me the favor of telling me (something).’ salang ha-ci anh-nun-ta-ko malha-ci mal-a love do-Comp Neg-Proc-Decl-Comp say-Comp IrNeg-Comp cwu-ci ma-sey-yo give-Comp IrNeg-Hon-Level ‘Please do not give me the favor of not telling me that you don’t love me.’ (‘Please don’t hold back from telling me that you don’t love me.’)
The structure of the predicate part of (25)b is in (26), with mal- underlined: (26)
V V
V
V
V
V
V
mal ha-ci
mal-a
cwu-ci
ma-sey-yo
There is no reason to think that the source for this example involves two [Neg] heads matched by two [Mood] heads, as suggested by (21), such that there are two combinations of each pair, spelling out as mal-. Rather, these are just uses of a form of negation that is associated with an outcome-type meaning. The examples in (27), modified from Martin (1992, 683), make the same point, and in (28), the negative form mal- is not even in the same clause as the verb with the imperative marker -la.
8
(27) a.
b.
(28)
nol-ci mal-ko il ha-ca loaf-Comp IrNeg-Conj work do-Propositive ‘Let’s cut out the loafing and get to work!’ pakk-ey naka-ci mal-ko cip-ey iss-kela outside-Dir go out-Comp IrNeg-Conj home-Loc be-Imp ‘Don’t go out and stay home!’ icey il hay-la! nol-ci mal-ko now work-Imp loaf around-Comp IrNeg-Conj ‘Get to work now! No loafing around.’
Consider (27)b, where the imperative suffix in the final conjunct scopes over both conjuncts. In other words, conjunction must be lower in the structure than [ImpOp] in (21). If the left conjunct is a ModP, then [Neg] and [Mood] will be adjacent and after head-movement will spell-out as mal-. However, nothing prevents an analysis where the conjunction is lower than ModP: for example, the left conjunct could be simply NegP, in the scope of [Mood], but not adjacent to it. In such a structure, [Neg] would spell-out as anh-, incorrectly. These problems do not arise in an approach where mal- classifies its clause’s denotation in a certain way (to be an outcome). mal- might itself ‘be responsible’ for that classification (as in the fragment nol-ci mal-ko in (28)), or it may simply unify with information coming from elsewhere (as in (17)b).
4. Conclusion In this short paper I have argued that the special negative form mal- is associated semantically with a deontic interpretation because it classifies the meaning of the clause containing it to be of type outcome. The nature of the syntax-semantics interface in HPSG allows this classification to be formalized without the need for a [Mood] head in syntax, a head for which there is no overt evidence in Korean. Future work must determine whether this approach can extend to other cases of unusual morpho-syntax in negative imperatives, including those discussed in Zanuttini (1997) and Han (2000).
References Davies, Eirlys. 1986. The English Imperative. Dover, NH, Croom Helm. Ginzburg, Jonathan, and Ivan Sag. 2000. Interrogative Investigations: The Form, Meaning and Use of English Interrogatives. Stanford, CSLI Publications. Han, Chung-Hye. 2000. The Structure and Interpretation of Imperatives. New York, Garland Publishing, Inc.
9
Han, Chung-Hye. 2001. Force, negation and imperatives. The Linguistic Review 18, 289–325. Han, Chung-Hye, and Chung-min Lee. 2002. On negative imperatives in Korean. In Proceedings of the 16th Pacific Asia Conference on Language, Information and Computation. The Korean Society for Language and Information. Kim, Jong-Bok. 2000. The Grammar of Negation: A Lexicalist, Constraint-Based Perspective. Stanford, Dissertations in Linguistics, CSLI Publications. Lascarides, Alex, and Nicholas Asher. 2002. Imperatives in dialogue. In P. Kuehnlein, H. Rieser, and H. Zeevat (eds.), The Semantics and Pragmatics of Dialogue for the New Millenium. Amsterdam/Philadelphia, John Benjamins, to appear. Martin, Samuel. 1992. A Reference Grammar of Korean. Rutland, Vermont, Charles E. Tuttle. Portner, Paul. 1997. The semantics of mood, complementation, and conversational force. Natural Language Semantics 5, 167–212. Sadock, Jerrold, and Arnold Zwicky. 1985. Speech act distinctions in syntax. In Timothy Shopen (ed.), Language Typology and Syntactic Description: Volume I, Clause Structure. Cambridge, Cambridge University Press, 155– 196. Schmerling, Susan. 1982. How imperatives are special, and how they aren’t. In R. Schneider, K. Tuite, and R. Chametzky (eds.), Papers from the Parasession on Nondeclaratives. University of Chicago, The Chicago Linguistics Society, 202–218. Sells, Peter. 1998. Structural relationships within complex predicates. In ByungSoo Park and James Yoon (eds.), Proceedings of ICKL 11. Seoul, International Circle of Korean Linguists, 115–147. Sells, Peter. 2001. Three aspects of negation in Korean. Journal of Linguistic Studies 6, Cheju Linguistics Circle, 1–15. Sohn, Ho-Min. 1999. The Korean Language. Cambridge and New York, Cambridge University Press. Zanuttini, Raffaella. 1997. Negation and Clausal Structure: A Comparative Study of Romance Languages. New York, Oxford University Press. Zanuttini, Raffaella, and Paul Portner. 2003. Exclamative clauses: At the syntaxsemantics interface. Language 79, 39–81.
[email protected] http://www-csli.stanford.edu/˜sells/
10
