A REANALYSIS OF WH-QUESTIONS IN KOREAN
JINHEE PARK
Abstract
In this paper I will argue that Korean wh-words undergo focus movement at LF. The robust evidence for the argument is that Korean wh-words without focus are interpreted as indefinites. Although wh-words and indefinites manifest distinct interpretations and syntactic distributions it turns out that they are the same syntactic objects and that the difference between them is from a focus feature which induces focus movement of wh-words.
- Introduction
Since Huang (1982) argued that wh-words in wh-in-situ languages move covertly there have been many suggestions about wh-in-situ languages. One of them is the argument against Huang (1982), which is suggested by Aoun and Li (1993a, b) among others. According to Aoun and Li (1993a, b), wh-words which are expected to move covertly under Huang (1982) do not move at any level of grammar, rather they are bound unselectively by a null-operator which is base-generated in Spec, CP.
This paper will argue against the unselective binding thesis. Specifically I will argue that this strategy cannot be applied to Koreanwh-questions. In the course of advancing this view I will examine the uneven interpretations of wh-words, the parallel behavior of wh-words to the other focused phrases and intervention effects triggered by indefinites. The structure of this paper is as follows. In section 2 I will investigate the ambiguity of Korean wh-words. Adopting previous analyses I will propose that wh-words have a focus feature. Section 3 will argue that wh-words and indefinites are the same syntactic entity by comparing them to quantifiers. Meanwhile it is shown that indefinites do not move because of their lack of focus feature. Section 4 will argue that wh-words undergo focus movement, and will explain that the intervention effect that is induced by indefinites is a case of Principle C violation.
My argument is advanced within the framework of Minimalism. Under Minimalism we assume that there is a computational component, i.e. narrow syntax, which maps lexical array to a derivation. A derivation in turn feeds an interpretation in the sensorimotor system (historically phonetic form) and an interpretation in the conceptual-intentional system (historically logical form).
2. Two interpretations of wh-words
The Korean wh-word can have two interpretations depending on focus:awh-word reading and an indefinite reading.[1] [2]
(1) Jinsu-ga (a. nugu[3]/ b. nugu)-reul joaha-ni?
Jinsu-NOM (a. who/ b. someone)-ACC like-Q
a. ‘Who does Jinsu like?’
b. ‘Does Jinsu like someone?’
Among two possible readings, nugu in (1a) has only a wh-word reading, i.e. ‘who’, due to prosodic prominence which marks focus. On the other hand nugu in (1b) has only an indefinite reading, i.e. ‘someone’, because it does not have focus. Because Korean also has a complex form of quantifier comprising a wh-word and the question marker nga and because this complex form is interpreted only as a quantifier regardless of focus, the fact that a bare wh-word has an indefinite reading is significant when we compare Korean to other wh-in-situ languages.[4][5] In other words Korean has two forms for existential quantifiers: a bare wh-word and a wh+Q complex. For simplicity, I will refer to the wh-word interpreted as an existential quantifier as an indefinite, and the complex form as a quantifier.
To explain the ambiguity of the wh-words, Suh(1987) suggested a Focus Assign Rule, which was based on the phonological difference between wh-words andindefinites. Following Suh (1987), if we consider prosodicprominence as focus we expect one distribution of a wh-word which to be parallel to other focused phrases.
(2) a. amudo nugu-reul an-chodaeha-ass-ni?
no one someone-ACC NEG-invite-TEN-Q
‘Does no one invite someone?’
b. *amudo nugu-reul an-chodaeha-ass-ni?
no one who-ACC NEG-invite-TEN-Q
‘Who does no one invite?’
c. amudo geu saram-reul an-chodaeha-ass-ni?
no one that person -ACC NEG-invite-TEN-Q
‘Does no one invite that person?’
d. *amudo geu saram-man an-chodaeha-ass-ni?
no one that person-only NEG-invite-TEN-Q
‘Does no one invite only that person?’
From the ungrammaticality of (2b) and (2d) we see that a wh-word and a focused phrase cannot be c-commanded by a negative quantifier as Kim, S. S. (to appear) pointed out. According to Beck and Kim (1997) and Kim,S. S. (to appear) this can be considered as an island effect. That is, the node which immediately dominates a negative quantifier is a barrier which blocksthe movement of a scope bearing element. Adopting Beck and Kim (1997) and Kim,S. S. (to appear) we assume that the reason why (2b) is ungrammatical is because the wh-word nugu in (2b) moves covertly. In contrast, the reason why (2a) is a grammatical sentence is that the object nugu does not move like the noun phrase geu saram ‘that person’ in (2c). Meanwhile from (2d) we can surmise that a focused phrase undergoes movement.
Following from this reasoning, where by wh-words exhibit island effects parallel to other focused phrases, I assume that wh-words bear focus feature. Under this assumption the asymmetric grammaticality of (2a) and (2b) is predictable. A further prediction of my assumption is that wh-words with focus cannot occur in a declarative sentence. The data in (3) beat this prediction out.
(3) a.*Jinsu-ga (dodaeche) nugu-reul joaha-n-da.
Jinsu-NOM (on earth) who-ACC like-TEN-C
‘Jinsu likes who.’
b. Jinsu-ga nugu-reul joahan-da.
Jinsu-NOM someone-ACC like-C
‘Jinsu likes someone.’
The object nugu in (3a) has focus which is marked by the adverb dodaeche‘on earth’. Therefore the reason for the apparently exotic phenomenon whereby wh-words can occur ina declarative sentence such as (3b) turns out to be that the wh-words in this context are always indefinites which do not have focus.
The focus property of wh-words in Korean leads us to postulate a free focus assignment rule which goes back at least to Jackendoff (1972) and which has been adopted by Rochemont (1986) and Horvath(1986).
(4) Free Focus Assignment Rule
Assign focus to
To reflect the fact that the focus feature affects both the phonological and the semantic content of the relevant syntactic object, the free focus assign rule should apply to each lexical item before derivation starts. The ambiguity of wh-words, and the asymmetric behavior of wh-words and indefinites in an island construction, suggest that the distinction between wh-words and indefinites is decided not by any structure-based condition on their distribution, but by the morphological feature assigned by the free focus assignment rule, in the sense that focus is assigned to any kind of syntactic object arbitrarily.
Before moving to the next section in which I will discuss indefinites in more detail, it should be stressed that the fact that a wh-word can have a wh-word interpretation only in case it has focus leads us to this question: If wh-words in Korean undergo focus movement, does it mean that wh-words do not undergowh-movement or that wh-words experience both focus movement and wh-movement? This question will be discussed in Section 4, but let me point out the implication of the fact that the free focus assignment rule feeds a wh-word reading. The significant consequence of the free focus assignment rule for Korean is that a [wh] feature cannot take a role in deciding the interpretation of a wh-word because under the traditional conception all wh-words should have a [wh] feature irrelevant to their interpretation.
3. Indefinites, existential quantifiers and wh-words
In the previous section we saw that the focus feature plays a crucial role in distinguishinga wh-word reading and an indefinite reading, and this can be explained by the free focus assignment rule.The potential argument against this proposal is that a bare wh-word might be the same entity as a quantifier. In other words indefinites are the result of deletion of the question morpheme nga in the wh+Q complex.[6] The immediate problem with this potential argument is thatit cannot explain why wh-words manifest prosodic prominence unlike indefinites as we have already seen in (1). Another falsification of the deletion hypothesis comes from the distinct syntactic distribution between indefinites and quantifiers.
(5) Intervention effects
a. nugunga-ga Jinsu-ege mueos-eul ju-ass-ni?
someone-NOM Jinsu-DAT what-ACC give-TEN-Q
‘What did someone give to Jinsu?’
b. *nugu-ga Jinsu-ege mueos-eul ju-ass-ni?
someone-NOM Jinsu-DAT what-ACC give-TEN-Q
‘What did someone give to Jinsu?’
(5b) shows that, unlike quantifiers, indefinites cannot c-command wh-words. At this point let me explain why the example (5) is analyzed as an intervention effect. Within the Minimalist framework,an intervention effect is understood as a sub-case of the economy principle that requires minimization of computational complexity. A version of economy principle is the minimal link condition that requires minimization of the searching domain of a head, in the sense that a head attracts the closest syntactic object.[7] The following English data exemplify a case of violation of the economy principle under the minimal link condition.
(6) a. Do you know who likes what?
b. *Do you know what who likes?
c. Do you know [C+wh who likes what]
At some point during the derivation the head C+wh in (6c) must attract a [wh] feature. Whois the closest potential attractee. If the head C+wh attracts what over who, which is (6b), it results in ungrammaticality due to violation of the economy principle. Therefore in English only (6a) is acceptable.
Now let me consider example (5b). To the extent that (5b) is understood as an intervention effect it is equivalent to the English example (6b). When a question morpheme is merged it starts to search for a candidate which will be moved. By top-down search the closest element to Q is nugu which does not have focus. Therefore, to obey the economy principle, Q should attract nugu but the problem is thatnugu does not bear focus. On the other hand if mueos which has a focus feature moves, then it violates the economy principle. Therefore the sentence (5b) crashes. Note that the distance is not decided by linear order but by hierarchy. Because Korean is a head-final language it is not easy to see what is the closest syntactic object. For clarity, see the following tree.
(7) a. *nugu-ga Jinsu-ege mueos-eul ju-ass-ni?
someone-NOM Jinsu-DAT what-ACC give-TEN-Q
‘What did someone give to Jinsu?’[8]
b.
* CP
2
TP C
2
nugu T’
2
vP T
2
mueos v’
[foc] 5
The fact that indefinites interfere with the focus movement of wh-words makes it seem plausible that wh-words and indefinites are the same syntactic objects in the sense that both of them have a feature which is relevant to Q. In fact, this is the traditional analysis:wh-words and indefinites in Korean belong to the same morphological category. However as we have seen in Section 2, indefinites are also different from wh-words. An apparent difference is their interpretation. A further difference between indefinites and wh-words is that indefinites are irrelevant to a negation-induced barrier.
(8) Negation Induced Barrier (Beck and Kim 1997: 347)
The first node that dominates a negative quantifier, its restriction, and its nuclear scope is
a Negation Induced Barrier.
For convenience let me rephrase (8). According to Beck and Kim (1997), a negative quantifier such as no one forms a barrier which prevents scope bearing elements(including quantifiers[9]and wh-words) which are c-commanded by no one from moving out of its scope.
(9) Island effects
a. *amudo nugu-reul an-chodaeha-ass-ni?
no one who-ACC NEG-invite-TEN-Q
‘Whom does no one invite?’
b. amudo nugu-reul an-chodaeha-ass-ni?
no one someone-ACC NEG-invite-TEN-Q
‘Does no one invite someone?’
The licit status of the sentence (9b) shows that, unlike a wh-word nugu, an indefinite nugu is not extracted from the scope of a negative quantifier. That is, considering the fact that a negation-induced barrier blocks movement, it seems natural to assume that indefinites do not move at all. To make this assumption more plausible, let me examine another island construction. Following Ross(1967), a transformation rule must apply to all conjuncts in a coordinate structure.
(10) *Whati did you see [who and ti]?
The ungrammaticality of the example (10) comes from an improper application of a transformational rule, in this case wh-movement. That is, it applied to only one element of the conjunct pair who and what. As shown in (11a) Korean also obeys the coordinate structure constraint which bans uneven application of Move on coordinate structures.
(11) Coordinate structure
a. *Jinsu-neun nugu-reul manna-go Yunho-neun jip-e ga-ass-ni?
Jinsu-TOP who-ACC meet-conj Yunho-TOP home-LOC go-TEN-Q
‘Whomi did Jinsu meet tiand did Yunho go home?’
b. Jinsu-neun nugu-reul manna-go Yunho-neun jip-e ga-ass-ni?
Jinsu-TOP someone-ACC meet-conj Yunho-TOP home-LOC go-TEN-Q
‘Did Jinsu meet someone and Yunho go home?’
On the other handsentence (11b) is grammatical,with the indefinite nugu contained in one conjunct. This means that there is no improper movement in (11b), which in turn implies that indefinites in Korean do not move.
In this section we have seen that indefinites are different both from quantifiers and from wh-words. For the distinction between indefinites and quantifiers, it was pointed out that indefinites display an intervention effect when they intervene between Q and wh-words. From the intervention effect and the morphological homogeneity of wh-words and indefinites, I followed the traditional analysis that wh-words and indefinites in Korean are the same syntactic object. Turning to the difference between indefinites and wh-words we saw that indefinites do not move. To the extent that the negative induced barrier and the coordinate structure constraint are diagnostics of movement, it seems natural to assume that indefinites do not move, unlike wh-words which are visible to the island effect and the coordinate structure constraint on movement.
4. Focus movement and intervention effects
4.1. Focus movement of wh-words
In this section I will reinforce the current proposal of focus movement. The previous section provided evidence for the proposal by showing that wh-words manifest island effects and obey the coordinate structure constraint. Yet it should be noted that island effects and the coordinate structure constraint can also be evidence for the unselective binding thesis, in the sense that they can block proper binding. Therefore, to provide further evidence for focus movement of wh-words,I will investigate the uneven interpretations of wh-words which the unselective binding thesis cannot capture.
The relevant evidence comes from a condition on the intervention effect which requires that an intervener and a moved element should occur in the same clause. To begin let us reconsider the intervention effect which was introduced in Section 2. As already noted, wh-words in Korean cannot be c-commanded by indefinites. Therefore, among possible interpretations of a sentence including an indefinite and a wh-word, one reading ((12c) in the example below) is not available.
(12)a. nugu-ga Jinsu-ege mueos-eul ju-ass-ni?
who-NOM Jinsu-DAT what-ACC give-TEN-Q
‘Who gave what to Jinsu?’
b. nugu-ga Jinsu-ege mueos-eul ju-ass-ni?
who-NOM Jinsu-DAT something-ACC give-TEN-Q
‘Who gave something to Jinsu?’
c. *nugu-ga Jinsu-ege mueos-eul ju-ass-ni?
someone-NOM Jinsu-DAT what-ACC give-TEN-Q
‘What did someone give to Jinsu?’
d. nugu-ga Jinsu-ege mueos-eul ju-ass-ni?
someone-NOM Jinsu-DAT something-ACC give-TEN-Q
‘Did someone give something to Jinsu?’
Here our interest is in (12b). (12b) is grammatical sentence in which the indefinite mueos is not followed by the wh-wordnugu. But if the indefinite mueos is scrambled to the left of the sentence in (12b) the sentence becomes ungrammatical. The interesting fact related to the scrambling of the constructions such as (12b) is that the degradation of grammaticality seems to depend whether indefinites and wh-words sit in the same clause or not. To be more precise, if an indefinite and a wh-word occur in separate clauses, respectively scrambling an indefinite over a wh-word does not affect the grammaticality of the sentence.
(13)a. nugu-ga Jinsu-ege mueos-eul ju-ass-ni?
who-NOM Jinsu-DAT something-ACC give-TEN-Q?
b. *mueos-eul i nugu-ga Jinsu-ege t i ju-ass-ni?
something-ACC who-NOM Jinsu-DAT give-TEN-Q?
‘Who gave what to Jinsu?’
(14)a. nugu-ga [Jinsu-ga mueos-eul sa-ass-dago] malha-ass-ni?
who-NOM Jinsu-NOM something-ACC buy-TEN-C say-TEN-Q?
b. [Jinsu-ga mueos-eul sa-ass-dago] i nugu-ga ti malha-ass-ni?
Jinsu-NOM something-ACC buy-TEN-C who-NOM say-TEN-Q?
c. *mueos-euli nugu-ga [Jinsu-ga ti sa-ass-dago] malha-ass-ni?
something-ACC who-NOM Jinsu-NOM buy-TEN-C say-TEN-Q?
‘Who said that Jinsu bought something?’
The ungrammaticality of (13b) and (14c) shows that the degradation of the sentence occurs either in short scrambling, where a syntactic object moves within a clause boundary, or in long scrambling, which crosses a clause boundary. In other words regardless of the type of scrambling, moving an indefinite over a wh-word degrades the grammaticality of the sentence. On the other hand, as shown in (14b), scrambling the embedded sentence to the left periphery does not affect the legitimate status of (14a). Comparing (14b) to (14c), it seems that being clausemates is responsible for the ungrammaticality of the sentences which display the intervention effect. However the intervention effect is not limited to the case in which indefinites and wh-words sit in the same clause.