Negation, Focus, and do-Support in Korean
Paul Hagstrom, MIT
Summer 1995

ABSTRACT

I. Movement to specifier of functional categories; Korean negation

The recent literature contains many analyses which posit movement into specifiers
of functional categories, driven by a need to check morphological features.  I
will demonstrate that Korean supports these views of movement.  In particular, we
see that in negation constructions, an object will shift into a specifier above a
negation phrase (NegP) unless a closer nominal element intervenes.  Positing
object shift in negation constructions coupled with the independent ability to
nominalize a VP in Korean predicts the two negation constructions which in fact
occur.  In (1), the two forms are given; the first, Short-Form Negation ("SFN")
involves a negative preverbal affix, while the second, Long-Form Negation ("LFN")
involves a postverbal negation morpheme and morphological "do-support" analogous
to that seen in the English gloss of the same sentence.

(1)	a.	Bill-i   chayk-ul an-ilk-ess-ta
		Bill-nom book-acc neg-read-pst-dcl
		'Bill did not read the book.'
		
	b.	Bill-i   chayk-ul ilk-ci  ani ha-ess-ta
		Bill-nom book-acc read-ci neg do-pst-dcl
		'Bill did not read the book.'

I argue that both cases in (1) involve object shift.  In (1a), the object shifts
from a base position between an- (the Neg morpheme) and the verb, moving into a
specifier position above negation.  In (1b), the entire verb-object complex has
been nominalized by the morpheme -ci, causing this entire phrase to shift as it
were the object.  This movement of the "ci"-phrase is due to the fact that it
represents the closest nominal element which is capable of satisfying the strong
features of Neg-0 head that drive object shift.  The analysis adopts the "Split-VP
Hypothesis" (variations of which appear in Koizumi (1993), Travis (1992), Bowers
(1993), Chomsky (1995), among other places), and the freedom of specifiers to
proliferate (Chomsky (1995), Ura (1994)).

II. Support for the proposed analysis

The proposed analysis receives support from an otherwise unexplained phenomenon
observed in child acquisition of Korean.  While in adult SFN (1a) the negative
morpheme an- is strictly a verbal prefix, children near age 2 will frequently
utter sentences like those in (2), where an- is separated from the verb by the
object (Cho & Hong (1988), Kim (1992)).  This is neatly explained by supposing
that these children simply fail to execute the object shift under discussion.

(2)	a.	*na an  pap  mek-e
		 I  neg rice eat-dcl
		('I do not eat rice')

	b.	*an  phikul coa-hay
		 neg pickle like-dcl
		('I do not like pickles')

	c.	*an  kyelan mek-e
		 neg egg    eat-dcl
		('(I) wonÍt eat (my) eggs.')

Further support comes from the surface form and allowable combinations in
multiple negation constructions.  While any number of LFNs may be used in a
sentence, no more than one instance of SFN is allowable.  This pattern derives
from the inability of a Neg-0 head to take a NegP complement, while allowing a
complement nominalized with -ci.  The surface order of these constructions falls
out of the proposed analysis in a straightforward way. Other supporting evidence
includes some unusual behavior of Negative Polarity Items which both support the
proposed structure and draw an interesting parallel to wh-movement in English.
Lastly, we examine some properties of a "tense doubling" phenomenon, previously
undiscussed in the Korean syntax literature (to my knowledge).


[Cited references can be found in the paper]